kwin_x11: The X11 connection broke: I/O error (code 1)

[Kai Kasurinen]

[Impact]

kwin might crash after running some time

Two commits have been reverted upstream since 22.2.x branch was closed, needs those backported to fix this.

[Test case]

Run kwin for a day or so, which is usually enough time to hit this.

Crash happens mostly on a notification popups, so system must be actively receiving notifications to test the crash. Without that crash may not happen even in a week of runtime.

[Where things could go wrong]

This just reverts two commits, and they have been upstream for a few months now, so these causing a regression is unlikely.

[Vasvir-p (vasvir-p)]

Really really sorry for this but no backtrace.

Kwin is getting killed by X with (2 last lines)
The X11 connection broke: I/O error (code 1)
X connection to :0 broken (explicit kill or server shutdown).

I trust that I dind't kill my kwin on purpose. Can any keyboard combination do this?
This time it took 2-3 days to be killed like this. Here are all the messages that it spits in these 3 days. I really hope that you will be able to find something valuable.

I mentioned this bug in https://bugs.kde.org/show_bug.cgi?id=461052 (which is another bug)

After the kwin crash I have a 50% change for a plasmashell crash. This is of course is another bug but it is entirely possible that the plasmashell crash comes first and lead to the kwin death mentioned here. I can't tell because when kwin is killed the desktop is unresponsive.

One other observation. Some times kwin is killed (cannot be found with ps - no titlebars in the windows) and sometimes it hangs (unresponsive desktop, titlebars in windows and can be found by ps).

Next time (in 2 - 3 days) I will try to attach a strace.

bill@beyonder:~$ kwin --replace &
[1] 1185549
bill@beyonder:~$ kwin_xkbcommon: XKB: inet:334:58: unrecognized keysym "XF86EmojiPicker"
bill@beyonder:~$ OpenGL vendor string: Intel
OpenGL renderer string: Mesa Intel(R) HD Graphics 4000 (IVB GT2)
OpenGL version string: 4.2 (Compatibility Profile) Mesa 22.2.2
OpenGL shading language version string: 4.20
Driver: Intel
GPU class: IvyBridge
OpenGL version: 4.2
GLSL version: 4.20
Mesa version: 22.2.2
X server version: 1.21.1
Linux kernel version: 6.0
Requires strict binding: yes
GLSL shaders: yes
Texture NPOT support: yes
Virtual Machine: no

bill@beyonder:~$
bill@beyonder:~$
bill@beyonder:~$
bill@beyonder:~$ jobs
[1]+ Running kwin --replace &
bill@beyonder:~$ disown
bill@beyonder:~$ jobs
bill@beyonder:~$
bill@beyonder:~$
bill@beyonder:~$ qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 5075, resource id: 69206021, major code: 18 (ChangeProperty), minor code: 0

bill@beyonder:~$ kwin_core: XCB error: 152 (BadDamage), sequence: 39, resource id: 29365899, major code: 143 (DAMAGE), minor code: 3 (Subtract)
kwin_core: XCB error: 152 (BadDamage), sequence: 7894, resource id: 29366594, major code: 143 (DAMAGE), minor code: 3 (Subtract)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 10273, resource id: 29366843, major code: 3 (GetWindowAttributes), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 9 (BadDrawable), sequence: 10274, resource id: 29366843, major code: 14 (GetGeometry), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 10277, resource id: 29366843, major code: 3 (GetWindowAttributes), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 9 (BadDrawable), sequence: 10278, resource id: 29366843, major code: 14 (GetGeometry), minor code: 0
...

[Vasvir-p (vasvir-p)]

Now I saw the about system:

Operating System: Debian GNU/Linux
KDE Plasma Version: 5.26.0
KDE Frameworks Version: 5.98.0
Qt Version: 5.15.6
Kernel Version: 6.0.0-2-amd64 (64-bit)
Graphics Platform: X11
Processors: 4 × Intel® Core™ i5-3570K CPU @ 3.40GHz
Memory: 31.3 GiB of RAM
Graphics Processor: Mesa Intel® HD Graphics 4000
Manufacturer: Gigabyte Technology Co., Ltd.

[Grzegorz-alibozek (grzegorz-alibozek)]

I have the same problem

[Vasvir-p (vasvir-p)]

ok it did it again. Same error but different message. kwin got killed so no backtrace.

The X11 connection broke: I/O error (code 1)
XIO: fatal IO error 0 (Success) on X server ":0"

Here is the combined output in the console from kwin and plasmashell.

I don't want to spam this bug so I won't post any more of these outputs unless somebody finds them useful.

caching records from 6725 to 6925
Export in 0 msecs
kwin_core: XCB error: 152 (BadDamage), sequence: 48054, resource id: 69786721, major code: 143 (DAMAGE), minor code: 3 (Subtract)
kwin_core: XCB error: 152 (BadDamage), sequence: 50025, resource id: 69786857, major code: 143 (DAMAGE), minor code: 3 (Subtract)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 60404, resource id: 69787609, major code: 3 (GetWindowAttributes), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 9 (BadDrawable), sequence: 60405, resource id: 69787609, major code: 14 (GetGeometry), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 60408, resource id: 69787609, major code: 3 (GetWindowAttributes), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 9 (BadDrawable), sequence: 60409, resource id: 69787609, major code: 14 (GetGeometry), minor code: 0
kwin_core: XCB error: 152 (BadDamage), sequence: 1323, resource id: 69788124, major code: 143 (DAMAGE), minor code: 3 (Subtract)
libkcups: Renew-Subscription last error: 0 successful-ok
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 10656, resource id: 94371894, major code: 18 (ChangeProperty), minor code: 0
kwin_core: XCB error: 3 (BadWindow), sequence: 13117, resource id: 69781020, major code: 129 (SHAPE), minor code: 6 (Input)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 13118, resource id: 69781020, major code: 2 (ChangeWindowAttributes), minor code: 0
QFont::setPointSizeF: Point size <= 0 (0.000000), must be greater than 0
kwin_core: XCB error: 3 (BadWindow), sequence: 27544, resource id: 69789767, major code: 129 (SHAPE), minor code: 6 (Input)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 27545, resource id: 69789767, major code: 2 (ChangeWindowAttributes), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 29605, resource id: 29362200, major code: 18 (ChangeProperty), minor code: 0
kwin_core: XCB error: 3 (BadWindow), sequence: 31735, resource id: 69791250, major code: 129 (SHAPE), minor code: 6 (Input)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 31736, resource id: 69791250, major code: 2 (ChangeWindowAttributes), minor code: 0
kwin_core: XCB error: 3 (BadWindow), sequence: 36739, resource id: 69791729, major code: 129 (SHAPE), minor code: 6 (Input)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 36740, resource id: 69791729, major code: 2 (ChangeWindowAttributes), minor code: 0
kwin_core: XCB error: 3 (BadWindow), sequence: 28091, resource id: 69792410, major code: 129 (SHAPE), minor code: 6 (Input)
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 28092, resource id: 69792410, major code: 2 (ChangeWindowAttributes), minor code: 0
kwin_core: XCB error: 152 (BadDamage), sequence: 46384...

[Jed-bugs-kde-org (jed-bugs-kde-org)]

This has been happening to me as well. Today, it got so bad that kwin was crashing every few minutes. Interestingly, things stablized after I killed and restarted plasmashell.

Operating System: Arch Linux
KDE Plasma Version: 5.26.2
KDE Frameworks Version: 5.99.0
Qt Version: 5.15.7
Kernel Version: 5.15.76-1-lts (64-bit)
Graphics Platform: X11
Processors: 8 × Intel® Core™ i7-10510U CPU @ 1.80GHz
Memory: 15.3 GiB of RAM
Graphics Processor: Mesa Intel® UHD Graphics
Manufacturer: LENOVO
Product Name: 20UBCTO1WW
System Version: ThinkPad X1 Yoga Gen 5

[Vasvir-p (vasvir-p)]

It did it a again this on (remote with xpra) unlock time with error code 22.
I am posting only a few lines of intermixed output (plasmashell & kwin) just in case somebody may spot something.

qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 16004, resource id: 60817412, major code: 18 (ChangeProperty), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 3 (BadWindow), sequence: 16008, resource id: 60817421, major code: 18 (ChangeProperty), minor code: 0
QFont::setPointSizeF: Point size <= 0 (0.000000), must be greater than 0
Could not find the Plasmoid for Plasma::FrameSvgItem(0x563937d58010) QQmlContext(0x563934b35570) QUrl("file:///usr/share/plasma/plasmoids/org.kde.plasma.notifications/contents/ui/global/Globals.qml")
Could not find the Plasmoid for Plasma::FrameSvgItem(0x563937d58010) QQmlContext(0x563934b35570) QUrl("file:///usr/share/plasma/plasmoids/org.kde.plasma.notifications/contents/ui/global/Globals.qml")
file:///usr/share/plasma/plasmoids/org.kde.plasma.notifications/contents/ui/NotificationItem.qml:248:21: QML SelectableLabel: Binding loop detected for property "implicitWidth"
file:///usr/share/plasma/plasmoids/org.kde.plasma.notifications/contents/ui/NotificationItem.qml:248:21: QML SelectableLabel: Binding loop detected for property "implicitWidth"
file:///usr/share/plasma/plasmoids/org.kde.plasma.notifications/contents/ui/NotificationItem.qml:248:21: QML SelectableLabel: Binding loop detected for property "implicitHeight"
file:///usr/share/plasma/plasmoids/org.kde.plasma.notifications/contents/ui/SelectableLabel.qml:40:5: QML TextArea: Binding loop detected for property "implicitHeight"
The X11 connection broke: I/O error (code 1)
XIO: fatal IO error 22 (Invalid argument) on X server ":0"
      after 6963214 requests (6963214 known processed) with 0 events remaining.
qt.qpa.xcb: QXcbConnection: XCB error: 2 (BadValue), sequence: 32106, resource id: 0, major code: 53 (CreatePixmap), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 9 (BadDrawable), sequence: 32107, resource id: 29364601, major code: 55 (CreateGC), minor code: 0
qt.qpa.xcb: QXcbConnection: XCB error: 9 (BadDrawable), sequence: 32108, resource id: 29364601, major code: 72 (PutImage), minor code: 0

[Vasvir-p (vasvir-p)]

So far the X errors are:

1. Local: The X11 connection broke: I/O error (code 1) X connection to :0 broken (explicit kill or server shutdown). --> kwin killed
2. Local: The X11 connection broke: I/O error (code 1) XIO: fatal IO error 0 (Success) on X server ":0" --> kwin killed
3. Remote: The X11 connection broke: I/O error (code 1) XIO: fatal IO error 22 (Invalid argument) on X server ":0" after 6963214 requests (6963214 known processed) with 0 events remaining. --> kwin killed

I will keep updating the list.

Note: It looks like the hung case is much rarer. Maybe it happens when the plasmashell is busted and it is actually a different bug. When kwin dies user input are dying too so I can't be really sure that plasmashell is actually alive before I restart kwin.

Maybe all 3 above are different bugs. I certainly hope that this isn't the case though because it looks terribly difficult to diagnose this without a backtrace let alone 3 of them.

[Trilader+kdebugs (trilader+kdebugs)]

I also have encountered this (quite a lot) in the last days (since I updated to 5.26.x). Sometimes I need to `kwin_x11 --replace` multiple times a day - sometimes I also then need to `pkill plasmashell && kstart5 plasmashell` as it becomes non-responsive.

Messages I've seen so far:
The X11 connection broke: I/O error (code 1) X connection to :0 broken (explicit kill or server shutdown)
The X11 connection broke: I/O error (code 1) XIO: fatal IO error 22 (Invalid argument) on X server ":0" after 171015 requests (171015 known processed) with 0 events remaining.
The X11 connection broke: I/O error (code 1) XIO: fatal IO error 2 (No such file or directory) on X server ":0" after 78026 requests (78026 known processed) with 0 events remaining.

There are more which I don't have anything saved for as they are from re-running kwin_x11 from a terminal.
There isn't anything interesting in dmesg or Xorg.0.log.

System information:
Operating System: Arch Linux
KDE Plasma Version: 5.26.2
KDE Frameworks Version: 5.99.0
Qt Version: 5.15.7
Kernel Version: 6.0.6-arch1-1 (64-bit)
Graphics Platform: X11
Processors: 8 × Intel® Xeon® CPU E3-1245 v5 @ 3.50GHz
Memory: 31,1 GiB of RAM
Graphics Processor: AMD Radeon RX 580 Series
Manufacturer: Supermicro
Product Name: Super Server
System Version: 0123456789

[Vasvir-p (vasvir-p)]

I also got the 'error 2: No such file or directory' error:

The X11 connection broke: I/O error (code 1)
XIO: fatal IO error 2 (No such file or directory) on X server ":0"
      after 5589828 requests (5589828 known processed) with 0 events remaining.

kwin killed so no strace.

[Vasvir-p (vasvir-p)]

In my home. I upgraded to 5.26.3 (debian unstable) and was hit again by this . Not sure which error code though, because it was the first kwin instance and not one that I run from terminal so I didn't have stderr output.

[Oleg Solovyov (mcpain)]

Try reverting Mesa commit 768238fd

[Vasvir-p (vasvir-p)]

This one? https://gitlab.freedesktop.org/mesa/mesa/-/commit/768238fdc06eed3dce36da3baf811cb70db42b5c

[Oleg Solovyov (mcpain)]

(In reply to Vassilis Virvilis from comment #11)
> This one?
> https://gitlab.freedesktop.org/mesa/mesa/-/commit/
> 768238fdc06eed3dce36da3baf811cb70db42b5c

Yes. Before this commit kwin never exited for me (not a crash)

[Oleg Solovyov (mcpain)]

(In reply to Oleg Solovyov from comment #12)
> (In reply to Vassilis Virvilis from comment #11)
> > This one?
> > https://gitlab.freedesktop.org/mesa/mesa/-/commit/
> > 768238fdc06eed3dce36da3baf811cb70db42b5c
>
> Yes. Before this commit kwin never exited for me (not a crash)

just exited on commit 768238fdc06eed3dce36da3baf811cb70db42b5c

[Oleg Solovyov (mcpain)]

(In reply to Oleg Solovyov from comment #13)
> (In reply to Oleg Solovyov from comment #12)
> > (In reply to Vassilis Virvilis from comment #11)
> > > This one?
> > > https://gitlab.freedesktop.org/mesa/mesa/-/commit/
> > > 768238fdc06eed3dce36da3baf811cb70db42b5c
> >
> > Yes. Before this commit kwin never exited for me (not a crash)
>
> just exited on commit 768238fdc06eed3dce36da3baf811cb70db42b5c

commit 151aa19c21575ba498a19c48e84474107a4eb304*

[silvan (silvan-calarco)]

(In reply to Oleg Solovyov from comment #14)
> (In reply to Oleg Solovyov from comment #13)
> > (In reply to Oleg Solovyov from comment #12)
> > > (In reply to Vassilis Virvilis from comment #11)
> > > > This one?
> > > > https://gitlab.freedesktop.org/mesa/mesa/-/commit/
> > > > 768238fdc06eed3dce36da3baf811cb70db42b5c
> > >
> > > Yes. Before this commit kwin never exited for me (not a crash)
> >
> > just exited on commit 768238fdc06eed3dce36da3baf811cb70db42b5c
>
> commit 151aa19c21575ba498a19c48e84474107a4eb304*

Hello, I'm facing this problem too.
If I understand the problem might be resolved by reverting commit 768238fdc06eed3dce36da3baf811cb70db42b5c, is that correct? Or the two last comments state something different? Thanks!

[Oleg Solovyov (mcpain)]

(In reply to Silvan Calarco from comment #15)
> Hello, I'm facing this problem too.
> If I understand the problem might be resolved by reverting commit
> 768238fdc06eed3dce36da3baf811cb70db42b5c, is that correct? Or the two last
> comments state something different? Thanks!

7682 is bad, not sure if it's the same about 151a

[silvan (silvan-calarco)]

(In reply to Oleg Solovyov from comment #16)
> (In reply to Silvan Calarco from comment #15)
> > Hello, I'm facing this problem too.
> > If I understand the problem might be resolved by reverting commit
> > 768238fdc06eed3dce36da3baf811cb70db42b5c, is that correct? Or the two last
> > comments state something different? Thanks!
>
> 7682 is bad, not sure if it's the same about 151a

Ok, I'm testing now Mesa 22.2.3 with 7682 reverted.
In my case the problem happens seldomly usually upon the creation of a notification box which is rendered black and it may take some days before I may say that the issue is fixed.
As you may have noticed an upstream bug has been opened here: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7674

[Oleg Solovyov (mcpain)]

(In reply to Silvan Calarco from comment #17)
> As you may have noticed an upstream bug has been opened here:
> https://gitlab.freedesktop.org/mesa/mesa/-/issues/7674

It's mine

[Xxmlud (xxmlud)]

Hi, this same thing happens to me, Kwin dies randomly, what I do to get back to normal I run alt+f2 and type "systemctl restart display-manager".

I have a bug created, but this one seems more extensive. My graphics card is an Intel

Traducido con DeepL (https://www.deepl.com/app/?utm_source=ios&utm_medium=app&utm_campaign=share-translation

[Oleg Solovyov (mcpain)]

Created attachment 153667
how to crash

Additionally, put

[Notifications]
PopupTimeout=100

into ~/.config/plasmanotifyrc and relogin

[Vasvir-p (vasvir-p)]

Aa a reproducer. Thanks for that.

I will try it at home and report back.

I hope that you are onto something. If memory serves a lot of the crashes are related with virtual dekstop switches - not sure if that is a Knotification though...

We 'll see. Much appreciated anyway...

[Vasvir-p (vasvir-p)]

This is probably another bug. I looked around but can't anything. Probable bad keywords. (no keyboard input)

This time my desktop hanged. (Not sure if I had mouse pointer). Certainly not keyboard. Windows decorations were there.

I switch to console and back to X. Now I have mouse pointer but no click. No keyboard.

Now I can't Ctrl-Alt-F2. I ssh and do sudo chvt 2

I restart kwin from console (DISPLAY=:0 kwin --replace)

Now I have mouse pointer + click but still no keyboard.

I restarted plasmashell. Kate crashed as per: https://bugs.kde.org/show_bug.cgi?id=461318

Still no keyboard input.

*** Do you happen to know what process I need to restart to gain keyboard again?. I had to reboot. and I lost a lot of state.

In the meantime as I promised I straced both kwin and plasmashell. Everything looks ok. They are waiting for input:

kwin:
futex(0x56164c8eaec4, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, FUTEX_BITSET_MATCH_ANY

plasmashell:
futex(0x55796b5c3f14, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 0, NULL, FUTEX_BITSET_MATCH_ANY

This bug is more severe and more difficult / rare to trigger. If somebody has a pointer for this bug I will post there. Otherwise I will open a new bug later today.

[Oleg Solovyov (mcpain)]

(In reply to Vassilis Virvilis from comment #22)
> This is probably another bug. I looked around but can't anything. Probable
> bad keywords. (no keyboard input)

Unfortunately, if it is, it makes difficult to bisect the original one.

> *** Do you happen to know what process I need to restart to gain keyboard
> again?. I had to reboot. and I lost a lot of state.

Dunno, sorry. Just making tests to figure out what happens and why

[Oleg Solovyov (mcpain)]

Created attachment 153678
reproducer v3

Better reproducer, gave 31b04e420b0eb080084c6323066ea0b83929d59e on Mesa

[Oleg Solovyov (mcpain)]

(In reply to Oleg Solovyov from comment #24)
> Created attachment 153678 [details]
> reproducer v3
>
> Better reproducer, gave 31b04e420b0eb080084c6323066ea0b83929d59e on Mesa

aaaand 768238fd should be reverted too.

[Vasvir-p (vasvir-p)]

Thanks for looking so hard at this. Much appreciated.

Yesterday I had another instance of this (or another bug). This time kwin wasn't killed (windows have their titles) but kwin was unresponsive. However I had keyboard and I was able to run 'kwin --replace' and then everything got ok.

I don't remember how I got to the terminal though. Either I was already there (konsole had focus) or Alt-Tab worked which strikes me as somewhat weird.

Unfortunately no logs for this new one.

I am not sure I am helping anyone with these awfully incomplete reports. I will try my best to run the reproducer. Maybe I would be able to report something more useful.

[Oleg Solovyov (mcpain)]

(In reply to Vassilis Virvilis from comment #26)
> I am not sure I am helping anyone with these awfully incomplete reports. I
> will try my best to run the reproducer. Maybe I would be able to report
> something more useful.

Older version would spam notifications so hard it could have killed any version of Mesa

[Xxmlud (xxmlud)]

Hello, Vassilis Virvilis, I am in the same situation as you. But I don't know how to provide some logs that can help with this problem.
One question, do you have two monitors?

[Vasvir-p (vasvir-p)]

(In reply to xxmlud from comment #28)
> Hello, Vassilis Virvilis, I am in the same situation as you. But I don't
> know how to provide some logs that can help with this problem.
> One question, do you have two monitors?

No. I have one monitor per machine. But I have work and home computer. Some times I use xpra to access the work machine.

Sometimes I got a local crash. Sometimes I got a remote crash. Sometimes I have keyboard other I have mouse. Sometimes the keyboard input is borked like a modifier (like Ctrl or Alt) is pressed. The borked modifier is probably a remote thing only but I am not 100% sure. It's all over the place really.

[Vasvir-p (vasvir-p)]

(In reply to Oleg Solovyov from comment #27)
> Older version would spam notifications so hard it could have killed any
> version of Mesa

<humor>
But but... a client shouldn't be able to crash a library... My world shutters...
</humor>

Seriously though. If I try to open thousands or million of files I wouldn't expect a crash in libc or in the kernel. I would expect an error "out of resources" but then I guess my program would crash because I wouldn't have been checking for errors. Is something like that goes on here?

[Xxmlud (xxmlud)]

I do work with two monitors, and I wanted to find some relation.
 When kwin breaks and I have the Konsole in focus I restart the service "display-manager", otherwise I connect with alt+f2. The truth is that the logs are not enlightening. I am lost with this issue

[Oleg Solovyov (mcpain)]

(In reply to Vassilis Virvilis from comment #30)
> (In reply to Oleg Solovyov from comment #27)
> > Older version would spam notifications so hard it could have killed any
> > version of Mesa
>
> <humor>
> But but... a client shouldn't be able to crash a library... My world
> shutters...
> </humor>
>
> Seriously though. If I try to open thousands or million of files I wouldn't
> expect a crash in libc or in the kernel. I would expect an error "out of
> resources" but then I guess my program would crash because I wouldn't have
> been checking for errors. Is something like that goes on here?

I suspect something happens with rendering process

[Trilader+kdebugs (trilader+kdebugs)]

Not sure how much of this is just a coincidence or if it is actually related:
Before I had the "KWin just quits" issue I had problems where KWin sometimes just "forgot to update a window", meaning a window starts flickering between some old window states instead of rendering new ones. The issue was "fixable" by toggling compositing off and on again (using Ctrl+Shift+F12 by default I think). I had that issue for a few years but never really investigated as just pressing the keys was an easy work around for me. How often this happened varied from once every few days to multiple times an hour, though subjectively it was occurring more often lately than a year ago.

At the beginning of November I changed my default browser from Chrome to Firefox and that issue didn't happen anymore for a few days so I thought maybe it was a wired interaction between Chrome and KWin. Then this issue started happening.

What both issues/behaviors have in common is that I noticed that usually there are notifications being shown when it happens. The "old issue" with stuck rendering showed the notifications just fine and also still updated a small (maybe 30 px?) region around the notification popup. With this issue I saw a notification pop up appear as a black rectangle and then KWin crashing (and sometimes also Plasmashell hanging after the KWin restart).

[silvan (silvan-calarco)]

I also used to have the problems reported by Daniel but in my case it seems they stopped happening since Qt update to 5.15.7 (with KDE patches applied to qtbase and qtdeclarative modules).
Currently I didn't have any issues at all in the last 5 days on two computers (with Intel and AMD/ATI chipsets) where they happened once every 1-2 days. I also run many times Oleg's reproducer without any issue. It may just be a coincidence, BTW I also updated Mesa to 22.2.3 and KDE Plasma to 5.26.3, are you all using such releases?

[Oleg Solovyov (mcpain)]

(In reply to Silvan Calarco from comment #34)
> I also used to have the problems reported by Daniel but in my case it seems
> they stopped happening since Qt update to 5.15.7 (with KDE patches applied
> to qtbase and qtdeclarative modules).
> I also run many times Oleg's reproducer without any issue. It may
> just be a coincidence, BTW I also updated Mesa to 22.2.3 and KDE Plasma to
> 5.26.3, are you all using such releases?

All the same, except Qt release (5.15.6)

[Trilader+kdebugs (trilader+kdebugs)]

(In reply to Silvan Calarco from comment #34)
> I also used to have the problems reported by Daniel but in my case it seems
> they stopped happening since Qt update to 5.15.7 (with KDE patches applied
> to qtbase and qtdeclarative modules).
> Currently I didn't have any issues at all in the last 5 days on two
> computers (with Intel and AMD/ATI chipsets) where they happened once every
> 1-2 days. I also run many times Oleg's reproducer without any issue. It may
> just be a coincidence, BTW I also updated Mesa to 22.2.3 and KDE Plasma to
> 5.26.3, are you all using such releases?
I'm running these version except that all the Qt Packages have the KDE patches, as that is what ArchLinux ships at the moment.

For me the v3 reproducer works. I'm, thinking of putting a check in there to see if KWin is still alive before sending the next notification and quit if not. Maybe it's somewhat consistent how many notifications it takes for KWin to crash.

[Xxmlud (xxmlud)]

Hello,

These are the versions that I use.

KDE Plasma: 5.26.3
KDE Frameworks: 5.99.0
QT Version: 5.15.7

Is there any way to put kwin_x11 in debug mode? I have seen that in Wayland it is possible, but in X11 I have not found anything.

[Vlad-zahorodnii (vlad-zahorodnii)]

Do you know what tends to cause

kwin_core: XCB error: 152 (BadDamage), sequence: 48054, resource id: 69786721, major code: 143 (DAMAGE), minor code: 3 (Subtract)

e.g. did you close some window?

[Vlad-zahorodnii (vlad-zahorodnii)]

It looks like https://invent.kde.org/plasma/kwin/-/commit/7eca5f44dbc2b094468e7430f784ff1e7f372989 was not backported to 5.26. I backported it just now, hopefully it's going to reduce the amount of BadDamage errors.

[Oleg Solovyov (mcpain)]

(In reply to Silvan Calarco from comment #34)
> I also used to have the problems reported by Daniel but in my case it seems
> they stopped happening since Qt update to 5.15.7 (with KDE patches applied
> to qtbase and qtdeclarative modules).

It didn't help for me

[Xxmlud (xxmlud)]

*** Bug 461333 has been marked as a duplicate of this bug. ***

[Xxmlud (xxmlud)]

Hello,

In my case, the "crash" comes after a notification comes up.

Could it be related to some bug in "(org.kde.plasma.notifications)"?
Can you check if it appears in the same situation?

For the moment I have disabled notifications in the application that gives me more notifications on the screen.

Regards

[Oleg Solovyov (mcpain)]

(In reply to xxmlud from comment #42)
> Could it be related to some bug in "(org.kde.plasma.notifications)"?
> Can you check if it appears in the same situation?

No, it is a Mesa bug, which has been bisected and tested with reproducer (see attachments)

[Xxmlud (xxmlud)]

Hello,

So far since I disabled notifications in my application, I haven't had any crashes. Are you sure it's a Mesa bug?

I will report next Tuesday if I had any new crashes.

Regards

[Oleg Solovyov (mcpain)]

(In reply to xxmlud from comment #44)
> So far since I disabled notifications in my application, I haven't had any
> crashes. Are you sure it's a Mesa bug?

Retry with hardware acceleration disabled.
It crashes after updating Mesa 22.1 -> 22.2 branch

[Kde-bugs-io-drsf32 (kde-bugs-io-drsf32)]

I keep having this for weeks now and can confirm the last line before X goes down always is the BadDamage one.
I think it started like 6-8 weeks ago. Once it starts to happen, restarting kwin_x11 mostly won't help for any longer period of time. In the best cases it lasts long enough to allow for a controlled Desktop shutdown, package update and reboot - sometimes not even that. After reboot it keeps going for another 24 to 48 hours before kwin crashes start again. I

I don't think notifications are the one culprit. Yes, notifications sometimes seem to trigger the problem (notably causing latte dock to hang until killed), but not always. Sometimes a mere desktop switch apparently is enough to get it down again. In the worst cases just coming back to the x11 VT (remapping the desktop) did end it. I'll keep re-starting and updating until it gets better since I get it there is a promising patch now. But the general instability really is getting to me :-)
If you want logs of anything you don't already have, drop me a line.
Thanks for looking into this.

[Xxmlud (xxmlud)]

Hi, <email address hidden>. Is there any way we users can help?

I have disabled notifications for several days and have not had any crashes.

Regards

[Vlad-zahorodnii (vlad-zahorodnii)]

I can reproduce the issue with the attached test app. Thank you, Oleg, for providing it and also bisecting mesa. :)

kwin 5.26 didn't have any significant X11 changes besides one fix to prevent frozen windows due to xcb_generate_id() returning a bad id, however this bug is reproducible even after reverting those changes, so I think that we can rule out kwin.

One peculiar thing about this bug is that if compositing is disabled, reproducer v3 can "hang" xorg. That's it, at some point, notification windows will stop popping up and other X11 connections will hang in general. Restarting plasmashell helps to fix that.

kwin gets killed when it repaints the screen, specifically binds the window texture

Thread 1 "kwin_x11" received signal SIGTERM, Terminated.
0x00007fc511503120 in ?? () from /usr/lib/libqaccessibilityclient-qt5.so.0
(gdb) bt
#0 0x00007fc511503120 in () at /usr/lib/libqaccessibilityclient-qt5.so.0
#1 0x00007fc51168eaae in () at /lib64/ld-linux-x86-64.so.2
#2 0x00007fc50de53fa5 in () at /usr/lib/libc.so.6
#3 0x00007fc50de54120 in () at /usr/lib/libc.so.6
#4 0x00007fc50ec05be6 in _XDefaultIOError () at /usr/lib/libX11.so.6
#5 0x00007fc50ec08eb3 in _XIOError () at /usr/lib/libX11.so.6
#6 0x00007fc50ec0c5c9 in _XFlush () at /usr/lib/libX11.so.6
#7 0x00007fc50ec0c85e in _XGetRequest () at /usr/lib/libX11.so.6
#8 0x00007fc50ec010ee in XSync () at /usr/lib/libX11.so.6
#9 0x00007fc5082e00cd in () at /usr/lib/libGLX_mesa.so.0
#10 0x00005577ad004ec4 in KWin::GlxPixmapTexturePrivate::onDamage() (this=0x5577af38c5d0)
    at /home/vlad/Workspace/kwin/src/backends/x11/standalone/x11_standalone_glx_backend.cpp:921
#11 0x00007fc50efb0b29 in KWin::GLTexture::bind() (this=<optimized out>) at /home/vlad/Workspace/kwin/src/libkwineffects/kwingltexture.cpp:454
#12 0x00007fc511235a78 in KWin::SceneOpenGL::render(KWin::Item*, int, QRegion const&, KWin::WindowPaintData const&)

so if plasmashell manages to "break" xorg, it sort of explains why kwin_x11 gets killed

I built mesa 22.1.7 and the issue is not reproducible there. I didn't bisect mesa changes, but the ones mentioned in https://gitlab.freedesktop.org/mesa/mesa/-/issues/7674 look like good candidates.

It looks like glXMakeCurrent(Display, Window, GLXContext) got broken in Mesa 22.2. It's worth noting that Qt uses glXMakeCurrent(Display, Window, GLXContext) rather than glXMakeCurrent(Display, GLXWindow, GLXContext). I made a quick and dirty patch to test that theory https://invent.kde.org/-/snippets/2423. I ran reproducer v3 a couple of times, the bug doesn't seem to occur.

Regardless, glXMakeCurrent(Display, Window, GLXContext) is still a valid case and it must work.

[Xaver Hugl (xaver-hugl)]

*** Bug 462143 has been marked as a duplicate of this bug. ***

[Eliashogstvedt (eliashogstvedt)]

So it might be finally solved?

I made a duplicate issue, but having read the comments here I can add that it seems/feels to happen when:

- launching an application that creates an opengl context or something that triggers the desktop turning off the compositor (for example me developing an opengl application)
- screen lock due to inactivity. (not sure if happens on lock out or logging back in though)
- abnormally high gpu usage (ie running something very gpu intensive by mistake when programming)

[João Vidal da Silva (joao-vidal-silva)]

Following the last comment by CapsAdmin: it happened to me during the process of opening +/- 20 files at once in LibreOffice Writer, using Dolphin.

[Vlad-zahorodnii (vlad-zahorodnii)]

Fixed in Mesa.

[Xxmlud (xxmlud)]

Hello, in which version of Mesa is the problem solved?

[Bugs-kde-org-1 (bugs-kde-org-1)]

(In reply to xxmlud from comment #53)
> Hello, in which version of Mesa is the problem solved?

The URL @ the top of the ticket points to [this](https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19972) issue which is part of the `22.3` release which looks to be _almost_ ready to ship.

[Vasvir-p (vasvir-p)]

Out of curiosity,

I have no experience with opengl/mesa so the question may be invalid.

The main culprit that is being reversed is a patch that was adjusting reference counting of ... something (opengl context? not sure).

The question is: Are we sure that the patch being reversed is buggy? We know for sure that it caused a regression and kwin crashes. However, isn't it possible that this is a failure of Qt/KDE/Plasma/kwin to handle the new "correct" behavior that the patch introduced?

Don't get me wrong I am thrilled and grateful that you guys found it and have the mesa guys reverting it but I am just curious...

[B-sam-g (b-sam-g)]

(In reply to Vassilis Virvilis from comment #55)
> Out of curiosity,
>
> I have no experience with opengl/mesa so the question may be invalid.
>
> The main culprit that is being reversed is a patch that was adjusting
> reference counting of ... something (opengl context? not sure).
>

Anything to do with resource management is tricky :)

> The question is: Are we sure that the patch being reversed is buggy? We know
> for sure that it caused a regression and kwin crashes. However, isn't it
> possible that this is a failure of Qt/KDE/Plasma/kwin to handle the new
> "correct" behavior that the patch introduced?
>

I'm sure that the Mesa folks wouldn't have reverted it unless they thought
it looked like an issue on their end.

[ValdikSS (valdikss)]

*** Bug 463231 has been marked as a duplicate of this bug. ***

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Fabrice Boyrie (fboyrie)]

Hello

I have around one kwin_X11 crash by day, and I suppose it is the same bug. Same problem on my laptop and my desktop.

Is the mesa patch backported to Ubuntu 22.04 ?

[bayron calderon (kirbayron)]

I was experiencing the same issue on Kubuntu 22.04.2. It ships 22.0.1 mesa version and this bug was fixed on 22.3.0 (as the tag says).

Search for kisak-mesa stable ppa to upgrade your mesa version to 22.3.7, issue is gone there.

[sir phobos (sir-phobos)]

will the fix be available in 22.04 lts without using ppa any time soon?

[sir phobos (sir-phobos)]

noticed that if disabling compositing right after kwin restart - then it seem to crash much less often (3 days so far without crash), but plasmashell still freezes approx once a day

(default ubuntu 22.04.2 with kept back 5.15 kernel (because of amdgpu bugs in 5.19))
KDE Plasma: 5.24.7
KDE Frameworks: 5.92.0
QT Version: 5.15.3
MESA: 22.2.5
amdgpu

[Andreas Hasenack (ahasenack)]

The revert is not in Kinetic, implying it could be affected by this bug too. Do you intend to skip kinetic, or is it not affected?

[Andreas Hasenack (ahasenack)]

Hello Kai, or anyone else affected,

Accepted mesa into jammy-proposed. The package will build now and be available at https://launchpad.net/ubuntu/+source/mesa/22.2.5-0ubuntu0.1~22.04.2 in a few hours, and then in the -proposed repository.

Please help us by testing this new package. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation on how to enable and use -proposed. Your feedback will aid us getting this update out to other Ubuntu users.

If this package fixes the bug for you, please add a comment to this bug, mentioning the version of the package you tested, what testing has been performed on the package and change the tag from verification-needed-jammy to verification-done-jammy. If it does not fix the bug for you, please add a comment stating that, and change the tag to verification-failed-jammy. In either case, without details of your testing we will not be able to proceed.

Further information regarding the verification process can be found at https://wiki.ubuntu.com/QATeam/PerformingSRUVerification . Thank you in advance for helping!

N.B. The updated package will be released to -updates after the bug(s) fixed by this package have been verified and the package has been in -proposed for a minimum of 7 days.

[Timo Aaltonen (tjaalton)]

yes, now that lunar is out, I'd skip kinetic

[Gunter Ohrner (gohrner)]

I can confirm the fix for jammy - I'm now running the fixed version since it hit jammy-proposed on Friday.

I wanted to wait a few days to be sure, but so far it looks very promising - no kwin crash / termination since then! :)

[Ubuntu SRU Bot (ubuntu-sru-bot)]

All autopkgtests for the newly accepted mesa (22.2.5-0ubuntu0.1~22.04.2) for jammy have finished running.
The following regressions have been reported in tests triggered by the package:

gtk+3.0/3.24.33-1ubuntu2 (i386)
mir/2.7.0-0ubuntu3 (armhf)
mutter/42.5-0ubuntu1 (amd64)

Please visit the excuses page listed below and investigate the failures, proceeding afterwards as per the StableReleaseUpdates policy regarding autopkgtest regressions [1].

https://people.canonical.com/~ubuntu-archive/proposed-migration/jammy/update_excuses.html#mesa

[1] https://wiki.ubuntu.com/StableReleaseUpdates#Autopkgtest_Regressions

Thank you!

[Launchpad Janitor (janitor)]

This bug was fixed in the package mesa - 22.2.5-0ubuntu0.1~22.04.2

---------------
mesa (22.2.5-0ubuntu0.1~22.04.2) jammy; urgency=medium

  * patches: Fix VA-API on DCN 3.1.4. (LP: #2017142)
  * patches: Fix a freeze with iris under stress test. (LP: #1990089)
  * patches: Revert two commits causing kwin to eventually crash. (LP:
    #2003339)

 -- Timo Aaltonen <email address hidden> Fri, 21 Apr 2023 14:51:10 +0300

[Timo Aaltonen (tjaalton)]

The verification of the Stable Release Update for mesa has completed successfully and the package is now being released to -updates. Subsequently, the Ubuntu Stable Release Updates Team is being unsubscribed and will not receive messages about this bug report. In the event that you encounter a regression using the package from -updates please report a new bug using ubuntu-bug and tag the bug report regression-update so we can easily find any regressions.

[Konstantin Petrov (iozho1)]

Unfortunately this is still happening for me - though in much lower rate. My vga is: VGA compatible controller: Intel Corporation Alder Lake-P Integrated Graphics Controller (rev 0c)

[sir phobos (sir-phobos)]

How often does it happen for you?
i'm currently running 4days without the bug.
previously it was consistently happening every 24 hours at least once

[Konstantin Petrov (iozho1)]

Perhaps every 2 days. Initially it is working find (let's say 2-3 days) and once it hangs (note no GPU hang kernel logs) perhaps every day. I have teams open in chrome browser with notifications on.

[Gunter Ohrner (gohrner)]

On my machine the crashes also seem to be solved. I was using Intel driver until the weekend (since I use this laptop) and then switched to the nVidia binary drivers because I wanted to use CUDA for some experiments.

kwin seems to run properly with both so far.

@Konstantin Petrov: You write about a "hang", even referring to "GPU hangs" in the logs, which sometimes happen with Intel cards but on my system then actually cause a "hang" / "freeze", not a crash with kwin disappearing.

This issue here is about kwin suddenly just quitting / disappearing, not about kwin freezing.

Are we talking about the same symptoms?

[Konstantin Petrov (iozho1)]

Yes Gunter you might be correct. For me it is a freeze but seems related since other people have reported the same issue stating the revert of the two commits solved the freeze issue. There are no GPU hangs in the kernel logs (I just wanted to stress that out). I guess I will open a new bug report.

[Konstantin Petrov (iozho1)]

I actually failed to mention that before the update I was experiencing the crash. Not instead of the crashing UI is freezing.

[Jamie Scott (jamiescott)]

According to the launchpad page for mesa 22.2.5-0ubuntu0.1~22.04.2 (https://launchpad.net/ubuntu/+source/mesa/22.2.5-0ubuntu0.1~22.04.2) the package is still in proposed, not updates. @tjaalton 's message in #72 suggests it has been released to updates but this appears incorrect.

I'm on 22.04 with the updates archive enabled but not proposed and I'm not seeing the available update. Temporarily enabling proposed and looking at upgradable packages shows 22.2.5-0ubuntu0.1~22.04.2 which matches up which the launchpad page.

Is there an ETA for this package to land in updates?

[Timo Aaltonen (tjaalton)]

it was pulled from updates because it caused bug 2020604, and that has a fix pending on the SRU queue

[sir phobos (sir-phobos)]

it seems that i've managed to update to it before it was pulled from updates.

I just hope that 2020604 fix won't revert this fix.

choosing between a fix for kwin crash that often leads to the need to relogin alltoghether, and a fix for minor chromium graphics glitches - i'd prefer a fix for kwin

[Ubuntu SRU Bot (ubuntu-sru-bot)]

All autopkgtests for the newly accepted mesa (22.2.5-0ubuntu0.1~22.04.2) for jammy have finished running.
The following regressions have been reported in tests triggered by the package:

gtk+3.0/3.24.33-1ubuntu2 (i386)
mutter/42.5-0ubuntu1 (amd64)

Please visit the excuses page listed below and investigate the failures, proceeding afterwards as per the StableReleaseUpdates policy regarding autopkgtest regressions [1].

https://people.canonical.com/~ubuntu-archive/proposed-migration/jammy/update_excuses.html#mesa

[1] https://wiki.ubuntu.com/StableReleaseUpdates#Autopkgtest_Regressions

Thank you!

[autra (autra)]

Still reproducing on mesa 22.2.5-0ubuntu0.1~22.04.3

[sir phobos (sir-phobos)]

i've been running previous version 22.2.5-0ubuntu0.1~22.04.2 for a month straight and saw no issues.

has the bug been reintroduced in 22.2.5-0ubuntu0.1~22.04.3 ?

[Gunter Ohrner (gohrner)]

I'm running mesa 22.2.5-0ubuntu0.1~22.04.3 since it was released without any issues any more.

Previously, I had between one and a dozend kwin crashes a day, so for me it's definitely fixed in this version.

@autra: Do you really see the "kwin_x11: The X11 connection broke: I/O error (code 1)" kwin quits (technically, it's not even a crash as I understand), or maybe something differently which causes kwin to suddenly disappear?

[Konstantin Petrov (iozho1)]

Even though it is not crashing it definitely freezes randomly with KDE Kwin.

[Konstantin Petrov (iozho1)]

And I see another IO error:

qt.qpa.xcb: xcb_shm_create_segment() failed for size 1779648
The X11 connection broke: I/O error (code 1)
XIO: fatal IO error 62 (Timer expired) on X server ":0"^M
      after 4027461 requests (4027461 known processed) with 0 events remaining.^M

[Konstantin Petrov (iozho1)]

Also I can easily reproduce by repeatedly calling this command: kdialog --passivepopup 'text' 5