On Thu, 2017-08-10 at 20:32 +1000, Michael Ellerman wrote: > "Martin K. Petersen" <martin.petersen@xxxxxxxxxx> writes: > > > One of the two scsi-mq functions that requeue a request unprepares a > > > request before requeueing (scsi_io_completion()) but the other > > > function not (__scsi_queue_insert()). Make sure that a request is > > > unprepared before requeuing it. > > > > Applied to 4.13/scsi-fixes. Thanks much! > > This seems to be preventing my Power8 box, which uses IPR, from booting. > > Bisect said so: > > # first bad commit: [270065e92c317845d69095ec8e3d18616b5b39d5] scsi: scsi-mq: Always unprepare before requeuing a request > > And if I revert that on top of next-20170809 my system boots again. > > The symptom is that it just gets "stuck" during boot and never gets to > mounting root, full log below, the end is: > > . > ready > ready > sd 0:2:4:0: [sde] 554287104 512-byte logical blocks: (284 GB/264 GiB) > sd 0:2:4:0: [sde] 4096-byte physical blocks > sd 0:2:5:0: [sdf] 272646144 512-byte logical blocks: (140 GB/130 GiB) > sd 0:2:5:0: [sdf] 4096-byte physical blocks > sd 0:2:4:0: [sde] Write Protect is off > sd 0:2:4:0: [sde] Mode Sense: 0b 00 00 08 > sd 0:2:5:0: [sdf] Write Protect is off > sd 0:2:5:0: [sdf] Mode Sense: 0b 00 00 08 > > > And it just sits there for at least hours. > > I compared a good and bad boot log and there appears to be essentially > no difference. Certainly nothing that looks suspicous. Hello Michael, Thanks for having reported this early. Is there any chance that you can reproduce this state, press SysRq-w on the console and collect the task overview that is reported on the console (see also Documentation/admin-guide/ sysrq.rst)? If this is not possible or if that task overview does not report any blocked tasks, can you add scsi_mod.scsi_logging_level=-1 to the kernel command line (either through /etc/default/grub or in /boot/grub2/grub.cfg when using GRUB), reboot the system, capture the console output and report that output as a reply to this e-mail? Thanks, Bart.
