Update 2026-07-24 — 2 additional occurrences, new kernel and firmware do
not fix the issue

The crash has now occurred 2 times today, all with an identical
signature.

Occurrences:

2026-06-29 09:53:56 (original report)
2026-07-24 08:48:15 (second occurrence)
2026-07-24 16:40:08 (third occurrence — same session, different Chrome PID)
New kernel and firmware installed today did not help:

This morning I ran sudo apt upgrade, which installed:

Kernel 7.0.0-28-generic (previously 7.0.0-27)
linux-firmware-amd-graphics 20260319.git217ca6e4-0ubuntu3.1 (previously 
0ubuntu3, firmware files updated Jun 25)
After rebooting into the new kernel and firmware, the crash occurred again 
within the same session — twice.

The crash pattern is identical across all four occurrences:

Chrome GPU process submits an invalid address to the GPU (PERMISSION_FAULTS: 
0x3, client SQC (data), UTCL2 page fault on gfxhub)
gfx_0.1.0 ring times out with gnome-shell as a victim process
Ring-level reset fails → full MODE2 GPU reset required
gnome-shell loses its GPU context and exits with SIGABRT
Entire GNOME/Wayland session is lost; GDM login screen appears
On the third occurrence (kernel 7.0.0-28), a compute ring (comp_1.1.0) also 
timed out but its reset succeeded — the gfx_0.1.0 reset still failed and 
required the full GPU reset.

Fault addresses seen (all PERMISSION_FAULTS: 0x3):

0x000000003f919000 (Jun 29)
0x0000400040004000 (Jul 24 08:48 — suspicious repeating pattern)
0x000049aa22454000 (Jul 24 16:40)
Saved artefacts:

AMD GPU coredump from 08:48 crash: amdgpu_coredump_20260724_085213.bin 
(available on request)
Full journalctl -b output covering all three Jul 24 events available on request
System: AMD Granite Ridge iGPU (1002:13c0), kernel 7.0.0-28-generic, Mesa 
26.0.3, Chrome 150.0.7871.186, Ubuntu 26.04

Chrome was in use (GitLab in browser tab) during all crashes. No other
GPU-intensive workload was running. The crash does not require a
Teams/video call — ordinary browser use is sufficient to trigger it.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2158614

Title:
  amdgpu: Granite Ridge iGPU gfx ring timeout after Chrome GPU page
  fault causes gnome-shell SIGABRT and session loss

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2158614/+subscriptions


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

Reply via email to