Same symptom on different hardware — RX 9060 XT (RDNA4)

Adding a data point from a different GPU generation, since this doesn't
appear to be RX 7900 XT/RDNA3-specific.

Hardware: AMD RX 9060 XT / Navi 44 (RDNA4), PCI ID [1002:7590]
Kernel: 7.1.2-070102-generic (mainline build, #202606271039 SMP
PREEMPT_DYNAMIC, Sat Jun 27 2026)

Symptom: Identical "dm_handle_vmin_vmax_update [amdgpu] hogged CPU for
>10000us N times" workqueue warnings, recurring across boots.

Example from one boot (kernel monotonic timestamps, wall-clock range
Jul 11 19:21:52 - Jul 12 00:37:29 — note the count climbs across the
boot rather than resetting):

[  370.797119] workqueue: dm_handle_vmin_vmax_update [amdgpu] hogged CPU for 
>10000us 4 times, consider switching to WQ_UNBOUND
[  455.037142] workqueue: dm_handle_vmin_vmax_update [amdgpu] hogged CPU for 
>10000us 5 times, consider switching to WQ_UNBOUND
[ 4037.810198] workqueue: dm_handle_vmin_vmax_update [amdgpu] hogged CPU for 
>10000us 7 times, consider switching to WQ_UNBOUND
[ 5131.830194] workqueue: dm_handle_vmin_vmax_update [amdgpu] hogged CPU for 
>10000us 11 times, consider switching to WQ_UNBOUND
[19307.125552] workqueue: dm_handle_vmin_vmax_update [amdgpu] hogged CPU for 
>10000us 19 times, consider switching to WQ_UNBOUND

Ruled out via systematic testing (12+ hours real usage under all three
conditions simultaneously):

1. VRR/FreeSync — confirmed disabled at both the Xorg driver level
   (VariableRefresh: disabled) and the KWin compositor level
   (vrrPolicy: Never on both outputs). Warning still fired.

2. DPM/mclk power-state switching — GPU pinned to
   power_dpm_force_performance_level=high for the entire session
   (static top pstate, no automatic transitions). Warning still fired.

3. Refresh rate/modeset switching — both monitors static at
   3440x1440@60Hz for the whole boot, no modeset or hotplug events in
   kernel or Xorg logs. Warning still fired.

Since all three of the obvious triggers were independently ruled out,
this points toward the underlying mechanism rather than any of the
usual suspects (adaptive sync, power states, or display timing
changes).

Looked into the driver code and found dm_handle_vmin_vmax_update is a
workqueue handler introduced specifically to offload
dc_stream_adjust_vmin_vmax() off the interrupt handler so it can wait
on dc_lock instead of racing it (see the amd-gfx patch thread on "fix
dmub access race condition"):
https://www.mail-archive.com/[email protected]/msg126413.html

So the warning itself looks like the race-condition fix's own offload
mechanism hitting contention, rather than being triggered by any
user-facing setting.

Happy to provide full journalctl -k output or test a debug kernel
build if that'd help narrow it down further.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2150776

Title:
  Ubuntu 26.04 GNOME Wayland: random short display/presentation freezes
  on AMD RX 7900 XT while apps continue running

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2150776/+subscriptions


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

Reply via email to