Re: [BUG] Crda oopses the system

[Maxim Levitsky <maximlevitsky@gmail.com>]

On Thu, 2009-05-21 at 17:36 -0700, Luis R. Rodriguez wrote:
> On Thu, May 21, 2009 at 5:20 PM, Maxim Levitsky <maximlevitsky@gmail.com> wrote:
> > On Tue, 2009-05-19 at 17:17 +0300, Maxim Levitsky wrote:
> >> On Thu, 2009-05-14 at 10:20 -0400, Bob Copeland wrote:
> >> > On Wed, May 13, 2009 at 8:07 PM, Luis R. Rodriguez <mcgrof@gmail.com> wrote:
> >> > >>
> >> > >> I wish I had time to set up, an ath5k in AP mode. There are rumors that
> >> > >> it works more or less now, then I could use the same beacon frame to
> >> > >> test.
> >> > >
> >> > > It does and if it doesn't its a bug.
> >> >
> >> > Just FYI, it's known to have problems if the STA uses power saving --
> >> > ath5k never updates the beacon.  I have a patch that should work in the
> >> > relevant bugzilla but I haven't been back to retest it (at first I
> >> > thought it was causing hangs, but later realized the hangs were due to
> >> > other bugs in w-t).
> >> >
> >>
> >
> > I was just able to reproduce that bug on latest wireless-testing that I
> > belive contains the patches you sent for me to test.
> 
> Latest wireless-testing *does* have those patches.
Yes I know, just to be sure.

> 
> > I reproduced it against ath5k running in AP mode at home.
> >
> > I finally made the ath5k send beacons (although not much works besides
> > this).
> >
> >
> > Note that I noticed that this bug
> 
> You can get an oops when trying to associate to your ath5k AP? And its
> easily reproducible? What driver to you use as the STA? Can you
> provide a trace?
> 
> > happens once at boot, if I set NM to
> > use system settings,
> 
> What does this mean?
Well, I can't say for sure, but it appears that if I asccociate with AP,
then I repeat this, I don't see this bug again.


> 
> > when I try to connect again it doesn't happen. Its
> > a race condition after all.
> 
> Is the oops not crashing your box? How are you able to try again? What
> does trying again mean?
Well, its not an oops on my system long ago.

I have converted the BUG_ON to printk, and so far I havent see any side
effects.
However I don't want you just to remove this check, unless it is bogus.
(Because it will bury the bug deeper).

On the other hand I see no reason to convert it to WARN_ON so it won't
oops user systems.

I am talking about 

BUG_ON(!country_ie_regdomain);
in net/wireless/reg.c



> 
>   Luis