Re: [RFT] Intel 3945abg wireless driver (wpi)

On Tue, 2007-11-06 at 18:31 +0100, Frank Staals wrote:
> Sam Leffler wrote:
> > Frank Staals wrote:
> >> Benjamin Close wrote:
> >>> Frank Staals wrote:
> >> <snip>
> >>>>
> >>>>
> >>>>
> >>>> Everything works fine with the connection itself. Allthough 
> >>>> sometimes when switching from tty9 to tty0 and back the system 
> >>>> locks up. I've had it before when switching  from tty1 to tty0. 
> >>>> Anyone with the same problems ?
> >>>>
> >>>> Anyway; Great work on the driver so far :D
> >>>>
> >>> I've similar issues and believe it might be due to the amount of 
> >>> kernel printfs that are happening. Can you sysctl debug.wpi=0  and 
> >>> see if the problem still exists?
> >>> By chance are you using ZFS? I caught a memory modified after free 
> >>> panic in zfs the other day pid was from syslogd. I'm trying to work 
> >>> out if it's related.
> >>>
> >>> Cheers,
> >>>    Benjamin
> >>>
> >> When setting debug.wpi to 0 it seems like the problem is gone. I'm 
> >> not using ZFS by the way.  I did have a problem connecting to the AP 
> >> at my university though; the driver wouldn't assosicate whatever I 
> >> tried. I didn't have a chance to do some extensive testing though. It 
> >> might be because of the WEP+wpa_supplicant + ca certificate method 
> >> that is required to authenticate. Anyway I'll let it know if there is 
> >> an actual problem with the driver itself
> >
> > Rule of thumb in debugging wireless issues (and most others for that 
> > matter): simplify your config if at all possible.  In this case try 
> > checking things out w/o wpa_supplicant (i.e. no crypto).
> >
> > When debugging, start at the top and work your way down to rule out 
> > problems at each layer.  In this case collect a wpa_supplicant debug 
> > log first, then check for issues at the net80211 layer (wlandebug, 
> > wlanstats, etc.), then finally check at the driver level.
> >
> >    Sam
> >
> For at home wpi is working fine, and I indeed used the method you 
> described above. Unfortunately I don't have the luxury of that extended 
> testing methods at my university since I can't turn off encryption and 
> the weird authentication methods there ;)
> 
Hello,

Here it works for a while with debug enabled even with wpa_supplicant,
but after a few minutes the console output stops working, but the
keyboard is still handled ok. The Caps lock led blinks when the key is
pressed and ctrl-alt-del can reboot the box.

When debug is disabled and wpa_supplicant is on, it crashes after a few
seconds and gets me into the debugger.
I'll try later with unsecured networks and see if that happens again.

Cristi

-- 
Cristi Magherusan,
Universitatea Tehnica din Cluj - Napoca
Centrul de Comunicatii "Pusztai Kalman"
Tel. 0264/401247  http://cc.utcluj.ro