cyring / CoreFreq

CoreFreq : CPU monitoring and tuning software designed for 64-bit processors.
https://www.cyring.fr
GNU General Public License v2.0
1.97k stars 126 forks source link

System hang when registering Clock Source twice #229

Closed ppascher closed 3 years ago

ppascher commented 3 years ago

I registered "Clock Source" in the settings menu which worked fine. Next I wanted to "unregister" but accidentally hit "register" again. This caused my system to hang. Not sure if this is due to corefreq or a bug with amdgpu? Journal only shows me this:

Apr 28 16:53:24 kernel: clocksource: corefreq: mask: 0xffffffffffffffff max_cycles: 0x3102c246e1e, max_idle_ns: 440795343192 ns
Apr 28 16:57:34 kernel: clocksource: corefreq: mask: 0xffffffffffffffff max_cycles: 0x3102c2dd02c, max_idle_ns: 440795323298 ns
Apr 28 16:57:38 kernel: [drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* Waiting for fences timed out!
Apr 28 16:57:39 kernel: [drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* Waiting for fences timed out!

after which I hit the reset button since switching tty was not possible. This is on kernel 5.12 with a 5950x CPU and 290x GPU using corefreq-git (split packages) on ArchLinux.

Thank you for your work cyring.

cyring commented 3 years ago

Hello

ppascher commented 3 years ago

No, I did not build with DELAY_TSC (using AUR corefreq-dkms-git). Neither did I register as idle, governor or cpufreq. I should have checked the README. Closing this issue.

cyring commented 3 years ago

No, I did not build with DELAY_TSC (using AUR corefreq-dkms-git). Neither did I register as idle, governor or cpufreq. I should have checked the README. Closing this issue.

Just to make code better: you mean that with prerequisites fully set up, you don't encounter the issue anymore ?

If true then I will add those registrations check before any attempt to enable my clock source algorithm.