[PATCH] don't continually reschedule specific scans

[Pavel Polischouk]

Dan Williams wrote:
> On Mon, 2008-06-02 at 15:20 -0400, Dan Williams wrote:
>   
>> On Mon, 2008-06-02 at 15:01 -0400, Pavel Polischouk wrote:
>>     
>>> Dan Williams wrote:  
>>>       
>>> Tested the patch as a possible way to fix "failure scenario 1" of this 
>>> bug: https://bugzilla.redhat.com/show_bug.cgi?id=449215
>>> The situation does improve - it is more often able to break the scanning 
>>> deadlock, associate and proceed to authentication stage.
>>>       
>> You'll need to pair this patch with the ipw2200 patch I just re-posted
>> for ipw2200 direct scan queuing (attached so you don't have to find it).
>> The combination makes ipw2200 + wpa_supplicant a lot happier with hidden
>> SSIDs.
>>     
>
> I haven't tested the patch out for poor link quality like you mention in
> the bug report, but it certainly can't hurt and should probably help
> keep the AP in the supplicant's scan list, which would help the
> re-connection.
>
> Dan
>   

Tested both patches together. The number of association timeouts has 
been dramatically reduced, but not eliminated - they used to happen 
almost every time, now they happen once in a blue moon.

The patches don't fix the second problem, where the re-authentication 
fails. I'll try to compare the good and bad segments of log file and 
figure out the difference.

The poor link quality is not required to reproduce the failure, the same 
result can be achieved by killing nm-applet and reloading it.

Thanks,
Pavel