SUMMARY There is a memory leak in plasma that causes it to eventually consume an unbounded quantity of memory, resulting in system slowness / potential instability. STEPS TO REPRODUCE 1. Run plasmashell (i.e. "the KDE desktop") 2. Use the computer for several days OBSERVED RESULT The plasmashell process is consuming many GB of memory (16GB, 32GB, more...) and is >45% RES. EXPECTED RESULT Memory usage is relatively low and stable. SOFTWARE/OS VERSIONS bash-4.4.23-5.fc29.x86_64 coreutils-8.30-6.fc29.x86_64 dbus-x11-1.12.10-1.fc29.x86_64 glibc-2.28-26.fc29.x86_64 iso-codes-3.79-2.fc29.noarch kactivitymanagerd-5.14.4-1.fc29.x86_64 kde-settings-plasma-29.0-1.fc29.noarch kf5-attica-5.53.0-1.fc29.x86_64 kf5-baloo-5.53.0-1.fc29.x86_64 kf5-baloo-libs-5.53.0-1.fc29.x86_64 kf5-filesystem-5.53.0-1.fc29.x86_64 kf5-frameworkintegration-5.53.0-3.fc29.x86_64 kf5-kactivities-5.53.0-1.fc29.x86_64 kf5-kauth-5.53.0-1.fc29.x86_64 kf5-kbookmarks-5.53.0-1.fc29.x86_64 kf5-kcodecs-5.53.0-1.fc29.x86_64 kf5-kcompletion-5.53.0-1.fc29.x86_64 kf5-kconfig-core-5.53.0-1.fc29.x86_64 kf5-kconfig-gui-5.53.0-1.fc29.x86_64 kf5-kconfigwidgets-5.53.0-1.fc29.x86_64 kf5-kcoreaddons-5.53.0-1.fc29.x86_64 kf5-kcrash-5.53.0-1.fc29.x86_64 kf5-kdbusaddons-5.53.0-1.fc29.x86_64 kf5-kdeclarative-5.53.0-3.fc29.x86_64 kf5-kded-5.53.0-1.fc29.x86_64 kf5-kdelibs4support-libs-5.53.0-1.fc29.x86_64 kf5-kdoctools-5.53.0-1.fc29.x86_64 kf5-kfilemetadata-5.53.0-1.fc29.x86_64 kf5-kglobalaccel-5.53.0-1.fc29.x86_64 kf5-kglobalaccel-libs-5.53.0-1.fc29.x86_64 kf5-kguiaddons-5.53.0-1.fc29.x86_64 kf5-ki18n-5.53.0-1.fc29.x86_64 kf5-kiconthemes-5.53.0-1.fc29.x86_64 kf5-kinit-5.53.0-1.fc29.x86_64 kf5-kio-core-libs-5.53.0-1.fc29.x86_64 kf5-kio-file-widgets-5.53.0-1.fc29.x86_64 kf5-kio-widgets-libs-5.53.0-1.fc29.x86_64 kf5-kitemmodels-5.53.0-1.fc29.x86_64 kf5-kitemviews-5.53.0-1.fc29.x86_64 kf5-kjobwidgets-5.53.0-1.fc29.x86_64 kf5-knewstuff-5.53.0-1.fc29.x86_64 kf5-knotifications-5.53.0-1.fc29.x86_64 kf5-kpackage-5.53.0-1.fc29.x86_64 kf5-kparts-5.53.0-1.fc29.x86_64 kf5-krunner-5.53.0-1.fc29.x86_64 kf5-kservice-5.53.0-1.fc29.x86_64 kf5-ktexteditor-5.53.0-1.fc29.x86_64 kf5-ktextwidgets-5.53.0-1.fc29.x86_64 kf5-kunitconversion-5.53.0-1.fc29.x86_64 kf5-kwayland-5.53.0-3.fc29.x86_64 kf5-kwidgetsaddons-5.53.0-1.fc29.x86_64 kf5-kwindowsystem-5.53.0-1.fc29.x86_64 kf5-kxmlgui-5.53.0-3.fc29.x86_64 kf5-kxmlrpcclient-5.53.0-1.fc29.x86_64 kf5-plasma-5.53.0-1.fc29.x86_64 kf5-prison-5.53.0-1.fc29.x86_64 kf5-solid-5.53.0-1.fc29.x86_64 kf5-sonnet-ui-5.53.0-1.fc29.x86_64 khotkeys-5.14.4-2.fc29.x86_64 kscreenlocker-5.14.4-1.fc29.x86_64 ksysguardd-5.14.4-1.fc29.x86_64 libICE-1.0.9-14.fc29.x86_64 libSM-1.2.3-1.fc29.x86_64 libX11-1.6.7-1.fc29.x86_64 libXext-1.3.3-10.fc29.x86_64 libXrender-0.9.10-8.fc29.x86_64 libXtst-1.2.3-8.fc29.x86_64 libgcc-8.2.1-6.fc29.x86_64 libksysguard-5.14.4-1.fc29.x86_64 libkworkspace5-5.14.4-1.fc29.x86_64 libstdc++-8.2.1-6.fc29.x86_64 libxcb-1.13.1-1.fc29.x86_64 oxygen-sound-theme-5.14.4-1.fc29.noarch phonon-qt5-4.10.1-2.fc29.x86_64 plasma-desktop-5.14.4-2.fc29.x86_64 plasma-lookandfeel-fedora-5.14.4-1.fc29.noarch plasma-milou-5.14.4-1.fc29.x86_64 plasma-pa-5.14.4-1.fc29.x86_64 plasma-workspace-5.14.4-1.fc29.x86_64 plasma-workspace-common-5.14.4-1.fc29.x86_64 plasma-workspace-libs-5.14.4-1.fc29.x86_64 polkit-kde-5.14.4-1.fc29.x86_64 powerdevil-5.14.4-1.fc29.x86_64 qt5-qtbase-5.11.3-1.fc29.x86_64 qt5-qtbase-gui-5.11.3-1.fc29.x86_64 qt5-qtdeclarative-5.11.3-1.fc29.x86_64 qt5-qtgraphicaleffects-5.11.3-1.fc29.x86_64 qt5-qtquickcontrols-5.11.3-1.fc29.x86_64 qt5-qttools-5.11.3-1.fc29.x86_64 qt5-qtx11extras-5.11.3-1.fc29.x86_64 socat-1.7.3.2-7.fc29.x86_64 systemd-239-8.gite339eae.fc29.x86_64 xcb-util-0.4.0-11.fc29.x86_64 xcb-util-image-0.4.0-11.fc29.x86_64 xorg-x11-apps-7.7-22.fc29.x86_64 xorg-x11-server-utils-7.7-26.fc29.x86_64 xorg-x11-utils-7.5-29.fc29.x86_64 zlib-1.2.11-14.fc29.x86_64 ADDITIONAL INFORMATION I'm using slideshow wallpaper on a 5 minute rotation, 2.5 screens (one "screen" is 1080p entirely within the bounds of a 1920x1200, but I've noticed plasma sometimes seems to use it as a separate screen). This seems like #368838 again, but is still happening despite that bug supposedly being "fixed". (Note: I was advised on that bug to open a new report.)
>Note: I was advised on that bug to open a new report. Yes, thanks. I'm afraid I'm going to need a lot of info to narrow this down. 1) Does it mostly go away if you don't have a slideshow wallpaper? 2) Can I have the output of: qdbus org.kde.KWin /KWin org.kde.KWin.supportInformation 3) If you run "kcmshell5 qtquicksettings" does it go away if you set Render loop is set to "basic" and run "plasmashell --replace" 4) If you run "kcmshell5 qtquicksettings" does it go away if rendering backend is set to "software" and run "plasmashell --replace"
Created attachment 117657 [details] kwin info Unfortunately, it's going to be hard to answer "does it go away" questions, since it usually takes at least a few days before the issue becomes apparent. At best, I can try experiments and report back if some weeks pass without issue. Kwin info is attached.
Okay... after initially making the first change (render loop = basic), it ran fine for a while, then X crashed. This morning, however: virt: 15.2G res: 11.8G shr: 3.8M mem: 37.6% time: 1h23m Is it possible there is something different about the instance that's auto-started? I'm not entirely certain but what I may only be seeing problems with the initial auto-started instance...
...and again: v: 14.9G r: 13.0G m: 41% This was with basic/software.
Information was provided with comment #2; changing status for inspection.
Can you attach your ~/.config/plasmarc ~/.config/plasmashellrc ~/.config/plasma-org.kde.plasma.desktop-appletsrc
Created attachment 118227 [details] plasmarc
Created attachment 118228 [details] plasmashellrc
Created attachment 118229 [details] plasma-org.kde.plasma.desktop-appletsrc
Dear Bug Submitter, This bug has been in NEEDSINFO status with no change for at least 15 days. Please provide the requested information as soon as possible and set the bug status as REPORTED. Due to regular bug tracker maintenance, if the bug is still in NEEDSINFO status with no change in 30 days the bug will be closed as RESOLVED > WORKSFORME due to lack of needed information. For more information about our bug triaging procedures please read the wiki located here: https://community.kde.org/Guidelines_and_HOWTOs/Bug_triaging If you have already provided the requested information, please mark the bug as REPORTED so that the KDE team knows that the bug is ready to be confirmed. Thank you for helping us make KDE software even better for everyone!
Requested information was provided AFAICT.
Hi. I see that there are many similar reports about Plasmashell Memory Leak problem occurring even to this day. I don't use slideshow wallpaper. I never power off my PC, I always sleep/suspend to RAM. I can suspend to RAM and resume the PC several times without plasmashell memory leak problem happening, but suddenly sometime when I resume the PC the problem starts to occur and I can notice it when I resume from suspend to RAM and I get a little freeze in Plasma until it does it becomes responsive after a few seconds. So from this point I start to check Plasmashell RAM usage, and from this point the problem after suspending RAM and Resume is always reproducible. From this point every time you suspend to RAM and Resume, plasmashell increases RAM usage up to a lot and plasma becomes slower every time. I'm not sure how to debug the problem when it occurs in order to provide some useful information.
Bump x5. I'd appreciate if this bug takes priority - it might be causing other accidental threads being created in relation to memory leaks and mistaking side effects in other places (such as I did with plasma widgets). It makes it impossible to keep a desktop linux distro running KDE for several days and it's making the desktop environment give a really poor impression and experience.
Soooo ... is this going to be fixed or does your users have to change to another DE? I mean it's 3 years now and I still get memory leaks on a daily basis (with a sliding desktop wallpaper - not sure why this memory leak issue isn't fixed after 8 years...)
Will ever this bug will be fixed ? It is really annoying when you have to restart your computer several times a day because plasmashell has a huge memory leak... Isn't is possible to mark this bug as important, not as "normal" priority ? It litterally ruins the user experience. I personnally don't use the slideshow background but I have these memory leaks anyways.
This is the issue that comes up when Googling "plasmashell memory leak", so I'll post here. There are usually multiple root causes of these types of problems, but without better ways to debug / investigate these issues than "try it with this configuration" it is really hard for users to determine whether we are talking about the same issue or not. On my Fedora 36 system with 128 GB of real memory, systemd-oomd has killed plasmashell twice in two days for using almost 133 GB of memory (real memory + considerable swap). There is no conceivable situation in which plasmashell should be using that much memory. Oct 23 03:04:40 edison systemd-oomd[1888]: Killed /user.slice/user-1000.slice/user@1000.service/session.slice/plasma-plasmashell.service due to memory used (132956889088) / total (134229807104) and swap used (27105050624) / total (30064758784) being more than 90.00% Oct 23 03:04:40 edison systemd[2255]: plasma-plasmashell.service: systemd-oomd killed 199 process(es) in this unit. and then almost exactly 24 hours later: Oct 24 03:05:46 edison systemd-oomd[1888]: Killed /user.slice/user-1000.slice/user@1000.service/session.slice/plasma-plasmashell.service due to memory used (132685852672) / total (134229807104) and swap used (27061297152) / total (30064758784) being more than 90.00% Oct 24 03:05:46 edison systemd[2255]: plasma-plasmashell.service: systemd-oomd killed 156 process(es) in this unit. I do find it suspicious that both instances happened at almost exactly the same time overnight -- the cron log does not show any process triggering at or around that time. If there is a way to effectively debug this sort of thing, that would be helpful. I don't use slideshow wallpaper. Output of qdbus supportInformation: https://invent.kde.org/-/snippets/2387
Created attachment 159213 [details] Settings an KWin support info For the past few weeks plasmashell (version plasma-workspace-x11-5.27.4.1-1.fc37.x86_64 from fedora RPMs) has been growing to tens of gigabytes while idle and getting killed by the oomkiller. I've attached my ~/config/plasma*rc and the kwin support info as requested above. I've tried setting the Render loop to basic and will see if that changes anything.
Well that didn't help: top - 00:29:30 up 8:52, 6 users, load average: 3.24, 2.35, 2.20 Tasks: 412 total, 2 running, 410 sleeping, 0 stopped, 0 zombie %Cpu(s): 1.9 us, 7.7 sy, 0.0 ni, 26.6 id, 63.3 wa, 0.3 hi, 0.3 si, 0.0 st MiB Mem : 64229.3 total, 380.3 free, 63408.5 used, 440.5 buff/cache MiB Swap: 69618.0 total, 22149.9 free, 47468.1 used. 141.8 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 76 root 20 0 0 0 0 R 38.5 0.0 22:44.61 kswapd0 114347 root 20 0 25.4g 39608 25700 S 7.6 0.1 1:32.63 Xorg 114631 jwakely 20 0 1515740 20804 15968 S 2.7 0.0 0:04.83 kded5 3092 jwakely 9 -11 473512 12616 8144 S 2.3 0.0 0:02.24 wireplumber 114634 jwakely 20 0 1323284 53276 44564 D 2.3 0.1 1:02.43 kwin_x11 115789 jwakely 20 0 3071832 86556 30860 S 2.0 0.1 3:54.37 Isolated Web Co 115756 jwakely 20 0 3063488 87384 30896 S 1.7 0.1 3:52.92 Isolated Web Co 118895 jwakely 20 0 100.4g 54.9g 33756 D 1.7 87.5 30:56.40 plasmashell It's about 4 hours since I restarted plasmashell (then left the system idle), and it's reached 100gb and the system is performing badly due to swapping. I'll try software rendering tomorrow.
> I'll try software rendering tomorrow. That didn't make any difference. The last time I had the problem the plasmashell process grew to 78gb in about 20 minutes. But I've noticed that when I turn off my display and watch plasmashell using `top` in an ssh session on another machine, it spikes to 100% CPU usage every 15-20 seconds. When the display is on, that doesn't happen, it stays at about 0% all the time. This observation seems correlated to the fact that the memory leaks happen when I'm away from the machine and it's idle, because I always turn the screen off when I leave the desk.
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 22860 jwakely 20 0 2048664 399404 187656 S 96.3 0.6 0:32.10 QSGRenderThread 3222 jwakely 20 0 2048664 399404 187656 S 0.3 0.6 0:07.03 plasmashell 3238 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.17 QDBusConnection 3239 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.51 QXcbEventQueue 3352 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.31 QQmlThread 3363 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.31 Qt bearer threa 3508 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:01.50 QQuickPixmapRea 3577 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.00 CPMMListener 3588 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.07 plasmashell 3728 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.00 KCupsConnection 4218 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.65 QSGRenderThread 8537 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.34 QSGRenderThread 21405 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.05 QSGRenderThread 21413 jwakely 20 0 2048664 399404 187656 S 0.0 0.6 0:00.38 QSGRenderThread That thread is running code from the nvidia drivers, so maybe that's where the problem is: Thread 14 (Thread 0x7fdd7d5ba6c0 (LWP 22860) "QSGRenderThread"): #0 0x00007fddcc52221f in __GI___poll (fds=0x7fdd7d5b95b8, nfds=1, timeout=1000) at ../sysdeps/unix/sysv/linux/poll.c:29 #1 0x00007fddb2912a50 in () at /lib64/libnvidia-glcore.so.530.41.03 #2 0x00007fddb29c3e51 in () at /lib64/libnvidia-glcore.so.530.41.03 #3 0x00007fddb9aefe1c in () at /lib64/libGLX_nvidia.so.0 #4 0x00007fddb9abfe30 in () at /lib64/libGLX_nvidia.so.0 #5 0x00007fddc874fe21 in QGLXContext::swapBuffers(QPlatformSurface*) (this=0x7fdd582c79d0, surface=0x55b4cb438680) at qglxintegration.cpp:637 #6 0x00007fddcea479db in QSGRenderThread::syncAndRender(QImage*) (this=this@entry=0x55b4cd1bc1c0, grabImage=grabImage@entry=0x0) at scenegraph/qsgthreadedrenderloop.cpp:869 #7 0x00007fddcea480eb in QSGRenderThread::run() (this=0x55b4cd1bc1c0) at scenegraph/qsgthreadedrenderloop.cpp:1042 #8 0x00007fddccae855b in operator() (__closure=<optimized out>) at thread/qthread_unix.cpp:350 #9 (anonymous namespace)::terminate_on_exception<QThreadPrivate::start(void*)::<lambda()> > (t=<optimized out>) at thread/qthread_unix.cpp:287 #10 QThreadPrivate::start(void*) (arg=0x55b4cd1bc1c0) at thread/qthread_unix.cpp:310 #11 0x00007fddcc4ae12d in start_thread (arg=<optimized out>) at pthread_create.c:442 #12 0x00007fddcc52fbc0 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81
I don't know whether it's related, but Bug 465441 has heaptrack report attached with specific line number causing the leak.
(In reply to YAFU from comment #12) > Hi. I see that there are many similar reports about Plasmashell Memory Leak > problem occurring even to this day. > I don't use slideshow wallpaper. I never power off my PC, I always > sleep/suspend to RAM. I can suspend to RAM and resume the PC several times > without plasmashell memory leak problem happening, but suddenly sometime > when I resume the PC the problem starts to occur and I can notice it when I > resume from suspend to RAM and I get a little freeze in Plasma until it does > it becomes responsive after a few seconds. So from this point I start to > check Plasmashell RAM usage, and from this point the problem after > suspending RAM and Resume is always reproducible. From this point every time > you suspend to RAM and Resume, plasmashell increases RAM usage up to a lot > and plasma becomes slower every time. > I'm not sure how to debug the problem when it occurs in order to provide > some useful information. I am having the exact same issue (2 years later...). The leak becomes apparent after about 20 sleep/resume cycles or so, then plasmashell memory consumption increases exponentially each time. On Ubuntu Studio 22.04, KDE Plasma 5.24.7, X11 with NVidia RTX3060.