129 lines
3.8 KiB
Bash
129 lines
3.8 KiB
Bash
#!/usr/bin/env bash
|
|
|
|
# NVIDIA GPU Integration for Proxmox LXC
|
|
# Author: CanbiZ
|
|
# License: MIT
|
|
|
|
set -euo pipefail
|
|
|
|
function nvidia_exit() {
|
|
printf "⚠️ User exited script\n"
|
|
exit 0
|
|
}
|
|
|
|
function msg() {
|
|
local type="$1"
|
|
shift
|
|
case "$type" in
|
|
info) printf " \033[36m➤\033[0m %s\n" "$@" ;;
|
|
ok) printf " \033[32m✔\033[0m %s\n" "$@" ;;
|
|
warn) printf " \033[33m⚠\033[0m %s\n" "$@" >&2 ;;
|
|
err) printf " \033[31m✘\033[0m %s\n" "$@" >&2 ;;
|
|
esac
|
|
}
|
|
|
|
function nvidia_check_driver_installed() {
|
|
command -v nvidia-smi &>/dev/null
|
|
}
|
|
|
|
function nvidia_get_driver_version() {
|
|
nvidia-smi --query-gpu=driver_version --format=csv,noheader,nounits 2>/dev/null | head -n1
|
|
}
|
|
|
|
function nvidia_get_cuda_version() {
|
|
nvidia-smi --query-gpu=cuda_version --format=csv,noheader,nounits 2>/dev/null | head -n1
|
|
}
|
|
|
|
function nvidia_validate_driver_version() {
|
|
if ! nvidia_check_driver_installed; then
|
|
msg err "NVIDIA drivers not found"
|
|
nvidia_exit
|
|
fi
|
|
local ver major
|
|
ver=$(nvidia_get_driver_version)
|
|
major=${ver%%.*}
|
|
if ((major < 500)); then
|
|
msg warn "Detected old NVIDIA driver version: $ver"
|
|
read -rp "Continue anyway? [y/N]: " confirm
|
|
[[ "${confirm,,}" =~ ^(y|yes)$ ]] || nvidia_exit
|
|
fi
|
|
}
|
|
|
|
function nvidia_validate_cuda_version() {
|
|
if ! nvidia_check_driver_installed; then
|
|
msg err "NVIDIA drivers not found"
|
|
nvidia_exit
|
|
fi
|
|
local ver major
|
|
ver=$(nvidia_get_cuda_version)
|
|
major=${ver%%.*}
|
|
if ((major < 11)); then
|
|
msg warn "Detected old CUDA version: $ver"
|
|
read -rp "Continue anyway? [y/N]: " confirm
|
|
[[ "${confirm,,}" =~ ^(y|yes)$ ]] || nvidia_exit
|
|
fi
|
|
}
|
|
|
|
function nvidia_setup_kernel_modules() {
|
|
local modfile="/etc/modules-load.d/nvidia.conf"
|
|
local udevfile="/etc/udev/rules.d/70-nvidia.rules"
|
|
|
|
printf "nvidia\nnvidia_uvm\nnvidia_drm\n" >"$modfile"
|
|
|
|
cat <<EOF >"$udevfile"
|
|
KERNEL=="nvidia", RUN+="/bin/bash -c '/usr/bin/nvidia-smi -L && chmod 666 /dev/nvidia*'"
|
|
KERNEL=="nvidia_uvm", RUN+="/bin/bash -c '/usr/bin/nvidia-modprobe -c0 -u && chmod 0666 /dev/nvidia-uvm*'"
|
|
EOF
|
|
|
|
msg ok "NVIDIA modules configured"
|
|
msg warn "Reboot the host to apply kernel changes"
|
|
}
|
|
|
|
function nvidia_select_gpu_minor() {
|
|
local menu=() max=0
|
|
while IFS= read -r path; do
|
|
local dev="${path##*/}"
|
|
local info="/proc/driver/nvidia/gpus/${dev}/information"
|
|
[[ -f "$info" ]] || continue
|
|
local model minor
|
|
model=$(awk -F': ' '/Model:/ {print $2}' "$info")
|
|
minor=$(awk '/Device Minor/ {print $NF}' "$info")
|
|
menu+=("$minor" "$model" "OFF")
|
|
((${#model} > max)) && max=${#model}
|
|
done < <(find /proc/driver/nvidia/gpus -mindepth 1 -type d)
|
|
|
|
[[ ${#menu[@]} -eq 0 ]] && msg err "No NVIDIA GPU found" && return 1
|
|
[[ ${#menu[@]} -eq 3 ]] && printf "%s" "${menu[0]}" && return
|
|
|
|
whiptail --title "NVIDIA GPU Selection" --radiolist \
|
|
"Select GPU for passthrough:" 15 $((max + 40)) 6 \
|
|
"${menu[@]}" 3>&1 1>&2 2>&3
|
|
}
|
|
|
|
function nvidia_lxc_passthrough() {
|
|
local ctid="$1" minor="$2"
|
|
local conf="/etc/pve/lxc/${ctid}.conf"
|
|
local devices=(
|
|
"/dev/nvidia${minor}"
|
|
"/dev/nvidiactl"
|
|
"/dev/nvidia-uvm"
|
|
"/dev/nvidia-uvm-tools"
|
|
)
|
|
|
|
local devnums=()
|
|
for dev in "${devices[@]}"; do
|
|
[[ -e "$dev" ]] || continue
|
|
local major_hex
|
|
major_hex=$(stat -c '%t' "$dev")
|
|
devnums+=($((16#$major_hex)))
|
|
echo "lxc.mount.entry: $dev ${dev##*/} none bind,optional,create=file" >>"$conf"
|
|
done
|
|
|
|
echo "lxc.mount.entry: /dev/dri dev/dri none bind,optional,create=dir" >>"$conf"
|
|
|
|
for n in "${devnums[@]}"; do
|
|
echo "lxc.cgroup2.devices.allow: c ${n}:* rwm" >>"$conf"
|
|
done
|
|
msg ok "Installed NVIDIA GPU tools in $ctid"
|
|
}
|