Skip to content

[spark-rapids] Fix MIG driver installation on Debian 12 #1413

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
nvliyuan wants to merge 3 commits into
base: main
Choose a base branch
from
Open

[spark-rapids] Fix MIG driver installation on Debian 12 #1413

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
25e8852
[spark-rapids] Fix MIG driver install on Debian 12
nvliyuan Oct 27, 2026
bad1ca8
Update spark-rapids/mig.sh
nvliyuan Oct 27, 2026
bf02dd5
fix: address MIG install review comments
nvliyuan Oct 27, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
76 changes: 54 additions & 22 deletions spark-rapids/mig.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -172,11 +172,11 @@ function configure_dkms_certs() {

function clear_dkms_key {
if [[ -z "${PSN}" ]]; then
echo "No signing secret provided. skipping" >2
echo "No signing secret provided. skipping" >&2
return 0
fi
echo "WARN -- PURGING SIGNING MATERIAL -- WARN" >2
echo "future dkms runs will not use correct signing key" >2
echo "WARN -- PURGING SIGNING MATERIAL -- WARN" >&2
echo "future dkms runs will not use correct signing key" >&2
rm -rf "${CA_TMPDIR}" /var/lib/dkms/mok.key /var/lib/shim-signed/mok/MOK.priv
}

Expand Down Expand Up @@ -217,6 +217,17 @@ NVIDIA_DRIVER_VERSION=$(get_metadata_attribute 'driver-version' '550.54.15') #53
CUDA_VERSION_MAJOR="${CUDA_VERSION%.*}" #12.2

# EXCEPTIONS
# Debian 12 security kernels use the four-argument pci_resize_resource API,
# which is incompatible with NVIDIA 550 open modules.
if is_debian12 ; then
NVIDIA_DRIVER_VERSION=$(get_metadata_attribute 'driver-version' '580.95.05')
if [[ "${NVIDIA_DRIVER_VERSION%%.*}" == "550" ]]; then
echo "WARNING: Driver version 550 is incompatible with Debian 12 kernel. Overriding to 580.95.05" >&2
NVIDIA_DRIVER_VERSION='580.95.05'
fi
USE_REPO_INSTALL="true"
fi

# Change CUDA version for Ubuntu 18 (Cuda 12.1.1 - Driver v530.30.02 is the latest version supported by Ubuntu 18)
if [[ "${OS_NAME}" == "ubuntu" ]]; then
if is_ubuntu18 ; then
Expand Down Expand Up @@ -296,30 +307,51 @@ function install_nvidia_gpu_driver() {

execute_with_retries "apt-get install -y -q 'linux-headers-$(uname -r)'"

curl -fsSL --retry-connrefused --retry 3 --retry-max-time 5 \
"${LOCAL_DEB_URL}" -o /tmp/local-installer.deb
if [[ "${USE_REPO_INSTALL:-false}" == "true" ]]; then
execute_with_retries \
"curl -fsSL --retry-connrefused --retry 3 --retry-max-time 5 https://developer.download.nvidia.com/compute/cuda/repos/${shortname}/x86_64/cuda-keyring_1.1-1_all.deb -o /tmp/cuda-keyring_1.1-1_all.deb"
execute_with_retries "dpkg -i /tmp/cuda-keyring_1.1-1_all.deb"
rm -f /tmp/cuda-keyring_1.1-1_all.deb
add_contrib_components
execute_with_retries "apt-get update"
Comment thread
nvliyuan marked this conversation as resolved.

dpkg -i /tmp/local-installer.deb
rm /tmp/local-installer.deb
cp ${DIST_KEYRING_DIR}/cuda-*-keyring.gpg /usr/share/keyrings/
execute_with_retries "apt-get install -y -q --no-install-recommends dkms"
execute_with_retries \
"apt-get install -y -q --no-install-recommends nvidia-driver-pinning-${NVIDIA_DRIVER_VERSION_PREFIX}"

add_contrib_components
configure_dkms_certs
execute_with_retries \
"apt-get install -y -q --no-install-recommends nvidia-kernel-open-dkms nvidia-driver-cuda"
clear_dkms_key
Comment thread
nvliyuan marked this conversation as resolved.

execute_with_retries "apt-get update"
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-toolkit-${CUDA_VERSION_MAJOR//./-}"
else
curl -fsSL --retry-connrefused --retry 3 --retry-max-time 5 \
"${LOCAL_DEB_URL}" -o /tmp/local-installer.deb

## EXCEPTION
if is_debian10 ; then
apt-get remove -y libglvnd0
apt-get install -y ca-certificates-java
fi
dpkg -i /tmp/local-installer.deb
rm /tmp/local-installer.deb
cp ${DIST_KEYRING_DIR}/cuda-*-keyring.gpg /usr/share/keyrings/

configure_dkms_certs
execute_with_retries "apt-get install -y -q nvidia-kernel-open-dkms"
clear_dkms_key
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-drivers-${NVIDIA_DRIVER_VERSION_PREFIX}"
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-toolkit-${CUDA_VERSION_MAJOR//./-}"
add_contrib_components

execute_with_retries "apt-get update"

## EXCEPTION
if is_debian10 ; then
apt-get remove -y libglvnd0
apt-get install -y ca-certificates-java
fi

configure_dkms_certs
execute_with_retries "apt-get install -y -q nvidia-kernel-open-dkms"
clear_dkms_key
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-drivers-${NVIDIA_DRIVER_VERSION_PREFIX}"
execute_with_retries \
"apt-get install -y -q --no-install-recommends cuda-toolkit-${CUDA_VERSION_MAJOR//./-}"
fi

modprobe nvidia

Expand Down
Loading