feat: support lts and production nvidia modules

Support LTS and production versions of NVIDIA kernel modules as per https://docs.nvidia.com/datacenter/tesla/drivers/index.html#lifecycle Part of: siderolabs/talos#9086 Signed-off-by: Noel Georgi <git@frezbo.dev>
siderolabs · Aug 2, 2024 · 75adbde · 75adbde
1 parent a97d58f
commit 75adbde
Show file tree

Hide file tree

Showing 10 changed files with 165 additions and 72 deletions.
diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
@@ -1,6 +1,6 @@
 # THIS FILE WAS AUTOMATICALLY GENERATED, PLEASE DO NOT EDIT.
 #
-# Generated on 2024-07-03T10:20:21Z by kres 2c4ff8a.
+# Generated on 2024-07-31T18:33:29Z by kres faf91e3.
 
 name: default
 concurrency:
@@ -33,7 +33,7 @@ jobs:
       labels: ${{ steps.retrieve-pr-labels.outputs.result }}
     services:
       buildkitd:
-        image: moby/buildkit:v0.14.1
+        image: moby/buildkit:v0.15.0
         options: --privileged
         ports:
           - 1234:1234
@@ -85,6 +85,7 @@ jobs:
         run: |
           make
       - name: Build nonfree
+        if: github.event_name == 'pull_request'
         run: |
           make nonfree
       - name: Login to registry
@@ -136,7 +137,7 @@ jobs:
       - default
     services:
       buildkitd:
-        image: moby/buildkit:v0.14.1
+        image: moby/buildkit:v0.15.0
         options: --privileged
         ports:
           - 1234:1234

diff --git a/.github/workflows/weekly.yaml b/.github/workflows/weekly.yaml
@@ -1,6 +1,6 @@
 # THIS FILE WAS AUTOMATICALLY GENERATED, PLEASE DO NOT EDIT.
 #
-# Generated on 2024-07-02T13:34:14Z by kres 582671e.
+# Generated on 2024-07-31T18:33:29Z by kres faf91e3.
 
 name: weekly
 concurrency:
@@ -16,7 +16,7 @@ jobs:
       - pkgs
     services:
       buildkitd:
-        image: moby/buildkit:v0.14.1
+        image: moby/buildkit:v0.15.0
         options: --privileged
         ports:
           - 1234:1234

diff --git a/.kres.yaml b/.kres.yaml
@@ -40,11 +40,13 @@ spec:
     - kernel
     - drbd-pkg
     - gasket-driver-pkg
-    - nvidia-open-gpu-kernel-modules-pkg
+    - nvidia-open-gpu-kernel-modules-lts-pkg
+    - nvidia-open-gpu-kernel-modules-production-pkg
     - zfs-pkg
   additionalTargets:
     nonfree:
-      - nonfree-kmod-nvidia-pkg
+      - nonfree-kmod-nvidia-lts-pkg
+      - nonfree-kmod-nvidia-production-pkg
   reproducibleTargetName: reproducibility
 ---
 kind: auto.CustomSteps

diff --git a/Makefile b/Makefile
@@ -1,6 +1,6 @@
 # THIS FILE WAS AUTOMATICALLY GENERATED, PLEASE DO NOT EDIT.
 #
-# Generated on 2024-06-25T05:52:46Z by kres 4c9f215.
+# Generated on 2024-08-02T11:55:34Z by kres faf91e3.
 
 # common variables
 
@@ -79,9 +79,11 @@ TARGETS += xfsprogs
 TARGETS += kernel
 TARGETS += drbd-pkg
 TARGETS += gasket-driver-pkg
-TARGETS += nvidia-open-gpu-kernel-modules-pkg
+TARGETS += nvidia-open-gpu-kernel-modules-lts-pkg
+TARGETS += nvidia-open-gpu-kernel-modules-production-pkg
 TARGETS += zfs-pkg
-NONFREE_TARGETS = nonfree-kmod-nvidia-pkg
+NONFREE_TARGETS = nonfree-kmod-nvidia-lts-pkg
+NONFREE_TARGETS += nonfree-kmod-nvidia-production-pkg
 
 # help menu
 

diff --git a/Pkgfile b/Pkgfile
@@ -132,10 +132,13 @@ vars:
   musl_sha256: a9a118bbe84d8764da0ea0d28b3ab3fae8477fc7e4085d90102b8596fc7c75e4
   musl_sha512: 7bb7f7833923cd69c7a1a9b8a5f1784bfd5289663eb6061dcd43d583e45987df8a68a1be05d75cc1c88a3f5b610653d1a70f4a9cff4d8f7fd41ae73ee058c17c
 
-  # renovate: datasource=github-releases depName=nvidia/open-gpu-kernel-modules
-  nvidia_driver_version: 535.129.03
-  nvidia_driver_sha256: 170ee91b6cd6c801bba0de91e0d62af8e574adde38120bc762784755b5a9821d
-  nvidia_driver_sha512: c8104103f34365b7a20dcebb64e1eb2e64a7b056b44329fa305f81e267759f2b1bb1c09923e617d4f5f4af3be01c3b8e2e6d739f03d4d0f129b31331fec3a4d8
+  # renovate: datasource=github-releases extractVersion=^\d+\.(?<version>\d+\.\d+)$ depName=nvidia/open-gpu-kernel-modules
+  # NOTE: Use the version that's also available under fabricmanager at https://developer.download.nvidia.com/compute/cuda/redist/fabricmanager/linux-x86_64/
+  nvidia_driver_version_lts: 535.104.05
+
+  # renovate: datasource=github-releases extractVersion=^\d+\.(?<version>\d+\.\d+)$ depName=nvidia/open-gpu-kernel-modules
+  # NOTE: Use the version that's also available under fabricmanager at https://developer.download.nvidia.com/compute/cuda/redist/fabricmanager/linux-x86_64/
+  nvidia_driver_version_production: 550.54.15
 
   # renovate: datasource=git-tags extractVersion=^openssl-(?<version>.*)$ depName=git://git.openssl.org/openssl.git
   openssl_version: 3.3.0

diff --git a/nonfree/kmod-nvidia/pkg.yaml → nonfree/kmod-nvidia/lts/pkg.yaml b/nonfree/kmod-nvidia/pkg.yaml → nonfree/kmod-nvidia/lts/pkg.yaml
@@ -1,40 +1,34 @@
-name: nonfree-kmod-nvidia-pkg
+name: nonfree-kmod-nvidia-lts-pkg
 variant: scratch
 shell: /toolchain/bin/bash
 dependencies:
   - stage: kernel-build
 steps:
   - sources:
     # {{ if eq .ARCH "aarch64" }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
-      - url: https://download.nvidia.com/XFree86/Linux-aarch64/{{ .nvidia_driver_version }}/NVIDIA-Linux-aarch64-{{ .nvidia_driver_version }}.run
-        destination: nvidia.run
-        sha256: 8ba8d961457a241bcdf91b76d6fe2f36cb473c8bbdb02fb6650a622ce2e85b33
-        sha512: 706de7e53b81f909d8bc6a12a39c594754a164c49f5d23c7939dc3abcfc04f5d5b12b7d65762ae574582149a098f06ee5fe95be4f8ad1056a3307a6ce93f3c00
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-sbsa/nvidia_driver-linux-sbsa-{{ .nvidia_driver_version_lts }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: eb089f52ddda09f4b2bcf0eb661d0bb7881a3d63cbca558c9530ab50b4ef84ee
+        sha512: cbc575b8f72211c54b962c3c12a671d4b993436f5c292181c5db098a5b5fe909ca3e001e679430dfa393cd63a626d4c2cee1c2dd3e2b79464eea4d72fc06a3ff
     # {{ else }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
-      - url: https://download.nvidia.com/XFree86/Linux-x86_64/{{ .nvidia_driver_version }}/NVIDIA-Linux-x86_64-{{ .nvidia_driver_version }}.run
-        destination: nvidia.run
-        sha256: e6dca5626a2608c6bb2a046cfcb7c1af338b9e961a7dd90ac09bb8a126ff002e
-        sha512: 3d7142658fe836e1debf7786857bdb293490ef33351e9b7d39face245fe8596b0f46052b86fae08350fcda1e2a9fd68d7309b94e107d1b016bd529d8fc37e31f
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-x86_64/nvidia_driver-linux-x86_64-{{ .nvidia_driver_version_lts }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: f19c11dfda9e1e5c3e0a2ca775cdaa851431648540f4e8e2916c53735debd450
+        sha512: 7505eda598719325b757375e25a98b17fb0e31d355e42277c32897648f63769222ac32a50f7ae89998fea52bd858160c83114e473e36cd36cfef17d57f8fdde6
     # {{ end }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
     env:
       ARCH: {{ if eq .ARCH "aarch64"}}arm64{{ else if eq .ARCH "x86_64" }}x86_64{{ else }}unsupported{{ end }}
     prepare:
       - |
-        export PATH=/toolchain/bin:$PATH
-        export GUESS_MD5_PATH=/toolchain/bin
-
-        rm -f /dev/tty && ln -s /dev/stdout /dev/tty
-        ln -s /toolchain/bin/echo /toolchain/bin/which
-
-        /toolchain/bin/bash nvidia.run --extract-only
+        tar xf nvidia.tar.xz --strip-components=1
     build:
       - |
-        cd NVIDIA-Linux-*/kernel
+        cd kernel
 
         make -j $(nproc) SYSSRC=/src
     install:
       - |
-        cd NVIDIA-Linux-*/kernel
+        cd kernel
 
         mkdir -p /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
         cp /src/modules.order /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/

diff --git a/nonfree/kmod-nvidia/production/pkg.yaml b/nonfree/kmod-nvidia/production/pkg.yaml
@@ -0,0 +1,41 @@
+name: nonfree-kmod-nvidia-production-pkg
+variant: scratch
+shell: /toolchain/bin/bash
+dependencies:
+  - stage: kernel-build
+steps:
+  - sources:
+    # {{ if eq .ARCH "aarch64" }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-sbsa/nvidia_driver-linux-sbsa-{{ .nvidia_driver_version_production }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: b39a8cb837fe3846284d89ab2202e93b14804349d3fc9aa8a8384fd2759ec2f5
+        sha512: 164636addea50d4d28c4630bf7eaf6060c6392b4aac670c0f5e43996fb309203efc1fa95d464eed8a9d3a22fa2aa5f446195c0f3e4dafa9e32a532447810a2d7
+    # {{ else }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-x86_64/nvidia_driver-linux-x86_64-{{ .nvidia_driver_version_production }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: e1416ebc1e7b4a14d393ca584e3f9c5e22f9750f4ba502a8ba2c5c5ad943bbb5
+        sha512: 188e742796ace5a48b6cdbedae0943c2be593f9e48fb3107ac9f844ed443369cbe977aa084abf17604dd789f475633cd419ad80d63c27429d2760a4018c57586
+    # {{ end }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+    env:
+      ARCH: {{ if eq .ARCH "aarch64"}}arm64{{ else if eq .ARCH "x86_64" }}x86_64{{ else }}unsupported{{ end }}
+    prepare:
+      - |
+        tar xf nvidia.tar.xz --strip-components=1
+    build:
+      - |
+        cd kernel
+
+        make -j $(nproc) SYSSRC=/src
+    install:
+      - |
+        cd kernel
+
+        mkdir -p /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.order /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.builtin /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.builtin.modinfo /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+
+        make -j $(nproc) modules_install SYSSRC=/src DEPMOD=/toolchain/bin/depmod INSTALL_MOD_PATH=/rootfs INSTALL_MOD_DIR=extras INSTALL_MOD_STRIP=1
+finalize:
+  - from: /rootfs
+    to: /
diff --git a/nvidia-open-gpu-kernel-modules/lts/pkg.yaml b/nvidia-open-gpu-kernel-modules/lts/pkg.yaml
@@ -0,0 +1,45 @@
+name: nvidia-open-gpu-kernel-modules-lts-pkg
+variant: scratch
+shell: /toolchain/bin/bash
+dependencies:
+  - stage: kernel-build
+steps:
+  - sources:
+    # {{ if eq .ARCH "aarch64" }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-sbsa/nvidia_driver-linux-sbsa-{{ .nvidia_driver_version_lts }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: eb089f52ddda09f4b2bcf0eb661d0bb7881a3d63cbca558c9530ab50b4ef84ee
+        sha512: cbc575b8f72211c54b962c3c12a671d4b993436f5c292181c5db098a5b5fe909ca3e001e679430dfa393cd63a626d4c2cee1c2dd3e2b79464eea4d72fc06a3ff
+    # {{ else }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-x86_64/nvidia_driver-linux-x86_64-{{ .nvidia_driver_version_lts }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: f19c11dfda9e1e5c3e0a2ca775cdaa851431648540f4e8e2916c53735debd450
+        sha512: 7505eda598719325b757375e25a98b17fb0e31d355e42277c32897648f63769222ac32a50f7ae89998fea52bd858160c83114e473e36cd36cfef17d57f8fdde6
+    # {{ end }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+    env:
+      ARCH: {{ if eq .ARCH "aarch64"}}arm64{{ else if eq .ARCH "x86_64" }}x86_64{{ else }}unsupported{{ end }}
+    prepare:
+      - |
+        tar xf nvidia.tar.xz --strip-components=1
+    build:
+      - |
+        cd kernel-open
+
+        make -j $(nproc) SYSSRC=/src
+    install:
+      - |
+        cd kernel-open
+
+        mkdir -p /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.order /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.builtin /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.builtin.modinfo /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+
+        make -j $(nproc) modules_install SYSSRC=/src DEPMOD=/toolchain/bin/depmod INSTALL_MOD_PATH=/rootfs INSTALL_MOD_STRIP=1
+    test:
+      - |
+        # https://www.kernel.org/doc/html/v4.15/admin-guide/module-signing.html#signed-modules-and-stripping
+        find /rootfs/lib/modules -name '*.ko' -exec grep -FL '~Module signature appended~' {} \+
+finalize:
+  - from: /rootfs
+    to: /
diff --git a/nvidia-open-gpu-kernel-modules/pkg.yaml b/nvidia-open-gpu-kernel-modules/pkg.yaml
diff --git a/nvidia-open-gpu-kernel-modules/production/pkg.yaml b/nvidia-open-gpu-kernel-modules/production/pkg.yaml
@@ -0,0 +1,45 @@
+name: nvidia-open-gpu-kernel-modules-production-pkg
+variant: scratch
+shell: /toolchain/bin/bash
+dependencies:
+  - stage: kernel-build
+steps:
+  - sources:
+    # {{ if eq .ARCH "aarch64" }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-sbsa/nvidia_driver-linux-sbsa-{{ .nvidia_driver_version_production }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: b39a8cb837fe3846284d89ab2202e93b14804349d3fc9aa8a8384fd2759ec2f5
+        sha512: 164636addea50d4d28c4630bf7eaf6060c6392b4aac670c0f5e43996fb309203efc1fa95d464eed8a9d3a22fa2aa5f446195c0f3e4dafa9e32a532447810a2d7
+    # {{ else }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+      - url: https://developer.download.nvidia.com/compute/cuda/redist/nvidia_driver/linux-x86_64/nvidia_driver-linux-x86_64-{{ .nvidia_driver_version_production }}-archive.tar.xz
+        destination: nvidia.tar.xz
+        sha256: e1416ebc1e7b4a14d393ca584e3f9c5e22f9750f4ba502a8ba2c5c5ad943bbb5
+        sha512: 188e742796ace5a48b6cdbedae0943c2be593f9e48fb3107ac9f844ed443369cbe977aa084abf17604dd789f475633cd419ad80d63c27429d2760a4018c57586
+    # {{ end }} This in fact is YAML comment, but Go templating instruction is evaluated by bldr
+    env:
+      ARCH: {{ if eq .ARCH "aarch64"}}arm64{{ else if eq .ARCH "x86_64" }}x86_64{{ else }}unsupported{{ end }}
+    prepare:
+      - |
+        tar xf nvidia.tar.xz --strip-components=1
+    build:
+      - |
+        cd kernel-open
+
+        make -j $(nproc) SYSSRC=/src
+    install:
+      - |
+        cd kernel-open
+
+        mkdir -p /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.order /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.builtin /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+        cp /src/modules.builtin.modinfo /rootfs/lib/modules/$(cat /src/include/config/kernel.release)/
+
+        make -j $(nproc) modules_install SYSSRC=/src DEPMOD=/toolchain/bin/depmod INSTALL_MOD_PATH=/rootfs INSTALL_MOD_STRIP=1
+    test:
+      - |
+        # https://www.kernel.org/doc/html/v4.15/admin-guide/module-signing.html#signed-modules-and-stripping
+        find /rootfs/lib/modules -name '*.ko' -exec grep -FL '~Module signature appended~' {} \+
+finalize:
+  - from: /rootfs
+    to: /