Too many open files crashing kwin_wayland

I have seen multiple crashes in the last 1-2 days, all mentioning too many open files. These crashes kill kwin_wayland with a segfault, and the stack trace always mentions libspa-videoconvert.so, though that might be a red herring.
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x5624a365d5f0
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: Could not create EGL surface (EGL error 0x300b)
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: eglSwapBuffers failed with 0x300d, surface: 0x0
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: error marshalling arguments for import_timeline: dup failed: Too many open files
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: Error marshalling request for wp_linux_drm_syncobj_manager_v1.import_timeline: Too many open files
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x5624a5d98b80
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: The Wayland connection experienced a fatal error: Too many open files
Oct 27 14:28:34 DESKTOP-VV1DUT5 systemd-coredump[82470]: [šŸ”•] Process 1909 (kwin_wayland) of user 1000 dumped core.

Module libspa-videoconvert.so from rpm pipewire-1.4.9-1.fc42.x86_64
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x5624a365d5f0
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: Could not create EGL surface (EGL error 0x300b)
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: eglSwapBuffers failed with 0x300d, surface: 0x0
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: error marshalling arguments for import_timeline: dup failed: Too many open files
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: Error marshalling request for wp_linux_drm_syncobj_manager_v1.import_timeline: Too many open files
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x5624a5d98b80
Oct 27 14:28:29 DESKTOP-VV1DUT5 plasmashell[2080]: The Wayland connection experienced a fatal error: Too many open files
Oct 27 14:28:34 DESKTOP-VV1DUT5 systemd-coredump[82470]: [šŸ”•] Process 1909 (kwin_wayland) of user 1000 dumped core.

Module libspa-videoconvert.so from rpm pipewire-1.4.9-1.fc42.x86_64
70 Replies
AgentElrond
AgentElrondOP•3w ago
In particular, I notice that ulimit -Sn and ulimit -Hn show soft and hard open file limits of 1024 and 1048576, respectively. I want to track my actual open file (descriptor?) count so I can identify which process(es) or driver(s) may be leaking descriptors and causing these crashes. Currently, the only data I have is lsof | wc -l which shows, at the moment, 503755 For anyone who sees this, what do you currently see from lsof | wc -l? I'm curious if that is typical, or way higher than normal. I have Steam, Firefox with several windows, several terminals, a Wine application, and Discord running
CaramelCorn
CaramelCorn•3w ago
I'm showing 745,489 right now
AgentElrond
AgentElrondOP•3w ago
wow, so I guess that's normal That seems uncomfortably close to the 1048576 hard limit, given that it varies substantially depending on what's going on (right now Vulkan shaders are compiling)
CaramelCorn
CaramelCorn•3w ago
yeah I'm not even running anything intensive, just my standard work from home setup of like.. discord, spotify, teams, vs code, terminal, openoffice, bunch of browser tabs
AgentElrond
AgentElrondOP•3w ago
@CaramelCorn before anyone else asks, nVidia or AMD? (I have nvidia)
CaramelCorn
CaramelCorn•3w ago
AMD
AgentElrond
AgentElrondOP•3w ago
I've filtered it by process. The big ones are roughly 40k from steam roughly 40k from plasmashell/krunner over 240k from Firefox 35k from Discord and about 80k from "Isolated Web Container" not sure if that's a Firefox thing or a Bazzite thing A couple hours later, seeing about 520k-550k Ugh, it's getting more frequent I had just checked the open file count and it was around 520k, then within a couple minutes I clicked a taskbar icon System freeze, then kwin_wayland crash, too many open files coredump After the KDE restart, now it's showing around 260k open file descriptors, random note
CaramelCorn
CaramelCorn•3w ago
is it clicking the Taskbar icon every time that does it?
AgentElrond
AgentElrondOP•3w ago
POSSIBLY, that's when I usually notice it I go to click a (stacked?) icon to expand it, then the system freezes for a few seconds while still showing windows, then kwin_wayland dies and kills all applications I see 2-3 coredump processes running after the fact, so it may also be that there are multiple coredumps. Is it safe to kill those processes? It takes forever because my hard drive is slow, so I'd prefer to just skip them It sucks that the Vulkan shader compilation completely got its progress reset due to the kwin_wayland crash. 😭 And it takes hours on my system. It looks like clicking "Cancel" also resets Vulkan compilation. Only "Skip" might preserve existing progress? Next time it crashes, I'll try to do a more thorough log scan to see if there are 2 or 3 segfaults 3 hours later, I'm back up at almost 500k open files. I do have Vulkan shaders compiling yet again. Just recording data points.
ēš‡å®¶ęµ·ę˜Ÿ
shot in the dark, but try turning off file indexing
AgentElrond
AgentElrondOP•3w ago
I didn't know Bazzite did that, I'll search how to do so
ēš‡å®¶ęµ·ę˜Ÿ
it's in kde settings, Search -> File Search or just type index into the search box
AgentElrond
AgentElrondOP•3w ago
Thanks, I see baloo_file stopped running, but baloorunner is still active
ēš‡å®¶ęµ·ę˜Ÿ
see if it persists after a restart
AgentElrond
AgentElrondOP•3w ago
I'll let this sit for a while, currently at 400k open files while compiling Vulkan ooooooof even clicking "Skip" resets progress, so it seems there is literally no way to compile Vulkan shaders without it running 100% to completion without a KDE crash New day, Steam downloaded an update (to ProtonGE 10-17? to their own shaders? no idea) so I'm attempting to make it through Vulkan again without a crash. 333k open files as of this message, will monitor through the day 7 hours later, 450k open files, no crashes today I've started using konsole instead of ptyxis because I heard konsole can survive a KDE compositor crash This happened again today, too many open files running a few terminals, Firefox, System Monitor Steam with shader compilation kwin_wayland crash, lost all apps including konsole terminals that I hoped would survive compositor handoff
Nov 01 15:16:24 plasmashell[2053]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x55cefe166ac0
Nov 01 15:16:24 plasmashell[2053]: The Wayland connection experienced a fatal error: Too many open files
Nov 01 15:16:24 systemd[1713]: plasma-plasmashell.service: Main process exited, code=exited, status=255/EXCEPTION
Nov 01 15:16:24 systemd[1713]: plasma-plasmashell.service: Failed with result 'exit-code'.
Nov 01 15:16:24 plasmashell[2053]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x55cefe166ac0
Nov 01 15:16:24 plasmashell[2053]: The Wayland connection experienced a fatal error: Too many open files
Nov 01 15:16:24 systemd[1713]: plasma-plasmashell.service: Main process exited, code=exited, status=255/EXCEPTION
Nov 01 15:16:24 systemd[1713]: plasma-plasmashell.service: Failed with result 'exit-code'.
Nov 01 15:16:29 systemd-coredump[73248]: [šŸ”•] Process 1840 (kwin_wayland) of user 1000 dumped core.

Module libmpg123.so.0 from rpm mpg123-1.32.10-1.fc42.x86_64
Module libopus.so.0 from rpm opus-1.5.2-2.fc42.x86_64
Module libvorbisenc.so.2 from rpm libvorbis-1.3.7-12.fc42.x86_64
Module libFLAC.so.12 from rpm flac-1.4.3-6.fc42.x86_64
Module libgsm.so.1 from rpm gsm-1.0.22-8.fc42.x86_64
Nov 01 15:16:29 systemd-coredump[73248]: [šŸ”•] Process 1840 (kwin_wayland) of user 1000 dumped core.

Module libmpg123.so.0 from rpm mpg123-1.32.10-1.fc42.x86_64
Module libopus.so.0 from rpm opus-1.5.2-2.fc42.x86_64
Module libvorbisenc.so.2 from rpm libvorbis-1.3.7-12.fc42.x86_64
Module libFLAC.so.12 from rpm flac-1.4.3-6.fc42.x86_64
Module libgsm.so.1 from rpm gsm-1.0.22-8.fc42.x86_64
Bizarrely, a segfault shows up in libFLAC when I wasn't listening to any music, not sure what's up with that
Nov 01 15:16:29 systemd[1713]: app-org.kde.konsole@8c134d5567c24009aa4505277e0399b1.service: Main process exited, code=exited, status=1/FAILURE
Nov 01 15:16:29 systemd[1713]: app-org.kde.konsole@8c134d5567c24009aa4505277e0399b1.service: Failed with result 'exit-code'.
Nov 01 15:16:29 systemd[1713]: app-flatpak-org.kde.konsole-1124774152.scope: Consumed 1min 24.571s CPU time, 448.9M memory peak.
Nov 01 15:16:29 systemd[1713]: app-org.kde.konsole@8c134d5567c24009aa4505277e0399b1.service: Main process exited, code=exited, status=1/FAILURE
Nov 01 15:16:29 systemd[1713]: app-org.kde.konsole@8c134d5567c24009aa4505277e0399b1.service: Failed with result 'exit-code'.
Nov 01 15:16:29 systemd[1713]: app-flatpak-org.kde.konsole-1124774152.scope: Consumed 1min 24.571s CPU time, 448.9M memory peak.
And here's konsole dying. I wonder if it being a flatpak instead of a normal installed application prevents the compositor handoff?
renner
renner•3w ago
need the stuff further down
AgentElrond
AgentElrondOP•3w ago
below which section? That's the top of the crash I think
renner
renner•3w ago
I need a stack trace it doesn't have debug symbols but ti's fine
AgentElrond
AgentElrondOP•3w ago
Nov 01 15:09:20 audit[71671]: USER_AUTH pid=71671 uid=1000 auid=1000 ses=3 subj=unconfined_u:unconfined_r:chkpwd_t:s0-s0:c0.c1023 msg='op=PAM:unix_chkpwd acct="elro>
Nov 01 15:09:20 kscreenlocker_greet[66370]: qt.qpa.wayland: Could not create EGL surface (EGL error 0x3003)
Nov 01 15:09:20 kscreenlocker_greet[66370]: qt.qpa.wayland: Could not create EGL surface (EGL error 0x3003)
Nov 01 15:09:20 kscreenlocker_greet[66370]: Failed to write to the pipe: Bad file descriptor.
Nov 01 15:09:25 plasmashell[2053]: kpipewire_logging: Could not connect to stream
Nov 01 15:09:25 kernel: [drm:nv_drm_semsurf_fence_wait_ioctl [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Attempt to wait on invalid sync FD: -24
Nov 01 15:09:25 plasmashell[2053]: qt.qpa.wayland: eglSwapBuffers failed with 0x300c, surface: 0x55cefac292c0
Nov 01 15:09:43 systemd[1]: fprintd.service: Deactivated successfully.
Nov 01 15:09:43 audit[1]: SERVICE_STOP pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=fprintd comm="systemd" exe="/usr/lib/sy>
Nov 01 15:09:43 audit: BPF prog-id=218 op=UNLOAD
Nov 01 15:09:48 plasmashell[2053]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x55cf05dfa1e0
Nov 01 15:09:54 flatpak[54909]: 15:09:54.791 › [WindowVisibilityVideoManager] Incoming video enabled changed, incomingVideoEnabled = true
Nov 01 15:09:20 audit[71671]: USER_AUTH pid=71671 uid=1000 auid=1000 ses=3 subj=unconfined_u:unconfined_r:chkpwd_t:s0-s0:c0.c1023 msg='op=PAM:unix_chkpwd acct="elro>
Nov 01 15:09:20 kscreenlocker_greet[66370]: qt.qpa.wayland: Could not create EGL surface (EGL error 0x3003)
Nov 01 15:09:20 kscreenlocker_greet[66370]: qt.qpa.wayland: Could not create EGL surface (EGL error 0x3003)
Nov 01 15:09:20 kscreenlocker_greet[66370]: Failed to write to the pipe: Bad file descriptor.
Nov 01 15:09:25 plasmashell[2053]: kpipewire_logging: Could not connect to stream
Nov 01 15:09:25 kernel: [drm:nv_drm_semsurf_fence_wait_ioctl [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Attempt to wait on invalid sync FD: -24
Nov 01 15:09:25 plasmashell[2053]: qt.qpa.wayland: eglSwapBuffers failed with 0x300c, surface: 0x55cefac292c0
Nov 01 15:09:43 systemd[1]: fprintd.service: Deactivated successfully.
Nov 01 15:09:43 audit[1]: SERVICE_STOP pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=fprintd comm="systemd" exe="/usr/lib/sy>
Nov 01 15:09:43 audit: BPF prog-id=218 op=UNLOAD
Nov 01 15:09:48 plasmashell[2053]: qt.qpa.wayland: eglSwapBuffers failed with 0x3000, surface: 0x55cf05dfa1e0
Nov 01 15:09:54 flatpak[54909]: 15:09:54.791 › [WindowVisibilityVideoManager] Incoming video enabled changed, incomingVideoEnabled = true
This is also weird, it's showing an nvidia driver error with invalid sync file descriptor, maybe because of too many open files? And also video, maybe that's due to Discord? I didn't see any video playing anywhere
renner
renner•3w ago
also make a new user account and see if you can reproduce the issue there? Nov 01 15:09:25 kernel: [drm:nv_drm_semsurf_fence_wait_ioctl [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] I thought you were on AMD
AgentElrond
AgentElrondOP•3w ago
no, nvidia I'm aware it has issues, just trying to identify if it's causing this out of files error specifically
renner
renner•3w ago
No description
AgentElrond
AgentElrondOP•3w ago
I meant to say I have nvidia
renner
renner•3w ago
delete the last 2, useless
AgentElrond
AgentElrondOP•3w ago
let me try to get the entire stack
AgentElrond
AgentElrondOP•3w ago
the only other crash dump I see is irefox, I assume crashing because everything else died A bunch of stuff in the middle of that stack trace references plasma, cursor, decorations, Qt etc. so I'm wondering if it's related to me clicking the taskbar icon stack twice and triggering something
renner
renner•3w ago
no idea do you have a reliable way to reproduce the issue
AgentElrond
AgentElrondOP•3w ago
"Too many open files" happens just BEFORE the crash and mentions wp_linux_drm_syncobj_manager_v1.import_timeline Sadly no, I think every time this specific crash has happened, with "Too many open files", it's been right after I click on the taskbar to change between applications. And that's likely always a stacked set of identical windows, e.g. a set of terminals.
renner
renner•3w ago
it would be nice if you make a new user account and try to reproduce the issue third party plasmoids?
AgentElrond
AgentElrondOP•3w ago
I could do that but I'd lose all customization -- may I respectfully ask what might be linked to my accout which could affect things? No plasma/KDE customization that I'm aware of I've removed widgets like the media player which had crash reports, disabled the desktop background, minimized as much as I could
renner
renner•3w ago
ruling out configuration, makes reproduction 100 times easier
AgentElrond
AgentElrondOP•3w ago
The default configuration crashed repeatedly for me, sadly
renner
renner•3w ago
then it's just bazzite defaults
AgentElrond
AgentElrondOP•3w ago
By default, the taskbar has a media player widget, I have crash logs elsewhere I think I can go dig it up
renner
renner•3w ago
rpm-ostree status
AgentElrond
AgentElrondOP•3w ago
It seemed to be a known issue supposed to be replaced in a later Plasma version, with the workaround being, remove the systray media player widget I'm base Bazzite 42, holding off upgrading until 44 or later
State: idle
Deployments:
ā— ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-nvidia-open:stable
Digest: sha256:5d86b5f6db5361d4c74f4ed2c9503a694502d91e61f872ceb3c93e48503f2401
Version: 42.20251019 (2025-10-19T16:29:32Z)
Pinned: yes

ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-nvidia-open:stable
Digest: sha256:33b539973873e3ea41aa417c71121b12a7669c7249e02a77e5cb23fbf4229f66
Version: 42.20250916.1 (2025-09-16T17:32:52Z)
elrond@DESKTOP-VV1DUT5:/var/home/elrond$
State: idle
Deployments:
ā— ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-nvidia-open:stable
Digest: sha256:5d86b5f6db5361d4c74f4ed2c9503a694502d91e61f872ceb3c93e48503f2401
Version: 42.20251019 (2025-10-19T16:29:32Z)
Pinned: yes

ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-nvidia-open:stable
Digest: sha256:33b539973873e3ea41aa417c71121b12a7669c7249e02a77e5cb23fbf4229f66
Version: 42.20250916.1 (2025-09-16T17:32:52Z)
elrond@DESKTOP-VV1DUT5:/var/home/elrond$
renner
renner•3w ago
update your system
AgentElrond
AgentElrondOP•3w ago
Due to known issues and me needing the system for work, that may not be feasible until Bazzite 44, but if I do I'll see if I notice the same crash
renner
renner•3w ago
respectfully, I don't think you know what that means
AgentElrond
AgentElrondOP•3w ago
Can you clarify which part I have decades of Linux experience, just new to Bazzite, so apologies if I use the wrong terms Which part you feel I misunderstand, that is.
renner
renner•3w ago
the fedora version bazzite is based on shouldn't matter, F44 comes out in 6 months there was a media player widget crash caused by chromium
AgentElrond
AgentElrondOP•3w ago
I meant Bazzite, not Fedora
renner
renner•3w ago
that was resolved in kde and you get the fix by upgrading
AgentElrond
AgentElrondOP•3w ago
Point is, I will likely wait a week or two to upgrade for the next Bazzite push, not sure yet Links for my own / others' reference: https://github.com/NixOS/nixpkgs/issues/442630 https://bugreports.qt.io/browse/QTBUG-140018 https://bugs.kde.org/show_bug.cgi?id=509192 I can confirm Bazzite 42 has Plasma 6.4.5, Qt 6.9.2. The bugs.kde.or thread says fixed in Plasma 6.4.6 and later, Qt 6.10.1 and later. I'm not sure which versions Bazzite 43 uses I'll have to dig up the change log
renner
renner•3w ago
have you actually upgraded your system and looked if it worked rpm -qi plasma-desktop qt6-qtbase should be 6.4.5 and 6.9.3 or are you purely speculating on this you can rollback if it's broken
AgentElrond
AgentElrondOP•3w ago
I have not tried Bazzite 43 on my system yet, but due to schedule and work needs I wanted to avoid experimenting for a few days unless crashes are too frequent. I'm aware rollbacks exist, yes. 6.4.5 and 6.9.2 in Bazzite 42, also shown in the info screen https://discord.com/channels/1072614816579063828/1087140957096517672/1434266666325512193 looks like 6.4.5 and 6.9.2 are also the versions in Bazzite 43, per that user's screenshot So I''ll watch for Plasma 6.4.6 and Qt 6.10.1 becoming available in a Bazzite release, then give those a shot.
renner
renner•3w ago
yes, I'm not sure if fedora backported the patch it will report the same version
AgentElrond
AgentElrondOP•3w ago
touche I'm not aware either, but thanks for your patience. I have to step away.
renner
renner•3w ago
the next bazzite release will have 6.5 and QT 6.10 it hit fedora repos today
AgentElrond
AgentElrondOP•3w ago
awesome
AgentElrond
AgentElrondOP•3w ago
NVIDIA Developer Forums
Fd leak with explicit sync and kde plasma
I can trivially reproduce this as well, by sending notifications with notify-send . I am on 570.86.16 and KDE Plasma 6.3. The Upstream KDE issue here: 497424 – fd leak with explicit sync (nvidia) , which was linked in the OP, claims this is an Nvidia driver bug. Please let me know if any other information would be useful
AgentElrond
AgentElrondOP•3w ago
ah heck, this sounds eerily familiar Yeah, I just tried clicking on and off a stacked Firefox taskbar icon a bunch of times, and promptly saw lsof -p $(pidof plasmashell) | grep sync_file | wc -l output jump by 100 this seems bad Looks like nVidia made an internal bug track, suggested that egl-wayland2 fixed things, but someone reported it didn't (for them). So if anyone wants to test this, you can run lsof -p $(pidof plasmashell) | grep sync_file | wc -l after mousing over a taskbar icon It seems to add 3 new handles for me, every time I move my mouse over a taskbar icon. Update: It doesn't seem exact, I see 2-5 new handles each time I mouse over a taskbar icon. Important thing, you have to wait until the tooltip appears. I don't have graphical previews enabled, but it still seems to leak descriptors if the tooltip appears. Alt-tab window previews don't affect the file descriptor count, only the taskbar tooltips I'm attempting ujust update today due to the new Plasma version. I'll test the taskbar behavior if it succeeds.
AgentElrond
AgentElrondOP•3w ago
No description
AgentElrond
AgentElrondOP•3w ago
nVidia driver version is still 580.95.05 file descriptor leak still exists, I immediately saw 3-5 new descriptors created every time I hover over a taskbar icon
lsof -p $(pidof plasmashell) | grep sync_file | wc -l
lsof -p $(pidof plasmashell) | grep sync_file | wc -l
I would still love if anyone with an nVidia card (or even AMD) wants to test the above. The process is still: 1. Run the command. 2. Hover your mouse over a few taskbar icons, waiting each time for a tooltip to appear. 3. Run the command again. Did the file descriptor count increase? Update: I confirmed that the leaking file descriptors do indeed correspond to sync_file (see https://www.kernel.org/doc/html/v6.18-rc3/driver-api/sync_file.html). This is a kernel interface so it would indeed be possible for the nVidia driver to be leaking these. What's curious to me is that they only seem to leak in the very specific situation of a taskbar tooltip popup (or maybe more generally, a plasmoid opening, but I haven't messed around with that much). I wonder what API calls the taskbar popups use that other logic (like alt-tab) doesn't. Update: Also from the nVidia dev forum thread, I noticed that the leak is more general and happens ANY time a popup is produced from KDE, e.g. by using the command notify-send test So anyone who often switches applications using the taskbar in KDE with nVidia, or often sees notifications, is presumably going to crash eventually. I'm surprised this isn't a more common problem, unless people just don't click stacked taskbar icons or get notifications much.
š•¾š–™š–Šš–‘š–‘š–†
mine just stays at 0
No description
AgentElrond
AgentElrondOP•3w ago
nvidia or AMD? (thanks so much for testing, by the way)
š•¾š–™š–Šš–‘š–‘š–†
Nvidia 1660TI on Nvidia-open, 43.20251102
AgentElrond
AgentElrondOP•3w ago
Some nvidia rep posted to ask if egl-wayland2 fixes it, but a couple replies said it didn't, so I didn't bother trying. you're KDE, I assume?
AgentElrond
AgentElrondOP•3w ago
I'm trying to find out how to get the 43.20251102 value you did
š•¾š–™š–Šš–‘š–‘š–†
i suppose by rebooting or updating manually
AgentElrond
AgentElrondOP•3w ago
I notice one (key?) difference, you have an integrated graphics card. I wonder if KDE is somehow using that for plasmoid stuff. Since we have the same Bazzite version I assume we have the same nvidia driver version, I didn't layer anything
š•¾š–™š–Šš–‘š–‘š–†
wait didn't you just say that your image is slightly out of date?
AgentElrond
AgentElrondOP•3w ago
rpm ostree status works I upgraded to the latest today because of the new Plasma version. so yeah I'm now 43.20251102
AgentElrond
AgentElrondOP•3w ago
Let me pin an update in case anyone else helpful comes along
š•¾š–™š–Šš–‘š–‘š–†
seems like it's only running steam and distroshelf on the Nvidia card :ThonkFedora:
No description
AgentElrond
AgentElrondOP•2w ago
This is the original test I'm chasing, that results in a KDE crash over time, just from using the GUI. https://discord.com/channels/1072614816579063828/1432840141101338654/1434931367069749312 I only have 1 GPU (GTX 1650), and I see the handles constantly increase after a notification. Another use has an integrated Intel and dedicated nVidia, and they don't see any handles at all. oh I can't pin lol nice @š•¾š–™š–Šš–‘š–‘š–† I don't know how to check if you're using explicit sync, but I assume Bazzite always does by default Working theory seems to be, this happens due to a nVidia driver bug involving explicit sync. I just had futile hopes I could get some workaround from the KDE side. Silly workaround for now: Once I notice the open file descriptor count getting dangerously high, like over 600, I can just run plasmashell --replace & or whatever to start a new instance Almost all the kernel sync_file descriptors get dropped at that point, now I'm back to 17 open descriptors šŸ™Œ This should at least extend my uptime. uptime is 4 days so far, with this workaround I'm still hoping the nVidia rep eventually answers on the forum I actually got a brief reply from nVidia. Apparently engineering is still working on this bug.

Did you find this page helpful?