Skip to content

Updates to vGPU Manager build#610

Open
mattwittwer wants to merge 1 commit intoNVIDIA:mainfrom
mattwittwer:mwittwer/vgpu-manager-rhel9.6-build-and-pciutils-for-airgap
Open

Updates to vGPU Manager build#610
mattwittwer wants to merge 1 commit intoNVIDIA:mainfrom
mattwittwer:mwittwer/vgpu-manager-rhel9.6-build-and-pciutils-for-airgap

Conversation

@mattwittwer
Copy link

Fix vGPU Manager build for OCP 4.19+ and pciutils availability in air-gapped environments

The documented OS_TAG=rhel9.6 for OpenShift 4.19+ fails because the Makefile SUBDIR overrides only match the "rhcos" prefix, not "rhel9". Add SUBDIR mappings for rhel9-prefixed targets across driver, vgpu-guest, and vgpu-host build targets.

The ocp_dtk_entrypoint script copies the lspci binary to the shared directory for the DTK container, but does not copy its library dependency (libpci.so) or the setpci binary also required by sriov-manage. In air-gapped environments where the dnf fallback cannot reach cdn-ubi.redhat.com, this causes VF creation to fail silently, preventing vGPU mediated devices from being created.

Copy libpci.so* and setpci alongside lspci into the shared directory, and export LD_LIBRARY_PATH in the DTK container so the binaries can find their libraries.

@copy-pr-bot
Copy link

copy-pr-bot bot commented Feb 13, 2026

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

# Copy pciutils library dependencies for the DTK container
# (the DTK image does not have libpci, and dnf fallback fails in air-gapped networks)
mkdir -p "$DRIVER_TOOLKIT_SHARED_DIR/lib64"
cp /usr/lib64/libpci.so* "$DRIVER_TOOLKIT_SHARED_DIR/lib64/"
Copy link
Contributor

Choose a reason for hiding this comment

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

Question -- do we know if libpci.so will always be installed at /usr/lib64?

Copy link
Author

Choose a reason for hiding this comment

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

Specifically for RHEL9 this is the standard library path and should be the same. But there may be a better approach to find this dynamically for compatibility against other OS versions or architectures.

This change just allowed us to make progress in accessing libpci within the vgpu-manager image at runtime.

@mvalsecchi-nv
Copy link
Contributor

mvalsecchi-nv commented Feb 16, 2026

This seems to supersede #566 . Thanks @mattwittwer !

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.

3 participants