[bug report] NVMe/IB: nvme disconnect -n testnqn need more than 1min

Yi Zhang yi.zhang at redhat.com
Sun Dec 12 19:26:13 PST 2021


On Sun, Dec 12, 2021 at 5:32 PM Sagi Grimberg <sagi at grimberg.me> wrote:
>
>
> > Hello
> >
> > I found the nvme disconnect always needs more than 1min to be
> > finished, could anyone help check this issue?
>
> Yi,
>
> Does this happen with rxe? didn't see this in my test VM lately.

Got "rdma_rxe: check_rkey: no MW matches rkey" error when I use rdma_rxe.
but it cannot be reproduced when I change to use siw and
NVMe/ROCE(Ethernet: Mellanox MT27710 CX-4 Lx])

>
> >
> > Reproducer:
> > + nvme connect -t rdma -a 172.31.0.202 -s 4420 -n testnqn
> >
> > real    0m4.673s
> > user    0m0.000s
> > sys     0m0.005s
> > + nvme disconnect -n testnqn
> > NQN:testnqn disconnected 1 controller(s)
> >
> > real    1m9.028s
> > user    0m0.000s
> > sys     0m0.187s
> >
> > client:
> > [  283.966930] nvme nvme0: creating 40 I/O queues.
> > [  288.527028] nvme nvme0: mapped 40/0/0 default/read/poll queues.
> > [  288.530501] nvme nvme0: new ctrl: NQN "testnqn", addr 172.31.0.202:4420
> > [  288.533203] nvme nvme0: Removing ctrl: NQN "testnqn"
> > [  357.534154] nvme nvme0: I/O 18 QID 0 timeout
> > [  357.539644] nvme nvme0: Property Set error: 881, offset 0x14target:
> >
> > target:
> > [  229.016490] nvmet: adding nsid 1 to subsystem testnqn
> > [  229.044861] nvmet_rdma: enabling port 2 (172.31.0.202:4420)
> > [  280.697930] nvmet: creating nvm controller 1 for subsystem testnqn
> > for NQN nqn.2014-08.org.nvmexpress:uuid:4c4c4544-0056-4c10-8058-b7c04f383432.
> > [  291.155204] nvmet: ctrl 1 keep-alive timer (5 seconds) expired!
> > [  291.162238] nvmet: ctrl 1 fatal error occurred!--
>
> Will have another look.

Thanks.

>


--
Best Regards,
  Yi Zhang




More information about the Linux-nvme mailing list