Merge ~bladernr/plainbox-provider-certification-server:gpu-setup-update into plainbox-provider-certification-server:master

[Jeff Lane ]

Multiple improvements to gpu-setup.sh

Multiple improvements to gpu-setup.sh

[Jeff Lane  (bladernr)]

Self=approved to get this in and fix the issues discuvered when really using the original version of the script.

This version has been tested with Xenial and Bionic on my local test machine as well as a multiple GPU box hosted at a partner lab remotely.

review: Approve

diff --git a/tools/gpu-setup.sh b/tools/gpu-setup.sh
index c6909ca..22a8430 100755
--- a/tools/gpu-setup.sh
+++ b/tools/gpu-setup.sh
@@ -1,73 +1,79 @@
 #!/bin/bash
 
 if [[ $EUID -ne 0 ]]; then
-   echo "This script must be run as root"
+   echo "ERROR: This script must be run as root"
    exit 1
 fi
 
 echo "Configuring system for GPU Testing"
 echo "**********************************"
-echo
+echo "*"
 
 # For now we need internet access for this to work.  Future versions will
 # remove this necessity
+echo "*  Testing network connectivity"
 [[ ! `ping -c 1 www.ubuntu.com` ]] \
-    && echo "This script requires internet access to function correctly" \
+    && echo "ERROR: This script requires internet access to function correctly" \
     && exit 1
-
-# Disable Nouveau and reset initramfs
-if lsmod |grep -q nouveau; then
-    echo "*  Disabling nouveau driver"
-    echo blacklist nouveau > /etc/modprobe.d/blacklist-nvidia-nouveau.conf
-    echo options nouveau modeset=0 >> /etc/modprobe.d/blacklist-nvidia-nouveau.conf
-    echo "*  Updating initramfs"
-    update-initramfs -u
-    echo "*  Finished disabling nouveau driver"
-else
-    echo "*  nouveau not detected, proceeding..."
-fi
-
 echo "*"
 echo "**********************************"
+echo "*"
+echo "*  Adding nVidia package repository"
+## Leave the following bits for now, it's useful code to template if we sort
+## out local hosting in the future.
+#file_pattern="cuda-repo-ubuntu$(lsb_release -r | cut -f 2 | sed -es/\\.//)*.deb"
+#gw_ip=`ip route | awk '/default/ { print $3 }'`
 
-# Add required PPAs
-echo "*  Adding Canonical graphis-drivers PPA" 
-add-apt-repository -y ppa:graphics-drivers/ppa
+# attempt a download from our MAAS server (presume the SUT's gw IP is MAAS)
+#wget -r -l1 -nd --no-parent -A "$file_pattern" http://$gw_ip/nvidia/
+#nvidia_pkg=`find / -name $file_pattern`
 
-echo "*  Adding nVidia package repository"
+#if [ -z "$nvidia_pkg" ]; then
+#    echo "*  No local package found, installing remote installer"
+#    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
+#    dpkg -i cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
+#    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
+#else
+#    echo "*  Local package detected at $nvidia_pkg... installing"
+#    key_cmd=`dpkg -i $nvidia_pkg |awk '/To install the key, run this command/{getline; print}'`
+#    $key_cmd
+#fi    
 
-file_pattern="cuda-repo-ubuntu$(lsb_release -r | cut -f 2 | sed -es/\\.//)*.deb"
-gw_ip=`ip route | awk '/default/ { print $3 }'`
+## For now, require internet access and installing directly from nVidia
+# SAUCE: https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1804&target_type=debnetwork
 
-# attempt a download from our MAAS server (presume the SUT's gw IP is MAAS)
-wget -r -l1 -nd --no-parent -A "$file_pattern" http://$gw_ip/nvidia/
-nvidia_pkg=`find / -name $file_pattern`
+OSRELEASE=ubuntu`lsb_release -r | cut -f2 |sed -e 's/\.//'`
 
-if [ -z "$nvidia_pkg" ]; then
-    echo "*  No local package found, installing remote installer"
-    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
-    dpkg -i cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
-    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
-else
-    echo "*  Local package detected at $nvidia_pkg... installing"
-    key_cmd=`dpkg -i $nvidia_pkg |awk '/To install the key, run this command/{getline; print}'`
-    $key_cmd
-fi    
+wget https://developer.download.nvidia.com/compute/cuda/repos/$OSRELEASE/x86_64/cuda-$OSRELEASE.pin
+mv cuda-$OSRELEASE.pin /etc/apt/preferences.d/cuda-repository-pin-600
+# Ran into a case where the 16.04 repo uses http while 18.04 uses https :/
+KEYSERVER_URL_TAIL="://developer.download.nvidia.com/compute/cuda/repos/$OSRELEASE/x86_64/7fa2af80.pub"
+SUCCESS=1
+for KEYSERVER_URL_HEAD in https http; do
+    KEYSERVER_URL=$KEYSERVER_URL_HEAD$KEYSERVER_URL_TAIL
+    apt-key adv --fetch-keys $KEYSERVER_URL
+    SUCCESS=$?
+    [ $SUCCESS == 0 ] && break
+done
 
-echo "*  Updating apt cache"
-apt update
+add-apt-repository "deb http://developer.download.nvidia.com/compute/cuda/repos/$OSRELEASE/x86_64/ /"
 
 # Install necessary files
+apt update
 echo "*  Installing necessary pacakges"
-apt install -y cuda nvidia-cuda-toolkit build-essential git
+apt install -y build-essential git cuda
 
+#fix the path to get nvcc from the cuda package
+CUDA_PATH=$(find /usr/local -maxdepth 1 -type d -iname "cuda*")/bin
+export PATH=$PATH:$CUDA_PATH
 
 # get the gpu-burn repo and build it
 echo "*  Cloning gpu-burn repo"
-git clone https://github.com/wilicc/gpu-burn.git
-cd gpu-burn
+GPU_BURN_DIR=/opt/gpu-burn
+git clone https://github.com/wilicc/gpu-burn.git $GPU_BURN_DIR
+cd $GPU_BURN_DIR
 echo "*  Building gpu-burn"
 make && echo "*  Build completed..."
-
-echo "*  Completed installatiion. Please reboot the machine now"
+echo "*"
+echo "*  Completed installation. Please reboot the machine now"
 echo "*  to load the nVidia proprietary drivers"