[LU-12729] lnet_selftest not work in multi-net and multi-rail client

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Affects Version/s: Lustre 2.12.2
• Severity: 3
• Assignee: Serguei Smirnov
• Reporter: Mahmoud Hanafi

Description

A multi-rail client can't setup lnet selftest to a router with multiple interface like this:

# lctl list_nids 10.151.26.148@o2ib 10.151.25.170@o2ib 10.141.26.148@o2ib417 10.141.25.170@o2ib417 

Here is what peer show looks like

    - primary nid: 10.151.26.148@o2ib
      Multi-Rail: True
      peer ni:
        - nid: 10.141.26.148@o2ib417
          state: up
        - nid: 10.141.25.170@o2ib417
          state: NA
        - nid: 10.151.26.148@o2ib
          state: NA
        - nid: 10.151.25.170@o2ib
          state: NA

~ # modprobe lnet_selftest
~ # export LST_SESSION=99
~ # lst new_session test
SESSION: test FEATURES: 1 TIMEOUT: 300 FORCE: No
nbp16-srv1 ~ # lst add_group servers 10.151.26.148@o2ib                                        
create session RPC failed on 12345-10.151.26.148@o2ib: Unknown error -110
No nodes added successfully, deleting group servers
Group is deleted