| Summary: | KDE Plasma segfaults in QSGAreaAllocator::deallocateInNode() | ||
|---|---|---|---|
| Product: | [Frameworks and Libraries] libplasma | Reporter: | James Beddek <telans> |
| Component: | libplasma | Assignee: | Marco Martin <notmart> |
| Status: | RESOLVED FIXED | ||
| Severity: | crash | CC: | account4reg, andysem, aronkvh, bouclier, bryan, charles.christie.ii, d7ghrrkcz, Daniel.Cardenas, Debraym46, devguy.ca, eiskboy, gabriel, jesaenh, johannes.derrer, kde, michaeljensendk, midenok+kdebugs, nate, nicolas.fella, oded, philspam, plasma-bugs, postix, stephen, totok, vkrevs, vladimir.menshakov, zenorbi |
| Priority: | VHI | Keywords: | drkonqi |
| Version: | 5.88.0 | ||
| Target Milestone: | --- | ||
| Platform: | Compiled Sources | ||
| OS: | Linux | ||
| See Also: | https://bugs.kde.org/show_bug.cgi?id=491788 | ||
| Latest Commit: | https://invent.kde.org/frameworks/plasma-framework/-/commit/258cecbf0fab27ee84ad6c43ff3cfe3647215afd | Version Fixed In: | 6.0 |
| Sentry Crash Report: | |||
| Attachments: |
New crash information added by DrKonqi
New crash information added by DrKonqi New crash information added by DrKonqi New crash information added by DrKonqi New crash information added by DrKonqi New crash information added by DrKonqi |
||
|
Description
James Beddek
2021-11-22 05:55:49 UTC
Created attachment 143821 [details]
New crash information added by DrKonqi
DrKonqi auto-attaching complete backtrace.
#14 Plasma::WindowTextureProvider::~WindowTextureProvider (this=0x7efc345ac5b0) at src/declarativeimports/core/corebindingsplugin_autogen/EWIEGA46WW/../../../../../../plasma-framework-5.88.0/src/declarativeimports/core/windowthumbnail.h:149 #15 Plasma::WindowTextureProvider::~WindowTextureProvider (this=0x7efc345ac5b0) at src/declarativeimports/core/corebindingsplugin_autogen/EWIEGA46WW/../../../../../../plasma-framework-5.88.0/src/declarativeimports/core/windowthumbnail.h:149 *** Bug 447407 has been marked as a duplicate of this bug. *** *** Bug 447236 has been marked as a duplicate of this bug. *** *** Bug 446822 has been marked as a duplicate of this bug. *** *** Bug 448827 has been marked as a duplicate of this bug. *** *** Bug 449585 has been marked as a duplicate of this bug. *** Created attachment 147680 [details]
New crash information added by DrKonqi
plasmashell (5.24.3) using Qt 5.15.2
- What I was doing when the application crashed:
Unlocked a locked session by entering password.
-- Backtrace (Reduced):
#4 QSGAreaAllocatorNode::isLeaf (this=0x20) at scenegraph/util/qsgareaallocator.cpp:93
#5 QSGAreaAllocator::deallocateInNode (this=this@entry=0x7f002c010b10, pos=..., node=0x20) at scenegraph/util/qsgareaallocator.cpp:170
#6 0x00007f0039f93826 in QSGAreaAllocator::deallocate (this=this@entry=0x7f002c010b10, rect=...) at scenegraph/util/qsgareaallocator.cpp:116
#7 0x00007f0039f99d3f in QSGOpenGLAtlasTexture::AtlasBase::remove (this=0x7f002c010b00, t=t@entry=0x7effe0240350) at scenegraph/util/qsgopenglatlastexture.cpp:286
#8 0x00007f0039f99e8d in QSGOpenGLAtlasTexture::TextureBase::~TextureBase (this=0x7effe0240350, __in_chrg=<optimized out>) at scenegraph/util/qsgopenglatlastexture.cpp:516
*** Bug 451880 has been marked as a duplicate of this bug. *** *** Bug 453059 has been marked as a duplicate of this bug. *** Created attachment 148398 [details]
New crash information added by DrKonqi
plasmashell (5.24.4) using Qt 5.15.3
- What I was doing when the application crashed:
- Clicked task bar, then plasmashell exited and restarted. I didn't do anything else
-- Backtrace (Reduced):
#4 0x00007f062f64fe90 in QSGAreaAllocator::deallocateInNode(QPoint const&, QSGAreaAllocatorNode*) () at /usr/lib64/libQt5Quick.so.5
#5 0x00007f062f64ff06 in QSGAreaAllocator::deallocate(QRect const&) () at /usr/lib64/libQt5Quick.so.5
#6 0x00007f062f65642f in QSGOpenGLAtlasTexture::AtlasBase::remove(QSGOpenGLAtlasTexture::TextureBase*) () at /usr/lib64/libQt5Quick.so.5
#7 0x00007f062f65657d in QSGOpenGLAtlasTexture::Texture::~Texture() () at /usr/lib64/libQt5Quick.so.5
#8 0x00007f062f656599 in QSGOpenGLAtlasTexture::Texture::~Texture() () at /usr/lib64/libQt5Quick.so.5
Always crashes after awake from sleep. Package: plasma-workspace Version: 4:5.24.4-0ubuntu1 Package: libqt5quick5 Version: 5.15.3+dfsg-1 [KCrash Handler] #4 0x00007f64366aeb10 in QSGAreaAllocator::deallocateInNode(QPoint const&, QSGAreaAllocatorNode*) () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #5 0x00007f64366aeb8a in QSGAreaAllocator::deallocate(QRect const&) () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #6 0x00007f64366b5353 in () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #7 0x00007f64366b54a1 in () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #8 0x00007f64366b54bd in () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #9 0x00007f6427f4e3d1 in () at /usr/lib/x86_64-linux-gnu/qt5/qml/org/kde/plasma/core/libcorebindingsplugin.so #10 0x00007f6436700012 in () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #11 0x00007f64366f996f in () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #12 0x00007f64366fbe49 in () at /usr/lib/x86_64-linux-gnu/libQt5Quick.so.5 #13 0x00007f6434b03ca1 in () at /usr/lib/x86_64-linux-gnu/libQt5Core.so.5 #14 0x00007f6434677b43 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442 #15 0x00007f6434709a00 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81 Please change the priority to critical. This is very annoying bug. "Critical" means 100% easily reproducible with data loss. "Crash" is the correct severity here. See https://community.kde.org/Guidelines_and_HOWTOs/Bug_triaging#Severity if there is any remaining confusion about what the severity levels mean If we get many more duplicates though, I'll consider raising the priority to HI and making it a 15-minute bug. (In reply to Nate Graham from comment #14) > "Critical" means 100% easily reproducible with data loss. "Crash" is the > correct severity here. > > See https://community.kde.org/Guidelines_and_HOWTOs/Bug_triaging#Severity if > there is any remaining confusion about what the severity levels mean > > If we get many more duplicates though, I'll consider raising the priority to > HI and making it a 15-minute bug. Hi, I think the same than Aleksey Midenkov Its very annoying. I don't know if this bug is crash or critical, but I have this bug for a long time (4- 5 years). I decided to disable drkonqi so I was tired to send reports without answering or fixing. Sometimes plasma restart by itself, other I have to restart with "plasmashell &" PD. Sorry for my english, It's is not my first language Created attachment 149062 [details]
New crash information added by DrKonqi
plasmashell (5.24.4) using Qt 5.15.3
- What I was doing when the application crashed:
Waking up from sleep, using nvidia drivers. This happens frequently, the cause is the nullpointer, originating from the ~Texture().
- Unusual behavior I noticed:
Plasmashell behaved erratic, slow window movement, then crash - and plasma, restarted again, launching DrKonqi.
-- Backtrace (Reduced):
#4 QSGAreaAllocatorNode::isLeaf() (this=0x0) at scenegraph/util/qsgareaallocator.cpp:93
#5 QSGAreaAllocator::deallocateInNode(QPoint const&, QSGAreaAllocatorNode*) (this=this@entry=0x55561cda0c20, pos=..., node=0x0) at scenegraph/util/qsgareaallocator.cpp:170
#6 0x00007fecae5dcb8a in QSGAreaAllocator::deallocate(QRect const&) (this=this@entry=0x55561cda0c20, rect=...) at scenegraph/util/qsgareaallocator.cpp:116
#7 0x00007fecae5e3353 in QSGOpenGLAtlasTexture::AtlasBase::remove(QSGOpenGLAtlasTexture::TextureBase*) (this=0x55561cda0c10, t=t@entry=0x7fec601cc710) at scenegraph/util/qsgopenglatlastexture.cpp:286
#8 0x00007fecae5e34a1 in QSGOpenGLAtlasTexture::TextureBase::~TextureBase() (this=0x7fec601cc710, __in_chrg=<optimized out>) at scenegraph/util/qsgopenglatlastexture.cpp:516
Created attachment 149731 [details]
New crash information added by DrKonqi
plasmashell (5.24.4) using Qt 5.15.3
- Unusual behavior I noticed:
Unlocking the screen, this background went black, and plasma restarted.
-- Backtrace (Reduced):
#4 QSGAreaAllocatorNode::isLeaf() (this=0x0) at scenegraph/util/qsgareaallocator.cpp:93
#5 QSGAreaAllocator::deallocateInNode(QPoint const&, QSGAreaAllocatorNode*) (this=this@entry=0x7f337c135c60, pos=..., node=0x0) at scenegraph/util/qsgareaallocator.cpp:170
#6 0x00007f33d4aa4b8a in QSGAreaAllocator::deallocate(QRect const&) (this=this@entry=0x7f337c135c60, rect=...) at scenegraph/util/qsgareaallocator.cpp:116
#7 0x00007f33d4aab353 in QSGOpenGLAtlasTexture::AtlasBase::remove(QSGOpenGLAtlasTexture::TextureBase*) (this=0x7f337c135c50, t=t@entry=0x7f337c26a460) at scenegraph/util/qsgopenglatlastexture.cpp:286
#8 0x00007f33d4aab4a1 in QSGOpenGLAtlasTexture::TextureBase::~TextureBase() (this=0x7f337c26a460, __in_chrg=<optimized out>) at scenegraph/util/qsgopenglatlastexture.cpp:516
Created attachment 149752 [details]
New crash information added by DrKonqi
plasmashell (5.24.4) using Qt 5.15.3
- Unusual behavior I noticed:
Coming out of suspend, using nvidia drivers, the desktop turned black and plasmashell segfaulted. This is a common symptom. That the system comes out of suspend and then plasma segfaults shortly thereafter, using nvidia-drivers.
-- Backtrace (Reduced):
#4 QSGAreaAllocatorNode::isLeaf() (this=0xbec56f0000000000) at scenegraph/util/qsgareaallocator.cpp:93
#5 QSGAreaAllocator::deallocateInNode(QPoint const&, QSGAreaAllocatorNode*) (this=this@entry=0x564bb29f3d50, pos=..., node=0xbec56f0000000000) at scenegraph/util/qsgareaallocator.cpp:170
#6 0x00007f3d99e3cb8a in QSGAreaAllocator::deallocate(QRect const&) (this=this@entry=0x564bb29f3d50, rect=...) at scenegraph/util/qsgareaallocator.cpp:116
#7 0x00007f3d99e43353 in QSGOpenGLAtlasTexture::AtlasBase::remove(QSGOpenGLAtlasTexture::TextureBase*) (this=0x564bb29f3d40, t=t@entry=0x7f3d3416bcc0) at scenegraph/util/qsgopenglatlastexture.cpp:286
#8 0x00007f3d99e434a1 in QSGOpenGLAtlasTexture::TextureBase::~TextureBase() (this=0x7f3d3416bcc0, __in_chrg=<optimized out>) at scenegraph/util/qsgopenglatlastexture.cpp:516
*** Bug 467575 has been marked as a duplicate of this bug. *** *** Bug 467819 has been marked as a duplicate of this bug. *** If i remember correctly after reporting the bug, it was a version upgrade (plasma 5.26.x to 5.27.x) dependency issue in fedora kde. When you ran the update before the bug appeared some packages needed replacement and their filenames changed (those were some purely qt related packages) but since their filenames changed you had to accept the replacement of those new dependencies overwriting other existing ones with different filenames. I initially did not accept those as i did not know if it would break my system after I saw the warning in dnf. After that I experienced this bug. After upgrading and overwriting the dependency upon experiencing the bug KDE plasma 5.27.x runs smoothly without those filed bug issues. I hope this helps! CORRECTION: please scrap the above comment for some reason after logging in i submitted it to the wrong bug number! Above comment was meant for Bug 456531 !!! No worries. *** Bug 470943 has been marked as a duplicate of this bug. *** *** Bug 472396 has been marked as a duplicate of this bug. *** *** Bug 472536 has been marked as a duplicate of this bug. *** This is clearly related to Nvidia Drivers / Hardware. Having switched to Radeon, i have not seen segfaults for this codepath in approx 6 months. A possibly relevant merge request was started @ https://invent.kde.org/frameworks/plasma-framework/-/merge_requests/840 Git commit 258cecbf0fab27ee84ad6c43ff3cfe3647215afd by David Edmundson. Committed on 01/09/2023 at 15:32. Pushed by davidedmundson into branch 'master'. windowthumbnail: Cleanup on scene graph invalidation The documentation (https://doc.qt.io/qt-6/qquickitem.html#graphics-resource-handling) states that if we are doing our own low level handling of graphic resources we need to perform cleanup in 3 places: - on item destruction - when an item is removed from the scene (via the virtual releaseResources) - when the scenegraph gets invalidated (via a slot invalidateSceneGraph) We only handled the first two. This would cause a crash if the window thumbnail is showing just an icon instead of a thumbnail and we hit one of the nvidia context loss resets. The atlas texture would be deleted, but we would keep a QSGTexture refernecing something inside that for longer which would linger. I believe we only need the texture provider to reset, but we may as well do it all. BUG: M +27 -0 src/declarativeimports/core/windowthumbnail.cpp M +3 -0 src/declarativeimports/core/windowthumbnail.h https://invent.kde.org/frameworks/plasma-framework/-/commit/258cecbf0fab27ee84ad6c43ff3cfe3647215afd *** Bug 474259 has been marked as a duplicate of this bug. *** *** Bug 485048 has been marked as a duplicate of this bug. *** *** Bug 483024 has been marked as a duplicate of this bug. *** *** Bug 472616 has been marked as a duplicate of this bug. *** *** Bug 475384 has been marked as a duplicate of this bug. *** *** Bug 475778 has been marked as a duplicate of this bug. *** *** Bug 478510 has been marked as a duplicate of this bug. *** *** Bug 482188 has been marked as a duplicate of this bug. *** |