From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; Q312461)
Description of problem:
I would like to alert users to a critical problem with the Intel SRCMR RAID
Controller under RedHat 7.3 on a Intel SCB2 server board.
Upon rebooting or halting the system, the gdth drivers crash under a seg fault
when executing "halt -i -b" or "reboot -i -b" - the EIP trace leads back to
scsi_mod, which in turn is required to run the gdth driver. It appears just as
the gdth driver tries to flush data to the raid drives.
This issue only appears to occur under RH compiled kernels - recompiling the
kernel yourself using minimal scsi config options appears to overcome the
segfault. The array is 1x Mirrored, so two drives, RAID 1.
Wish I could provide more information than this, except for a lack of
mouse/syslog etc (as the segfault occurs literally seconds before "System
halted" (which in turn fails to appear) - all devices are unmounted by this
stage, so filesystem corruption should be minimal)
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Use a RedHat 7.3 SMP Kernel prior to this date.
2. Use an Intel SRCMR controller card, with 1x Mirrored array (gdth drivers)
3. Use an Intel SCB2 server board
4. Shutdown/halt/reboot the system.
Actual Results: The system segfaults with EIP errors (unlistable due to lack
of screen capture at this point).
Data loss seems to be zero by this point as the system appears to have
unmounted successfully; however on "gdth: flushing drives", and data not yet
written to the RAID drives will be lost.
Expected Results: "System Halted" should display enabling a system shutdown.
Forgot to mention, that once the seg fault occurs, the system will hang. This
is rather disasterous for servers that need a reboot, as it will never come
back up without a remote reboot.
I've looked at the code in question quite a bit but haven't found any obvious bugs..
this got fixed several errata ago