Re: pcmcia wireless adaptor, hw fault or kernel bug?

[Juraj Ziegler <next@trip.sk>]

Gabriele Persia wrote:
> Hi all,
>
> I'm running debian/testing on a Travelmate 525txv.
> When downloading large files, the prismGT (prism54) cardbus card 
> "freezes" (and sometime hangs the system).
>
>
> What works:
> - system boots fine with the card plugged;
> - when I plug [unplug] the card, hotplug brings up [remove] the interface (as 
> expected);
> - browsing the lan/Internet;
> - UPLOADING files of ANY size;
> - downloading of "small" files;
>
> What doesn't work:
> - downloading of big file (>3Mb)
> (buffer-to-disk problem?)
>
> Tried with many combination of
> - kernels (2.4.27, 2.6.8, 2.6.10)
> - firmware version (isl3890 v1.0.3.0 and v1.0.4.3)
> - driver settings (txpower, channels, key on/off)
> ...they all shows the same behaviour.
>
> Mepis3.3 and Knoppix3.7 don't work either.
>
> If I unplug the card, the frozen system resumes working.
>
>
> Follows a kernel log I saw once, when the system freezed:
> May  9 00:27:12 localhost kernel:  [__report_bad_irq+42/160] 
> __report_bad_irq+0x2a/0xa0
> May  9 00:27:12 localhost kernel:  [handle_IRQ_event+48/112] 
> handle_IRQ_event+0x30/0x70
> May  9 00:27:12 localhost kernel:  [note_interrupt+112/176] 
> note_interrupt+0x70/0xb0
> May  9 00:27:12 localhost kernel:  [__do_IRQ+296/320] __do_IRQ+0x128/0x140
> May  9 00:27:12 localhost kernel:  [do_IRQ+25/48] do_IRQ+0x19/0x30
> May  9 00:27:12 localhost kernel:  [common_interrupt+26/32] 
> common_interrupt+0x1a/0x20
> May  9 00:27:12 localhost kernel:  [__do_softirq+46/144] 
> __do_softirq+0x2e/0x90
> May  9 00:27:12 localhost kernel:  [do_softirq+39/48] do_softirq+0x27/0x30
> May  9 00:27:12 localhost kernel:  [irq_exit+53/64] irq_exit+0x35/0x40
> May  9 00:27:12 localhost kernel:  [do_IRQ+30/48] do_IRQ+0x1e/0x30
> May  9 00:27:12 localhost kernel:  [common_interrupt+26/32] 
> common_interrupt+0x1a/0x20
> May  9 00:27:12 localhost kernel:  [pg0+273215471/1069863936] 
> unix_poll+0x2f/0xa0 [unix]
> May  9 00:27:12 localhost kernel:  [tty_poll+137/192] tty_poll+0x89/0xc0
> May  9 00:27:12 localhost kernel:  [sock_poll+41/64] sock_poll+0x29/0x40
> May  9 00:27:12 localhost kernel:  [do_select+607/720] do_select+0x25f/0x2d0
> May  9 00:27:12 localhost kernel:  [__pollwait+0/208] __pollwait+0x0/0xd0
> May  9 00:27:12 localhost kernel:  [sys_select+731/1264] 
> sys_select+0x2db/0x4f0
> May  9 00:27:12 localhost kernel:  [syscall_call+7/11] syscall_call+0x7/0xb

Interesting. You at least get something in the log.
I have na zcom xg-300 card, which also uses the prism54. My experience is, 
that during big downloads, the transfer stalled for cca. 30 seconds. First 
I suspected the card, but the logs were empty and it came back to life by 
itself. Then I noticed, that the failing part was my AP - Linksys WRT54G.

With not enough time on my hands and the AP being in different place than 
me at the moment, I have not investigated this issue further.

Things I'm planning to check out.
* upgrade the  firmware in the AP
* try the newest drivers for the card, from prism54.org
* use a htb shaper to limit the data flow through the card.

I'm open to suggestions from the list :)


j.

--
__________________________________________________________________________
"We did a risk management review.  We concluded that there was no risk of
any management." -- Dilbert
:wq