Bug 512289 - Kernel 6.18.0 performance degradation with Firefox
Summary: Kernel 6.18.0 performance degradation with Firefox
Status: RESOLVED UPSTREAM
Alias: None
Product: kwin
Classification: Plasma
Component: wayland-generic (other bugs)
Version First Reported In: 6.5.2
Platform: Other Linux
: NOR normal
Target Milestone: ---
Assignee: KWin default assignee
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2025-11-18 11:47 UTC by Gurenko Alex
Modified: 2025-12-02 17:27 UTC (History)
1 user (show)

See Also:
Latest Commit:
Version Fixed/Implemented In:
Sentry Crash Report:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Gurenko Alex 2025-11-18 11:47:51 UTC
SUMMARY

I'm trying out a 6.18.0-rc.5/6 kernel and I see performance degradation in Plasma, there is a lag for mouse movement, specifically when entering and leaving Firefox (145.0). I can see a block of warning in the log when it happens:

```
...
Nov 18 12:40:10 kwin_wayland[3283]: atomic commit failed: Device or resource busy
Nov 18 12:40:18 kwin_wayland[3283]: atomic commit failed: Device or resource busy
...
```

STEPS TO REPRODUCE
1. Install 6.18.0-0.rc6.51.fc44.x86_64 kernel
2. Boot into new kernel
3. Launch Firefox
4. Try opening several tabs and move cursor

OBSERVED RESULT

Cursor movement lags, especially when entering and leaving Firefox window

EXPECTED RESULT

No change in performance

SOFTWARE/OS VERSIONS

Operating System: Fedora Linux 43
KDE Plasma Version: 6.5.2
KDE Frameworks Version: 6.20.0
Qt Version: 6.10.0
Kernel Version: 6.18.0-0.rc6.51.fc44.x86_64 (64-bit)
Graphics Platform: Wayland
Processors: 32 × AMD Ryzen 9 9950X3D 16-Core Processor
Memory: 64 GiB of RAM (60.4 GiB usable)
Graphics Processor 1: AMD Radeon RX 7900 XTX
Graphics Processor 2: AMD Ryzen 9 9950X3D 16-Core Processor

ADDITIONAL INFORMATION
The issue does not reproduce on current 6.17.x kernel branch, nothing else changes between affected and non-affected boot. Main reason I'm testing 6.18.0 kernel is to test the fix for https://gitlab.freedesktop.org/drm/amd/-/issues/4458 which is fixed by https://lore.kernel.org/amd-gfx/20251017130401.3196276-1-kevinyang.wang@amd.com/ and I'm wondering if the said fix is causing the lag.
Comment 1 Gurenko Alex 2025-11-18 13:01:02 UTC
I've updated title as 6.17.8 kernel also have some atomic commit failed: Device or resource busy errors too (not as many though), but it works as expected. Also seems like Firefox is the culprit. With Firefox opened some other random windows also affected by lag, but with Firefox closed, does not look like in my limited testing, no other application cause this lag
Comment 2 Zamundaaa 2025-11-18 15:46:07 UTC
Please report kernel regressions to the kernel / at https://gitlab.freedesktop.org/drm/amd/-/issues
Comment 3 Gurenko Alex 2025-11-24 19:58:35 UTC
(In reply to Zamundaaa from comment #2)
> Please report kernel regressions to the kernel / at
> https://gitlab.freedesktop.org/drm/amd/-/issues

I've opened a bug here: https://gitlab.freedesktop.org/drm/amd/-/issues/4725
Comment 4 Gurenko Alex 2025-11-25 09:52:14 UTC
I've been trying to help isolating a problem and found an interesting development. This only reproduces with Colour Profile set to None. If I set it to the ICC Profile or Built-In, then kwin_wayland[3283]: atomic commit failed: Device or resource busy messages stop appearing and lag if not completely but almost completely gone (still looking into it)
Comment 5 Zamundaaa 2025-11-25 11:54:13 UTC
Great, that means you can add to the amd issue that it only really happens with direct scanout. A kernel regression is still a kernel regression though, not something we can fix.
Comment 6 Gurenko Alex 2025-12-02 12:36:01 UTC
Hm, I've been re-testing today released 6.18.0 kernel and I know it's not officially enabled, but I forgot that I've added KWIN_USE_OVERLAYS=1 to my /etc/environment and this is the culprit. Issue only exists with it enabled, with removed OVERLAYS, both non-ICC and ICC profiles perform identically to the 6.17.x kernel and there are no atomic commit fail messages in system logs.
Comment 7 Zamundaaa 2025-12-02 17:27:38 UTC
That's great information, but doesn't change that it's a kernel regression.