This affects me too, on an Ada GPU. On the same machine the failure only
happens in the Wayland session. "Ubuntu on Xorg" has never hit it.
System: Ubuntu 24.04, GeForce RTX 4060 Ti (AD106), nvidia-driver-580
proprietary DKMS (580.173.02, now 580.178.04), HWE kernels 7.0.0-30 and
7.0.0-38, GNOME Shell 46.0 / mutter 46.2, one DisplayPort monitor,
nvidia_drm modeset=1, NVreg_PreserveVideoMemoryAllocations=1, nvidia-
suspend/resume services enabled, mem_sleep "s2idle [deep]".
All six failures have the signature from the original report:
- "Freezing user space processes failed after 20 seconds", with
task:gnome-shell in state:R in nv_drm_mmap -> __nv_drm_gem_nvkms_mmap ->
__nv_drm_gem_nvkms_map -> nvkms_call_rm -> rm_kernel_rmapi_op
- the s2idle fallback fails the same way, then "Failed to put system to sleep.
System resumed again: Device or resource busy"
- in half of them the kernel also logs nvidia-sleep.sh "blocked for more than
122 seconds" on an rw-semaphore "likely owned by task gnome-shell", and "Failed
to map NvKmsKapiMemory"
Failures by session type, counted from the journal (2026-04-21 to 2026-10-24).
Each logind "The system will suspend now!" is attributed to the session type
reported by the logged-in user's gnome-shell ("as a Wayland display server" vs
"as a X11 window and compositing manager"):
- Xorg session: 0 of 104 suspends failed this way (kernels 6.17.0-22 to
6.17.0-35, 7.0.0-28 and 7.0.0-31; drivers 580.126.09 to 580.173.02)
- Wayland session: 6 of 50 suspends failed (2026-08-27, 08-29, 09-20, 10-07,
10-18, 10-24)
- Closest like-for-like: with 580.173.02, Xorg on 7.0.0-31 was 0/18 and Wayland
on 7.0.0-30 was 3/34. Wayland on 7.0.0-38 + 580.178.04 was 3/16.
(One Xorg boot, on 2026-07-03, ended silently after "PM: suspend entry
(deep)" with no freezer failure logged. I don't count it as this bug.)
Other observations:
- All six were manual suspends from the GNOME menu (idle suspend is off). The
screen lock is disabled on this machine, so there are no unlock events, but the
monitor may have just woken from blanking. Compare the "suspend within ~10 s of
unlock" trigger reported at https://forums.developer.nvidia.com/t/385082
- In 3 of the 6 (08-29, 09-20, 10-18), logind gave up waiting on gnome-shell's
delay lock ("Delay lock is active ... but inhibitor timeout is reached",
InhibitDelayMaxSec=30) before it started nvidia-suspend.service. In those cases
gnome-shell was already unresponsive before the NVIDIA suspend helper ran.
- Outcome varies. In 3 of the 6 (09-20, 10-07, 10-18) the journal stops within
about 2 minutes and the machine needed a hard power-off. In the other 3 it kept
running. On 10-24, systemd sent SIGKILL to nvidia-sleep.sh at the
systemd-suspend.service stop timeout; about 20 s later it exited,
nvidia-resume.service ran, and the session recovered (gnome-shell logged
"Failed to set hardware cursor ... using OpenGL from now on").
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2165319
Title:
GNOME Shell intermittently refuses to freeze during suspend while
blocked in NVIDIA NVKMS with nvidia-driver-580
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-580/+bug/2165319/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
