https://bugzilla.suse.com/show_bug.cgi?id=1220119 https://bugzilla.suse.com/show_bug.cgi?id=1220119#c20 --- Comment #20 from Jiri Wiesner <jwiesner@suse.com> --- (In reply to Ivan Ivanov from comment #19)
Created attachment 872923 [details] Hm, happens here too on Thinkpad T495. Screenshot attached
Since this is a kernel running on a bare metal machine it is a different issue. Yours is actually a hardware issue. Either the HPET or the TSC is wrong. I bet it's the HPET. You could pass tsc=nowatchdog to the kernel to disable the checks (on any kernel). If you ran the latest kernel on Leap 15.4/5/6 you could pass tsc=watchdog to force the TSC to act as a watchdog (implemented by my out-of-tree patch that was ignored upstream). Since this is a 1-socket machine and you probably do not need TSC warp detection, you could also use "tsc=reliable tsc=watchdog" to accomplish a similar effect as with tsc=watchdog and my out-of-tree patch in Leap. AFAIK, the clocksource watchdog error often tends to happen during boot and on machines that are under thermal stress. It's rather unnecessary to have the TSC marked unstable when there is a good chance that it is the HPET readout what is wrong. That's why I implemented the semantic change of tsc=watchdog. Another option would be to postpone the clocksource watchdog checks until after the system fully boots up. I can just see the cold reception a patch like this would receive upstream. They do like their clocksource watchdog (even if its actions - switching to the HPET and causing a performance degradation - are uncalled for). -- You are receiving this mail because: You are on the CC list for the bug.