Bug Impact Update: Catastrophic Display Server Crashes & Workspace Wipes

Following up on my previous comments regarding the rare recurrences and
Nathan's attempt to recreate this in comment #18.

A new, suspected-related/derivative bug has manifested twice over the
last 30 days on my end, but with a highly disruptive behavior: it isn't
triggering a full hardware reboot, but rather a catastrophic graphical
display server crash that forcibly kicks me out to the login screen and
obliterates my active workspace session.

To answer Nathan's reproduction questions: this does not happen with a
simple single-package downgrade or install. It requires a heavy
multitasking stack and specific idle states to trip.

Incident 1 (August 16): 
I left the machine entirely idle to run errands for several hours. While away, 
background updates ran and/or pending updates existed. journalctl shows that at 
7:31:39 AM, the graphical session completely imploded under the hood. When I 
returned at 4:36 PM (16:36:31), I was greeted by a fresh login screen and a 
completely wiped workspace.

Incident 2 (August 31): 
I was running Chrome, Brave (installed as a snap per comment #21), gedit, 
Writer, and a VirtualBox Win10 VM. Immediately after executing a file-save 
operation inside the VM (saving a compressed 3D slicing .gcode.3mf file), I 
stepped away for a restroom break. At 11:11:29 AM, the exact same crash 
triggered, kicking me back to the login screen by the time I returned at 11:16 
AM.

In both instances, running uptime reveals the system has been up
continuously for 28+ days—proving the core Linux kernel did not power-
cycle or reboot. Instead, journalctl error logs for the desktop
environment show a catastrophic display server crash tracking exactly to
those timestamps:

Aug 16 07:31:39 intern gdm3[2211]: Gdm: on_display_added: assertion 
'GDM_IS_REMOTE_DISPLAY (display)' failed
Aug 16 07:31:39 intern gdm3[2211]: Gdm: on_display_removed: assertion 
'GDM_IS_REMOTE_DISPLAY (display)' failed
...
Aug 31 11:11:29 intern gdm3[2211]: Gdm: on_display_added: assertion 
'GDM_IS_REMOTE_DISPLAY (display)' failed
Aug 31 11:11:29 intern gdm3[2211]: Gdm: on_display_removed: assertion 
'GDM_IS_REMOTE_DISPLAY (display)' failed

The graphical session (GNOME Shell/Wayland) completely crashed,
instantly terminating all open user processes and wiping out unsaved
data in editors, browsers, and active virtual machines. Immediately
after signing back in post-incident 2, a massive flood of pending
updates arrived that had been silently bottlenecked behind an invisible
reboot flag.

There appears to be an active "ghost lock state" happening here. Because
update-manager fails to properly release its sleep/shutdown inhibitors
when background updates are pending, it corrupts the desktop power
management layer. When the system hits its however-many-second idle
delay, the desynchronization between the locked power hooks, the desktop
idle state, and active virtualization workloads causes a fatal display
server crash instead of a clean screen-blanking.

This issue needs its importance escalated from "Low." This is no longer
just an annoying power-consumption or screen-wear bug; the interaction
between the unreleased inhibitor and active user workloads is actively
causing desktop session terminations and data loss on a production LTS
release (24.04, to preclude scrolling up).

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2148306

Title:
  Idle inhibition persists when updates are pending, preventing screen
  blanking

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/update-manager/+bug/2148306/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to