[67] in linux-scsi channel archive
Guaranteed kernel panic
daemon@ATHENA.MIT.EDU (gpg109)
Sat Feb 11 08:49:22 1995
To: linux-scsi@vger.rutgers.edu
Date: Sat, 11 Feb 1995 23:55:00 +1100 (EST)
From: "gpg109" <gpg109@rsphy1.anu.edu.au>
Cc: linux-activists@niksula.hut.fi (linux)
Ok - don't try this at home (do it at work instead ;-) And
remount your filesystem(s) as ro first. (You have been warned)
In one vc, cd to a dir that is on your SCSI disk (preferrably large) and do:
tar -cf - * > /dev/null
Now while that is going, go to another vc, and do a:
scsiinfo -a /dev/sda | less
Result: GPF, killed IRQ handler, kfree of non-kmalloc, swapper killed, panic.
You can still do <Shift>-<Scrl-Lck> etc. and change vc's but that is
about it.
It is repeatable as the sunrise. It blows up in requeue_sd_request,
which means host adapters that can't queue will not be affected. This
was with the BusLogic driver of 1.1.90 -- haven't tried to see if the
Adaptec driver does the same. "scsiinfo" was version 1.4, if that
matters. The intense disk activity caused by the "tar" is required, else
nothing happens. I won't bother putting the stack, call trace, etc. up
here, as it is easy to get on demand. :-)
Paul.
