[LU-16850] socklnd: ksocknal_shutdown() ASSERTION( net->ksnn_interface.ksni_nroutes == 0 ) failed

• Status: Resolved
• Project: Lustre
• Fix Version/s: Lustre 2.16.0
• Type: Bug
• Priority: Minor
• Reporter: Serguei Smirnov
• Assignee: Serguei Smirnov
• Resolution: Fixed
• Labels: lnet, socklnd
• Severity: 3

Description

This showed up in Janitor testing for LU-16709 patch, https://review.whamcloud.com/50530/

The sanity-lnet test introduced in that patch causes Janitor to run tests on that specific test ~20 times exercising the touched test only. All of these runs failed on the same assert.

Running the same test as part of full sanity-lnet suite execution doesn't trigger the assertion failure. Local testing also failed to reproduce the issue.

The assertion implies incomplete clean-up of socklnd peer on NI shutdown or, less likely, memory corruption.

It needs to be determined whether the issue causing the assertion to fail is introduced by LNet code changes LU-16709 patch or only discovered by the test changes introduced in the same patch.

Comments

[Gerrit Updater]

"Serguei Smirnov <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51148
Subject: LU-16850 socklnd: debug assert on ksni_nroutes == 0
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 3797f4867e43ed6edfa51501d5d3cecddb74e8be

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/51148/
Subject: LU-16850 socklnd: remove ksnr_myiface from ksock_conn_cb
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: f6be07c457385cfacd9b802e4cade9f6f6ab7d6f

[Peter Jones]

Landed for 2.16