[GB206] RTX 5060 Ti: display engine wedges on power transitions under Wayland (Xid 56 + GSP watchdog + nvResumeDevEvo oops)

[kreigiron]

NVIDIA Open GPU Kernel Modules Version

NVIDIA Open GPU Kernel Modules Version 610.57.04 (also reproduced on 595.71.05)

Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.

• I confirm that this does not happen with the proprietary driver package.

Operating System and Version

Arch Linux (rolling), up to date as of 2026-08-18

Kernel Release

Linux 7.1.8-arch1-3 (also reproduced on 6.18.44-1-lts), stock distro kernels

Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.

• I am running on a stable kernel release.

Hardware: GPU

GPU 0: NVIDIA GeForce RTX 2060 (UUID: GPU-8c654f45-99de-273f-468d-3ffdb2d1e96c) GPU 1: NVIDIA GeForce RTX 5060 Ti (UUID: GPU-37e396cb-6b69-5ca6-0a09-f21df81af63c)

Describe the bug

Suspend/resume/logout/shutdown all hang or wedge the display engine, but only when a Wayland compositor is the DRM master. X11 sessions, the X11-based SDDM greeter, and TTY all suspend and resume cleanly.

Four failure modes on the same display engine:

1. Suspend entry hang (always). systemctl suspend from a Wayland session hangs at PM: suspend entry (deep) and never reaches S3. Same second:
   • Xid 56 (CMDre 00000000 00000200 00000001 00000005 00200031)
   • Lost display notification
   • Flip event timeout on head 0/1 every ~3 s
   • GSP task watchdog timeout @ partition:4#0 task:3 + GSP-CrashCat every ~30 s
   • Hard power-cycle required.
2. Resume wedge. After a clean suspend from an X11 path, a later Wayland suspend/resume produced Xid 56 ~42 s after PM: suspend exit, followed by Idling display engine timed out: 0x0000ca7e:4:0:1174 spam.
3. Resume oops (610.57.04). After resume, repeated Assertion failed: !pGpu->getProperty(... PDB_PROP_GPU_IN_PM_CODEPATH) @ mem_mapper.c:49, then kernel tried to execute NX-protected page — Oops 0011 in nvShutDownApiHeads+0x211, called from nvResumeDevEvo+0x136 [nvidia_modeset].
4. Logout/shutdown. Logging out of a Wayland session gives a black screen (hard reset); shutdown/reboot logs Idling display engine timed out ~20x over ~38 s.

The same crash signature reproduces with two different GSP bin buildIds (595: 6332050f... vs 610: 4f09703c...), so it is not a single firmware blob.

To Reproduce

• CPU AMD Ryzen 9 5900XT, ASUS PRIME X570-P, BIOS 5044, two monitors (Samsung DP-2 + Gigabyte HDMI-A-3) on the 5060 Ti.
• Arch Linux, nvidia-open 610.57.04, kernel 7.1.8.
• Log into KDE Plasma (Wayland) or GNOME (Wayland).
• Run systemctl suspend.
• Expected: suspend to S3 and resume.
• Actual: hard hang at suspend entry, Xid 56 + GSP watchdog, no resume, power-cycle required.
• Control: the same suspend from XFCE (X11), the SDDM greeter (X11), or a TTY suspends and resumes without error.

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz

More Info

Ruled out:

• Second GPU (bound RTX 2060 to vfio-pci — identical failure)
• Kernel version (6.18 LTS and 7.1 both fail)
• Driver version (595 and 610)
• Suspend mode (deep and s2idle)
• pcie_aspm=off
• BIOS 5031 → 5044
• Clock lock (nvidia-smi -lgc)
• nvidia-sleep.sh VT-switch timing (manual chvt 63 + sleep 3 hook)
• systemd SYSTEMD_SLEEP_FREEZE_USER_SESSIONS=true

Note on the proprietary-driver checkbox: GB206 does not support the closed driver, so an A/B comparison is not possible on this hardware.

[Counta]

Another reproduction — RTX 5060 Laptop (GB206M), nearly identical env.

Hardware/software:

• Laptop: MECHREVO JIAOLONG Series, hybrid graphics (AMD Raphael iGPU + NVIDIA dGPU)
• GPU: NVIDIA GeForce RTX 5060 Max-Q / Mobile (GB206M, PCI 10de:2d19, rev a1)
• OS: Arch Linux (rolling), kernel 7.1.8-arch1-3
• Driver: nvidia-open 610.57.04 (distro prebuilt package, not DKMS)
• Compositor: niri (Wayland)
• Suspend: s2idle only (no S3 available on this laptop)

Same symptom: suspend/resume hangs; after resume the log floods with

nvidia-modeset: ERROR: GPU:0: Error while waiting for GPU progress: 0x0000ca7d:0 2:2:0:4040

repeating every ~5 s indefinitely (error code 0x0000ca7d:0 2:2:0:4040, same family as the 0x0000ca7e:4:0:1174 in the report).

Journal chain around the failed resume (may help narrowing it down):

NVRM: kfspWaitForResponse: FSP command timed out
NVRM: kfspPrepareBootCommands_GH100: Clock boost feature cmd 1 via FSP failed with error 0x65
NVRM: kfspPrepareBootCommands_GH100: Preparing FSP boot cmds failed. RM cannot boot.
NVRM: gpuPowerManagementResume: GSP boot preparation failed at resume (bootMode 0x1): 0x65
NVRM: GPU0 gpuHandleSanityCheckRegReadError_GH100: Possible bad register read: addr 0x110094, regvalue 0xbadf4100
NVRM: GPU0 _kgspIsHeartbeatTimedOut: Heartbeat timed out
NVRM: GPU0 nvAssertFailedNoLog: Assertion failed: GPU marked for reset @ kernel_gsp.c:2778

So on this machine the GSP firmware fails to re-initialize on resume (RM cannot boot), after which nvidia-modeset waits for GPU progress forever. Happy to attach a full nvidia-bug-report.log.gz if useful.