Add support for GPU-acceleration

dist/platforms/ubuntu/entrypoint.sh

@@ -10,6 +10,7 @@ mkdir -p "$ACTIVATE_LICENSE_PATH"
 #
 # Run steps
 #
+source /steps/startx.sh
 source /steps/set_extra_git_configs.sh
 source /steps/set_gitcredential.sh
 source /steps/activate.sh

dist/platforms/ubuntu/steps/startx.sh

@@ -0,0 +1,212 @@
+#!/bin/bash -e
+
+# Source:
+# https://github.com/selkies-project/docker-nvidia-glx-desktop/blob/main/entrypoint.sh
+# This Source Code Form is subject to the terms of the Mozilla Public
+# License, v. 2.0. If a copy of the MPL was not distributed with this
+# file, You can obtain one at https://mozilla.org/MPL/2.0/.
+
+# Make all NVIDIA GPUs visible by default
+export NVIDIA_VISIBLE_DEVICES=all
+export DEBIAN_FRONTEND=noninteractive
+export NVIDIA_DRIVER_CAPABILITIES=all
+export APPIMAGE_EXTRACT_AND_RUN=1
+
+# System defaults that should not be changed
+export DISPLAY=:0
+export XDG_RUNTIME_DIR=/tmp/runtime-user
+export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:/usr/lib/i386-linux-gnu${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
+
+# Default environment variables (password is "mypasswd")
+export TZ=UTC
+export SIZEW=1920
+export SIZEH=1080
+export REFRESH=60
+export DPI=96
+export CDEPTH=24
+export VIDEO_PORT=DFP
+export PASSWD=mypasswd
+
+
+init(){
+        # Start DBus without systemd
+        sudo /etc/init.d/dbus start
+
+        # Change time zone from environment variable
+        sudo ln -snf "/usr/share/zoneinfo/$TZ" /etc/localtime && echo "$TZ" | sudo tee /etc/timezone > /dev/null
+
+        # This symbolic link enables running Xorg inside a container with `-sharevts`
+        sudo ln -snf /dev/ptmx /dev/tty7
+
+        # Allow starting Xorg from a pseudoterminal instead of strictly on a tty console
+        if [ ! -f /etc/X11/Xwrapper.config ]; then
+            echo -e "allowed_users=anybody\nneeds_root_rights=yes" | sudo tee /etc/X11/Xwrapper.config > /dev/null
+        fi
+        if grep -Fxq "allowed_users=console" /etc/X11/Xwrapper.config; then
+          sudo sed -i "s/allowed_users=console/allowed_users=anybody/;$ a needs_root_rights=yes" /etc/X11/Xwrapper.config
+        fi
+
+        # Remove existing Xorg configuration
+        if [ -f "/etc/X11/xorg.conf" ]; then
+          sudo rm -f "/etc/X11/xorg.conf"
+        fi
+}
+
+install_driver() {
+        # Install NVIDIA userspace driver components including X graphic libraries
+        if ! command -v nvidia-xconfig &> /dev/null; then
+          # Driver version is provided by the kernel through the container toolkit
+          export DRIVER_VERSION=$(head -n1 </proc/driver/nvidia/version | awk '{print$8}')
+          cd /tmp
+          # If version is different, new installer will overwrite the existing components
+          if [ ! -f "/tmp/NVIDIA-Linux-x86_64-$DRIVER_VERSION.run" ]; then
+            # Check multiple sources in order to probe both consumer and datacenter driver versions
+            curl --progress-bar -fL -O "https://us.download.nvidia.com/XFree86/Linux-x86_64/$DRIVER_VERSION/NVIDIA-Linux-x86_64-$DRIVER_VERSION.run" || curl --progress-bar -fL -O "https://us.download.nvidia.com/tesla/$DRIVER_VERSION/NVIDIA-Linux-x86_64-$DRIVER_VERSION.run" || { echo "Failed NVIDIA GPU driver download. Exiting."; exit 1; }
+          fi
+          # Extract installer before installing
+          sudo sh "NVIDIA-Linux-x86_64-$DRIVER_VERSION.run" -x
+          cd "NVIDIA-Linux-x86_64-$DRIVER_VERSION"
+          # Run installation without the kernel modules and host components
+          sudo ./nvidia-installer --silent \
+                            --no-kernel-module \
+                            --install-compat32-libs \
+                            --no-nouveau-check \
+                            --no-nvidia-modprobe \
+                            --no-rpms \
+                            --no-backup \
+                            --no-check-for-alternate-installs || true
+          sudo rm -rf /tmp/NVIDIA* && cd ~
+        fi
+}
+
+find_gpu(){
+        # Get first GPU device if all devices are available or `NVIDIA_VISIBLE_DEVICES` is not set
+        if [ "$NVIDIA_VISIBLE_DEVICES" == "all" ]; then
+          export GPU_SELECT=$(sudo nvidia-smi --query-gpu=uuid --format=csv | sed -n 2p)
+        elif [ -z "$NVIDIA_VISIBLE_DEVICES" ]; then
+          export GPU_SELECT=$(sudo nvidia-smi --query-gpu=uuid --format=csv | sed -n 2p)
+        # Get first GPU device out of the visible devices in other situations
+        else
+          export GPU_SELECT=$(sudo nvidia-smi --id=$(echo "$NVIDIA_VISIBLE_DEVICES" | cut -d ',' -f1) --query-gpu=uuid --format=csv | sed -n 2p)
+          if [ -z "$GPU_SELECT" ]; then
+            export GPU_SELECT=$(sudo nvidia-smi --query-gpu=uuid --format=csv | sed -n 2p)
+          fi
+        fi
+
+        if [ -z "$GPU_SELECT" ]; then
+          echo "No NVIDIA GPUs detected or nvidia-container-toolkit not configured. Exiting."
+          exit 1
+        fi
+
+        # Setting `VIDEO_PORT` to none disables RANDR/XRANDR, do not set this if using datacenter GPUs
+        if [ "${VIDEO_PORT,,}" = "none" ]; then
+          export CONNECTED_MONITOR="--use-display-device=None"
+        # The X server is otherwise deliberately set to a specific video port despite not
+        # being plugged to enable RANDR/XRANDR, monitor will display the screen if plugged to the
+        # specific port
+        else
+          export CONNECTED_MONITOR="--connected-monitor=${VIDEO_PORT}"
+        fi
+}
+
+create_xorg_conf(){
+        # Bus ID from nvidia-smi is in hexadecimal format, should be converted to decimal
+        # format which Xorg understands, required because nvidia-xconfig doesn't work as intended in
+        # a container
+        HEX_ID=$(sudo nvidia-smi --query-gpu=pci.bus_id --id="$GPU_SELECT" --format=csv |sed -n 2p)
+        IFS=":." ARR_ID=($HEX_ID)
+        unset IFS
+        BUS_ID=PCI:$((16#${ARR_ID[1]})):$((16#${ARR_ID[2]})):$((16#${ARR_ID[3]}))
+
+        # A custom modeline should be generated because there is no monitor to fetch this
+        # information normally
+        export MODELINE=$(cvt -r "${SIZEW}" "${SIZEH}" "${REFRESH}" | sed -n 2p)
+
+        # Generate /etc/X11/xorg.conf with nvidia-xconfig
+        sudo nvidia-xconfig --virtual="${SIZEW}x${SIZEH}" --depth="$CDEPTH" --mode=$(echo "$MODELINE" | awk '{print $2}' | tr -d '"') \
+                --allow-empty-initial-configuration \
+                --no-probe-all-gpus \
+                --busid="$BUS_ID" \
+                --no-multigpu \
+                --no-sli \
+                --no-base-mosaic \
+                --only-one-x-screen ${CONNECTED_MONITOR}
+
+        # Guarantee that the X server starts without a monitor by adding more options to
+        # the configuration
+        sudo sed -i '/Driver\s\+"nvidia"/a\    Option         "ModeValidation" "NoMaxPClkCheck, NoEdidMaxPClkCheck, NoMaxSizeCheck, NoHorizSyncCheck, NoVertRefreshCheck, NoVirtualSizeCheck, NoExtendedGpuCapabilitiesCheck, NoTotalSizeCheck, NoDualLinkDVICheck, NoDisplayPortBandwidthCheck, AllowNon3DVisionModes, AllowNonHDMI3DModes, AllowNonEdidModes, NoEdidHDMI2Check, AllowDpInterlaced"\n    Option         "HardDPMS" "False"' /etc/X11/xorg.conf
+
+        # Add custom generated modeline to the configuration
+        sudo sed -i '/Section\s\+"Monitor"/a\    '"$MODELINE" /etc/X11/xorg.conf
+}
+
+start_app(){
+        # Since we dont boot the system with systemd, starting and maintaining long-running
+        # processes is tricky and has multiple approaches. You can do as the Nvidia-GLX-Desktop
+        # does and use supervisord to start process as shown in the link below.
+        # - https://github.com/selkies-project/docker-nvidia-glx-desktop/blob/main/supervisord.conf
+        #
+        # You can also use this neat project to simulate systemd/systemctl in a container:
+        # - https://github.com/gdraheim/docker-systemctl-replacement
+        #
+        # I'm using another option, which is to use tmux to manage sessions of the individual application.
+        # this method is janky and fragile but I think it's easier to use during development because
+        # it only relies on tmux (no special dependancies or extra config files) and attaching to the
+        # sessions via tmux attach-session -t <app> is easier and more intuitive than trying to find
+        # the process ID and re-attach a shell. Additionally it prevents the main user-shell from being
+        # monopoloized by stdout messages from supervisord andallows for additional commands to be run after
+        # entrypoint.sh has finished.
+        #
+        # I will probably switch to using supervisord or docker-systemctl-replacement later as a more
+        # stable solution after I'm done making constant changes to the project.
+
+        # Starts an empty Xorg session on DDISPLAY:0
+        tmux new-session -d -s "xorg"
+        tmux send-keys -t "xorg" "export DISPLAY=:0 && \
+        Xorg vt7 \
+        -noreset \
+        -novtswitch \
+        -sharevts \
+        -dpi ${DPI} \
+        +extension GLX \
+        +extension RANDR \
+        +extension RENDER \
+        +extension MIT-SHM ${DISPLAY}" ENTER
+
+        echo "Waiting for X socket"
+        until [ -S "/tmp/.X11-unix/X${DISPLAY/:/}" ]; do sleep 1; done
+        echo "X socket is ready"
+
+        # Start an x11vnc session that brodcasts the contents our X session
+        tmux new-session -d -s "x11vnc"
+        tmux send-keys -t "x11vnc" "export DISPLAY=:0 && \
+        sudo x11vnc -display ${DISPLAY} \
+        -shared \
+        -loop \
+        -repeat \
+        -xkb \
+        -snapfb \
+        -threads \
+        -xrandr resize \
+        -passwd "${BASIC_AUTH_PASSWORD:-$PASSWD}" \
+        -rfbport 5900 ${NOVNC_VIEWONLY}" ENTER
+
+        # Start the no-vnc session that exposes x11vnc over websocket
+        tmux new-session -d -s "novnc"
+        tmux send-keys -t "novnc" "export DISPLAY=:0 && \
+        /opt/noVNC/utils/novnc_proxy \
+        --vnc localhost:5900 \
+        --listen 8080 \
+        --heartbeat 10" ENTER
+
+        # Start the desktop session
+        tmux new-session -d -s "app"
+        tmux send-keys -t "app" "export DISPLAY=:0 && \
+        startxfce4" ENTER
+}
+
+init
+find_gpu
+install_driver
+create_xorg_conf
+start_app

src/model/docker.ts

@@ -59,6 +59,7 @@ class Docker {
     return `docker run \
             --workdir ${dockerWorkspacePath} \
             --rm \
+            --gpus all \
             ${ImageEnvironmentFactory.getEnvVarString(parameters, additionalVariables)} \
             --env UNITY_SERIAL \
             --env GITHUB_WORKSPACE=${dockerWorkspacePath} \