Created attachment 103804 [details] strace -t Using Neon Wayland Dev Unstable (git master), kwin_wayland is taking 100 of CPU core, and memory usage wet up to 2G then down again, but CPU stayed 100% (about an hour now), I asked in #kwin and they suggested using strace, so this is some output, something wrong with sendmsg and recvmsg.
Created attachment 103805 [details] strace
Created attachment 103806 [details] strace -c
What might be better is a gdb backtrace of the stuck process. In general it works in neon wayland dev unstable, I'm just writing in such a session.
It happens sometimes, I didn't use gdb before, I used: gdb -p `pidof kwin_wayland` -ex bt -ex 'thread apply all bt full' -ex detach -ex quit &> kwin_wayland_trace to create the attached file, hope it helps, please let me know if there is anything else I can provide
Created attachment 103904 [details] kwin_wayland trace
Difficult to say. We have many "No symbol table info available." which means we don't see the actual code line. What we see is it is dispatching the Wayland server event loop.
Created attachment 103912 [details] trace with debug packages installed I installed all debug packages I could find in the file and generated a new trace.
It hit a completely different place. This time we don't have Wayland server event loop in the backtrace.
It is the same process, still running until now if you want something, maybe different timing can cause different gdb output :)
> It is the same process, still running until now if you want something, > maybe > different timing can cause different gdb output :) sure, sure. But if we have 100 % cpu usage cases, it's quite often, that gdb ends in the same area. Which is why I was interested in gdb in the first place. I just hoped, that it ends up always in the same spot - that would make investigating easier.
Getting this too. On Arch Linux, Plasma 5.9.3, Kwin_wayland suddenly took 100% of one core of my CPU after a few hours of use.
Hi, I'm seeing this also on Fedora 25 x86_64 with Plasma 5.9.4, choosing the wayland sesssion and logging in or starting a nested wayland session shows kwin_wayland using 100% cpu. This on a optimus Acer laptop (intel 530 skylake and nouveau for nvidia gtx950m).
Same here with gentoo, unmasking plasma 5.9.5. Is there any interesting info I could share to ease the debugging? I'm afraid debugging symbols won't be there either. Thanks! José.
What I need is to know where the problem happens. For that I ideally get a backtrace with debug symbols. Without that all is just guessing unfortunately.
(In reply to Martin Flöser from comment #14) > What I need is to know where the problem happens. For that I ideally get a > backtrace with debug symbols. Without that all is just guessing > unfortunately. I'll try to provide another backtrace with debug symbols (I think I already provided one), but the session is very unstable these days (crash when I traverse the menu), so maybe later.
Is this issue still reproducible?
(In reply to Alexander Mentyu from comment #16) > Is this issue still reproducible? I don't think so, at least not for me, I don't mind closing it, it's your call.
It stopped happening here too.
I'm already far from the versions this happened to me too, so not reproducing anymore, it can be closed.
Thanks for the update; closing.