On 12/14/2021 12:39 PM, Sagi Grimberg wrote:

quoted

quoted

quoted

Hi Sagi
It is still reproducible with the change, here is the log:

# time nvme reset /dev/nvme0

real    0m12.973s
user    0m0.000s
sys     0m0.006s
# time nvme reset /dev/nvme0

real    1m15.606s
user    0m0.000s
sys     0m0.007s

Does it speed up if you use less queues? (i.e. connect with -i 4) ?

Yes, with -i 4, it has stablee 1.3s
# time nvme reset /dev/nvme0

So it appears that destroying a qp takes a long time on
IB for some reason...

quoted

real 0m1.225s
user 0m0.000s
sys 0m0.007s

quoted

quoted

# dmesg | grep nvme
[  900.634877] nvme nvme0: resetting controller
[  909.026958] nvme nvme0: creating 40 I/O queues.
[  913.604297] nvme nvme0: mapped 40/0/0 default/read/poll queues.
[  917.600993] nvme nvme0: resetting controller
[  988.562230] nvme nvme0: I/O 2 QID 0 timeout
[  988.567607] nvme nvme0: Property Set error: 881, offset 0x14
[  988.608181] nvme nvme0: creating 40 I/O queues.
[  993.203495] nvme nvme0: mapped 40/0/0 default/read/poll queues.

BTW, this issue cannot be reproduced on my NVME/ROCE environment.

Then I think that we need the rdma folks to help here...

Max?

It took me 12s to reset a controller with 63 IO queues with 5.16-rc3+.

Can you try repro with latest versions please ?

Or give the exact scenario ?