softirq kernel panic in Centos 7.9

[wc1000]

I am seeing kernel panics on stock Centos 7.9 similar to the attached back trace (captured within crash via kdump).
Currently seen these on three devices after they were running for 2, 4 and 20 days, once each.
The servers are dual socket running: E5-2695 v4 (2x18 cores) with hyperthreading enabled.

The servers are running a DPDK application but the crash is not on one of the DPDK cores.
I did not see these crashes running earlier versions of Centos 7.

Has anyone seen anything similar or have any suggestions of what might be causing this?

My current working theory is a hardware interrupt corrupting a register used for Spectre mitigation trampoline...
Did hardware interrupt routines need to be modified to support Spectre mitigation?

[TrevorH]

Your 3.10.0-1160.31.1.el7 is from May 2021 and the current kernel 3.10.0-1160.59.1.el7 rpm changelog has 269 lines of fixes listed in it since yours. Run yum update to get up to date and then try to recreate the problem to see if it is already fixed. If it still crashes then you need to report the problem on bugzilla.redhat.com as CentOS only rebuilds what Red Hat release for RHEL 7.

[wc1000]

Thanks. I've upgraded one of the systems to latest Centos 7.9 and will leave it running.