[PATCH blktests 0/2] Test multipath and marginal ports

[Jesse Taube]

On Mon, Aug 17, 2026 at 3:39 AM Shin'ichiro Kawasaki
<shinichiro.kawasaki at wdc.com> wrote:
>
> On Aug 14, 2026 / 10:18, Jesse Taube wrote:
> ...
> > > I applied the series on top of v7.2-rc7 kernel. Then I ran the test case added
> > > by this blktests series and observed the failurea below:
> > >
> > > nvme/070 (tr=fc) (test nvme-fc marginal path handling)       [failed]
> > >     runtime  33.581s  ...  33.028s
> > >     --- tests/nvme/070.out      2026-08-14 12:03:21.338000000 +0900
> > >     +++ /home/shin/Blktests/blktests/results/nodev_tr_fc/nvme/070.out.bad       2026-08-14 20:17:04.731000000 +0900
> > >     @@ -2,7 +2,8 @@
> > >      Starting background I/O
> > >      Testing iopolicy: numa
> > >      Changing FC links to online
> > >     -Set all online: pass
> > >     +No FC ports are being used, expected atleast one in use when all are online in numa mode.

This is should pass even without my patches. I had this happen once
when a link got stuck in connecting.
it could be solved by adding a longer delat after _run_fio_verify_io
as it sometimes takes a bit to connect

> > >     +Set all online: fail
> > >      One host marginal: pass
> > >     ...
> > >     (Run 'diff -u tests/nvme/070.out /home/shin/Blktests/blktests/results/nodev_tr_fc/nvme/070.out.bad' to see the entire diff)
> > >
> > > Is this failure expected?
> >
> > No. I can't reproduce the issue either.
> > Does this happen consistently?
>
> Yes, it happnes always I run the test case.
> After I rebased to v7.2 kernel, it shows a bit different message:
>
> nvme/070 (tr=fc) (test nvme-fc marginal path handling)       [failed]
>     runtime  37.200s  ...  38.926s
>     --- tests/nvme/070.out      2026-08-17 11:13:53.073000000 +0900
>     +++ /home/shin/Blktests/blktests/results/nodev_tr_fc/nvme/070.out.bad       2026-08-17 16:34:17.997000000 +0900
>     @@ -3,10 +3,14 @@
>      Testing iopolicy: numa
>      Changing FC links to online
>      Set all online: pass
>     -One host marginal: pass
>     +FC port on (/sys/class/nvme-subsystem/nvme-subsys6/nvme6) is being used, expected no use.
>     +One host marginal: fail
>      All marginal: pass
>     ...
>     (Run 'diff -u tests/nvme/070.out /home/shin/Blktests/blktests/results/nodev_tr_fc/nvme/070.out.bad' to see the entire diff)
>
> > Maybe the links are still connecting?
>
> Should I try to add some sleeps in the test script? If so, could you suggest
> where to put the sleeps?

maybe add a delay before the cat command in _rport_in_use it might take a second
for the io to move paths

> If it helps, I will share the kernel config I use. If you want it, please let
> me know.

That would be good. Im using the base centos10 config with
https://github.com/Mr-Bossman/linux/tree/dev/jesse/fpin_v11k

Thanks,
Jesse Taube