[PATCH v2] RDMA/cma: prevent rdma id destroy during cma_iw_handler

[Shinichiro Kawasaki]

On Jun 15, 2023 / 09:45, Shin'ichiro Kawasaki wrote:
[...]
> I stop my fix attempt here, since it looks beyond my bandwidth now. If anyone
> provides fix patches, I'm willing to test them.

FYI, I tried to recreate this failure on my test system this week, and found it
is no longer recreated. In case your TODO list includes a work item for this
issue, please put lower priority on it.

Two month ago, the failure had been recreated by repeating the blktests test
cases nvme/030 and nvme/031 with SIW tranposrt around 30 times. Now I do not see
it when I repeat them 100 times. I suspected some kernel changes between v6.4
and v6.5 would have fixed the issue, but even when I use both kernel versions,
the failure is no longer seen. I reverted nvme-cli and libnvme to older version,
but still do not see the failure. I just guess some change on my test system
affected the symptom (Fedora 38 on QEMU).