NVIDIA suspend fails with video-memory preservation · Linux Fix Lab

Symptoms & scope

Sleep is aborted or returns immediately.
After resume, textures are corrupt or GPU applications fail; this symptom alone does not identify the interface mismatch.

Relevant environment

NVIDIA systems using NVreg_PreserveVideoMemoryAllocations and the driver’s documented system power-management interfaces.

Recognizable messages (synthetic examples)

NVRM: GPU 0000:01:00.0: PreserveVideoMemoryAllocations module parameter is set. System Power Management attempted without driver procfs suspend interface.

Reconcile the driver’s integration instead of assuming the platform cannot suspend.

Possible causes

These are possible explanations, not a confirmed diagnosis. Several independent faults can coexist.

Full preservation may be enabled without the required suspend interface integration.
The configured backing filesystem may not support the required temporary files or lack capacity; other resume bugs remain possible.

Diagnose safely

Run one command at a time in the relevant session. Read the explanation first. Uppercase placeholders need your own values; tools and privileges vary by distribution. These commands are displayed here and never executed by the website.

Check 1

Read active NVIDIA module parameters; note PreserveVideoMemoryAllocations and TemporaryFilePath.

cat /proc/driver/nvidia/params

Interpret the result: A preservation value of 1 requires the matching documented integration for this driver version. The temporary path must be evaluated on this machine.

Check 2

Inspect distribution-provided sleep units without starting or enabling them; names may differ by packaging.

systemctl status nvidia-suspend.service nvidia-resume.service nvidia-hibernate.service --no-pager

Interpret the result: Missing or failed units warrant checking the package’s integration. Inactive oneshot units after execution can be normal; read their journal and dependencies.

Evidence-guided next steps

Align parameters with packaged sleep integration

If the driver explicitly reports preservation without its required procfs interface, use the distribution’s documented suspend integration for that installed driver. Ensure suspend, resume and hibernate hooks are configured together before a controlled test.

Precautions: The cited NVIDIA README describes one release, not universal defaults. Save work and test suspend before hibernation; do not write commands directly to the procfs interface.

Recovery / rollback: Restore the recorded package default parameters and unit configuration, then reboot if changed module parameters must be reloaded.

Did this solution help you?

Share this solution #

Provide suitable preservation backing space

If full preservation is needed and space is insufficient, choose a supported backing filesystem with room for the combined GPU memory plus the README’s margin, and configure its path through distribution settings.

Precautions: A small tmpfs can consume RAM and prevent saving. Keep encrypted data handling and early mount availability consistent with the system’s existing setup.

Recovery / rollback: Restore the previous temporary path and preservation setting if the new backing store fails; remove only files created by the test after use.