Dear Maintainer,


CONTEXT
=======

I upgraded this machine from Debian 11 to Debian 12 to Debian 13. On Debian
12
I do not know whether I was running Xorg or Wayland, but everything seemed
to
work fine, which is why I went ahead and upgraded to 13. Debian 13 gave me a
Wayland session by default, and that is when I started having trouble with
this
graphics card.


WHAT I OBSERVE
==============

On a GeForce 210 (GT218, Tesla) using nouveau, the kernel logs PGRAPH faults
attributed to gst-plugin-scan every time the GStreamer plugin registry is
rebuilt. This happens a few seconds into every session since the upgrade. No
application prints anything; I only see it in dmesg.

I do not know which plugin or which Mesa code path is responsible.

In a Wayland session my whole session freezes when this happens, which I
have
reported separately as Debian bug #1147096. I cannot say whether the two are
actually related: the faults below also occur under Xorg, where nothing bad
happens.


HARDWARE AND VERSIONS
=====================

  01:00.0 VGA compatible controller: NVIDIA Corporation GT218 [GeForce 210]
(rev a2)
  Dell Inc. OptiPlex 390/0M5DCD, BIOS A07 05/27/2012
  nouveau 0000:01:00.0: NVIDIA GT218 (0a8280b1), bios 70.18.49.00.00

  libgl1-mesa-dri:                  25.0.7-2+deb13u1
  libglx-mesa0:                     25.0.7-2+deb13u1
  libdrm2:                          2.4.124-2
  gstreamer1.0-gl:                  1.26.2-1+deb13u1
  gstreamer1.0-plugins-base:        1.26.2-1+deb13u1
  linux-image-6.12.107+deb13-amd64: 6.12.107-1

Also seen with linux-image-6.1.0-52-amd64 (6.1.180-1).


REPRODUCER
==========

Running gst-inspect-1.0 reproduces the error every time.

With MESA_LOADER_DRIVER_OVERRIDE=llvmpipe or LIBGL_ALWAYS_SOFTWARE=1 no
fault occurs.


WHAT THE LOGS CONTAIN
=====================

  nouveau 0000:01:00.0: gst-plugin-scan[1386]: channel failed to
initialise, -17
  nouveau 0000:01:00.0: gr: TRAP_PROP - TP 0 - 00000040 [RT_FAULT] -
Address 0000424000
  nouveau 0000:01:00.0: gr: TRAP_PROP - TP 0 - e0c: 00000000, e18:
00000000, e1c: 00000020, e20: 00001800, e24: 00030000
  nouveau 0000:01:00.0: gr: 00200000 [] ch 7 [003f384000
gst-plugin-scan[1386]] subc 3 class 8597 mthd 1b0c data 1000f010
  nouveau 0000:01:00.0: fb: trapped write at 0000424000 on channel 7
[3f384000 gst-plugin-scan[1386]] engine 00 [PGRAPH] client 0b [PROP]
subclient 00 [RT0] reason 00000002 [PAGE_NOT_PRESENT]

and in some runs, additionally:

  nouveau 0000:01:00.0: gr: TRAP_PROP - TP 0 - 00001000
[RT_LINEAR_MISMATCH] - Address 0000000000
  nouveau 0000:01:00.0: gr: TRAP_PROP - TP 0 - e0c: 00000000, e18:
00000000, e1c: 00000020, e20: 00001800, e24: 00220000

The faults come from the VA-API driver:
with LIBVA_DRIVERS_PATH=/nonexistent, or with mesa-va-drivers removed
altogether, the same command produces no TRAP_PROP at all.

The "channel failed to initialise, -17" line is unaffected by any of this
and appears to be a separate issue; I have no explanation for it.