Thanks for looking into this. Requested info attached: - journal.txt — `journalctl -b0` from a fresh boot today (no leak yet, uptime only minutes). - journal-boot-with-leak-2026-10-10_12.txt.xz — full journal of the boot where the leak happened (2026-10-10 17:35 → 2026-10-12 14:34, +07), containing the repeated OOM kills of gnome-shell. - packages.txt — `dpkg -l` - settings.txt — `dconf dump /org/gnome/shell/` - hotplug.log — my own logger: every DP-1/HDMI-A-1 connector status/DPMS change as it happens, plus a MEM line once a minute (MemAvailable, Shmem from /proc/meminfo, compositor RSS, count of "Cursor update failed" messages; "phút" = minute). It shows DP-1 flapping connected/disconnected ~470-520 times/hour while DPMS is off, with Shmem rising ~25 GB/h until OOM, and Shmem being released when the screens wake.
System: Ubuntu 26.04.1, Intel Alder Lake-S UHD 730 (i915), mesa 26.0.8-1ubuntu0.3, mutter 50.1-0ubuntu2.4, gnome-shell 50.1-0ubuntu1.3. Monitors: LG UltraGear on DP-1, generic HDMI panel on HDMI-A-1. Regarding the other points: * Local extensions: ~/.local/share/gnome-shell/extensions does not exist — I only have the default Ubuntu system extensions (ding, ubuntu-dock, ubuntu-appindicators, tiling-assistant, snapd-prompting, snapd-search-provider, web-search-provider). * Bug 2143073: possibly related, since this is i915 and the leaking memory is Shmem. But note that after my user session was OOM-killed (2026-10-10 21:47), the leak continued in the *GDM greeter* with no user apps running: the greeter's gnome-shell (UID 605xx) was OOM-killed and restarted roughly every 1.7 h (06:09, 07:50, 11:12, 12:53, 02:18 ...). So no app redrawing in the background is needed; the DP-1 hotplug flapping while DPMS is off seems to be the trigger. The problem started right after mutter was updated to 50.1-0ubuntu2.4 on 2026-10-08. MUTTER_DEBUG_DISABLE_HW_CURSORS=1 removes the "Cursor update failed: drmModeAtomicCommit" spam but does NOT stop the leak. * I will now test with mutter 50.1-0ubuntu2.2 and report back (needs an overnight idle test). Apps running when the leak happened on 2026-10-10: Claude Code (CLI) in ptyxis, VS Code, Google Chrome (~10 tabs) — all native Wayland, all idle while the screen was locked. The same apps were open this morning (2026-10-13): screen locked ~2 h, DP-1 still flapped ~500 times/h, but Shmem stayed at 1–4 GB with no OOM, so the leak is intermittent. The leak also continued in the GDM greeter with no user apps running at all, so I don't think an app redrawing in the background is required. ** Attachment added: "log.zip" https://bugs.launchpad.net/ubuntu/+source/mutter/+bug/2170511/+attachment/6008610/+files/log.zip -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2170511 Title: Shmem leak (~25 GB/h) leading to OOM when a DisplayPort monitor repeatedly hotplugs while DPMS is off (regression in 50.1-0ubuntu2.4) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/mutter/+bug/2170511/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
