Bug#1111037: mutter-common: Misbehaviour on AMD A6-6310 APU with AMD Radeon R4 Graphics

[Simon Waters]

Package: mutter-common
Version: 48.4-2
Severity: important
X-Debbugs-Cc: simon at technocool.net

Dear Maintainer,

I did a clean install of Trixie a few days before it became stable on HP
450-a160na. I selected GNOME for my desktop, and UK locale and mirror, but
otherwise a vanilla Debian install. Some user files were restored from
Bookworm, but the only settings restored were in "~/.ssh".

After installation the GNOME desktop was unstable, some actions would
occasionally cause the graphics driver to get into some sort of reset loop,
where the screen would go black, then reappear and work for a few seconds, then
disappear again. Kernel errors are logged, and on one occasion a kernel crash
resulted, Oops caused by corruption in the radeon driver.

After some investigation I established that gnome-shell was unstable when the
file /usr/share/glib-2.0/schemas/10_mutter-common.gschema.override existed.
This instability seemed to also apply when using GNOME on Xorg, I note mutter
still reported it was loading the experimental xwayland scaling module if this
file exists, and Xorg is selected, hopefully it isn't used.

However after removing this file, switching to GNOME on wayland, it seemed
stable for a few days, then after suspending the machine on resuming from
suspend the primary monitor (of a two monitor set up) had the display in a
small rectangle in the middle of the screen surrounded by black, which resisted
all attempts to resize it back to 1920x1080, short of rebooting the machine.
The second monitor wass displaying its usual content at full size. I tried
removing monitors.xml to no available.

At this point I returned to GNOME using Xorg, which seems stable without the
experimental modules, and able to suspend successfully.

This APU is using the radeon driver, this, I understand, is the correct driver
for this APU, I tried blacklisting to get it to use amdgpu to no benefit, and
GNOME on wayland was still unstable.

KDE on Xorg was fine, Bookworm has worked successfully on the same hardware
(less the SSD installed for this reinstall) for many years.

I understand AMD aren't supporting the drivers for this hardware any more, but
feels like we ought to be able to exclude it from experimental features, or
more gracefully handle cases where expected features aren't present. I believe
the same or similar APUs are widely used in older laptops. It is variously
reported as KABINI or MULLINS, lspci reports the graphics as Mullins and the
PSP Security Processor as Kabini/Mullins.

Manually selected error from log

Jul 30 16:51:54 rodney gnome-character[7118]: Vulkan:
../src/amd/vulkan/radv_physical_device.c:2011: Device '/dev/dri/renderD128' is
not using the AMDGPU kernel driver: Invalid argument
(VK_ERROR_INCOMPATIBLE_DRIVER)

Jul 30 16:53:27 rodney gnome-shell[1965]: g_closure_unref: assertion
'closure->ref_count > 0' failed
Jul 30 16:53:27 rodney kernel: radeon 0000:00:01.0: GPU fault detected: 147
0x020a2001
Jul 30 16:53:27 rodney kernel: radeon 0000:00:01.0:
VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x04000810
Jul 30 16:53:27 rodney kernel: radeon 0000:00:01.0:
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0A020001
Jul 30 16:53:27 rodney kernel: VM fault (0x01, vmid 5) at page 67110928, read
from 'TC0' (0x54433000) (32)

Jul 30 16:53:39 rodney kernel: radeon 0000:00:01.0: ring 0 stalled for more
than 10488msec
Jul 30 16:53:39 rodney kernel: radeon 0000:00:01.0: GPU lockup (current fence
id 0x0000000000036c5d last fence id 0x0000000000036c5f on ring 0)
Jul 30 16:53:39 rodney kernel: radeon 0000:00:01.0: ring 2 stalled for more
than 10488msec
Jul 30 16:53:39 rodney kernel: radeon 0000:00:01.0: GPU lockup (current fence
id 0x0000000000000961 last fence id 0x0000000000000965 on ring 2)



-- System Information:
Debian Release: 13.0
  APT prefers stable-updates
  APT policy: (500, 'stable-updates'), (500, 'stable-security'), (500, 'stable')
Architecture: amd64 (x86_64)

Kernel: Linux 6.12.38+deb13-amd64 (SMP w/4 CPU threads; PREEMPT)
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8), LANGUAGE=en_GB:en
Shell: /bin/sh linked to /usr/bin/dash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled

Versions of packages mutter-common depends on:
ii  dconf-gsettings-backend [gsettings-backend]  0.40.0-5

mutter-common recommends no packages.

mutter-common suggests no packages.

-- no debconf information