[bug report] blktests nvme/022 lead kernel WARNING and NULL pointer

[Yi Zhang]

On Sat, May 1, 2021 at 5:58 PM Yi Zhang <yi.zhang at redhat.com> wrote:
>
> On Sat, May 1, 2021 at 8:55 AM Sagi Grimberg <sagi at grimberg.me> wrote:
> >
> >
> > > Hello
> > > Recently CKI reproduced this WARNING and NULL pointer with
> > > linux-block/for-next on aarch64, seems it's one regression, I will try
> > > if I can bisect the culprit.
> > >
> > > blktests: nvme/022 (test NVMe reset command on NVMeOF file-backed ns)
> > >
> > > [ 1879.759978] run blktests nvme/022 at 2021-04-30 12:30:36
> > > [ 1879.804283] nvmet: adding nsid 1 to subsystem blktests-subsystem-1
> > > [ 1879.819087] nvmet: creating controller 1 for subsystem
> > > blktests-subsystem-1 for NQN
> > > nqn.2014-08.org.nvmexpress:uuid:0da758a0-4d84-4133-82dd-9801235b55cd.
> > > [ 1879.833081] nvmet: unhandled identify cns 6 on qid 0
> > > [ 1879.838079] nvme nvme0: creating 128 I/O queues.
> > > [ 1879.852353] nvme nvme0: new ctrl: "blktests-subsystem-1"
> > > [ 1880.879731] nvme nvme0: resetting controller
> > > [ 1889.940458] nvmet: ctrl 1 keep-alive timer (5 seconds) expired!
> > > [ 1889.946377] nvmet: ctrl 1 fatal error occurred!
> > > [ 1889.950928] nvme nvme0: Removing ctrl: NQN "blktests-subsystem-1"
> >
> > It appears that we are somehow now expire the kato after/during a reset
> > sequence and then seem to race reset and remove...
> >
> > bisection will help definitely.
> >
> Hi Sagi
>
> Bisect shows bellow commits was the first bad commit.
>
> commit a70b81bd4d9d2d6c05cfe6ef2a10bccc2e04357a (refs/bisect/bad)
> Author: Hannes Reinecke <hare at suse.de>
> Date:   Fri Apr 16 13:46:20 2021 +0200
>
>     nvme: sanitize KATO setting
>

Hi Sagi
This issue has been consistently reproduced with blktests on recent
linux-block/for-next, do you have a chance to check it?


> Thanks
> Yi