| Summary: | System-freeze when watching YouTube Videos in Firefox or Chromium | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | Phillip Schichtel <phillip> |
| Component: | core | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | RESOLVED UPSTREAM | ||
| Severity: | crash | CC: | linx.system.adm, xaver.hugl |
| Priority: | NOR | ||
| Version First Reported In: | 6.0.4 | ||
| Target Milestone: | --- | ||
| Platform: | Arch Linux | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed/Implemented In: | ||
| Sentry Crash Report: | |||
|
Description
Phillip Schichtel
2024-04-21 13:53:08 UTC
I forgot to mention: My GPU is a vega64 using AMDGPU on kernel `6.8.7-arch1-1` with `amdgpu.gpu_recovery=0 amdgpu.ppfeaturemask=0xffffffff`. It looks like this is a bug in amdgpu driver https://gitlab.freedesktop.org/drm/amd/-/issues/2156 interesting, I might try that patch. Though a few concerns: 1. I'm not sure that change applies to my hardware 2. I don't see any of the relevant log lines that the original reporter there observed 3. most importantly: why does it hang the system in KDE, but not in Gnome? Does KDE allow for more hardware decoding compared to Gnome and has thus a higher chance of hitting this? The patch indeed seems to help. I'll monitor it for a few more days, but I think a freeze would already have happened. > most importantly: why does it hang the system in KDE, but not in Gnome? Does KDE allow for more hardware decoding compared to Gnome and has thus a higher chance of hitting this?
I don't think so, but GPU driver bugs are sometimes triggered by specific patterns in how rendering is done, or by buffer formats (like 10 bits per color with KWin vs 8 with other compositors), there can be many differences that can make one happen or not happen.
Anyways, total system freezes like that are always kernel bugs; if it seems GPU driver related you can directly create the an issue on drm,/amd about it in the future
Ok. Just a final update: The patch mentioned by @username was indeed not related, according to Alex Deucher it only affects navi21 and I also wasn't able to reproduce the issue after reverting the an unpatched kernel. Not sure what else changed, but this went from "happens on every video even across reboots" to "doesn't happen at all" basically overnight. Maybe some package update snug in that I didn't notice. |