[dm-devel] dm-rq queue stalls

[Laurence Oberman]

On Wed, 2018-01-17 at 19:29 +0000, Bart Van Assche wrote:
> On Wed, 2018-01-17 at 13:29 -0500, Laurence Oberman wrote:
> > If you can share your exact fio run scriptparameters I can try this
> > on
> > my mlx5 test bed using the same kernel you are using.
> > Ming has access to the same test bed as well.
> > Is this direct to devices, or on a file system etc. etc.
> > 
> > Are these from the tests directory in the srp test kit of yours.
> 
> Hello Laurence,
> 
> This issue can be reproduced as follows:
> * Clone the block-scsi-for-next branch from https://github.com/bvanas
> sche/linux
>   (commit cd0b6b9f8779).
> * Clone the srp-test project from https://github.com/bvanassche/srp-t
> est.
> * Modify /etc/multipath.conf as indicated in the README.md file of
> the srp-test
>   project and restart multipathd.
> 
> If you prefer the rdma_rxe driver, run the following command:
> 
> srp-test/run_tests -c -d -r 10 -q 1 -t 02-mq -e kyber
> 
> Or if you prefer to run the test against an IB driver (mlx4, mlx5,
> ..) on an IB
> system, run the following command:
> 
> srp-test/run_tests -d -r 10 -q 1 -t 02-mq
> 
> If running one of the above commands once would not be sufficient to
> trigger the
> lockup, rerun the command a few times.
> 
> Bart.

Thanks
Bart
I will take care of this test and come back to you.
First going to test the scatter/gather fix.
Regards
Laurence