Bug 524840 - KWin 6.7.4 crashes with SIGSEGV on NVIDIA 610.57.04 under Wayland
Summary: KWin 6.7.4 crashes with SIGSEGV on NVIDIA 610.57.04 under Wayland
Status: RESOLVED FIXED
Alias: None
Product: kwin
Classification: Plasma
Component: wayland-generic (other bugs)
Version First Reported In: 6.7.4
Platform: Other Linux
: NOR crash
Target Milestone: ---
Assignee: KWin default assignee
URL:
Keywords:
: 523771 524325 525354 526342 (view as bug list)
Depends on:
Blocks:
 
Reported: 2026-08-27 14:52 UTC by actimel323
Modified: 2026-10-07 10:56 UTC (History)
7 users (show)

See Also:
Latest Commit:
Version Fixed/Implemented In: 6.7.6
Sentry Crash Report:


Attachments
Kernel journal from the previous boot (135.86 KB, text/plain)
2026-08-27 14:52 UTC, actimel323
Details
Full GDB backtrace from the KWin crash, including all threads, registers and local variables. (15.45 KB, text/plain)
2026-08-27 14:52 UTC, actimel323
Details
NVIDIA bug report (795.29 KB, application/gzip)
2026-08-27 14:53 UTC, actimel323
Details
KWin user journal from the previous boot, covering the crash (67.53 KB, text/plain)
2026-08-27 14:54 UTC, actimel323
Details
Hardware and software configuration summary generated with inxi (5.97 KB, text/plain)
2026-08-27 14:54 UTC, actimel323
Details

Note You need to log in before you can comment on or make changes to this bug.
Description actimel323 2026-08-27 14:52:10 UTC
Created attachment 195526 [details]
Kernel journal from the previous boot

DESCRIPTION
KWin 6.7.4 intermittently crashes with SIGSEGV on Fedora 44 using NVIDIA driver 610.57.04 and a Wayland session.

STEPS TO REPRODUCE
I have not yet found a 100% reproducible sequence of actions.
The issue occurs during normal desktop use under KDE Wayland, specfically while I was playing Counter-Strike 2 without proton (native), with custom args: "SDL_VIDEO_DRIVER=wayland -w 2560 -h 1440"

OBSERVED RESULT
KWin crashes and generates a core dump.
- The core dump reports that Program terminated with signal SIGSEGV, Segmentation fault.
- The crashed process is /usr/bin/kwin_wayland
- The available core dump currently does not contain a useful resolved frame stack beyond the crashing thread, despite debug information being downloaded.

EXPECTED RESULT
KWin should remain running and continue compositing the Wayland session.

SOFTWARE/OS VERSIONS
Operating System: Fedora Linux 44
Qt Version: 6.11.1-1
Kernel: 7.1.10-200
Desktop: KDE Plasma
KWin: 6.7.4
KDE Plasma Version: 6.7.4
KDE Frameworks Version: 6.29.0
Session: Wayland
GPU: NVIDIA GeForce RTX 3060 Ti
NVIDIA driver: 610.57.04
KMD version: 610.57.04
KWin is running on the 3060ti GPU and nvidia-drm is used for the display.

ADDITIONAL INFORMATION
nvidia-smi reports:
GPU: NVIDIA GeForce RTX 3060 Ti
Driver Version: 610.57.04
KMD Version: 610.57.04
Display Active: Yes

KWin is listed as a GPU process. At the time of investigation, dmesg did not show an NVIDIA Xid error associated with this particular crash.
Relevant journal output immediately before/around the crash contains no obvious pageflip/Xid error in the excerpt I currently have.
I can provide the complete KWin core/backtrace, journal, and NVIDIA bug report if required.

This does not appear to be an isolated configuration issue, because a very simillar issue has been reported with:
- Fedora 44
- KDE Plasma / KWin 6.7.4
- Wayland
- NVIDIA 610.57.04
- multiple DisplayPort monitors

There are also reports of "Pageflip timed out!" / nvidia-drm failures continuing on NVIDIA 610.57.04 with KDE Wayland.
My configuration differs from those reports in some details, including GPU model and refresh rate, but the software stack is closely related.
Comment 1 actimel323 2026-08-27 14:52:45 UTC
Created attachment 195527 [details]
Full GDB backtrace from the KWin crash, including all threads, registers and local variables.
Comment 2 actimel323 2026-08-27 14:53:11 UTC
Created attachment 195528 [details]
NVIDIA bug report
Comment 3 actimel323 2026-08-27 14:54:07 UTC
Created attachment 195529 [details]
KWin user journal from the previous boot, covering the crash
Comment 4 actimel323 2026-08-27 14:54:48 UTC
Created attachment 195530 [details]
Hardware and software configuration summary generated with inxi
Comment 5 Zamundaaa 2026-08-27 16:08:52 UTC
I'm relatively confident your crash is caused by a GPU reset, and should be fixed with https://invent.kde.org/plasma/kwin/-/commit/067646a4571a9e75d6cf5bfb64df796c29a79fcb in 6.7.5

If you see it again then, just reopen this with a new backtrace
Comment 6 Zamundaaa 2026-08-27 16:46:59 UTC
*** Bug 524325 has been marked as a duplicate of this bug. ***
Comment 7 TraceyC 2026-08-27 20:09:25 UTC
*** Bug 523771 has been marked as a duplicate of this bug. ***
Comment 8 TraceyC 2026-09-08 15:34:59 UTC
*** Bug 525354 has been marked as a duplicate of this bug. ***
Comment 9 Zamundaaa 2026-09-29 13:05:11 UTC
*** Bug 526342 has been marked as a duplicate of this bug. ***
Comment 10 Zamundaaa 2026-09-29 13:06:35 UTC
There's another case where we're missing makeCurrent entirely, rather than just the check if it was successful.
Comment 11 Bug Janitor Service 2026-09-29 13:07:16 UTC
A possibly relevant merge request was started @ https://invent.kde.org/plasma/kwin/-/merge_requests/10003
Comment 12 Vlad Zahorodnii 2026-10-01 06:39:49 UTC
Git commit ad655ed764a61995776abd6b59d98cec1814531f by Vlad Zahorodnii, on behalf of Xaver Hugl.
Committed on 01/10/2026 at 06:19.
Pushed by vladz into branch 'master'.

plugins/screenshot: make the context current after prePaint

prePaint can change the context, and changing it back can fail. The compositor
doesn't hit this issue because it calls beginFrame on the layer after prePaint,
which makes the context current as well.

M  +3    -0    src/plugins/screenshot/screenshot.cpp

https://invent.kde.org/plasma/kwin/-/commit/ad655ed764a61995776abd6b59d98cec1814531f
Comment 13 Vlad Zahorodnii 2026-10-01 07:17:34 UTC
Git commit 0e062dad8b206cf830ce1432d4aeeb850b23abc4 by Vlad Zahorodnii.
Committed on 01/10/2026 at 06:41.
Pushed by vladz into branch 'Plasma/6.7'.

plugins/screenshot: make the context current after prePaint

prePaint can change the context, and changing it back can fail. The compositor
doesn't hit this issue because it calls beginFrame on the layer after prePaint,
which makes the context current as well.


(cherry picked from commit 3f13a0b86fbc2782e6036b811a52c3a484a3ad62)

Co-authored-by: Xaver Hugl <xaver.hugl@kde.org>

M  +3    -0    src/plugins/screenshot/screenshot.cpp

https://invent.kde.org/plasma/kwin/-/commit/0e062dad8b206cf830ce1432d4aeeb850b23abc4
Comment 14 Vlad Zahorodnii 2026-10-01 07:58:16 UTC
Git commit 785a9e6aff5099d671190cd83574b73045888337 by Vlad Zahorodnii.
Committed on 01/10/2026 at 07:28.
Pushed by vladz into branch 'Plasma/6.8'.

plugins/screenshot: make the context current after prePaint

prePaint can change the context, and changing it back can fail. The compositor
doesn't hit this issue because it calls beginFrame on the layer after prePaint,
which makes the context current as well.


(cherry picked from commit ad655ed764a61995776abd6b59d98cec1814531f)

Co-authored-by: Xaver Hugl <xaver.hugl@kde.org>

M  +3    -0    src/plugins/screenshot/screenshot.cpp

https://invent.kde.org/plasma/kwin/-/commit/785a9e6aff5099d671190cd83574b73045888337
Comment 15 Vlad Zahorodnii 2026-10-07 10:56:11 UTC
Git commit 3d4845a2f1e4627dadfba1b18211e218581b6829 by Vlad Zahorodnii.
Committed on 07/10/2026 at 10:17.
Pushed by vladz into branch 'Plasma/6.6'.

plugins/screenshot: make the context current after prePaint

prePaint can change the context, and changing it back can fail. The compositor
doesn't hit this issue because it calls beginFrame on the layer after prePaint,
which makes the context current as well.


(cherry picked from commit ad655ed764a61995776abd6b59d98cec1814531f)

Co-authored-by: Xaver Hugl <xaver.hugl@kde.org>

M  +3    -0    src/plugins/screenshot/screenshot.cpp

https://invent.kde.org/plasma/kwin/-/commit/3d4845a2f1e4627dadfba1b18211e218581b6829