Skip to content

ib_write_bw --rdma_cm loopback test intermittently fails with "rdma_bind_addr failed" #383

@panjason26-eng

Description

@panjason26-eng

Hi,

Environment:

  • OS: Ubuntu 22.04 (Linux 6.5)
  • perftest version: 25.01.0-0.70

Reproduction:
I run a loopback test using two consecutive ib_write_bw commands (referencing line 219 in the attached file):

sudo ib_write_bw -d rocep77s0f1 -i 1 --rdma_cm -D 3 -s 4M --run_infinitely -F --report_gbits &
sleep 3;
sudo ib_write_bw -d rocep77s0f0 -i 1 --rdma_cm -D 3 -s 4M --run_infinitely -F --report_gbits 11.7.7.111

Observed behavior:

  • The first time I run the command pair, nothing happens (it exits without output).
  • The second time I run the same commands, it gets stuck with the error: "rdma_bind_addr failed".
  • After interrupting and trying a third time, the commands succeed.
  • If I omit the --rdma_cm option, the commands work consistently and do not show this intermittent failure.

Questions:

  1. Is this intermittent behavior expected when using --rdma_cm for loopback tests, or am I using the commands incorrectly?
  2. If this is a bug or an unexpected behavior, what additional logs or information would be helpful for debugging?

Attached: the referenced file (please see line 229).

Thanks for any guidance!

Jason

cx7_loopback_rdma.txt

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions