> crashes

Please provide a backtrace that includes debugging information. Please read the following links [1][2][3]. It's usually as simple as (with gdb installed):

$ coredumpctl gdb (then answer y when it asks "Enable debuginfod for this session?")
(gdb) set logging enabled
(gdb) bt (or bt full)

Then post gdb.txt

[1] https://blogs.gnome.org/mcatanzaro/2021/09/18/creating-quality-backtraces-for-crash-reports/
[2] https://wiki.archlinux.org/title/Debugging/Getting_traces
[3] https://wiki.archlinux.org/title/Core_dump#Analyzing_a_core_dump

I can confirm that gnome shell is crashing. Disabling the extensions prevents crashes.

Same issue here. I tried getting a backtrace, but for some reason gdb doesn't seem to find the symbols (even though debuginfod is enabled and it downloaded them).

@Benjamin - were you able to narrow it down to a particular extension? Or did you just disable them all?

One of these activated extensions. I have no time for a deeper research.

The only ones I have in common with that are:
- AppIndicator and KStatusNotifierItem Support
- Places Status Indicator
- Workspace Indicator

I don't have time to test further right now, but if disabling extensions works then perhaps it's one of those 3 that's triggering the crash.

It appears as if uninstalling the gnome-shell-extension-appindicator package and then restarting the GNOME session stops the crashes from occurring.

[EDIT]
I've now experienced a crash even without the gnome-shell-extension-appindicator package installed/enabled, so chances are the crashes are not caused by that extension after all.

My latest crash occured whilst trying to enable a VPN connection from gnome-shell's menu in the upper right corner of the screen. Everything froze for a few seconds, then the gdm login screen appeared and my gnome session was gone.
[/EDIT]

I've also had a brief/superficial look at the git commits between the last and the current version of our gnome-shell package (https://gitlab.gnome.org/GNOME/gnome-shell/-/compare/ebf2f80360f152c9d071a16c5caeecb4bd0927c4...45.1) and tried reverting upstream commit 5de44898d9d86378e1ebfecc4d69e314268c40a0, but that didn't stop the crashes.

Reverting upstream commits 902c373249c86e3ce570d1fbd64bacef6261a8df, ef7a14c69dd6d85237b2f2262cef04fc1a33f446, 7d2014da71c0ba8e06a1d791cd9170017ae5b781 and ae5e8fa924fb073901a6b51a6d099a72733ba792 in gnome-shell has fixed the reproducible and immediate crashes when activating a VPN connection from the gnome shell menu in the upper right screen corner on my machine.

I should add though that I use a somewhat non-standard setup, because I don't use gnome-keyring, but rather keepassxc, which is configured to open a "do you really want to allow this application to access this credential" dialogue when enabling the VPN connection.

I'll report back if this fixes all of the crashes, or only the specific case of enabling the VPN connection.

I have a plethora of dumps from gnome-shell thanks to systemd-coredump. A few of them size range 100MB contained to me what just looked like numbers but I found a smaller one ~27.MB that gave me some output in gdb.txt.

Here's another upstream report that matches more closely
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/7158

Merging  FS#80154  here

Toolybird you closed my other issue but mine happens to be specific to the Linux testing kernel package which does not seem to be the case here. Reverting either Linux or using the LTS kernel resolved the issue in my case.

> mine happens to be specific to the Linux testing kernel package

Everything applicable is within this ticket. Hint: please provide a backtrace containing debugging information as per above.

Please find attached the trace.

Problem exists only with kernel 6.6 (in testing), not observed with 6.5.9 (in core) or with linux-lts.

Unfortunately, the crashing thread (Thread 1) had no frame information at all, so it's useless.

Upstream report https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/7158 happens on `linux-lts` and `linux` kernels.

To me it looks now like both upstream reports
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/7158
https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/7152
are unrelated to the issue in this bug.

I also experienced gnome-shell constantly crashing after the update. It would always crash after opening a few Firefox tabs and windows and sometimes after opening other programs too. In my case disabling the Workspace Indicator extension fixed the issue.

Adding that GNOME 45 as a whole might be unstable, I'm experiencing frequent crashes, e.g at the following:
- Wacom Tablet first input or extended usage
- Opening OpenGL apps
- Opening Rider
- Opening the lid of my laptop
only to list the few that I've been experiencing after specific actions so far, with even more happening.
I cannot reproduce most of these, as they aren't consistent.
Extensions don't seem to matter.

Currently cannot get anything done.

If you have amdgpu try `export MUTTER_DEBUG_KMS_THREAD_TYPE=user` to `/etc/profile.d/gnome-bugs.sh` and rebooting.

I found it there https://bugs.launchpad.net/ubuntu/+source/mutter/+bug/2034619

it helped me

It's tricky because there is no surefire way to repro. And of course there could easily be multiple issues at play here..

Anyway, while playing around in a VM (Nvidia passthrough -> Nouveau drivers) I managed to experience a crash after approx' 10 mins of usage. Not sure if relevant, but it only happened after enabling the built-in Workspace Indicator extension. t apply all bt attached.

Edit: https://gitlab.gnome.org/GNOME/gnome-shell-extensions/-/issues/469

For me this is definitely being trigger by the "Workspace Indicator" extension that is bundled with gnome. With that disabled, I've so far not seen any crashes. With it enabled, I can cause a crash just by dragging a window around the screen for a couple of seconds. Sometimes even just toggling on the extension will cause a crash immediately.

I haven't seen any evidence that kernel makes a difference. I have tried on linux-rt-6.5.2.8.realtime1-1 and linux-6.5.9.arch2-1 kernels. I have not tested on the lts kernel though.

The crashes seem to be happening in libmozjs-115.so. I'm still not getting a full set of symbols in the backtrace even with debuginfod. However, I have attached one that does show some symbols.

In case it's relevant - I'm using nvidia-dkms-545.29.02-2 with an RTX 3070.

> For me this is definitely being trigger by the "Workspace Indicator"

Yes, I can repro this one now pretty much at will. I've retested again with latest "gjs-2:1.78.0-3" (thinking it might help) and it still repro's, but now unfortunately the backtraces are all useless because the crashing "Thread 1" frames are all empty. Frustrating...

> crashing "Thread 1" frames are all empty. Frustrating...

It's extremely random whether a meaningful trace is produced or not. Managed to get one just now that seems a bit more "juicy". Hopefully someone can make sense of all this...

@everyone

Is this for you on Wayland, X11, both?

Do you maybe have an experimental feature active

dconf read /org/gnome/mutter/experimental-features

Whats your Hardware and driver?

For me it seems to happen only 'kms-modifiers' is enabled as experimental feature in AMDGPU with both an Ryzen 5 4500U integrated as well as an Rx5600 dedicated graphics on the desktop. After disabling the experimental features I have not experienced a crash so far (last 3h)

There seem to be at least two separate issues currently plaguing people:

- workspace indicator extension causing a crash in js115 since gnome-shell 45.1
- AMDGPU syscalls taking too long and causing the shell to get killed for exceeding its RLIMIT_RTTIME

This bug is about the former.

Does this package prevent the extension crashes? https://pkgbuild.com/~heftig/FS%2380132/gnome-shell-extensions-45.1-1.1-any.pkg.tar.zst

> Does this package prevent the extension crashes?

It definitely seems better. With the repo pkg and workspace indicator extension enabled, I can make it crash within about 1 or 2 minutes. No crashes so far with your pkg.

I also note the other changes you made to "js115" and "gjs". I'm glad you got rid of the -O3 opt because it definitely seemed like a red flag to my untrained eye. Will test these 2 updated pkgs next...

Ok, reverted back to the repo version of g-s-e then updated to latest "js115" and "gjs". It took a little longer but it still crashed :(

So the big question is, what changes were made to the special g-s-e pkg you made? Because it seems like a winner so far..

It reverts https://gitlab.gnome.org/GNOME/gnome-shell-extensions/-/merge_requests/280

That 45.1-1.1 package has been released as 45.1-2

> It reverts

Ah, ok. BTW, concur with your remark in the upstream ticket about something smelling fishy in GJS. I would *really* love to know why all the backtraces are so inconsistent/erratic.

So at this point, we can hopefully conclude that at least the workspace indicator extension issue is resolved.

Is anyone still getting gnome-shell crashes? Are you on AMDGPU?

Just a few minutes ago, I've upgraded again gnome-shell from 1:45.0+r17+gebf2f8036-1 to 1:45.1-1, and gnome-shell-extensions from 45.0-2 to 45.1-2, and until now, I haven't had any crashes. I'll report back if I encounter any issues/crashes.

The laptop on which I'm testing/using this has both an Intel and an AMD GPU, and I *think* only the Intel GPU is actually used, but I'm not 100% certain, and if you can show me a way to tell which of both GPUs is currently used, I'll happily try that and report back. :)

Not a single crash until now, so gnome-shell-extensions 45.1-2 seems to have indeed solved the problem. :)

I've got two computers with amdgpu. One was affected by bug with RLIMIT_RTIME (with rt-scheduler experimental feature) and the other is still affected by crashes on login. Stack traces hinted at user-themes extension. It happened after updating gnome-shell-extensions to 45.1-2 and before.
I've got (mostly) same extensions installed on both computers but only one is affected.

I updated to gnome-shell-extensions 45.1-2 and reenabled the Workspace Indicator extension. It's been 6 hours and no crash so far.

dGPU: Intel ARC A770 (used for display)
iGPU: AMD Raphael (Ryzen 9 7900 CPU)

Updated to gnome-shell-extensions 45.1-2. No crashes so far in almost 3 hours.

45.1-3 contains upstream's fix instead of a revert.

45.1-3 is working for me as well. On 45.1-1, I could easily reproduce the crash (enable workspace indicator extension and drag a window around). No crashes at all on 45.1-3, despite my best efforts to cause it.

Hi!

I have just upgraded gnome-shell-extensions to 45.1-3 and cannot enable extensions anymore. It crashes instantly. I'm on Wayland.

```
[ 22.366888] warning: `ThreadPoolForeg' uses wireless extensions which will stop working for Wi-Fi 7 hardware; use nl80211
[ 52.837171] dleyna-renderer[3903]: segfault at 5607e9af4160 ip 00005607e9af4160 sp 00007ffdf41ddb68 error 15 likely on CPU 22 (core 6, socket 0)
[ 52.837179] Code: 00 00 21 00 00 00 00 00 00 00 02 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 81 00 00 00 00 00 00 00 <00> 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00 af 46 ab e9 07 56
[ 56.709689] gnome-shell[1548]: segfault at 56297ae2e000 ip 00007f111117ec95 sp 00007ffe0021b9d0 error 4 in libst-13.so[7f1111149000+6b000] likely on CPU 2 (core 2, socket 0)
[ 56.709702] Code: 05 00 48 85 db 74 5b 4c 63 e8 4c 89 e6 48 89 df ff 15 c7 c8 05 00 48 85 c0 74 3c 48 8b 15 5b ca 05 00 48 8b 2a 48 39 c3 74 0b <0f> b6 50 ff f6 44 55 01 01 74 10 42 0f b6 14 28 84 d2 74 19 f6 44
[ 61.203276] rfkill: input handler enabled
[ 63.597040] rfkill: input handler disabled
[ 69.412486] rfkill: input handler enabled
[ 70.465118] rfkill: input handler disabled
[ 79.176753] gnome-shell[5537]: segfault at 558904964000 ip 00007f71c53bfc95 sp 00007fff1223c4e0 error 4 in libst-13.so[7f71c538a000+6b000] likely on CPU 8 (core 8, socket 0)
[ 79.176766] Code: 05 00 48 85 db 74 5b 4c 63 e8 4c 89 e6 48 89 df ff 15 c7 c8 05 00 48 85 c0 74 3c 48 8b 15 5b ca 05 00 48 8b 2a 48 39 c3 74 0b <0f> b6 50 ff f6 44 55 01 01 74 10 42 0f b6 14 28 84 d2 74 19 f6 44
[ 81.206594] rfkill: input handler enabled
[ 83.638957] rfkill: input handler disabled
[ 95.202438] rfkill: input handler enabled
[ 96.232090] rfkill: input handler disabled
```

> cannot enable extensions anymore. It crashes instantly

This looks completely unrelated to the issues in this ticket. Please do not post *obviously* useless backtraces. Please reread the first comment to learn how to produce a useful backtrace. (Although as we have seen above, sometimes the backtraces produced for gnome-shell are randomly not helpful, for unknown reasons).

Anyway, please visit the Arch support channels for troubleshooting assistance with your issue.

This ticket has gone on for too long so I'm closing it here and now because the main culprit has been fixed. If anyone experiences further gnome-shell crashes, please reproduce persistently until you can provide a quality backtrace, then report it upstream (unless it is proved to be an Arch packaging issue).