RTX 5070 Ti (GB203, sm_120): `nvidia-smi -q -d SUPPORTED_CLOCKS` crashes GSP firmware (Xid 120 page fault on control 0x20800a4c → Xid 119 RPC timeouts → unrecoverable hang)

[vhrabar]

NVIDIA Open GPU Kernel Modules Version

610.43.02

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

Kubuntu 26.04 LTS

Kernel Release

7.0.0-27-generic # 27-Ubuntu SMP PREEMPT_DYNAMIC Thu Jun 18 19:13:49 UTC 2026 x86_64

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

NVIDIA GeForce RTX 5070 Ti

Describe the bug

On a desktop RTX 5070 Ti (GB203-A, sm_120) with the open kernel module, issuing the read-only clock query

nvidia-smi -q -d SUPPORTED_CLOCKS

crashes the GSP firmware and hard-hangs the machine. The GSP task page-faults while servicing the clock/perf control GSP_RM_CONTROL 0x20800a4c (Xid 120), after which every RPC, including the nvidia-smi clock query itself, times out (Xid 119), the GSP heartbeat freezes and never recovers, and the host locks up. A warm reboot stalls on shutdown (for hours if not interrupted) because the GPU is wedged; only a full power cycle recovers.

To Reproduce

1. Boot normally, open module loaded.
2. Run nvidia-smi -q -d SUPPORTED_CLOCKS.
3. GSP firmware faults; the command hangs and the host becomes unresponsive (display frozen if GPU-driven, subsequent nvidia-smi never returns, etc.).
4. Recovery requires a hard power cycle (warm reboot hangs on shutdown).

Reproduced repeatedly.

NOTE: The issu does not reproduce on pre-610 drivers.

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz

GSP firmware page-faults servicing control 0x20800a4c:

NVRM: Xid (PCI:0000:01:00): 120, GSP task exception: load access page fault (cause:0xd) @ pc:0x13b86d6, partition:4#0, task:3, gfid: 0
NVRM: GPU0 GSP RPC buffer contains function 76 (GSP_RM_CONTROL) sequence 10802 and data 0x0000000020800a4c 0x0000000000000004.
NVRM: GPU0 krcRcAndNotifyAllChannels_IMPL: RC all user channels for critical error 120.

GSP heartbeat then freezes (heartbeat value stuck at 3059099, diff grows unbounded):

NVRM: GPU0 _kgspIsHeartbeatTimedOut: Heartbeat timed out, currentTimeMs 1195377247 heartbeat 3059099 heartbeatWithOffsetMs 1195370837 diff 6410 timeout 5200

nvidia-smi's clock-query RPC times out on the same control, repeatedly, and never recovers:

NVRM: Xid (PCI:0000:01:00): 119, pid=8304, name=nvidia-smi, Timeout after 6s of waiting for RPC response from GPU0 GSP! Expected function 76 (GSP_RM_CONTROL) sequence 8788 (0x20800a4c 0x4).
NVRM: Xid (PCI:0000:01:00): 119, Timeout after 6s of waiting for RPC response from GPU0 GSP! Expected function 76 (GSP_RM_CONTROL) sequence 8789 (0x20808530 0x3d4).
NVRM: Xid (PCI:0000:01:00): 119, pid=8304, name=nvidia-smi, Timeout after 6s of waiting for RPC response from GPU0 GSP! Expected function 76 (GSP_RM_CONTROL) sequence 8790 (0x20800a4c 0x4).
NVRM: GPU0 _kgspIsHeartbeatTimedOut: Heartbeat timed out, ... heartbeat 3059099 ... diff 19216 timeout 5200

More Info

• Not a hardware/power/thermal issue: hours of 4K path-traced gaming or ML workloads are stable, OCCT stress-test passes, and Windows is stable for the same workloads. The fault is specific to the open-module GSP control path.
• Warm reboot hangs on shutdown; only a hard power cycle recovers.