Intermittent desktop lockup: Xid 62/45 -> Xid 119 (GSP timeout) -> Xid 154 on RTX 5080 (Arch Linux, nvidia-open 590.48.01)

[dlukt]

Suggested title:
Intermittent desktop lockup: Xid 62/45 -> Xid 119 (GSP timeout) -> Xid 154 on RTX 5080 (Arch Linux, nvidia-open 590.48.01)

NVIDIA Open GPU Kernel Modules Version
590.48.01 (Arch package: nvidia-open-dkms 590.48.01-4)

Please confirm this issue does not happen with the proprietary driver (of the same version)
I have not yet validated this with the proprietary kernel module of the same version.
(Arch repos currently provide nvidia-open* packages on this system; I can test proprietary .run driver if required.)

Operating System and Version
Arch Linux (rolling)

Kernel Release
Linux dev.luketic 6.18.13-arch1-1 #1 SMP PREEMPT_DYNAMIC Wed, 25 Feb 2026 23:12:35 +0000 x86_64 GNU/Linux

Please confirm you are running a stable release kernel
Yes (stable, non-rc kernel release).

Hardware: GPU
GPU 0: NVIDIA GeForce RTX 5080 (UUID: GPU-cccfc25d-fc66-0d40-e289-3b1a1f5d7896)

Describe the bug
System intermittently hard-locks (desktop becomes unusable and requires hard reset).
In affected boots, the kernel logs the same NVRM sequence:

1. Xid 62
2. Xid 45 on a user-space channel (often zed-editor/cinnamon/Xorg)
3. Xid 119 (GSP RPC timeout)
4. Xid 154 (GPU reset required)

Latest incident (absolute timestamps):

• 2026-03-01 12:41:30 CET: Xid 62
• 2026-03-01 12:42:00 CET: Xid 45 (pid=15041, name=zed-editor, channels 0x6/0x7/0x8)
• 2026-03-01 12:42:15 CET: Xid 119 ("Timeout after 45s of waiting for RPC response from GPU0 GSP")
• 2026-03-01 12:42:15 CET: Xid 154 ("GPU recovery action changed ... GPU Reset Required")

After reboot from hard reset, the new boot shows unclean-shutdown recovery indicators:

• journald: system.journal/user-1000.journal "corrupted or uncleanly shut down"
• filesystem journal/log recovery

To Reproduce

1. Boot into X11 Cinnamon session on Arch Linux.
2. Use desktop normally (browser/editor; zed-editor has appeared in recent failing logs).
3. After variable runtime, GPU faults appear in dmesg/journal and desktop locks up.
4. Hard reset required.

Bug Incidence
Sometimes (recurring; seen in multiple separate boots)

nvidia-bug-report.log.gz
I will attach nvidia-bug-report.log.gz generated right after reproducing.
Command used:
sudo nvidia-bug-report.sh

More Info

• Driver/userspace:
  • nvidia-open-dkms 590.48.01-4
  • nvidia-utils 590.48.01-4
  • lib32-nvidia-utils 590.48.01-1
• GPU details:
  • NVIDIA GeForce RTX 5080
  • Driver Version: 590.48.01
  • VBIOS: 98.03.3B.C0.0B
  • PCI Bus: 00000000:01:00.0
• Kernel cmdline:
  • root=UUID=d2f3d84b-3ae0-4098-a3ab-e4961a94a116 rw nvidia-drm.modeset=1 pci=realloc initrd=\initramfs-linux.img
• Repeated historical incidents with same driver branch:
  • 2026-02-11 (kernel 6.18.8): heavy recurrence including Xid 62/45/119/120/154 plus many Xid 109 CTX SWITCH TIMEOUT
  • 2026-02-14 (kernel 6.18.9): Xid 62/109/119/120/154
  • 2026-02-16 (kernel 6.18.9): Xid 62/109/119/120/154
  • 2026-03-01 (kernel 6.18.13): Xid 62/45/119/154

Attachments planned:

• nvidia-bug-report.log.gz
• nvidia-xid-incident-2026-03-01.log (kernel excerpt around incident)
• nvidia-env-2026-03-01.txt (system + package versions)
• nvidia-xid-history-2026-03-01.txt (cross-boot Xid summary)

[dlukt]

Attaching NVIDIA bug report log as requested by the template:

• nvidia-bug-report.log (generated on 2026-03-01 after reboot): https://gist.github.com/dlukt/f5130d3c8e56e634ec138487022a4f4c

If you specifically need the compressed nvidia-bug-report.log.gz artifact instead of plaintext log, I can provide it via an alternate upload method.

[dlukt]

New incident after issue creation; crash frequency appears to be increasing.

Latest crash timeline (boot -1, 2026-03-02 CET):

• 01:53:13: Xid 62
• 01:53:13: Xid 154 (GPU Reset Required)
• 01:53:22: Xid 109 (Xorg, CTX SWITCH TIMEOUT)
• 01:53:28: Xid 119 (GSP Timeout, 6s RPC timeout)
• 01:53:34: Xid 119 again (Xorg)
• 01:53:42: Xid 109 again
• 01:53:43: Xid 120 (cinnamon, GSP task exception)
• 01:53:50 onward: repeated "RC watchdog: GPU is probably locked"

This ended in another hard reset.

Attachment with full kernel excerpt for this crash:

• /home/darko/nvidia-xid-incident-2026-03-02.log

Note: if preferred, I can also attach this excerpt as a gist like the bug report log.

[dlukt]

Mitigation test in progress on affected system:

• Rolled back from 590.48.01 to 580.119.02 (nvidia-open-dkms, nvidia-utils, lib32-nvidia-utils)
• DKMS rebuilt for kernel 6.18.13-arch1-1
• Packages pinned temporarily to avoid re-upgrade during stability testing

Will report whether Xid 62/109/119/120/154 hard-lock sequence still reproduces on this branch.