Re: WLAN disabled by state file

[Ozan Çağlayan <ozan pardus org tr>]

On 09.02.2011 18:40, Dan Williams wrote:

> In current NM 0.8.x git, if you uncheck "Enable Wireless" from the menu,
> NM writes that to the state file.  That will then be in-force until you
> re-check "Enable Wireless", even across reboots.  This takes precedence
> over rfkill because it was an explicit user choice to disable wifi.
> This fix was included in NM 0.8.2.
>
> I don't think we should be relying on rfkill here, because then it's
> simply magic what happens on reboot.  Some laptops expose multiple
> rfkill "switches" in the kernel, others expose one, some chain them
> together, etc.  It's a big mess really, and trying to rely on the kernel
> behavior here isn't going to help much.  Instead, I think it's a lot
> clearer that "If you turn off wifi it stays disabled til you turn it
> back on".  Or?

I think there are multiple issues in here. The one that I've mentioned 
in the original thread and you've commented about may be triggered by 
the bug that I've replied.

So with NM 0.8.2 on my Toshiba Portege R700 (iwlagn, exposing only a 
soft rfkill which is correctly turning on/off on keypress), when I 
sw-kill the radio, this is correctly detected by NM. The "Enable 
Wireless" gets unchecked and WirelessEnabled=false is written to the 
state file.

But when I unblock, NM detects this to some point (looking at the 
debugged outputs and the code) but doesn't update the state file and 
doesn't enable the wireless networking. So one should explicitly check 
the "enable wireless" every time after unblocking the rfkill.

I think this should be fixed. I'd like to debug more but I'm really 
getting lost in the glib/gobject mechanisms and NM code which contains a 
lot of abstraction/callback stuff really hard to follow :(


--
Pardus Linux
http://www.pardus.org.tr/eng