On Thu, 7 Feb 2013, Bart Van Assche wrote:
> On 02/06/13 23:31, Joe Lawrence wrote:
> > crash> list scsi_device.siblings -H 0xffff8808513a4290 -s scsi_device
> >
> > ffff880851232520
> > struct scsi_device {
> > is_visible = 0x1,
> > sdev_state = SDEV_DEL,
> > }
> > ffff880851235388
> > struct scsi_device {
> > is_visible = 0x1,
> > sdev_state = SDEV_DEL,
> > }
>
> This is interesting. This probably means that one or more threads got stuck in
> __scsi_remove_device(). If you still have the crash dump available it would be
> appreciated if you could verify whether this is correct. If so, there might be
> an issue in the mpt2sas driver where scsi_done() does not get invoked for all
> outstanding commands after a surprise removal.
Hi Bart,
I haven't had time to rerun the test without the two patches that wait in
scsi_remove_host(), however I did rerun the test and verify the same
behavior as in my earlier mail. I didn't see any __scsi_remove_device()
instances running.
Some more investigation revealed that MD RAID was holding a reference to
the removed device. (In short, mdadm --remove had failed and left the
device as a faulty member of the array.) When I did finally manage to
kick that disk from the MD device, scsi host/device removal continued to
completion as expected.
There's a bit more context to the MD situation that I'll post to the raid
list once I get the details together for Neil. I will CC you if you are
interested in following.
Regards,
-- Joe
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
