Bugzilla – Bug 303834
NetworkManager: killswitch has inverted effect on intel WLAN chipsets
Last modified: 2007-09-08 06:48:27 UTC
My network is not working any longer since the upgrade to iwlwifi instead of ipw3945. Today I started debugging it an found the following in the NetworkManager-logs. On enabling the wireless with the mechanical swich I get the following: Aug 23 14:41:00 linux-c19q NetworkManager: <info> Wireless now disabled by radio killswitch Aug 23 14:41:00 linux-c19q NetworkManager: <info> Deactivating device wlan0. On disabling the WLAN i get: Aug 23 14:42:00 linux-c19q NetworkManager: <info> Wireless now enabled by radio killswitch Aug 23 14:42:14 linux-c19q NetworkManager: <WARN> nm_device_802_11_wireless_scan(): could not trigger wireless scan on device wlan0: Network is down
Well, NetworkManager and HAL have different opinions what the kill setting actually means.
It's really confusing in HAL code and the spec: the interface for this is called: 'org.freedesktop.Hal.Device.KillSwitch' which indicates the interface would do things with and about the Killswitch, but the methodes in the interface are: 'Get/SetPower' which make no sense for the killswitch. And the spec say about this interface: This interface provides a mechanism for both querying whether a radio is on as well as turning it on and off. The following methods are available: and the methodes: * GetPower(): Returns 1 if, and only if, the power is on. * SetPower(): Set the power of the radio. In fact the name of the interface imply the something like get status of the killswitch and set the (SW)killswitch, but actually it get/set the status of the WLAN (or e.g. bluetooth) interface. We need to clear/discuss this upstream. Until we have a decision I fixed the ipw related GetPower() method to return the correct state of the WLAN instead of the state of the switch in the upstream git repo, it should be in beta3.
Created attachment 159568 [details] workaround for intel WLAN devices This is a workaround for the ipw intel WLAN network devices. This diff/patch is for x86_64, you can use it also for i386, simply change the path for the file in the patch
Thanks Danny I applied your patch to my i386 system and the problem is gone now. Sadly wireless is still not working. :-( but thats another problem.
The next problem is the kernel I think. I am writing into this report as quiet some wlan people are subscribed. On Turning of the wireless with the mechanical switch I get in /var/log/messages: Aug 23 23:09:06 linux-c19q kernel: iwl3945: Radio Frequency Kill Switch is On: Aug 23 23:09:06 linux-c19q kernel: Kill switch must be turned off for wireless networking to work. On turning on I get: Aug 23 23:10:10 linux-c19q kernel: iwl3945: ipw going down NetworkManager is never able to find any network it detects the card now. But shows just "Kein drahtloses Netzwerk gefunden" (No wireless network found).
Oops sorry: Should the above be handled in another report?
The same happens here with ipw 2100 (openSUSE 10.3 beta2) on a centrino x86. Is the workaround applied to factory? Or will it be?
This might be interesting: 1. I boot the computer with the wireless switch on; 2. (K)networkManager reports that the switch is off, and does not allow to enable the wireless; 3. I turn the switch off, and now I can enable the wireless with (k)networkmanager, as it thinks the switch is on; 4. I try to connect to an wireless network, but it can't, since the switch is off; 5. If I turn the switch on while it is trying to connect to a wireless network, it seems to ignore the fact that the switch has changed state, and I am able to have both the switch on and (k)networkmanager with the wireless enabled, and thus the wireless works.
Could someone provide for a x86_64 binary with the diff aplied, because this bug is driving me n*
(In reply to comment #9 from Fred van Zwieten) > Could someone provide for a x86_64 binary with the diff aplied, because this > bug is driving me n* This patch is against a shell script! You can just edit manually the file and change the "1" to a "0" and the "0" to a "1". You do not have to compile anything.
*** Bug 304012 has been marked as a duplicate of this bug. ***
setback Priority and Severity
Christoph, we should consider releasing an update for this one. NM wireless support is virtually non-existent in beta2 otherwise.
*** Bug 303758 has been marked as a duplicate of this bug. ***
Joe, I'd propose to just post a note to openSUSE-factory@openSUSE.org -- could you please take care of this?
Oops!! So sorry for that, didn't think it was a patch for a script for one moment. Anyway, I applied it and it's working now in one go. Great!! Thanx guys. And I agree, it's worth a small notice on the website _and_ a update to 10.3b2
submitted new (fixed) hal package to STABLE
*** Bug 304682 has been marked as a duplicate of this bug. ***
This problem is still in 10.3 beta 3.
Sorry, that doesn't mean much. The errors in the script in question (as per the attached patch) are still in beta 3.
Strange, I was severely affected by this bug, but I have no problems with it now. My kill switch if OFF (so wireless is on) and I have wireless. Sep 8 08:34:20 pegasus NetworkManager: <info> Found radio killswitch /org/freedesktop/Hal/devices/ipw_wlan_switch Sep 8 08:34:21 pegasus NetworkManager: <info> Wireless now enabled by radio killswitch