[LU-12402] LNet Health: lnet_finalize() recursion

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.13.0, Lustre 2.12.5
• Severity: 3
• Assignee: Amir Shehata (Inactive)
• Reporter: Amir Shehata (Inactive)

Description

When there are many messages being dropped, health feature introduced a path where it is possible to enter into a deep recursion path.

lnet_finalize()->lnet_health_check()->lnet_msg_decommit_tx()->
lnet_return_tx_credits_locked()->lnet_post_send_locked()->lnet_finalize()

This was dealth with in lnet_finalize() via keeping track of the finalizers thread using msc_finalizers. And returning if all slots are busy.

The above path doesn't have the same mechanism, therefore is susceptible to this problem.

Issue Links

• is related to: LU-13483 Apparently infinite recursion in lnet_finalize()