Oddly enough, there was an issue last month with CrowdStrike and RHEL 9 kernel where they were triggering a kernel panic when attempting to load a bpf program from their newer bpf sensor. One of the workarounds was to switch to their kernel driver mode.

This was obviously a bug in RHEL kernel because even if the bpf program was bunk it should not cause the kernel to panic. However, it's almost like CrowdStrike does zero testing of their software and looks at their end users as Test/QA.

https://access.redhat.com/solutions/7068083

> 4bb7ea946a37 bpf: fix precision backtracking instruction iteration