fix precompiled driver container failures when enabling OpenRM

Signed-off-by: Tariq Ibrahim <[email protected]>
NVIDIA · Dec 12, 2024 · 070e32e · 070e32e
1 parent 6b78622
commit 070e32e
Show file tree

Hide file tree

Showing 4 changed files with 70 additions and 12 deletions.
diff --git a/ubuntu22.04/precompiled/Dockerfile b/ubuntu22.04/precompiled/Dockerfile
@@ -49,14 +49,9 @@ RUN if [ -n "${CVE_UPDATES}" ]; then \
     fi
 
 # update pkg cache and install pkgs for userspace driver libs
-RUN apt-get update && apt-get install -y --no-install-recommends nvidia-driver-${DRIVER_BRANCH}-server \
+RUN apt-get update && apt-get install -y --download-only --no-install-recommends nvidia-driver-${DRIVER_BRANCH}-server \
     nvidia-fabricmanager-${DRIVER_BRANCH}=${DRIVER_VERSION}-1 \
     libnvidia-nscq-${DRIVER_BRANCH}=${DRIVER_VERSION}-1 && \
-    apt-get purge -y \
-    libnvidia-egl-wayland1 \
-    nvidia-dkms-${DRIVER_BRANCH}-server \
-    nvidia-kernel-source-${DRIVER_BRANCH}-server \
-    xserver-xorg-video-nvidia-${DRIVER_BRANCH}-server && \
     rm -rf /var/lib/apt/lists/*;
 
 # update pkg cache and download pkgs for driver module installation during runtime.

diff --git a/ubuntu22.04/precompiled/nvidia-driver b/ubuntu22.04/precompiled/nvidia-driver
@@ -101,6 +101,25 @@ _load_driver() {
     echo "Parsing kernel module parameters..."
     _get_module_params
 
+    local nv_fw_search_path="$RUN_DIR/driver/lib/firmware"
+    local set_fw_path="true"
+    local fw_path_config_file="/sys/module/firmware_class/parameters/path"
+    for param in "${NVIDIA_MODULE_PARAMS[@]}"; do
+        if [[ "$param" == "NVreg_EnableGpuFirmware=0" ]]; then
+          set_fw_path="false"
+        fi
+    done
+
+    if [[ "$set_fw_path" == "true" ]]; then
+        echo "Configuring the following firmware search path in '$fw_path_config_file': $nv_fw_search_path"
+        if [[ ! -z $(grep '[^[:space:]]' $fw_path_config_file) ]]; then
+            echo "WARNING: A search path is already configured in $fw_path_config_file"
+            echo "         Retaining the current configuration"
+        else
+            echo -n "$nv_fw_search_path" > $fw_path_config_file || echo "WARNING: Failed to configure firmware search path"
+        fi
+    fi
+
     echo "Loading ipmi and i2c_core kernel modules..."
     modprobe -a i2c_core ipmi_msghandler ipmi_devintf
 
@@ -124,6 +143,10 @@ _load_driver() {
 
     DRIVER_VERSION=$(nvidia-smi -q | grep "Driver Version" | awk -F: '{print $2}' | xargs)
     if _assert_nvswitch_system; then
+        echo "Installing NVIDIA fabric manager and libnvidia NSCQ packages..."
+        apt-get install -y --no-install-recommends \
+            nvidia-fabricmanager-${DRIVER_BRANCH}=${DRIVER_VERSION}-1 \
+            libnvidia-nscq-${DRIVER_BRANCH}=${DRIVER_VERSION}-1
 
         echo "Starting NVIDIA fabric manager daemon..."
         nv-fabricmanager -c /usr/share/nvidia/nvswitch/fabricmanager.cfg
@@ -211,6 +234,17 @@ _unload_driver() {
 
 # Link and install the kernel modules from a precompiled packages
 _install_driver() {
+    # Install necessary userspace, fabric manager and libnvidia-nscq packages
+    apt-get install -y --no-install-recommends nvidia-driver-${DRIVER_BRANCH}-server
+
+    # Uninstall unnecessary packages installed as a part of the nvidia-driver-${DRIVER_BRANCH}-server package
+    apt-get purge -y \
+        libnvidia-egl-wayland1 \
+        nvidia-dkms-${DRIVER_BRANCH}-server \
+        nvidia-kernel-source-${DRIVER_BRANCH}-server \
+        xserver-xorg-video-nvidia-${DRIVER_BRANCH}-server
+
+    # Now install the precompiled kernel module packages signed by Canonical
     if [ "$OPEN_KERNEL_MODULES_ENABLED" = true ]; then
         echo "Installing Open NVIDIA driver kernel modules..."
         apt-get install --no-install-recommends -y \

diff --git a/ubuntu24.04/precompiled/Dockerfile b/ubuntu24.04/precompiled/Dockerfile
@@ -41,14 +41,9 @@ RUN if [ -n "${CVE_UPDATES}" ]; then \
     fi
 
 # update pkg cache and install pkgs for userspace driver libs
-RUN apt-get update && apt-get install -y --no-install-recommends nvidia-driver-${DRIVER_BRANCH}-server \
+RUN apt-get update && apt-get install -y --download-only --no-install-recommends nvidia-driver-${DRIVER_BRANCH}-server \
     nvidia-fabricmanager-${DRIVER_BRANCH}=${DRIVER_VERSION}-1 \
     libnvidia-nscq-${DRIVER_BRANCH}=${DRIVER_VERSION}-1 && \
-    apt-get purge -y \
-    libnvidia-egl-wayland1 \
-    nvidia-dkms-${DRIVER_BRANCH}-server \
-    nvidia-kernel-source-${DRIVER_BRANCH}-server \
-    xserver-xorg-video-nvidia-${DRIVER_BRANCH}-server && \
     rm -rf /var/lib/apt/lists/*;
 
 # update pkg cache and download pkgs for driver module installation during runtime.

diff --git a/ubuntu24.04/precompiled/nvidia-driver b/ubuntu24.04/precompiled/nvidia-driver
@@ -101,6 +101,25 @@ _load_driver() {
     echo "Parsing kernel module parameters..."
     _get_module_params
 
+    local nv_fw_search_path="$RUN_DIR/driver/lib/firmware"
+    local set_fw_path="true"
+    local fw_path_config_file="/sys/module/firmware_class/parameters/path"
+    for param in "${NVIDIA_MODULE_PARAMS[@]}"; do
+        if [[ "$param" == "NVreg_EnableGpuFirmware=0" ]]; then
+          set_fw_path="false"
+        fi
+    done
+
+    if [[ "$set_fw_path" == "true" ]]; then
+        echo "Configuring the following firmware search path in '$fw_path_config_file': $nv_fw_search_path"
+        if [[ ! -z $(grep '[^[:space:]]' $fw_path_config_file) ]]; then
+            echo "WARNING: A search path is already configured in $fw_path_config_file"
+            echo "         Retaining the current configuration"
+        else
+            echo -n "$nv_fw_search_path" > $fw_path_config_file || echo "WARNING: Failed to configure firmware search path"
+        fi
+    fi
+
     echo "Loading ipmi and i2c_core kernel modules..."
     modprobe -a i2c_core ipmi_msghandler ipmi_devintf
 
@@ -124,6 +143,10 @@ _load_driver() {
 
     DRIVER_VERSION=$(nvidia-smi -q | grep "Driver Version" | awk -F: '{print $2}' | xargs)
     if _assert_nvswitch_system; then
+        echo "Installing NVIDIA fabric manager and libnvidia NSCQ packages..."
+        apt-get install -y --no-install-recommends \
+            nvidia-fabricmanager-${DRIVER_BRANCH}=${DRIVER_VERSION}-1 \
+            libnvidia-nscq-${DRIVER_BRANCH}=${DRIVER_VERSION}-1
 
         echo "Starting NVIDIA fabric manager daemon..."
         nv-fabricmanager -c /usr/share/nvidia/nvswitch/fabricmanager.cfg
@@ -211,6 +234,17 @@ _unload_driver() {
 
 # Link and install the kernel modules from a precompiled packages
 _install_driver() {
+    # Install necessary driver userspace packages
+    apt-get install -y --no-install-recommends nvidia-driver-${DRIVER_BRANCH}-server
+
+    # Uninstall unnecessary packages installed as a part of the nvidia-driver-${DRIVER_BRANCH}-server package
+    apt-get purge -y \
+        libnvidia-egl-wayland1 \
+        nvidia-dkms-${DRIVER_BRANCH}-server \
+        nvidia-kernel-source-${DRIVER_BRANCH}-server \
+        xserver-xorg-video-nvidia-${DRIVER_BRANCH}-server
+
+    # Now install the precompiled kernel module packages signed by Canonical
     if [ "$OPEN_KERNEL_MODULES_ENABLED" = true ]; then
         echo "Installing Open NVIDIA driver kernel modules..."
         apt-get install --no-install-recommends -y \