On Wed, Apr 13, 2022 at 06:03:07PM -0700, Luis Chamberlain wrote: > On Wed, Apr 13, 2022 at 05:50:38PM -0700, Bart Van Assche wrote: > > On 4/13/22 17:18, Luis Chamberlain wrote: > > > I've started to work on expanding coverage of testing with blktests > > > on kdevops [0] to other test groups. srp was one of them. But the amount > > > of failures I'm seeing seems to tell me I'm probably doing something > > > really stupid, so please help me review the setup. The baseline for > > > srp is listed below, each of these is a failure. I've used v5.17-rc7 > > > as a starting point. > > > > Do the SRP tests pass when using the Soft-iWARP driver instead of rdma_rxe? > > It looks *much* better. > > > I'm asking this because there are known issues with the rdma_rxe driver in > > kernel versions v5.17 and also in v5.18-rc1. > > OK thanks. > > > An example of how to select the > > Soft-iWARP driver: > > > > cd blktests && use_siw=1 ./check -q srp/001 > > I see failures on the first run: > > srp/005 A couple of more failures on the 3rd run, srp/006 (Direct I/O with large transfer sizes, cmd_sg_entries=255 and bs=8M) [failed] runtime 13.566s ... 9.712s --- tests/srp/006.out 2022-04-09 03:14:48.859579024 +0000 +++ /usr/local/blktests/results/nodev/srp/006.out.bad 2022-04-14 01:05:20.904694773 +0000 @@ -1,2 +1 @@ Configured SRP target driver -Passed srp/012 (dm-mpath on top of multiple I/O schedulers) [failed] runtime 5.601s ... 9.256s --- tests/srp/012.out 2022-04-09 03:14:48.859579024 +0000 +++ /usr/local/blktests/results/nodev/srp/012.out.bad 2022-04-14 01:06:46.136722414 +0000 @@ -1,2 +1 @@ Configured SRP target driver -Passed My exclusion list one-liner is getting longer, but hey, no crashes yet. i=0; while true; do use_siw=1 ./check -q srp -x srp/001 -x srp/005 -x srp/006 -x srp/011 -x srp/012 -x srp/013 ; if [[ $? -ne 0 ]]; then echo "BAD at $i"; break; else echo GOOOD $i ; fi; let i=$i+1; done; Luis
