●
Write(/tmp/claude-1000/-home-andreas-liebmann/1088df0d-040c-4d97-91b9-867cfdc236c6/scratchpad/launchpad_comment_draft.txt)
Wrote 19 lines to
/tmp/claude-1000/-home-andreas-liebmann/1088df0d-040c-4d97-91b9-867cfdc236c6/scratchpad/launchpad_comment_draft.txt
Re comment #12: ran the dpkg -l comparison you suggested, and also directly
re-tested with
a newer kernel that's landed since — writing both up together since they touch
the same "
is this actually a local/regression issue" question.
1) Kernel update tested: system is now on 7.0.0-28-generic (up from -27).
Started gdm.serv
ice manually and captured the journal — identical failure, same as every prior
kernel:
gnome-shell[...]: Failed to open gpu '/dev/dri/card1': Failed to initialize
render devic
e for /dev/dri/card1: Failed to create gbm device: Invalid argument, EGLStream
render devi
ce requires an EGL display
strace[...]: Failed to setup: No GPUs found
gdm3[...]: GdmLocalDisplayFactory: maximum number of display failures
reached. Giving up
.
So the kernel bump doesn't change anything either.
2) Package comparison: downloaded the official
ubuntu-26.04-desktop-amd64.manifest from re
leases.ubuntu.com and diffed it against `dpkg-query -W
-f='${Package}:${Architecture}\t${V
ersion}\n'` on the affected system (no need to re-boot the live USB for this).
- Every package in the
gnome-shell/mutter/mesa/libgbm/libdrm/libegl/systemd-udev stack mat
ches the live image, with the sole exception of routine point-release bumps
that have land
ed via -updates since the ISO was built (gnome-shell 50.1-0ubuntu1 ->
50.1-0ubuntu1.1, lib
mutter-18-0/mutter-common 50.1-0ubuntu2 -> 50.1-0ubuntu2.2). I'd already diffed
those spec
ific changelogs earlier in this bug and confirmed none of the entries touch
KMS/GBM/render
er code.
- 171 packages present on the live image are absent on the installed system.
Went through
the full list by hand: all fall into expected categories — installer-only
tooling that has
no business being on an installed system (casper, gparted, user-setup,
grub-pc-bin — this
machine boots UEFI, not BIOS), unused-language
packs/dictionaries/input-methods (fr/es/it
/pt/ru/zh, mozc, ibus-*), filesystem tooling for types not in use here (LVM,
ZFS, btrfs),
and one metapackage-naming false positive (initramfs-tools itself isn't
installed, but ini
tramfs-tools-core/-bin are, and update-initramfs works — confirmed, since I'd
used it dire
ctly in an earlier round of this bug).
- ~300 packages have differing versions, essentially all ordinary -updates SRU
bumps accum
ulated since the ISO was built.
None of this intersects with the actual failing code path. Earlier in this bug
I gdb'd dir
ectly into gbm_create_device() and traced it precisely: it opens /dev/dri,
reads the PCI d
evice's own sysfs attributes
(uevent/vendor/device/subsystem_vendor/subsystem_device) four
times identically, and returns EINVAL without ever issuing an ioctl or opening
any other
file (driver .so, config file, etc.). A missing/extra/differently-versioned
package would
need to sit somewhere in that exact path to explain the bug, and the diff above
doesn't turn one up — libgbm1, mesa-libgallium and the DRI driver blob are all
byte-identical in version to the live image and already passed dpkg -V.
Given the package inventory itself looks clean, I think the remaining
plausible direction is state/config drift that dpkg doesn't track at all
— /etc modifications, stale udev rules, or cached state left behind by
the mass package-purge/reinstall cascade this machine went through in
early July, rather than a wrong package version. Happy to compare
specific config paths (e.g. /lib/udev/rules.d,
/usr/lib/systemd/logind.conf.d) against the live image if that's a
useful next step, or to try whatever else you'd suggest.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2160652
Title:
gnome-shell/mutter 50.1 fails to start on Intel Alder Lake iGPU:
gbm_create_device() returns EINVAL ("Failed to open gpu ... No GPUs
found"
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/gnome-shell/+bug/2160652/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
