hooks/nvidia

   1 #! /bin/bash
   2 #
   3 # Copyright (c) 2017, 2018 NVIDIA CORPORATION.
   4 #
   5 # This library is free software; you can redistribute it and/or
   6 # modify it under the terms of the GNU Lesser General Public
   7 # License as published by the Free Software Foundation; either
   8 # version 2.1 of the License, or (at your option) any later version.
   9 #
  10 # This library is distributed in the hope that it will be useful,
  11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13 # Lesser General Public License for more details.
  14 #
  15 # You should have received a copy of the GNU Lesser General Public
  16 # License along with this library; if not, write to the Free Software
  17 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  18
  19 set -eu
  20
  21 # NVIDIA_VISIBLE_DEVICES="" *or* NVIDIA_VISIBLE_DEVICES="void"
  22 # GPU support was explicitly disabled, exit early.
  23 if [ -z "${NVIDIA_VISIBLE_DEVICES-x}" ] || [ "${NVIDIA_VISIBLE_DEVICES:-}" = "void" ]; then
  24     exit 0
  25 fi
  26
  27 # https://github.com/nvidia/nvidia-container-runtime#cuda_version
  28 if [ -n "${CUDA_VERSION:-}" ] && [ -z "${NVIDIA_REQUIRE_CUDA:-}" ]; then
  29     # Legacy CUDA image: default to all devices and all driver capabilities.
  30     if [ -z "${NVIDIA_VISIBLE_DEVICES+x}" ]; then
  31         NVIDIA_VISIBLE_DEVICES="all"
  32     fi
  33     if [ -z "${NVIDIA_DRIVER_CAPABILITIES:-}" ]; then
  34         NVIDIA_DRIVER_CAPABILITIES="all"
  35     fi
  36     if [[ "${CUDA_VERSION}" =~ ^[0-9]+\.[0-9]+ ]]; then
  37         NVIDIA_REQUIRE_CUDA="cuda>=${BASH_REMATCH[0]}"
  38     fi
  39 else
  40     # NVIDIA_VISIBLE_DEVICES unset and it's not a legacy CUDA image.
  41     # This is not a GPU image, exit early.
  42     if [ -z "${NVIDIA_VISIBLE_DEVICES+x}" ]; then
  43         exit 0
  44     fi
  45 fi
  46
  47 export PATH=$PATH:/usr/sbin:/usr/bin:/sbin:/bin
  48 if ! which nvidia-container-cli >/dev/null; then
  49     echo "ERROR: Missing tool nvidia-container-cli, see https://github.com/NVIDIA/libnvidia-container" >&2
  50     exit 1
  51 fi
  52
  53 in_userns() {
  54     [ -e /proc/self/uid_map ] || { echo no; return; }
  55     while read line; do
  56         fields=$(echo $line | awk '{ print $1 " " $2 " " $3 }')
  57         [ "$fields" = "0 0 4294967295" ] && { echo no; return; } || true
  58         echo $fields | grep -q " 0 1$" && { echo userns-root; return; } || true
  59     done < /proc/self/uid_map
  60
  61     if [ -e /proc/1/uid_map ]; then
  62         if [ "$(cat /proc/self/uid_map)" = "$(cat /proc/1/uid_map)" ]; then
  63             echo userns-root
  64             return
  65         fi
  66     fi
  67     echo yes
  68 }
  69
  70 get_ldconfig() {
  71     which "ldconfig.real" || which "ldconfig"
  72     return $?
  73 }
  74
  75 capability_to_cli() {
  76     case "$1" in
  77         compute)  echo "--compute";;
  78         compat32) echo "--compat32";;
  79         display)  echo "--display";;
  80         graphics) echo "--graphics";;
  81         utility)  echo "--utility";;
  82         video)    echo "--video";;
  83         *)        exit 1;;
  84     esac
  85     return
  86 }
  87
  88 # Same behavior as strconv.ParseBool in golang
  89 parse_bool() {
  90     case "$1" in
  91         1|t|T|TRUE|true|True)    echo "true";;
  92         0|f|F|FALSE|false|False) echo "false";;
  93         *)                       exit 1;;
  94     esac
  95     return
  96 }
  97
  98 usage() {
  99     cat <<EOF
 100 nvidia-container-cli hook for LXC
 101
 102 Special arguments:
 103 [ -h | --help ]: Print this help message and exit.
 104
 105 Optional arguments:
 106 [ --no-load-kmods ]: Do not try to load the NVIDIA kernel modules.
 107 [ --disable-require ]: Disable all the constraints of the form NVIDIA_REQUIRE_*.
 108 [ --debug <path> ]: The path to the log file.
 109 [ --ldcache <path> ]: The path to the host system's DSO cache.
 110 [ --root <path> ]: The path to the driver root directory.
 111 [ --ldconfig <path> ]: The path to the ldconfig binary, use a '@' prefix for a host path.
 112 EOF
 113     return 0
 114 }
 115
 116 options=$(getopt -o h -l help,no-load-kmods,disable-require,debug:,ldcache:,root:,ldconfig: -- "$@")
 117 if [ $? -ne 0 ]; then
 118     usage
 119     exit 1
 120 fi
 121 eval set -- "$options"
 122
 123 CLI_LOAD_KMODS="true"
 124 CLI_DISABLE_REQUIRE="false"
 125 CLI_DEBUG=
 126 CLI_LDCACHE=
 127 CLI_ROOT=
 128 CLI_LDCONFIG=
 129
 130 while :; do
 131     case "$1" in
 132         --help)             usage && exit 1;;
 133         --no-load-kmods)    CLI_LOAD_KMODS="false"; shift 1;;
 134         --disable-require)  CLI_DISABLE_REQUIRE="true"; shift 1;;
 135         --debug)            CLI_DEBUG=$2; shift 2;;
 136         --ldcache)          CLI_LDCACHE=$2; shift 2;;
 137         --root)             CLI_ROOT=$2; shift 2;;
 138         --ldconfig)         CLI_LDCONFIG=$2; shift 2;;
 139         --)                 shift 1; break;;
 140         *)                  break;;
 141     esac
 142 done
 143
 144 HOOK_SECTION=
 145 HOOK_TYPE=
 146 case "${LXC_HOOK_VERSION:-0}" in
 147     0) HOOK_SECTION="${2:-}"; HOOK_TYPE="${3:-}";;
 148     1) HOOK_SECTION="${LXC_HOOK_SECTION:-}"; HOOK_TYPE="${LXC_HOOK_TYPE:-}";;
 149     *) echo "ERROR: Unsupported hook version: ${LXC_HOOK_VERSION}." >&2; exit 1;;
 150 esac
 151
 152 if [ "${HOOK_SECTION}" != "lxc" ]; then
 153     echo "ERROR: Not running through LXC." >&2
 154     exit 1
 155 fi
 156
 157 if [ "${HOOK_TYPE}" != "mount" ]; then
 158     echo "ERROR: This hook must be used as a \"mount\" hook." >&2
 159     exit 1
 160 fi
 161
 162 USERNS=$(in_userns)
 163 if [ "${USERNS}" != "yes" ]; then
 164     # This is a limitation of libnvidia-container.
 165     echo "FIXME: This hook currently only works in unprivileged mode." >&2
 166     exit 1
 167 fi
 168
 169 if [ "${USERNS}" = "yes" ]; then
 170     CLI_LOAD_KMODS="false"
 171     if ! grep -q nvidia_uvm /proc/modules; then
 172         echo "WARN: Kernel module nvidia_uvm is not loaded, nvidia-container-cli might fail. Make sure the NVIDIA device driver is installed and loaded." >&2
 173     fi
 174 fi
 175
 176 # https://github.com/nvidia/nvidia-container-runtime#nvidia_disable_require
 177 if [ -n "${NVIDIA_DISABLE_REQUIRE:-}" ]; then
 178     if [ "$(parse_bool "${NVIDIA_DISABLE_REQUIRE}")" = "true" ]; then
 179         CLI_DISABLE_REQUIRE="true"
 180     fi
 181 fi
 182
 183 if [ -z "${CLI_DEBUG}" ]; then
 184     if [ "${LXC_LOG_LEVEL}" = "DEBUG" ] || [ "${LXC_LOG_LEVEL}" = "TRACE" ]; then
 185         rootfs_path="${LXC_ROOTFS_PATH#*:}"
 186         hookdir="${rootfs_path/%rootfs/hook}"
 187         if mkdir -p "${hookdir}"; then
 188             CLI_DEBUG="${hookdir}/nvidia.log"
 189         fi
 190     fi
 191 fi
 192
 193 # A '@' prefix means a host path.
 194 if [ -z "${CLI_LDCONFIG}" ]; then
 195     if host_ldconfig=$(get_ldconfig); then
 196         CLI_LDCONFIG="@${host_ldconfig}"
 197     fi
 198 fi
 199
 200 # https://github.com/nvidia/nvidia-container-runtime#nvidia_visible_devices
 201 CLI_DEVICES="${NVIDIA_VISIBLE_DEVICES}"
 202
 203 # https://github.com/nvidia/nvidia-container-runtime#nvidia_driver_capabilities
 204 CLI_CAPABILITIES=
 205 if [ -n "${NVIDIA_DRIVER_CAPABILITIES:-}" ]; then
 206      CLI_CAPABILITIES="${NVIDIA_DRIVER_CAPABILITIES//,/ }"
 207 fi
 208
 209 if [ "${CLI_CAPABILITIES}" = "all" ]; then
 210     CLI_CAPABILITIES="compute compat32 display graphics utility video"
 211 fi
 212
 213 if [ -z "${CLI_CAPABILITIES}" ]; then
 214     CLI_CAPABILITIES="utility"
 215 fi
 216
 217 global_args=()
 218 configure_args=()
 219
 220 if [ -n "${CLI_DEBUG}" ]; then
 221     echo "INFO: Writing nvidia-container-cli log at ${CLI_DEBUG}." >&2
 222     global_args+=("--debug=${CLI_DEBUG}")
 223 fi
 224
 225 if [ "${CLI_LOAD_KMODS}" = "true" ]; then
 226     global_args+=(--load-kmods)
 227 fi
 228
 229 if [ "${USERNS}" = "yes" ]; then
 230     global_args+=(--user)
 231     configure_args+=(--no-cgroups)
 232 fi
 233
 234 if [ -n "${CLI_LDCACHE}" ]; then
 235     global_args+=(--ldcache="${CLI_LDCACHE}")
 236 fi
 237
 238 if [ -n "${CLI_ROOT}" ]; then
 239     global_args+=(--root="${CLI_ROOT}")
 240 fi
 241
 242 if [ -n "${CLI_LDCONFIG}" ]; then
 243     configure_args+=(--ldconfig="${CLI_LDCONFIG}")
 244 fi
 245
 246 if [ -n "${CLI_DEVICES}" ] && [ "${CLI_DEVICES}" != "none" ]; then
 247     configure_args+=(--device="${CLI_DEVICES}")
 248 fi
 249
 250 for cap in ${CLI_CAPABILITIES}; do
 251     if arg=$(capability_to_cli "${cap}"); then
 252         configure_args+=("${arg}")
 253     else
 254         echo "ERROR: Unknown driver capability \"${cap}\"." >&2
 255         exit 1
 256     fi
 257 done
 258
 259 # https://github.com/nvidia/nvidia-container-runtime#nvidia_require_
 260 if [ "${CLI_DISABLE_REQUIRE}" = "false" ]; then
 261     for req in $(compgen -e "NVIDIA_REQUIRE_"); do
 262         configure_args+=("--require=${!req}")
 263     done
 264 fi
 265
 266 if [ -d "/sys/kernel/security/apparmor" ]; then
 267     # Try to transition to the unconfined AppArmor profile.
 268     echo "changeprofile unconfined" > /proc/self/attr/current || true
 269 fi
 270
 271 set -x
 272 exec nvidia-container-cli ${global_args[@]} configure "${configure_args[@]}" "${LXC_ROOTFS_MOUNT}"