gpu-manager: enable kernel module configuration via KernelModuleConfig

vgpu-manager/rhel8/nvidia-driver

@@ -7,6 +7,7 @@ DRIVER_VERSION=${DRIVER_VERSION:?"Missing driver version"}
 DRIVER_RESET_RETRIES=10
 DELAY_BEFORE_VF_CREATION=${DELAY_BEFORE_VF_CREATION:-15}
 RUN_DIR=/run/nvidia
+NVIDIA_MODULE_PARAMS=()
 
 # Mount the driver rootfs into the run directory with the exception of sysfs.
 _mount_rootfs() {
@@ -52,14 +53,52 @@ _set_fw_search_path() {
     echo -n "$nv_fw_search_path" > $fw_path_config_file
 }
 
+# For each kernel module configuration file mounted into the container,
+# parse the file contents and extract the custom module parameters that
+# are to be passed as input to 'modprobe'.
+#
+# Assumptions:
+# - Configuration file is named nvidia.conf
+# - Configuration file is mounted inside the container at /drivers.
+# - Each line in the file contains at least one parameter, where parameters on the same line
+#   are space delimited. It is up to the user to properly format the file to ensure
+#   the correct set of parameters are passed to 'modprobe'.
+_get_module_params() {
+    local base_path="/drivers"
+    # nvidia
+    if [ -f "${base_path}/nvidia.conf" ]; then
+       while IFS="" read -r param || [ -n "$param" ]; do
+           NVIDIA_MODULE_PARAMS+=("$param")
+       done <"${base_path}/nvidia.conf"
+       echo "Module parameters provided for nvidia: ${NVIDIA_MODULE_PARAMS[@]}"
+    fi
+}
+
 _install_driver() {
     local tmp_dir=$(mktemp -d)
 
     sh NVIDIA-Linux-${DRIVER_ARCH}-${DRIVER_VERSION}-vgpu-kvm.run --ui=none --no-questions --tmpdir ${tmp_dir} --no-systemd
 }
 
-# Currently _install_driver() takes care of loading nvidia modules. Just need to start necessary vgpu daemons
+# Load NVIDIA driver kernel modules with custom parameters and start vGPU daemons
 _load_driver() {
+    # Unload modules if they're already loaded so we can reload with custom parameters
+    if [ -f /sys/module/nvidia_vgpu_vfio/refcnt ] || [ -f /sys/module/nvidia/refcnt ]; then
+        echo "NVIDIA modules already loaded by installer, unloading to apply custom parameters..."
+        rmmod nvidia_vgpu_vfio 2>/dev/null || true
+        rmmod nvidia 2>/dev/null || true
+    fi
+
+    echo "Parsing kernel module parameters..."
+    _get_module_params
+
+    echo "Loading NVIDIA driver kernel modules..."
+    set -o xtrace +o nounset
+    modprobe nvidia "${NVIDIA_MODULE_PARAMS[@]}"
+    modprobe nvidia_vgpu_vfio
+    set +o xtrace -o nounset
+
+    # Start vGPU daemons
     /usr/bin/nvidia-vgpud
     /usr/bin/nvidia-vgpu-mgr &
 

vgpu-manager/rhel9/nvidia-driver

@@ -19,6 +19,7 @@ DRIVER_VERSION=${DRIVER_VERSION:?"Missing driver version"}
 DRIVER_RESET_RETRIES=10
 DELAY_BEFORE_VF_CREATION=${DELAY_BEFORE_VF_CREATION:-15}
 RUN_DIR=/run/nvidia
+NVIDIA_MODULE_PARAMS=()
 
 # Mount the driver rootfs into the run directory with the exception of sysfs.
 _mount_rootfs() {
@@ -64,14 +65,52 @@ _set_fw_search_path() {
     echo -n "$nv_fw_search_path" > $fw_path_config_file
 }
 
+# For each kernel module configuration file mounted into the container,
+# parse the file contents and extract the custom module parameters that
+# are to be passed as input to 'modprobe'.
+#
+# Assumptions:
+# - Configuration file is named nvidia.conf
+# - Configuration file is mounted inside the container at /drivers.
+# - Each line in the file contains at least one parameter, where parameters on the same line
+#   are space delimited. It is up to the user to properly format the file to ensure
+#   the correct set of parameters are passed to 'modprobe'.
+_get_module_params() {
+    local base_path="/drivers"
+    # nvidia
+    if [ -f "${base_path}/nvidia.conf" ]; then
+       while IFS="" read -r param || [ -n "$param" ]; do
+           NVIDIA_MODULE_PARAMS+=("$param")
+       done <"${base_path}/nvidia.conf"
+       echo "Module parameters provided for nvidia: ${NVIDIA_MODULE_PARAMS[@]}"
+    fi
+}
+
 _install_driver() {
     local tmp_dir=$(mktemp -d)
 
     sh NVIDIA-Linux-${DRIVER_ARCH}-${DRIVER_VERSION}-vgpu-kvm.run --ui=none --no-questions --tmpdir ${tmp_dir} --no-systemd
 }
 
-# Currently _install_driver() takes care of loading nvidia modules. Just need to start necessary vgpu daemons
+# Load NVIDIA driver kernel modules with custom parameters and start vGPU daemons
 _load_driver() {
+    # Unload modules if they're already loaded so we can reload with custom parameters
+    if [ -f /sys/module/nvidia_vgpu_vfio/refcnt ] || [ -f /sys/module/nvidia/refcnt ]; then
+        echo "NVIDIA modules already loaded by installer, unloading to apply custom parameters..."
+        rmmod nvidia_vgpu_vfio 2>/dev/null || true
+        rmmod nvidia 2>/dev/null || true
+    fi
+
+    echo "Parsing kernel module parameters..."
+    _get_module_params
+
+    echo "Loading NVIDIA driver kernel modules..."
+    set -o xtrace +o nounset
+    modprobe nvidia "${NVIDIA_MODULE_PARAMS[@]}"
+    modprobe nvidia_vgpu_vfio
+    set +o xtrace -o nounset
+
+    # Start vGPU daemons
     /usr/bin/nvidia-vgpud
     /usr/bin/nvidia-vgpu-mgr &
 

vgpu-manager/ubuntu22.04/nvidia-driver

@@ -8,6 +8,7 @@ DRIVER_RESET_RETRIES=10
 DELAY_BEFORE_VF_CREATION=${DELAY_BEFORE_VF_CREATION:-15}
 KERNEL_VERSION=$(uname -r)
 RUN_DIR=/run/nvidia
+NVIDIA_MODULE_PARAMS=()
 
 export DEBIAN_FRONTEND=noninteractive
 
@@ -133,8 +134,46 @@ _set_fw_search_path() {
     echo -n "$nv_fw_search_path" > $fw_path_config_file
 }
 
-# Currently _install_driver() takes care of loading nvidia modules. Just need to start necessary vgpu daemons
+# For each kernel module configuration file mounted into the container,
+# parse the file contents and extract the custom module parameters that
+# are to be passed as input to 'modprobe'.
+#
+# Assumptions:
+# - Configuration file is named nvidia.conf
+# - Configuration file is mounted inside the container at /drivers.
+# - Each line in the file contains at least one parameter, where parameters on the same line
+#   are space delimited. It is up to the user to properly format the file to ensure
+#   the correct set of parameters are passed to 'modprobe'.
+_get_module_params() {
+    local base_path="/drivers"
+    # nvidia
+    if [ -f "${base_path}/nvidia.conf" ]; then
+       while IFS="" read -r param || [ -n "$param" ]; do
+           NVIDIA_MODULE_PARAMS+=("$param")
+       done <"${base_path}/nvidia.conf"
+       echo "Module parameters provided for nvidia: ${NVIDIA_MODULE_PARAMS[@]}"
+    fi
+}
+
+# Load NVIDIA driver kernel modules with custom parameters and start vGPU daemons
 _load_driver() {
+    # Unload modules if they're already loaded so we can reload with custom parameters
+    if [ -f /sys/module/nvidia_vgpu_vfio/refcnt ] || [ -f /sys/module/nvidia/refcnt ]; then
+        echo "NVIDIA modules already loaded by installer, unloading to apply custom parameters..."
+        rmmod nvidia_vgpu_vfio 2>/dev/null || true
+        rmmod nvidia 2>/dev/null || true
+    fi
+
+    echo "Parsing kernel module parameters..."
+    _get_module_params
+
+    echo "Loading NVIDIA driver kernel modules..."
+    set -o xtrace +o nounset
+    modprobe nvidia "${NVIDIA_MODULE_PARAMS[@]}"
+    modprobe nvidia_vgpu_vfio
+    set +o xtrace -o nounset
+
+    # Start vGPU daemons
     /usr/bin/nvidia-vgpud
     /usr/bin/nvidia-vgpu-mgr &
 

vgpu-manager/ubuntu24.04/nvidia-driver

@@ -8,6 +8,7 @@ DRIVER_RESET_RETRIES=10
 DELAY_BEFORE_VF_CREATION=${DELAY_BEFORE_VF_CREATION:-15}
 KERNEL_VERSION=$(uname -r)
 RUN_DIR=/run/nvidia
+NVIDIA_MODULE_PARAMS=()
 
 export DEBIAN_FRONTEND=noninteractive
 
@@ -133,8 +134,46 @@ _install_driver() {
     sh NVIDIA-Linux-${DRIVER_ARCH}-${DRIVER_VERSION}-vgpu-kvm.run --ui=none --no-questions --tmpdir ${tmp_dir} --no-systemd
 }
 
-# Currently _install_driver() takes care of loading nvidia modules. Just need to start necessary vgpu daemons
+# For each kernel module configuration file mounted into the container,
+# parse the file contents and extract the custom module parameters that
+# are to be passed as input to 'modprobe'.
+#
+# Assumptions:
+# - Configuration file is named nvidia.conf
+# - Configuration file is mounted inside the container at /drivers.
+# - Each line in the file contains at least one parameter, where parameters on the same line
+#   are space delimited. It is up to the user to properly format the file to ensure
+#   the correct set of parameters are passed to 'modprobe'.
+_get_module_params() {
+    local base_path="/drivers"
+    # nvidia
+    if [ -f "${base_path}/nvidia.conf" ]; then
+       while IFS="" read -r param || [ -n "$param" ]; do
+           NVIDIA_MODULE_PARAMS+=("$param")
+       done <"${base_path}/nvidia.conf"
+       echo "Module parameters provided for nvidia: ${NVIDIA_MODULE_PARAMS[@]}"
+    fi
+}
+
+# Load NVIDIA driver kernel modules with custom parameters and start vGPU daemons
 _load_driver() {
+    # Unload modules if they're already loaded so we can reload with custom parameters
+    if [ -f /sys/module/nvidia_vgpu_vfio/refcnt ] || [ -f /sys/module/nvidia/refcnt ]; then
+        echo "NVIDIA modules already loaded by installer, unloading to apply custom parameters..."
+        rmmod nvidia_vgpu_vfio 2>/dev/null || true
+        rmmod nvidia 2>/dev/null || true
+    fi
+
+    echo "Parsing kernel module parameters..."
+    _get_module_params
+
+    echo "Loading NVIDIA driver kernel modules..."
+    set -o xtrace +o nounset
+    modprobe nvidia "${NVIDIA_MODULE_PARAMS[@]}"
+    modprobe nvidia_vgpu_vfio
+    set +o xtrace -o nounset
+
+    # Start vGPU daemons
     /usr/bin/nvidia-vgpud
     /usr/bin/nvidia-vgpu-mgr &