Description of problem:
kernel panicked. The message on the console mentioned rwsem.h.
Filesystem access was lost, so no panic log was recoverable, and the
colo admin didn't get a screenshot before rebooting. :(
Version-Release number of selected component (if applicable):
First time it's happened, hope it doesn't happen again.
Steps to Reproduce:
This is our second kernel panic on this box with this kernel, but
since the panic error was different this time I filed a new bug.
Because of the previous experience, we already had
kernel-smp-2.6.9-6.16.EL installed and configured to boot into, so we
picked up that kernel on the post-panic reboot. Previous issue with
this machine is bug 150044.
I don't know what I can do about this. Are you sure that "rwsem.h" was
mentioned and not "rwsem.c" or some function whose name contains "rwsem"?
We do have a patch to fix a problem with R/W semaphores that so far only crops
up in conjunction with AIO as far as I know. Are you using AIO?
I don't know of any way to get additional info, unfortunately the tech at the
scene at the time didn't have a camera to get a screenshot of the console before
rebooting it when it happened, and we were going from his memory to describe
what happened. It's been 5 months (a 3 kernel upgrades) since then, and it
hasn't happened again, and that machine's been under pretty serious load the
entire time. :)
I would just go ahead and close this out as unreproducable or something.
Okay. I suspect this will be fixed by a patch added to RHEL4 to make
rwsems interrupt-safe, assuming it wasn't caused indirectly by some other bug.
This patch will be in RHEL-4 U2. Let us know if it happens again.