Skip to content

Commit a137564

Browse files
lwinckerslwinckers
committed
Moved to Dockerfile from fedora driver repository
https://github.com/NVIDIA/gpu-driver-container/blob/main/fedora/Dockerfile Some minor changes made as well - driver version etc.
1 parent fc8fdfc commit a137564

File tree

1 file changed

+132
-81
lines changed

1 file changed

+132
-81
lines changed

‎Dockerfile‎

Lines changed: 132 additions & 81 deletions
Original file line numberDiff line numberDiff line change
@@ -1,94 +1,145 @@
1-
FROM nvcr.io/nvidia/cuda:12.9.1-base-ubi8 AS license
2-
3-
# Build nvidia-container-runtime binary
4-
FROM golang:1.24.4 AS build-runtime
5-
6-
WORKDIR /go/bin/nvidia-container-runtime
7-
COPY nvidia-container-runtime .
8-
RUN go mod vendor
9-
RUN go install -v .
10-
11-
# Build driver image
12-
FROM ubuntu:22.04
13-
14-
RUN dpkg --add-architecture i386 && \
15-
apt-get update && apt-get install -y --no-install-recommends \
16-
apt-transport-https \
17-
apt-utils \
18-
bc \
19-
binutils \
20-
build-essential \
21-
ca-certificates \
22-
curl \
23-
gnupg2 \
24-
jq \
25-
kmod \
26-
libc6:i386 \
27-
libelf-dev \
28-
libssl-dev \
29-
kmod \
30-
software-properties-common && \
31-
rm -rf /var/lib/apt/lists/*
32-
33-
RUN echo "deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic main" > /etc/apt/sources.list && \
34-
echo "deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic-updates main" >> /etc/apt/sources.list && \
35-
echo "deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic-security main" >> /etc/apt/sources.list && \
36-
usermod -o -u 0 -g 0 _apt
37-
38-
RUN curl -fsSL -o /usr/local/bin/donkey https://github.com/3XX0/donkey/releases/download/v1.1.0/donkey && \
39-
curl -fsSL -o /usr/local/bin/extract-vmlinux https://raw.githubusercontent.com/torvalds/linux/master/scripts/extract-vmlinux && \
40-
chmod +x /usr/local/bin/donkey /usr/local/bin/extract-vmlinux
1+
ARG FEDORA_VERSION=39
2+
3+
# Stage 1
4+
FROM nvcr.io/nvidia/cuda:12.6.1-base-ubi8 AS build
5+
6+
ARG TARGETARCH
7+
8+
SHELL ["/bin/bash", "-c"]
9+
10+
RUN dnf install -y git wget
11+
12+
ENV GOLANG_VERSION=1.22.7
13+
14+
# download appropriate binary based on the target architecture for multi-arch builds
15+
RUN OS_ARCH=${TARGETARCH/x86_64/amd64} && OS_ARCH=${OS_ARCH/aarch64/arm64} && \
16+
curl https://dl.google.com/go/go${GOLANG_VERSION}.linux-${OS_ARCH}.tar.gz \
17+
| tar -C /usr/local -xz
18+
19+
ENV PATH=/usr/local/go/bin:$PATH
20+
21+
WORKDIR /work
22+
23+
RUN git clone https://github.com/NVIDIA/gpu-driver-container driver && \
24+
cd driver/vgpu/src && \
25+
go build -o vgpu-util && \
26+
mv vgpu-util /work
27+
28+
# Stage 2
29+
FROM fedora:${FEDORA_VERSION}
30+
31+
LABEL org.opencontainers.image.source https://github.com/maastrichtu-library/dsri-nvidia-driver
32+
33+
ARG TARGETARCH
34+
ENV TARGETARCH=$TARGETARCH
35+
36+
ARG KERNEL_TYPE
37+
ENV KERNEL_TYPE=$KERNEL_TYPE
38+
39+
SHELL ["/bin/bash", "-c"]
4140

42-
#ARG BASE_URL=http://us.download.nvidia.com/XFree86/Linux-x86_64
4341
ARG BASE_URL=https://us.download.nvidia.com/tesla
44-
ARG DRIVER_VERSION=450.80.02
42+
ENV BASE_URL=${BASE_URL}
43+
ARG DRIVER_VERSION=570.158.01
4544
ENV DRIVER_VERSION=$DRIVER_VERSION
4645

47-
# Install the userspace components and copy the kernel module sources.
48-
RUN cd /tmp && \
49-
curl -fSsl -O $BASE_URL/$DRIVER_VERSION/NVIDIA-Linux-x86_64-$DRIVER_VERSION.run && \
50-
sh NVIDIA-Linux-x86_64-$DRIVER_VERSION.run -x && \
51-
cd NVIDIA-Linux-x86_64-$DRIVER_VERSION* && \
52-
./nvidia-installer --silent \
53-
--no-kernel-module \
54-
--install-compat32-libs \
55-
--no-nouveau-check \
56-
--no-nvidia-modprobe \
57-
--no-rpms \
58-
--no-backup \
59-
--no-check-for-alternate-installs \
60-
--no-libglx-indirect \
61-
--no-install-libglvnd \
62-
--x-prefix=/tmp/null \
63-
--x-module-path=/tmp/null \
64-
--x-library-path=/tmp/null \
65-
--x-sysconfig-path=/tmp/null && \
66-
mkdir -p /usr/src/nvidia-$DRIVER_VERSION && \
67-
mv LICENSE mkprecompiled kernel /usr/src/nvidia-$DRIVER_VERSION && \
68-
sed '9,${/^\(kernel\|LICENSE\)/!d}' .manifest > /usr/src/nvidia-$DRIVER_VERSION/.manifest && \
69-
rm -rf /tmp/*
70-
71-
# Install and configure nvidia-container-runtime
72-
ENV NVIDIA_VISIBLE_DEVICES void
73-
74-
COPY --from=build-runtime /go/bin/nvidia-container-runtime /usr/bin/nvidia-container-runtime
75-
76-
RUN curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | apt-key add - && \
77-
distribution=$(. /etc/os-release;echo $ID$VERSION_ID) && \
78-
curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.list | tee /etc/apt/sources.list.d/nvidia-container-runtime.list && \
79-
apt-get update && \
80-
apt-get install -y nvidia-container-runtime-hook && \
81-
cd /usr && ln -s lib/x86_64-linux-gnu lib64 && cd - && \
82-
sed -i 's/^#root/root/; s;@/sbin/ldconfig.real;@/run/nvidia/driver/sbin/ldconfig.real;' /etc/nvidia-container-runtime/config.toml
46+
# Arg to indicate if driver type is either of passthrough/baremetal or vgpu
47+
ARG DRIVER_TYPE=passthrough
48+
ENV DRIVER_TYPE=$DRIVER_TYPE
49+
ARG VGPU_LICENSE_SERVER_TYPE=NLS
50+
ENV VGPU_LICENSE_SERVER_TYPE=$VGPU_LICENSE_SERVER_TYPE
51+
# Enable vGPU version compability check by default
52+
ARG DISABLE_VGPU_VERSION_CHECK=true
53+
ENV DISABLE_VGPU_VERSION_CHECK=$DISABLE_VGPU_VERSION_CHECK
54+
# Avoid dependency of container-toolkit for driver container
55+
ENV NVIDIA_VISIBLE_DEVICES=void
56+
57+
# Utils, download plugin, and patch package
58+
RUN dnf install -y util-linux 'dnf-command(download)'
59+
RUN dnf install -y patch
60+
61+
ADD install.sh /tmp/
62+
63+
RUN NVIDIA_GPGKEY_SUM=d0664fbbdb8c32356d45de36c5984617217b2d0bef41b93ccecd326ba3b80c87 && \
64+
OS_ARCH=${TARGETARCH/amd64/x86_64} && OS_ARCH=${OS_ARCH/arm64/sbsa} && \
65+
curl -fsSL "https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$OS_ARCH/D42D0685.pub" | sed '/^Version/d' > /etc/pki/rpm-gpg/RPM-GPG-KEY-NVIDIA && \
66+
echo "$NVIDIA_GPGKEY_SUM /etc/pki/rpm-gpg/RPM-GPG-KEY-NVIDIA" | sha256sum -c --strict -
67+
68+
RUN sh /tmp/install.sh depinstall && \
69+
curl -fsSL -o /usr/local/bin/donkey https://github.com/3XX0/donkey/releases/download/v1.1.0/donkey && \
70+
curl -fsSL -o /usr/local/bin/extract-vmlinux https://raw.githubusercontent.com/torvalds/linux/master/scripts/extract-vmlinux && \
71+
chmod +x /usr/local/bin/donkey /usr/local/bin/extract-vmlinux && \
72+
ln -s /sbin/ldconfig /sbin/ldconfig.real
73+
74+
ADD drivers drivers/
75+
76+
# Fetch the installer automatically for passthrough/baremetal types
77+
# RUN echo $BASE_URL/$DRIVER_VERSION/NVIDIA-Linux-$DRIVER_ARCH-$DRIVER_VERSION.run
78+
RUN if [ "$DRIVER_TYPE" != "vgpu" ]; then \
79+
cd drivers && \
80+
DRIVER_ARCH=${TARGETARCH/amd64/x86_64} && DRIVER_ARCH=${DRIVER_ARCH/arm64/aarch64} && \
81+
curl -fSsl -O $BASE_URL/$DRIVER_VERSION/NVIDIA-Linux-$DRIVER_ARCH-$DRIVER_VERSION.run && \
82+
chmod +x NVIDIA-Linux-$DRIVER_ARCH-$DRIVER_VERSION.run; fi
83+
84+
# ******************************************************************************
85+
# REMOVED FABRIC MANAGER RHEL WORK FOR NOW - HAVEN'T ATTEMPTED INCLUSION
86+
# ******************************************************************************
87+
# Install fabric-manager packages
88+
# RUN if [ "$DRIVER_TYPE" != "vgpu" ] && [ "$TARGETARCH" != "arm64" ]; then \
89+
# versionArray=(${DRIVER_VERSION//./ }); \
90+
# DRIVER_BRANCH=${versionArray[0]}; \
91+
# if [ ${versionArray[0]} -ge 470 ] || ([ ${versionArray[0]} == 460 ] && [ ${versionArray[1]} -ge 91 ]); then \
92+
# fmPackage=nvidia-fabric-manager-${DRIVER_VERSION}-1; \
93+
# else \
94+
# fmPackage=nvidia-fabricmanager-${DRIVER_BRANCH}-${DRIVER_VERSION}-1; \
95+
# fi; \
96+
# nscqPackage=libnvidia-nscq-${DRIVER_BRANCH}-${DRIVER_VERSION}-1; \
97+
# dnf module enable -y nvidia-driver:${DRIVER_BRANCH} && \
98+
# dnf install -y ${fmPackage} ${nscqPackage}; fi
8399

84100
COPY nvidia-driver /usr/local/bin
85101

102+
RUN chmod 755 /usr/local/bin/nvidia-driver
103+
104+
#RUN cat /usr/local/bin/nvidia-driver
105+
86106
WORKDIR /usr/src/nvidia-$DRIVER_VERSION
87107

108+
COPY ocp_dtk_entrypoint /usr/local/bin
109+
RUN chmod +x /usr/local/bin/ocp_dtk_entrypoint
110+
COPY common.sh /usr/local/bin
111+
RUN chmod +x /usr/local/bin/common.sh
112+
ENV PATH="$PATH:/user/local/bin"
113+
114+
COPY --from=build /work/vgpu-util /usr/local/bin
115+
116+
WORKDIR /drivers
117+
88118
ARG PUBLIC_KEY=empty
89119
COPY ${PUBLIC_KEY} kernel/pubkey.x509
90120

91-
# Add NGC DL license
92-
COPY --from=license /NGC-DL-CONTAINER-LICENSE /licenses/NGC-DL-CONTAINER-LICENSE
121+
#ARG PRIVATE_KEY
122+
ARG KERNEL_VERSION=latest
123+
124+
LABEL io.k8s.display-name="NVIDIA Driver Container"
125+
LABEL name="NVIDIA Driver Container"
126+
LABEL vendor="NVIDIA"
127+
LABEL version="${DRIVER_VERSION}"
128+
LABEL release="N/A"
129+
LABEL summary="Provision the NVIDIA driver through containers"
130+
LABEL description="See summary"
131+
132+
# Add NGC DL license from the CUDA image
133+
COPY --from=build /NGC-DL-CONTAINER-LICENSE /licenses/NGC-DL-CONTAINER-LICENSE
134+
135+
# Install / upgrade packages here that are required to resolve CVEs
136+
ARG CVE_UPDATES
137+
RUN if [ -n "${CVE_UPDATES}" ]; then \
138+
yum update -y ${CVE_UPDATES} && \
139+
rm -rf /var/cache/yum/*; \
140+
fi
141+
142+
# Remove cuda repository to avoid GPG errors
143+
RUN rm -f /etc/yum.repos.d/cuda.repo
93144

94-
ENTRYPOINT ["nvidia-driver"]
145+
ENTRYPOINT ["nvidia-driver", "init"]

0 commit comments

Comments
 (0)