Re: Apparent regression in blktests since 5.18-rc1+

[Bob Pearson <rpearsonhpe@gmail.com>]

On 5/12/22 22:40, Bart Van Assche wrote:
> On 5/12/22 17:41, Bob Pearson wrote:
>> OK thanks. I don't exactly know how to merge "the latest master branch". I did merge the tag v5.18-rc2 and am now trying
>> v5.18-rc4. As far as I can tell rc6 is the latest branch. When I tried to merge the git repo it didn't work.
> 
> Hmm ... what didn't work? Merging or running blktests?
> 
> Anyway, this is how I merge Linus' latest branch:
> 
> $ git remote add origin \
>     git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> $ git fetch origin
> $ git merge origin/master
> 
> I also found which commit I merged from the master branch (see also my email from May 7):
> $ git merge 4b97bac0756a
> 
> Thanks,
> 
> Bart.

Thanks Bart. I was able to follow your steps above. But unfortunately not much has changed.
I still see hangs in siw (with no code changes by me) and also in rxe (but here I have
fixed some lockdep warnings so it will run in a debug kernel.) There are two test cases that
cause the most problems. srp/002 and srp/011. 011 always fails solidly. 002 sometimes hangs and
sometimes completes but with failed status. The rest of the tests all pass.

Both tests hang at a line that looks like
	scsi host6: ib_srp: Already connected to target port with id_ext=...

When 002 completes but fails there are 14 second gaps at some of the same lines in the trace.
This has the feel of the earlier problem with the 3 minute timeout that was fixed by the
patch (revert ... scsi_debug.c) that you sent and is applied here.

I really don't know how to make progress here. If anyone knows what is happening at the
already connected lines let me know. They seem normal except for the long gaps and hangs
when they occur.

Thanks, Bob