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.
|
|
|
|
|
*
|
|
|
|
|
* (C) Copyright 2004 Red Hat, Inc.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#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>
|
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"
|
|
|
|
|
#include "NetworkManagerDHCP.h"
|
2004-06-24 14:18:37 +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
|
|
|
/* 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_activate (gpointer user_data);
|
2004-12-17 17:16:22 +00:00
|
|
|
static gboolean nm_device_activation_configure_ip (NMDevice *dev, gboolean do_only_autoip);
|
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);
|
2005-02-07 23:04:05 +00:00
|
|
|
static gboolean supports_mii_carrier_detect (NMDevice *dev);
|
|
|
|
|
static gboolean supports_ethtool_carrier_detect (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
|
|
|
|
|
|
|
|
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
|
|
|
|
|
|
|
|
|
|
|
|
|
/******************************************************/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/******************************************************/
|
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
|
|
|
{
|
2004-10-28 19:49:55 +00:00
|
|
|
int sk;
|
2004-10-29 16:37:43 +00:00
|
|
|
int err;
|
2004-10-28 19:49:55 +00:00
|
|
|
char ioctl_buf[64];
|
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;
|
|
|
|
|
strncpy(ioctl_buf, nm_device_get_iface(dev), 63);
|
|
|
|
|
|
|
|
|
|
sk = iw_sockets_open ();
|
2004-10-29 16:37:43 +00:00
|
|
|
err = ioctl(sk, SIOCGIWNAME, ioctl_buf);
|
2004-10-28 19:49:55 +00:00
|
|
|
close (sk);
|
2004-10-29 16:37:43 +00:00
|
|
|
return (err == 0);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_supports_wireless_scan
|
|
|
|
|
*
|
|
|
|
|
* Test whether a given device is a wireless one or not.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_supports_wireless_scan (NMDevice *dev)
|
|
|
|
|
{
|
2004-10-29 16:37:43 +00:00
|
|
|
int sk;
|
|
|
|
|
int err;
|
|
|
|
|
gboolean can_scan = TRUE;
|
|
|
|
|
wireless_scan_head scan_data;
|
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
|
|
|
g_return_val_if_fail (dev->type == DEVICE_TYPE_WIRELESS_ETHERNET, FALSE);
|
|
|
|
|
|
|
|
|
|
/* A test wireless device can always scan (we generate fake scan data for it) */
|
|
|
|
|
if (dev->test_device)
|
|
|
|
|
return (TRUE);
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2004-10-28 19:49:55 +00:00
|
|
|
sk = iw_sockets_open ();
|
2004-10-29 16:37:43 +00:00
|
|
|
err = iw_scan (sk, (char *)nm_device_get_iface (dev), WIRELESS_EXT, &scan_data);
|
2004-06-24 14:18:37 +00:00
|
|
|
nm_dispose_scan_results (scan_data.result);
|
2004-10-29 16:37:43 +00:00
|
|
|
if ((err == -1) && (errno == EOPNOTSUPP))
|
2004-06-24 14:18:37 +00:00
|
|
|
can_scan = FALSE;
|
2004-10-28 19:49:55 +00:00
|
|
|
close (sk);
|
2004-06-24 14:18:37 +00:00
|
|
|
return (can_scan);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* 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
|
|
|
{
|
2005-02-10 04:39:40 +00:00
|
|
|
dev = (NMDevice *)(elt->data);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (dev)
|
|
|
|
|
{
|
|
|
|
|
if (nm_null_safe_strcmp (nm_device_get_udi (dev), udi) == 0)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (dev);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* 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
|
|
|
/*****************************************************************************/
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* 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;
|
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
|
|
|
{
|
|
|
|
|
syslog (LOG_ERR, "nm_device_new(): attempt to create a test device, but test devices were not enabled"
|
|
|
|
|
" on the command line. Will not create the device.\n");
|
|
|
|
|
return (NULL);
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
dev = g_malloc0 (sizeof (NMDevice));
|
2004-06-24 14:18:37 +00:00
|
|
|
if (!dev)
|
|
|
|
|
{
|
2004-08-23 19:20:49 +00:00
|
|
|
syslog (LOG_ERR, "nm_device_new() could not allocate a new device... Not enough memory?");
|
2004-06-24 14:18:37 +00:00
|
|
|
return (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
|
|
|
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;
|
2004-10-21 17:42:14 +00:00
|
|
|
nm_device_set_udi (dev, udi);
|
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 */
|
|
|
|
|
nm_device_bring_up (dev);
|
2005-02-07 23:04:05 +00:00
|
|
|
g_usleep (G_USEC_PER_SEC);
|
|
|
|
|
dev->driver_support_level = nm_get_driver_support_level (dev->app_data->hal_ctx, dev);
|
2004-12-16 04:24:20 +00:00
|
|
|
|
2004-09-08 18:14:42 +00:00
|
|
|
/* Initialize wireless-specific options */
|
2004-08-05 18:54:29 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-02-07 23:04:05 +00:00
|
|
|
int sk;
|
|
|
|
|
NMDeviceWirelessOptions *opts = &(dev->options.wireless);
|
2004-12-16 04:24:20 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
nm_device_set_mode (dev, NETWORK_MODE_INFRA);
|
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
|
|
|
opts->scan_interval = 20;
|
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
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
opts->scan_mutex = g_mutex_new ();
|
|
|
|
|
opts->best_ap_mutex = g_mutex_new ();
|
|
|
|
|
opts->ap_list = nm_ap_list_new (NETWORK_TYPE_DEVICE);
|
|
|
|
|
if (!opts->scan_mutex || !opts->best_ap_mutex || !opts->ap_list)
|
|
|
|
|
goto err;
|
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
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
nm_register_mutex_desc (opts->scan_mutex, "Scan Mutex");
|
|
|
|
|
nm_register_mutex_desc (opts->best_ap_mutex, "Best AP Mutex");
|
2004-09-12 03:06:44 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
opts->supports_wireless_scan = nm_device_supports_wireless_scan (dev);
|
2004-12-16 04:24:20 +00:00
|
|
|
|
|
|
|
|
if ((sk = iw_sockets_open ()) >= 0)
|
|
|
|
|
{
|
2005-01-31 19:53:38 +00:00
|
|
|
iwrange range;
|
|
|
|
|
if (iw_get_range_info (sk, nm_device_get_iface (dev), &range) >= 0)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
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;
|
2005-01-31 19:53:38 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
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;
|
2005-01-31 19:53:38 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
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]));
|
2005-01-31 19:53:38 +00:00
|
|
|
}
|
2004-12-16 04:24:20 +00:00
|
|
|
close (sk);
|
|
|
|
|
}
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
2005-02-07 23:04:05 +00:00
|
|
|
else if (nm_device_is_wired (dev))
|
|
|
|
|
{
|
|
|
|
|
if (supports_ethtool_carrier_detect (dev) || supports_mii_carrier_detect (dev))
|
|
|
|
|
dev->options.wired.has_carrier_detect = TRUE;
|
|
|
|
|
}
|
2004-10-21 17:42:14 +00:00
|
|
|
|
2004-10-21 19:00:13 +00:00
|
|
|
if (nm_device_get_driver_support_level (dev) != NM_DRIVER_UNSUPPORTED)
|
|
|
|
|
{
|
2005-02-07 23:04:05 +00:00
|
|
|
nm_device_update_link_active (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 */
|
|
|
|
|
nm_system_device_update_config_info (dev);
|
2004-10-21 19:00:13 +00:00
|
|
|
}
|
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
|
|
|
if (!g_thread_create (nm_device_worker, dev, FALSE, &error))
|
|
|
|
|
{
|
|
|
|
|
syslog (LOG_CRIT, "nm_device_new (): could not create device worker thread. (glib said: '%s')", error->message);
|
|
|
|
|
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. */
|
|
|
|
|
syslog (LOG_ERR, "nm_device_new(): waiting for device's worker thread to start.\n");
|
|
|
|
|
while (dev->worker_started == FALSE)
|
|
|
|
|
g_usleep (G_USEC_PER_SEC / 2);
|
|
|
|
|
syslog (LOG_ERR, "nm_device_new(): device's worker thread started, continuing.\n");
|
|
|
|
|
|
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
|
|
|
|
|
|
|
|
dev->refcount--;
|
2004-08-05 18:54:29 +00:00
|
|
|
if (dev->refcount <= 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
if (dev->loop)
|
|
|
|
|
g_main_loop_quit (dev->loop);
|
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 (dev->worker_done == FALSE)
|
|
|
|
|
g_usleep (300);
|
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);
|
2005-01-18 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 (dev->options.wireless.best_ap)
|
|
|
|
|
nm_ap_unref (dev->options.wireless.best_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
|
|
|
g_mutex_free (dev->options.wireless.best_ap_mutex);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
2005-02-10 04:39:40 +00:00
|
|
|
/* Get rid of DHCP state data */
|
|
|
|
|
if (dev->dhcp_iface)
|
|
|
|
|
{
|
|
|
|
|
dhcp_interface_free (dev->dhcp_iface);
|
|
|
|
|
dev->dhcp_iface = NULL;
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
g_free (dev->udi);
|
|
|
|
|
g_free (dev->iface);
|
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
|
|
|
}
|
2005-01-07 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)
|
|
|
|
|
{
|
|
|
|
|
syslog (LOG_CRIT, "nm_device_worker(): received NULL device object, NetworkManager cannot continue.\n");
|
|
|
|
|
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
|
|
|
|
2005-01-24 21:53:39 +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
|
|
|
/* Do an initial wireless scan */
|
|
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
GSource *source = g_idle_source_new ();
|
|
|
|
|
guint source_id = 0;
|
|
|
|
|
|
|
|
|
|
g_source_set_callback (source, nm_device_wireless_scan, dev, NULL);
|
|
|
|
|
source_id = g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
g_main_loop_run (dev->loop);
|
|
|
|
|
|
2005-02-10 04:39:40 +00:00
|
|
|
/* Remove any DHCP timeouts that might have been running */
|
2005-02-22 00:30:46 +00:00
|
|
|
if (nm_device_config_get_use_dhcp (dev))
|
|
|
|
|
nm_device_dhcp_remove_timeouts (dev);
|
2005-02-10 04:39:40 +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
|
|
|
g_main_loop_unref (dev->loop);
|
|
|
|
|
g_main_context_unref (dev->context);
|
|
|
|
|
|
|
|
|
|
dev->loop = NULL;
|
|
|
|
|
dev->context = NULL;
|
|
|
|
|
|
|
|
|
|
dev->worker_done = TRUE;
|
|
|
|
|
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);
|
|
|
|
|
|
|
|
|
|
g_main_loop_quit (dev->loop);
|
|
|
|
|
while (dev->worker_done == FALSE)
|
|
|
|
|
g_usleep (G_USEC_PER_SEC / 2);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
/*
|
|
|
|
|
* nm_device_open_sock
|
|
|
|
|
*
|
|
|
|
|
* Get a control socket for network operations.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
int nm_device_open_sock (void)
|
|
|
|
|
{
|
|
|
|
|
int fd;
|
|
|
|
|
|
|
|
|
|
/* Try to grab a control socket */
|
|
|
|
|
fd = socket(PF_INET, SOCK_DGRAM, 0);
|
|
|
|
|
if (fd >= 0)
|
|
|
|
|
return (fd);
|
|
|
|
|
fd = socket(PF_PACKET, SOCK_DGRAM, 0);
|
|
|
|
|
if (fd >= 0)
|
|
|
|
|
return (fd);
|
|
|
|
|
fd = socket(PF_INET6, SOCK_DGRAM, 0);
|
|
|
|
|
if (fd >= 0)
|
|
|
|
|
return (fd);
|
|
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
syslog (LOG_ERR, "nm_device_open_sock () could not get network control socket.");
|
2004-11-16 02:41:53 +00:00
|
|
|
return (-1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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.
|
|
|
|
|
*/
|
2005-01-13 19:04:26 +00:00
|
|
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
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
|
|
|
/*
|
|
|
|
|
* Accessor for driver support level
|
|
|
|
|
*/
|
|
|
|
|
NMDriverSupportLevel nm_device_get_driver_support_level (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NM_DRIVER_UNSUPPORTED);
|
|
|
|
|
|
|
|
|
|
return (dev->driver_support_level);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-06-24 14:18:37 +00:00
|
|
|
/*
|
|
|
|
|
* Get/set functions for link_active
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_get_link_active (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
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);
|
|
|
|
|
|
|
|
|
|
dev->link_active = link_active;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/*
|
|
|
|
|
* Get/set functions for now_scanning
|
|
|
|
|
*/
|
|
|
|
|
gboolean nm_device_get_now_scanning (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->options.wireless.now_scanning);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_set_now_scanning (NMDevice *dev, const gboolean now_scanning)
|
|
|
|
|
{
|
|
|
|
|
gboolean old_val;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
|
|
|
|
old_val = nm_device_get_now_scanning (dev);
|
|
|
|
|
dev->options.wireless.now_scanning = now_scanning;
|
|
|
|
|
if (old_val != now_scanning)
|
|
|
|
|
nm_dbus_schedule_device_status_change (dev, DEVICE_STATUS_CHANGE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
return (dev->options.wireless.supports_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);
|
|
|
|
|
|
|
|
|
|
return (dev->options.wired.has_carrier_detect);
|
|
|
|
|
}
|
|
|
|
|
|
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)
|
|
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
2005-01-13 19:04:26 +00:00
|
|
|
int sk;
|
2005-01-07 18:07:06 +00:00
|
|
|
gboolean associated = FALSE;
|
|
|
|
|
|
|
|
|
|
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)
|
|
|
|
|
return (nm_device_get_link_active (dev));
|
|
|
|
|
|
|
|
|
|
if ((sk = iw_sockets_open ()) < 0)
|
|
|
|
|
return (FALSE);
|
|
|
|
|
|
|
|
|
|
/* Some cards, for example ipw2x00 cards, can short-circuit the MAC
|
|
|
|
|
* address check using this check on IWNAME. Its faster.
|
|
|
|
|
*/
|
|
|
|
|
if (iw_get_ext (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.
|
|
|
|
|
*/
|
2005-01-07 18:07:06 +00:00
|
|
|
if (iw_get_ext (sk, nm_device_get_iface (dev), SIOCGIWAP, &wrq) >= 0)
|
|
|
|
|
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:
|
2005-01-07 18:07:06 +00:00
|
|
|
close (sk);
|
2005-01-13 19:04:26 +00:00
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
return (associated);
|
|
|
|
|
}
|
|
|
|
|
|
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_wireless_link_active
|
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
|
|
|
* Gets the link state of a wireless device
|
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
|
|
|
static gboolean nm_device_wireless_link_active (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)
|
|
|
|
|
return (nm_device_get_link_active (dev));
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
if (!nm_device_wireless_is_associated (dev))
|
2004-12-16 04:24:20 +00:00
|
|
|
return (FALSE);
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* If we don't have a "best" ap, we can't logically have a valid link
|
|
|
|
|
* that we want to use.
|
|
|
|
|
*/
|
|
|
|
|
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
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
if (!nm_device_need_ap_switch (dev))
|
|
|
|
|
link = TRUE;
|
|
|
|
|
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
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (link);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_wired_link_active
|
|
|
|
|
*
|
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
|
|
|
*
|
|
|
|
|
*/
|
2005-02-07 23:04:05 +00:00
|
|
|
static gboolean nm_device_wired_link_active (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;
|
|
|
|
|
|
|
|
|
|
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)
|
|
|
|
|
return (nm_device_get_link_active (dev));
|
|
|
|
|
|
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.
|
|
|
|
|
*/
|
|
|
|
|
if (dev->options.wired.has_carrier_detect != TRUE)
|
|
|
|
|
link = TRUE;
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Device has carrier detect, yay! */
|
|
|
|
|
if (hal_device_property_exists (dev->app_data->hal_ctx, nm_device_get_udi (dev), "net.80203.link"))
|
|
|
|
|
link = hal_device_get_property_bool (dev->app_data->hal_ctx, nm_device_get_udi (dev), "net.80203.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
|
|
|
|
|
|
|
|
return (link);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_update_link_active
|
|
|
|
|
*
|
|
|
|
|
* Updates the link state for a particular device.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-02-07 23:04:05 +00:00
|
|
|
void nm_device_update_link_active (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;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
switch (nm_device_get_type (dev))
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2004-08-05 18:54:29 +00:00
|
|
|
case DEVICE_TYPE_WIRELESS_ETHERNET:
|
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
|
|
|
link = nm_device_wireless_link_active (dev);
|
2004-09-13 17:43:16 +00:00
|
|
|
/* Update our current signal strength too */
|
|
|
|
|
nm_device_update_signal_strength (dev);
|
2004-06-24 14:18:37 +00:00
|
|
|
break;
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
case DEVICE_TYPE_WIRED_ETHERNET:
|
2005-02-07 23:04:05 +00:00
|
|
|
link = nm_device_wired_link_active (dev);
|
2004-09-08 18:14:42 +00:00
|
|
|
break;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
|
|
|
|
default:
|
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
|
|
|
link = nm_device_get_link_active (dev); /* Can't get link info for this device, so don't change link status */
|
2004-06-24 14:18:37 +00:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Update device link status and global state variable if the status changed */
|
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 (link != nm_device_get_link_active (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
|
|
|
nm_device_set_link_active (dev, link);
|
2005-01-21 19:54:18 +00:00
|
|
|
nm_dbus_schedule_device_status_change (dev, DEVICE_STATUS_CHANGE);
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_policy_schedule_state_update (dev->app_data);
|
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)
|
|
|
|
|
{
|
2005-01-13 19:04:26 +00:00
|
|
|
int 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);
|
|
|
|
|
}
|
|
|
|
|
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
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2004-09-13 21:00:35 +00:00
|
|
|
wireless_config info;
|
|
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
err = iw_get_basic_config(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
|
2004-12-17 21:39:23 +00:00
|
|
|
syslog (LOG_ERR, "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
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
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)
|
|
|
|
|
{
|
2005-01-10 02:14:04 +00:00
|
|
|
int sk;
|
2004-06-24 14:18:37 +00:00
|
|
|
int err;
|
|
|
|
|
struct iwreq wreq;
|
|
|
|
|
unsigned char safe_essid[IW_ESSID_MAX_SIZE + 1] = "\0";
|
|
|
|
|
|
|
|
|
|
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
|
|
|
|
|
{
|
|
|
|
|
strncpy (safe_essid, essid, IW_ESSID_MAX_SIZE);
|
|
|
|
|
safe_essid[IW_ESSID_MAX_SIZE] = '\0';
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
|
|
|
|
wreq.u.essid.pointer = (caddr_t) safe_essid;
|
|
|
|
|
wreq.u.essid.length = strlen (safe_essid) + 1;
|
|
|
|
|
wreq.u.essid.flags = 1; /* Enable essid on card */
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCSIWESSID, &wreq);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (err == -1)
|
2004-12-17 21:39:23 +00:00
|
|
|
syslog (LOG_ERR, "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
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
close (sk);
|
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.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
double nm_device_get_frequency (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
int sk;
|
|
|
|
|
int err;
|
|
|
|
|
double freq = 0;
|
|
|
|
|
|
|
|
|
|
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;
|
|
|
|
|
|
|
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
|
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
|
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCGIWFREQ, &wrq);
|
|
|
|
|
if (err >= 0)
|
|
|
|
|
freq = iw_freq2float (&wrq.u.freq);
|
|
|
|
|
if (err == -1)
|
|
|
|
|
syslog (LOG_ERR, "nm_device_get_frequency(): error getting frequency for device %s. errno = %d", nm_device_get_iface (dev), errno);
|
|
|
|
|
|
|
|
|
|
close (sk);
|
|
|
|
|
}
|
|
|
|
|
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
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_set_frequency (NMDevice *dev, const double freq)
|
|
|
|
|
{
|
|
|
|
|
int sk;
|
|
|
|
|
int 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
|
|
|
/* 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;
|
|
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
|
|
|
|
{
|
|
|
|
|
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);
|
|
|
|
|
}
|
2004-12-17 17:16:22 +00:00
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCSIWFREQ, &wrq);
|
|
|
|
|
if (err == -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;
|
|
|
|
|
if (iw_set_ext (sk, nm_device_get_iface (dev), SIOCSIWFREQ, &wrq) != -1)
|
|
|
|
|
success = TRUE;
|
|
|
|
|
}
|
|
|
|
|
}
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
close (sk);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_get_bitrate
|
|
|
|
|
*
|
|
|
|
|
* For wireless devices, get the bitrate to broadcast/receive at.
|
|
|
|
|
* Returned value is rate in KHz.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
int nm_device_get_bitrate (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
int sk;
|
|
|
|
|
int err = -1;
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
|
|
|
|
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;
|
|
|
|
|
|
|
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
|
|
|
|
{
|
|
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCGIWRATE, &wrq);
|
|
|
|
|
close (sk);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
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
|
|
|
*
|
|
|
|
|
*/
|
2005-01-21 19:32:08 +00:00
|
|
|
void nm_device_set_bitrate (NMDevice *dev, const int Mbps)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
int sk;
|
|
|
|
|
|
|
|
|
|
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;
|
|
|
|
|
|
2004-12-17 17:16:22 +00:00
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
|
|
|
|
{
|
|
|
|
|
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) */
|
2004-12-22 18:42:37 +00:00
|
|
|
iw_set_ext (sk, nm_device_get_iface (dev), SIOCSIWRATE, &wrq);
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
close (sk);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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)
|
|
|
|
|
{
|
|
|
|
|
int iwlib_socket;
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
|
|
|
|
|
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
|
|
|
|
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} };
|
|
|
|
|
gboolean link = nm_device_get_link_active (dev);
|
|
|
|
|
|
|
|
|
|
memcpy ((link ? &good_addr : &bad_addr), &(wrq.u.ap_addr.sa_data), sizeof (struct ether_addr));
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
iwlib_socket = iw_sockets_open ();
|
|
|
|
|
if (iw_get_ext (iwlib_socket, nm_device_get_iface (dev), SIOCGIWAP, &wrq) >= 0)
|
|
|
|
|
memcpy (addr, &(wrq.u.ap_addr.sa_data), sizeof (struct ether_addr));
|
|
|
|
|
else
|
|
|
|
|
memset (addr, 0, sizeof (struct ether_addr));
|
|
|
|
|
close (iwlib_socket);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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
|
|
|
{
|
2005-01-10 02:14:04 +00:00
|
|
|
int sk;
|
2004-06-24 14:18:37 +00:00
|
|
|
int err;
|
|
|
|
|
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
|
|
|
|
|
{
|
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
|
|
|
strncpy (safe_key, key, IW_ENCODING_TOKEN_MAX);
|
2004-06-24 14:18:37 +00:00
|
|
|
safe_key[IW_ENCODING_TOKEN_MAX] = '\0';
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
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
|
|
|
|
|
|
|
|
if (strlen (safe_key) == 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_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];
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
keylen = iw_in_key_full (sk, nm_device_get_iface (dev), 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)
|
|
|
|
|
{
|
2005-01-10 02:14:04 +00:00
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCSIWENCODE, &wreq);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (err == -1)
|
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
|
|
|
syslog (LOG_ERR, "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
|
|
|
}
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
close (sk);
|
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
|
|
|
} else syslog (LOG_ERR, "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)
|
|
|
|
|
{
|
|
|
|
|
gboolean has_range;
|
2005-01-10 02:14:04 +00:00
|
|
|
int sk;
|
2004-09-13 17:43:16 +00:00
|
|
|
iwrange range;
|
|
|
|
|
iwstats stats;
|
|
|
|
|
int percent = -1;
|
|
|
|
|
|
|
|
|
|
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.
|
|
|
|
|
*/
|
|
|
|
|
if (dev != dev->app_data->active_device)
|
|
|
|
|
{
|
|
|
|
|
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
|
|
|
}
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
has_range = (iw_get_range_info (sk, nm_device_get_iface (dev), &range) >= 0);
|
|
|
|
|
if (iw_get_stats (sk, nm_device_get_iface (dev), &stats, &range, has_range) == 0)
|
2004-09-13 17:43:16 +00:00
|
|
|
{
|
2005-01-31 19:53:38 +00:00
|
|
|
percent = nm_wireless_qual_to_percent (&stats.qual, (const iwqual *)(&dev->options.wireless.max_qual),
|
|
|
|
|
(const iwqual *)(&dev->options.wireless.avg_qual));
|
2004-09-13 17:43:16 +00:00
|
|
|
}
|
2005-01-10 02:14:04 +00:00
|
|
|
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;
|
|
|
|
|
|
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;
|
2005-01-10 02:14:04 +00:00
|
|
|
int 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;
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
if ((sk = nm_device_open_sock ()) < 0)
|
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)));
|
2005-01-10 02:14:04 +00:00
|
|
|
err = ioctl (sk, SIOCGIFADDR, &req);
|
|
|
|
|
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 the new address is different, send an IP4AddressChanged signal on the bus */
|
|
|
|
|
if (new_address != nm_device_get_ip4_address (dev))
|
|
|
|
|
{
|
2004-07-25 02:40:19 +00:00
|
|
|
nm_dbus_signal_device_ip4_address_change (dev->app_data->dbus_connection, dev);
|
2004-07-06 04:45:00 +00:00
|
|
|
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
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_get_hw_address(NMDevice *dev, unsigned char hw_addr[ETH_ALEN])
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
memcpy (hw_addr, dev->hw_addr, ETH_ALEN);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_update_hw_address (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
struct ifreq req;
|
2005-01-10 02:14:04 +00:00
|
|
|
int 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)
|
|
|
|
|
{
|
|
|
|
|
memset (dev->hw_addr, 0, ETH_ALEN);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
if ((sk = nm_device_open_sock ()) < 0)
|
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)));
|
2005-01-10 02:14:04 +00:00
|
|
|
err = ioctl (sk, SIOCGIFHWADDR, &req);
|
|
|
|
|
close (sk);
|
2004-11-16 02:41:53 +00:00
|
|
|
if (err != 0)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
memcpy (dev->hw_addr, req.ifr_hwaddr.sa_data, ETH_ALEN);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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)
|
|
|
|
|
{
|
|
|
|
|
struct ifreq ifr;
|
2005-01-10 02:14:04 +00:00
|
|
|
int sk;
|
2004-06-24 14:18:37 +00:00
|
|
|
int err;
|
|
|
|
|
guint32 flags = up ? IFF_UP : ~IFF_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;
|
|
|
|
|
}
|
|
|
|
|
|
2004-10-21 19:00:13 +00:00
|
|
|
if (nm_device_get_driver_support_level (dev) == NM_DRIVER_UNSUPPORTED)
|
|
|
|
|
return;
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
sk = nm_device_open_sock ();
|
|
|
|
|
if (sk < 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
/* Get flags already there */
|
|
|
|
|
strcpy (ifr.ifr_name, nm_device_get_iface (dev));
|
2005-01-10 02:14:04 +00:00
|
|
|
err = ioctl (sk, SIOCGIFFLAGS, &ifr);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (!err)
|
|
|
|
|
{
|
|
|
|
|
/* If the interface doesn't have those flags already,
|
|
|
|
|
* set them on it.
|
|
|
|
|
*/
|
|
|
|
|
if ((ifr.ifr_flags^flags) & IFF_UP)
|
|
|
|
|
{
|
|
|
|
|
ifr.ifr_flags &= ~IFF_UP;
|
|
|
|
|
ifr.ifr_flags |= IFF_UP & flags;
|
2005-01-10 02:14:04 +00:00
|
|
|
if ((err = ioctl (sk, SIOCSIFFLAGS, &ifr)))
|
2004-08-23 19:20:49 +00:00
|
|
|
syslog (LOG_ERR, "nm_device_set_up_down() could not bring device %s %s. errno = %d", nm_device_get_iface (dev), (up ? "up" : "down"), errno );
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
2005-02-11 14:54:40 +00:00
|
|
|
/* Make sure we have a valid MAC address, some cards reload firmware when they
|
|
|
|
|
* are brought up.
|
|
|
|
|
*/
|
|
|
|
|
if (!nm_ethernet_address_is_valid((struct ether_addr *)dev->hw_addr))
|
|
|
|
|
nm_device_update_hw_address(dev);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
else
|
2004-08-23 19:20:49 +00:00
|
|
|
syslog (LOG_ERR, "nm_device_set_up_down() could not get flags for device %s. errno = %d", nm_device_get_iface (dev), errno );
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
close (sk);
|
2004-06-24 14:18:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Interface state functions: bring up, down, check
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_bring_up (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
nm_device_set_up_down (dev, TRUE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_bring_down (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
nm_device_set_up_down (dev, FALSE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
gboolean nm_device_is_up (NMDevice *dev)
|
|
|
|
|
{
|
2005-01-10 02:14:04 +00:00
|
|
|
int 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);
|
|
|
|
|
|
2005-01-10 02:14:04 +00:00
|
|
|
sk = nm_device_open_sock ();
|
|
|
|
|
if (sk < 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
return (FALSE);
|
|
|
|
|
|
|
|
|
|
/* Get device's flags */
|
|
|
|
|
strcpy (ifr.ifr_name, nm_device_get_iface (dev));
|
2005-01-10 02:14:04 +00:00
|
|
|
err = ioctl (sk, SIOCGIFFLAGS, &ifr);
|
|
|
|
|
close (sk);
|
2004-06-24 14:18:37 +00:00
|
|
|
if (!err)
|
|
|
|
|
return (!((ifr.ifr_flags^IFF_UP) & IFF_UP));
|
|
|
|
|
|
2004-08-23 19:20:49 +00:00
|
|
|
syslog (LOG_ERR, "nm_device_is_up() could not get flags for device %s. errno = %d", nm_device_get_iface (dev), errno );
|
2004-06-24 14:18:37 +00:00
|
|
|
return (FALSE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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)
|
|
|
|
|
{
|
|
|
|
|
int sk;
|
|
|
|
|
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 */
|
|
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
|
|
|
|
{
|
|
|
|
|
struct iwreq wrq;
|
|
|
|
|
int err;
|
|
|
|
|
|
|
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCGIWMODE, &wrq);
|
|
|
|
|
if (err == 0)
|
|
|
|
|
{
|
|
|
|
|
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
|
|
|
|
|
syslog (LOG_ERR, "nm_device_get_mode (%s): error setting card to Infrastructure mode. errno = %d", nm_device_get_iface (dev), errno);
|
|
|
|
|
close (sk);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
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
|
|
|
{
|
|
|
|
|
int sk;
|
2004-12-16 04:24:20 +00:00
|
|
|
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 */
|
|
|
|
|
sk = iw_sockets_open ();
|
|
|
|
|
if (sk >= 0)
|
|
|
|
|
{
|
2004-12-16 04:24:20 +00:00
|
|
|
struct iwreq wreq;
|
|
|
|
|
int err;
|
|
|
|
|
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)
|
|
|
|
|
{
|
|
|
|
|
err = iw_set_ext (sk, nm_device_get_iface (dev), SIOCSIWMODE, &wreq);
|
|
|
|
|
if (err == 0)
|
|
|
|
|
success = TRUE;
|
|
|
|
|
else
|
2004-12-17 17:16:22 +00:00
|
|
|
syslog (LOG_ERR, "nm_device_set_mode (%s): error setting card to Infrastructure mode. errno = %d", nm_device_get_iface (dev), errno);
|
2004-12-16 04:24:20 +00:00
|
|
|
}
|
2004-10-12 16:44:58 +00:00
|
|
|
close (sk);
|
|
|
|
|
}
|
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
|
|
|
/*
|
|
|
|
|
* nm_device_activation_schedule_finish
|
|
|
|
|
*
|
|
|
|
|
* Schedule an idle routine in the main thread to finish the activation.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2005-01-21 19:32:08 +00:00
|
|
|
void nm_device_activation_schedule_finish (NMDevice *dev, DeviceStatus activation_result)
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
|
|
|
|
GSource *source = NULL;
|
|
|
|
|
NMActivationResult *result = NULL;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
|
|
|
|
|
result = g_malloc0 (sizeof (NMActivationResult));
|
|
|
|
|
nm_device_ref (dev); /* Ref device for idle handler */
|
|
|
|
|
result->dev = dev;
|
2005-01-21 19:32:08 +00:00
|
|
|
result->result = activation_result;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
|
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, nm_policy_activation_finish, (gpointer)result, NULL);
|
|
|
|
|
g_source_attach (source, dev->app_data->main_context);
|
|
|
|
|
g_source_unref (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_activation_schedule_start
|
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
|
|
|
* Tell the device thread to begin activation.
|
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
|
|
|
*
|
|
|
|
|
* Returns: TRUE on success activation beginning
|
|
|
|
|
* FALSE on error beginning activation (bad params, couldn't create 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
|
|
|
gboolean nm_device_activation_schedule_start (NMDevice *dev)
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
NMData *data = 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
|
|
|
GSource *source = 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);
|
2004-10-23 06:43:01 +00:00
|
|
|
g_return_val_if_fail (!dev->activating, TRUE); /* Return if activation has already begun */
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
data = dev->app_data;
|
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
|
|
|
g_return_val_if_fail (data != NULL, 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
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Reset communication flags between worker and main thread */
|
|
|
|
|
dev->activating = TRUE;
|
|
|
|
|
dev->quit_activation = FALSE;
|
|
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_device_set_now_scanning (dev, TRUE);
|
2005-01-07 18:07:06 +00:00
|
|
|
dev->options.wireless.user_key_received = FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (nm_device_get_driver_support_level (dev) == NM_DRIVER_UNSUPPORTED)
|
|
|
|
|
{
|
|
|
|
|
dev->activating = FALSE;
|
|
|
|
|
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
|
|
|
|
2005-01-18 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 ();
|
|
|
|
|
g_source_set_callback (source, nm_device_activate, dev, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
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
|
|
|
|
2004-08-26 20:05:24 +00:00
|
|
|
nm_dbus_signal_device_status_change (data->dbus_connection, dev, DEVICE_ACTIVATING);
|
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
|
|
|
|
2004-08-11 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 (TRUE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-09-09 20:02:59 +00:00
|
|
|
/*
|
2004-12-16 04:24:20 +00:00
|
|
|
* nm_device_activation_handle_cancel
|
2004-09-09 20:02:59 +00:00
|
|
|
*
|
|
|
|
|
* Check whether we should stop activation, and if so clean up flags
|
|
|
|
|
* and other random things.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-12-16 04:24:20 +00:00
|
|
|
static gboolean nm_device_activation_handle_cancel (NMDevice *dev)
|
2004-09-09 20:02:59 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, TRUE);
|
|
|
|
|
|
|
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
|
|
|
|
if (dev->quit_activation)
|
|
|
|
|
{
|
|
|
|
|
syslog (LOG_DEBUG, "nm_device_activation_worker(%s): activation canceled.", nm_device_get_iface (dev));
|
2005-01-21 19:32:08 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
nm_device_set_now_scanning (dev, FALSE);
|
2004-09-09 20:02:59 +00:00
|
|
|
return (TRUE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (FALSE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
/*
|
|
|
|
|
* 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)
|
|
|
|
|
{
|
|
|
|
|
struct timeval end_time;
|
|
|
|
|
struct timeval cur_time;
|
|
|
|
|
gboolean link = FALSE;
|
|
|
|
|
double last_freq = 0;
|
|
|
|
|
guint assoc_count = 0;
|
|
|
|
|
gint pause_value;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (time > 0, FALSE);
|
|
|
|
|
|
|
|
|
|
pause_value = nm_device_get_association_pause_value (dev);
|
|
|
|
|
if (pause_value < 1)
|
|
|
|
|
return FALSE;
|
|
|
|
|
|
|
|
|
|
gettimeofday (&end_time, NULL);
|
|
|
|
|
end_time.tv_sec += pause_value;
|
|
|
|
|
|
|
|
|
|
/* We more or less keep asking the driver for the frequency the card is on, and
|
|
|
|
|
* when the frequency has stabilized (the driver has to scan channels to find the AP,
|
|
|
|
|
* and when it finds the AP it stops scanning) and the MAC is valid, we think we
|
|
|
|
|
* have a link.
|
|
|
|
|
*/
|
|
|
|
|
gettimeofday (&cur_time, NULL);
|
|
|
|
|
while (cur_time.tv_sec < end_time.tv_sec)
|
|
|
|
|
{
|
|
|
|
|
double cur_freq = nm_device_get_frequency (dev);
|
|
|
|
|
gboolean assoc = nm_device_wireless_is_associated (dev);
|
|
|
|
|
char *cur_essid = nm_device_get_essid (dev);
|
|
|
|
|
|
|
|
|
|
if ((cur_freq == last_freq) && assoc && !strcmp (essid, cur_essid))
|
|
|
|
|
assoc_count++;
|
|
|
|
|
else
|
|
|
|
|
assoc_count = 0;
|
|
|
|
|
last_freq = cur_freq;
|
|
|
|
|
|
|
|
|
|
g_usleep (G_USEC_PER_SEC / 2);
|
|
|
|
|
if (nm_device_activation_should_cancel (dev))
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
gettimeofday (&cur_time, NULL);
|
|
|
|
|
if ((cur_time.tv_sec >= end_time.tv_sec) && (cur_time.tv_usec >= end_time.tv_usec))
|
|
|
|
|
break;
|
|
|
|
|
|
|
|
|
|
/* Assume that if we've been associated this long, we might as well just stop. */
|
|
|
|
|
if (assoc_count >= 9)
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* If we've had a reasonable association count, we say we have a link */
|
|
|
|
|
if (assoc_count > 6)
|
|
|
|
|
link = TRUE;
|
|
|
|
|
|
|
|
|
|
return (link);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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
|
|
|
|
2004-10-29 16:37:43 +00:00
|
|
|
/* Force the card into Managed/Infrastructure mode */
|
2004-11-17 Dan Williams <dcbw@redhat.com>
* Cache access point MAC addresses in NetworkManagerInfo after you've explicitly
connected to them. Then, after a scan, match up non-ESSID-broadcasting access
points with any cached MAC addresses from NetworkManagerInfo. Allows us to
show known access points that don't broadcast their ESSID in the menus without
any user intervention whatsoever.
* info-daemon/NetworkManagerInfoDbus.c
- (nmi_dbus_get_network_addresses, nmi_dbus_add_network_address): new functions
for dbus method calls "getNetworkAddresses" and "addNetworkAddress"
* src/NetworkManagerAP.[ch]
- Add a "user_addresses" data member to the NMAccessPoint structure
- (nm_ap_get_user_addresses, nm_ap_set_user_addresses): new functions for accessing
the user_addresses data member
* src/NetworkManagerAPList.c
- (nm_ap_list_get_ap_by_address): check user_addresses list too, instead of just
the AP's reported address
- (nm_ap_list_update_network): grab the user_addresses list from NetworkManagerInfo
* src/NetworkManagerDHCP.c
- Increase DHCP timeout from 25s -> 30s
* src/NetworkManagerDbus.[ch]
- (nm_dbus_get_network_addresses, nm_dbus_add_network_address): have NMI get/set
user addresses
* src/NetworkManagerDevice.c
- (nm_device_set_wireless_config): bring down the interface, wait 4s, bring it up,
wait 2s, then configure it. Sometimes Prism54 cards will freeze up with
"mgnt tx queue full", seemingly in response to NM controlling the card too much.
So, we take the card down to clear it out.
- (nm_device_do_normal_scan): Copy over AP ESSIDs from the allowed access point list
too, since that's where the user_addresses are
* src/NetworkManagerPolicy.c
- (nm_state_modification_monitor): Tell NMI to add an AP's hardware address to
that wireless networks' user_addresses list upon successful activation
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@319 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-11-17 17:51:36 +00:00
|
|
|
nm_device_bring_down (dev);
|
2005-01-10 06:41:02 +00:00
|
|
|
g_usleep (G_USEC_PER_SEC * 2);
|
2004-11-17 Dan Williams <dcbw@redhat.com>
* Cache access point MAC addresses in NetworkManagerInfo after you've explicitly
connected to them. Then, after a scan, match up non-ESSID-broadcasting access
points with any cached MAC addresses from NetworkManagerInfo. Allows us to
show known access points that don't broadcast their ESSID in the menus without
any user intervention whatsoever.
* info-daemon/NetworkManagerInfoDbus.c
- (nmi_dbus_get_network_addresses, nmi_dbus_add_network_address): new functions
for dbus method calls "getNetworkAddresses" and "addNetworkAddress"
* src/NetworkManagerAP.[ch]
- Add a "user_addresses" data member to the NMAccessPoint structure
- (nm_ap_get_user_addresses, nm_ap_set_user_addresses): new functions for accessing
the user_addresses data member
* src/NetworkManagerAPList.c
- (nm_ap_list_get_ap_by_address): check user_addresses list too, instead of just
the AP's reported address
- (nm_ap_list_update_network): grab the user_addresses list from NetworkManagerInfo
* src/NetworkManagerDHCP.c
- Increase DHCP timeout from 25s -> 30s
* src/NetworkManagerDbus.[ch]
- (nm_dbus_get_network_addresses, nm_dbus_add_network_address): have NMI get/set
user addresses
* src/NetworkManagerDevice.c
- (nm_device_set_wireless_config): bring down the interface, wait 4s, bring it up,
wait 2s, then configure it. Sometimes Prism54 cards will freeze up with
"mgnt tx queue full", seemingly in response to NM controlling the card too much.
So, we take the card down to clear it out.
- (nm_device_do_normal_scan): Copy over AP ESSIDs from the allowed access point list
too, since that's where the user_addresses are
* src/NetworkManagerPolicy.c
- (nm_state_modification_monitor): Tell NMI to add an AP's hardware address to
that wireless networks' user_addresses list upon successful activation
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@319 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
2004-11-17 17:51:36 +00:00
|
|
|
nm_device_bring_up (dev);
|
|
|
|
|
g_usleep (G_USEC_PER_SEC * 2);
|
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
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
char *hashed_key = nm_ap_get_enc_key_hashed (ap);
|
|
|
|
|
|
|
|
|
|
if (auth == NM_DEVICE_AUTH_METHOD_NONE)
|
|
|
|
|
{
|
|
|
|
|
nm_ap_set_auth_method (ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
|
|
|
|
syslog (LOG_ERR, "Activation (%s/wireless): AP '%s' said it was encrypted, but had "
|
|
|
|
|
"'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
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
syslog (LOG_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
|
|
|
|
2005-01-13 19:04:26 +00:00
|
|
|
/* Bring the device up and pause to allow card to associate. */
|
|
|
|
|
g_usleep (G_USEC_PER_SEC * 2);
|
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 */
|
|
|
|
|
|
2005-01-13 19:04:26 +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
|
|
|
/*
|
|
|
|
|
* nm_device_activate_wireless_adhoc
|
|
|
|
|
*
|
|
|
|
|
* Create an ad-hoc network (rather than associating with one).
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_activate_wireless_adhoc (NMDevice *dev, NMAccessPoint *ap)
|
|
|
|
|
{
|
|
|
|
|
gboolean success = FALSE;
|
|
|
|
|
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;
|
2005-01-31 19:53:38 +00:00
|
|
|
iwrange range;
|
|
|
|
|
int sk;
|
2004-12-17 17:16:22 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (ap != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
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);
|
|
|
|
|
|
2005-01-31 19:53:38 +00:00
|
|
|
if ((sk = iw_sockets_open ()) < 0)
|
|
|
|
|
return FALSE;
|
|
|
|
|
|
|
|
|
|
if (iw_get_range_info (sk, nm_device_get_iface (dev), &range) < 0)
|
|
|
|
|
{
|
|
|
|
|
close (sk);
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
|
|
|
|
close (sk);
|
|
|
|
|
|
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;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (freq_to_use)
|
|
|
|
|
{
|
|
|
|
|
nm_ap_set_freq (ap, freq_to_use);
|
|
|
|
|
|
|
|
|
|
syslog (LOG_INFO, "Will create network '%s' with frequency %f.\n", nm_ap_get_essid (ap), nm_ap_get_freq (ap));
|
2005-01-21 19:32:08 +00:00
|
|
|
if ((success = nm_device_set_wireless_config (dev, ap)))
|
2004-12-17 17:16:22 +00:00
|
|
|
success = nm_device_activation_configure_ip (dev, TRUE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (success);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
static gboolean AP_NEED_KEY (NMDevice *dev, NMAccessPoint *ap)
|
2004-10-29 16:37:43 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
char *essid;
|
|
|
|
|
gboolean need_key = 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
|
|
|
|
2004-10-29 16:37:43 +00:00
|
|
|
g_return_val_if_fail (ap != NULL, 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
|
|
|
essid = nm_ap_get_essid (ap);
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
if (!nm_ap_get_encrypted (ap))
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
|
|
|
|
syslog (LOG_NOTICE, "Activation (%s/wireless): access point '%s' is unencrypted, no key needed.",
|
|
|
|
|
nm_device_get_iface (dev), essid ? essid : "(null)");
|
|
|
|
|
}
|
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))
|
|
|
|
|
{
|
|
|
|
|
syslog (LOG_NOTICE, "Activation (%s/wireless): access point '%s' is encrypted, and a key exists. No new key needed.",
|
|
|
|
|
nm_device_get_iface (dev), essid ? essid : "(null)");
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
else
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
|
|
|
|
syslog (LOG_NOTICE, "Activation (%s/wireless): access point '%s' is encrypted, but NO valid key exists. New key needed.",
|
|
|
|
|
nm_device_get_iface (dev), essid ? essid : "(null)");
|
|
|
|
|
need_key = TRUE;
|
|
|
|
|
}
|
2005-01-07 18:07:06 +00:00
|
|
|
}
|
2004-10-29 16:37:43 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
return (need_key);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* get_initial_auth_method
|
|
|
|
|
*
|
2005-01-31 19:53:38 +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.
|
2005-01-21 19:32:08 +00:00
|
|
|
*
|
|
|
|
|
*/
|
2005-01-31 19:53:38 +00:00
|
|
|
static NMDeviceAuthMethod get_initial_auth_method (NMAccessPoint *ap, NMAccessPointList *allowed_list)
|
2005-01-21 19:32:08 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (ap != NULL, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
|
|
|
|
|
|
|
|
|
if (nm_ap_get_encrypted (ap))
|
|
|
|
|
{
|
2005-01-31 19:53:38 +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);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
if ( (auth == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM)
|
|
|
|
|
|| (auth == NM_DEVICE_AUTH_METHOD_SHARED_KEY))
|
2005-01-31 19:53:38 +00:00
|
|
|
return (auth);
|
2005-01-21 19:32:08 +00:00
|
|
|
else
|
|
|
|
|
return (NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return (NM_DEVICE_AUTH_METHOD_NONE);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
void invalidate_ap (NMDevice *dev, NMAccessPoint *ap)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
g_return_if_fail (ap != NULL);
|
|
|
|
|
|
|
|
|
|
/* If its an AP the user forced, notify the user it failed. */
|
|
|
|
|
/* FIXME: we dont' set ap's that are in our scan list as "artificial",
|
|
|
|
|
* so we won't be able to signal the user when a connection to on of them
|
|
|
|
|
* failed.
|
|
|
|
|
*/
|
|
|
|
|
if (nm_ap_get_artificial (ap))
|
|
|
|
|
nm_dbus_schedule_network_not_found_signal (dev->app_data, nm_ap_get_essid (ap));
|
|
|
|
|
|
|
|
|
|
nm_ap_set_invalid (ap, TRUE);
|
|
|
|
|
nm_ap_list_append_ap (dev->app_data->invalid_ap_list, ap);
|
|
|
|
|
nm_ap_unref (ap);
|
|
|
|
|
nm_device_update_best_ap (dev);
|
2004-10-29 16:37:43 +00:00
|
|
|
}
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_activate_wireless
|
|
|
|
|
*
|
2005-01-10 06:41:02 +00:00
|
|
|
* Activate a wireless ethernet device. Locking could be confusing here, pay attention to it.
|
|
|
|
|
* We grab the scan mutex because scanning requires us to set certain state on the card,
|
|
|
|
|
* like mode, which could screw up device activation link state checks.
|
2004-11-16 02:41:53 +00:00
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
static gboolean nm_device_activate_wireless (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *best_ap;
|
|
|
|
|
gboolean success = FALSE;
|
|
|
|
|
guint8 attempt = 1;
|
|
|
|
|
char last_essid [50] = "\0";
|
2005-01-21 19:32:08 +00:00
|
|
|
gboolean need_key = FALSE;
|
2005-01-29 23:34:31 +00:00
|
|
|
gboolean found_ap = FALSE;
|
2004-11-16 02:41:53 +00:00
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
2004-11-16 15:06:46 +00:00
|
|
|
|
2005-01-10 06:41:02 +00:00
|
|
|
/* Grab the scan mutex, we don't want the scan thread to mess up our settings
|
|
|
|
|
* during activation and link detection.
|
|
|
|
|
*/
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2004-11-16 15:06:46 +00:00
|
|
|
if (!nm_device_is_up (dev))
|
|
|
|
|
nm_device_bring_up (dev);
|
|
|
|
|
g_usleep (G_USEC_PER_SEC);
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
get_ap:
|
|
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
2004-12-16 04:24:20 +00:00
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
2004-11-16 02:41:53 +00:00
|
|
|
goto out;
|
|
|
|
|
|
2005-01-10 06:41:02 +00:00
|
|
|
/* Get a valid "best" access point we should connect to. We don't hold the scan
|
|
|
|
|
* lock here because this might take a while.
|
|
|
|
|
*/
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2004-11-16 02:41:53 +00:00
|
|
|
while (!(best_ap = nm_device_get_best_ap (dev)))
|
|
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_device_set_now_scanning (dev, TRUE);
|
2005-01-29 23:34:31 +00:00
|
|
|
if (!found_ap)
|
|
|
|
|
syslog (LOG_ERR, "Activation (%s/wireless): waiting for an access point.", nm_device_get_iface (dev));
|
2004-11-16 02:41:53 +00:00
|
|
|
g_usleep (G_USEC_PER_SEC * 2);
|
|
|
|
|
|
|
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
2004-12-16 04:24:20 +00:00
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
2005-01-10 06:41:02 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Wierd as it may seem, we lock here to balance the unlock in "out:" */
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2004-11-16 02:41:53 +00:00
|
|
|
goto out;
|
2005-01-10 06:41:02 +00:00
|
|
|
}
|
2005-01-29 23:34:31 +00:00
|
|
|
found_ap = TRUE;
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
2005-01-29 23:34:31 +00:00
|
|
|
if (found_ap)
|
|
|
|
|
syslog (LOG_ERR, "Activation (%s/wireless): found access point '%s' to use.", nm_device_get_iface (dev), nm_ap_get_essid (best_ap));
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
/* Set ESSID early so that when we send out the DeviceStatusChanged signal below,
|
|
|
|
|
* we are able to respond correctly to queries for "getActiveNetwork" against
|
|
|
|
|
* our device. nm_device_get_path_for_ap() uses the /card's/ AP, not the best_ap.
|
|
|
|
|
*/
|
|
|
|
|
nm_device_set_essid (dev, nm_ap_get_essid (best_ap));
|
|
|
|
|
|
2005-02-10 04:39:40 +00:00
|
|
|
/* We grab the scan mutex so that scanning cannot screw up our link detection, since
|
|
|
|
|
* a scan can change most any attribute on the card for a period of time.
|
|
|
|
|
*/
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_device_set_now_scanning (dev, FALSE);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (nm_ap_get_artificial (best_ap))
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
/* 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
|
2005-02-10 04:39:40 +00:00
|
|
|
* AP was encrypted. We have to update their encryption status on the fly.
|
2004-11-16 02:41:53 +00:00
|
|
|
*/
|
2005-01-21 19:32:08 +00:00
|
|
|
if (nm_ap_get_encrypted (best_ap) || nm_ap_is_enc_key_valid (best_ap))
|
2004-12-17 17:37:12 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_encrypted (best_ap, TRUE);
|
|
|
|
|
nm_ap_set_auth_method (best_ap, NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM);
|
2004-12-17 17:37:12 +00:00
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
}
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
need_key = AP_NEED_KEY (dev, best_ap);
|
|
|
|
|
|
|
|
|
|
need_key:
|
|
|
|
|
if (need_key)
|
|
|
|
|
{
|
|
|
|
|
char *essid = nm_ap_get_essid (best_ap);
|
|
|
|
|
if (strcmp (essid, last_essid) != 0)
|
|
|
|
|
attempt = 1;
|
|
|
|
|
strncpy (&last_essid[0], essid, 49);
|
|
|
|
|
|
|
|
|
|
/* Don't hold the mutex while waiting for a key */
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
/* Get a wireless key */
|
|
|
|
|
dev->options.wireless.user_key_received = FALSE;
|
|
|
|
|
nm_dbus_get_user_key_for_network (dev->app_data->dbus_connection, dev, best_ap, attempt);
|
|
|
|
|
attempt++;
|
|
|
|
|
need_key = FALSE;
|
|
|
|
|
|
|
|
|
|
/* Wait for the key to come back */
|
|
|
|
|
syslog (LOG_DEBUG, "Activation (%s/wireless): asking for user key.", nm_device_get_iface (dev));
|
|
|
|
|
while (!dev->options.wireless.user_key_received && !dev->quit_activation)
|
|
|
|
|
g_usleep (G_USEC_PER_SEC / 2);
|
|
|
|
|
|
|
|
|
|
syslog (LOG_DEBUG, "Activation (%s/wireless): user key received.", nm_device_get_iface (dev));
|
|
|
|
|
|
|
|
|
|
/* Done waiting, grab lock again */
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
/* User may have cancelled the key request, so we need to update our best AP again. */
|
|
|
|
|
nm_ap_unref (best_ap);
|
|
|
|
|
|
|
|
|
|
goto get_ap;
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
|
|
|
|
|
if (nm_ap_get_mode (best_ap) == NETWORK_MODE_ADHOC)
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Only do auto-ip on Ad-Hoc connections for now. We technically
|
|
|
|
|
* could do DHCP on them though.
|
|
|
|
|
*/
|
|
|
|
|
success = nm_device_activation_configure_ip (dev, TRUE);
|
|
|
|
|
goto connect_done;
|
|
|
|
|
}
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
try_connect:
|
|
|
|
|
/* Initial authentication method */
|
2005-01-31 19:53:38 +00:00
|
|
|
nm_ap_set_auth_method (best_ap, get_initial_auth_method (best_ap, dev->app_data->allowed_ap_list));
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
while (success == FALSE)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *tmp_ap = NULL;
|
|
|
|
|
gboolean link = FALSE;
|
|
|
|
|
gboolean adhoc = (nm_ap_get_mode (best_ap) == NETWORK_MODE_ADHOC);
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
|
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
|
|
|
|
goto out;
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_device_set_wireless_config (dev, best_ap);
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
link = nm_device_wireless_wait_for_link (dev, nm_ap_get_essid (best_ap));
|
2004-11-16 02:41:53 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
|
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
|
|
|
|
goto out;
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (!link)
|
|
|
|
|
{
|
|
|
|
|
if (nm_ap_get_auth_method (best_ap) == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM)
|
|
|
|
|
{
|
|
|
|
|
syslog (LOG_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 (best_ap) ? nm_ap_get_essid (best_ap) : "(none)");
|
|
|
|
|
/* Back down to Shared Key mode */
|
|
|
|
|
nm_ap_set_auth_method (best_ap, NM_DEVICE_AUTH_METHOD_SHARED_KEY);
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
else if (nm_ap_get_auth_method (best_ap) == NM_DEVICE_AUTH_METHOD_SHARED_KEY)
|
|
|
|
|
{
|
|
|
|
|
/* Must be in Open System mode and it still didn't work, so
|
|
|
|
|
* we'll invalidate the current "best" ap and get another one */
|
|
|
|
|
syslog (LOG_DEBUG, "Activation (%s/wireless): no hardware link to '%s' in Shared Key mode, trying another access point.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (best_ap) ? nm_ap_get_essid (best_ap) : "(none)");
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
syslog (LOG_DEBUG, "Activation (%s/wireless): no hardware link to '%s' in non-encrypted mode.",
|
|
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (best_ap) ? nm_ap_get_essid (best_ap) : "(none)");
|
|
|
|
|
}
|
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
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
/* All applicable modes failed, invalidate current best_ap and get a new one */
|
|
|
|
|
invalidate_ap (dev, best_ap);
|
2004-11-16 02:41:53 +00:00
|
|
|
goto get_ap;
|
|
|
|
|
}
|
|
|
|
|
|
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.
|
|
|
|
|
*/
|
|
|
|
|
if ((success = nm_device_activation_configure_ip (dev, adhoc)))
|
|
|
|
|
{
|
|
|
|
|
/* Cache the last known good auth method in both NetworkManagerInfo and our allowed AP list */
|
|
|
|
|
nm_dbus_update_network_auth_method (dev->app_data->dbus_connection, nm_ap_get_essid (best_ap), nm_ap_get_auth_method (best_ap));
|
|
|
|
|
if ((tmp_ap = nm_ap_list_get_ap_by_essid (dev->app_data->allowed_ap_list, nm_ap_get_essid (best_ap))))
|
|
|
|
|
nm_ap_set_auth_method (tmp_ap, nm_ap_get_auth_method (best_ap));
|
|
|
|
|
}
|
|
|
|
|
else
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2004-11-22 14:42:34 +00:00
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
2004-12-16 04:24:20 +00:00
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
2004-11-22 14:42:34 +00:00
|
|
|
goto out;
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if ((nm_ap_get_auth_method (best_ap) == NM_DEVICE_AUTH_METHOD_OPEN_SYSTEM) && !adhoc)
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Back down to Shared Key mode */
|
|
|
|
|
syslog (LOG_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 (best_ap) ? nm_ap_get_essid (best_ap) : "(none)");
|
|
|
|
|
nm_ap_set_auth_method (best_ap, NM_DEVICE_AUTH_METHOD_SHARED_KEY);
|
|
|
|
|
continue;
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
2005-01-21 19:32:08 +00:00
|
|
|
else if ((nm_ap_get_auth_method (best_ap) == NM_DEVICE_AUTH_METHOD_SHARED_KEY) && !adhoc)
|
2004-11-16 02:41:53 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
/* Shared Key mode failed, we must have bad WEP key */
|
|
|
|
|
syslog (LOG_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 (best_ap) ? nm_ap_get_essid (best_ap) : "(none)");
|
|
|
|
|
need_key = TRUE;
|
|
|
|
|
goto need_key;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* All applicable modes failed, invalidate current best_ap and get a new one */
|
|
|
|
|
invalidate_ap (dev, best_ap);
|
|
|
|
|
goto get_ap;
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
connect_done:
|
|
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
|
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
|
|
|
|
goto out;
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
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
|
|
|
syslog (LOG_DEBUG, "Activation (%s/wireless): Success! Connected to access point '%s' and got an IP address.",
|
2004-11-16 02:41:53 +00:00
|
|
|
nm_device_get_iface (dev), nm_ap_get_essid (best_ap) ? nm_ap_get_essid (best_ap) : "(none)");
|
2004-11-16 15:06:46 +00:00
|
|
|
nm_ap_unref (best_ap);
|
2004-11-16 02:41:53 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
out:
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_device_set_now_scanning (dev, FALSE);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2004-11-16 02:41:53 +00:00
|
|
|
return (success);
|
|
|
|
|
}
|
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
|
|
|
/*
|
|
|
|
|
* nm_device_activation_configure_ip
|
|
|
|
|
*
|
|
|
|
|
* Perform any IP-based configuration on a device, like running DHCP
|
|
|
|
|
* or manually setting up the IP address, gateway, and default route.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-12-17 17:16:22 +00:00
|
|
|
static gboolean nm_device_activation_configure_ip (NMDevice *dev, gboolean do_only_autoip)
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2004-10-08 05:22:17 +00:00
|
|
|
gboolean success = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2004-09-08 18:14:42 +00:00
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
nm_system_delete_default_route ();
|
2004-12-17 17:16:22 +00:00
|
|
|
if (do_only_autoip)
|
|
|
|
|
{
|
|
|
|
|
success = nm_device_do_autoip (dev);
|
|
|
|
|
}
|
|
|
|
|
else if (nm_device_config_get_use_dhcp (dev))
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
2004-11-16 02:41:53 +00:00
|
|
|
int err;
|
|
|
|
|
|
2004-11-22 14:42:34 +00:00
|
|
|
err = nm_device_dhcp_request (dev);
|
2004-11-16 02:41:53 +00:00
|
|
|
if (err == RET_DHCP_BOUND)
|
2004-10-08 05:22:17 +00:00
|
|
|
success = TRUE;
|
|
|
|
|
else
|
2004-09-08 18:14:42 +00:00
|
|
|
{
|
|
|
|
|
/* Interfaces cannot be down if they are the active interface,
|
|
|
|
|
* otherwise we cannot use them for scanning or link detection.
|
|
|
|
|
*/
|
|
|
|
|
if (nm_device_is_wireless (dev))
|
|
|
|
|
{
|
|
|
|
|
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);
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
if (!nm_device_is_up (dev))
|
|
|
|
|
nm_device_bring_up (dev);
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
/* Manually set up the device */
|
2004-10-12 11:15:47 +00:00
|
|
|
success = nm_system_device_setup_static_ip4_config (dev);
|
2004-09-08 18:14:42 +00:00
|
|
|
}
|
2004-10-08 05:22:17 +00:00
|
|
|
|
2004-12-16 04:24:20 +00:00
|
|
|
if (success)
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
2005-02-10 21:16:03 +00:00
|
|
|
nm_system_device_add_ip6_link_address (dev);
|
2004-12-17 17:16:22 +00:00
|
|
|
nm_system_flush_arp_cache ();
|
2004-12-16 04:24:20 +00:00
|
|
|
nm_system_restart_mdns_responder ();
|
2004-12-17 17:16:22 +00:00
|
|
|
}
|
2004-12-16 04:24:20 +00:00
|
|
|
|
2004-10-08 05:22:17 +00:00
|
|
|
return (success);
|
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
|
|
|
/*
|
2005-01-18 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_activate
|
|
|
|
|
*
|
|
|
|
|
* Activate a device, done from the device's worker thread.
|
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
|
|
|
*
|
|
|
|
|
*/
|
2005-01-18 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_activate (gpointer user_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
|
|
|
{
|
2005-01-07 18:07:06 +00:00
|
|
|
NMDevice *dev = (NMDevice *)user_data;
|
|
|
|
|
gboolean success = FALSE;
|
2005-01-21 19:32:08 +00:00
|
|
|
gboolean finished = FALSE;
|
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
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, 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
|
|
|
|
2005-01-29 23:34:31 +00:00
|
|
|
syslog (LOG_ERR, "Activation (%s) started...", nm_device_get_iface (dev));
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
/* Bring the device up */
|
|
|
|
|
if (!nm_device_is_up (dev));
|
|
|
|
|
nm_device_bring_up (dev);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-10-29 16:37:43 +00:00
|
|
|
if (nm_device_is_wireless (dev))
|
2004-12-17 17:16:22 +00:00
|
|
|
{
|
|
|
|
|
gboolean create_network = FALSE;
|
|
|
|
|
NMAccessPoint *best_ap = nm_device_get_best_ap (dev);
|
|
|
|
|
|
|
|
|
|
if (best_ap)
|
|
|
|
|
{
|
|
|
|
|
if (nm_ap_get_user_created (best_ap))
|
|
|
|
|
{
|
|
|
|
|
create_network = TRUE;
|
|
|
|
|
syslog (LOG_INFO, "Creating wireless network '%s'.\n", nm_ap_get_essid (best_ap));
|
|
|
|
|
success = nm_device_activate_wireless_adhoc (dev, best_ap);
|
|
|
|
|
syslog (LOG_INFO, "Wireless network creation for '%s' was %s.\n", nm_ap_get_essid (best_ap), success ? "successful" : "unsuccessful");
|
|
|
|
|
}
|
|
|
|
|
nm_ap_unref (best_ap);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!create_network)
|
|
|
|
|
success = nm_device_activate_wireless (dev);
|
|
|
|
|
}
|
2004-11-16 02:41:53 +00:00
|
|
|
else if (nm_device_is_wired (dev))
|
2004-12-17 17:16:22 +00:00
|
|
|
success = nm_device_activation_configure_ip (dev, FALSE);
|
2004-07-06 01:34:10 +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
|
|
|
/* If we were told to quit activation, stop the thread and return */
|
2004-12-16 04:24:20 +00:00
|
|
|
if (nm_device_activation_handle_cancel (dev))
|
2004-10-23 06:43:01 +00:00
|
|
|
goto out;
|
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
|
|
|
|
2005-01-18 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 (success)
|
|
|
|
|
syslog (LOG_DEBUG, "Activation (%s) IP configuration/DHCP successful!\n", nm_device_get_iface (dev));
|
|
|
|
|
else
|
|
|
|
|
syslog (LOG_DEBUG, "Activation (%s) IP configuration/DHCP unsuccessful! Ending activation...\n", nm_device_get_iface (dev));
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
finished = TRUE;
|
2004-11-22 14:42:34 +00:00
|
|
|
|
2004-12-05 21:28:42 +00:00
|
|
|
out:
|
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
|
|
|
syslog (LOG_DEBUG, "Activation (%s) ended.\n", nm_device_get_iface (dev));
|
2005-01-21 19:32:08 +00:00
|
|
|
dev->activating = FALSE;
|
|
|
|
|
dev->quit_activation = FALSE;
|
|
|
|
|
if (finished)
|
|
|
|
|
nm_device_activation_schedule_finish (dev, success ? DEVICE_NOW_ACTIVE : DEVICE_ACTIVATION_FAILED);
|
|
|
|
|
|
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;
|
2004-08-11 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
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->activating);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
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
|
|
|
{
|
2004-09-12 03:06:44 +00:00
|
|
|
syslog (LOG_DEBUG, "nm_device_activation_cancel(%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
|
|
|
|
|
|
|
|
/* 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.
|
|
|
|
|
*/
|
2004-10-11 21:32:19 +00:00
|
|
|
while (nm_device_is_activating (dev))
|
2005-01-09 23:15:36 +00:00
|
|
|
{
|
|
|
|
|
/* Nice race here between quit activation and dhcp. We may not have
|
|
|
|
|
* started DHCP when we're told to quit activation, so we need to keep
|
|
|
|
|
* signalling dhcp to quit, which it will pick up whenever it starts.
|
|
|
|
|
* This should really be taken care of a better way.
|
|
|
|
|
*/
|
|
|
|
|
if (dev->dhcp_iface)
|
|
|
|
|
nm_device_dhcp_cease (dev);
|
|
|
|
|
|
2004-09-12 03:06:44 +00:00
|
|
|
g_usleep (G_USEC_PER_SEC / 2);
|
2005-01-09 23:15:36 +00:00
|
|
|
}
|
2004-12-17 17:16:22 +00:00
|
|
|
syslog (LOG_DEBUG, "nm_device_activation_cancel(%s): cancelled.", 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
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_deactivate
|
|
|
|
|
*
|
|
|
|
|
* Remove a device's routing table entries and IP address.
|
|
|
|
|
*
|
|
|
|
|
*/
|
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_deactivate (NMDevice *dev, gboolean just_added)
|
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
|
|
|
|
2004-09-12 03:06:44 +00:00
|
|
|
nm_device_activation_cancel (dev);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-10-21 19:00:13 +00:00
|
|
|
if (nm_device_get_driver_support_level (dev) == NM_DRIVER_UNSUPPORTED)
|
|
|
|
|
return (TRUE);
|
|
|
|
|
|
2005-02-22 00:30:46 +00:00
|
|
|
/* Remove any DHCP timeouts that might have been running */
|
|
|
|
|
if (nm_device_config_get_use_dhcp (dev))
|
|
|
|
|
nm_device_dhcp_remove_timeouts (dev);
|
2005-02-10 04:39:40 +00:00
|
|
|
|
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
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (!just_added && (dev == dev->app_data->active_device))
|
2004-08-26 20:05:24 +00:00
|
|
|
nm_dbus_signal_device_status_change (dev->app_data->dbus_connection, dev, DEVICE_NO_LONGER_ACTIVE);
|
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);
|
2005-01-18 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->options.wireless.scan_interval = 20;
|
2004-07-06 04:45:00 +00:00
|
|
|
}
|
|
|
|
|
|
2004-08-24 00:31:47 +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.
|
|
|
|
|
*
|
|
|
|
|
*/
|
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
|
|
|
void nm_device_set_user_key_for_network (NMDevice *dev, NMAccessPointList *invalid_list,
|
2004-10-08 05:22:17 +00:00
|
|
|
unsigned char *network, unsigned char *key,
|
2005-01-21 19:32:08 +00:00
|
|
|
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
|
|
|
{
|
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
|
|
|
NMAccessPoint *best_ap;
|
|
|
|
|
const char *cancel_message = "***canceled***";
|
|
|
|
|
|
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 (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (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_return_if_fail (network != 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 (key != 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
|
|
|
/* 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
|
|
|
{
|
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
|
|
|
NMAccessPoint *ap;
|
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
|
|
|
|
2004-08-11 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 ((ap = nm_device_ap_list_get_ap_by_essid (dev, network)))
|
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
|
|
|
{
|
2004-08-11 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
|
|
|
NMAccessPoint *invalid_ap = nm_ap_new_from_ap (ap);
|
|
|
|
|
if (invalid_list)
|
|
|
|
|
nm_ap_list_append_ap (invalid_list, invalid_ap);
|
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
|
|
|
}
|
2004-07-19 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_update_best_ap (dev);
|
|
|
|
|
}
|
|
|
|
|
else if ((best_ap = nm_device_get_best_ap (dev)))
|
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
|
|
|
/* Make sure the "best" ap matches the essid we asked for the key of,
|
|
|
|
|
* then set the new key on the access point.
|
|
|
|
|
*/
|
2004-08-20 18:18:37 +00:00
|
|
|
if (nm_null_safe_strcmp (network, nm_ap_get_essid (best_ap)) == 0)
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_ap_set_enc_key_source (best_ap, key, enc_type);
|
2004-10-29 16:37:43 +00:00
|
|
|
|
|
|
|
|
nm_ap_unref (best_ap);
|
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
|
|
|
dev->options.wireless.user_key_received = TRUE;
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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
|
|
|
}
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
/*
|
|
|
|
|
* Get/Set functions for "best" access point
|
|
|
|
|
*
|
2004-10-29 16:37:43 +00:00
|
|
|
* Caller MUST unref returned access point when done with it.
|
|
|
|
|
*
|
2004-07-06 01:34:10 +00:00
|
|
|
*/
|
|
|
|
|
NMAccessPoint *nm_device_get_best_ap (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
|
|
|
NMAccessPoint *best_ap;
|
|
|
|
|
|
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);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
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
|
|
|
best_ap = dev->options.wireless.best_ap;
|
2004-10-29 16:37:43 +00:00
|
|
|
/* Callers get a reffed AP */
|
|
|
|
|
if (best_ap) nm_ap_ref (best_ap);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
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 (best_ap);
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_set_best_ap (NMDevice *dev, NMAccessPoint *ap)
|
|
|
|
|
{
|
|
|
|
|
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));
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
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
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (dev->options.wireless.best_ap)
|
|
|
|
|
nm_ap_unref (dev->options.wireless.best_ap);
|
|
|
|
|
|
|
|
|
|
if (ap)
|
|
|
|
|
nm_ap_ref (ap);
|
|
|
|
|
|
|
|
|
|
dev->options.wireless.best_ap = ap;
|
2004-09-09 20:02:59 +00:00
|
|
|
nm_device_unfreeze_best_ap (dev);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
2004-08-05 18:54:29 +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
|
|
|
/*
|
|
|
|
|
* Freeze/unfreeze best ap
|
|
|
|
|
*
|
|
|
|
|
* If the user explicitly picks a network to associate with, we don't
|
|
|
|
|
* change the active network until it goes out of range.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_freeze_best_ap (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
|
|
|
|
dev->options.wireless.freeze_best_ap = TRUE;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_unfreeze_best_ap (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
|
|
|
|
dev->options.wireless.freeze_best_ap = FALSE;
|
|
|
|
|
}
|
|
|
|
|
|
2004-10-11 21:32:19 +00:00
|
|
|
gboolean nm_device_is_best_ap_frozen (NMDevice *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
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->options.wireless.freeze_best_ap);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-11-16 02:41:53 +00:00
|
|
|
/*
|
|
|
|
|
* Accessor for dhcp_interface
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
struct dhcp_interface *nm_device_get_dhcp_iface (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
|
|
|
|
|
return (dev->dhcp_iface);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_set_dhcp_iface (NMDevice *dev, struct dhcp_interface *dhcp_iface)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
/* NOTE: this function should only be used from the activation worker thread
|
|
|
|
|
* which will take care of shutting down any active DHCP threads and cleaning
|
|
|
|
|
* up the dev->dhcp_iface structure.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
dev->dhcp_iface = dhcp_iface;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
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_get_path_for_ap
|
|
|
|
|
*
|
|
|
|
|
* Return the object path for an access point.
|
|
|
|
|
*
|
|
|
|
|
* NOTE: assumes the access point is actually in the device's access point list.
|
|
|
|
|
*
|
|
|
|
|
*/
|
2004-08-05 18:54:29 +00:00
|
|
|
char * nm_device_get_path_for_ap (NMDevice *dev, NMAccessPoint *ap)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, NULL);
|
|
|
|
|
g_return_val_if_fail (ap != NULL, NULL);
|
|
|
|
|
|
2004-10-13 20:57:23 +00:00
|
|
|
if (nm_ap_get_essid (ap))
|
|
|
|
|
return (g_strdup_printf ("%s/%s/Networks/%s", NM_DBUS_PATH_DEVICES, nm_device_get_iface (dev), nm_ap_get_essid (ap)));
|
|
|
|
|
else
|
|
|
|
|
return (NULL);
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_need_ap_switch
|
|
|
|
|
*
|
|
|
|
|
* Returns TRUE if the essid of the card does not match the essid
|
|
|
|
|
* of the "best" access point it should be associating with.
|
|
|
|
|
*
|
|
|
|
|
*/
|
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_need_ap_switch (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPoint *ap;
|
|
|
|
|
gboolean need_switch = FALSE;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (nm_device_is_wireless (dev), FALSE);
|
|
|
|
|
|
2005-01-29 23:34:31 +00:00
|
|
|
|
2005-02-02 21:49:14 +00:00
|
|
|
/* Since the card's ESSID may change during a scan, we can't really
|
|
|
|
|
* rely on checking the ESSID during that time.
|
2005-01-29 23:34:31 +00:00
|
|
|
*/
|
2005-02-02 21:49:14 +00:00
|
|
|
if (!nm_try_acquire_mutex (dev->options.wireless.scan_mutex, __FUNCTION__))
|
|
|
|
|
return FALSE;
|
2005-01-29 23:34:31 +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
|
|
|
ap = nm_device_get_best_ap (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
|
|
|
if (nm_null_safe_strcmp (nm_device_get_essid (dev), (ap ? nm_ap_get_essid (ap) : NULL)) != 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
|
|
|
need_switch = TRUE;
|
|
|
|
|
|
2005-01-29 23:34:31 +00:00
|
|
|
if (ap)
|
|
|
|
|
nm_ap_unref (ap);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.scan_mutex, __FUNCTION__);
|
2004-10-29 16:37:43 +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
|
|
|
return (need_switch);
|
|
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_update_best_ap
|
|
|
|
|
*
|
2004-09-09 20:02:59 +00:00
|
|
|
* Recalculate the "best" access point we should be associating with. This
|
|
|
|
|
* function may disrupt the current connection, so it should be called only
|
|
|
|
|
* when necessary, ie when the current access point is no longer in range
|
|
|
|
|
* or is for some other reason invalid and should no longer be used.
|
2004-08-05 18:54:29 +00:00
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
void nm_device_update_best_ap (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
NMAccessPointList *ap_list;
|
|
|
|
|
NMAPListIter *iter;
|
2004-10-29 16:37:43 +00:00
|
|
|
NMAccessPoint *scan_ap = NULL;
|
2004-08-05 18:54:29 +00:00
|
|
|
NMAccessPoint *best_ap = NULL;
|
2004-09-08 18:14:42 +00:00
|
|
|
NMAccessPoint *trusted_best_ap = NULL;
|
|
|
|
|
NMAccessPoint *untrusted_best_ap = NULL;
|
|
|
|
|
GTimeVal trusted_latest_timestamp = {0, 0};
|
|
|
|
|
GTimeVal untrusted_latest_timestamp = {0, 0};
|
2004-08-05 18:54:29 +00:00
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
g_return_if_fail (nm_device_is_wireless (dev));
|
|
|
|
|
|
|
|
|
|
if (!(ap_list = nm_device_ap_list_get (dev)))
|
|
|
|
|
return;
|
|
|
|
|
|
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
|
|
|
/* Iterate over the device's ap list to make sure the current
|
|
|
|
|
* "best" ap is still in the device's ap list (so that if its
|
|
|
|
|
* not, we can "unfreeze" the best ap if its been frozen already).
|
|
|
|
|
* If it is, we don't change the best ap here.
|
|
|
|
|
*/
|
2004-10-11 21:32:19 +00:00
|
|
|
if (nm_device_is_best_ap_frozen (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
|
|
|
{
|
2004-10-29 16:37:43 +00:00
|
|
|
best_ap = nm_device_get_best_ap (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
|
|
|
|
|
|
|
|
/* If its in the device's ap list still, don't change the
|
|
|
|
|
* best ap, since its frozen.
|
|
|
|
|
*/
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_lock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
2004-10-29 16:37:43 +00:00
|
|
|
if (best_ap)
|
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
|
|
|
{
|
2004-10-29 16:37:43 +00:00
|
|
|
char *essid = nm_ap_get_essid (best_ap);
|
2004-12-17 17:16:22 +00:00
|
|
|
/* Two reasons to keep the current best_ap:
|
|
|
|
|
* 1) Its still valid and we see it in our scan data
|
|
|
|
|
* 2) Its an ad-hoc network that we've created (and therefore its not in our scan data)
|
|
|
|
|
*/
|
|
|
|
|
if ( ( !nm_ap_list_get_ap_by_essid (dev->app_data->invalid_ap_list, essid)
|
|
|
|
|
&& nm_device_ap_list_get_ap_by_essid (dev, essid))
|
|
|
|
|
|| nm_ap_get_user_created (best_ap))
|
2004-10-29 16:37:43 +00:00
|
|
|
{
|
|
|
|
|
nm_ap_unref (best_ap);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
2004-10-29 16:37:43 +00:00
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
nm_ap_unref (best_ap);
|
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
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Otherwise, its gone away and we don't care about it anymore */
|
|
|
|
|
nm_device_unfreeze_best_ap (dev);
|
2005-02-02 21:49:14 +00:00
|
|
|
nm_unlock_mutex (dev->options.wireless.best_ap_mutex, __FUNCTION__);
|
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)))
|
|
|
|
|
return;
|
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
|
|
|
|
2004-09-08 18:14:42 +00:00
|
|
|
if (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
|
|
|
}
|
2004-09-08 18:14:42 +00:00
|
|
|
else if (!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
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
nm_device_set_best_ap (dev, best_ap);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
typedef struct NMDeviceForceData
|
2004-10-13 20:57:23 +00:00
|
|
|
{
|
2005-01-21 19:32:08 +00:00
|
|
|
NMDevice *dev;
|
|
|
|
|
const char *net;
|
|
|
|
|
const char *key;
|
|
|
|
|
NMEncKeyType key_type;
|
|
|
|
|
} NMDeviceForceData;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
static gboolean nm_device_wireless_force_use (NMDevice *dev, const char *essid, const char *key, NMEncKeyType key_type)
|
|
|
|
|
{
|
|
|
|
|
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
|
|
|
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
2005-01-21 19:32:08 +00:00
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
2004-10-13 20:57:23 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (!essid)
|
|
|
|
|
return FALSE;
|
2004-10-13 20:57:23 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
syslog (LOG_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
|
|
|
}
|
|
|
|
|
|
2005-01-21 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
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_device_set_best_ap (dev, ap);
|
|
|
|
|
nm_device_freeze_best_ap (dev);
|
2004-10-27 19:02:07 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
return TRUE;
|
2004-10-13 20:57:23 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
gboolean nm_device_wired_force_use (NMDevice *dev)
|
2004-10-14 20:31:35 +00:00
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
return TRUE;
|
|
|
|
|
}
|
2005-01-10 06:41:02 +00:00
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
gboolean nm_device_force_use (gpointer user_data)
|
|
|
|
|
{
|
|
|
|
|
NMDeviceForceData *cb_data = (NMDeviceForceData *)user_data;
|
|
|
|
|
NMData *app_data = NULL;
|
|
|
|
|
gboolean success = FALSE;
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
g_return_val_if_fail (cb_data != NULL, FALSE);
|
2004-11-08 19:00:04 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (!cb_data->dev || !cb_data->dev->app_data)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
app_data = cb_data->dev->app_data;
|
|
|
|
|
if (nm_device_is_wireless (cb_data->dev))
|
|
|
|
|
success = nm_device_wireless_force_use (cb_data->dev, cb_data->net, cb_data->key, cb_data->key_type);
|
|
|
|
|
else if (nm_device_is_wired (cb_data->dev))
|
|
|
|
|
success = nm_device_wired_force_use (cb_data->dev);
|
2005-01-10 06:41:02 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
if (success)
|
|
|
|
|
nm_policy_schedule_device_switch (cb_data->dev, cb_data->dev->app_data);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
/* Function that scheduled us must ref the device */
|
|
|
|
|
nm_device_unref (cb_data->dev);
|
2004-10-14 20:31:35 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
app_data->forcing_device = FALSE;
|
|
|
|
|
g_free (cb_data);
|
|
|
|
|
return FALSE;
|
|
|
|
|
}
|
2004-11-08 19:00:04 +00:00
|
|
|
|
|
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
void nm_device_schedule_force_use (NMDevice *dev, const char *network, const char *key, NMEncKeyType key_type)
|
|
|
|
|
{
|
|
|
|
|
NMDeviceForceData *cb_data;
|
|
|
|
|
GSource *source;
|
|
|
|
|
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data != NULL);
|
|
|
|
|
g_return_if_fail (dev->app_data->main_context != NULL);
|
|
|
|
|
|
|
|
|
|
cb_data = g_malloc0 (sizeof (NMDeviceForceData));
|
|
|
|
|
cb_data->dev = dev;
|
|
|
|
|
cb_data->net = network ? g_strdup (network) : NULL;
|
|
|
|
|
cb_data->key = key ? g_strdup (key) : NULL;
|
|
|
|
|
cb_data->key_type = key_type;
|
2004-10-14 20:31:35 +00:00
|
|
|
|
2005-01-21 19:32:08 +00:00
|
|
|
source = g_idle_source_new ();
|
|
|
|
|
g_source_set_callback (source, nm_device_force_use, cb_data, NULL);
|
|
|
|
|
g_source_attach (source, dev->context);
|
|
|
|
|
g_source_unref (source);
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* nm_device_do_pseudo_scan
|
|
|
|
|
*
|
|
|
|
|
* Brute-force the allowed access point list to find one that works, if any.
|
|
|
|
|
*
|
|
|
|
|
* FIXME
|
|
|
|
|
* There's probably a better way to do the non-scanning access point discovery
|
|
|
|
|
* than brute forcing it like this, but that makes the state machine here oh so
|
|
|
|
|
* much more complicated.
|
|
|
|
|
*/
|
|
|
|
|
static void nm_device_do_pseudo_scan (NMDevice *dev)
|
|
|
|
|
{
|
2004-08-05 18:54:29 +00:00
|
|
|
NMAPListIter *iter;
|
|
|
|
|
NMAccessPoint *ap;
|
|
|
|
|
|
2004-07-06 01:34:10 +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 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 shouldn't get here since we fake the AP list earlier */
|
|
|
|
|
g_return_if_fail (!dev->test_device);
|
|
|
|
|
|
2004-07-06 01:34:10 +00:00
|
|
|
nm_device_ref (dev);
|
|
|
|
|
|
2004-10-01 03:05:16 +00:00
|
|
|
if (!(iter = nm_ap_list_iter_new (dev->app_data->allowed_ap_list)))
|
2004-08-05 18:54:29 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
nm_device_set_essid (dev, "");
|
|
|
|
|
while ((ap = nm_ap_list_iter_next (iter)))
|
2004-07-06 01:34:10 +00:00
|
|
|
{
|
2004-08-05 18:54:29 +00:00
|
|
|
gboolean valid = FALSE;
|
|
|
|
|
struct ether_addr save_ap_addr;
|
|
|
|
|
struct ether_addr cur_ap_addr;
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (!nm_device_is_up (dev));
|
|
|
|
|
nm_device_bring_up (dev);
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
/* Save the MAC address */
|
|
|
|
|
nm_device_get_ap_address (dev, &save_ap_addr);
|
2004-07-06 01:34:10 +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 (nm_ap_get_enc_key_source (ap))
|
2004-10-08 05:22:17 +00:00
|
|
|
{
|
|
|
|
|
char *hashed_key = nm_ap_get_enc_key_hashed (ap);
|
2004-11-03 07:47:32 +00:00
|
|
|
nm_device_set_enc_key (dev, hashed_key, NM_DEVICE_AUTH_METHOD_SHARED_KEY);
|
2004-10-08 05:22:17 +00:00
|
|
|
g_free (hashed_key);
|
|
|
|
|
}
|
2004-08-05 18:54:29 +00:00
|
|
|
else
|
2004-11-03 07:47:32 +00:00
|
|
|
nm_device_set_enc_key (dev, NULL, NM_DEVICE_AUTH_METHOD_NONE);
|
2004-12-16 04:24:20 +00:00
|
|
|
nm_device_set_essid (dev, nm_ap_get_essid (ap));
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
/* Wait a bit for association */
|
2004-12-16 04:24:20 +00:00
|
|
|
g_usleep (G_USEC_PER_SEC * nm_device_get_association_pause_value (dev));
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
/* Do we have a valid MAC address? */
|
|
|
|
|
nm_device_get_ap_address (dev, &cur_ap_addr);
|
|
|
|
|
valid = nm_ethernet_address_is_valid (&cur_ap_addr);
|
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
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
/* If the ap address we had before, and the ap address we
|
|
|
|
|
* have now, are the same, AP is invalid. Certain cards (orinoco)
|
|
|
|
|
* will let the essid change, but the the card won't actually de-associate
|
|
|
|
|
* from the previous access point if it can't associate with the new one
|
|
|
|
|
* (ie signal too weak, etc).
|
|
|
|
|
*/
|
|
|
|
|
if (valid && (memcmp (&save_ap_addr, &cur_ap_addr, sizeof (struct ether_addr)) == 0))
|
|
|
|
|
valid = FALSE;
|
2004-07-06 01:34:10 +00:00
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
if (valid)
|
|
|
|
|
{
|
2004-08-23 19:20:49 +00:00
|
|
|
syslog(LOG_INFO, "%s: setting AP '%s' best", nm_device_get_iface (dev), nm_ap_get_essid (ap));
|
2004-08-05 18:54:29 +00:00
|
|
|
|
|
|
|
|
nm_device_set_best_ap (dev, ap);
|
2005-01-21 19:32:08 +00:00
|
|
|
nm_policy_schedule_state_update (dev->app_data);
|
2004-08-05 18:54:29 +00:00
|
|
|
break;
|
|
|
|
|
}
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
2004-08-05 18:54:29 +00:00
|
|
|
nm_ap_list_iter_free (iter);
|
2004-07-06 01:34:10 +00:00
|
|
|
nm_device_unref (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_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
|
|
|
|
|
|
|
|
|
|
int i;
|
|
|
|
|
NMAccessPointList *old_ap_list = nm_device_ap_list_get (dev);
|
|
|
|
|
|
|
|
|
|
char *fake_essids[NUM_FAKE_APS] = { "green", "bay", "packers", "rule" };
|
|
|
|
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (dev == dev->app_data->active_device)
|
|
|
|
|
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
|
|
|
{
|
2005-01-18 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
|
|
|
GSource *wscan_source;
|
|
|
|
|
guint wscan_source_id;
|
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));
|
|
|
|
|
|
|
|
|
|
wscan_source = g_timeout_source_new (dev->options.wireless.scan_interval * 1000);
|
|
|
|
|
g_source_set_callback (wscan_source, nm_device_wireless_scan, dev, NULL);
|
|
|
|
|
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;
|
|
|
|
|
gboolean have_blank_essids = FALSE;
|
|
|
|
|
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
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Test devices get their info faked */
|
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)
|
|
|
|
|
{
|
|
|
|
|
nm_device_fake_ap_list (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
|
|
|
return 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
|
|
|
}
|
|
|
|
|
|
2005-01-07 18:07:06 +00:00
|
|
|
/* Devices that don't support scanning have their pseudo-scanning done in
|
|
|
|
|
* the main thread anyway.
|
|
|
|
|
*/
|
2005-02-10 04:39:40 +00:00
|
|
|
if (!nm_device_get_supports_wireless_scan (dev))
|
2005-01-07 18:07:06 +00:00
|
|
|
{
|
|
|
|
|
nm_device_do_pseudo_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
|
|
|
return FALSE;
|
2005-01-07 18:07:06 +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
|
|
|
{
|
2005-02-27 06:11:16 +00:00
|
|
|
/* Handle dbus signals that we need to broadcast when the AP is added to the list or changes
|
|
|
|
|
* strength.
|
|
|
|
|
*/
|
|
|
|
|
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
|
|
|
{
|
2005-01-18 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
|
|
|
NMAccessPoint *outdated_ap;
|
|
|
|
|
GSList *outdated_list = NULL;
|
2005-02-10 04:39:40 +00:00
|
|
|
GSList *elt;
|
2005-01-29 23:34:31 +00:00
|
|
|
NMAccessPoint *best_ap = nm_device_get_best_ap (dev);
|
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)
|
2005-02-27 06:11:16 +00:00
|
|
|
&& (best_ap && (nm_null_safe_strcmp (nm_ap_get_essid (best_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);
|
|
|
|
|
|
2005-01-29 23:34:31 +00:00
|
|
|
/* nm_device_get_best_ap() refs the ap */
|
|
|
|
|
if (best_ap)
|
|
|
|
|
nm_ap_unref (best_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
|
|
|
/* 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
|
|
|
|
2005-01-18 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 the list changed, decrease our wireless scanning interval */
|
|
|
|
|
if (list_changed)
|
|
|
|
|
dev->options.wireless.scan_interval = 20;
|
|
|
|
|
else
|
|
|
|
|
dev->options.wireless.scan_interval = MIN (60, dev->options.wireless.scan_interval + 10);
|
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
|
|
|
return FALSE;
|
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
|
|
|
{
|
2005-01-18 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
|
|
|
NMDevice *dev = (NMDevice *)(user_data);
|
|
|
|
|
int sk;
|
|
|
|
|
NMWirelessScanResults *scan_results = 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
|
|
|
|
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_val_if_fail (dev != NULL, FALSE);
|
|
|
|
|
g_return_val_if_fail (dev->app_data != NULL, FALSE);
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/* We don't scan on test devices or devices that don't have scanning support */
|
2005-02-10 04:39:40 +00:00
|
|
|
if (dev->test_device || !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
|
|
|
return FALSE;
|
2005-01-07 18:07:06 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
/* Just reschedule ourselves if scanning or all wireless is disabled */
|
|
|
|
|
if ( (dev->app_data->scanning_enabled == FALSE)
|
|
|
|
|
|| (dev->app_data->wireless_enabled == FALSE))
|
|
|
|
|
{
|
|
|
|
|
dev->options.wireless.scan_interval = 10;
|
|
|
|
|
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__))
|
|
|
|
|
{
|
|
|
|
|
/* Device must be up before we can scan */
|
|
|
|
|
if (!nm_device_is_up (dev))
|
|
|
|
|
nm_device_bring_up (dev);
|
|
|
|
|
g_usleep (G_USEC_PER_SEC);
|
|
|
|
|
|
|
|
|
|
if ((sk = iw_sockets_open ()) >= 0)
|
|
|
|
|
{
|
|
|
|
|
int err;
|
|
|
|
|
NMNetworkMode orig_mode = NETWORK_MODE_INFRA;
|
2005-01-10 06:41:02 +00:00
|
|
|
double orig_freq = 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
|
|
|
int orig_rate = 0;
|
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));
|
|
|
|
|
err = iw_scan (sk, (char *)nm_device_get_iface (dev), WIRELESS_EXT, &(scan_results->scan_head));
|
2005-01-07 18:07:06 +00:00
|
|
|
if ((err == -1) && (errno == ENODATA))
|
|
|
|
|
{
|
|
|
|
|
/* 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
|
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
|
|
|
* give up.
|
2005-01-07 18:07:06 +00:00
|
|
|
*/
|
|
|
|
|
g_usleep ((G_USEC_PER_SEC * nm_device_get_association_pause_value (dev)) / 2);
|
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
|
|
|
err = iw_scan (sk, (char *)nm_device_get_iface (dev), WIRELESS_EXT, &(scan_results->scan_head));
|
2005-01-07 18:07:06 +00:00
|
|
|
if (err == -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
|
|
|
scan_results->scan_head.result = NULL;
|
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
|
|
|
else if ((err == -1) && (errno == ETIME))
|
2005-01-21 19:32:08 +00:00
|
|
|
syslog (LOG_ERR, "Warning: the wireless card (%s) requires too much time for scans. Its driver needs to be fixed.", 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
|
|
|
|
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
|
|
|
|
|
|
|
|
close (sk);
|
|
|
|
|
}
|
|
|
|
|
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
|
|
|
|
|
|
|
|
/* 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 ();
|
|
|
|
|
|
|
|
|
|
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);
|
|
|
|
|
}
|
|
|
|
|
|
2005-02-07 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 */
|
|
|
|
|
nm_device_wireless_schedule_scan (dev);
|
|
|
|
|
|
|
|
|
|
return FALSE;
|
2004-07-06 01:34:10 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
2004-09-08 18:14:42 +00:00
|
|
|
/* System config data accessors */
|
|
|
|
|
|
|
|
|
|
gboolean nm_device_config_get_use_dhcp (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
|
|
|
|
|
return (dev->config_info.use_dhcp);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_config_set_use_dhcp (NMDevice *dev, gboolean use_dhcp)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->config_info.use_dhcp = use_dhcp;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
guint32 nm_device_config_get_ip4_address (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
|
|
|
|
|
return (dev->config_info.ip4_address);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_config_set_ip4_address (NMDevice *dev, guint32 addr)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->config_info.ip4_address = addr;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
guint32 nm_device_config_get_ip4_gateway (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
|
|
|
|
|
return (dev->config_info.ip4_gateway);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_config_set_ip4_gateway (NMDevice *dev, guint32 gateway)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->config_info.ip4_gateway = gateway;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
guint32 nm_device_config_get_ip4_netmask (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
|
|
|
|
|
return (dev->config_info.ip4_netmask);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_config_set_ip4_netmask (NMDevice *dev, guint32 netmask)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->config_info.ip4_netmask = netmask;
|
|
|
|
|
}
|
2004-10-15 00:06:14 +00:00
|
|
|
guint32 nm_device_config_get_ip4_broadcast (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
g_return_val_if_fail (dev != NULL, 0);
|
|
|
|
|
|
|
|
|
|
return (dev->config_info.ip4_broadcast);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void nm_device_config_set_ip4_broadcast (NMDevice *dev, guint32 broadcast)
|
|
|
|
|
{
|
|
|
|
|
g_return_if_fail (dev != NULL);
|
|
|
|
|
|
|
|
|
|
dev->config_info.ip4_broadcast = broadcast;
|
|
|
|
|
}
|
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
|
|
|
{
|
2005-02-07 23:04:05 +00:00
|
|
|
int sk;
|
|
|
|
|
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
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
if ((sk = socket (AF_INET, SOCK_DGRAM, 0)) < 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-02-07 23:04:05 +00:00
|
|
|
syslog (LOG_ERR, "cannot open socket on interface %s for MII detect; errno=%d", nm_device_get_iface (dev), errno);
|
|
|
|
|
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;
|
|
|
|
|
if (ioctl(sk, SIOCETHTOOL, &ifr) == -1)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
supports_ethtool = TRUE;
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
close (sk);
|
|
|
|
|
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>
|
|
|
|
|
|
|
|
|
|
static int mdio_read (int 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;
|
|
|
|
|
|
|
|
|
|
g_return_val_if_fail (sk < 0, -1);
|
|
|
|
|
g_return_val_if_fail (ifr != NULL, -1);
|
|
|
|
|
|
|
|
|
|
mii = (struct mii_ioctl_data *) &(ifr->ifr_data);
|
|
|
|
|
mii->reg_num = location;
|
|
|
|
|
|
|
|
|
|
if (ioctl (sk, SIOCGMIIREG, &ifr) < 0)
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
return (mii->val_out);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static gboolean supports_mii_carrier_detect (NMDevice *dev)
|
|
|
|
|
{
|
|
|
|
|
int 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;
|
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
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
if ((sk = socket (AF_INET, SOCK_DGRAM, 0)) < 0)
|
2004-06-24 14:18:37 +00:00
|
|
|
{
|
2005-02-07 23:04:05 +00:00
|
|
|
syslog (LOG_ERR, "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);
|
|
|
|
|
if (ioctl(sk, SIOCGMIIPHY, &ifr) < 0)
|
|
|
|
|
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 */
|
|
|
|
|
bmsr = mdio_read(sk, &ifr, MII_BMSR);
|
|
|
|
|
supports_mii = (bmsr != -1) ? TRUE : FALSE;
|
2004-06-24 14:18:37 +00:00
|
|
|
|
2005-02-07 23:04:05 +00:00
|
|
|
out:
|
|
|
|
|
close (sk);
|
|
|
|
|
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);
|
|
|
|
|
}
|