Tighten up handling of wireless devices that don't support wireless
scanning (ie, Orinoco). Due to restructuring of code, these devices
hadn't been doing pseudo-scanning for a while either and would just
spin waiting for an access point. They are now manual devices where
the user must choose the access point from the menu every time. All
"allowed" access points are listed in the applet's menu regardless
of whether or not they can be seen by the card, since it can't scan
anyway.
* src/NetworkManager.c
- (nm_wireless_link_state_handle): new function, but only update
the "best" ap for non-scanning devices when its not activating,
and when no device is being forced on the card
- (nm_link_state_monitor): split wireless link state handling out
into separate function
* src/NetworkManagerDevice.c
- (nm_device_copy_allowed_to_dev_list): new function
- (nm_device_new): populate non-scanning cards' AP lists with
access points from the "allowed" list
- (nm_device_new): don't start a scanning timeout for devices that
can't scan
- (nm_device_activation_schedule_finish): new parameter, should be
the AP that failed to be connected to, pass it on to the
activation finish function in NetworkManagerPolicy.c
- (nm_device_activate_wireless): don't ever try to get a new AP
for non-scanning devices, just fail. The user must choose
a new access point manually.
- (nm_device_activate): grab the AP that failed connection and
pass it on
- (nm_device_update_best_ap): Clear the best AP if we don't have
a link to it, user must manually choose a new one
- (nm_device_do_pseudo_scan): remove function
- (nm_device_wireless_process_scan_results): remove bits for non-
scanning cards since they never get here
- (nm_device_wireless_scan): remove bits for non-scanning devices,
and fake the scan list for test devices a bit earlier
* src/NetworkManagerPolicy.c
- (nm_policy_activation_finish): use the failed_ap that we get
passed rather than getting the best_ap from the card, which
may have changed since we were scheduled
- (nm_policy_allowed_ap_list_update): for non-scanning devices,
update their scan list directly from the allowed list when
we get updates to the allowed list from NetworkManagerInfo
* src/NetworkManagerPolicy.h
- New member for failed access point in NMActivationResult
-------------------------------------
Driver Notification patch: notifies the user when their driver
sucks. Gives them the option to ignore further insertions
of the card that has the sucky driver.
* NetworkManager.h
- Remove the SEMI_SUPPORTED member from the NMDriverSupportLevel
enum and replace it with NO_CARRIER_DETECT and
NO_WIRELESS_SCAN
* panel-applet/NMWirelessApplet.[ch]
- Merge essid.glade -> wireless-applet.glade
- Implement the "Your driver sucks" notification dialog
* panel-applet/NMWirelessAppletDbus.c
- Change stuff from getSupportsCarrierDetect->getDriverSupportLevel
- Grab hardware address for each device from NM too
- Check whether the driver for each device sucks or not whenever
a new device is noticed
* panel-applet/NMWirelessAppletOtherNetworkDialog.c
- Deal with stuff being in wireless-applet.glade now rather than essid.glade
* src/NetworkManager.c
- Fix a double-unref on device removal
* src/NetworkManagerUtils.c
- Set appropriate driver support level on a device that doesn't
support scanning or carrier detection
* src/nm-dbus-device.c
- New "getHWAddress" dbus method on devices
- getSupportsCarrierDetect -> getDriverSupportLevel
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@534 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
* configure.in
- Check DBUS version in configure, and set the C macros
DBUS_VERSION_[MAJOR,MINOR,MICRO]
* info-daemon/NetworkManagerInfoDbus.c
- Remove #if 0-d section of code that quit NMI if NM went away.
* panel-applet/NMWirelessAppletDbus.c
- Trap the "ServiceOwnerChanged" signal that's new in dbus-0.23
* src/NetworkManager.c
src/NetworkManagerMain.h
src/NetworkManagerDbus.c
- Trap the "ServiceOwnerChanged" signal that's new in dbus-0.23
- Make updating of our Allowed Wireless Network lists from NMI
an idle function in the main thread now, with a high priority.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@398 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
* Cache last-known-good wireless authentication method in
NetworkManagerInfo, and use that method first during
wireless device activation. Should speed up devices that
need Shared Key authentication method since Open System is
now the default.
* Remove the hack to not do full activation on wired connections
that are active when we launch, it causes too many problems
with name resolution and was a hack in the first place.
* Re-work wireless device activation again somewhat to have a
clearer chain of events and to use last-known-good
authentication method of the access point. Also provide
better status throughout activation to ensure the applet
can tell the user exactly what's going on.
* Remove the "find wireless network" code and now simply attempt
to activate with that access point. This reduces the delay
between selecting "Other wireless Network" and actually
connecting to that network.
* Correctly stop the device's worker thread when its removed.
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@388 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
* dhcpcd/client.c
- Rework the DHCP client code to be much less chatty when
it receives non-DHCP UDP packets during the DHCP run
(reported by and preliminary patches from Bill Moss)
* Move wireless scanning to a separate thread. This thread forwards the
results to the main thread when done where they are integrated into
the device's access point lists. This keeps the main thread (which
does all the DBUS communication) from being blocked for long periods
of time by wireless scanning.
* Make state modification an idle routine in the main loop, and trigger
state changes rather than polling for them.
* src/backends/NetworkManagerGentoo.c
- Fix up invalid C90 code (reported by Christoph Ruessler)
* src/NetworkManagerDevice.c
- Revert IPv6 patch for wired devices from 2004-12-22 for
router advertisements, causing problems and infinite loop
during "best" device determination due to link going up/down
(reported by Bill Moss)
Apply patch from Peter Jones
* src/NetworkManagerDevice.c
- Shortcut for link-checking for ipw2x00 cards
- Split out association check into separate routine
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@360 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
* dispatcher-daemon/NetworkManagerDispatcher.c
- A bit more descriptive state message
- Don't segfault when reading directory
* src/NetworkManager.h
- Remove NMData desired_ap member, its now
per-device rather than global
* src/NetworkManager.c
- Remove references to desired_ap
- Move the allowed AP list refresh stuff into a thread
* src/NetworkManagerDevice.c
src/NetworkManagerDevice.h
- Each wireless device now has a "best ap"
- Make device activate/deactivate functions per-device
- Make wireless scanning per-device
- Add IPv4 address discover functions, stub IPv6 ones
- Move ethernet address validation functions to NetworkManagerUtils.c
- Add wireless access point accessor function
- Get/Set functions for "best ap"
* src/NetworkManagerPolicy.c
- Move activate/deactivate stuff into NetworkManagerDevice.c, per-device
- Deal with per-device "best ap" rather than data->desired_apa
- Implement allowed access point worker thread
- Add nm_policy_essid_is_allowed() function
* src/NetworkManagerUtils.c
src/NetworkManagerUtils.h
- Add nm_ethernet_address_is_valid() function
- Add IPv4/IPv6 address get functions
* src/NetworkManagerWireless.c
src/NetworkManagerWireless.h
- Move scanning stuff into NetworkManagerDevice.c, per-device
git-svn-id: http://svn-archive.gnome.org/svn/NetworkManager/trunk@12 4912f4e0-d625-0410-9fb7-b9a5a253dbdc
