Re comment #5 (overlap with bug 2143073): I have a measurement that separates
the two.
In my case the memory is released when the displays leave power saving, not
when the
session is unlocked (2026-10-02, mutter 50.1-0ubuntu2, kernel 7.0.0-34):
- Session locked, displays in power saving, external monitor hotplugging.
At 00:11, gnome-shell's i915 DRM client (fdinfo drm-total-system0) was 11.2
GB and
Shmem was 11.7 GB, so almost all of the Shmem growth is gnome-shell's.
- 00:13, `loginctl unlock-session`: the session unlocked
(ScreenSaver.GetActive=false),
but PowerSaveMode stayed 3. Shmem kept climbing, 14.8 -> 16.4 GB by
00:14:48.
- 00:15, set org.gnome.Mutter.DisplayConfig PowerSaveMode=0 (still unlocked):
Shmem 16.4 -> 1.06 GB within seconds.
The memory belongs to gnome-shell's DRM client and is freed by the mode set on
wake,
not by unlocking. That matches #4101 (detached_onscreens only cleared in
post_mode_set_updates).
ENVIRONMENT
-----------
Ubuntu 26.04.1, kernel 7.0.0-38-generic, Intel Arrow Lake-P [8086:7d51], i915
gnome-shell 50.1-0ubuntu1.3, mutter 50.1-0ubuntu2.4, GNOME on Wayland.
Lid closed, 4 external monitors: 3 via a Thunderbolt dock (MST), plus a
portable USB-C
monitor. When the portable monitor has power but no signal, its firmware
cycles through
its inputs indefinitely. While it does that, its connector disconnects and
reconnects
over and over (4-18 times a minute in my measurements). I'd expect any
monitor that
auto-scans its inputs like this to trigger it too, but I've only tested this
one.
STOCK vs FIX (same machine, session locked, displays in power saving)
------------------------------------------------------------------
50.1-0ubuntu2.4: 14 hotplugs in 2m20s, 464 -> 1410 MB (~68 MB per hotplug),
still growing.
50.1-0ubuntu2.4 + gnome-50 af0c980, 83c465f, c145106 (built locally):
10 min test: 57 hotplugs, 304-411 MB, no growth
overnight: 6h45m locked and blanked, monitor flapping throughout,
340-470 MB
Since mutter falls under the GNOME micro-release exception, is a 50.5 update
for resolute
planned? I'm happy to test it from -proposed on this setup. In case it helps a
cherry-pick
instead: the three commits apply to 50.1-0ubuntu2.4, but c145106 needs fuzz,
and three
test-exported symbols have to be added to debian/libmutter-18-0.symbols.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2167374
Title:
gnome-shell leaks scanout dma-bufs while locked (fixed in mutter 50.5)
To manage notifications about this bug go to:
https://bugs.launchpad.net/mutter/+bug/2167374/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
