Skip to content

feat: add gpu container support to microvm driver#1143

Open
drew wants to merge 2 commits intomainfrom
dn/vcaux-brisebo/vm-container-support
Open

feat: add gpu container support to microvm driver#1143
drew wants to merge 2 commits intomainfrom
dn/vcaux-brisebo/vm-container-support

Conversation

@drew
Copy link
Copy Markdown
Collaborator

@drew drew commented May 4, 2026

Summary

Related Issue

Changes

Testing

  • mise run pre-commit passes
  • Unit tests added/updated
  • E2E tests added/updated (if applicable)

Checklist

  • Follows Conventional Commits
  • Commits are signed off (DCO)
  • Architecture docs updated (if applicable)

@drew drew requested a review from a team as a code owner May 4, 2026 04:36
@drew drew force-pushed the vm-container-support branch from 1de95c9 to 13bae67 Compare May 4, 2026 05:59
Base automatically changed from vm-container-support to main May 4, 2026 06:23
@drew drew force-pushed the dn/vcaux-brisebo/vm-container-support branch from ea7a545 to 5943739 Compare May 4, 2026 06:27
# TODO(gpu): Pin SHA-256 checksum for reproducible builds. Compute with:
# curl -fsSL <url> | sha256sum
RUN curl -fsSL \
"https://us.download.nvidia.com/XFree86/Linux-x86_64/${NVIDIA_DRIVER_VERSION}/NVIDIA-Linux-x86_64-${NVIDIA_DRIVER_VERSION}.run" \
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems brittle since the user mode driver version needs to match the kernel mode driver (in the VM) exactly. Why is the user mode driver not installed in the VM itself?

/// image like `ubuntu:latest` instead of the GPU sandbox Dockerfile.
fn warn_missing_gpu_userspace(rootfs: &Path) {
let nvidia_smi_candidates = [
"usr/bin/nvidia-smi",
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

On some systems nvidia-smi is installed at /usr/sbin/nvidia-smi.

Comment on lines +63 to +65
# Must match NVIDIA_DRIVER_VERSION in sandboxes/nvidia-gpu/versions.env
# and sandboxes/nvidia-gpu/Dockerfile ARG NVIDIA_DRIVER_VERSION
NVIDIA_OPEN_VERSION="${NVIDIA_OPEN_VERSION:-580.159.03}"
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If it must match this version, why not just source the env file?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants