Support Data Center precompiled driver container for Arm (Ubuntu 24.04)

.github/workflows/image.yaml

@@ -198,5 +198,11 @@ jobs:
           PRECOMPILED: "true"
           DIST: signed_${{ matrix.dist }}
         run: |
-          source kernel_version.txt && \
+          source kernel_version.txt
+          # arm64 does not support azure-fde (package linux-objects-nvidia-*-azure-fde not available for arm64)
+          if [[ "${{ matrix.dist }}" == "ubuntu24.04" ]] && [[ "${{ matrix.flavor }}" != "azure-fde" ]]; then
+            export DOCKER_BUILD_PLATFORM_OPTIONS="--platform=linux/amd64,linux/arm64"
+          else
+            export DOCKER_BUILD_PLATFORM_OPTIONS="--platform=linux/amd64"
+          fi
           make DRIVER_VERSIONS=${DRIVER_VERSIONS} DRIVER_BRANCH=${{ matrix.driver }} build-${DIST}-${DRIVER_VERSION}

Makefile

@@ -160,12 +160,13 @@ build-%: DOCKERFILE = $(CURDIR)/$(SUBDIR)/Dockerfile
 # build-ubuntu20.04-$(DRIVER_VERSION) triggers a build for a specific $(DRIVER_VERSION)
 $(DISTRIBUTIONS): %: build-%
 $(BUILD_TARGETS): %: $(foreach driver_version, $(DRIVER_VERSIONS), $(addprefix %-, $(driver_version)))
+DRIVER_BUILD_TAG = $(if $(findstring type=oci,$(DOCKER_BUILD_OPTIONS)),,--tag $(IMAGE))
 $(DRIVER_BUILD_TARGETS):
 	DOCKER_BUILDKIT=1 \
 		$(DOCKER) $(BUILDX) build --pull \
 				$(DOCKER_BUILD_OPTIONS) \
 				$(DOCKER_BUILD_PLATFORM_OPTIONS) \
-				--tag $(IMAGE) \
+				$(DRIVER_BUILD_TAG) \
 				--build-arg DRIVER_VERSION="$(DRIVER_VERSION)" \
 				--build-arg GOLANG_VERSION="$(GOLANG_VERSION)" \
 				--build-arg DRIVER_BRANCH="$(DRIVER_BRANCH)" \
@@ -215,6 +216,7 @@ $(BASE_BUILD_TARGETS):
 	DOCKER_BUILDKIT=1 \
 		$(DOCKER) $(BUILDX) build --pull --no-cache \
 				$(DOCKER_BUILD_OPTIONS) \
+				$(DOCKER_BUILD_PLATFORM_OPTIONS) \
 				--tag $(IMAGE)  \
 				--target $(TARGET) \
 				--build-arg CUDA_VERSION="$(CUDA_VERSION)" \

multi-arch.mk

@@ -27,5 +27,4 @@ $(DRIVER_PUSH_TARGETS): push-%:
 build-ubuntu18.04%: DOCKER_BUILD_PLATFORM_OPTIONS = --platform=linux/amd64
 build-signed_ubuntu20.04%: DOCKER_BUILD_PLATFORM_OPTIONS = --platform=linux/amd64
 build-signed_ubuntu22.04%: DOCKER_BUILD_PLATFORM_OPTIONS = --platform=linux/amd64
-build-signed_ubuntu24.04%: DOCKER_BUILD_PLATFORM_OPTIONS = --platform=linux/amd64
 build-sles%: DOCKER_BUILD_PLATFORM_OPTIONS = --platform=linux/amd64

tests/holodeck_ubuntu22.04.yaml → tests/holodeck_ubuntu.yaml

@@ -11,19 +11,9 @@ spec:
   instance:
     type: g4dn.xlarge
     region: us-west-1
-    ingressIpRanges:
-    - 18.190.12.32/32
-    - 3.143.46.93/32
-    - 52.15.119.136/32
-    - 35.155.108.162/32
-    - 35.162.190.51/32
-    - 54.201.61.24/32
-    - 52.24.205.48/32
-    - 44.235.4.62/32
-    - 44.230.241.223/32
+    os: ubuntu-22.04
     image:
       architecture: amd64
-      imageId: ami-0007a86be89339c9f
   containerRuntime:
     install: true
     name: containerd

tests/scripts/ci-precompiled-helpers.sh

@@ -1,18 +1,18 @@
 get_kernel_versions_to_test() {
-    if [[ "$#" -ne 4 ]]; then
-	    echo " Error:$0 must be called with KERNEL_FLAVORS DRIVER_BRANCHES DIST LTS_KERNEL" >&2
+    if [[ "$#" -lt 4 || "$#" -gt 5 ]]; then
+	    echo " Error:$0 must be called with KERNEL_FLAVORS DRIVER_BRANCHES DIST LTS_KERNEL or KERNEL_FLAVORS DRIVER_BRANCHES DIST LTS_KERNEL PLATFORM_SUFFIX" >&2
 	    exit 1
     fi
 
     local -a KERNEL_FLAVORS=("${!1}")
     local -a DRIVER_BRANCHES=("${!2}")
     local DIST="$3"
     local LTS_KERNEL="$4"
-
+    local PLATFORM_SUFFIX="$5"
     kernel_versions=()
     for kernel_flavor in "${KERNEL_FLAVORS[@]}"; do
         for DRIVER_BRANCH in "${DRIVER_BRANCHES[@]}"; do
-            source ./tests/scripts/findkernelversion.sh "${kernel_flavor}" "$DRIVER_BRANCH" "$DIST" "$LTS_KERNEL" >&2
+            source ./tests/scripts/findkernelversion.sh "${kernel_flavor}" "$DRIVER_BRANCH" "$DIST" "$LTS_KERNEL" "$PLATFORM_SUFFIX" >&2
             if [[ "$should_continue" == true ]]; then
                 break
             fi
@@ -25,7 +25,7 @@ get_kernel_versions_to_test() {
     # Remove duplicates
     kernel_versions=($(printf "%s\n" "${kernel_versions[@]}" | sort -u))
     for i in "${!kernel_versions[@]}"; do
-        kernel_versions[$i]="${kernel_versions[$i]}-$DIST"
+        kernel_versions[$i]="${kernel_versions[$i]}-$DIST$PLATFORM_SUFFIX"
     done
     echo "${kernel_versions[@]}"
 }

tests/scripts/findkernelversion.sh

@@ -1,14 +1,15 @@
 #!/bin/bash
 
-if [[ $# -ne 4 ]]; then
-	echo " KERNEL_FLAVOR DRIVER_BRANCH DIST LTS_KERNEL are required"
+if [[ $# -lt 4 || $# -gt 5 ]]; then
+	echo " KERNEL_FLAVOR DRIVER_BRANCH DIST LTS_KERNEL or KERNEL_FLAVOR DRIVER_BRANCH DIST LTS_KERNEL PLATFORM_SUFFIX are required"
 	exit 1
 fi
 
 export KERNEL_FLAVOR="${1}"
 export DRIVER_BRANCH="${2}"
 export DIST="${3}"
 export LTS_KERNEL="${4}"
+export PLATFORM_SUFFIX="${5}"
 
 export REGCTL_VERSION=v0.7.1
 mkdir -p bin
@@ -21,21 +22,21 @@ prefix="kernel-version-${DRIVER_BRANCH}-${LTS_KERNEL}"
 suffix="${KERNEL_FLAVOR}-${DIST}"
 
 artifact_dir="./kernel-version-artifacts"
-artifact=$(find "$artifact_dir" -maxdepth 1 -type d -name "${prefix}*-${suffix}" | head -1)
-if [ -n "$artifact" ]; then
-    artifact_name=$(basename "$artifact")
-    if [ -f "$artifact/${artifact_name}.tar" ]; then
-        tar -xf "$artifact/${artifact_name}.tar" -C ./
-        export $(grep -oP 'KERNEL_VERSION=[^ ]+' ./kernel_version.txt)
-        rm -f kernel_version.txt
-    fi
+artifact_file=$(find "$artifact_dir" -maxdepth 1 -type f -name "${prefix}*-${suffix}.tar" | head -1)
+if [ -n "$artifact_file" ]; then
+    tar -xf "$artifact_file" -C ./
+    export $(grep -oE 'KERNEL_VERSION=[^ ]+' ./kernel_version.txt)
+    rm -f kernel_version.txt
 fi
 
 # calculate driver tag
 status_nvcr=0
 status_ghcr=0
-regctl tag ls  nvcr.io/nvidia/driver | grep "^${DRIVER_BRANCH}-${KERNEL_VERSION}-${DIST}$" || status_nvcr=$?
-regctl tag ls  ghcr.io/nvidia/driver | grep "^${DRIVER_BRANCH}-${KERNEL_VERSION}-${DIST}$" || status_ghcr=$?
+PLATFORM=$(echo "${PLATFORM_SUFFIX}" | sed 's/-//')
+[ -z "$PLATFORM" ] && PLATFORM=amd64
+regctl manifest inspect nvcr.io/nvidia/driver:${DRIVER_BRANCH}-${KERNEL_VERSION}-${DIST} --platform=linux/${PLATFORM} > /dev/null 2>&1; status_nvcr=$?
+regctl manifest inspect ghcr.io/nvidia/driver:${DRIVER_BRANCH}-${KERNEL_VERSION}-${DIST} --platform=linux/${PLATFORM} > /dev/null 2>&1; status_ghcr=$?
+
 if [[ $status_nvcr -eq 0 || $status_ghcr -eq 0 ]]; then
     export should_continue=false
 else

ubuntu24.04/precompiled/Dockerfile

@@ -17,7 +17,8 @@ ENV NVIDIA_VISIBLE_DEVICES=void
 
 RUN echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections
 
-RUN dpkg --add-architecture i386 && \
+# Add i386 architecture only for amd64 builds (not available on ARM)
+RUN if [ "$TARGETARCH" = "amd64" ]; then dpkg --add-architecture i386; fi && \
     apt-get update && apt-get install -y --no-install-recommends \
         apt-utils \
         build-essential \
@@ -31,11 +32,12 @@ RUN dpkg --add-architecture i386 && \
         pkg-config && \
     rm -rf /var/lib/apt/lists/*
 
-# Fetch GPG keys for CUDA repo
-RUN apt-key del 3bf863cc && \
+# Fetch GPG keys for CUDA repo (architecture-specific)
+RUN CUDA_ARCH=$([ "$TARGETARCH" = "arm64" ] && echo "sbsa" || echo "x86_64") && \
+    apt-key del 3bf863cc && \
     rm /etc/apt/sources.list.d/cuda.list && \
-    curl -fsSL https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/3bf863cc.pub | gpg --dearmor -o /etc/apt/keyrings/cuda.pub && \
-    echo "deb [signed-by=/etc/apt/keyrings/cuda.pub] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64 /" > /etc/apt/sources.list.d/cuda.list
+    curl -fsSL https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/${CUDA_ARCH}/3bf863cc.pub | gpg --dearmor -o /etc/apt/keyrings/cuda.pub && \
+    echo "deb [signed-by=/etc/apt/keyrings/cuda.pub] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/${CUDA_ARCH} /" > /etc/apt/sources.list.d/cuda.list
 
 RUN usermod -o -u 0 -g 0 _apt
 

ubuntu24.04/precompiled/local-repo.sh

@@ -68,15 +68,23 @@ download_driver_package_deps () {
   pushd ${LOCAL_REPO_DIR}
 
   download_apt_with_dep linux-objects-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION}
-  download_apt_with_dep linux-signatures-nvidia-${KERNEL_VERSION}
+
+  # linux-signatures-nvidia (secure boot signatures) is not available for arm64
+  if [ "$TARGETARCH" = "amd64" ]; then
+    download_apt_with_dep linux-signatures-nvidia-${KERNEL_VERSION}
+  fi
+
   download_apt_with_dep linux-modules-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION}
   download_apt_with_dep linux-modules-nvidia-${DRIVER_BRANCH}-server-open-${KERNEL_VERSION}
   download_apt_with_dep nvidia-utils-${DRIVER_BRANCH}-server
   download_apt_with_dep nvidia-headless-no-dkms-${DRIVER_BRANCH}-server
   download_apt_with_dep libnvidia-decode-${DRIVER_BRANCH}-server
   download_apt_with_dep libnvidia-extra-${DRIVER_BRANCH}-server
   download_apt_with_dep libnvidia-encode-${DRIVER_BRANCH}-server
-  download_apt_with_dep libnvidia-fbc1-${DRIVER_BRANCH}-server
+  # libnvidia-fbc1 (FrameBuffer Capture) is not available for arm64
+  if [ "$TARGETARCH" = "amd64" ]; then
+    download_apt_with_dep libnvidia-fbc1-${DRIVER_BRANCH}-server
+  fi
   download_apt_with_dep libnvidia-gl-${DRIVER_BRANCH}-server
 
   fabricmanager_download

ubuntu24.04/precompiled/nvidia-driver

@@ -416,22 +416,37 @@ _install_driver() {
         nvidia-headless-no-dkms-${DRIVER_BRANCH}-server \
         libnvidia-decode-${DRIVER_BRANCH}-server \
         libnvidia-extra-${DRIVER_BRANCH}-server \
-        libnvidia-encode-${DRIVER_BRANCH}-server \
-        libnvidia-fbc1-${DRIVER_BRANCH}-server \
-        libnvidia-gl-${DRIVER_BRANCH}-server
+        libnvidia-encode-${DRIVER_BRANCH}-server
+        # libnvidia-fbc1 (FrameBuffer Capture) is not available for arm64
+    if [ "$TARGETARCH" = "amd64" ]; then
+        apt-get install -y --no-install-recommends libnvidia-fbc1-${DRIVER_BRANCH}-server
+    fi
+    apt-get install -y --no-install-recommends libnvidia-gl-${DRIVER_BRANCH}-server
 
     # Now install the precompiled kernel module packages signed by Canonical
+    # linux-signatures-nvidia (secure boot signatures) is not available for arm64
     if [ "$KERNEL_TYPE" = "kernel-open" ]; then
         echo "Installing Open NVIDIA driver kernel modules..."
-        apt-get install --no-install-recommends -y \
-            linux-signatures-nvidia-${KERNEL_VERSION} \
-            linux-modules-nvidia-${DRIVER_BRANCH}-server-open-${KERNEL_VERSION}
+        if [ "$TARGETARCH" = "arm64" ]; then
+            apt-get install --no-install-recommends -y \
+                linux-modules-nvidia-${DRIVER_BRANCH}-server-open-${KERNEL_VERSION}
+        else
+            apt-get install --no-install-recommends -y \
+                linux-signatures-nvidia-${KERNEL_VERSION} \
+                linux-modules-nvidia-${DRIVER_BRANCH}-server-open-${KERNEL_VERSION}
+        fi
     else
         echo "Installing Closed NVIDIA driver kernel modules..."
-        apt-get install --no-install-recommends -y \
-            linux-objects-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION} \
-            linux-signatures-nvidia-${KERNEL_VERSION} \
-            linux-modules-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION}
+        if [ "$TARGETARCH" = "arm64" ]; then
+                apt-get install --no-install-recommends -y \
+                    linux-objects-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION} \
+                    linux-modules-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION}
+        else
+                apt-get install --no-install-recommends -y \
+                    linux-objects-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION} \
+                    linux-signatures-nvidia-${KERNEL_VERSION} \
+                    linux-modules-nvidia-${DRIVER_BRANCH}-server-${KERNEL_VERSION}
+        fi
     fi
 }