Ok, weird one, updated to plasma 6.7, any program that tries to use the GPU crashes plasma hard, black screen no desktop, nothing, however having the kwin debug window open prevents the crash, close it it immediately crashes.
Ok, so keeping the debug window open still eventually leads to a crash, this makes me suspect some gpu memory miss-management in the gpu is at fault.
Update 6.7.0-1.1 didn’t fix it, I have no snapper to fallback to as this is a fresh install.
AMD 5900x
nvidia rtx 3060
edit: add bug-report link
edit2: add more info
edit3: add my hw info
bug-report log: b793cac
from dmesg
kwin_wayland[5158]: segfault at 38 ip 00007f81505d381c sp 00007ffd39d84940 error 4 in libkwin.so.6.7.0[50981c,7f8150232000+5d6000] likely on CPU 1 (core 1, socket 0)
[ 245.235924] Code: c0 ff 15 af 67 41 00 4c 8b 85 50 ff ff ff e9 e8 fc ff ff ff 15 85 94 40 00 90 0f 1f 40 00 f3 0f 1e fa 55 48 89 e5 48 83 c4 80 <8b> 46 38 4c 89 65 f0 64 4c 8b 24 25 28 00 00 00 4c 89 65 d8 49 89
[ 247.672579] NVRM: VM: invalid mmap context
[ 247.672663] NVRM: VM: invalid mmap context
[ 247.672676] NVRM: VM: invalid mmap context
[ 247.672692] NVRM: VM: invalid mmap context
[ 247.672704] NVRM: VM: invalid mmap context
[ 247.672792] NVRM: VM: invalid mmap context
[ 247.672818] NVRM: VM: invalid mmap context
[ 247.672869] NVRM: VM: invalid mmap context
[ 247.672940] NVRM: VM: invalid mmap context
[ 247.672958] NVRM: VM: invalid mmap context
[ 247.673094] NVRM: VM: invalid mmap context
[ 247.673301] NVRM: VM: invalid mmap context
[ 247.673336] NVRM: VM: invalid mmap context
[ 247.673522] NVRM: VM: invalid mmap context
[ 247.673608] NVRM: VM: invalid mmap context
[ 247.673659] NVRM: VM: invalid mmap context
[ 247.673714] NVRM: VM: invalid mmap context
[ 247.673767] NVRM: VM: invalid mmap context
[ 247.673829] NVRM: VM: invalid mmap context
[ 247.673844] NVRM: VM: invalid mmap context
[ 247.673862] NVRM: VM: invalid mmap context
[ 247.673915] NVRM: VM: invalid mmap context
[ 247.673951] NVRM: VM: invalid mmap context
[ 247.673963] NVRM: VM: invalid mmap context
[ 247.674027] NVRM: VM: invalid mmap context
[ 247.674040] NVRM: VM: invalid mmap context
[ 247.674659] NVRM: VM: invalid mmap context
[ 247.674825] NVRM: VM: invalid mmap context
[ 247.674946] NVRM: VM: invalid mmap context
[ 247.675028] NVRM: VM: invalid mmap context
[ 247.675079] NVRM: VM: invalid mmap context
[ 247.675092] NVRM: VM: invalid mmap context
[ 247.675127] NVRM: VM: invalid mmap context
[ 247.675242] NVRM: VM: invalid mmap context
[ 247.675286] NVRM: VM: invalid mmap context
[ 247.675300] NVRM: VM: invalid mmap context
[ 247.676613] NVRM: VM: invalid mmap context
[ 247.676628] NVRM: VM: invalid mmap context
I think I may be hitting the same issue.
My system is CachyOS KDE Plasma Wayland with NVIDIA, currently on:
kernel: 7.0.12-1-cachyos
kwin: 6.7.0-1.1
plasma-desktop: 6.7.0-1.1
linux-cachyos-nvidia-open: 7.0.12-1
GPU: RTX 3090
After the Plasma/KWin update on 2026-06-19, I got repeated kwin_wayland crashes. The first happened after the live update, but the important part is that it reproduced again after reboot/relogin, so it does not look like only a stale in-memory library issue.
The post-reboot coredumps resolve to this stack area:
KWin::DrmOutput::testPresentation(...)
KWin::Compositor::setupLayers(...)
KWin::Compositor::composite(...)
Symptoms after the compositor crash:
yakuake: The Wayland connection broke. Did the Wayland compositor die?
yakuake: There are no outputs - creating placeholder screen
plasmashell: requesting unexisting screen available rect -1
Firefox child processes exit with channel error
On the latest crash I also saw repeated:
NVRM: VM: invalid mmap context
but those appeared after kwin_wayland had already died, so I am not sure whether NVIDIA is the trigger or just part of the fallout.
Full report/logs: c2028bb
Does this match the same Plasma 6.7 / KWin / NVIDIA issue being tracked here? Also, is nvidia_drm.color_pipeline=0 currently the recommended workaround to test on CachyOS?
Hey, do you also have the behavior where it doesn’t crash if debug window is open?
Happens to me too. Console says “did wayland compositor die?”.
I only get a black screen when login in.
I haven’t tested with the debug window open. It happens kind of randomly a few minutes after logging in, and all I have after the crash is the firefox crash reporting window. But that’s because all programs are killed when the desktop environment crashes.
yeah, open the kwin debug console window, it will give you some time of desktop use, for me I can still use the alt+space shortcut
It seems like there was an update to nvidia drivers that fixed this for me.
Maybe a dumb question, but did you reboot or at least logged out and back in after the upgrade to Plasma 6.7 and kwin? Btw there was a newer version of kwin released like a day after the Plasma 6.7 was shipped IIRC, have you updated to the latest 6.7.0-3.1 version of kwin?
Yes. It happened consistently a few minutes after rebooting, for three times, each after a reboot. A few minutes in. Debugging with AI reading the logs it suggests it might be related to my screen blanking after a period of inactivity, which is consistent with my behavior.
I was working on another machine (my or the work macbook) for a while and when I turned back to my desktop I had just the Firefox crash report window open instead of all my other windows. I’m pretty sure Firefox is unrelated other than it popping up the crash report window when the desktop starts back up.
What’s the GPU? Are you using NVidia, Plasma and the screen is high resolution, high frequency monitor? If that’s the case the issue is almost certainly related to either Display Stream Compression or VRR (or the combination of both).
To prevent the “screen blanking” you can either lower the refresh rate to say 144Hz (Assuming 4K display) or depending on your max refresh rate, leaving high refresh rate can be possible if you disable VRR in Plasma settings. I personally don’t see/feel any benefit in VRR at least not in a combo: good enough GPU, high refresh rates. With high refresh rates it’s hard to notice tearing simply because our eyes/brains can’t or barely can notice it. Of course this it’s different with super demanding games/settings where even the best GPUs struggle to render 50 FPS.
This is not screen banking, where one specific screen goes black because of the transmission bandwidth or cable. This is the entire desktop on all screens crashing and taking down all running desktop applications. VS Code, Firefox, Steam, etc. All are gone without a trace, other than Firefox which has the automatic crash reporting service pop up.
Same here. Downgrading NVIDIA to 580xx did not work…
This hasn’t been happening in the last couple of days. Hopefully it was patched in one of the recent updates. Though since I can’t trigger it on demand I can’t be 100% certain it’s fixed. It’s been stable for up to 12h a day until I shut down my PC.
I am experiencing the same issue on KDE/Nvidia setup. The first sign I get is often times I will try to open a web page and the browser will say the tab crashed. Then I just start having cascading failures of anything that is running until I am left with a black screen and no start menu. Attempting to click Restart (before the start menu crashes) does nothing and that is unresponsive at this state. I have switched over to LTS for now. I am unsure if this solved my issue or not since I usually only get this crash about once per day, but hopefully that will help until this issue gets resolved.