| Summary: | Wayland native apps close often when the computer is slower due to high i/o | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | Patrick Silva <bugseforuns> |
| Component: | wayland-generic | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | CC: | fanzhuyifan, me, voidpointertonull+bugskdeorg |
| Priority: | NOR | Keywords: | qt6 |
| Version: | 6.0.2 | ||
| Target Milestone: | --- | ||
| Platform: | Arch Linux | ||
| OS: | Linux | ||
| URL: | https://gitlab.freedesktop.org/wayland/wayland/-/issues/159 | ||
| See Also: | https://bugs.kde.org/show_bug.cgi?id=392376 | ||
| Latest Commit: | Version Fixed In: | ||
|
Description
Patrick Silva
2024-03-25 22:58:06 UTC
It happened again: mar 25 20:06:56 Arch-PC kwin_wayland_wrapper[926]: error in client communication (pid 158020) mar 25 20:06:56 Arch-PC thunderbird[158020]: Lost connection to Wayland compositor. mar 25 20:06:56 Arch-PC plasmashell[158106]: Exiting due to channel error. mar 25 20:06:56 Arch-PC systemd[829]: app-Thunderbird-57cdae19694241bbb1fb043f5c46afa2.scope: Consumed 22.272s CPU time. I would say this is upstream wayland issue: https://gitlab.freedesktop.org/wayland/wayland/-/issues/159 The Wayland-specific part is already covered by #392376 . The issue with background I/O being capable of slowing everything down to almost a halt is a Linux kernel problem with a long history, and it doesn't look like it's getting improvements any soon, but you can surely make it worse by using extra features added over time like Btrfs with really heavy compression. KDE overall is getting quite resilient to it, at least as long as the I/O queue is progressing, not a complete halt as an NFS mount not responding will unfortunately still get many KDE components unusable as there's a whole lot of mount point poking everywhere with questionable necessity. Apparently I can't set CLOSED, so setting DUPLICATE while keeping the RESOLVED status, even though the issue itself isn't resolved. *** This bug has been marked as a duplicate of bug 392376 *** (In reply to Pedro V from comment #3) > The issue with background I/O being capable of slowing everything down to > almost a halt is a Linux kernel problem with a long history, and it doesn't > look like it's getting improvements any soon, but you can surely make it > worse by using extra features added over time like Btrfs with really heavy > compression. I don't think this is really something that can be fixed on the Linux kernel side and is a problem on any OS in some form (though other OSes don't have such a disastrous reaction to it). This is caused by the wayland server killing the socket connection to clients whose event buffer fills up, very often caused by the application doing blocking I/O or other long operations on the UI thread, but could also happen when the system is generally overloaded and the client process doesn't get scheduled in time to empty its event buffer. It has to be fixed in the wayland protocol lib (to handle a full socket buffer more gracefully) and also in applications (don't block the UI thread). |