gnome-shell crashes (traps) with int3 on launch using nvidia proprietary drivers

[Sidney Kelley]

I have a working backup with no nvidia drivers. Adding nvidia proprietary driver (nvidia-driver-535) and rebooting causes a generic "contact a system administrator" screen when one would expect the greeter. The ubuntu-desktop and nvidia driver metapackages were installed with --no-install-recommends. A bug report has been filed with gnome-shell, and someone there asked me to file it here as well.

https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/7721

ProblemType: Bug
DistroRelease: Ubuntu 24.04
Package: gnome-shell 46.0-0ubuntu5.1
ProcVersionSignature: Ubuntu 6.8.0-35.35.1-lowlatency 6.8.4
Uname: Linux 6.8.0-35-lowlatency x86_64
NonfreeKernelModules: nvidia_modeset nvidia zfs
ApportVersion: 2.28.1-0ubuntu3
Architecture: amd64
CasperMD5CheckResult: unknown
Date: Fri Jun 14 21:04:31 2024
DisplayManager: gdm3
RelatedPackageVersions: mutter-common 46.0-1ubuntu9
ShellJournal:
 Hint: You are currently not seeing messages from other users and the system.
       Users in groups 'adm', 'systemd-journal' can see all messages.
       Pass -q to turn off this notice.
 -- No entries --
SourcePackage: gnome-shell
UpgradeStatus: No upgrade log present (probably fresh install)

[Sidney Kelley (memtha)]

attaching dmesg dump

[Sidney Kelley (memtha)]

adding journal dump. This includes a few attempts to restart the service manually.

[Sidney Kelley (memtha)]

attaching prevjournal dump

[Sidney Kelley (memtha)]

Installing the kubuntu-desktop metapackage (also with --no-install-recommends) and then switching the default display-manager back to gdm3 fixes this problem. I reverted to a backup to reproduce a problem setup to aid in diagnosing the gnome issue because honestly, that's a pretty dumb solution. My expectation is that we'll figure out which specific packages fix the issue and add it to some documentation somewhere so future users can find it.

[Sidney Kelley (memtha)]

list of packages that would be installed by `apt install kubuntu-desktop --no-install-recommends`

[Daniel van Vugt (vanvugt)]

There are two separate issues here, and both seem to result in an int3:

[ 14.366966] simple-framebuffer simple-framebuffer.0: swiotlb buffer is full (sz: 770048 bytes), total 32768 (slots), used 288 (slots)
[ 14.440052] traps: gnome-shell[1377] trap int3 ip:76f0445210df sp:7ffe06de1b10 error:0 in libglib-2.0.so.0.8000.0[76f0444dd000+a0000]
[ 14.958767] simple-framebuffer simple-framebuffer.0: swiotlb buffer is full (sz: 360448 bytes), total 32768 (slots), used 22 (slots)
[ 15.015203] traps: gnome-shell[1397] trap int3 ip:7f04827000df sp:7ffc222971e0 error:0 in libglib-2.0.so.0.8000.0[7f04826bc000+a0000]

^^^ That's bug 2062426 with workarounds listed in bug 2060268.

Jun 14 20:45:57 acrylic gnome-shell[1377]: No cursor theme available, please install a cursor theme
Jun 14 20:45:57 acrylic gnome-shell[1377]: Received an X Window System error.
                                           This probably reflects a bug in the program.
                                           The error was 'BadCursor (invalid Cursor parameter)'.
                                             (Details: serial 925 error_code 6 request_code 95 (core protocol) minor_code 0)
                                             (Note to programmers: normally, X errors are reported asynchronously;
                                              that is, you will receive the error a while after causing it.
                                              To debug your program, run it with the MUTTER_SYNC environment
                                              variable to change this behavior. You can then get a meaningful
                                              backtrace from your debugger if you break on the mtk_x_error() function.)
Jun 14 20:45:57 acrylic kernel: traps: gnome-shell[1377] trap int3 ip:76f0445210df sp:7ffe06de1b10 error:0 in libglib-2.0.so.0.8000.0[76f0444dd000+a0000]

^^^ That's bug 2062377.

[Daniel van Vugt (vanvugt)]

Due to the delayed nature of X errors, they could both be the same bug 2062377. It's possible that's the only real crash here.