raid5d oopses during hot-removal of disk (2.4.27)

Hello all,

I am experiencing kernel "oops" errors in raid5d.

This is on a Linux 2.4.27 kernel, running RAID-5 across 4 SATA drives
using the Marvell Linux driver (not libata) which acts as a SCSI host.

The oopses occur when a disk is physically hot-removed from the system
and Software RAID finds the disk failed.

I have analyzed the error to occur inside a ITERATE_RDEV loop in
md_update_sb() in the drivers/md/md.c code.

Please note that concurrent to the Software RAID-5 marking the
disk as failed, another process is performing a
"scsi-remove-single-device" action.

I am suspecting that this *might* trigger the fault, but I have not
been able to confirm this. However, I hate guessing here.

I have included the oops, together with a text file in which I
described the steps (ksymoops et al) taken so far to pin-point
the occurence.

Could someone more knowledgeable about md/raid look at this pls,
and tell me how to proceed?

Regards,

Leon Woestenberg.