NVIDIA / NVIDIA/open-gpu-kernel-modules
Hard kernel crash: nvidia-open 580.159.03 on dual RTX 2080 Ti (kexec/kdump triggered on GPU idle transition)
Nobody has claimed this yet.
- Dominant language
- C
- Stars
- 17.4k
- Forks
- 1.9k
- PR merge metrics
- No merged PRs in 30d
Description
Environment
- OS: Ubuntu 26.04, kernel 7.0.0-22-generic
- Driver: nvidia-open 580.159.03
- GPU: 2× RTX 2080 Ti 22GB + NVLink
- CPU: i5-12400, 48GB RAM
Behavior
System hard-crashes (kernel panic → kexec/kdump → reboot) when GPU transitions from heavy load to idle. Reproduced 3 times in 24 hours, always at the same point: after vLLM inference server finishes processing and GPU utilization drops from 100% to 0%.
No Xid errors, no coredump, no kernel log entries before crash. /var/crash/ contains only kexec_cmd and kdump_lock (0 bytes — dump not generated).
Related
Similar report with same driver version on RTX 5070: https://github.com/pop-os/cosmic-comp/issues/2341
Workaround
Switching to closed-source nvidia-driver-580-server resolves the issue (suggests bug is in the open kernel module, not the driver layer).
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reproducing the vLLM inference workload on the listed Ubuntu kernel and dual RTX 2080 Ti setup, focusing on the transition from full load to idle. Inspect /var/crash/ and the kexec/kdump results; done is a confirmed cause and an appropriate fix or actionable diagnosis for the nvidia-open module.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- c, linux
- Domain
- operating-systems
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100