2004-06-24 14:18:37 +00:00
|
|
|
/* NetworkManager -- Network link manager
|
|
|
|
|
*
|
|
|
|
|
* Dan Williams <dcbw@redhat.com>
|
|
|
|
|
*
|
|
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
|
* it under the terms of the GNU General Public License as published by
|
|
|
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
|
|
|
* (at your option) any later version.
|
|
|
|
|
*
|
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
|
* GNU General Public License for more details.
|
|
|
|
|
*
|
|
|
|
|
* You should have received a copy of the GNU General Public License
|
|
|
|
|
* along with this program; if not, write to the Free Software
|
|
|
|
|
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
|
|
|
|
|
*
|
2005-04-15 15:43:42 +00:00
|
|
|
* (C) Copyright 2005 Red Hat, Inc.
|
2004-06-24 14:18:37 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include <errno.h>
|
|
|
|
|
#include <glib.h>
|
|
|
|
|
#include <dbus/dbus-glib.h>
|
2005-02-18 18:15:54 +00:00
|
|
|
#include <libhal.h>
|
2004-06-24 14:18:37 +00:00
|
|
|
#include <iwlib.h>
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
#include <signal.h>
|
|
|
|
|
#include <string.h>
|
2005-09-15 14:47:17 +00:00
|
|
|
#include <netinet/ether.h>
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-06-21 15:07:01 +00:00
|
|
|
#include "autoip.h"
|
2004-06-24 14:18:37 +00:00
|
|
|
#include "NetworkManager.h"
|
2004-10-21 17:42:14 +00:00
|
|
|
#include "NetworkManagerMain.h"
|
2004-06-24 14:18:37 +00:00
|
|
|
#include "NetworkManagerDevice.h"
|
2004-11-22 14:42:34 +00:00
|
|
|
#include "NetworkManagerDevicePrivate.h"
|
2004-06-24 14:18:37 +00:00
|
|
|
#include "NetworkManagerUtils.h"
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
#include "NetworkManagerDbus.h"
|
|
|
|
|
#include "NetworkManagerWireless.h"
|
|
|
|
|
#include "NetworkManagerPolicy.h"
|
2004-08-02 21:12:40 +00:00
|
|
|
#include "NetworkManagerAPList.h"
|
2004-11-16 02:41:53 +00:00
|
|
|
#include "NetworkManagerSystem.h"
|
2005-04-15 15:43:42 +00:00
|
|
|
#include "nm-ip4-config.h"
|
|
|
|
|
#include "nm-vpn-manager.h"
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
#include "nm-dhcp-manager.h"
|
|
|
|
|
#include "nm-activation-request.h"
|
2005-03-11 20:12:57 +00:00
|
|
|
#include "nm-utils.h"
|
|
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
/* Local static prototypes */
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
static gpointer nm_device_worker (gpointer user_data);
|
|
|
|
|
static gboolean nm_device_wireless_scan (gpointer user_data);
|
2005-02-07 23:04:05 +00:00
|
|
|
static gboolean supports_mii_carrier_detect (NMDevice *dev);
|
|
|
|
|
static gboolean supports_ethtool_carrier_detect (NMDevice *dev);
|
2005-03-26 03:42:05 +00:00
|
|
|
static gboolean nm_device_bring_up_wait (NMDevice *dev, gboolean cancelable);
|
2005-05-05 16:52:04 +00:00
|
|
|
static gboolean link_to_specific_ap (NMDevice *dev, NMAccessPoint *ap, gboolean default_link);
|
2005-10-09 04:39:49 +00:00
|
|
|
static guint32 nm_device_discover_capabilities (NMDevice *dev);
|
|
|
|
|
static gboolean nm_is_driver_supported (NMDevice *dev);
|
|
|
|
|
static guint32 nm_device_wireless_discover_capabilities (NMDevice *dev);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
static void nm_device_activate_schedule_stage1_device_prepare (NMActRequest *req);
|
|
|
|
|
static void nm_device_activate_schedule_stage2_device_config (NMActRequest *req);
|
|
|
|
|
static void nm_device_activate_schedule_stage3_ip_config_start (NMActRequest *req);
|
|
|
|
|
static void nm_device_activate_schedule_stage5_ip_config_commit (NMActRequest *req);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
|
|
|
|
typedef struct
|
|
|
|
|
{
|
|
|
|
|
NMDevice *dev;
|
|
|
|
|
struct wireless_scan_head scan_head;
|
|
|
|
|
} NMWirelessScanResults;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-04-01 21:30:12 +00:00
|
|
|
typedef struct
|
|
|
|
|
{
|
2005-09-02 03:47:19 +00:00
|
|
|
NMDevice * dev;
|
|
|
|
|
gboolean force;
|
2005-04-01 21:30:12 +00:00
|
|
|
} NMWirelessScanCB;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
|
|
|
|
/******************************************************/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/******************************************************/
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
/*
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
* nm_device_test_wireless_extensions
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
|
|
|
|
* Test whether a given device is a wireless one or not.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
static gboolean nm_device_test_wireless_extensions (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
int err = -1;
|
|
|
|
|
char ioctl_buf[64];
|
|
|
|
|
NMSock *sk;
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
/* We obviously cannot probe test devices (since they don't
|
|
|
|
|
* actually exist in hardware).
|
|
|
|
|
*/
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
return (FALSE);
|
|
|
|
|
|
2004-10-28 19:49:55 +00:00
|
|
|
ioctl_buf[63] = 0;
|
2005-04-06 16:45:48 +00:00
|
|
|
strncpy (ioctl_buf, nm_device_get_iface(dev), 63);
|
2004-10-28 19:49:55 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
|
|
|
|
{
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: About to GET IWNAME\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = ioctl (nm_dev_sock_get_fd (sk), SIOCGIWNAME, ioctl_buf);
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with GET IWNAME\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
|
|
|
|
}
|
2004-10-29 16:37:43 +00:00
|
|
|
return (err == 0);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
2005-10-09 04:39:49 +00:00
|
|
|
* nm_get_device_driver_name
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
2005-10-09 04:39:49 +00:00
|
|
|
* Get the device's driver name from HAL.
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-10-09 04:39:49 +00:00
|
|
|
static char *nm_get_device_driver_name (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-10-09 04:39:49 +00:00
|
|
|
char * udi = NULL;
|
|
|
|
|
char * driver_name = NULL;
|
|
|
|
|
LibHalContext *ctx = NULL;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, NULL);
|
2005-04-06 16:45:48 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
ctx = dev->app_data->hal_ctx;
|
|
|
|
|
g_return_val_if_fail (ctx != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
if ((udi = nm_device_get_udi (dev)))
|
2005-04-06 16:45:48 +00:00
|
|
|
{
|
2005-10-11 16:18:12 +00:00
|
|
|
char *physdev_udi = libhal_device_get_property_string (ctx, udi, "net.physical_device", NULL);
|
2005-10-09 04:39:49 +00:00
|
|
|
|
2005-10-11 16:18:12 +00:00
|
|
|
if (physdev_udi && libhal_device_property_exists (ctx, physdev_udi, "info.linux.driver", NULL))
|
2005-10-09 04:39:49 +00:00
|
|
|
{
|
2005-10-11 16:18:12 +00:00
|
|
|
char *drv = libhal_device_get_property_string (ctx, physdev_udi, "info.linux.driver", NULL);
|
2005-10-09 04:39:49 +00:00
|
|
|
driver_name = g_strdup (drv);
|
|
|
|
|
g_free (drv);
|
|
|
|
|
}
|
2005-10-11 16:18:12 +00:00
|
|
|
g_free (physdev_udi);
|
2005-04-06 16:45:48 +00:00
|
|
|
}
|
2005-10-09 04:39:49 +00:00
|
|
|
|
|
|
|
|
return driver_name;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Blacklist of unsupported drivers */
|
|
|
|
|
static char * driver_blacklist[] =
|
|
|
|
|
{
|
|
|
|
|
NULL
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_is_driver_supported
|
|
|
|
|
*
|
|
|
|
|
* Check device's driver against a blacklist of unsupported drivers.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_is_driver_supported (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
char ** drv = NULL;
|
2005-10-11 20:13:28 +00:00
|
|
|
gboolean supported = TRUE;
|
2005-10-09 04:39:49 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->driver != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
for (drv = &driver_blacklist[0]; *drv; drv++)
|
|
|
|
|
{
|
|
|
|
|
if (!strcmp (*drv, dev->driver))
|
2005-10-11 20:13:28 +00:00
|
|
|
{
|
|
|
|
|
supported = FALSE;
|
|
|
|
|
break;
|
|
|
|
|
}
|
2005-10-09 04:39:49 +00:00
|
|
|
}
|
2005-10-11 20:13:28 +00:00
|
|
|
|
|
|
|
|
/* Check for Wireless Extensions support >= 16 for wireless devices */
|
|
|
|
|
if (supported && nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
NMSock * sk = NULL;
|
|
|
|
|
guint8 we_ver = 0;
|
|
|
|
|
|
|
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
|
|
|
|
{
|
|
|
|
|
iwrange range;
|
|
|
|
|
if (iw_get_range_info (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), &range) >= 0)
|
|
|
|
|
we_ver = range.we_version_compiled;
|
|
|
|
|
nm_dev_sock_close (sk);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (we_ver < 16)
|
|
|
|
|
{
|
|
|
|
|
nm_warning ("%s: driver's Wireless Extensions version (%d) is too old. Can't use device.",
|
|
|
|
|
nm_device_get_iface (dev), we_ver);
|
|
|
|
|
supported = FALSE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return supported;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_get_device_by_udi
|
|
|
|
|
*
|
|
|
|
|
* Search through the device list for a device with a given UDI.
|
|
|
|
|
*
|
|
|
|
|
* NOTE: the caller MUST hold the device list mutex already to make
|
|
|
|
|
* this routine thread-safe.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMDevice *nm_get_device_by_udi (NMData *data, const char *udi)
|
|
|
|
|
{
|
|
|
|
|
NMDevice *dev = NULL;
|
2005-02-10 04:39:40 +00:00
|
|
|
GSList *elt;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (data != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (udi != NULL, NULL);
|
|
|
|
|
|
2005-02-10 04:39:40 +00:00
|
|
|
for (elt = data->dev_list; elt; elt = g_slist_next (elt))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-04-27 18:05:16 +00:00
|
|
|
if ((dev = (NMDevice *)(elt->data)))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
if (nm_null_safe_strcmp (nm_device_get_udi (dev), udi) == 0)
|
2005-09-06 19:35:50 +00:00
|
|
|
return dev;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2005-09-06 19:35:50 +00:00
|
|
|
return NULL;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_get_device_by_iface
|
|
|
|
|
*
|
|
|
|
|
* Search through the device list for a device with a given iface.
|
|
|
|
|
*
|
|
|
|
|
* NOTE: the caller MUST hold the device list mutex already to make
|
|
|
|
|
* this routine thread-safe.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMDevice *nm_get_device_by_iface (NMData *data, const char *iface)
|
|
|
|
|
{
|
|
|
|
|
NMDevice *iter_dev = NULL;
|
|
|
|
|
NMDevice *found_dev = NULL;
|
2005-02-10 04:39:40 +00:00
|
|
|
GSList *elt;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (data != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (iface != NULL, NULL);
|
|
|
|
|
|
2005-02-10 04:39:40 +00:00
|
|
|
for (elt = data->dev_list; elt; elt = g_slist_next (elt))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-02-10 04:39:40 +00:00
|
|
|
iter_dev = (NMDevice *)(elt->data);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (iter_dev)
|
|
|
|
|
{
|
|
|
|
|
if (nm_null_safe_strcmp (nm_device_get_iface (iter_dev), iface) == 0)
|
|
|
|
|
{
|
|
|
|
|
found_dev = iter_dev;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (found_dev);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*****************************************************************************/
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
/* NMDevice object routines */
|
2004-06-24 14:18:37 +00:00
|
|
|
/*****************************************************************************/
|
|
|
|
|
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_copy_allowed_to_dev_list
|
|
|
|
|
*
|
|
|
|
|
* For devices that don't support wireless scanning, copy
|
|
|
|
|
* the allowed AP list to the device's ap list.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_copy_allowed_to_dev_list (NMDevice *dev, NMAccessPointList *allowed_list)
|
|
|
|
|
{
|
|
|
|
|
NMAPListIter *iter;
|
|
|
|
|
NMAccessPoint *src_ap;
|
|
|
|
|
NMAccessPointList *dev_list;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
if (allowed_list == NULL)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
nm_device_ap_list_clear (dev);
|
|
|
|
|
dev->options.wireless.ap_list = nm_ap_list_new (NETWORK_TYPE_ALLOWED);
|
|
|
|
|
|
|
|
|
|
if (!(iter = nm_ap_list_iter_new (allowed_list)))
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
dev_list = nm_device_ap_list_get (dev);
|
|
|
|
|
while ((src_ap = nm_ap_list_iter_next (iter)))
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *dst_ap = nm_ap_new_from_ap (src_ap);
|
|
|
|
|
|
|
|
|
|
/* Assume that if the allowed list AP has a saved encryption
|
|
|
|
|
* key that the AP is encrypted.
|
|
|
|
|
*/
|
|
|
|
|
if ( (nm_ap_get_auth_method (src_ap) == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM)
|
|
|
|
|
|| (nm_ap_get_auth_method (src_ap) == NM_DEVICE_AUTH_METHOD_SHARED_KEY))
|
|
|
|
|
nm_ap_set_encrypted (dst_ap, TRUE);
|
|
|
|
|
|
|
|
|
|
nm_ap_list_append_ap (dev_list, dst_ap);
|
|
|
|
|
nm_ap_unref (dst_ap);
|
|
|
|
|
}
|
|
|
|
|
nm_ap_list_iter_free (iter);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_init
|
|
|
|
|
*
|
|
|
|
|
* Initialize a new wireless device with wireless-specific settings.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_wireless_init (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
NMSock *sk;
|
|
|
|
|
NMDeviceWirelessOptions *opts = &(dev->options.wireless);
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), FALSE);
|
|
|
|
|
|
|
|
|
|
opts->scan_mutex = g_mutex_new ();
|
|
|
|
|
opts->ap_list = nm_ap_list_new (NETWORK_TYPE_DEVICE);
|
|
|
|
|
if (!opts->scan_mutex || !opts->ap_list)
|
|
|
|
|
return FALSE;
|
|
|
|
|
|
|
|
|
|
nm_register_mutex_desc (opts->scan_mutex, "Scan Mutex");
|
|
|
|
|
nm_wireless_set_scan_interval (dev->app_data, dev, NM_WIRELESS_SCAN_INTERVAL_ACTIVE);
|
|
|
|
|
|
|
|
|
|
nm_device_set_mode (dev, NETWORK_MODE_INFRA);
|
|
|
|
|
|
|
|
|
|
/* Non-scanning devices show the entire allowed AP list as their
|
|
|
|
|
* available networks.
|
|
|
|
|
*/
|
|
|
|
|
if (!(dev->capabilities & NM_DEVICE_CAP_WIRELESS_SCAN))
|
|
|
|
|
nm_device_copy_allowed_to_dev_list (dev, dev->app_data->allowed_ap_list);
|
|
|
|
|
|
2005-10-11 20:13:28 +00:00
|
|
|
opts->we_version = 0;
|
2005-10-09 04:39:49 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
|
|
|
|
{
|
|
|
|
|
iwrange range;
|
|
|
|
|
if (iw_get_range_info (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), &range) >= 0)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
|
|
|
|
opts->max_qual.qual = range.max_qual.qual;
|
|
|
|
|
opts->max_qual.level = range.max_qual.level;
|
|
|
|
|
opts->max_qual.noise = range.max_qual.noise;
|
|
|
|
|
opts->max_qual.updated = range.max_qual.updated;
|
|
|
|
|
|
|
|
|
|
opts->avg_qual.qual = range.avg_qual.qual;
|
|
|
|
|
opts->avg_qual.level = range.avg_qual.level;
|
|
|
|
|
opts->avg_qual.noise = range.avg_qual.noise;
|
|
|
|
|
opts->avg_qual.updated = range.avg_qual.updated;
|
|
|
|
|
|
|
|
|
|
opts->num_freqs = MIN (range.num_frequency, IW_MAX_FREQUENCIES);
|
|
|
|
|
for (i = 0; i < opts->num_freqs; i++)
|
|
|
|
|
opts->freqs[i] = iw_freq2float (&(range.freq[i]));
|
2005-10-11 20:13:28 +00:00
|
|
|
|
|
|
|
|
opts->we_version = range.we_version_compiled;
|
2005-10-09 04:39:49 +00:00
|
|
|
}
|
|
|
|
|
nm_dev_sock_close (sk);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_new
|
|
|
|
|
*
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
* Creates and initializes the structure representation of an NM device. For test
|
|
|
|
|
* devices, a device type other than DEVICE_TYPE_DONT_KNOW must be specified, this
|
|
|
|
|
* argument is ignored for real hardware devices since they are auto-probed.
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2004-10-21 17:42:14 +00:00
|
|
|
NMDevice *nm_device_new (const char *iface, const char *udi, gboolean test_dev, NMDeviceType test_dev_type, NMData *app_data)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
NMDevice *dev;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
GError *error = NULL;
|
2005-04-28 02:50:54 +00:00
|
|
|
nm_completion_args args;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (iface != NULL, NULL);
|
2004-08-24 00:31:47 +00:00
|
|
|
g_return_val_if_fail (strlen (iface) > 0, NULL);
|
2004-10-21 17:42:14 +00:00
|
|
|
g_return_val_if_fail (app_data != NULL, NULL);
|
2004-08-24 00:31:47 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/* Test devices must have a valid type specified */
|
|
|
|
|
if (test_dev && !(test_dev_type != DEVICE_TYPE_DONT_KNOW))
|
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
|
|
/* Another check to make sure we don't create a test device unless
|
|
|
|
|
* test devices were enabled on the command line.
|
|
|
|
|
*/
|
2004-10-21 17:42:14 +00:00
|
|
|
if (!app_data->enable_test_devices && test_dev)
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
{
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_warning ("attempt to create a test device, but test devices were not enabled "
|
2005-03-26 03:42:05 +00:00
|
|
|
"on the command line. Will not create the device.");
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
return (NULL);
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
dev = g_malloc0 (sizeof (NMDevice));
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
dev->refcount = 2; /* 1 for starters, and another 1 for the worker thread */
|
2004-07-25 02:40:19 +00:00
|
|
|
dev->app_data = app_data;
|
2004-06-24 14:18:37 +00:00
|
|
|
dev->iface = g_strdup (iface);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
dev->test_device = test_dev;
|
2005-10-09 04:39:49 +00:00
|
|
|
dev->udi = g_strdup (udi);
|
|
|
|
|
dev->driver = nm_get_device_driver_name (dev);
|
2005-04-15 15:43:42 +00:00
|
|
|
dev->use_dhcp = TRUE;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
/* Real hardware devices are probed for their type, test devices must have
|
|
|
|
|
* their type specified.
|
|
|
|
|
*/
|
|
|
|
|
if (test_dev)
|
|
|
|
|
dev->type = test_dev_type;
|
|
|
|
|
else
|
|
|
|
|
dev->type = nm_device_test_wireless_extensions (dev) ?
|
2004-08-05 18:54:29 +00:00
|
|
|
DEVICE_TYPE_WIRELESS_ETHERNET : DEVICE_TYPE_WIRED_ETHERNET;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/* Device thread's main loop */
|
|
|
|
|
dev->context = g_main_context_new ();
|
|
|
|
|
dev->loop = g_main_loop_new (dev->context, FALSE);
|
|
|
|
|
|
|
|
|
|
if (!dev->context || !dev->loop)
|
|
|
|
|
goto err;
|
|
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
/* Have to bring the device up before checking link status and other stuff */
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_device_bring_up_wait (dev, 0);
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
/* First check for driver support */
|
|
|
|
|
if (nm_is_driver_supported (dev))
|
|
|
|
|
dev->capabilities |= NM_DEVICE_CAP_NM_SUPPORTED;
|
2005-01-31 19:53:38 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
/* Then discover devices-specific capabilities */
|
|
|
|
|
if (dev->capabilities & NM_DEVICE_CAP_NM_SUPPORTED)
|
2005-02-07 23:04:05 +00:00
|
|
|
{
|
2005-10-09 04:39:49 +00:00
|
|
|
dev->capabilities |= nm_device_discover_capabilities (dev);
|
2004-10-21 17:42:14 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
/* Initialize wireless-specific options */
|
|
|
|
|
if (nm_device_is_wireless (dev) && !nm_device_wireless_init (dev))
|
|
|
|
|
goto err;
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_set_link_active (dev, nm_device_probe_link_state (dev));
|
2004-11-16 02:41:53 +00:00
|
|
|
nm_device_update_ip4_address (dev);
|
|
|
|
|
nm_device_update_hw_address (dev);
|
|
|
|
|
|
|
|
|
|
/* Grab IP config data for this device from the system configuration files */
|
2005-04-15 15:43:42 +00:00
|
|
|
dev->system_config_data = nm_system_device_get_system_config (dev);
|
2005-06-15 12:15:34 +00:00
|
|
|
dev->use_dhcp = nm_system_device_get_use_dhcp (dev);
|
2004-10-21 19:00:13 +00:00
|
|
|
}
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
nm_print_device_capabilities (dev);
|
|
|
|
|
|
2005-04-15 20:00:28 +00:00
|
|
|
dev->worker = g_thread_create (nm_device_worker, dev, TRUE, &error);
|
|
|
|
|
if (!dev->worker)
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_error ("could not create device worker thread. (glib said: '%s')", error->message);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
g_error_free (error);
|
2005-02-07 23:04:05 +00:00
|
|
|
goto err;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
}
|
|
|
|
|
|
2005-01-24 21:53:39 +00:00
|
|
|
/* Block until our device thread has actually had a chance to start. */
|
2005-04-28 02:50:54 +00:00
|
|
|
args[0] = &dev->worker_started;
|
2005-07-13 16:58:13 +00:00
|
|
|
args[1] = (gpointer) "nm_device_new(): waiting for device's worker thread to start";
|
|
|
|
|
args[2] = GINT_TO_POINTER (LOG_INFO);
|
|
|
|
|
args[3] = GINT_TO_POINTER (0);
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_wait_for_completion (NM_COMPLETION_TRIES_INFINITY,
|
2005-04-28 02:50:54 +00:00
|
|
|
G_USEC_PER_SEC / 20, nm_completion_boolean_test, NULL, args);
|
|
|
|
|
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_info ("nm_device_new(): device's worker thread started, continuing.");
|
2005-01-24 21:53:39 +00:00
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
return (dev);
|
2005-02-07 23:04:05 +00:00
|
|
|
|
|
|
|
|
err:
|
|
|
|
|
/* Initial refcount is 2 */
|
|
|
|
|
nm_device_unref (dev);
|
|
|
|
|
nm_device_unref (dev);
|
|
|
|
|
return NULL;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Refcounting functions
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_ref (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->refcount++;
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_unref
|
|
|
|
|
*
|
|
|
|
|
* Decreases the refcount on a device by 1, and if the refcount reaches 0,
|
|
|
|
|
* deallocates memory used by the device.
|
|
|
|
|
*
|
|
|
|
|
* Returns: FALSE if device was not deallocated
|
|
|
|
|
* TRUE if device was deallocated
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_unref (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
gboolean deleted = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (dev->refcount == 1)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_device_worker_thread_stop (dev);
|
|
|
|
|
nm_device_bring_down (dev);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
nm_device_ap_list_clear (dev);
|
|
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
g_mutex_free (dev->options.wireless.scan_mutex);
|
2004-10-11 21:32:19 +00:00
|
|
|
if (dev->options.wireless.ap_list)
|
|
|
|
|
nm_ap_list_unref (dev->options.wireless.ap_list);
|
2005-02-10 04:39:40 +00:00
|
|
|
}
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
nm_system_device_free_system_config (dev, dev->system_config_data);
|
|
|
|
|
if (dev->ip4_config)
|
|
|
|
|
nm_ip4_config_unref (dev->ip4_config);
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (dev->act_request)
|
|
|
|
|
nm_act_request_unref (dev->act_request);
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
g_free (dev->udi);
|
|
|
|
|
g_free (dev->iface);
|
2005-10-09 04:39:49 +00:00
|
|
|
g_free (dev->driver);
|
2005-01-31 19:53:38 +00:00
|
|
|
memset (dev, 0, sizeof (NMDevice));
|
2004-08-05 18:54:29 +00:00
|
|
|
g_free (dev);
|
2005-01-07 18:07:06 +00:00
|
|
|
deleted = TRUE;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
else
|
|
|
|
|
dev->refcount--;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
return deleted;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_worker
|
|
|
|
|
*
|
|
|
|
|
* Main thread of the device.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gpointer nm_device_worker (gpointer user_data)
|
|
|
|
|
{
|
|
|
|
|
NMDevice *dev = (NMDevice *)user_data;
|
|
|
|
|
|
2005-01-24 21:53:39 +00:00
|
|
|
if (!dev)
|
|
|
|
|
{
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_error ("received NULL device object, NetworkManager cannot continue.");
|
2005-01-24 21:53:39 +00:00
|
|
|
exit (1);
|
|
|
|
|
}
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
/* Start the scanning timeout for devices that can do scanning */
|
|
|
|
|
if (nm_device_is_wireless (dev) && nm_device_get_supports_wireless_scan (dev))
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-04-01 21:30:12 +00:00
|
|
|
GSource *source = g_idle_source_new ();
|
|
|
|
|
guint source_id = 0;
|
|
|
|
|
NMWirelessScanCB *scan_cb;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-04-01 21:30:12 +00:00
|
|
|
scan_cb = g_malloc0 (sizeof (NMWirelessScanCB));
|
|
|
|
|
scan_cb->dev = dev;
|
2005-09-02 03:47:19 +00:00
|
|
|
scan_cb->force = TRUE;
|
2005-04-01 21:30:12 +00:00
|
|
|
|
|
|
|
|
g_source_set_callback (source, nm_device_wireless_scan, scan_cb, NULL);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
source_id = g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
}
|
|
|
|
|
|
2005-03-26 03:42:05 +00:00
|
|
|
dev->worker_started = TRUE;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
g_main_loop_run (dev->loop);
|
|
|
|
|
|
|
|
|
|
g_main_loop_unref (dev->loop);
|
|
|
|
|
g_main_context_unref (dev->context);
|
|
|
|
|
|
|
|
|
|
dev->loop = NULL;
|
|
|
|
|
dev->context = NULL;
|
|
|
|
|
|
|
|
|
|
nm_device_unref (dev);
|
|
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
void nm_device_worker_thread_stop (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
2005-03-26 03:42:05 +00:00
|
|
|
if (dev->loop)
|
|
|
|
|
g_main_loop_quit (dev->loop);
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
if (dev->worker)
|
|
|
|
|
{
|
|
|
|
|
g_thread_join (dev->worker);
|
|
|
|
|
dev->worker = NULL;
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_discover_capabilities
|
|
|
|
|
*
|
|
|
|
|
* Figure out wireless-specific capabilities
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static guint32 nm_device_wireless_discover_capabilities (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
NMSock * sk;
|
|
|
|
|
int err;
|
|
|
|
|
wireless_scan_head scan_data;
|
|
|
|
|
guint32 caps = NM_DEVICE_CAP_NONE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NM_DEVICE_CAP_NONE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NM_DEVICE_CAP_NONE);
|
|
|
|
|
|
|
|
|
|
/* A test wireless device can always scan (we generate fake scan data for it) */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
caps |= NM_DEVICE_CAP_WIRELESS_SCAN;
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
|
|
|
|
{
|
2005-10-11 20:13:28 +00:00
|
|
|
guint8 we_ver = dev->options.wireless.we_version;
|
|
|
|
|
err = iw_scan (nm_dev_sock_get_fd (sk), (char *) nm_device_get_iface (dev), we_ver, &scan_data);
|
2005-10-09 04:39:49 +00:00
|
|
|
nm_dispose_scan_results (scan_data.result);
|
|
|
|
|
if (!((err == -1) && (errno == EOPNOTSUPP)))
|
|
|
|
|
caps |= NM_DEVICE_CAP_WIRELESS_SCAN;
|
|
|
|
|
nm_dev_sock_close (sk);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return caps;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_discover_capabilities
|
|
|
|
|
*
|
|
|
|
|
* Figure out wireless-specific capabilities
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static guint32 nm_device_wired_discover_capabilities (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
guint32 caps = NM_DEVICE_CAP_NONE;
|
|
|
|
|
const char * udi = NULL;
|
|
|
|
|
char * usb_test = NULL;
|
|
|
|
|
LibHalContext *ctx = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NM_DEVICE_CAP_NONE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wired (dev), NM_DEVICE_CAP_NONE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, NM_DEVICE_CAP_NONE);
|
|
|
|
|
|
|
|
|
|
/* cipsec devices are also explicitly unsupported at this time */
|
|
|
|
|
if (strstr (nm_device_get_iface (dev), "cipsec"))
|
|
|
|
|
return NM_DEVICE_CAP_NONE;
|
|
|
|
|
|
|
|
|
|
/* Ignore Ethernet-over-USB devices too for the moment (Red Hat #135722) */
|
|
|
|
|
ctx = dev->app_data->hal_ctx;
|
|
|
|
|
udi = nm_device_get_udi (dev);
|
|
|
|
|
if ( libhal_device_property_exists (ctx, udi, "usb.interface.class", NULL)
|
|
|
|
|
&& (usb_test = libhal_device_get_property_string (ctx, udi, "usb.interface.class", NULL)))
|
|
|
|
|
{
|
|
|
|
|
libhal_free_string (usb_test);
|
|
|
|
|
return NM_DEVICE_CAP_NONE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (supports_ethtool_carrier_detect (dev) || supports_mii_carrier_detect (dev))
|
|
|
|
|
caps |= NM_DEVICE_CAP_CARRIER_DETECT;
|
|
|
|
|
|
|
|
|
|
return caps;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_discover_capabilities
|
|
|
|
|
*
|
|
|
|
|
* Called only at device initialization time to discover device-specific
|
|
|
|
|
* capabilities.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static guint32 nm_device_discover_capabilities (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
guint32 caps = NM_DEVICE_CAP_NONE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NM_DEVICE_CAP_NONE);
|
|
|
|
|
|
|
|
|
|
/* Don't touch devices that we already don't support */
|
|
|
|
|
if (!(dev->capabilities & NM_DEVICE_CAP_NM_SUPPORTED))
|
|
|
|
|
return NM_DEVICE_CAP_NONE;
|
|
|
|
|
|
|
|
|
|
if (nm_device_is_wired (dev))
|
|
|
|
|
caps |= nm_device_wired_discover_capabilities (dev);
|
|
|
|
|
else if (nm_device_is_wireless (dev))
|
|
|
|
|
caps |= nm_device_wireless_discover_capabilities (dev);
|
|
|
|
|
|
|
|
|
|
return caps;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_app_data
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMData *nm_device_get_app_data (const NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->app_data);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/*
|
|
|
|
|
* Get/Set for "removed" flag
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_get_removed (const NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
|
|
|
|
|
return (dev->removed);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_set_removed (NMDevice *dev, const gboolean removed)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->removed = removed;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
/*
|
|
|
|
|
* Return the amount of time we should wait for the device
|
|
|
|
|
* to get a link, based on the # of frequencies it has to
|
|
|
|
|
* scan.
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static gint nm_device_get_association_pause_value (NMDevice *dev)
|
2004-12-16 04:24:20 +00:00
|
|
|
{
|
2005-01-13 19:04:26 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, -1);
|
2004-12-16 04:24:20 +00:00
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), -1);
|
|
|
|
|
|
|
|
|
|
/* If the card supports more than 14 channels, we should probably wait
|
|
|
|
|
* around 10s so it can scan them all. After we set the ESSID on the card, the card
|
|
|
|
|
* has to scan all channels to find our requested AP (which can take a long time
|
|
|
|
|
* if it is an A/B/G chipset like the Atheros 5212, for example).
|
|
|
|
|
*/
|
2005-01-31 19:53:38 +00:00
|
|
|
if (dev->options.wireless.num_freqs > 14)
|
2005-02-02 21:49:14 +00:00
|
|
|
return 8;
|
2004-12-16 04:24:20 +00:00
|
|
|
else
|
|
|
|
|
return 5;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
|
|
|
|
* Get/set functions for UDI
|
|
|
|
|
*/
|
|
|
|
|
char * nm_device_get_udi (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
return (dev->udi);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_set_udi (NMDevice *dev, const char *udi)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (udi != NULL);
|
|
|
|
|
|
|
|
|
|
if (dev->udi)
|
|
|
|
|
g_free (dev->udi);
|
|
|
|
|
|
|
|
|
|
dev->udi = g_strdup (udi);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Get/set functions for iface
|
|
|
|
|
*/
|
2004-10-21 17:42:14 +00:00
|
|
|
const char * nm_device_get_iface (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
return (dev->iface);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
/*
|
|
|
|
|
* Get/set functions for driver
|
|
|
|
|
*/
|
|
|
|
|
const char * nm_device_get_driver (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
return (dev->driver);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
2004-08-05 18:54:29 +00:00
|
|
|
* Get/set functions for type
|
2004-06-24 14:18:37 +00:00
|
|
|
*/
|
2004-08-05 18:54:29 +00:00
|
|
|
guint nm_device_get_type (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2004-08-05 18:54:29 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, DEVICE_TYPE_DONT_KNOW);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
return (dev->type);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
gboolean nm_device_is_wireless (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
return (dev->type == DEVICE_TYPE_WIRELESS_ETHERNET);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
gboolean nm_device_is_wired (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
return (dev->type == DEVICE_TYPE_WIRED_ETHERNET);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
}
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-10-21 17:42:14 +00:00
|
|
|
/*
|
2005-10-09 04:39:49 +00:00
|
|
|
* Accessor for device capabilities
|
2004-10-21 17:42:14 +00:00
|
|
|
*/
|
2005-10-09 04:39:49 +00:00
|
|
|
guint32 nm_device_get_capabilities (NMDevice *dev)
|
2004-10-21 17:42:14 +00:00
|
|
|
{
|
2005-10-09 04:39:49 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, NM_DEVICE_CAP_NONE);
|
2004-10-21 17:42:14 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
return dev->capabilities;
|
2004-10-21 17:42:14 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
|
|
|
|
* Get/set functions for link_active
|
|
|
|
|
*/
|
2005-03-11 20:12:57 +00:00
|
|
|
gboolean nm_device_has_active_link (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->link_active);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_set_link_active (NMDevice *dev, const gboolean link_active)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
2005-05-05 16:52:04 +00:00
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-03-11 20:12:57 +00:00
|
|
|
if (dev->link_active != link_active)
|
|
|
|
|
{
|
|
|
|
|
dev->link_active = link_active;
|
2005-05-05 16:52:04 +00:00
|
|
|
|
|
|
|
|
/* Deactivate a currently active device */
|
|
|
|
|
if (!link_active && nm_device_get_act_request (dev))
|
|
|
|
|
{
|
2005-05-16 01:28:45 +00:00
|
|
|
nm_device_deactivate (dev);
|
2005-05-05 16:52:04 +00:00
|
|
|
nm_policy_schedule_device_change_check (dev->app_data);
|
|
|
|
|
}
|
|
|
|
|
else if (link_active && !nm_device_get_act_request (dev))
|
|
|
|
|
{
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
NMDevice * act_dev = nm_get_active_device (dev->app_data);
|
|
|
|
|
NMActRequest * act_dev_req = act_dev ? nm_device_get_act_request (act_dev) : NULL;
|
2005-05-05 16:52:04 +00:00
|
|
|
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
/* Should we switch to this device now that it has a link?
|
|
|
|
|
*
|
|
|
|
|
* Only auto-switch for wired devices, AND...
|
|
|
|
|
*
|
|
|
|
|
* only switch to fully-supported devices, since ones that don't have carrier detection
|
|
|
|
|
* capability usually report the carrier as "always on" even if its not really on. User
|
|
|
|
|
* must manually choose semi-supported devices.
|
|
|
|
|
*
|
2005-05-05 16:52:04 +00:00
|
|
|
*/
|
2005-10-09 04:39:49 +00:00
|
|
|
if (nm_device_is_wired (dev) && (nm_device_get_capabilities (dev) & NM_DEVICE_CAP_CARRIER_DETECT))
|
2005-05-05 16:52:04 +00:00
|
|
|
{
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
gboolean do_switch = act_dev ? FALSE : TRUE; /* If no currently active device, switch to this one */
|
|
|
|
|
NMActRequest * act_req;
|
|
|
|
|
|
|
|
|
|
/* If active device is wireless, switch to this one */
|
|
|
|
|
if (act_dev && nm_device_is_wireless (act_dev) && act_dev_req && !nm_act_request_get_user_requested (act_dev_req))
|
|
|
|
|
do_switch = TRUE;
|
2005-05-05 16:52:04 +00:00
|
|
|
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
if (do_switch && (act_req = nm_act_request_new (dev->app_data, dev, NULL, TRUE)))
|
2005-05-05 16:52:04 +00:00
|
|
|
{
|
|
|
|
|
nm_info ("Will activate wired connection '%s' because it now has a link.", nm_device_get_iface (dev));
|
|
|
|
|
nm_policy_schedule_device_activation (act_req);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
nm_dbus_schedule_device_status_change_signal (dev->app_data, dev, NULL, link_active ? DEVICE_CARRIER_ON : DEVICE_CARRIER_OFF);
|
2005-03-11 20:12:57 +00:00
|
|
|
}
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Get function for supports_wireless_scan
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_get_supports_wireless_scan (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
if (!nm_device_is_wireless (dev))
|
|
|
|
|
return (FALSE);
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
return (dev->capabilities & NM_DEVICE_CAP_WIRELESS_SCAN);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_supports_carrier_detect
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_get_supports_carrier_detect (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
if (!nm_device_is_wired (dev))
|
|
|
|
|
return (FALSE);
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
return (dev->capabilities & NM_DEVICE_CAP_CARRIER_DETECT);
|
2005-02-07 23:04:05 +00:00
|
|
|
}
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_is_associated
|
|
|
|
|
*
|
|
|
|
|
* Figure out whether or not we're associated to an access point
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_wireless_is_associated (NMDevice *dev)
|
|
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
struct iwreq wrq;
|
|
|
|
|
NMSock *sk;
|
|
|
|
|
gboolean associated = FALSE;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
/* Test devices have their link state set through DBUS */
|
|
|
|
|
if (dev->test_device)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return nm_device_has_active_link (dev);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)) == NULL)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return FALSE;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
/* Some cards, for example ipw2x00 cards, can short-circuit the MAC
|
|
|
|
|
* address check using this check on IWNAME. Its faster.
|
|
|
|
|
*/
|
2005-04-27 18:05:16 +00:00
|
|
|
memset (&wrq, 0, sizeof (struct iwreq));
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IWNAME.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_get_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCGIWNAME, &wrq) >= 0)
|
2005-01-13 19:04:26 +00:00
|
|
|
{
|
|
|
|
|
if (!strcmp(wrq.u.name, "unassociated"))
|
|
|
|
|
{
|
|
|
|
|
associated = FALSE;
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
if (!associated)
|
2005-01-13 19:04:26 +00:00
|
|
|
{
|
|
|
|
|
/*
|
|
|
|
|
* For all other wireless cards, the best indicator of a "link" at this time
|
|
|
|
|
* seems to be whether the card has a valid access point MAC address.
|
|
|
|
|
* Is there a better way? Some cards don't work too well with this check, ie
|
|
|
|
|
* Lucent WaveLAN.
|
|
|
|
|
*/
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IWAP.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_get_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCGIWAP, &wrq) >= 0)
|
2005-01-07 18:07:06 +00:00
|
|
|
if (nm_ethernet_address_is_valid ((struct ether_addr *)(&(wrq.u.ap_addr.sa_data))))
|
|
|
|
|
associated = TRUE;
|
2005-01-13 19:04:26 +00:00
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
out:
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return associated;
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
2005-03-11 20:12:57 +00:00
|
|
|
* nm_device_probe_wireless_link_state
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Gets the link state of a wireless device. WARNING: results are not
|
|
|
|
|
* conclusive if the device is currently activating.
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-03-11 20:12:57 +00:00
|
|
|
static gboolean nm_device_probe_wireless_link_state (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
gboolean link = FALSE;
|
|
|
|
|
NMAccessPoint *best_ap;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
/* Test devices have their link state set through DBUS */
|
|
|
|
|
if (dev->test_device)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return nm_device_has_active_link (dev);
|
2004-12-16 04:24:20 +00:00
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
if ((best_ap = nm_device_get_best_ap (dev)))
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
{
|
2005-05-05 16:52:04 +00:00
|
|
|
link = link_to_specific_ap (dev, best_ap, TRUE);
|
2005-05-04 18:11:40 +00:00
|
|
|
nm_ap_unref (best_ap);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return link;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
2005-03-11 20:12:57 +00:00
|
|
|
* nm_device_probe_wired_link_state
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
*
|
2005-02-07 23:04:05 +00:00
|
|
|
*
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-03-11 20:12:57 +00:00
|
|
|
static gboolean nm_device_probe_wired_link_state (NMDevice *dev)
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
{
|
|
|
|
|
gboolean link = FALSE;
|
2005-03-11 20:12:57 +00:00
|
|
|
gchar *contents, *carrier_path;
|
|
|
|
|
gsize length;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2005-02-07 23:04:05 +00:00
|
|
|
g_return_val_if_fail (nm_device_is_wired (dev) == TRUE, FALSE);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
/* Test devices have their link state set through DBUS */
|
|
|
|
|
if (dev->test_device)
|
2005-05-08 16:31:03 +00:00
|
|
|
return nm_device_has_active_link (dev);
|
2005-03-11 20:12:57 +00:00
|
|
|
|
|
|
|
|
if (dev->removed)
|
|
|
|
|
return FALSE;
|
|
|
|
|
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
carrier_path = g_strdup_printf ("/sys/class/net/%s/carrier", dev->iface);
|
2005-05-08 16:31:03 +00:00
|
|
|
if (g_file_get_contents (carrier_path, &contents, &length, NULL))
|
|
|
|
|
{
|
2005-03-11 20:12:57 +00:00
|
|
|
link = (gboolean) atoi (contents);
|
|
|
|
|
g_free (contents);
|
2005-04-15 20:00:28 +00:00
|
|
|
}
|
2005-05-08 16:31:03 +00:00
|
|
|
g_free (carrier_path);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/* We say that non-carrier-detect devices always have a link, because
|
|
|
|
|
* they never get auto-selected by NM. User has to force them on us,
|
|
|
|
|
* so we just hope the user knows whether or not the cable's plugged in.
|
|
|
|
|
*/
|
2005-10-09 04:39:49 +00:00
|
|
|
if (!(dev->capabilities & NM_DEVICE_CAP_CARRIER_DETECT))
|
2005-02-07 23:04:05 +00:00
|
|
|
link = TRUE;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
2005-05-05 16:52:04 +00:00
|
|
|
return link;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_probe_link_state
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Return the current link state of the device.
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
gboolean nm_device_probe_link_state (NMDevice *dev)
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
gboolean link = FALSE;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (!nm_device_is_up (dev))
|
|
|
|
|
nm_device_bring_up (dev);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
link = nm_device_probe_wireless_link_state (dev);
|
|
|
|
|
nm_device_update_signal_strength (dev);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
else if (nm_device_is_wired (dev))
|
|
|
|
|
link = nm_device_probe_wired_link_state (dev);
|
|
|
|
|
|
|
|
|
|
return link;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_get_essid
|
|
|
|
|
*
|
|
|
|
|
* If a device is wireless, return the essid that it is attempting
|
|
|
|
|
* to use.
|
|
|
|
|
*
|
|
|
|
|
* Returns: allocated string containing essid. Must be freed by caller.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
char * nm_device_get_essid (NMDevice *dev)
|
|
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
int err;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NULL);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
/* Test devices return the essid of their "best" access point
|
|
|
|
|
* or if there is none, the contents of the cur_essid field.
|
|
|
|
|
*/
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
2004-10-29 16:37:43 +00:00
|
|
|
NMAccessPoint *best_ap = nm_device_get_best_ap (dev);
|
|
|
|
|
char *essid = dev->options.wireless.cur_essid;
|
|
|
|
|
|
|
|
|
|
/* Or, if we've got a best ap, use that ESSID instead */
|
|
|
|
|
if (best_ap)
|
|
|
|
|
{
|
|
|
|
|
essid = nm_ap_get_essid (best_ap);
|
|
|
|
|
nm_ap_unref (best_ap);
|
|
|
|
|
}
|
2005-05-04 18:11:40 +00:00
|
|
|
return essid;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
}
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2004-09-13 21:00:35 +00:00
|
|
|
wireless_config info;
|
|
|
|
|
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET 'basic config' for ESSID.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = iw_get_basic_config (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), &info);
|
2004-09-13 21:00:35 +00:00
|
|
|
if (err >= 0)
|
|
|
|
|
{
|
|
|
|
|
if (dev->options.wireless.cur_essid)
|
|
|
|
|
g_free (dev->options.wireless.cur_essid);
|
|
|
|
|
dev->options.wireless.cur_essid = g_strdup (info.essid);
|
|
|
|
|
}
|
|
|
|
|
else
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_warning ("nm_device_get_essid(): error getting ESSID for device %s. errno = %d", nm_device_get_iface (dev), errno);
|
2004-09-13 21:00:35 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
return (dev->options.wireless.cur_essid);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_set_essid
|
|
|
|
|
*
|
|
|
|
|
* If a device is wireless, set the essid that it should use.
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_set_essid (NMDevice *dev, const char *essid)
|
|
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
int err;
|
|
|
|
|
struct iwreq wreq;
|
|
|
|
|
unsigned char safe_essid[IW_ESSID_MAX_SIZE + 1] = "\0";
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/* Test devices directly set cur_essid */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
|
|
|
|
if (dev->options.wireless.cur_essid)
|
|
|
|
|
g_free (dev->options.wireless.cur_essid);
|
|
|
|
|
dev->options.wireless.cur_essid = g_strdup (essid);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/* Make sure the essid we get passed is a valid size */
|
|
|
|
|
if (!essid)
|
|
|
|
|
safe_essid[0] = '\0';
|
|
|
|
|
else
|
|
|
|
|
{
|
2005-10-09 04:46:36 +00:00
|
|
|
strncpy ((char *) safe_essid, essid, IW_ESSID_MAX_SIZE);
|
2004-06-24 14:18:37 +00:00
|
|
|
safe_essid[IW_ESSID_MAX_SIZE] = '\0';
|
|
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
wreq.u.essid.pointer = (caddr_t) safe_essid;
|
2005-10-09 04:46:36 +00:00
|
|
|
wreq.u.essid.length = strlen ((char *) safe_essid) + 1;
|
2004-06-24 14:18:37 +00:00
|
|
|
wreq.u.essid.flags = 1; /* Enable essid on card */
|
|
|
|
|
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to SET IWESSID.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
|
|
|
|
if ((err = iw_set_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCSIWESSID, &wreq)) == -1)
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
{
|
|
|
|
|
if (errno != ENODEV)
|
|
|
|
|
nm_warning ("nm_device_set_essid(): error setting ESSID '%s' for device %s. errno = %d", safe_essid, nm_device_get_iface (dev), errno);
|
|
|
|
|
}
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2005-03-29 22:15:28 +00:00
|
|
|
|
|
|
|
|
/* Orinoco cards seem to need extra time here to not screw
|
|
|
|
|
* up the firmware, which reboots when you set the ESSID.
|
|
|
|
|
* Unfortunately, there's no way to know when the card is back up
|
|
|
|
|
* again. Sigh...
|
|
|
|
|
*/
|
|
|
|
|
sleep (2);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_frequency
|
|
|
|
|
*
|
|
|
|
|
* For wireless devices, get the frequency we broadcast/receive on.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static double nm_device_get_frequency (NMDevice *dev)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
int err;
|
|
|
|
|
double freq = 0;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), 0);
|
|
|
|
|
|
|
|
|
|
/* Test devices don't really have a frequency, they always succeed */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
return 703000000;
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IWFREQ.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = iw_get_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCGIWFREQ, &wrq);
|
2004-12-17 17:16:22 +00:00
|
|
|
if (err >= 0)
|
|
|
|
|
freq = iw_freq2float (&wrq.u.freq);
|
|
|
|
|
if (err == -1)
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_warning ("nm_device_get_frequency(): error getting frequency for device %s. errno = %d", nm_device_get_iface (dev), errno);
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
return (freq);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_set_frequency
|
|
|
|
|
*
|
|
|
|
|
* For wireless devices, set the frequency to broadcast/receive on.
|
2005-01-10 06:41:02 +00:00
|
|
|
* A frequency <= 0 means "auto".
|
2004-12-17 17:16:22 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static void nm_device_set_frequency (NMDevice *dev, const double freq)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
int err;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
/* HACK FOR NOW */
|
|
|
|
|
if (freq <= 0)
|
|
|
|
|
return;
|
|
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
|
|
|
|
/* Test devices don't really have a frequency, they always succeed */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
return;
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
if (nm_device_get_frequency (dev) == freq)
|
|
|
|
|
return;
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
2005-01-10 06:41:02 +00:00
|
|
|
if (freq <= 0)
|
|
|
|
|
{
|
|
|
|
|
/* Auto */
|
|
|
|
|
/* People like to make things hard for us. Even though iwlib/iwconfig say
|
|
|
|
|
* that wrq.u.freq.m should be -1 for "auto" mode, nobody actually supports
|
|
|
|
|
* that. Madwifi actually uses "0" to mean "auto". So, we'll try 0 first
|
|
|
|
|
* and if that doesn't work, fall back to the iwconfig method and use -1.
|
|
|
|
|
*
|
|
|
|
|
* As a further note, it appears that Atheros/Madwifi cards can't go back to
|
|
|
|
|
* any-channel operation once you force set the channel on them. For example,
|
|
|
|
|
* if you set a prism54 card to a specific channel, but then set the ESSID to
|
|
|
|
|
* something else later, it will scan for the ESSID and switch channels just fine.
|
|
|
|
|
* Atheros cards, however, just stay at the channel you previously set and don't
|
|
|
|
|
* budge, no matter what you do to them, until you tell them to go back to
|
|
|
|
|
* any-channel operation.
|
|
|
|
|
*/
|
|
|
|
|
wrq.u.freq.m = 0;
|
|
|
|
|
wrq.u.freq.e = 0;
|
|
|
|
|
wrq.u.freq.flags = 0;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Fixed */
|
|
|
|
|
wrq.u.freq.flags = IW_FREQ_FIXED;
|
|
|
|
|
iw_float2freq (freq, &wrq.u.freq);
|
|
|
|
|
}
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to SET IWFREQ.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
|
|
|
|
if ((err = iw_set_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCSIWFREQ, &wrq)) == -1)
|
2005-01-10 06:41:02 +00:00
|
|
|
{
|
|
|
|
|
gboolean success = FALSE;
|
2005-01-10 15:22:21 +00:00
|
|
|
if ((freq <= 0) && ((errno == EINVAL) || (errno == EOPNOTSUPP)))
|
2005-01-10 06:41:02 +00:00
|
|
|
{
|
|
|
|
|
/* Ok, try "auto" the iwconfig way if the Atheros way didn't work */
|
2005-01-10 15:22:21 +00:00
|
|
|
wrq.u.freq.m = -1;
|
2005-01-10 06:41:02 +00:00
|
|
|
wrq.u.freq.e = 0;
|
|
|
|
|
wrq.u.freq.flags = 0;
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_set_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCSIWFREQ, &wrq) != -1)
|
2005-01-10 06:41:02 +00:00
|
|
|
success = TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_get_bitrate
|
|
|
|
|
*
|
|
|
|
|
* For wireless devices, get the bitrate to broadcast/receive at.
|
|
|
|
|
* Returned value is rate in KHz.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static int nm_device_get_bitrate (NMDevice *dev)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
int err = -1;
|
|
|
|
|
struct iwreq wrq;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), 0);
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
/* Test devices don't really have a bitrate, they always succeed */
|
2004-12-17 17:16:22 +00:00
|
|
|
if (dev->test_device)
|
|
|
|
|
return 11;
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IWRATE.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = iw_get_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCGIWRATE, &wrq);
|
|
|
|
|
nm_dev_sock_close (sk);
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return ((err >= 0) ? wrq.u.bitrate.value / 1000 : 0);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_set_bitrate
|
|
|
|
|
*
|
|
|
|
|
* For wireless devices, set the bitrate to broadcast/receive at.
|
2005-01-21 19:32:08 +00:00
|
|
|
* Rate argument should be in Mbps (mega-bits per second), or 0 for automatic.
|
2004-12-17 17:16:22 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static void nm_device_set_bitrate (NMDevice *dev, const int Mbps)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Test devices don't really have a bitrate, they always succeed */
|
2004-12-17 17:16:22 +00:00
|
|
|
if (dev->test_device)
|
|
|
|
|
return;
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (nm_device_get_bitrate (dev) == Mbps)
|
2005-01-10 02:14:04 +00:00
|
|
|
return;
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (Mbps != 0)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
wrq.u.bitrate.value = Mbps * 1000;
|
2004-12-17 17:16:22 +00:00
|
|
|
wrq.u.bitrate.fixed = 1;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
2004-12-22 18:42:37 +00:00
|
|
|
/* Auto bitrate */
|
2004-12-17 17:16:22 +00:00
|
|
|
wrq.u.bitrate.value = -1;
|
|
|
|
|
wrq.u.bitrate.fixed = 0;
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Silently fail as not all drivers support setting bitrate yet (ipw2x00 for example) */
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to SET IWRATE.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
iw_set_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCSIWRATE, &wrq);
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_ap_address
|
|
|
|
|
*
|
|
|
|
|
* If a device is wireless, get the access point's ethernet address
|
|
|
|
|
* that the card is associated with.
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_get_ap_address (NMDevice *dev, struct ether_addr *addr)
|
|
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
struct iwreq wrq;
|
2004-07-06 01:34:10 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (addr != NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
memset (addr, 0, sizeof (struct ether_addr));
|
|
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/* Test devices return an invalid address when there's no link,
|
|
|
|
|
* and a made-up address when there is a link.
|
|
|
|
|
*/
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
|
|
|
|
struct ether_addr good_addr = { {0x70, 0x37, 0x03, 0x70, 0x37, 0x03} };
|
|
|
|
|
struct ether_addr bad_addr = { {0x00, 0x00, 0x00, 0x00, 0x00, 0x00} };
|
2005-03-11 20:12:57 +00:00
|
|
|
gboolean link = nm_device_has_active_link (dev);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
memcpy ((link ? &good_addr : &bad_addr), &(wrq.u.ap_addr.sa_data), sizeof (struct ether_addr));
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
|
|
|
|
{
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IWAP.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_get_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCGIWAP, &wrq) >= 0)
|
|
|
|
|
memcpy (addr, &(wrq.u.ap_addr.sa_data), sizeof (struct ether_addr));
|
|
|
|
|
nm_dev_sock_close (sk);
|
|
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
2004-08-24 Dan Williams <dcbw@redhat.com>
* src/NetworkManagerAP.[ch]
- Add a "enc_method_good" member and accessors to an Access Point
to signal when we've found the correct encryption method
for an access point
- Add a "timestamp" member and accessors, remove "priority" member
and accessors (use timestamps instead)
- Rename "wep_key"->"enc_key"
- (nm_ap_get_enc_key_hashed): new, return the correct mangled key
for a specified encryption method using the access points
source encryption key/passphrase
* src/NetworkManagerAPList.c
- When updating a network with dbus, grab timestamp now instead of
priority
* src/NetworkManagerDBus.[ch]
- Add signal for "DeviceActivating"
- Switch priority->timestamp
* src/NetworkManagerDevice.c
- Change references of "wep_key" -> "enc_key" or "key"
- Signal DeviceActivating when starting activation
- When activating a wireless device, if the access point we are connecting
to is encrypted, and we have a source key, try to generate a mangled
key and use that (ie, generate real WEP key from a passphrase)
- Rework device activation to fallback to other encryption methods if
a previous one didn't work (ie, try mangling a key as a 104-bit passphrase
first, then if that doesn't work fall back to direct hex key).
- (nm_device_update_best_ap): fix a deadlock, and use timestamps instead of
priority. We now prefer the latest access point used, rather than using
a priority scheme
- (nm_device_do_normal_scan): make the encryption method "unknown" on access
points we've just discovered, and merge in correct info from the global
access point lists
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@68 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-25 22:41:12 +00:00
|
|
|
* nm_device_set_enc_key
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
2004-08-24 Dan Williams <dcbw@redhat.com>
* src/NetworkManagerAP.[ch]
- Add a "enc_method_good" member and accessors to an Access Point
to signal when we've found the correct encryption method
for an access point
- Add a "timestamp" member and accessors, remove "priority" member
and accessors (use timestamps instead)
- Rename "wep_key"->"enc_key"
- (nm_ap_get_enc_key_hashed): new, return the correct mangled key
for a specified encryption method using the access points
source encryption key/passphrase
* src/NetworkManagerAPList.c
- When updating a network with dbus, grab timestamp now instead of
priority
* src/NetworkManagerDBus.[ch]
- Add signal for "DeviceActivating"
- Switch priority->timestamp
* src/NetworkManagerDevice.c
- Change references of "wep_key" -> "enc_key" or "key"
- Signal DeviceActivating when starting activation
- When activating a wireless device, if the access point we are connecting
to is encrypted, and we have a source key, try to generate a mangled
key and use that (ie, generate real WEP key from a passphrase)
- Rework device activation to fallback to other encryption methods if
a previous one didn't work (ie, try mangling a key as a 104-bit passphrase
first, then if that doesn't work fall back to direct hex key).
- (nm_device_update_best_ap): fix a deadlock, and use timestamps instead of
priority. We now prefer the latest access point used, rather than using
a priority scheme
- (nm_device_do_normal_scan): make the encryption method "unknown" on access
points we've just discovered, and merge in correct info from the global
access point lists
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@68 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-25 22:41:12 +00:00
|
|
|
* If a device is wireless, set the encryption key that it should use.
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
2004-08-24 Dan Williams <dcbw@redhat.com>
* src/NetworkManagerAP.[ch]
- Add a "enc_method_good" member and accessors to an Access Point
to signal when we've found the correct encryption method
for an access point
- Add a "timestamp" member and accessors, remove "priority" member
and accessors (use timestamps instead)
- Rename "wep_key"->"enc_key"
- (nm_ap_get_enc_key_hashed): new, return the correct mangled key
for a specified encryption method using the access points
source encryption key/passphrase
* src/NetworkManagerAPList.c
- When updating a network with dbus, grab timestamp now instead of
priority
* src/NetworkManagerDBus.[ch]
- Add signal for "DeviceActivating"
- Switch priority->timestamp
* src/NetworkManagerDevice.c
- Change references of "wep_key" -> "enc_key" or "key"
- Signal DeviceActivating when starting activation
- When activating a wireless device, if the access point we are connecting
to is encrypted, and we have a source key, try to generate a mangled
key and use that (ie, generate real WEP key from a passphrase)
- Rework device activation to fallback to other encryption methods if
a previous one didn't work (ie, try mangling a key as a 104-bit passphrase
first, then if that doesn't work fall back to direct hex key).
- (nm_device_update_best_ap): fix a deadlock, and use timestamps instead of
priority. We now prefer the latest access point used, rather than using
a priority scheme
- (nm_device_do_normal_scan): make the encryption method "unknown" on access
points we've just discovered, and merge in correct info from the global
access point lists
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@68 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-25 22:41:12 +00:00
|
|
|
* key: encryption key to use, or NULL or "" to disable encryption.
|
|
|
|
|
* NOTE that at this time, the key must be the raw HEX key, not
|
|
|
|
|
* a passphrase.
|
2004-06-24 14:18:37 +00:00
|
|
|
*/
|
2004-11-03 07:47:32 +00:00
|
|
|
void nm_device_set_enc_key (NMDevice *dev, const char *key, NMDeviceAuthMethod auth_method)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
2004-06-24 14:18:37 +00:00
|
|
|
struct iwreq wreq;
|
|
|
|
|
int keylen;
|
2004-07-06 04:45:00 +00:00
|
|
|
unsigned char safe_key[IW_ENCODING_TOKEN_MAX + 1];
|
2004-06-24 14:18:37 +00:00
|
|
|
gboolean set_key = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/* Test devices just ignore encryption keys */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
return;
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/* Make sure the essid we get passed is a valid size */
|
2004-08-24 Dan Williams <dcbw@redhat.com>
* src/NetworkManagerAP.[ch]
- Add a "enc_method_good" member and accessors to an Access Point
to signal when we've found the correct encryption method
for an access point
- Add a "timestamp" member and accessors, remove "priority" member
and accessors (use timestamps instead)
- Rename "wep_key"->"enc_key"
- (nm_ap_get_enc_key_hashed): new, return the correct mangled key
for a specified encryption method using the access points
source encryption key/passphrase
* src/NetworkManagerAPList.c
- When updating a network with dbus, grab timestamp now instead of
priority
* src/NetworkManagerDBus.[ch]
- Add signal for "DeviceActivating"
- Switch priority->timestamp
* src/NetworkManagerDevice.c
- Change references of "wep_key" -> "enc_key" or "key"
- Signal DeviceActivating when starting activation
- When activating a wireless device, if the access point we are connecting
to is encrypted, and we have a source key, try to generate a mangled
key and use that (ie, generate real WEP key from a passphrase)
- Rework device activation to fallback to other encryption methods if
a previous one didn't work (ie, try mangling a key as a 104-bit passphrase
first, then if that doesn't work fall back to direct hex key).
- (nm_device_update_best_ap): fix a deadlock, and use timestamps instead of
priority. We now prefer the latest access point used, rather than using
a priority scheme
- (nm_device_do_normal_scan): make the encryption method "unknown" on access
points we've just discovered, and merge in correct info from the global
access point lists
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@68 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-25 22:41:12 +00:00
|
|
|
if (!key)
|
2004-06-24 14:18:37 +00:00
|
|
|
safe_key[0] = '\0';
|
|
|
|
|
else
|
|
|
|
|
{
|
2005-10-09 04:46:36 +00:00
|
|
|
strncpy ((char *) safe_key, key, IW_ENCODING_TOKEN_MAX);
|
2004-06-24 14:18:37 +00:00
|
|
|
safe_key[IW_ENCODING_TOKEN_MAX] = '\0';
|
|
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
wreq.u.data.pointer = (caddr_t) NULL;
|
|
|
|
|
wreq.u.data.length = 0;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
wreq.u.data.flags = IW_ENCODE_ENABLED;
|
|
|
|
|
|
|
|
|
|
/* Unfortunately, some drivers (Cisco) don't make a distinction between
|
|
|
|
|
* Open System authentication mode and whether or not to use WEP. You
|
|
|
|
|
* DON'T have to use WEP when using Open System, but these cards force
|
|
|
|
|
* it. Therefore, we have to set Open System mode when using WEP.
|
|
|
|
|
*/
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-10-09 04:46:36 +00:00
|
|
|
if (strlen ((char *) safe_key) == 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
wreq.u.data.flags |= IW_ENCODE_DISABLED | IW_ENCODE_NOKEY;
|
2004-06-24 14:18:37 +00:00
|
|
|
set_key = TRUE;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
2004-07-06 04:45:00 +00:00
|
|
|
unsigned char parsed_key[IW_ENCODING_TOKEN_MAX + 1];
|
|
|
|
|
|
2005-10-09 04:46:36 +00:00
|
|
|
keylen = iw_in_key_full (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev),
|
|
|
|
|
(char *) safe_key, &parsed_key[0], &wreq.u.data.flags);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (keylen > 0)
|
|
|
|
|
{
|
2004-11-03 07:47:32 +00:00
|
|
|
switch (auth_method)
|
|
|
|
|
{
|
|
|
|
|
case NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM:
|
|
|
|
|
wreq.u.data.flags |= IW_ENCODE_OPEN;
|
|
|
|
|
break;
|
|
|
|
|
case NM_DEVICE_AUTH_METHOD_SHARED_KEY:
|
|
|
|
|
wreq.u.data.flags |= IW_ENCODE_RESTRICTED;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
wreq.u.data.flags |= IW_ENCODE_RESTRICTED;
|
|
|
|
|
break;
|
|
|
|
|
}
|
2004-07-06 04:45:00 +00:00
|
|
|
wreq.u.data.pointer = (caddr_t) &parsed_key;
|
2004-06-24 14:18:37 +00:00
|
|
|
wreq.u.data.length = keylen;
|
|
|
|
|
set_key = TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (set_key)
|
|
|
|
|
{
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to SET IWENCODE.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_set_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCSIWENCODE, &wreq) == -1)
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
{
|
|
|
|
|
if (errno != ENODEV)
|
|
|
|
|
nm_warning ("nm_device_set_enc_key(): error setting key for device %s. errno = %d", nm_device_get_iface (dev), errno);
|
|
|
|
|
}
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
} else nm_warning ("nm_device_set_enc_key(): could not get wireless control socket.");
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-09-13 17:43:16 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_signal_strength
|
|
|
|
|
*
|
|
|
|
|
* Get the current signal strength of a wireless device. This only works when
|
|
|
|
|
* the card is associated with an access point, so will only work for the
|
|
|
|
|
* active device.
|
|
|
|
|
*
|
|
|
|
|
* Returns: -1 on error
|
|
|
|
|
* 0 - 100 strength percentage of the connection to the current access point
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
gint8 nm_device_get_signal_strength (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, -1);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), -1);
|
|
|
|
|
|
|
|
|
|
return (dev->options.wireless.strength);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_update_signal_strength
|
|
|
|
|
*
|
|
|
|
|
* Update the device's idea of the strength of its connection to the
|
|
|
|
|
* current access point.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_update_signal_strength (NMDevice *dev)
|
|
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
gboolean has_range = FALSE;
|
|
|
|
|
NMSock * sk;
|
|
|
|
|
iwrange range;
|
|
|
|
|
iwstats stats;
|
|
|
|
|
int percent = -1;
|
2004-09-13 17:43:16 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
/* Grab the scan lock since our strength is meaningless during a scan. */
|
|
|
|
|
if (!nm_try_acquire_mutex (dev->options.wireless.scan_mutex, __FUNCTION__))
|
|
|
|
|
return;
|
|
|
|
|
|
2004-09-13 17:43:16 +00:00
|
|
|
/* If we aren't the active device, we don't really have a signal strength
|
|
|
|
|
* that would mean anything.
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (!dev->act_request)
|
2004-09-13 17:43:16 +00:00
|
|
|
{
|
|
|
|
|
dev->options.wireless.strength = -1;
|
2005-01-31 19:53:38 +00:00
|
|
|
goto out;
|
2004-09-13 17:43:16 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Fake a value for test devices */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
|
|
|
|
dev->options.wireless.strength = 75;
|
2005-01-31 19:53:38 +00:00
|
|
|
goto out;
|
2004-09-13 17:43:16 +00:00
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-09-13 17:43:16 +00:00
|
|
|
{
|
2005-04-27 18:05:16 +00:00
|
|
|
memset (&range, 0, sizeof (iwrange));
|
|
|
|
|
memset (&stats, 0, sizeof (iwstats));
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET 'iwrange'.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
has_range = (iw_get_range_info (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), &range) >= 0);
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET 'iwstats'.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_get_stats (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), &stats, &range, has_range) == 0)
|
|
|
|
|
{
|
|
|
|
|
percent = nm_wireless_qual_to_percent (&stats.qual, (const iwqual *)(&dev->options.wireless.max_qual),
|
|
|
|
|
(const iwqual *)(&dev->options.wireless.avg_qual));
|
|
|
|
|
}
|
|
|
|
|
nm_dev_sock_close (sk);
|
2004-09-13 17:43:16 +00:00
|
|
|
}
|
|
|
|
|
|
2004-11-03 17:05:16 +00:00
|
|
|
/* Try to smooth out the strength. Atmel cards, for example, will give no strength
|
|
|
|
|
* one second and normal strength the next.
|
|
|
|
|
*/
|
|
|
|
|
if ((percent == -1) && (++dev->options.wireless.invalid_strength_counter <= 3))
|
|
|
|
|
percent = dev->options.wireless.strength;
|
|
|
|
|
else
|
|
|
|
|
dev->options.wireless.invalid_strength_counter = 0;
|
|
|
|
|
|
2005-09-06 19:04:09 +00:00
|
|
|
if (percent != dev->options.wireless.strength)
|
|
|
|
|
nm_dbus_signal_device_strength_change (dev->app_data->dbus_connection, dev, percent);
|
|
|
|
|
|
2004-09-13 17:43:16 +00:00
|
|
|
dev->options.wireless.strength = percent;
|
2004-10-04 15:44:00 +00:00
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
out:
|
|
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2004-10-04 15:44:00 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_ip4_address
|
|
|
|
|
*
|
|
|
|
|
* Get a device's IPv4 address
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
guint32 nm_device_get_ip4_address(NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
2004-07-06 04:45:00 +00:00
|
|
|
|
|
|
|
|
return (dev->ip4_address);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_update_ip4_address (NMDevice *dev)
|
|
|
|
|
{
|
2004-07-27 16:15:36 +00:00
|
|
|
guint32 new_address;
|
|
|
|
|
struct ifreq req;
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
2004-07-27 16:15:36 +00:00
|
|
|
int err;
|
2004-07-06 04:45:00 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
2004-07-25 02:40:19 +00:00
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
2004-07-06 04:45:00 +00:00
|
|
|
g_return_if_fail (nm_device_get_iface (dev) != NULL);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/* Test devices get a nice, bogus IP address */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
|
|
|
|
dev->ip4_address = 0x07030703;
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_GENERAL, __FUNCTION__, NULL)) == NULL)
|
2004-07-06 04:45:00 +00:00
|
|
|
return;
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
memset (&req, 0, sizeof (struct ifreq));
|
|
|
|
|
strncpy ((char *)(&req.ifr_name), nm_device_get_iface (dev), strlen (nm_device_get_iface (dev)));
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IFADDR.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = ioctl (nm_dev_sock_get_fd (sk), SIOCGIFADDR, &req);
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with GET IFADDR.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-07-27 16:15:36 +00:00
|
|
|
if (err != 0)
|
2004-07-06 04:45:00 +00:00
|
|
|
return;
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-07-06 04:45:00 +00:00
|
|
|
new_address = ((struct sockaddr_in *)(&req.ifr_addr))->sin_addr.s_addr;
|
|
|
|
|
if (new_address != nm_device_get_ip4_address (dev))
|
|
|
|
|
dev->ip4_address = new_address;
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_get_ip6_address
|
|
|
|
|
*
|
|
|
|
|
* Get a device's IPv6 address
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_get_ip6_address(NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
/* FIXME
|
|
|
|
|
* Implement
|
|
|
|
|
*/
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_hw_address
|
|
|
|
|
*
|
|
|
|
|
* Get a device's hardware address
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
void nm_device_get_hw_address (NMDevice *dev, struct ether_addr *addr)
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (addr != NULL);
|
2004-11-16 02:41:53 +00:00
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
memcpy (addr, &(dev->hw_addr), sizeof (struct ether_addr));
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_update_hw_address (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
struct ifreq req;
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
2004-11-16 02:41:53 +00:00
|
|
|
int err;
|
2005-01-10 02:14:04 +00:00
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_get_iface (dev) != NULL);
|
|
|
|
|
|
|
|
|
|
/* Test devices get a nice, bogus IP address */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
memset (&(dev->hw_addr), 0, sizeof (struct ether_addr));
|
2004-11-16 02:41:53 +00:00
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_GENERAL, __FUNCTION__, NULL)) == NULL)
|
2004-11-16 02:41:53 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
memset (&req, 0, sizeof (struct ifreq));
|
|
|
|
|
strncpy ((char *)(&req.ifr_name), nm_device_get_iface (dev), strlen (nm_device_get_iface (dev)));
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IFHWADDR.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = ioctl (nm_dev_sock_get_fd (sk), SIOCGIFHWADDR, &req);
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with GET IFHWADDR.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-11-16 02:41:53 +00:00
|
|
|
if (err != 0)
|
|
|
|
|
return;
|
|
|
|
|
|
2005-10-04 15:54:04 +00:00
|
|
|
memcpy (&(dev->hw_addr), &(req.ifr_hwaddr.sa_data), sizeof (struct ether_addr));
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_set_up_down
|
|
|
|
|
*
|
|
|
|
|
* Set the up flag on the device on or off
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static void nm_device_set_up_down (NMDevice *dev, gboolean up)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/* Test devices do whatever we tell them to do */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
|
|
|
|
dev->test_device_up = up;
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
nm_system_device_set_up_down (dev, up);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
/* Make sure we have a valid MAC address, some cards reload firmware when they
|
|
|
|
|
* are brought up.
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (!nm_ethernet_address_is_valid (&(dev->hw_addr)))
|
2005-04-15 15:43:42 +00:00
|
|
|
nm_device_update_hw_address (dev);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Interface state functions: bring up, down, check
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_is_up (NMDevice *dev)
|
|
|
|
|
{
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
NMSock * sk;
|
2004-06-24 14:18:37 +00:00
|
|
|
struct ifreq ifr;
|
|
|
|
|
int err;
|
|
|
|
|
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
if (dev->test_device)
|
|
|
|
|
return (dev->test_device_up);
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_GENERAL, __FUNCTION__, NULL)) == NULL)
|
2004-06-24 14:18:37 +00:00
|
|
|
return (FALSE);
|
|
|
|
|
|
|
|
|
|
/* Get device's flags */
|
|
|
|
|
strcpy (ifr.ifr_name, nm_device_get_iface (dev));
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IFFLAGS.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
err = ioctl (nm_dev_sock_get_fd (sk), SIOCGIFFLAGS, &ifr);
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with GET IFFLAGS.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (!err)
|
|
|
|
|
return (!((ifr.ifr_flags^IFF_UP) & IFF_UP));
|
|
|
|
|
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
if (errno != ENODEV)
|
|
|
|
|
nm_warning ("nm_device_is_up() could not get flags for device %s. errno = %d", nm_device_get_iface (dev), errno );
|
|
|
|
|
|
|
|
|
|
return FALSE;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
2005-03-26 03:42:05 +00:00
|
|
|
/* I really wish nm_v_wait_for_completion_or_timeout could translate these
|
2005-04-28 02:50:54 +00:00
|
|
|
* to first class args instead of a all this void * arg stuff, so these
|
|
|
|
|
* helpers could be nice and _tiny_. */
|
2005-06-21 15:07:01 +00:00
|
|
|
static gboolean nm_completion_device_is_up_test (int tries, nm_completion_args args)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-04-28 02:50:54 +00:00
|
|
|
NMDevice *dev = args[0];
|
|
|
|
|
gboolean *err = args[1];
|
2005-06-10 03:47:47 +00:00
|
|
|
gboolean cancelable = GPOINTER_TO_INT (args[2]);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-04-27 18:05:16 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
g_return_val_if_fail (err != NULL, TRUE);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
*err = FALSE;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (cancelable && nm_device_activation_should_cancel (dev)) {
|
2005-03-26 03:42:05 +00:00
|
|
|
*err = TRUE;
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
if (nm_device_is_up (dev))
|
|
|
|
|
return TRUE;
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_bring_up (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
nm_device_set_up_down (dev, TRUE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
gboolean nm_device_bring_up_wait (NMDevice *dev, gboolean cancelable)
|
|
|
|
|
{
|
|
|
|
|
gboolean err = FALSE;
|
2005-04-28 02:50:54 +00:00
|
|
|
nm_completion_args args;
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
|
|
|
|
|
nm_device_bring_up (dev);
|
2005-04-28 02:50:54 +00:00
|
|
|
|
|
|
|
|
args[0] = dev;
|
|
|
|
|
args[1] = &err;
|
2005-06-10 03:47:47 +00:00
|
|
|
args[2] = GINT_TO_POINTER (cancelable);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_wait_for_completion (400, G_USEC_PER_SEC / 200, NULL, nm_completion_device_is_up_test, args);
|
2005-03-26 03:42:05 +00:00
|
|
|
if (err)
|
2005-06-26 17:18:35 +00:00
|
|
|
nm_info ("failed to bring up device %s", dev->iface);
|
2005-03-26 03:42:05 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_bring_down (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
nm_device_set_up_down (dev, FALSE);
|
|
|
|
|
}
|
|
|
|
|
|
2005-06-21 15:07:01 +00:00
|
|
|
static gboolean nm_completion_device_is_down_test (int tries, nm_completion_args args)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-04-28 02:50:54 +00:00
|
|
|
NMDevice *dev = args[0];
|
|
|
|
|
gboolean *err = args[1];
|
2005-06-10 03:47:47 +00:00
|
|
|
gboolean cancelable = GPOINTER_TO_INT (args[2]);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-04-27 18:05:16 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
g_return_val_if_fail (err != NULL, TRUE);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
*err = FALSE;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (cancelable && nm_device_activation_should_cancel (dev)) {
|
2005-03-26 03:42:05 +00:00
|
|
|
*err = TRUE;
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
if (!nm_device_is_up (dev))
|
|
|
|
|
return TRUE;
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
2005-06-21 15:07:01 +00:00
|
|
|
static gboolean nm_device_bring_down_wait (NMDevice *dev, gboolean cancelable)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
|
|
|
|
gboolean err = FALSE;
|
2005-04-28 02:50:54 +00:00
|
|
|
nm_completion_args args;
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
|
|
|
|
|
nm_device_bring_down (dev);
|
2005-04-28 02:50:54 +00:00
|
|
|
|
|
|
|
|
args[0] = dev;
|
|
|
|
|
args[1] = &err;
|
2005-06-10 03:47:47 +00:00
|
|
|
args[2] = GINT_TO_POINTER (cancelable);
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_wait_for_completion(400, G_USEC_PER_SEC / 200, NULL,
|
2005-04-28 02:50:54 +00:00
|
|
|
nm_completion_device_is_down_test, args);
|
2005-03-26 03:42:05 +00:00
|
|
|
if (err)
|
2005-06-26 17:18:35 +00:00
|
|
|
nm_info ("failed to bring down device %s", dev->iface);
|
2005-03-26 03:42:05 +00:00
|
|
|
return err;
|
|
|
|
|
}
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_mode
|
|
|
|
|
*
|
|
|
|
|
* Get managed/infrastructure/adhoc mode on a device (currently wireless only)
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMNetworkMode nm_device_get_mode (NMDevice *dev)
|
|
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
2004-12-17 17:16:22 +00:00
|
|
|
NMNetworkMode mode = NETWORK_MODE_UNKNOWN;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NETWORK_MODE_UNKNOWN);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NETWORK_MODE_UNKNOWN);
|
|
|
|
|
|
|
|
|
|
/* Force the card into Managed/Infrastructure mode */
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
2005-04-27 18:05:16 +00:00
|
|
|
memset (&wrq, 0, sizeof (struct iwreq));
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to GET IWMODE.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_get_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCGIWMODE, &wrq) == 0)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
switch (wrq.u.mode)
|
|
|
|
|
{
|
|
|
|
|
case IW_MODE_INFRA:
|
|
|
|
|
mode = NETWORK_MODE_INFRA;
|
|
|
|
|
break;
|
|
|
|
|
case IW_MODE_ADHOC:
|
|
|
|
|
mode = NETWORK_MODE_ADHOC;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
else
|
2005-03-29 21:44:38 +00:00
|
|
|
nm_warning ("nm_device_get_mode (%s): error getting card mode. errno = %d", nm_device_get_iface (dev), errno);
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (mode);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-10-12 16:44:58 +00:00
|
|
|
/*
|
2004-12-16 04:24:20 +00:00
|
|
|
* nm_device_set_mode
|
2004-10-12 16:44:58 +00:00
|
|
|
*
|
2004-12-16 04:24:20 +00:00
|
|
|
* Set managed/infrastructure/adhoc mode on a device (currently wireless only)
|
2004-10-12 16:44:58 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2004-12-16 04:24:20 +00:00
|
|
|
gboolean nm_device_set_mode (NMDevice *dev, const NMNetworkMode mode)
|
2004-10-12 16:44:58 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
gboolean success = FALSE;
|
2004-10-12 16:44:58 +00:00
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), FALSE);
|
|
|
|
|
g_return_val_if_fail ((mode == NETWORK_MODE_INFRA) || (mode == NETWORK_MODE_ADHOC), FALSE);
|
2004-10-12 16:44:58 +00:00
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
if (nm_device_get_mode (dev) == mode)
|
|
|
|
|
return TRUE;
|
|
|
|
|
|
2004-10-12 16:44:58 +00:00
|
|
|
/* Force the card into Managed/Infrastructure mode */
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2004-10-12 16:44:58 +00:00
|
|
|
{
|
2004-12-16 04:24:20 +00:00
|
|
|
struct iwreq wreq;
|
|
|
|
|
gboolean mode_good = FALSE;
|
2004-10-12 16:44:58 +00:00
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
switch (mode)
|
|
|
|
|
{
|
|
|
|
|
case NETWORK_MODE_INFRA:
|
|
|
|
|
wreq.u.mode = IW_MODE_INFRA;
|
|
|
|
|
mode_good = TRUE;
|
|
|
|
|
break;
|
|
|
|
|
case NETWORK_MODE_ADHOC:
|
|
|
|
|
wreq.u.mode = IW_MODE_ADHOC;
|
|
|
|
|
mode_good = TRUE;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
mode_good = FALSE;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
if (mode_good)
|
|
|
|
|
{
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("%s: About to SET IWMODE.", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (iw_set_ext (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), SIOCSIWMODE, &wreq) == 0)
|
2004-12-16 04:24:20 +00:00
|
|
|
success = TRUE;
|
|
|
|
|
else
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
{
|
|
|
|
|
if (errno != ENODEV)
|
|
|
|
|
nm_warning ("nm_device_set_mode (%s): error setting card to %s mode. errno = %d",
|
|
|
|
|
nm_device_get_iface (dev),
|
|
|
|
|
mode == NETWORK_MODE_INFRA ? "Infrastructure" : (mode == NETWORK_MODE_ADHOC ? "adhoc" : "unknown"),
|
|
|
|
|
errno);
|
|
|
|
|
}
|
2004-12-16 04:24:20 +00:00
|
|
|
}
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2004-10-12 16:44:58 +00:00
|
|
|
}
|
2004-12-16 04:24:20 +00:00
|
|
|
|
|
|
|
|
return (success);
|
2004-10-12 16:44:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activation_start
|
2005-01-07 18:07:06 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Tell the device thread to begin activation.
|
|
|
|
|
*
|
|
|
|
|
* Returns: TRUE on success activation beginning
|
|
|
|
|
* FALSE on error beginning activation (bad params, couldn't create thread)
|
2005-01-07 18:07:06 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
gboolean nm_device_activation_start (NMActRequest *req)
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
g_assert (data);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
g_assert (dev);
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (!nm_device_is_activating (dev), TRUE); /* Return if activation has already begun */
|
|
|
|
|
|
|
|
|
|
nm_act_request_ref (req);
|
|
|
|
|
dev->act_request = req;
|
|
|
|
|
dev->quit_activation = FALSE;
|
|
|
|
|
|
|
|
|
|
nm_info ("Activation (%s) started...", nm_device_get_iface (dev));
|
|
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_DEVICE_PREPARE);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_activate_schedule_stage1_device_prepare (req);
|
|
|
|
|
|
|
|
|
|
nm_schedule_state_change_signal_broadcast (data);
|
2005-05-06 Dan Williams <dcbw@redhat.com>
* gnome/applet/applet-dbus-device.c
gnome/applet/applet-dbus-info.c
gnome/applet/applet-dbus.c
gnome/applet/applet.c
gnome/applet/applet.h
- (nmwa_get_device_for_nm_device) -> (nmwa_get_device_for_nm_path)
* gnome/applet/applet-dbus.c
- (nmwa_dbus_filter): trap DeviceCarrierOn/DeviceCarrierOff signals
so we notice when wired device's carriers come back on. Should
fix issue with wired devices being grayed out even if the cable
is in, for devices that support carrier detection.
* gnome/applet/applet.c
- (nmwa_driver_notify): bash focus-stealing prevention in the face
- (nmwa_act_stage_to_pixbuf): Clarify wireless ACT_STAGE_DEVICE_CONFIG
tooltip message
- (nmwa_menu_item_activate, nmwa_menu_add_device_item, nmwa_menu_item_data_free):
Fix situation where applet wouldn't respond to menu selections
* src/NetworkManager.c
src/NetworkManagerDevice.c
src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- (nm_dbus_signal_device_status_change) -> (nm_dbus_schedule_device_status_change_signal)
* src/NetworkManagerDbus.c
- (nm_dbus_send_network_not_found, nm_dbus_schedule_network_not_found_signal):
Remove, no longer used or relevant
- (nm_dbus_signal_device_status_change): Better signal enum->string matching
- (nm_dbus_schedule_device_status_change_signal): add
* src/NetworkManagerDevice.c
- (nm_device_worker_thread_stop): don't try to join a NULL worker thread
- (nm_device_set_link_active): Fix up switching for non-carrier-detect devices,
ie don't deactivate them unless explicitly told to by the user. Also send
CARRIER_OFF / CARRIER_ON signals when link changes
- (nm_device_set_essid, nm_device_set_enc_key, nm_device_is_up, nm_device_set_mode):
Don't print error message when device is no longer around
- (nm_device_deactivate): kill any current DHCP process attached to this device,
not just during activation
* src/NetworkManagerPolicy.c
- (nm_policy_auto_get_best_device): Ignore semi-supported devices completely from
auto-device-selection.
- (nm_policy_device_change_check): Don't interrupt semi-supported devices
* src/NetworkManagerSystem.c
- (nm_system_device_set_up_down_with_iface): Quiet first warning message when device
is no longer present (Bill Moss)
* src/backends/shvar.c
- (svOpenFile): Open read-only to make SELinux happy
* src/backends/NetworkManagerRedHat.c
- (nm_system_device_get_system_config): Use SYSCONFDIR rather than hardcoding
the path to the ifcfg-* files
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@613 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-06 21:20:42 +00:00
|
|
|
nm_dbus_schedule_device_status_change_signal (data, dev, NULL, DEVICE_ACTIVATING);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
return TRUE;
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activation_handle_cancel
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Cancel activation on a device and clean up.
|
2004-07-06 01:34:10 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static gboolean nm_device_activation_handle_cancel (NMActRequest *req)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMDevice * dev;
|
|
|
|
|
NMData * data;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if ((req = nm_device_get_act_request (dev)) && nm_device_is_activating (dev))
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev->act_request = NULL;
|
|
|
|
|
nm_act_request_unref (req);
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
2005-05-05 16:52:04 +00:00
|
|
|
nm_schedule_state_change_signal_broadcast (dev->app_data);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("Activation (%s) cancellation handled.", nm_device_get_iface (dev));
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_schedule_activation_handle_cancel
|
|
|
|
|
*
|
|
|
|
|
* Schedule the activation cancel handler
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static void nm_device_schedule_activation_handle_cancel (NMActRequest *req)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
{
|
|
|
|
|
NMDevice * dev;
|
|
|
|
|
NMData * data;
|
|
|
|
|
GSource * source;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (req != NULL);
|
|
|
|
|
|
|
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
nm_info ("Activation (%s) cancellation handler scheduled...", nm_device_get_iface (dev));
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
source = g_idle_source_new ();
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activation_handle_cancel, req, NULL);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
}
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2004-08-24 Dan Williams <dcbw@redhat.com>
* src/NetworkManagerAP.[ch]
- Add a "enc_method_good" member and accessors to an Access Point
to signal when we've found the correct encryption method
for an access point
- Add a "timestamp" member and accessors, remove "priority" member
and accessors (use timestamps instead)
- Rename "wep_key"->"enc_key"
- (nm_ap_get_enc_key_hashed): new, return the correct mangled key
for a specified encryption method using the access points
source encryption key/passphrase
* src/NetworkManagerAPList.c
- When updating a network with dbus, grab timestamp now instead of
priority
* src/NetworkManagerDBus.[ch]
- Add signal for "DeviceActivating"
- Switch priority->timestamp
* src/NetworkManagerDevice.c
- Change references of "wep_key" -> "enc_key" or "key"
- Signal DeviceActivating when starting activation
- When activating a wireless device, if the access point we are connecting
to is encrypted, and we have a source key, try to generate a mangled
key and use that (ie, generate real WEP key from a passphrase)
- Rework device activation to fallback to other encryption methods if
a previous one didn't work (ie, try mangling a key as a 104-bit passphrase
first, then if that doesn't work fall back to direct hex key).
- (nm_device_update_best_ap): fix a deadlock, and use timestamps instead of
priority. We now prefer the latest access point used, rather than using
a priority scheme
- (nm_device_do_normal_scan): make the encryption method "unknown" on access
points we've just discovered, and merge in correct info from the global
access point lists
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@68 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-25 22:41:12 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_act_request
|
|
|
|
|
*
|
|
|
|
|
* Return the devices activation request, if any.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMActRequest *nm_device_get_act_request (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
return dev->act_request;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-09-09 20:02:59 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* get_initial_auth_method
|
2004-09-09 20:02:59 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Update the auth method of the AP from the last-known-good one saved in the allowed list
|
|
|
|
|
* (which is found from NMI) and ensure that its valid with the encryption status of the AP.
|
2004-09-09 20:02:59 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static NMDeviceAuthMethod get_initial_auth_method (NMAccessPoint *ap, NMAccessPointList *allowed_list)
|
2004-09-09 20:02:59 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (ap != NULL, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
2004-09-09 20:02:59 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_ap_get_encrypted (ap))
|
2004-09-09 20:02:59 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMDeviceAuthMethod auth = nm_ap_get_auth_method (ap);
|
|
|
|
|
NMAccessPoint *allowed_ap = nm_ap_list_get_ap_by_essid (allowed_list, nm_ap_get_essid (ap));
|
|
|
|
|
|
|
|
|
|
/* Prefer default auth method if we found one for this AP in our allowed list. */
|
|
|
|
|
if (allowed_ap)
|
|
|
|
|
auth = nm_ap_get_auth_method (allowed_ap);
|
|
|
|
|
|
|
|
|
|
if ( (auth == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM)
|
|
|
|
|
|| (auth == NM_DEVICE_AUTH_METHOD_SHARED_KEY))
|
|
|
|
|
return (auth);
|
|
|
|
|
else
|
|
|
|
|
return (NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
2004-09-09 20:02:59 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return (NM_DEVICE_AUTH_METHOD_NONE);
|
2004-09-09 20:02:59 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_stage1_device_prepare
|
|
|
|
|
*
|
|
|
|
|
* Prepare for device activation
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_activate_stage1_device_prepare (NMActRequest *req)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMDevice * dev;
|
|
|
|
|
NMData * data;
|
|
|
|
|
NMAccessPoint * ap;
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
nm_info ("Activation (%s) Stage 1 (Device Prepare) started...", nm_device_get_iface (dev));
|
|
|
|
|
|
|
|
|
|
if (nm_device_is_wireless (dev))
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
ap = nm_act_request_get_ap (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
g_assert (ap);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
if (nm_ap_get_artificial (ap))
|
|
|
|
|
{
|
|
|
|
|
/* Some Cisco cards (340/350 PCMCIA) don't return non-broadcasting APs
|
|
|
|
|
* in their scan results, so we can't know beforehand whether or not the
|
|
|
|
|
* AP was encrypted. We have to update their encryption status on the fly.
|
|
|
|
|
*/
|
|
|
|
|
if (nm_ap_get_encrypted (ap) || nm_ap_is_enc_key_valid (ap))
|
|
|
|
|
{
|
|
|
|
|
nm_ap_set_encrypted (ap, TRUE);
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Initial authentication method */
|
|
|
|
|
nm_ap_set_auth_method (ap, get_initial_auth_method (ap, data->allowed_ap_list));
|
2005-03-26 03:42:05 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
else
|
|
|
|
|
nm_device_activate_schedule_stage2_device_config (req);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("Activation (%s) Stage 1 (Device Prepare) complete.", nm_device_get_iface (dev));
|
2005-03-26 03:42:05 +00:00
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activate_schedule_stage1_device_prepare
|
2005-01-13 19:04:26 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Prepare a device for activation
|
2005-01-13 19:04:26 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
void nm_device_activate_schedule_stage1_device_prepare (NMActRequest *req)
|
2005-01-13 19:04:26 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
GSource * source = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_DEVICE_PREPARE);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("Activation (%s) Stage 1 (Device Prepare) scheduled...", nm_device_get_iface (dev));
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activate_stage1_device_prepare, req, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
2005-03-26 03:42:05 +00:00
|
|
|
}
|
2005-01-13 19:04:26 +00:00
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static gboolean nm_device_link_test (int tries, nm_completion_args args)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-04-28 02:50:54 +00:00
|
|
|
NMDevice *dev = args[0];
|
|
|
|
|
gboolean *err = args[1];
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-04-27 18:05:16 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
g_return_val_if_fail (err != NULL, TRUE);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
if (nm_device_wireless_is_associated (dev) && nm_device_get_essid (dev))
|
|
|
|
|
{
|
|
|
|
|
*err = FALSE;
|
|
|
|
|
return TRUE;
|
2005-01-13 19:04:26 +00:00
|
|
|
}
|
2005-03-26 03:42:05 +00:00
|
|
|
*err = TRUE;
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static gboolean nm_device_is_up_and_associated_wait (NMDevice *dev, int timeout, int interval)
|
|
|
|
|
{
|
|
|
|
|
gboolean err;
|
|
|
|
|
const gint delay = (G_USEC_PER_SEC * nm_device_get_association_pause_value (dev)) / interval;
|
|
|
|
|
const gint max_cycles = timeout * interval;
|
2005-04-28 02:50:54 +00:00
|
|
|
nm_completion_args args;
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-03-26 03:42:05 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-04-28 02:50:54 +00:00
|
|
|
args[0] = dev;
|
|
|
|
|
args[1] = &err;
|
|
|
|
|
nm_wait_for_completion (max_cycles, delay, NULL, nm_device_link_test, args);
|
2005-03-26 03:42:05 +00:00
|
|
|
return !err;
|
2005-01-13 19:04:26 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
/*
|
2005-01-29 23:34:31 +00:00
|
|
|
* nm_device_set_wireless_config
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
*
|
|
|
|
|
* Bring up a wireless card with the essid and wep key of its "best" ap
|
|
|
|
|
*
|
|
|
|
|
* Returns: TRUE on successful activation
|
|
|
|
|
* FALSE on unsuccessful activation (ie no best AP)
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-01-21 19:32:08 +00:00
|
|
|
static gboolean nm_device_set_wireless_config (NMDevice *dev, NMAccessPoint *ap)
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
NMDeviceAuthMethod auth;
|
|
|
|
|
const char *essid = NULL;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), FALSE);
|
2004-10-29 16:37:43 +00:00
|
|
|
g_return_val_if_fail (ap != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_ap_get_essid (ap) != NULL, FALSE);
|
2005-01-21 19:32:08 +00:00
|
|
|
g_return_val_if_fail (nm_ap_get_auth_method (ap) != NM_DEVICE_AUTH_METHOD_UNKNOWN, FALSE);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-05-05 16:52:04 +00:00
|
|
|
dev->options.wireless.failed_link_count = 0;
|
|
|
|
|
|
2004-10-29 16:37:43 +00:00
|
|
|
/* Force the card into Managed/Infrastructure mode */
|
2005-03-26 03:42:05 +00:00
|
|
|
nm_device_bring_down_wait (dev, 0);
|
|
|
|
|
nm_device_bring_up_wait (dev, 0);
|
|
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
nm_device_set_mode (dev, NETWORK_MODE_INFRA);
|
2004-10-12 11:39:08 +00:00
|
|
|
|
2004-10-29 16:37:43 +00:00
|
|
|
essid = nm_ap_get_essid (ap);
|
2005-01-21 19:32:08 +00:00
|
|
|
auth = nm_ap_get_auth_method (ap);
|
|
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
nm_device_set_mode (dev, nm_ap_get_mode (ap));
|
2004-12-17 17:16:22 +00:00
|
|
|
nm_device_set_bitrate (dev, 0);
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
if (nm_ap_get_user_created (ap) || (nm_ap_get_freq (ap) && (nm_ap_get_mode (ap) == NETWORK_MODE_ADHOC)))
|
|
|
|
|
nm_device_set_frequency (dev, nm_ap_get_freq (ap));
|
2005-01-10 06:41:02 +00:00
|
|
|
else
|
|
|
|
|
nm_device_set_frequency (dev, 0); /* auto */
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (nm_ap_get_encrypted (ap) && nm_ap_is_enc_key_valid (ap))
|
2004-10-29 16:37:43 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
char * hashed_key = nm_ap_get_enc_key_hashed (ap);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
if (auth == NM_DEVICE_AUTH_METHOD_NONE)
|
|
|
|
|
{
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_warning ("Activation (%s/wireless): AP '%s' said it was encrypted, but had "
|
2005-01-21 19:32:08 +00:00
|
|
|
"'none' for authentication method. Using Open System authentication method.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (ap));
|
|
|
|
|
}
|
2004-11-16 02:41:53 +00:00
|
|
|
nm_device_set_enc_key (dev, hashed_key, auth);
|
2004-10-29 16:37:43 +00:00
|
|
|
g_free (hashed_key);
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
else
|
|
|
|
|
nm_device_set_enc_key (dev, NULL, NM_DEVICE_AUTH_METHOD_NONE);
|
2004-08-24 Dan Williams <dcbw@redhat.com>
* src/NetworkManagerAP.[ch]
- Add a "enc_method_good" member and accessors to an Access Point
to signal when we've found the correct encryption method
for an access point
- Add a "timestamp" member and accessors, remove "priority" member
and accessors (use timestamps instead)
- Rename "wep_key"->"enc_key"
- (nm_ap_get_enc_key_hashed): new, return the correct mangled key
for a specified encryption method using the access points
source encryption key/passphrase
* src/NetworkManagerAPList.c
- When updating a network with dbus, grab timestamp now instead of
priority
* src/NetworkManagerDBus.[ch]
- Add signal for "DeviceActivating"
- Switch priority->timestamp
* src/NetworkManagerDevice.c
- Change references of "wep_key" -> "enc_key" or "key"
- Signal DeviceActivating when starting activation
- When activating a wireless device, if the access point we are connecting
to is encrypted, and we have a source key, try to generate a mangled
key and use that (ie, generate real WEP key from a passphrase)
- Rework device activation to fallback to other encryption methods if
a previous one didn't work (ie, try mangling a key as a 104-bit passphrase
first, then if that doesn't work fall back to direct hex key).
- (nm_device_update_best_ap): fix a deadlock, and use timestamps instead of
priority. We now prefer the latest access point used, rather than using
a priority scheme
- (nm_device_do_normal_scan): make the encryption method "unknown" on access
points we've just discovered, and merge in correct info from the global
access point lists
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@68 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-25 22:41:12 +00:00
|
|
|
|
2004-10-29 16:37:43 +00:00
|
|
|
nm_device_set_essid (dev, essid);
|
2004-08-20 18:18:37 +00:00
|
|
|
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_info ("Activation (%s/wireless): using essid '%s', with %s authentication.",
|
2004-11-16 02:41:53 +00:00
|
|
|
nm_device_get_iface (dev), essid, (auth == NM_DEVICE_AUTH_METHOD_NONE) ? "no" :
|
|
|
|
|
((auth == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM) ? "Open System" :
|
|
|
|
|
((auth == NM_DEVICE_AUTH_METHOD_SHARED_KEY) ? "Shared Key" : "unknown")));
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-03-26 03:42:05 +00:00
|
|
|
/* Bring the device up and pause to allow card to associate. After we set the ESSID
|
|
|
|
|
* on the card, the card has to scan all channels to find our requested AP (which can
|
|
|
|
|
* take a long time if it is an A/B/G chipset like the Atheros 5212, for example).
|
|
|
|
|
*/
|
|
|
|
|
nm_device_is_up_and_associated_wait (dev, 2, 100);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
/* Some cards don't really work well in ad-hoc mode unless you explicitly set the bitrate
|
|
|
|
|
* on them. (Netgear WG511T/Atheros 5212 with madwifi drivers). Until we can get rate information
|
|
|
|
|
* from scanned access points out of iwlib, clamp bitrate for these cards at 11Mbps.
|
|
|
|
|
*/
|
|
|
|
|
if ((nm_ap_get_mode (ap) == NETWORK_MODE_ADHOC) && (nm_device_get_bitrate (dev) <= 0))
|
|
|
|
|
nm_device_set_bitrate (dev, 11000); /* In Kbps */
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return TRUE;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
|
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_wireless_configure_adhoc
|
2004-12-17 17:16:22 +00:00
|
|
|
*
|
|
|
|
|
* Create an ad-hoc network (rather than associating with one).
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static void nm_device_wireless_configure_adhoc (NMActRequest *req)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data;
|
|
|
|
|
NMDevice * dev;
|
|
|
|
|
NMAccessPoint * ap;
|
|
|
|
|
NMDeviceAuthMethod auth = NM_DEVICE_AUTH_METHOD_NONE;
|
|
|
|
|
NMAPListIter * iter;
|
|
|
|
|
NMAccessPoint * tmp_ap;
|
|
|
|
|
double card_freqs[IW_MAX_FREQUENCIES];
|
|
|
|
|
int num_freqs = 0, i;
|
|
|
|
|
double freq_to_use = 0;
|
|
|
|
|
iwrange range;
|
|
|
|
|
NMSock * sk;
|
|
|
|
|
int err;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
|
|
|
|
|
|
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
ap = nm_act_request_get_ap (req);
|
|
|
|
|
g_assert (ap);
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
if (nm_ap_get_encrypted (ap))
|
|
|
|
|
auth = NM_DEVICE_AUTH_METHOD_SHARED_KEY;
|
|
|
|
|
|
|
|
|
|
/* Build our local list of frequencies to whittle down until we find a free one */
|
|
|
|
|
memset (&card_freqs, 0, sizeof (card_freqs));
|
2005-01-31 19:53:38 +00:00
|
|
|
num_freqs = MIN (dev->options.wireless.num_freqs, IW_MAX_FREQUENCIES);
|
2004-12-17 17:16:22 +00:00
|
|
|
for (i = 0; i < num_freqs; i++)
|
2005-01-31 19:53:38 +00:00
|
|
|
card_freqs[i] = dev->options.wireless.freqs[i];
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
/* We need to find a clear wireless channel to use. We will
|
|
|
|
|
* only use 802.11b channels for now.
|
|
|
|
|
*/
|
|
|
|
|
iter = nm_ap_list_iter_new (nm_device_ap_list_get (dev));
|
|
|
|
|
while ((tmp_ap = nm_ap_list_iter_next (iter)))
|
|
|
|
|
{
|
|
|
|
|
double ap_freq = nm_ap_get_freq (tmp_ap);
|
|
|
|
|
for (i = 0; i < num_freqs && ap_freq; i++)
|
|
|
|
|
{
|
|
|
|
|
if (card_freqs[i] == ap_freq)
|
|
|
|
|
card_freqs[i] = 0;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
nm_ap_list_iter_free (iter);
|
|
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)) == NULL)
|
2005-05-04 18:11:40 +00:00
|
|
|
{
|
|
|
|
|
nm_policy_schedule_activation_failed (req);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return;
|
2005-05-04 18:11:40 +00:00
|
|
|
}
|
2005-01-31 19:53:38 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
err = iw_get_range_info (nm_dev_sock_get_fd (sk), nm_device_get_iface (dev), &range);
|
|
|
|
|
nm_dev_sock_close (sk);
|
|
|
|
|
if (err < 0)
|
2005-05-04 18:11:40 +00:00
|
|
|
{
|
|
|
|
|
nm_policy_schedule_activation_failed (req);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return;
|
2005-05-04 18:11:40 +00:00
|
|
|
}
|
2005-01-31 19:53:38 +00:00
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
/* Ok, find the first non-zero freq in our table and use it.
|
|
|
|
|
* For now we only try to use a channel in the 802.11b channel
|
|
|
|
|
* space so that most everyone can see it.
|
|
|
|
|
*/
|
|
|
|
|
for (i = 0; i < num_freqs; i++)
|
|
|
|
|
{
|
2005-01-31 19:53:38 +00:00
|
|
|
int channel = iw_freq_to_channel (card_freqs[i], &range);
|
2004-12-17 17:16:22 +00:00
|
|
|
if (card_freqs[i] && (channel > 0) && (channel < 15))
|
|
|
|
|
{
|
|
|
|
|
freq_to_use = card_freqs[i];
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Hmm, no free channels in 802.11b space. Pick one more or less randomly */
|
|
|
|
|
if (!freq_to_use)
|
|
|
|
|
{
|
|
|
|
|
double pfreq;
|
|
|
|
|
int channel = (int)(random () % 14);
|
|
|
|
|
int err;
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
err = iw_channel_to_freq (channel, &pfreq, &range);
|
2004-12-17 17:16:22 +00:00
|
|
|
if (err == channel)
|
|
|
|
|
freq_to_use = pfreq;
|
|
|
|
|
}
|
|
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
if (!freq_to_use)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-05-04 18:11:40 +00:00
|
|
|
nm_policy_schedule_activation_failed (req);
|
|
|
|
|
return;
|
|
|
|
|
}
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
nm_ap_set_freq (ap, freq_to_use);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
nm_info ("Will create network '%s' with frequency %f.", nm_ap_get_essid (ap), nm_ap_get_freq (ap));
|
|
|
|
|
nm_device_set_wireless_config (dev, ap);
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
{
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
return;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
}
|
2005-05-04 18:11:40 +00:00
|
|
|
|
|
|
|
|
nm_device_activate_schedule_stage3_ip_config_start (req);
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static gboolean nm_dwwfl_test (int tries, nm_completion_args args)
|
2004-10-29 16:37:43 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMDevice * dev = args[0];
|
|
|
|
|
guint * assoc_count = args[1];
|
|
|
|
|
double * last_freq = args[2];
|
|
|
|
|
char * essid = args[3];
|
2005-06-10 03:47:47 +00:00
|
|
|
int required = GPOINTER_TO_INT (args[4]);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
double cur_freq = nm_device_get_frequency (dev);
|
|
|
|
|
gboolean assoc = nm_device_wireless_is_associated (dev);
|
|
|
|
|
const char * cur_essid = nm_device_get_essid (dev);
|
2004-10-29 16:37:43 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* If we've been cancelled, return that we should stop */
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
return TRUE;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* If we're on the same frequency and essid, and we're associated,
|
|
|
|
|
* increment the count for how many iterations we've been associated;
|
|
|
|
|
* otherwise start over. */
|
|
|
|
|
/* XXX floating point comparison this way is dangerous, IIRC */
|
|
|
|
|
if ((cur_freq == *last_freq) && assoc && !strcmp (essid, cur_essid))
|
|
|
|
|
{
|
|
|
|
|
(*assoc_count)++;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
*assoc_count = 0;
|
|
|
|
|
*last_freq = cur_freq;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* If we're told to cancel, return that we're finished.
|
2005-08-17 16:04:17 +00:00
|
|
|
* If the card's frequency has been stable for more than the required
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* interval, return that we're finished.
|
|
|
|
|
* Otherwise, we're not finished. */
|
|
|
|
|
if (nm_device_activation_should_cancel (dev) || (*assoc_count >= required))
|
|
|
|
|
return TRUE;
|
|
|
|
|
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_wait_for_link
|
|
|
|
|
*
|
|
|
|
|
* Try to be clever about when the wireless card really has associated with the access point.
|
|
|
|
|
* Return TRUE when we think that it has, and FALSE when we thing it has not associated.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_wireless_wait_for_link (NMDevice *dev, const char *essid)
|
|
|
|
|
{
|
|
|
|
|
guint assoc = 0;
|
|
|
|
|
double last_freq = 0;
|
|
|
|
|
struct timeval timeout = { .tv_sec = 0, .tv_usec = 0 };
|
|
|
|
|
nm_completion_args args;
|
|
|
|
|
|
|
|
|
|
/* we want to sleep for a very short amount of time, to minimize
|
|
|
|
|
* hysteresis on the boundaries of our required time. But we
|
|
|
|
|
* also want the maximum to be based on what the card */
|
|
|
|
|
const guint delay = 30;
|
|
|
|
|
const guint required_tries = 10;
|
2005-06-23 20:10:41 +00:00
|
|
|
const guint min_delay = 2 * (delay / required_tries);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
/* for cards which don't scan many frequencies, this will return
|
|
|
|
|
* 5 seconds, which we'll bump up to 6 seconds below. Oh well. */
|
|
|
|
|
timeout.tv_sec = (time_t) nm_device_get_association_pause_value (dev);
|
|
|
|
|
|
|
|
|
|
/* Refuse to to have a timeout that's _less_ than twice the total time
|
|
|
|
|
* required before calling a link valid */
|
|
|
|
|
if (timeout.tv_sec < min_delay)
|
|
|
|
|
timeout.tv_sec = min_delay;
|
|
|
|
|
|
|
|
|
|
/* We more or less keep asking the driver for the frequency the
|
|
|
|
|
* card is listening on until it connects to an AP. Once it's
|
|
|
|
|
* associated, the driver stops scanning. To detect that, we look
|
|
|
|
|
* for the essid and frequency to remain constant for 3 seconds.
|
|
|
|
|
* When it remains constant, we assume it's a real link. */
|
|
|
|
|
args[0] = dev;
|
|
|
|
|
args[1] = &assoc;
|
|
|
|
|
args[2] = &last_freq;
|
|
|
|
|
args[3] = (void *)essid;
|
|
|
|
|
args[4] = (void *)(required_tries * 2);
|
|
|
|
|
nm_wait_for_timeout (&timeout, G_USEC_PER_SEC / delay, nm_dwwfl_test, nm_dwwfl_test, args);
|
|
|
|
|
|
|
|
|
|
/* If we've had a reasonable association count, we say we have a link */
|
|
|
|
|
if (assoc > required_tries)
|
|
|
|
|
return TRUE;
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
static gboolean ap_need_key (NMDevice *dev, NMAccessPoint *ap)
|
|
|
|
|
{
|
|
|
|
|
char *essid;
|
|
|
|
|
gboolean need_key = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (ap != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
essid = nm_ap_get_essid (ap);
|
|
|
|
|
|
|
|
|
|
if (!nm_ap_get_encrypted (ap))
|
|
|
|
|
{
|
2005-05-04 18:11:40 +00:00
|
|
|
nm_info ("Activation (%s/wireless): access point '%s' is unencrypted, no key needed.",
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_device_get_iface (dev), essid ? essid : "(null)");
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
else
|
|
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
if (nm_ap_is_enc_key_valid (ap))
|
|
|
|
|
{
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_info ("Activation (%s/wireless): access point '%s' "
|
|
|
|
|
"is encrypted, and a key exists. No new key needed.",
|
|
|
|
|
nm_device_get_iface (dev), essid ? essid : "(null)");
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
else
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_info ("Activation (%s/wireless): access point '%s' "
|
|
|
|
|
"is encrypted, but NO valid key exists. New key needed.",
|
|
|
|
|
nm_device_get_iface (dev),
|
|
|
|
|
essid ? essid : "(null)");
|
2005-01-21 19:32:08 +00:00
|
|
|
need_key = TRUE;
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
2004-10-29 16:37:43 +00:00
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
return need_key;
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activate_wireless_configure
|
2005-01-21 19:32:08 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Configure a wireless device for association with a particular access point.
|
2005-01-21 19:32:08 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static void nm_device_wireless_configure (NMActRequest *req)
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data;
|
|
|
|
|
NMDevice * dev;
|
|
|
|
|
NMAccessPoint * ap;
|
|
|
|
|
gboolean success = FALSE;
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
|
|
|
|
|
|
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
ap = nm_act_request_get_ap (req);
|
|
|
|
|
g_assert (ap);
|
|
|
|
|
|
|
|
|
|
nm_device_bring_up_wait (dev, 1);
|
|
|
|
|
|
2005-09-21 15:32:33 +00:00
|
|
|
nm_info ("Activation (%s/wireless) Stage 2 (Device Configure) will connect to access point '%s'.", nm_device_get_iface (dev), nm_ap_get_essid (ap));
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
if (ap_need_key (dev, ap))
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-08-17 16:04:17 +00:00
|
|
|
nm_dbus_get_user_key_for_network (data->dbus_connection, req, FALSE);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return;
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
while (success == FALSE)
|
|
|
|
|
{
|
|
|
|
|
gboolean link = FALSE;
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
nm_device_set_wireless_config (dev, ap);
|
|
|
|
|
|
|
|
|
|
success = link = nm_device_wireless_wait_for_link (dev, nm_ap_get_essid (ap));
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
if (!link)
|
|
|
|
|
{
|
|
|
|
|
if (nm_ap_get_auth_method (ap) == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM)
|
|
|
|
|
{
|
|
|
|
|
nm_debug ("Activation (%s/wireless): no hardware link to '%s' in Open System mode, trying Shared Key.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (ap) ? nm_ap_get_essid (ap) : "(none)");
|
|
|
|
|
/* Back down to Shared Key mode */
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_SHARED_KEY);
|
|
|
|
|
success = FALSE;
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
else if (nm_ap_get_auth_method (ap) == NM_DEVICE_AUTH_METHOD_SHARED_KEY)
|
|
|
|
|
{
|
|
|
|
|
/* Didn't work in Shared Key either. */
|
|
|
|
|
nm_debug ("Activation (%s/wireless): no hardware link to '%s' in Shared Key mode, need correct key?",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (ap) ? nm_ap_get_essid (ap) : "(none)");
|
2005-08-17 16:04:17 +00:00
|
|
|
nm_dbus_get_user_key_for_network (data->dbus_connection, req, TRUE);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
nm_debug ("Activation (%s/wireless): no hardware link to '%s' in non-encrypted mode.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (ap) ? nm_ap_get_essid (ap) : "(none)");
|
|
|
|
|
nm_policy_schedule_activation_failed (req);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (success)
|
|
|
|
|
{
|
|
|
|
|
nm_info ("Activation (%s/wireless) Stage 2 (Device Configure) successful. Connected to access point '%s'.", nm_device_get_iface (dev), nm_ap_get_essid (ap) ? nm_ap_get_essid (ap) : "(none)");
|
|
|
|
|
nm_device_activate_schedule_stage3_ip_config_start (req);
|
|
|
|
|
}
|
2005-05-06 02:58:06 +00:00
|
|
|
else if (!nm_device_activation_should_cancel (dev) && (nm_act_request_get_stage (req) != NM_ACT_STAGE_NEED_USER_KEY))
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_policy_schedule_activation_failed (req);
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_wired_configure
|
|
|
|
|
*
|
|
|
|
|
* Configure a wired device for activation
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static void nm_device_wired_configure (NMActRequest *req)
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data;
|
|
|
|
|
NMDevice * dev;
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
nm_info ("Activation (%s/wired) Stage 2 (Device Configure) successful.", nm_device_get_iface (dev));
|
|
|
|
|
nm_device_activate_schedule_stage3_ip_config_start (req);
|
2004-10-29 16:37:43 +00:00
|
|
|
}
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_stage2_device_config
|
|
|
|
|
*
|
|
|
|
|
* Determine device parameters and set those on the device, ie
|
|
|
|
|
* for wireless devices, set essid, keys, etc.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_activate_stage2_device_config (NMActRequest *req)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMDevice * dev;
|
|
|
|
|
NMData * data;
|
|
|
|
|
NMAccessPoint * ap;
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
ap = nm_act_request_get_ap (req);
|
|
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
g_assert (ap);
|
|
|
|
|
|
|
|
|
|
nm_info ("Activation (%s) Stage 2 (Device Configure) starting...", nm_device_get_iface (dev));
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* Bring the device up */
|
2005-06-15 11:59:43 +00:00
|
|
|
if (!nm_device_is_up (dev))
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_bring_up (dev);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_activation_should_cancel (dev))
|
2005-03-29 22:10:21 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
goto out;
|
2005-03-26 03:42:05 +00:00
|
|
|
}
|
2005-04-01 21:30:12 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
if (nm_ap_get_user_created (ap))
|
|
|
|
|
nm_device_wireless_configure_adhoc (req);
|
|
|
|
|
else
|
|
|
|
|
nm_device_wireless_configure (req);
|
2005-04-01 21:30:12 +00:00
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
else if (nm_device_is_wired (dev))
|
|
|
|
|
nm_device_wired_configure (req);
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
out:
|
|
|
|
|
nm_info ("Activation (%s) Stage 2 (Device Configure) complete.", nm_device_get_iface (dev));
|
2005-03-26 03:42:05 +00:00
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_schedule_stage2_device_config
|
|
|
|
|
*
|
|
|
|
|
* Schedule setup of the hardware device
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_activate_schedule_stage2_device_config (NMActRequest *req)
|
2005-03-29 22:10:21 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
GSource * source = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
2005-03-29 22:10:21 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
2005-03-29 22:10:21 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2005-03-29 22:10:21 +00:00
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_DEVICE_CONFIG);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activate_stage2_device_config, req, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
nm_info ("Activation (%s) Stage 2 (Device Configure) scheduled...", nm_device_get_iface (dev));
|
2005-03-29 22:10:21 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activate_stage3_ip_config_start
|
2004-11-16 02:41:53 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Begin IP configuration with either DHCP or static IP.
|
2004-11-16 02:41:53 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static gboolean nm_device_activate_stage3_ip_config_start (NMActRequest *req)
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
NMAccessPoint * ap = NULL;
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2004-11-16 15:06:46 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2004-11-16 15:06:46 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("Activation (%s) Stage 3 (IP Configure Start) started...", nm_device_get_iface (dev));
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
{
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
2004-11-16 02:41:53 +00:00
|
|
|
goto out;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
}
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
ap = nm_act_request_get_ap (req);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (!(ap && nm_ap_get_user_created (ap)) && nm_device_get_use_dhcp (dev))
|
|
|
|
|
{
|
|
|
|
|
/* Begin a DHCP transaction on the interface */
|
2005-05-04 18:11:40 +00:00
|
|
|
if (!nm_dhcp_manager_begin_transaction (data->dhcp_manager, req))
|
|
|
|
|
nm_policy_schedule_activation_failed (req);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
goto out;
|
|
|
|
|
}
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_activation_should_cancel (dev))
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
2005-03-26 03:42:05 +00:00
|
|
|
goto out;
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* Static IP and user-created wireless networks skip directly to IP configure stage */
|
|
|
|
|
nm_device_activate_schedule_stage4_ip_config_get (req);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
nm_info ("Activation (%s) Stage 3 (IP Configure Start) complete.", nm_device_get_iface (dev));
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_activate_schedule_stage3_ip_config_start
|
|
|
|
|
*
|
|
|
|
|
* Schedule IP configuration start
|
|
|
|
|
*/
|
|
|
|
|
static void nm_device_activate_schedule_stage3_ip_config_start (NMActRequest *req)
|
|
|
|
|
{
|
|
|
|
|
GSource * source = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (req != NULL);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_IP_CONFIG_START);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activate_stage3_ip_config_start, req, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
nm_info ("Activation (%s) Stage 3 (IP Configure Start) scheduled.", nm_device_get_iface (dev));
|
|
|
|
|
}
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_new_ip4_autoip_config
|
|
|
|
|
*
|
|
|
|
|
* Build up an IP config with a Link Local address
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static NMIP4Config *nm_device_new_ip4_autoip_config (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
struct in_addr ip;
|
|
|
|
|
NMIP4Config * config = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
if (get_autoip (dev, &ip))
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
#define LINKLOCAL_BCAST 0xa9feffff
|
|
|
|
|
|
|
|
|
|
config = nm_ip4_config_new ();
|
|
|
|
|
|
|
|
|
|
nm_ip4_config_set_address (config, (guint32)(ip.s_addr));
|
|
|
|
|
nm_ip4_config_set_netmask (config, (guint32)(ntohl (0xFFFF0000)));
|
|
|
|
|
nm_ip4_config_set_broadcast (config, (guint32)(ntohl (LINKLOCAL_BCAST)));
|
|
|
|
|
nm_ip4_config_set_gateway (config, 0);
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return config;
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_stage4_ip_config_get
|
|
|
|
|
*
|
|
|
|
|
* Retrieve the correct IP config.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_activate_stage4_ip_config_get (NMActRequest *req)
|
|
|
|
|
{
|
|
|
|
|
NMData * data = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
NMAccessPoint * ap = NULL;
|
|
|
|
|
NMIP4Config * ip4_config = NULL;
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
g_assert (data);
|
2005-04-28 02:50:54 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
g_assert (data);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
2005-05-04 18:11:40 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
ap = nm_act_request_get_ap (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
g_assert (ap);
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("Activation (%s) Stage 4 (IP Configure Get) started...", nm_device_get_iface (dev));
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
{
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
goto out;
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (ap && nm_ap_get_user_created (ap))
|
|
|
|
|
ip4_config = nm_device_new_ip4_autoip_config (dev);
|
|
|
|
|
else if (nm_device_get_use_dhcp (dev))
|
|
|
|
|
ip4_config = nm_dhcp_manager_get_ip4_config (data->dhcp_manager, req);
|
|
|
|
|
else
|
|
|
|
|
ip4_config = nm_system_device_new_ip4_system_config (dev);
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (ip4_config)
|
|
|
|
|
{
|
|
|
|
|
nm_act_request_set_ip4_config (req, ip4_config);
|
|
|
|
|
nm_device_activate_schedule_stage5_ip_config_commit (req);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Interfaces cannot be down if they are the active interface,
|
|
|
|
|
* otherwise we cannot use them for scanning or link detection.
|
2005-01-21 19:32:08 +00:00
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
nm_device_set_essid (dev, "");
|
|
|
|
|
nm_device_set_enc_key (dev, NULL, NM_DEVICE_AUTH_METHOD_NONE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!nm_device_is_up (dev))
|
|
|
|
|
nm_device_bring_up (dev);
|
|
|
|
|
|
|
|
|
|
nm_policy_schedule_activation_failed (req);
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
out:
|
|
|
|
|
nm_info ("Activation (%s) Stage 4 (IP Configure Get) complete.", nm_device_get_iface (dev));
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_schedule_stage4_ip_config_get
|
|
|
|
|
*
|
|
|
|
|
* Schedule creation of the IP config
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_activate_schedule_stage4_ip_config_get (NMActRequest *req)
|
|
|
|
|
{
|
|
|
|
|
GSource * source = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_IP_CONFIG_GET);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_info ("Activation (%s) Stage 4 (IP Configure Get) scheduled...", nm_device_get_iface (dev));
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activate_stage4_ip_config_get, req, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
}
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_stage4_ip_config_timeout
|
|
|
|
|
*
|
|
|
|
|
* Retrieve the correct IP config.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_activate_stage4_ip_config_timeout (NMActRequest *req)
|
|
|
|
|
{
|
|
|
|
|
NMData * data = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
NMIP4Config * ip4_config = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
nm_info ("Activation (%s) Stage 4 (IP Configure Timeout) started...", nm_device_get_iface (dev));
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
{
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (nm_device_is_wired (dev))
|
2005-07-05 15:05:33 +00:00
|
|
|
{
|
|
|
|
|
/* Wired network, no DHCP reply. Let's get an IP via Zeroconf. */
|
|
|
|
|
nm_info ("No DHCP reply received. Automatically obtaining IP via Zeroconf.");
|
|
|
|
|
ip4_config = nm_device_new_ip4_autoip_config (dev);
|
|
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
else if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *ap = nm_act_request_get_ap (req);
|
|
|
|
|
|
|
|
|
|
g_assert (ap);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* For those broken cards that report successful hardware link even when WEP key is wrong,
|
|
|
|
|
* and also for Open System mode (where you cannot know WEP key is wrong ever), we try to
|
|
|
|
|
* do DHCP and if that fails, fall back to next auth mode and try again.
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_ap_get_auth_method (ap) == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM)
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* Back down to Shared Key mode */
|
|
|
|
|
nm_debug ("Activation (%s/wireless): could not get IP configuration info for '%s' in Open System mode, trying Shared Key.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (ap) ? nm_ap_get_essid (ap) : "(none)");
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_SHARED_KEY);
|
|
|
|
|
nm_device_activate_schedule_stage2_device_config (req);
|
|
|
|
|
}
|
|
|
|
|
else if ((nm_ap_get_auth_method (ap) == NM_DEVICE_AUTH_METHOD_SHARED_KEY))
|
|
|
|
|
{
|
|
|
|
|
/* Shared Key mode failed, we must have bad WEP key */
|
|
|
|
|
nm_debug ("Activation (%s/wireless): could not get IP configuration info for '%s' in Shared Key mode, asking for new key.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (ap) ? nm_ap_get_essid (ap) : "(none)");
|
2005-08-17 16:04:17 +00:00
|
|
|
nm_dbus_get_user_key_for_network (data->dbus_connection, req, TRUE);
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
|
|
|
|
else
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-07-05 15:05:33 +00:00
|
|
|
/*
|
|
|
|
|
* Wireless, not encrypted, no DHCP Reply. Try Zeroconf. We do not do this in
|
|
|
|
|
* the encrypted case, because the problem could be (and more likely is) a bad key.
|
|
|
|
|
*/
|
|
|
|
|
nm_info ("No DHCP reply received. Automatically obtaining IP via Zeroconf.");
|
|
|
|
|
ip4_config = nm_device_new_ip4_autoip_config (dev);
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2005-07-05 15:05:33 +00:00
|
|
|
if (ip4_config)
|
|
|
|
|
{
|
|
|
|
|
nm_act_request_set_ip4_config (req, ip4_config);
|
|
|
|
|
nm_device_activate_schedule_stage5_ip_config_commit (req);
|
|
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
out:
|
|
|
|
|
nm_info ("Activation (%s) Stage 4 (IP Configure Timeout) complete.", nm_device_get_iface (dev));
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activate_schedule_stage4_ip_config_timeout
|
|
|
|
|
*
|
|
|
|
|
* Deal with a timed out DHCP transaction
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_activate_schedule_stage4_ip_config_timeout (NMActRequest *req)
|
|
|
|
|
{
|
|
|
|
|
GSource * source = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (req != NULL);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_IP_CONFIG_GET);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activate_stage4_ip_config_timeout, req, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
nm_info ("Activation (%s) Stage 4 (IP Configure Timeout) scheduled...", nm_device_get_iface (dev));
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
|
2004-09-08 18:14:42 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activate_stage5_ip_config_commit
|
2004-09-08 18:14:42 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Commit the IP config on the device
|
2004-09-08 18:14:42 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static gboolean nm_device_activate_stage5_ip_config_commit (NMActRequest *req)
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
NMIP4Config * ip4_config = NULL;
|
2004-10-08 05:22:17 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (req != NULL, FALSE);
|
2004-09-08 18:14:42 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2004-09-08 18:14:42 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
ip4_config = nm_act_request_get_ip4_config (req);
|
|
|
|
|
g_assert (ip4_config);
|
|
|
|
|
|
|
|
|
|
nm_info ("Activation (%s) Stage 5 (IP Configure Commit) started...", nm_device_get_iface (dev));
|
|
|
|
|
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
{
|
|
|
|
|
nm_device_schedule_activation_handle_cancel (req);
|
|
|
|
|
goto out;
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
nm_device_set_ip4_config (dev, ip4_config);
|
|
|
|
|
if (nm_system_device_set_from_ip4_config (dev))
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_update_ip4_address (dev);
|
|
|
|
|
nm_system_device_add_ip6_link_address (dev);
|
|
|
|
|
nm_system_restart_mdns_responder ();
|
|
|
|
|
nm_policy_schedule_activation_finish (req);
|
2005-05-08 16:31:03 +00:00
|
|
|
nm_device_set_link_active (dev, nm_device_probe_link_state (dev));
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
else
|
|
|
|
|
nm_policy_schedule_activation_failed (req);
|
2004-12-16 04:24:20 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
out:
|
|
|
|
|
nm_info ("Activation (%s) Stage 5 (IP Configure Commit) complete.", nm_device_get_iface (dev));
|
|
|
|
|
return FALSE;
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_activate_schedule_stage5_ip_config_commit
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Schedule commit of the IP config
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static void nm_device_activate_schedule_stage5_ip_config_commit (NMActRequest *req)
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
GSource * source = NULL;
|
|
|
|
|
NMDevice * dev = NULL;
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_if_fail (req != NULL);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
2004-11-22 14:42:34 +00:00
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
nm_act_request_set_stage (req, NM_ACT_STAGE_IP_CONFIG_COMMIT);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, (GSourceFunc) nm_device_activate_stage5_ip_config_commit, req, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
nm_info ("Activation (%s) Stage 5 (IP Configure Commit) scheduled...", nm_device_get_iface (dev));
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
2004-10-11 21:32:19 +00:00
|
|
|
* nm_device_is_activating
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
*
|
|
|
|
|
* Return whether or not the device is currently activating itself.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-10-11 21:32:19 +00:00
|
|
|
gboolean nm_device_is_activating (NMDevice *dev)
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMActRequest * req;
|
|
|
|
|
NMActStage stage;
|
2005-05-04 18:11:40 +00:00
|
|
|
gboolean activating = FALSE;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (!(req = nm_device_get_act_request (dev)))
|
|
|
|
|
return FALSE;
|
|
|
|
|
|
|
|
|
|
stage = nm_act_request_get_stage (req);
|
2005-05-04 18:11:40 +00:00
|
|
|
switch (stage)
|
|
|
|
|
{
|
2005-05-06 02:58:06 +00:00
|
|
|
case NM_ACT_STAGE_DEVICE_PREPARE:
|
|
|
|
|
case NM_ACT_STAGE_DEVICE_CONFIG:
|
|
|
|
|
case NM_ACT_STAGE_NEED_USER_KEY:
|
|
|
|
|
case NM_ACT_STAGE_IP_CONFIG_START:
|
|
|
|
|
case NM_ACT_STAGE_IP_CONFIG_GET:
|
|
|
|
|
case NM_ACT_STAGE_IP_CONFIG_COMMIT:
|
2005-05-04 18:11:40 +00:00
|
|
|
activating = TRUE;
|
|
|
|
|
break;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2005-05-06 02:58:06 +00:00
|
|
|
case NM_ACT_STAGE_ACTIVATED:
|
|
|
|
|
case NM_ACT_STAGE_FAILED:
|
|
|
|
|
case NM_ACT_STAGE_CANCELLED:
|
|
|
|
|
case NM_ACT_STAGE_UNKNOWN:
|
2005-05-04 18:11:40 +00:00
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return activating;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-06-09 Dan Williams <dcbw@redhat.com>
* NetworkManager.h
- Add NMWirelessScanMethod enum for scan methods
* gnome/applet/applet-dbus-devices.c
- (nmwa_dbus_update_scanning_enabled_cb): remove
- (nmwa_dbus_update_scanning_enabled): remove
- (nmwa_dbus_update_devices): don't call nmwa_dbus_update_scanning_enabled() anymore
since it got removed
- (nmwa_dbus_enable_scanning): remove
* gnome/applet/applet-dbus-info.c
- (nmi_dbus_signal_update_scan_method): new function, signal NetworkManager to
update the wireless scanning method from NMI
- (nmi_dbus_get_wireless_scan_method): new function, return wireless scanning
method value to NetworkManager
- (nmi_dbus_info_message_handler): respond to the "getWirelessScanMethod" method call
* gnome/applet/applet-dbus-info.h
- Add prototype for nmi_dbus_signal_update_scan_method
* gnome/applet/applet.c
- (scanning_menu_update): new function, update one GtkCheckMenuItem from the
Wireless Scanning menu based on current wireless scan method
- (nmwa_menu_scanning_item_activate): new function, callback for GTK "activate"
signal for Wireless Scanning menu items, tell NetworkManager the new method
and update our menu items to make sure the right one is checked
- (nmwa_set_scanning_enabled_cb): remove
- (nmwa_context_menu_update): remove references to pause_scanning_item
- (nmwa_context_menu_create): remove pause_scanning_item, and add new Wireless
Scanning menu item
- (nmwa_gconf_get_wireless_scan_method): new method, pull wireless scanning method
from GConf
- nmwa_gconf_networks_notify_callback -> nmwa_gconf_info_notify_callback: generalize
so we get notified of preference values too
- (nmwa_get_instance): monitor GCONF_PATH_WIRELESS rather than GCONF_PATH_WIRELESS_NETWORKS
* gnome/applet/applet.h
- GCONF_PATH_WIRELESS added, one level below GCONF_PATH_WIRELESS_NETWORKS
- Add wireless scan method member to applet data
- Remove pause_scanning_item, add Wireless Scanning submenu
* src/NetworkManager.c
- (nm_data_new): default to NM_SCAN_METHOD_ON
- (main): grab scanning method from NMI if we can
* src/NetworkManagerDbus.c
- (nm_dbus_update_wireless_scan_method_cb): new function, callback from
nm_dbus_update_wireless_scan_method()
- (nm_dbus_update_wireless_scan_method): new function to grab scanning method
from NMI
- (nm_dbus_nmi_is_running): redundant function, removed
- (nm_dbus_signal_filter): trap "WirelessScanMethodUpdate" signal, grab scanning method
when NMI comes back
* src/NetworkManagerDevice.c
- (nm_device_is_activated): return TRUE if the device is activated
- (nm_device_wireless_scan): don't scan if the scan method is OFF, or if its AUTO
and we are activated
* src/nm-dbus-nm.c
- (nm_dbus_nm_set_scanning_enabled): removed
- nm_dbus_nm_get_scanning_enabled -> nm_dbus_nm_get_wireless_scan_method
- (nm_dbus_nm_methods_setup): remove [get | set] ScanningEnabled and add "getWirelessScanMethod"
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@658 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-06-10 03:13:27 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_is_activated
|
|
|
|
|
*
|
|
|
|
|
* Return whether or not the device is successfully activated.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-06-21 15:07:01 +00:00
|
|
|
static gboolean nm_device_is_activated (NMDevice *dev)
|
2005-06-09 Dan Williams <dcbw@redhat.com>
* NetworkManager.h
- Add NMWirelessScanMethod enum for scan methods
* gnome/applet/applet-dbus-devices.c
- (nmwa_dbus_update_scanning_enabled_cb): remove
- (nmwa_dbus_update_scanning_enabled): remove
- (nmwa_dbus_update_devices): don't call nmwa_dbus_update_scanning_enabled() anymore
since it got removed
- (nmwa_dbus_enable_scanning): remove
* gnome/applet/applet-dbus-info.c
- (nmi_dbus_signal_update_scan_method): new function, signal NetworkManager to
update the wireless scanning method from NMI
- (nmi_dbus_get_wireless_scan_method): new function, return wireless scanning
method value to NetworkManager
- (nmi_dbus_info_message_handler): respond to the "getWirelessScanMethod" method call
* gnome/applet/applet-dbus-info.h
- Add prototype for nmi_dbus_signal_update_scan_method
* gnome/applet/applet.c
- (scanning_menu_update): new function, update one GtkCheckMenuItem from the
Wireless Scanning menu based on current wireless scan method
- (nmwa_menu_scanning_item_activate): new function, callback for GTK "activate"
signal for Wireless Scanning menu items, tell NetworkManager the new method
and update our menu items to make sure the right one is checked
- (nmwa_set_scanning_enabled_cb): remove
- (nmwa_context_menu_update): remove references to pause_scanning_item
- (nmwa_context_menu_create): remove pause_scanning_item, and add new Wireless
Scanning menu item
- (nmwa_gconf_get_wireless_scan_method): new method, pull wireless scanning method
from GConf
- nmwa_gconf_networks_notify_callback -> nmwa_gconf_info_notify_callback: generalize
so we get notified of preference values too
- (nmwa_get_instance): monitor GCONF_PATH_WIRELESS rather than GCONF_PATH_WIRELESS_NETWORKS
* gnome/applet/applet.h
- GCONF_PATH_WIRELESS added, one level below GCONF_PATH_WIRELESS_NETWORKS
- Add wireless scan method member to applet data
- Remove pause_scanning_item, add Wireless Scanning submenu
* src/NetworkManager.c
- (nm_data_new): default to NM_SCAN_METHOD_ON
- (main): grab scanning method from NMI if we can
* src/NetworkManagerDbus.c
- (nm_dbus_update_wireless_scan_method_cb): new function, callback from
nm_dbus_update_wireless_scan_method()
- (nm_dbus_update_wireless_scan_method): new function to grab scanning method
from NMI
- (nm_dbus_nmi_is_running): redundant function, removed
- (nm_dbus_signal_filter): trap "WirelessScanMethodUpdate" signal, grab scanning method
when NMI comes back
* src/NetworkManagerDevice.c
- (nm_device_is_activated): return TRUE if the device is activated
- (nm_device_wireless_scan): don't scan if the scan method is OFF, or if its AUTO
and we are activated
* src/nm-dbus-nm.c
- (nm_dbus_nm_set_scanning_enabled): removed
- nm_dbus_nm_get_scanning_enabled -> nm_dbus_nm_get_wireless_scan_method
- (nm_dbus_nm_methods_setup): remove [get | set] ScanningEnabled and add "getWirelessScanMethod"
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@658 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-06-10 03:13:27 +00:00
|
|
|
{
|
|
|
|
|
NMActRequest * req;
|
|
|
|
|
NMActStage stage;
|
|
|
|
|
gboolean activated = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
if (!(req = nm_device_get_act_request (dev)))
|
|
|
|
|
return FALSE;
|
|
|
|
|
|
|
|
|
|
stage = nm_act_request_get_stage (req);
|
|
|
|
|
switch (stage)
|
|
|
|
|
{
|
|
|
|
|
case NM_ACT_STAGE_ACTIVATED:
|
|
|
|
|
activated = TRUE;
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
case NM_ACT_STAGE_DEVICE_PREPARE:
|
|
|
|
|
case NM_ACT_STAGE_DEVICE_CONFIG:
|
|
|
|
|
case NM_ACT_STAGE_NEED_USER_KEY:
|
|
|
|
|
case NM_ACT_STAGE_IP_CONFIG_START:
|
|
|
|
|
case NM_ACT_STAGE_IP_CONFIG_GET:
|
|
|
|
|
case NM_ACT_STAGE_IP_CONFIG_COMMIT:
|
|
|
|
|
case NM_ACT_STAGE_FAILED:
|
|
|
|
|
case NM_ACT_STAGE_CANCELLED:
|
|
|
|
|
case NM_ACT_STAGE_UNKNOWN:
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return activated;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_activation_should_cancel
|
|
|
|
|
*
|
|
|
|
|
* Return whether or not we've been told to cancel activation
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_activation_should_cancel (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->quit_activation);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-04-28 02:50:54 +00:00
|
|
|
static gboolean nm_ac_test (int tries, nm_completion_args args)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-04-28 02:50:54 +00:00
|
|
|
NMDevice *dev = args[0];
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_is_activating (dev))
|
|
|
|
|
{
|
2005-03-26 03:42:05 +00:00
|
|
|
if (tries % 20 == 0)
|
2005-05-08 16:31:03 +00:00
|
|
|
nm_info ("Activation (%s): waiting for device to cancel activation.", nm_device_get_iface(dev));
|
2005-03-26 03:42:05 +00:00
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
/*
|
2004-09-12 03:06:44 +00:00
|
|
|
* nm_device_activation_cancel
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
*
|
|
|
|
|
* Signal activation worker that it should stop and die.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-09-12 03:06:44 +00:00
|
|
|
void nm_device_activation_cancel (NMDevice *dev)
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_completion_args args;
|
2005-04-28 02:50:54 +00:00
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
g_return_if_fail (dev != NULL);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_assert (dev->app_data);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2004-10-11 21:32:19 +00:00
|
|
|
if (nm_device_is_activating (dev))
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
{
|
2005-05-04 18:11:40 +00:00
|
|
|
NMActRequest * req = nm_device_get_act_request (dev);
|
|
|
|
|
gboolean clear_act_request = FALSE;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2005-08-18 17:34:56 +00:00
|
|
|
nm_info ("Activation (%s): cancelling...", nm_device_get_iface (dev));
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
dev->quit_activation = TRUE;
|
2004-09-12 03:06:44 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* If the device is waiting for DHCP or a user key, force its current request to stop. */
|
2005-05-06 02:58:06 +00:00
|
|
|
if (nm_act_request_get_stage (req) == NM_ACT_STAGE_NEED_USER_KEY)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
{
|
|
|
|
|
nm_dbus_cancel_get_user_key_for_network (dev->app_data->dbus_connection, req);
|
2005-05-04 18:11:40 +00:00
|
|
|
clear_act_request = TRUE;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
}
|
2005-05-06 02:58:06 +00:00
|
|
|
else if (nm_act_request_get_stage (req) == NM_ACT_STAGE_IP_CONFIG_START)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
{
|
|
|
|
|
nm_dhcp_manager_cancel_transaction (dev->app_data->dhcp_manager, req);
|
2005-05-04 18:11:40 +00:00
|
|
|
clear_act_request = TRUE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (clear_act_request)
|
|
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
dev->act_request = NULL;
|
|
|
|
|
nm_act_request_unref (req);
|
|
|
|
|
}
|
|
|
|
|
|
2004-09-12 03:06:44 +00:00
|
|
|
/* Spin until cancelled. Possible race conditions or deadlocks here.
|
|
|
|
|
* The other problem with waiting here is that we hold up dbus traffic
|
|
|
|
|
* that we should respond to.
|
|
|
|
|
*/
|
2005-04-28 02:50:54 +00:00
|
|
|
args[0] = dev;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_wait_for_completion (NM_COMPLETION_TRIES_INFINITY, G_USEC_PER_SEC / 20, nm_ac_test, NULL, args);
|
2005-08-18 17:34:56 +00:00
|
|
|
nm_info ("Activation (%s): cancelled.", nm_device_get_iface(dev));
|
2005-04-27 18:05:16 +00:00
|
|
|
nm_schedule_state_change_signal_broadcast (dev->app_data);
|
2005-09-04 17:02:41 +00:00
|
|
|
dev->quit_activation = FALSE;
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
2005-10-02 05:02:49 +00:00
|
|
|
* nm_device_deactivate_quickly
|
2004-07-06 01:34:10 +00:00
|
|
|
*
|
2005-10-02 05:02:49 +00:00
|
|
|
* Quickly deactivate a device, for things like sleep, etc. Doesn't
|
|
|
|
|
* clean much stuff up, and nm_device_deactivate() should be called
|
|
|
|
|
* on the device eventually.
|
2004-07-06 01:34:10 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-10-02 05:02:49 +00:00
|
|
|
gboolean nm_device_deactivate_quickly (NMDevice *dev)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-07-25 02:40:19 +00:00
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-10-02 05:02:49 +00:00
|
|
|
nm_vpn_manager_deactivate_vpn_connection (dev->app_data->vpn_manager, dev);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2005-08-22 22:49:23 +00:00
|
|
|
if (nm_device_is_activated (dev))
|
2005-05-16 01:28:45 +00:00
|
|
|
nm_dbus_schedule_device_status_change_signal (dev->app_data, dev, NULL, DEVICE_NO_LONGER_ACTIVE);
|
2005-08-22 22:49:23 +00:00
|
|
|
else if (nm_device_is_activating (dev))
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_activation_cancel (dev);
|
|
|
|
|
|
2005-08-22 22:49:23 +00:00
|
|
|
/* Tear down an existing activation request, which may not have happened
|
|
|
|
|
* in nm_device_activation_cancel() above, for various reasons.
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (dev->act_request)
|
|
|
|
|
{
|
2005-08-22 22:58:06 +00:00
|
|
|
nm_dhcp_manager_cancel_transaction (dev->app_data->dhcp_manager, dev->act_request);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_act_request_unref (dev->act_request);
|
|
|
|
|
dev->act_request = NULL;
|
|
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-10-02 05:02:49 +00:00
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_deactivate
|
|
|
|
|
*
|
|
|
|
|
* Remove a device's routing table entries and IP address.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_deactivate (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
NMIP4Config * config;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
nm_info ("Deactivating device %s.", nm_device_get_iface (dev));
|
|
|
|
|
|
|
|
|
|
nm_device_deactivate_quickly (dev);
|
|
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
if (!(nm_device_get_capabilities (dev) & NM_DEVICE_CAP_NM_SUPPORTED))
|
2005-05-16 01:28:45 +00:00
|
|
|
return TRUE;
|
2004-10-21 19:00:13 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
/* Remove any device nameservers and domains */
|
|
|
|
|
if ((config = nm_device_get_ip4_config (dev)))
|
|
|
|
|
{
|
2005-09-28 14:42:57 +00:00
|
|
|
nm_named_manager_remove_ip4_config (dev->app_data->named_manager, config);
|
2005-04-15 15:43:42 +00:00
|
|
|
nm_device_set_ip4_config (dev, NULL);
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Take out any entries in the routing table and any IP address the device had. */
|
2004-08-24 00:31:47 +00:00
|
|
|
nm_system_device_flush_routes (dev);
|
|
|
|
|
nm_system_device_flush_addresses (dev);
|
2005-02-10 04:39:40 +00:00
|
|
|
nm_device_update_ip4_address (dev);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-09-09 20:02:59 +00:00
|
|
|
/* Clean up stuff, don't leave the card associated */
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
2004-07-06 04:45:00 +00:00
|
|
|
{
|
|
|
|
|
nm_device_set_essid (dev, "");
|
2004-11-03 07:47:32 +00:00
|
|
|
nm_device_set_enc_key (dev, NULL, NM_DEVICE_AUTH_METHOD_NONE);
|
2004-12-17 17:16:22 +00:00
|
|
|
nm_device_set_mode (dev, NETWORK_MODE_INFRA);
|
2005-09-11 01:46:40 +00:00
|
|
|
nm_wireless_set_scan_interval (dev->app_data, dev, NM_WIRELESS_SCAN_INTERVAL_ACTIVE);
|
2004-07-06 04:45:00 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return TRUE;
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
/*
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
* nm_device_set_user_key_for_network
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
*
|
|
|
|
|
* Called upon receipt of a NetworkManagerInfo reply with a
|
|
|
|
|
* user-supplied key.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
void nm_device_set_user_key_for_network (NMActRequest *req, const char *key, const NMEncKeyType enc_type)
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMData * data;
|
|
|
|
|
NMDevice * dev;
|
|
|
|
|
NMAccessPoint * ap;
|
|
|
|
|
const char * cancel_message = "***canceled***";
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_if_fail (key != NULL);
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
data = nm_act_request_get_data (req);
|
|
|
|
|
g_assert (data);
|
|
|
|
|
|
|
|
|
|
dev = nm_act_request_get_dev (req);
|
|
|
|
|
g_assert (dev);
|
|
|
|
|
|
|
|
|
|
ap = nm_act_request_get_ap (req);
|
|
|
|
|
g_assert (ap);
|
|
|
|
|
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
/* If the user canceled, mark the ap as invalid */
|
|
|
|
|
if (strncmp (key, cancel_message, strlen (cancel_message)) == 0)
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_ap_list_append_ap (data->invalid_ap_list, ap);
|
2005-08-22 22:49:23 +00:00
|
|
|
nm_device_deactivate (dev);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_policy_schedule_device_change_check (data);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
else
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
{
|
2005-10-10 01:21:58 +00:00
|
|
|
NMAccessPoint * allowed_ap;
|
2005-05-04 18:11:40 +00:00
|
|
|
|
|
|
|
|
/* Start off at Open System auth mode with the new key */
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_ap_set_enc_key_source (ap, key, enc_type);
|
2005-05-04 18:11:40 +00:00
|
|
|
|
|
|
|
|
/* Be sure to update NMI with the new auth mode */
|
2005-10-10 01:25:48 +00:00
|
|
|
if ((allowed_ap = nm_ap_list_get_ap_by_essid (data->allowed_ap_list, nm_ap_get_essid (ap))))
|
2005-10-10 01:21:58 +00:00
|
|
|
nm_ap_set_auth_method (allowed_ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
2005-05-04 18:11:40 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_device_activate_schedule_stage1_device_prepare (req);
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
}
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
* nm_device_ap_list_add_ap
|
2004-06-24 14:18:37 +00:00
|
|
|
*
|
|
|
|
|
* Add an access point to the devices internal AP list.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
static void nm_device_ap_list_add_ap (NMDevice *dev, NMAccessPoint *ap)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (ap != NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
nm_ap_list_append_ap (dev->options.wireless.ap_list, ap);
|
|
|
|
|
/* Transfer ownership of ap to the list by unrefing it here */
|
|
|
|
|
nm_ap_unref (ap);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_ap_list_clear
|
|
|
|
|
*
|
|
|
|
|
* Clears out the device's internal list of available access points.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_ap_list_clear (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (!dev->options.wireless.ap_list)
|
2004-06-24 14:18:37 +00:00
|
|
|
return;
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
nm_ap_list_unref (dev->options.wireless.ap_list);
|
|
|
|
|
dev->options.wireless.ap_list = NULL;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_ap_list_get_ap_by_essid
|
|
|
|
|
*
|
|
|
|
|
* Get the access point for a specific essid
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMAccessPoint *nm_device_ap_list_get_ap_by_essid (NMDevice *dev, const char *essid)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *ret_ap = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NULL);
|
|
|
|
|
g_return_val_if_fail (essid != NULL, NULL);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (!dev->options.wireless.ap_list)
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
return (NULL);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
ret_ap = nm_ap_list_get_ap_by_essid (dev->options.wireless.ap_list, essid);
|
2004-07-19 Dan Williams <dcbw@redhat.com>
* Makefile.am
- Add info-daemon directory
* configure.in
- Check for glade libs and headers
- Add info-daemon directory
* src/NetworkManagerAP.c
- nm_ap_new_from_ap(): Fix bug that resulted in an APs encryption status not getting
copied over to the new AP.
* src/NetworkManagerDbus.c
src/NetworkManagerDbus.h
- Deal with nm_device_ap_list_get_ap()->nm_device_ap_list_get_ap_by_index() change
- Remove nm_dbus_signal_need_key_for_network()
- Add disabled code for asynchronous user wep key callbacks
- Add functions for getting, setting, and cancelling user key operations
- Remove "setKeyForNetwork" device dbus method call, its on NetworkManager object instead
- Add "setKeyForNetwork" dbus method call on NetworkManager object
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- nm_device_update_link_active(): revert changes for wireless link detection, the WEP-key-is-wrong
logic is in device activation now
- nm_device_activate(): for wireless devices, if we can't associate with access point (perhaps
key is wrong) trigger get-user-key pending action
- Implement get-user-key pending action stuff, tie to dbus messages
- Rename nm_device_ap_list_get_ap() -> nm_device_ap_list_get_ap_by_index()
- Add nm_device_ap_list_get_ap_by_essid()
- Instead of copying "best" access points, ref them instead so that the key we set
sticks around
* src/NetworkManagerPolicy.c
- Deal with wrong WEP key, but right access point (and if so, return link_active = TRUE)
- Don't cancel pending actions on a device if its the same device as last iteration
- Only promote pending_device->active_device if activation was successfull
* src/Makefile.am
- Rename nmclienttest->nmtest
* info-daemon/Makefile.am
info-daemon/NetworkManagerInfo.c
info-daemon/NetworkManagerInfo.h
info-daemon/NetworkManagerInfoDbus.c
info-daemon/NetworkManagerInfoDbus.h
info-daemon/passphrase.glade
info-daemon/NetworkManagerInfo.conf
info-daemon/keyring.png
- Import sources for info-daemon, which pops up dialog for passphrase/key when
NetworkManager asks for it, and also will (soon) provide "allowed" access point
lists to NetworkManager by proxying user's GConf
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@16 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-19 06:08:52 +00:00
|
|
|
|
|
|
|
|
return (ret_ap);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-10-13 20:57:23 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_ap_list_get_ap_by_address
|
|
|
|
|
*
|
|
|
|
|
* Get the access point for a specific MAC address
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMAccessPoint *nm_device_ap_list_get_ap_by_address (NMDevice *dev, const struct ether_addr *addr)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *ret_ap = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NULL);
|
|
|
|
|
g_return_val_if_fail (addr != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
if (!dev->options.wireless.ap_list)
|
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
|
|
ret_ap = nm_ap_list_get_ap_by_address (dev->options.wireless.ap_list, addr);
|
|
|
|
|
|
|
|
|
|
return (ret_ap);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-04-27 18:05:16 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_ap_list_get_ap_by_obj_path
|
|
|
|
|
*
|
|
|
|
|
* Get the access point for a dbus object path. Requires an _unescaped_
|
|
|
|
|
* object path.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMAccessPoint *nm_device_ap_list_get_ap_by_obj_path (NMDevice *dev, const char *obj_path)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint * ret_ap = NULL;
|
|
|
|
|
char * built_path;
|
|
|
|
|
char * dev_path;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NULL);
|
|
|
|
|
g_return_val_if_fail (obj_path != NULL, NULL);
|
|
|
|
|
|
|
|
|
|
if (!dev->options.wireless.ap_list)
|
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
|
|
dev_path = nm_dbus_get_object_path_for_device (dev);
|
|
|
|
|
dev_path = nm_dbus_unescape_object_path (dev_path);
|
|
|
|
|
built_path = g_strdup_printf ("%s/Networks/", dev_path);
|
|
|
|
|
g_free (dev_path);
|
|
|
|
|
|
|
|
|
|
if (strncmp (built_path, obj_path, strlen (built_path)) == 0)
|
|
|
|
|
{
|
|
|
|
|
char *essid = g_strdup (obj_path + strlen (built_path));
|
|
|
|
|
|
|
|
|
|
ret_ap = nm_ap_list_get_ap_by_essid (dev->options.wireless.ap_list, essid);
|
|
|
|
|
g_free (essid);
|
|
|
|
|
}
|
|
|
|
|
g_free (built_path);
|
|
|
|
|
|
|
|
|
|
return (ret_ap);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-08-02 21:12:40 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_ap_list_get
|
|
|
|
|
*
|
|
|
|
|
* Return a pointer to the AP list
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-08-05 18:54:29 +00:00
|
|
|
NMAccessPointList *nm_device_ap_list_get (NMDevice *dev)
|
2004-08-02 21:12:40 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NULL);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
return (dev->options.wireless.ap_list);
|
2004-08-02 21:12:40 +00:00
|
|
|
}
|
|
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
|
2005-05-05 16:52:04 +00:00
|
|
|
static gboolean link_to_specific_ap (NMDevice *dev, NMAccessPoint *ap, gboolean default_link)
|
2005-05-04 18:11:40 +00:00
|
|
|
{
|
|
|
|
|
gboolean link = FALSE;
|
|
|
|
|
|
2005-05-05 16:52:04 +00:00
|
|
|
/* Checking hardware's ESSID during a scan is doesn't work. */
|
|
|
|
|
nm_lock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
|
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
if (nm_device_wireless_is_associated (dev))
|
|
|
|
|
{
|
|
|
|
|
char * dev_essid = nm_device_get_essid (dev);
|
|
|
|
|
char * ap_essid = nm_ap_get_essid (ap);
|
|
|
|
|
|
2005-05-05 16:52:04 +00:00
|
|
|
if (dev_essid && ap_essid && !strcmp (dev_essid, ap_essid))
|
|
|
|
|
{
|
|
|
|
|
dev->options.wireless.failed_link_count = 0;
|
|
|
|
|
link = TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
|
|
|
|
|
|
|
|
|
if (!link)
|
|
|
|
|
{
|
|
|
|
|
dev->options.wireless.failed_link_count++;
|
|
|
|
|
if (dev->options.wireless.failed_link_count < 3)
|
|
|
|
|
link = default_link;
|
2005-05-04 18:11:40 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return link;
|
|
|
|
|
}
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
/*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* nm_device_update_best_ap
|
2004-07-06 01:34:10 +00:00
|
|
|
*
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
* Recalculate the "best" access point we should be associating with.
|
2004-10-29 16:37:43 +00:00
|
|
|
*
|
2004-07-06 01:34:10 +00:00
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMAccessPoint * nm_device_get_best_ap (NMDevice *dev)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
NMAccessPointList * ap_list;
|
|
|
|
|
NMAPListIter * iter;
|
|
|
|
|
NMAccessPoint * scan_ap = NULL;
|
|
|
|
|
NMAccessPoint * best_ap = NULL;
|
|
|
|
|
NMAccessPoint * cur_ap = NULL;
|
|
|
|
|
NMActRequest * req = NULL;
|
|
|
|
|
NMAccessPoint * trusted_best_ap = NULL;
|
|
|
|
|
NMAccessPoint * untrusted_best_ap = NULL;
|
|
|
|
|
GTimeVal trusted_latest_timestamp = {0, 0};
|
|
|
|
|
GTimeVal untrusted_latest_timestamp = {0, 0};
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
2004-07-15 Dan Williams <dcbw@redhat.com>
* src/Makefile.am
- Turn on warnings
* src/NetworkManager.c
- nm_create_device_and_add_to_list(): call nm_device_deactivate() rather
that doing the deactivation ourselves
- Cancel an pending actions on a device if its being removed
- Break up link state checking a bit, make non-active wireless cards
deactivated to save power
- Remove unused variables
* src/NetworkManager.h
- Add support for "pending" device
* src/NetworkManagerAP.h
src/NetworkManagerAP.c
- Add support for determining whether and AP has encryption enabled or not
- AP address is now "struct ether_addr" rather than a string
* src/NetworkManagerDbus.h
src/NetworkManagerDbus.c
- Add signal NeedKeyForNetwork, method SetKeyForNetwork (testing only)
- Changes for AP address from struct ether_addr->string
* src/NetworkManagerDevice.h
src/NetworkManagerDevice.c
- Remove unused variables, fix warnings
- Add support for Pending Actions (things that block a device from being "active"
until they are completed).
- First pending action: Get a WEP key from the user
- Add nm_device_is_wire[d|less](), rename nm_device_is_wireless()
- Clean up explicit testing of dev->iface_type to use nm_device_is_wireless()
- Update wireless link checking to try to determine if the AP we are associated
with is correct, but the WEP key we are using is just wrong. If its wrong,
trigger the GetUserKey pending action on the device
- If dhclient can't get an IP address, it brings the device down. Bring it back
up in that case, otherwise we can't scan or link-check on it
- Add IP address change notifications at appropriate points (still needs some work)
- Add nm_device_need_ap_switch(), checks whether we need to switch access points or not
* src/NetworkManagerPolicy.h
src/NetworkManagerPolicy.c
- Split out "best" access point determiniation into separate function
- Make device activation 2-stage: first the device is pending, then
in the next iteration through it becomes "active" unless it has
pending actions
* src/NetworkManagerUtils.h
src/NetworkManagerUtils.c
- Clean up unused variables and warnings
- Wrap our debug macros in {} to prevent possible confusion
* src/NetworkManagerWireless.c
- Forgot to return current best priority, which lead to last available AP always
being chosen no matter what its priority was. Corrected.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@15 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-07-15 16:51:48 +00:00
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), NULL);
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_assert (dev->app_data);
|
2004-08-05 18:54:29 +00:00
|
|
|
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
/* Devices that can't scan don't do anything automatic.
|
|
|
|
|
* The user must choose the access point from the menu.
|
|
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (!nm_device_get_supports_wireless_scan (dev) && !nm_device_has_active_link (dev))
|
|
|
|
|
return NULL;
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (!(ap_list = nm_device_ap_list_get (dev)))
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return NULL;
|
2004-08-05 18:54:29 +00:00
|
|
|
|
2005-05-04 18:11:40 +00:00
|
|
|
/* We prefer the currently selected access point if its user-chosen or if there
|
|
|
|
|
* is still a hardware link to it.
|
2004-08-12 Dan Williams <dcbw@redhat.com>
* info-daemon/passphrase.glade
- Set window title to " "
* panel-applet/Makefile.am
panel-applet/keyring.png
- Deliver to correct place
* panel-applet/NMWirelessApplet.[ch]
- Add comments
- Remove applet->have_active_device as its no longer used
- (nmwa_load_theme): load keyring.png too
- (error_dialog): remove
- (show_warning_dialog): subsume functionality of error dialog too
- (nmwa_destroy, nmwa_new): create and dispose of an application-wide GConfClient
- (nmwa_handle_network_choice): add to deal with user clicking on an item from
the networks menu
- (nmwa_menu_item_activated): GtkMenuItem "activate" signal handler
- (nmwa_button_clicked, nmwa_setup_widgets): create and populate the menu on startup
and when we get broadcasts of changed wireless access points only, not when the
user clicks on the button to display the menu (too long of a wait)
- (nmwa_add_menu_item): Make active network bold, and place a keyring icon beside
networks that are encrypted
- (nmwa_dispose_menu, nmwa_menu_item_data_free): dispose of the data we place on each
menu item with g_object_set_data()
* panel-applet/NMWirelessAppletDbus.[ch]
- (nmwa_dbus_get_bool): add method to return boolean value from dbus message
- (nmwa_dbus_get_active_network): add (nmwa_dbus_get_string() wrapper to get active network)
- (nmwa_dbus_add_networks_to_menu): clean up, only show one instance of each ESSID in the menu
- (nmwa_dbus_set_network): force NetworkManager to use a particular network for wireless cards
- (nmwa_dbus_init, nmwa_dbus_filter): Trap network appear/disappear and device
activation/deactivation signals and rebuild the menu when they happen
* src/NetworkManager.c
- (main): use new nm_spawn_process() rather than system()
* src/NetworkManagerDbus.c
- (nm_dbus_devices_handle_request): don't compare AP structure addresses directly, but essids
instead. Since we can now force best_aps to stick around, the AP structure to which
dev->options.wireless.best_ap points to won't necessarily be in the device's device list
if a scan has happened since the best_ap was frozen. Also add "setNetwork" method
to freeze the best_ap.
* src/NetworkManagerDevice.[ch]
- (nm_device_activation_worker): Use new nm_spawn_process() call rather than system()
- (nm_device_*_best_ap): add freeze/unfreeze/get_frozen functions, and don't really update
the best_ap in nm_device_update_best_ap() if the best_ap is frozen AND in the device's
ap list
* src/NetworkManagerUtils.[ch]
- (nm_spawn_process): add replacement for system() usage
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@48 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-12 19:58:01 +00:00
|
|
|
*/
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if ((req = nm_device_get_act_request (dev)))
|
2004-08-12 Dan Williams <dcbw@redhat.com>
* info-daemon/passphrase.glade
- Set window title to " "
* panel-applet/Makefile.am
panel-applet/keyring.png
- Deliver to correct place
* panel-applet/NMWirelessApplet.[ch]
- Add comments
- Remove applet->have_active_device as its no longer used
- (nmwa_load_theme): load keyring.png too
- (error_dialog): remove
- (show_warning_dialog): subsume functionality of error dialog too
- (nmwa_destroy, nmwa_new): create and dispose of an application-wide GConfClient
- (nmwa_handle_network_choice): add to deal with user clicking on an item from
the networks menu
- (nmwa_menu_item_activated): GtkMenuItem "activate" signal handler
- (nmwa_button_clicked, nmwa_setup_widgets): create and populate the menu on startup
and when we get broadcasts of changed wireless access points only, not when the
user clicks on the button to display the menu (too long of a wait)
- (nmwa_add_menu_item): Make active network bold, and place a keyring icon beside
networks that are encrypted
- (nmwa_dispose_menu, nmwa_menu_item_data_free): dispose of the data we place on each
menu item with g_object_set_data()
* panel-applet/NMWirelessAppletDbus.[ch]
- (nmwa_dbus_get_bool): add method to return boolean value from dbus message
- (nmwa_dbus_get_active_network): add (nmwa_dbus_get_string() wrapper to get active network)
- (nmwa_dbus_add_networks_to_menu): clean up, only show one instance of each ESSID in the menu
- (nmwa_dbus_set_network): force NetworkManager to use a particular network for wireless cards
- (nmwa_dbus_init, nmwa_dbus_filter): Trap network appear/disappear and device
activation/deactivation signals and rebuild the menu when they happen
* src/NetworkManager.c
- (main): use new nm_spawn_process() rather than system()
* src/NetworkManagerDbus.c
- (nm_dbus_devices_handle_request): don't compare AP structure addresses directly, but essids
instead. Since we can now force best_aps to stick around, the AP structure to which
dev->options.wireless.best_ap points to won't necessarily be in the device's device list
if a scan has happened since the best_ap was frozen. Also add "setNetwork" method
to freeze the best_ap.
* src/NetworkManagerDevice.[ch]
- (nm_device_activation_worker): Use new nm_spawn_process() call rather than system()
- (nm_device_*_best_ap): add freeze/unfreeze/get_frozen functions, and don't really update
the best_ap in nm_device_update_best_ap() if the best_ap is frozen AND in the device's
ap list
* src/NetworkManagerUtils.[ch]
- (nm_spawn_process): add replacement for system() usage
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@48 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-12 19:58:01 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if ((cur_ap = nm_act_request_get_ap (req)))
|
2004-08-12 Dan Williams <dcbw@redhat.com>
* info-daemon/passphrase.glade
- Set window title to " "
* panel-applet/Makefile.am
panel-applet/keyring.png
- Deliver to correct place
* panel-applet/NMWirelessApplet.[ch]
- Add comments
- Remove applet->have_active_device as its no longer used
- (nmwa_load_theme): load keyring.png too
- (error_dialog): remove
- (show_warning_dialog): subsume functionality of error dialog too
- (nmwa_destroy, nmwa_new): create and dispose of an application-wide GConfClient
- (nmwa_handle_network_choice): add to deal with user clicking on an item from
the networks menu
- (nmwa_menu_item_activated): GtkMenuItem "activate" signal handler
- (nmwa_button_clicked, nmwa_setup_widgets): create and populate the menu on startup
and when we get broadcasts of changed wireless access points only, not when the
user clicks on the button to display the menu (too long of a wait)
- (nmwa_add_menu_item): Make active network bold, and place a keyring icon beside
networks that are encrypted
- (nmwa_dispose_menu, nmwa_menu_item_data_free): dispose of the data we place on each
menu item with g_object_set_data()
* panel-applet/NMWirelessAppletDbus.[ch]
- (nmwa_dbus_get_bool): add method to return boolean value from dbus message
- (nmwa_dbus_get_active_network): add (nmwa_dbus_get_string() wrapper to get active network)
- (nmwa_dbus_add_networks_to_menu): clean up, only show one instance of each ESSID in the menu
- (nmwa_dbus_set_network): force NetworkManager to use a particular network for wireless cards
- (nmwa_dbus_init, nmwa_dbus_filter): Trap network appear/disappear and device
activation/deactivation signals and rebuild the menu when they happen
* src/NetworkManager.c
- (main): use new nm_spawn_process() rather than system()
* src/NetworkManagerDbus.c
- (nm_dbus_devices_handle_request): don't compare AP structure addresses directly, but essids
instead. Since we can now force best_aps to stick around, the AP structure to which
dev->options.wireless.best_ap points to won't necessarily be in the device's device list
if a scan has happened since the best_ap was frozen. Also add "setNetwork" method
to freeze the best_ap.
* src/NetworkManagerDevice.[ch]
- (nm_device_activation_worker): Use new nm_spawn_process() call rather than system()
- (nm_device_*_best_ap): add freeze/unfreeze/get_frozen functions, and don't really update
the best_ap in nm_device_update_best_ap() if the best_ap is frozen AND in the device's
ap list
* src/NetworkManagerUtils.[ch]
- (nm_spawn_process): add replacement for system() usage
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@48 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-12 19:58:01 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
char * essid = nm_ap_get_essid (cur_ap);
|
|
|
|
|
gboolean keep = FALSE;
|
|
|
|
|
|
|
|
|
|
if (nm_ap_get_user_created (cur_ap))
|
|
|
|
|
keep = TRUE;
|
|
|
|
|
else if (nm_act_request_get_user_requested (req))
|
|
|
|
|
keep = TRUE;
|
2005-05-05 16:52:04 +00:00
|
|
|
else if (link_to_specific_ap (dev, cur_ap, TRUE))
|
|
|
|
|
keep = TRUE;
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
|
|
|
|
/* Only keep if its not in the invalid list and its _is_ in our scaned list */
|
|
|
|
|
if ( keep
|
|
|
|
|
&& !nm_ap_list_get_ap_by_essid (dev->app_data->invalid_ap_list, essid)
|
|
|
|
|
&& nm_device_ap_list_get_ap_by_essid (dev, essid))
|
2004-10-29 16:37:43 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_ap_ref (cur_ap);
|
|
|
|
|
return cur_ap;
|
2004-10-29 16:37:43 +00:00
|
|
|
}
|
2004-08-12 Dan Williams <dcbw@redhat.com>
* info-daemon/passphrase.glade
- Set window title to " "
* panel-applet/Makefile.am
panel-applet/keyring.png
- Deliver to correct place
* panel-applet/NMWirelessApplet.[ch]
- Add comments
- Remove applet->have_active_device as its no longer used
- (nmwa_load_theme): load keyring.png too
- (error_dialog): remove
- (show_warning_dialog): subsume functionality of error dialog too
- (nmwa_destroy, nmwa_new): create and dispose of an application-wide GConfClient
- (nmwa_handle_network_choice): add to deal with user clicking on an item from
the networks menu
- (nmwa_menu_item_activated): GtkMenuItem "activate" signal handler
- (nmwa_button_clicked, nmwa_setup_widgets): create and populate the menu on startup
and when we get broadcasts of changed wireless access points only, not when the
user clicks on the button to display the menu (too long of a wait)
- (nmwa_add_menu_item): Make active network bold, and place a keyring icon beside
networks that are encrypted
- (nmwa_dispose_menu, nmwa_menu_item_data_free): dispose of the data we place on each
menu item with g_object_set_data()
* panel-applet/NMWirelessAppletDbus.[ch]
- (nmwa_dbus_get_bool): add method to return boolean value from dbus message
- (nmwa_dbus_get_active_network): add (nmwa_dbus_get_string() wrapper to get active network)
- (nmwa_dbus_add_networks_to_menu): clean up, only show one instance of each ESSID in the menu
- (nmwa_dbus_set_network): force NetworkManager to use a particular network for wireless cards
- (nmwa_dbus_init, nmwa_dbus_filter): Trap network appear/disappear and device
activation/deactivation signals and rebuild the menu when they happen
* src/NetworkManager.c
- (main): use new nm_spawn_process() rather than system()
* src/NetworkManagerDbus.c
- (nm_dbus_devices_handle_request): don't compare AP structure addresses directly, but essids
instead. Since we can now force best_aps to stick around, the AP structure to which
dev->options.wireless.best_ap points to won't necessarily be in the device's device list
if a scan has happened since the best_ap was frozen. Also add "setNetwork" method
to freeze the best_ap.
* src/NetworkManagerDevice.[ch]
- (nm_device_activation_worker): Use new nm_spawn_process() call rather than system()
- (nm_device_*_best_ap): add freeze/unfreeze/get_frozen functions, and don't really update
the best_ap in nm_device_update_best_ap() if the best_ap is frozen AND in the device's
ap list
* src/NetworkManagerUtils.[ch]
- (nm_spawn_process): add replacement for system() usage
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@48 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-12 19:58:01 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (!(iter = nm_ap_list_iter_new (ap_list)))
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return NULL;
|
2004-10-29 16:37:43 +00:00
|
|
|
while ((scan_ap = nm_ap_list_iter_next (iter)))
|
2004-08-05 18:54:29 +00:00
|
|
|
{
|
2004-09-08 18:14:42 +00:00
|
|
|
NMAccessPoint *tmp_ap;
|
2004-10-29 16:37:43 +00:00
|
|
|
char *ap_essid = nm_ap_get_essid (scan_ap);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2004-08-06 18:19:06 +00:00
|
|
|
/* Access points in the "invalid" list cannot be used */
|
2004-09-08 18:14:42 +00:00
|
|
|
if (nm_ap_list_get_ap_by_essid (dev->app_data->invalid_ap_list, ap_essid))
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
if ((tmp_ap = nm_ap_list_get_ap_by_essid (dev->app_data->allowed_ap_list, ap_essid)))
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2004-08-29 05:46:59 +00:00
|
|
|
const GTimeVal *curtime = nm_ap_get_timestamp (tmp_ap);
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-09-15 14:47:17 +00:00
|
|
|
gboolean blacklisted = nm_ap_has_manufacturer_default_essid (scan_ap);
|
|
|
|
|
if (blacklisted)
|
|
|
|
|
{
|
|
|
|
|
GSList *elt, *user_addrs;
|
|
|
|
|
const struct ether_addr *ap_addr;
|
|
|
|
|
char char_addr[20];
|
|
|
|
|
|
|
|
|
|
ap_addr = nm_ap_get_address (scan_ap);
|
|
|
|
|
user_addrs = nm_ap_get_user_addresses (tmp_ap);
|
|
|
|
|
|
|
|
|
|
memset (&char_addr[0], 0, 20);
|
|
|
|
|
ether_ntoa_r (ap_addr, &char_addr[0]);
|
|
|
|
|
|
|
|
|
|
for (elt = user_addrs; elt; elt = g_slist_next (elt))
|
|
|
|
|
{
|
|
|
|
|
if (elt->data && !strcmp (elt->data, &char_addr[0]))
|
|
|
|
|
{
|
|
|
|
|
blacklisted = FALSE;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
g_slist_foreach (user_addrs, (GFunc)g_free, NULL);
|
|
|
|
|
g_slist_free (user_addrs);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!blacklisted && nm_ap_get_trusted (tmp_ap) && (curtime->tv_sec > trusted_latest_timestamp.tv_sec))
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2004-09-08 18:14:42 +00:00
|
|
|
trusted_latest_timestamp = *nm_ap_get_timestamp (tmp_ap);
|
2004-10-29 16:37:43 +00:00
|
|
|
trusted_best_ap = scan_ap;
|
2004-09-08 18:14:42 +00:00
|
|
|
/* Merge access point data (mainly to get updated WEP key) */
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_enc_key_source (trusted_best_ap, nm_ap_get_enc_key_source (tmp_ap), nm_ap_get_enc_type (tmp_ap));
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
2005-09-15 14:47:17 +00:00
|
|
|
else if (!blacklisted && !nm_ap_get_trusted (tmp_ap) && (curtime->tv_sec > untrusted_latest_timestamp.tv_sec))
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
{
|
2004-09-08 18:14:42 +00:00
|
|
|
untrusted_latest_timestamp = *nm_ap_get_timestamp (tmp_ap);
|
2004-10-29 16:37:43 +00:00
|
|
|
untrusted_best_ap = scan_ap;
|
2004-09-08 18:14:42 +00:00
|
|
|
/* Merge access point data (mainly to get updated WEP key) */
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_enc_key_source (untrusted_best_ap, nm_ap_get_enc_key_source (tmp_ap), nm_ap_get_enc_type (tmp_ap));
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
}
|
2004-08-06 18:19:06 +00:00
|
|
|
}
|
|
|
|
|
}
|
2004-09-08 18:14:42 +00:00
|
|
|
best_ap = trusted_best_ap ? trusted_best_ap : untrusted_best_ap;
|
|
|
|
|
nm_ap_list_iter_free (iter);
|
2004-08-06 18:19:06 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (best_ap)
|
|
|
|
|
nm_ap_ref (best_ap);
|
2004-08-05 18:54:29 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return best_ap;
|
|
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_get_activation_ap
|
|
|
|
|
*
|
|
|
|
|
* Return an access point suitable for use in the device activation
|
|
|
|
|
* request.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
NMAccessPoint * nm_device_wireless_get_activation_ap (NMDevice *dev, const char *essid, const char *key, NMEncKeyType key_type)
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
|
|
|
|
gboolean encrypted = FALSE;
|
2004-10-27 19:02:07 +00:00
|
|
|
NMAccessPoint *ap = NULL;
|
2005-01-21 19:32:08 +00:00
|
|
|
NMAccessPoint *tmp_ap = NULL;
|
2004-10-13 20:57:23 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (essid != NULL, NULL);
|
2004-10-13 20:57:23 +00:00
|
|
|
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_debug ("Forcing AP '%s'", essid);
|
2004-10-13 20:57:23 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if ( key
|
|
|
|
|
&& strlen (key)
|
|
|
|
|
&& (key_type != NM_ENC_TYPE_UNKNOWN)
|
|
|
|
|
&& (key_type != NM_ENC_TYPE_NONE))
|
|
|
|
|
encrypted = TRUE;
|
2004-11-22 14:42:34 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Find the AP in our card's scan list first.
|
|
|
|
|
* If its not there, create an entirely new AP.
|
|
|
|
|
*/
|
|
|
|
|
if (!(ap = nm_ap_list_get_ap_by_essid (nm_device_ap_list_get (dev), essid)))
|
2004-11-16 15:06:46 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Okay, the card didn't see it in the scan, Cisco cards sometimes do this.
|
|
|
|
|
* So we make a "fake" access point and add it to the scan list.
|
2004-12-17 17:16:22 +00:00
|
|
|
*/
|
2005-01-21 19:32:08 +00:00
|
|
|
ap = nm_ap_new ();
|
|
|
|
|
nm_ap_set_essid (ap, essid);
|
|
|
|
|
nm_ap_set_encrypted (ap, encrypted);
|
|
|
|
|
if (encrypted)
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
|
|
|
|
else
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_NONE);
|
|
|
|
|
nm_ap_set_artificial (ap, TRUE);
|
|
|
|
|
nm_ap_list_append_ap (nm_device_ap_list_get (dev), ap);
|
|
|
|
|
nm_ap_unref (ap);
|
2004-11-16 15:06:46 +00:00
|
|
|
}
|
2005-04-04 21:24:58 +00:00
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* If the AP is in the ignore list, we have to remove it since
|
|
|
|
|
* the User Knows What's Best.
|
|
|
|
|
*/
|
|
|
|
|
nm_ap_list_remove_ap_by_essid (dev->app_data->invalid_ap_list, nm_ap_get_essid (ap));
|
|
|
|
|
}
|
2004-11-16 15:06:46 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Now that this AP has an essid, copy over encryption keys and whatnot */
|
|
|
|
|
if ((tmp_ap = nm_ap_list_get_ap_by_essid (dev->app_data->allowed_ap_list, nm_ap_get_essid (ap))))
|
2004-10-29 19:52:35 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_enc_key_source (ap, nm_ap_get_enc_key_source (tmp_ap), nm_ap_get_enc_type (tmp_ap));
|
|
|
|
|
nm_ap_set_auth_method (ap, nm_ap_get_auth_method (tmp_ap));
|
|
|
|
|
nm_ap_set_invalid (ap, nm_ap_get_invalid (tmp_ap));
|
|
|
|
|
nm_ap_set_timestamp (ap, nm_ap_get_timestamp (tmp_ap));
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Use the encryption key and type the user sent us if its valid */
|
|
|
|
|
if (encrypted)
|
|
|
|
|
nm_ap_set_enc_key_source (ap, key, key_type);
|
2004-10-15 17:04:57 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
return ap;
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_fake_ap_list
|
|
|
|
|
*
|
|
|
|
|
* Fake the access point list, used for test devices.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static void nm_device_fake_ap_list (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
#define NUM_FAKE_APS 4
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
int i;
|
|
|
|
|
NMAccessPointList * old_ap_list = nm_device_ap_list_get (dev);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
2005-07-13 16:58:13 +00:00
|
|
|
const char *fake_essids[NUM_FAKE_APS] = { "green", "bay", "packers", "rule" };
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
struct ether_addr fake_addrs[NUM_FAKE_APS] = {{{0x70, 0x37, 0x03, 0x70, 0x37, 0x03}},
|
|
|
|
|
{{0x12, 0x34, 0x56, 0x78, 0x90, 0xab}},
|
|
|
|
|
{{0xcd, 0xef, 0x12, 0x34, 0x56, 0x78}},
|
|
|
|
|
{{0x90, 0xab, 0xcd, 0xef, 0x12, 0x34}} };
|
|
|
|
|
guint8 fake_qualities[NUM_FAKE_APS] = { 150, 26, 200, 100 };
|
|
|
|
|
double fake_freqs[NUM_FAKE_APS] = { 3.1416, 4.1416, 5.1415, 6.1415 };
|
|
|
|
|
gboolean fake_enc[NUM_FAKE_APS] = { FALSE, TRUE, FALSE, TRUE };
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
|
|
|
|
|
dev->options.wireless.ap_list = nm_ap_list_new (NETWORK_TYPE_DEVICE);
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < NUM_FAKE_APS; i++)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *nm_ap = nm_ap_new ();
|
|
|
|
|
NMAccessPoint *list_ap;
|
|
|
|
|
|
|
|
|
|
/* Copy over info from scan to local structure */
|
|
|
|
|
nm_ap_set_essid (nm_ap, fake_essids[i]);
|
|
|
|
|
|
|
|
|
|
if (fake_enc[i])
|
|
|
|
|
nm_ap_set_encrypted (nm_ap, FALSE);
|
|
|
|
|
else
|
|
|
|
|
nm_ap_set_encrypted (nm_ap, TRUE);
|
|
|
|
|
|
|
|
|
|
nm_ap_set_address (nm_ap, (const struct ether_addr *)(&fake_addrs[i]));
|
2004-09-13 17:43:16 +00:00
|
|
|
nm_ap_set_strength (nm_ap, fake_qualities[i]);
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
nm_ap_set_freq (nm_ap, fake_freqs[i]);
|
|
|
|
|
|
2005-01-27 20:14:12 +00:00
|
|
|
/* Merge settings from wireless networks, mainly keys */
|
2004-09-08 18:14:42 +00:00
|
|
|
if ((list_ap = nm_ap_list_get_ap_by_essid (dev->app_data->allowed_ap_list, nm_ap_get_essid (nm_ap))))
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
{
|
|
|
|
|
nm_ap_set_timestamp (nm_ap, nm_ap_get_timestamp (list_ap));
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_enc_key_source (nm_ap, nm_ap_get_enc_key_source (list_ap), nm_ap_get_enc_type (list_ap));
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Add the AP to the device's AP list */
|
|
|
|
|
nm_device_ap_list_add_ap (dev, nm_ap);
|
|
|
|
|
}
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
if (nm_device_get_act_request (dev))
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
nm_ap_list_diff (dev->app_data, dev, old_ap_list, nm_device_ap_list_get (dev));
|
|
|
|
|
if (old_ap_list)
|
|
|
|
|
nm_ap_list_unref (old_ap_list);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
/*
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
* nm_device_wireless_schedule_scan
|
2004-07-06 01:34:10 +00:00
|
|
|
*
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
* Schedule a wireless scan in the /device's/ thread.
|
2004-07-06 01:34:10 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
static void nm_device_wireless_schedule_scan (NMDevice *dev)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2005-04-01 21:30:12 +00:00
|
|
|
GSource *wscan_source;
|
|
|
|
|
guint wscan_source_id;
|
|
|
|
|
NMWirelessScanCB *scan_cb;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
g_return_if_fail (dev != NULL);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
2005-04-01 21:30:12 +00:00
|
|
|
scan_cb = g_malloc0 (sizeof (NMWirelessScanCB));
|
|
|
|
|
scan_cb->dev = dev;
|
2005-09-02 03:47:19 +00:00
|
|
|
scan_cb->force = FALSE;
|
2005-04-01 21:30:12 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
wscan_source = g_timeout_source_new (dev->options.wireless.scan_interval * 1000);
|
2005-04-01 21:30:12 +00:00
|
|
|
g_source_set_callback (wscan_source, nm_device_wireless_scan, scan_cb, NULL);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
wscan_source_id = g_source_attach (wscan_source, dev->context);
|
|
|
|
|
g_source_unref (wscan_source);
|
|
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_wireless_process_scan_results
|
|
|
|
|
*
|
|
|
|
|
* Process results of an iwscan() into our own AP lists. We're an idle function,
|
|
|
|
|
* but we never reschedule ourselves.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_wireless_process_scan_results (gpointer user_data)
|
|
|
|
|
{
|
|
|
|
|
NMWirelessScanResults *results = (NMWirelessScanResults *)user_data;
|
|
|
|
|
NMDevice *dev;
|
|
|
|
|
wireless_scan *tmp_ap;
|
|
|
|
|
NMAPListIter *iter;
|
|
|
|
|
GTimeVal cur_time;
|
|
|
|
|
gboolean list_changed = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (results != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
dev = results->dev;
|
|
|
|
|
if (!dev || !results->scan_head.result)
|
|
|
|
|
return FALSE;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-02-27 06:11:16 +00:00
|
|
|
g_get_current_time (&cur_time);
|
|
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
/* Translate iwlib scan results to NM access point list */
|
2005-02-10 04:39:40 +00:00
|
|
|
for (tmp_ap = results->scan_head.result; tmp_ap; tmp_ap = tmp_ap->next)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
/* We need at least an ESSID or a MAC address for each access point */
|
|
|
|
|
if (tmp_ap->b.has_essid || tmp_ap->has_ap_addr)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
NMAccessPoint *nm_ap = nm_ap_new ();
|
2005-01-31 19:53:38 +00:00
|
|
|
int percent;
|
2005-02-27 06:11:16 +00:00
|
|
|
gboolean new = FALSE;
|
|
|
|
|
gboolean strength_changed = FALSE;
|
|
|
|
|
gboolean success = FALSE;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Copy over info from scan to local structure */
|
|
|
|
|
|
|
|
|
|
/* ipw2x00 drivers fill in an essid of "<hidden>" if they think the access point
|
|
|
|
|
* is hiding its MAC address. Sigh.
|
2004-12-17 17:16:22 +00:00
|
|
|
*/
|
2005-01-07 18:07:06 +00:00
|
|
|
if ( !tmp_ap->b.has_essid
|
|
|
|
|
|| (tmp_ap->b.essid && !strlen (tmp_ap->b.essid))
|
2005-02-27 06:11:16 +00:00
|
|
|
|| (tmp_ap->b.essid && !strcmp (tmp_ap->b.essid, "<hidden>"))) /* Stupid ipw drivers use <hidden> */
|
2005-01-07 18:07:06 +00:00
|
|
|
nm_ap_set_essid (nm_ap, NULL);
|
|
|
|
|
else
|
|
|
|
|
nm_ap_set_essid (nm_ap, tmp_ap->b.essid);
|
|
|
|
|
|
|
|
|
|
if (tmp_ap->b.has_key && (tmp_ap->b.key_flags & IW_ENCODE_DISABLED))
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
nm_ap_set_encrypted (nm_ap, FALSE);
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_auth_method (nm_ap, NM_DEVICE_AUTH_METHOD_NONE);
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
else
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
nm_ap_set_encrypted (nm_ap, TRUE);
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_auth_method (nm_ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
if (tmp_ap->has_ap_addr)
|
|
|
|
|
nm_ap_set_address (nm_ap, (const struct ether_addr *)(tmp_ap->ap_addr.sa_data));
|
|
|
|
|
|
|
|
|
|
if (tmp_ap->b.has_mode)
|
|
|
|
|
{
|
|
|
|
|
NMNetworkMode mode = NETWORK_MODE_INFRA;
|
|
|
|
|
switch (tmp_ap->b.mode)
|
|
|
|
|
{
|
|
|
|
|
case IW_MODE_INFRA:
|
|
|
|
|
mode = NETWORK_MODE_INFRA;
|
|
|
|
|
break;
|
|
|
|
|
case IW_MODE_ADHOC:
|
|
|
|
|
mode = NETWORK_MODE_ADHOC;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
mode = NETWORK_MODE_INFRA;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
nm_ap_set_mode (nm_ap, mode);
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
nm_ap_set_mode (nm_ap, NETWORK_MODE_INFRA);
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
percent = nm_wireless_qual_to_percent (&(tmp_ap->stats.qual),
|
|
|
|
|
(const iwqual *)(&dev->options.wireless.max_qual),
|
|
|
|
|
(const iwqual *)(&dev->options.wireless.avg_qual));
|
|
|
|
|
nm_ap_set_strength (nm_ap, percent);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
if (tmp_ap->b.has_freq)
|
|
|
|
|
nm_ap_set_freq (nm_ap, tmp_ap->b.freq);
|
|
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
nm_ap_set_last_seen (nm_ap, &cur_time);
|
|
|
|
|
|
2005-02-27 06:11:16 +00:00
|
|
|
/* If the AP is not broadcasting its ESSID, try to fill it in here from our
|
|
|
|
|
* allowed list where we cache known MAC->ESSID associations.
|
|
|
|
|
*/
|
|
|
|
|
if (!nm_ap_get_essid (nm_ap))
|
|
|
|
|
nm_ap_list_copy_one_essid_by_address (nm_ap, dev->app_data->allowed_ap_list);
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Add the AP to the device's AP list */
|
2005-02-27 06:11:16 +00:00
|
|
|
success = nm_ap_list_merge_scanned_ap (nm_device_ap_list_get (dev), nm_ap, &new, &strength_changed);
|
|
|
|
|
if (success)
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
/* Handle dbus signals that we need to broadcast when the AP is added to the list or changes strength */
|
2005-02-27 06:11:16 +00:00
|
|
|
if (new)
|
|
|
|
|
{
|
|
|
|
|
nm_dbus_signal_wireless_network_change (dev->app_data->dbus_connection, dev, nm_ap,
|
|
|
|
|
NETWORK_STATUS_APPEARED, -1);
|
|
|
|
|
list_changed = TRUE;
|
|
|
|
|
}
|
|
|
|
|
else if (strength_changed)
|
|
|
|
|
{
|
|
|
|
|
nm_dbus_signal_wireless_network_change (dev->app_data->dbus_connection, dev, nm_ap,
|
|
|
|
|
NETWORK_STATUS_STRENGTH_CHANGED, nm_ap_get_strength (nm_ap));
|
|
|
|
|
}
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
nm_ap_unref (nm_ap);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Once we have the list, copy in any relevant information from our Allowed list. */
|
|
|
|
|
nm_ap_list_copy_properties (nm_device_ap_list_get (dev), dev->app_data->allowed_ap_list);
|
|
|
|
|
|
2005-02-27 06:11:16 +00:00
|
|
|
/* Walk the access point list and remove any access points older than 120s */
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
g_get_current_time (&cur_time);
|
|
|
|
|
if (nm_device_ap_list_get (dev) && (iter = nm_ap_list_iter_new (nm_device_ap_list_get (dev))))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-05-04 18:11:40 +00:00
|
|
|
NMAccessPoint *outdated_ap;
|
|
|
|
|
GSList * outdated_list = NULL;
|
|
|
|
|
GSList * elt;
|
|
|
|
|
NMActRequest * req = nm_device_get_act_request (dev);
|
|
|
|
|
NMAccessPoint *cur_ap = NULL;
|
|
|
|
|
|
|
|
|
|
if (req)
|
|
|
|
|
{
|
|
|
|
|
cur_ap = nm_act_request_get_ap (req);
|
|
|
|
|
g_assert (cur_ap);
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
while ((outdated_ap = nm_ap_list_iter_next (iter)))
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
2005-01-29 23:34:31 +00:00
|
|
|
const GTimeVal *ap_time = nm_ap_get_last_seen (outdated_ap);
|
|
|
|
|
gboolean keep_around = FALSE;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-02-27 06:11:16 +00:00
|
|
|
/* Don't ever get prune the AP we're currently associated with */
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
if ( nm_ap_get_essid (outdated_ap)
|
2005-05-04 18:11:40 +00:00
|
|
|
&& (cur_ap && (nm_null_safe_strcmp (nm_ap_get_essid (cur_ap), nm_ap_get_essid (outdated_ap))) == 0))
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
keep_around = TRUE;
|
|
|
|
|
|
2005-02-27 06:11:16 +00:00
|
|
|
if (!keep_around && (ap_time->tv_sec + 120 < cur_time.tv_sec))
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
outdated_list = g_slist_append (outdated_list, outdated_ap);
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
|
|
|
|
nm_ap_list_iter_free (iter);
|
|
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
/* Ok, now remove outdated ones. We have to do it after the lock
|
|
|
|
|
* because nm_ap_list_remove_ap() locks the list too.
|
|
|
|
|
*/
|
2005-02-10 04:39:40 +00:00
|
|
|
for (elt = outdated_list; elt; elt = g_slist_next (elt))
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-02-10 04:39:40 +00:00
|
|
|
if ((outdated_ap = (NMAccessPoint *)(elt->data)))
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-02-27 06:11:16 +00:00
|
|
|
nm_dbus_signal_wireless_network_change (dev->app_data->dbus_connection, dev, outdated_ap, NETWORK_STATUS_DISAPPEARED, -1);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
nm_ap_list_remove_ap (nm_device_ap_list_get (dev), outdated_ap);
|
|
|
|
|
list_changed = TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
g_slist_free (outdated_list);
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
nm_policy_schedule_device_change_check (dev->app_data);
|
|
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
return FALSE;
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
static gboolean nm_completion_scan_has_results (int tries, nm_completion_args args)
|
2005-03-26 03:42:05 +00:00
|
|
|
{
|
2005-10-11 20:13:28 +00:00
|
|
|
NMDevice * dev = args[0];
|
|
|
|
|
gboolean * err = args[1];
|
|
|
|
|
NMSock * sk = args[2];
|
|
|
|
|
NMWirelessScanResults * scan_results = args[3];
|
|
|
|
|
int rc;
|
|
|
|
|
guint8 we_ver;
|
2005-03-26 03:42:05 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
g_return_val_if_fail (err != NULL, TRUE);
|
|
|
|
|
g_return_val_if_fail (scan_results != NULL, TRUE);
|
|
|
|
|
|
2005-10-11 20:13:28 +00:00
|
|
|
we_ver = dev->options.wireless.we_version;
|
|
|
|
|
rc = iw_scan (nm_dev_sock_get_fd (sk), (char *)nm_device_get_iface (dev), we_ver, &(scan_results->scan_head));
|
2005-04-04 16:22:03 +00:00
|
|
|
*err = FALSE;
|
2005-03-26 03:42:05 +00:00
|
|
|
if (rc == -1 && errno == ETIME)
|
|
|
|
|
{
|
2005-04-04 16:22:03 +00:00
|
|
|
/* Scans take time. iw_scan's timeout is 15 seconds, so if the card hasn't returned
|
|
|
|
|
* scan results after two consecutive runs of iw_scan(), the card sucks.
|
|
|
|
|
*/
|
|
|
|
|
if (tries >= 1)
|
|
|
|
|
{
|
|
|
|
|
nm_warning ("Warning: the wireless card (%s) requires too much time for scans. Its driver needs to be fixed.", nm_device_get_iface (dev));
|
|
|
|
|
scan_results->scan_head.result = NULL;
|
|
|
|
|
*err = TRUE;
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Give the card one more chance to return scan results */
|
|
|
|
|
scan_results->scan_head.result = NULL;
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
2005-03-26 03:42:05 +00:00
|
|
|
}
|
|
|
|
|
if ((rc == -1 && errno == ENODATA) || (rc == 0 && scan_results->scan_head.result == NULL))
|
|
|
|
|
{
|
|
|
|
|
/* Card hasn't had time yet to compile full access point list.
|
|
|
|
|
* Give it some more time and scan again. If that doesn't
|
|
|
|
|
* work, we eventually give up. */
|
|
|
|
|
scan_results->scan_head.result = NULL;
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
else if (rc == -1)
|
|
|
|
|
{
|
|
|
|
|
scan_results->scan_head.result = NULL;
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
return TRUE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/*
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
* nm_device_wireless_scan
|
2005-01-07 18:07:06 +00:00
|
|
|
*
|
|
|
|
|
* Get a list of access points this device can see.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
static gboolean nm_device_wireless_scan (gpointer user_data)
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
2005-10-09 04:39:49 +00:00
|
|
|
NMWirelessScanCB * scan_cb = (NMWirelessScanCB *)(user_data);
|
|
|
|
|
NMDevice * dev = NULL;
|
|
|
|
|
NMWirelessScanResults * scan_results = NULL;
|
|
|
|
|
guint32 caps;
|
2004-08-11 Dan Williams <dcbw@redhat.com>
* info-daemon/NetworkManagerInfo.c:
- (main): clean up Seth's code style
* info-daemon/NetworkManagerInfoDbus.c:
- Use the more aptly-named path/service/interface constants from NetworkManager
- Don't return empty strings ("") as object paths ever, instead return errors
* panel-applet/NMWirelessApplet.c:
- Clean up Seth's code style
* src/NetworkManager.[ch]
- (nm_remove_device_from_list): remove anything having to do with pending_device
- (main, nm_print_usage): change --daemon=[yes|no] -> --no-daemon
* src/NetworkManagerAPList.[ch]
- Move Iter struct right above the iter functions to preserve opacity
- (nm_ap_list_remove_ap): implement
- (nm_ap_list_update_network): deal with errors returned from nm_dbus_get_network_priority(),
remove AP if NetworkManagerInfo doesn't know anything about it
- (nm_ap_list_diff): user NMAPList iterators
- (nm_ap_list_print_members): implement debugging function
* src/NetworkManagerDbus.[ch]
- (nm_dbus_nm_get_active_device): remove anything to do with pending_device
- (nm_dbus_get_user_key_for_network): remove DBusPendingCall stuff (unused),
and move the actual key setting stuff into NetworkManagerDevice.c
- (nm_dbus_get_network_priority): return -1 now on errors
- (nm_dbus_nmi_filter): fix strcmp() error that caused PreferredNetworkUpdate signals to
get lost, and force the active device to update its "best" ap when AP lists change
- (nm_dbus_nm_message_handler): Update conditions for returning "connecting" for a "status"
method call due to pending_device member removal
* src/NetworkManagerDevice.[ch]
- Move NMDevice structure to the top
- Add a wireless scan mutex and a best_ap mutex to the Wireless Options structure
- Remove Pending Action stuff from everywhere
- (nm_device_activation_*): We now "begin" activation and start a thread to do the
activation for us. This thread blocks until all conditions for activation have
been met (ie for wireless devices, we need a valid WEP key and a "best" ap), and
then setup up the interface and runs dhclient. We have to do this because there
is no guaruntee how long dhclient takes, and while we are blocking on it, we cannot
run our main loop and respond to dbus method calls or HAL device removals/inserts
- (nm_device_set_user_key_for_network): Move logic here from NetworkManagerDbus.c so we
can tell nm_device_activation_worker() that we've got a key
- (nm_device_*_best_ap): lock access to best_ap member of Wireless Options structure
- (nm_device_get_path_for_ap): dumb it down so the list doesn't lock against itself when
diffing (AP appear/disappear signal functions make sure the AP is actually in the device's
list)
- (nm_device_update_best_ap): move logic from nm_wireless_is_ap_better() here
* src/NetworkManagerPolicy.c
- Remove anything to do with pending_device
- Adjust device activation to deal with activation-in-worker-thread
* src/NetworkManagerUtils.c
- Clean up locking debugging a bit
* src/NetworkManagerWireless.[ch]
- (nm_wireless_is_ap_better): remove, stick logic in nm_device_update_best_ap(). This function
was badly named and is better as a device function
* panel-applet/.cvsignore: add
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@46 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-11 18:14:02 +00:00
|
|
|
|
2005-04-01 21:30:12 +00:00
|
|
|
g_return_val_if_fail (scan_cb != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
dev = scan_cb->dev;
|
|
|
|
|
if (!dev || !dev->app_data)
|
|
|
|
|
{
|
|
|
|
|
g_free (scan_cb);
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-10-09 04:39:49 +00:00
|
|
|
caps = nm_device_get_capabilities (dev);
|
|
|
|
|
if (!(caps & NM_DEVICE_CAP_NM_SUPPORTED) || !(caps & NM_DEVICE_CAP_WIRELESS_SCAN))
|
|
|
|
|
{
|
|
|
|
|
g_free (scan_cb);
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
2005-06-09 Dan Williams <dcbw@redhat.com>
* NetworkManager.h
- Add NMWirelessScanMethod enum for scan methods
* gnome/applet/applet-dbus-devices.c
- (nmwa_dbus_update_scanning_enabled_cb): remove
- (nmwa_dbus_update_scanning_enabled): remove
- (nmwa_dbus_update_devices): don't call nmwa_dbus_update_scanning_enabled() anymore
since it got removed
- (nmwa_dbus_enable_scanning): remove
* gnome/applet/applet-dbus-info.c
- (nmi_dbus_signal_update_scan_method): new function, signal NetworkManager to
update the wireless scanning method from NMI
- (nmi_dbus_get_wireless_scan_method): new function, return wireless scanning
method value to NetworkManager
- (nmi_dbus_info_message_handler): respond to the "getWirelessScanMethod" method call
* gnome/applet/applet-dbus-info.h
- Add prototype for nmi_dbus_signal_update_scan_method
* gnome/applet/applet.c
- (scanning_menu_update): new function, update one GtkCheckMenuItem from the
Wireless Scanning menu based on current wireless scan method
- (nmwa_menu_scanning_item_activate): new function, callback for GTK "activate"
signal for Wireless Scanning menu items, tell NetworkManager the new method
and update our menu items to make sure the right one is checked
- (nmwa_set_scanning_enabled_cb): remove
- (nmwa_context_menu_update): remove references to pause_scanning_item
- (nmwa_context_menu_create): remove pause_scanning_item, and add new Wireless
Scanning menu item
- (nmwa_gconf_get_wireless_scan_method): new method, pull wireless scanning method
from GConf
- nmwa_gconf_networks_notify_callback -> nmwa_gconf_info_notify_callback: generalize
so we get notified of preference values too
- (nmwa_get_instance): monitor GCONF_PATH_WIRELESS rather than GCONF_PATH_WIRELESS_NETWORKS
* gnome/applet/applet.h
- GCONF_PATH_WIRELESS added, one level below GCONF_PATH_WIRELESS_NETWORKS
- Add wireless scan method member to applet data
- Remove pause_scanning_item, add Wireless Scanning submenu
* src/NetworkManager.c
- (nm_data_new): default to NM_SCAN_METHOD_ON
- (main): grab scanning method from NMI if we can
* src/NetworkManagerDbus.c
- (nm_dbus_update_wireless_scan_method_cb): new function, callback from
nm_dbus_update_wireless_scan_method()
- (nm_dbus_update_wireless_scan_method): new function to grab scanning method
from NMI
- (nm_dbus_nmi_is_running): redundant function, removed
- (nm_dbus_signal_filter): trap "WirelessScanMethodUpdate" signal, grab scanning method
when NMI comes back
* src/NetworkManagerDevice.c
- (nm_device_is_activated): return TRUE if the device is activated
- (nm_device_wireless_scan): don't scan if the scan method is OFF, or if its AUTO
and we are activated
* src/nm-dbus-nm.c
- (nm_dbus_nm_set_scanning_enabled): removed
- nm_dbus_nm_get_scanning_enabled -> nm_dbus_nm_get_wireless_scan_method
- (nm_dbus_nm_methods_setup): remove [get | set] ScanningEnabled and add "getWirelessScanMethod"
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@658 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-06-10 03:13:27 +00:00
|
|
|
/* Reschedule ourselves if all wireless is disabled, we're asleep,
|
|
|
|
|
* or we are currently activating.
|
|
|
|
|
*/
|
|
|
|
|
if ( (dev->app_data->wireless_enabled == FALSE)
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|| (dev->app_data->asleep == TRUE)
|
|
|
|
|
|| (nm_device_is_activating (dev) == TRUE))
|
2005-02-07 23:04:05 +00:00
|
|
|
{
|
2005-09-11 01:46:40 +00:00
|
|
|
nm_wireless_set_scan_interval (dev->app_data, dev, NM_WIRELESS_SCAN_INTERVAL_INIT);
|
|
|
|
|
goto reschedule;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* A/B/G cards should only scan if they are disconnected. Set the timeout to active
|
|
|
|
|
* for the case we lose this connection shortly, it will reach this point and then
|
|
|
|
|
* nm_device_is_activated will return FALSE, letting the scan proceed.
|
|
|
|
|
*/
|
|
|
|
|
if (dev->options.wireless.num_freqs > 14 && nm_device_is_activated (dev) == TRUE)
|
|
|
|
|
{
|
|
|
|
|
nm_wireless_set_scan_interval (dev->app_data, dev, NM_WIRELESS_SCAN_INTERVAL_ACTIVE);
|
2005-02-07 23:04:05 +00:00
|
|
|
goto reschedule;
|
|
|
|
|
}
|
|
|
|
|
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
/* Test devices get a fake ap list */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
{
|
|
|
|
|
nm_device_fake_ap_list (dev);
|
2005-09-11 01:46:40 +00:00
|
|
|
nm_wireless_set_scan_interval (dev->app_data, dev, NM_WIRELESS_SCAN_INTERVAL_ACTIVE);
|
2005-03-31 Dan Williams <dcbw@redhat.com>
Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-31 21:02:21 +00:00
|
|
|
goto reschedule;
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Grab the scan mutex */
|
|
|
|
|
if (nm_try_acquire_mutex (dev->options.wireless.scan_mutex, __FUNCTION__))
|
|
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
|
|
|
|
gboolean devup_err;
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Device must be up before we can scan */
|
2005-03-26 03:42:05 +00:00
|
|
|
devup_err = nm_device_bring_up_wait(dev, 1);
|
|
|
|
|
if (devup_err)
|
|
|
|
|
{
|
|
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2005-04-01 21:30:12 +00:00
|
|
|
goto reschedule;
|
2005-03-26 03:42:05 +00:00
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_WIRELESS, __FUNCTION__, NULL)))
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
2005-03-26 03:42:05 +00:00
|
|
|
int err;
|
|
|
|
|
NMNetworkMode orig_mode = NETWORK_MODE_INFRA;
|
|
|
|
|
double orig_freq = 0;
|
|
|
|
|
int orig_rate = 0;
|
2005-08-22 22:37:38 +00:00
|
|
|
const int interval = 20;
|
|
|
|
|
const int assoc_pause = nm_device_get_association_pause_value (dev);
|
2005-08-31 13:35:04 +00:00
|
|
|
const int delay = G_USEC_PER_SEC / interval;
|
2005-08-22 22:37:38 +00:00
|
|
|
const int max_tries = assoc_pause * interval;
|
2005-04-28 02:50:54 +00:00
|
|
|
nm_completion_args args;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
orig_mode = nm_device_get_mode (dev);
|
2005-01-10 06:41:02 +00:00
|
|
|
if (orig_mode == NETWORK_MODE_ADHOC)
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-01-10 06:41:02 +00:00
|
|
|
orig_freq = nm_device_get_frequency (dev);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
orig_rate = nm_device_get_bitrate (dev);
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
/* Must be in infrastructure mode during scan, otherwise we don't get a full
|
|
|
|
|
* list of scan results. Scanning doesn't work well in Ad-Hoc mode :(
|
|
|
|
|
*/
|
|
|
|
|
nm_device_set_mode (dev, NETWORK_MODE_INFRA);
|
2005-01-10 06:41:02 +00:00
|
|
|
nm_device_set_frequency (dev, 0);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
scan_results = g_malloc0 (sizeof (NMWirelessScanResults));
|
2005-04-28 02:50:54 +00:00
|
|
|
|
|
|
|
|
args[0] = dev;
|
|
|
|
|
args[1] = &err;
|
|
|
|
|
args[2] = sk;
|
|
|
|
|
args[3] = scan_results;
|
2005-08-22 22:37:38 +00:00
|
|
|
nm_wait_for_completion (max_tries, delay, nm_completion_scan_has_results, NULL, args);
|
2005-03-26 03:42:05 +00:00
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
nm_device_set_mode (dev, orig_mode);
|
2005-01-10 06:41:02 +00:00
|
|
|
/* Only set frequency if ad-hoc mode */
|
|
|
|
|
if (orig_mode == NETWORK_MODE_ADHOC)
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
{
|
2005-01-10 06:41:02 +00:00
|
|
|
nm_device_set_frequency (dev, orig_freq);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
nm_device_set_bitrate (dev, orig_rate);
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2005-03-31 16:13:06 +00:00
|
|
|
|
|
|
|
|
if (!scan_results->scan_head.result)
|
|
|
|
|
{
|
|
|
|
|
g_free (scan_results);
|
|
|
|
|
scan_results = NULL;
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
|
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
|
|
|
|
}
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-05-03 Dan Williams <dcbw@redhat.com>
* Kill dhcpcd. We now use "dhcdbd", a dbus daemon that controls dhclient.
This means that NetworkManager shouldn't have DHCP issues anymore. It also
means you need dhcdbd, which you can get here (get the latest one):
http://people.redhat.com/jvdias/dhcdbd/
Technically NetworkManager can use any DHCP daemon that uses the same DBUS
interface as dhcdbd.
* Rewrite device activation to facilitate the new DHCP infrastructure and
future improvements. Its now "activation request" based, ie there is a single
activation request composed of the device, access point, and other info which
follows the entire activation process. There are 5 stages of the activation
process which correspond to:
1) Device preparation
2) Device configuration (bring it up, set ESSID/Key/etc)
3) IP Config Start (fire off DHCP if we're using it)
4) IP Config Get (grab config from DHCP or static config files)
5) IP Config Commit (set device's IP address, DNS, etc)
Note that there is no longer a "scanning" step, since the access point must
be known _before_ activation starts. If the access point drops out or does
not exist for some reason, the entire activation process fails and must be
restarted for a different access point or device.
Patch from Bill Moss:
* gnome/applet/applet.c
- Fix type of vpn_failure dialog -> vpn_banner dialog
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@597 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-05-03 20:41:36 +00:00
|
|
|
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
/* We run the scan processing function from the main thread, since it must deliver
|
|
|
|
|
* messages over DBUS. Plus, that way the main thread is the only thread that has
|
|
|
|
|
* to modify the device's access point list.
|
|
|
|
|
*/
|
|
|
|
|
if ((scan_results != NULL) && (scan_results->scan_head.result != NULL))
|
|
|
|
|
{
|
|
|
|
|
guint scan_process_source_id = 0;
|
|
|
|
|
GSource *scan_process_source = g_idle_source_new ();
|
2005-04-01 21:30:12 +00:00
|
|
|
GTimeVal cur_time;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
|
|
|
|
scan_results->dev = dev;
|
|
|
|
|
g_source_set_callback (scan_process_source, nm_device_wireless_process_scan_results, scan_results, NULL);
|
|
|
|
|
scan_process_source_id = g_source_attach (scan_process_source, dev->app_data->main_context);
|
|
|
|
|
g_source_unref (scan_process_source);
|
2005-04-01 21:30:12 +00:00
|
|
|
|
|
|
|
|
g_get_current_time (&cur_time);
|
|
|
|
|
dev->options.wireless.last_scan = cur_time.tv_sec;
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
}
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
reschedule:
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
/* Make sure we reschedule ourselves so we keep scanning */
|
2005-09-02 03:47:19 +00:00
|
|
|
nm_device_wireless_schedule_scan (dev);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
|
2005-04-01 21:30:12 +00:00
|
|
|
g_free (scan_cb);
|
2005-01-18 Dan Williams <dcbw@redhat.com>
* dhcpcd/client.c
- Remove some debug messages
- Wrap others in #ifdef DEBUG/#endif
* src/NetworkManager.c
- Remove some debug messages
- Clarify some debug messages
- Remove code related to old single-thread wireless scanning
* src/NetworkManagerAP.[ch]
- New AP property "last_seen" to track how recently an AP was
found in a scan
- Start using 'const' more in function arguments
* src/NetworkManagerAPList.[ch]
- (nm_ap_list_merge_scanned_ap): new, selectively update attributes
of an AP found in an AP list from a source AP, or if not found
in the list add the source AP
- (nm_ap_list_combine): remove, no longer needed
* src/NetworkManagerDevice.c
- Each device now has a "worker" thread from start to end of its life.
Scanning for wireless devices now happens in that thread,
not in a single "wireless scanning thread" for all devices as
previously. Activation consists of adding an idle handler to the
thread's main loop/context, which gets run at the next available
opportunity.
- Wireless scanning is also simplified, there is now only one list of
access points per wireless device, and APs older than 60s are
removed from the list. Previously, we kept results for the last
3 scans and merged whole lists, which was complicated.
- Cleaned up activation debug messages.
- Wireless activation and access-point search routines now use Open System
authentication before trying Shared Key.
- Removed some code in nm_device_update_best_ap() that could cause cards
to loose their link to the access point.
- Scanning now uses a backoff algorithm, where the inverval becomes
progressively longer between scans when the list of scanned access
points doesn't change. A change will revert to the shortest scan
interval (20s).
* src/NetworkManagerWireless.[ch]
- Remove code related to old single-thread wireless scanning
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@382 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-01-18 22:08:22 +00:00
|
|
|
return FALSE;
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
/* IP Configuration stuff */
|
2004-09-08 18:14:42 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
gboolean nm_device_get_use_dhcp (NMDevice *dev)
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2005-04-15 15:43:42 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-09-08 18:14:42 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
return dev->use_dhcp;
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
void nm_device_set_use_dhcp (NMDevice *dev, gboolean use_dhcp)
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
dev->use_dhcp = use_dhcp;
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
NMIP4Config *nm_device_get_ip4_config (NMDevice *dev)
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2005-04-15 15:43:42 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
2004-09-08 18:14:42 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
return dev->ip4_config;
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
void nm_device_set_ip4_config (NMDevice *dev, NMIP4Config *config)
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2005-04-15 15:43:42 +00:00
|
|
|
NMIP4Config *old_config;
|
2004-09-08 18:14:42 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
old_config = dev->ip4_config;
|
|
|
|
|
if (config)
|
|
|
|
|
nm_ip4_config_ref (config);
|
|
|
|
|
dev->ip4_config = config;
|
|
|
|
|
if (old_config)
|
|
|
|
|
nm_ip4_config_unref (old_config);
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
2004-10-15 00:06:14 +00:00
|
|
|
|
2005-09-11 01:46:40 +00:00
|
|
|
void nm_device_set_wireless_scan_interval (NMDevice *dev, NMWirelessScanInterval interval)
|
|
|
|
|
{
|
|
|
|
|
guint seconds;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
switch (interval)
|
|
|
|
|
{
|
|
|
|
|
case NM_WIRELESS_SCAN_INTERVAL_INIT:
|
|
|
|
|
seconds = 10;
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
case NM_WIRELESS_SCAN_INTERVAL_INACTIVE:
|
|
|
|
|
seconds = 120;
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
case NM_WIRELESS_SCAN_INTERVAL_ACTIVE:
|
|
|
|
|
default:
|
|
|
|
|
seconds = 20;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
dev->options.wireless.scan_interval = seconds;
|
|
|
|
|
}
|
|
|
|
|
|
2004-10-15 00:06:14 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_get_system_config_data
|
|
|
|
|
*
|
|
|
|
|
* Return distro-specific system configuration data for this device.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void *nm_device_get_system_config_data (NMDevice *dev)
|
2004-10-15 00:06:14 +00:00
|
|
|
{
|
2005-04-15 15:43:42 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
2004-10-15 00:06:14 +00:00
|
|
|
|
2005-04-15 15:43:42 +00:00
|
|
|
return dev->system_config_data;
|
2004-10-15 00:06:14 +00:00
|
|
|
}
|
2004-09-08 18:14:42 +00:00
|
|
|
|
|
|
|
|
|
2005-02-10 20:59:10 +00:00
|
|
|
/* Define types for stupid headers */
|
|
|
|
|
typedef u_int8_t u8;
|
|
|
|
|
typedef u_int16_t u16;
|
|
|
|
|
typedef u_int32_t u32;
|
2005-02-11 21:44:35 +00:00
|
|
|
typedef u_int64_t u64;
|
2005-02-10 20:59:10 +00:00
|
|
|
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/**************************************/
|
|
|
|
|
/* Ethtool capability detection */
|
|
|
|
|
/**************************************/
|
|
|
|
|
#include <linux/sockios.h>
|
|
|
|
|
#include <linux/ethtool.h>
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
static gboolean supports_ethtool_carrier_detect (NMDevice *dev)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-04-06 16:45:48 +00:00
|
|
|
NMSock *sk;
|
2005-02-07 23:04:05 +00:00
|
|
|
struct ifreq ifr;
|
|
|
|
|
gboolean supports_ethtool = FALSE;
|
|
|
|
|
struct ethtool_cmd edata;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_GENERAL, __FUNCTION__, NULL)) == NULL)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-04-06 19:16:31 +00:00
|
|
|
nm_warning ("cannot open socket on interface %s for ethtool detect; errno=%d", nm_device_get_iface (dev), errno);
|
2005-02-07 23:04:05 +00:00
|
|
|
return (FALSE);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
2005-02-07 23:04:05 +00:00
|
|
|
|
|
|
|
|
strncpy (ifr.ifr_name, nm_device_get_iface (dev), sizeof(ifr.ifr_name)-1);
|
|
|
|
|
edata.cmd = ETHTOOL_GLINK;
|
|
|
|
|
ifr.ifr_data = (char *) &edata;
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: About to ETHTOOL\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
if (ioctl (nm_dev_sock_get_fd (sk), SIOCETHTOOL, &ifr) == -1)
|
2005-02-07 23:04:05 +00:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
supports_ethtool = TRUE;
|
|
|
|
|
|
|
|
|
|
out:
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with ETHTOOL\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2005-02-07 23:04:05 +00:00
|
|
|
return (supports_ethtool);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
/**************************************/
|
|
|
|
|
/* MII capability detection */
|
|
|
|
|
/**************************************/
|
|
|
|
|
#include <linux/mii.h>
|
|
|
|
|
|
2005-06-19 19:07:47 +00:00
|
|
|
static int mdio_read (NMDevice *dev, NMSock *sk, struct ifreq *ifr, int location)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-02-07 23:04:05 +00:00
|
|
|
struct mii_ioctl_data *mii;
|
2005-06-19 19:07:47 +00:00
|
|
|
int val = -1;
|
2005-02-07 23:04:05 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
g_return_val_if_fail (sk != NULL, -1);
|
2005-02-07 23:04:05 +00:00
|
|
|
g_return_val_if_fail (ifr != NULL, -1);
|
|
|
|
|
|
|
|
|
|
mii = (struct mii_ioctl_data *) &(ifr->ifr_data);
|
|
|
|
|
mii->reg_num = location;
|
|
|
|
|
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: About to GET MIIREG\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
|
|
|
|
if (ioctl (nm_dev_sock_get_fd (sk), SIOCGMIIREG, ifr) >= 0)
|
|
|
|
|
val = mii->val_out;
|
|
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with GET MIIREG\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
2005-02-07 23:04:05 +00:00
|
|
|
|
2005-06-19 19:07:47 +00:00
|
|
|
return val;
|
2005-02-07 23:04:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static gboolean supports_mii_carrier_detect (NMDevice *dev)
|
|
|
|
|
{
|
2005-06-19 19:07:47 +00:00
|
|
|
NMSock * sk;
|
2004-06-24 14:18:37 +00:00
|
|
|
struct ifreq ifr;
|
2005-02-07 23:04:05 +00:00
|
|
|
int bmsr;
|
|
|
|
|
gboolean supports_mii = FALSE;
|
2005-06-19 19:07:47 +00:00
|
|
|
int err;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-04-06 16:45:48 +00:00
|
|
|
if ((sk = nm_dev_sock_open (dev, DEV_GENERAL, __FUNCTION__, NULL)) == NULL)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-03-14 Ray Strode <rstrode@redhat.com>
Fourth (probably working) cut at porting to
dbus 0.30 api and new hal. This cut adds
some new logging macros to make debugging
easier.
* dispatcher-daemon/NetworkManagerDispatcher.c:
* info-daemon/NetworkmanagerInfo.c:
* info-daemon/NetworkManagerInfoPassphraseDialog.c:
* info-daemon/NetworkManagerInfoVPN.c:
* src/NetworkManager.c:
* src/NetworkManagerAP.c:
* src/NetworkManagerAPList.c:
* src/NetworkManagerDHCP.c:
* src/NetworkManagerDbus.c:
* src/NetworkManagerDevice.c:
* src/NetworkManagerPolicy.c:
* src/NetworkManagerSystem.c:
* src/NetworkManagerUtils.c:
* src/NetworkManagerWireless.c:
* src/autoip.c:
* src/nm-dbus-nm.c:
* src/backends/NetworkManagerDebian.c:
* src/backends/NetworkManagerGentoo.c:
* src/backends/NetworkManagerRedHat.c:
* src/backends/NetworkManagerSlackware.c:
use new logging macros.
* dispatcher-daemon/NetworkManagerDispatcher.c:
(nmd_dbus_filter): s/dbus_free/g_free/
* info-daemon/Makefile.am: link in utils library.
* info-daemon/NetworkmanagerInfo.c: use new logging
macros.
(nmi_dbus_get_network): don't assume enumerations
are 32-bit.
(nmi_dbus_nmi_message_handler): don't free what
doesn't belong to us.
* libnm_glib/libnm_glib.c:
(libnm_glib_get_nm_status):
(libnm_glib_init): don't free what doesn't
belong to us.
(libnm_glib_dbus): strdup result, so it doesn't get
lost when message is unref'd.
* panel-applet/NMWirelessAppletDbus.c:
(nmwa_dbus_update_devices): s/dbus_free/g_free/
* src/NetworkManager.c:
(nm_monitor_wired_link_state): request initial status
dump of all cards when we start up, instead of relying
on /sys/.../carrier.
(nm_info_handler), (nm_set_up_log_handlers):
log handlers to specify what syslog priorites
the logging macros default to.
* src/NetworkManagerAPList.c:
(nm_ap_list_populate_from_nmi):
s/dbus_free_string_array/g_strfreev/
* src/NetworkManagerDbus.c:
(nm_dbus_get_network_object):
validate d-bus message argument types.
Advance message iterator after reading argument,
prepend instead of append to GSList.
* src/NetworkManagerDevice.c:
(nm_device_probe_wired_link_status):
remove redundant /sys in /sys path. remove wrong
contents == NULL means has carrier assumption.
* src/nm-netlink-monitor.c
(nm_netlink_monitor_request_status): implement
function to ask kernel to dump interface link
status over netlink socket.
* test/*.c: s/dbus_free/g_free/
* utils/nm-utils.h:
(nm_print_backtrace): new macro to print backtrace.
(nm_get_timestamp): new macro to get sub-second precise
unix timestamp.
(nm_info), (nm_debug), (nm_warning), (nm_error):
new logging functions. nm_info just prints,
nm_debug includes timestamp and function,
nm_warning includes function, nm_error includes
backtrace and sigtrap.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@497 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2005-03-15 05:30:15 +00:00
|
|
|
nm_warning ("cannot open socket on interface %s for MII detect; errno=%d", nm_device_get_iface (dev), errno);
|
2004-06-24 14:18:37 +00:00
|
|
|
return (FALSE);
|
|
|
|
|
}
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
strncpy (ifr.ifr_name, nm_device_get_iface (dev), sizeof(ifr.ifr_name)-1);
|
2005-06-19 19:07:47 +00:00
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: About to GET MIIPHY\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
|
|
|
|
err = ioctl (nm_dev_sock_get_fd (sk), SIOCGMIIPHY, &ifr);
|
|
|
|
|
#ifdef IOCTL_DEBUG
|
|
|
|
|
nm_info ("%s: Done with GET MIIPHY\n", nm_device_get_iface (dev));
|
|
|
|
|
#endif
|
|
|
|
|
if (err < 0)
|
2005-02-07 23:04:05 +00:00
|
|
|
goto out;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/* If we can read the BMSR register, we assume that the card supports MII link detection */
|
2005-06-19 19:07:47 +00:00
|
|
|
bmsr = mdio_read (dev, sk, &ifr, MII_BMSR);
|
2005-02-07 23:04:05 +00:00
|
|
|
supports_mii = (bmsr != -1) ? TRUE : FALSE;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
out:
|
2005-04-06 16:45:48 +00:00
|
|
|
nm_dev_sock_close (sk);
|
2005-06-19 19:07:47 +00:00
|
|
|
return supports_mii;
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/****************************************/
|
|
|
|
|
/* End Code ripped from HAL */
|
|
|
|
|
/****************************************/
|
2004-08-31 Dan Williams <dcbw@redhat.com>
* Remove 'debug' extern global from all files since we now
use syslog()
* src/NetworkManager.[ch]
- Break out routine that get the net.interface property from HAL,
removing that logic from nm_create_device_and_add_to_list()
- (nm_create_device_and_add_to_list): make this a bit more general so
it doesn't do the talking to HAL. Also add arguments to facilitate
the create of test devices.
- (nm_data_mark_state_changed): rename from nm_data_set_state_modified()
- (nm_data_new, main, nm_print_usage): add new argument "--enable-test-devices"
which makes NetworkManager listen for dbus commands to create test
devices, which have no backing hardware. Use when you're on a plane
for example, and/or forgot your wireless card at home. Test devices
_cannot_ be created unless NM is started with --enable-test-devices.
* src/NetworkManagerDbus.[ch]
- New "getLinkActive" method for devices
- New "setLinkActive" method for devices (only works on test devices)
- New "createTestDevice" method on NetworkManager object to create a test
device of a specified type (ie wired, wireless). UDI is created from
scratch, as is the interface name. Only works when NM is started with
--enable-test-devices switch.
- New "removeTestDevice" method on NetworkManager object which removes a
test device. Only works when NM is started with --enable-test-devices
* src/NetworkManagerDevice.[ch]
- Logic to facilitate test devices. Add variables to NMDevice struct to indicate
whether a device is a test device or not, and what its link status is.
- Deal with test devices in most functions. For those that work directly on hardware
special-case test devices.
- (nm_device_new): don't create a test device if test devices weren't enabled on the
command-line.
- (nm_device_update_link_active): split out logic for wired and wireless device link
checking to separate functions to facilitate test device link checking.
- (nm_device_set_enc_key): Since some drivers for wireless cards are daft and
don't make a distinction between System Authentication and Encryption
(namely Cisco aironet), we use Open System auth when setting a WEP key
on the card. We don't deal with Shared Key auth yet.
- (nm_device_activation_worker): split the activation cancel check logic out into
a separate routine nm_device_activation_cancel_if_needed()
- (nm_device_activation_signal_cancel): rename from nm_device_activation_cancel()
- (nm_device_fake_ap_list): Test wireless devices obviously cannot scan, so create
a list of fake access points that they can "see"
- (nm_device_is_test_device): return whether or not a device is a test device
* src/NetworkManagerPolicy.c
- (nm_policy_get_best_device): attempt to deal with wireless network selection,
previously if you "locked"/forced NM to use a wireless device but then
selected a wireless network for NM to use, it would switch to a wired device.
So, if the active device is wireless and it has a "forced" best AP, use it
if the "forced" best AP is still valid
- (nm_state_modification_monitor): deal with NULL best devices, for example
there were no usable network devices, or the last one was removed
* src/backends/NetworkManager*.c
- Deal with test devices, mostly just return success for operations like getting
a DHCP address
* test/nmtestdevices.c
- Test tool to create/remove/link-switch test devices
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@112 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-08-31 16:09:15 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
/****************************************/
|
|
|
|
|
/* Test device routes */
|
|
|
|
|
/****************************************/
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_is_test_device
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_is_test_device (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->test_device);
|
|
|
|
|
}
|
2005-04-15 15:43:42 +00:00
|
|
|
|