ib_send_bw test stuck

[amitgeron]

When running the following, for example: ib_send_bw -b -e -q 16

I think that the root cause is wrong usage of notifications - the cq isn't emptied when the application is woken up. The argument passed to ibv_poll_cq is user_param->rx_depth, but we might have more completions than that, since we use multiple QPs.

There are two options I can think of for a fix:

1. Use a larger completion buffer when calling ibv_poll_cq (user_param->rx_depth * number_of_qps)
2. As done in run_iter_be_server - loop over ibv_poll_cq as long as there are completions in the cq

[sshaulnv]

Hi, The issue is not reproducing on my setup. can you please provide more information about your setup and chip , driver, etc? Thanks

[amitgeron]

Hi, you may see the setup information in the linked PR https://github.com/linux-rdma/perftest/pull/138 Don't know if it matters, but perftest is configured to use old post_send method.

[HassanKhadour]

Hi amitgeron, Thanks for your collaboration! as you said there is a linked PR for this issue that is waiting for code fix, So I'll close this issue