Most likely you're feeling placebo. By entering preempt=full into your boot parameters, you're priming your mind to expect a different behavior and seeing things you otherwise would ignore.

If you believe there is a difference, you'll need to construct a repeatable synthetic test to prove the parameter is changing anything.

As for why restoring CONFIG_PREEMPT_DYNAMIC did anything for you, it's most likely an implementation bug where not using dynamic ends up with some busted code being compiled.
Back to top

I thought, incorrectly, dynamic preemption and full preemption are different values for a same option. Actually, dynamic preemption is another thing not (very related). Using preempt=full do nothing (it's already using this value), but dynamic preemption do. I will try to execute the kernel 6.1.0-1.3 that hasn't dynamic preemption enabled, and I will check how things going.
I think keep dynamic preemption enabled is valid because the user could change the preemption value to full (the default), voluntary (Ubuntu default) or none (better for servers).
Back to top

I started with 6.1.0-1.3 and now everything worked fine (Firefox with few tabs and game), but the command you said, `journalctl -k|grep Preem` didn't show anything about the dynamic preemption level (full in that case). I think that is expected, but not desirable.
Back to top