Suspend/resume freeze (GSP panic XID 120) with RTX A5000 Mobile (Ampere) on nvidia-open-dkms 590.48.01/595

Graphics / Linux Linux
1

Hardware:

  • HP ZBook Fury 17 G8

  • CPU: Intel i7‑11850H

  • GPU1: Intel UHD Graphics (integrated)

  • GPU2: NVIDIA RTX A5000 Mobile (Ampere)

  • Driver: nvidia-open-dkms (open kernel modules)

  • Kernel: linux‑lts 6.12.66‑1‑lts (also reproduces on mainline)

Problem description:
When suspending the laptop (to s2idle) and then resuming, the system freezes and a lot of logs and text appears, completely within 1‑2 seconds after resume. Caps Lock LED blinks, no TTY, no mouse. Only forced power off works.
The freeze occurs reliably with driver version 590.48.01 (and likely any 590.x).

What does NOT fix the issue (on 590.48.01):

  • Adding kernel parameters nvidia.NVreg_EnableGpuFirmware=0 and nvidia.NVreg_DynamicPowerManagement=0x03

  • Switching from open modules to proprietary modules

  • Using only the integrated GPU (NVIDIA offloading still causes the same panic on resume)

What DOES fix the issue:
Downgrading to nvidia-open-dkms 580.95.05 (with matching nvidia‑utils and lib32‑nvidia‑utils). After downgrade, suspend/resume works perfectly.

The crash is triggered during resume, right after PM: suspend exit. The kernel log shows a GSP task panic followed by multiple assertion failures and a page fault in nvidia_modeset.

Jan 27 17:30:33 artbook kernel: PM: suspend exit
Jan 27 17:30:33 artbook kernel: NVRM: kgspHealthCheck_TU102: ****************************** GSP-CrashCat Report *******************************
Jan 27 17:30:33 artbook kernel: NVRM: kgspPrintGspBinBuildId_IMPL: GSP bin buildId: cf812cb3f2f1e8c8209dc2e446fdf536ba9ec88f
Jan 27 17:30:33 artbook kernel: NVRM: GPU at PCI:0000:01:00: GPU-d6d45493-ad27-63b7-7786-8e7c414c8bce
Jan 27 17:30:33 artbook kernel: NVRM: Xid (PCI:0000:01:00): 120, GSP task panic: unknown error (0) @ pc:0x19da988, aux:0x0, partition:2#0, task:3
[...]
Jan 27 17:30:33 artbook kernel: NVRM: GPU0 nvCheckOkFailedNoLog: Check failed: Reset required [NV_ERR_RESET_REQUIRED] (0x00000062) returned from kgspWaitForRmInitDone(...)
[...]
Jan 27 17:30:33 artbook kernel: BUG: unable to handle page fault for address: 0000000000002610
Jan 27 17:30:33 artbook kernel: #PF: supervisor read access in kernel mode
Jan 27 17:30:33 artbook kernel: RIP: 0010:DisableVblankSemControl+0x1f/0xa0 [nvidia_modeset]
[...]
Jan 27 17:30:33 artbook kernel: Call Trace:
Jan 27 17:30:33 artbook kernel:  nvEvoDisableVblankSemControl+0x37/0x60 [nvidia_modeset]
Jan 27 17:30:33 artbook kernel:  FreeDeviceReference+0x7d/0x180 [nvidia_modeset]
Jan 27 17:30:33 artbook kernel:  nvRevokeDevice+0x8c/0x90 [nvidia_modeset]
Jan 27 17:30:33 artbook kernel:  nvResumeDevEvo+0x1ef/0x200 [nvidia_modeset]
Jan 27 17:30:33 artbook kernel:  nvKmsResume+0x5c/0xc0 [nvidia_modeset]
Jan 27 17:30:33 artbook kernel:  nvkms_resume+0x1f/0x50 [nvidia_modeset]
Jan 27 17:30:33 artbook kernel:  nv_set_system_power_state+0x108/0x540 [nvidia]

Full logs available:

Additional observations

  • The issue is specific to driver version 590.x (including 590.48.01).
  • Driver 580.95.05 does not exhibit the problem.
  • The crash occurs regardless of whether the NVIDIA GPU is actively used before suspend (hybrid mode).
  • The GSP firmware build ID involved: cf812cb3f2f1e8c8209dc2e446fdf536ba9ec88f

For now, I am using the downgraded 580.95.05 driver and have locked the packages with IgnorePkg. I want to upgrade and solve this error

Note: I used AI to write this because i don’t understand a lot about GPUs and I don’t have the abilities to understand logs and read them and write about them, I am open to provide more logs or anything if that helps, I upgraded to 595 and same error happened, and I am sending this to see if the error is solved in version 610.