[ltp] T40: airo_mpi death under high load? (fwd)

[linux-thinkpad@linux-thinkpad.org]

I've been happily using my new T40 with the airo_mpi module thanks
to the good work of Fabrice Bellet and the airo-linux people (and
hints on this list!).  I'm however experiencing problems at high
network load.

Under normal use, the driver works fine.  When however I try and copy
a large amount of data across the wireless link to an SMB or NFS share,
I start getting the following in syslog and on all VT consoles:

Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: cmd= 21
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: status= 817
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Rsp0= 11f2
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Rsp1= 1b1
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Rsp2= 8ef1
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Max tries exceeded when issueing command 21
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Hung command reg = 8000
Aug 10 12:13:33 teapot kernel: airo_mpi: airo_kick
Aug 10 12:13:33 teapot kernel: airo_mpi: airo_clear: Could not clear command busy
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Could not clear command register
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: Was busy too long
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: cmd= 21
Aug 10 12:13:33 teapot kernel: airo_mpi: issuecommand_nolock: status= 2657
Aug 10 12:13:34 teapot kernel: airo_mpi: issuecommand_nolock: Rsp0= 81bf
Aug 10 12:13:34 teapot kernel: airo_mpi: issuecommand_nolock: Rsp1= a016
Aug 10 12:13:35 teapot kernel: airo_mpi: issuecommand_nolock: Rsp2= cf79
Aug 10 12:13:35 teapot kernel: airo_mpi: issuecommand_nolock: Was busy too long

The Rsp? and status hex entries thereafter appear the same in repeated
messages: I don't pretend to understand what this means!

What happens is that these messages are repeatedly blasted to syslog
(maybe 10-20 a second) and the whole machine more or less freezes.
If you're very quick/patient and can stop the network and rmmod
airo_mpi the messages stop: but then if you try and modprobe airo_mpi
again usually the system reports that the device is no longer present.
I don't know what to do at this point other than reboot.

Has anyone else seen this?  Any clues whatsoever to look at to fix this?

Some details: I'm using 2.4.21-ac4 and have followed Fabrice's
instructions at:

http://bellet.info/~bellet/laptop/README.airo_mpi

to the letter.  Using the current airo-net CVS files relevant and
have just got the latest.  I've tried this with airo_mpi.c-20030719
and the latest airo_mpi.c-20030802 from Fabrice's site.

Card operating in Ad-Hoc mode with WEP turned on.

Any suggestions?  Specifically, is it possible I'm looking at
intermittent hardware problems here in the card itself?