smp: Fix soft lockup problem in hard lockup test

Waiman Longrequested to merge
llong1/centos-stream-9:rhel-16867_smp into main

JIRA: https://issues.redhat.com/browse/RHEL-16867
MR: !6855 (merged)

The crasher tool mentioned in the Jira causes a deadlock in the CPU running the tool to cause a hard lockup panic. However, if another running task is sending IPIs to other CPUs and expecting responses from other CPUs at the same time, it may cause an avalanche of soft lockup messages in the kernel log which somehow prevents the watchdog from detecting the hard lockup.

This problem is likely to be solved by backporting some upstream patches to reduce redundant soft lockup messages and eventually panic the system if the CSD lock get stuck for too long. As the problem isn't easy to reproduce, we can't easily verify if this MR can really fix the problem. Let hope for the best and create another bug if this happens again in the future.

Signed-off-by: Waiman Long longman@redhat.com

Edited by Waiman Long

Merge request reports