cudaaaa

Allow specifying the variant for the ubuntu drivers

.gitignore

@@ -1 +1,2 @@
 *.retry
+.idea/

README.md

@@ -23,24 +23,24 @@ $ ansible-galaxy install nvidia.nvidia_driver
 ## Role variables
 
 
-| Variable | Default value | Description |
+| Variable                            | Default value                   | Description                                                                                                           |
-| -------- | ------------- | ----------- |
+|-------------------------------------|---------------------------------|-----------------------------------------------------------------------------------------------------------------------|
-| `nvidia_driver_package_state` | `"present"` | Package state for NVIDIA driver packages |
+| `nvidia_driver_package_state`       | `"present"`                     | Package state for NVIDIA driver packages                                                                              |
-| `nvidia_driver_package_version` | `""` | Package version to install. Note that this should match the actual version of the deb or RPM package to be installed. |
+| `nvidia_driver_package_version`     | `""`                            | Package version to install. Note that this should match the actual version of the deb or RPM package to be installed. |
-| `nvidia_driver_persistence_mode_on` | `yes` | Whether to enable persistence mode (boolean) |
+| `nvidia_driver_persistence_mode_on` | `yes`                           | Whether to enable persistence mode (boolean)                                                                          |
-| `nvidia_driver_skip_reboot` | `no` | Whether to skip rebooting the node during the install |
+| `nvidia_driver_skip_reboot`         | `no`                            | Whether to skip rebooting the node during the install                                                                 |
-| `nvidia_driver_module_file` | `"/etc/modprobe.d/nvidia.conf"` | Filename to use for NVIDIA driver parameters |
+| `nvidia_driver_module_file`         | `"/etc/modprobe.d/nvidia.conf"` | Filename to use for NVIDIA driver parameters                                                                          |
-| `nvidia_driver_module_params` | `""` | Parameters to pass to the NVIDIA driver |
+| `nvidia_driver_module_params`       | `""`                            | Parameters to pass to the NVIDIA driver                                                                               |
-| `nvidia_driver_branch` | `"510"` | Default driver branch to install |
+| `nvidia_driver_branch`              | `"515"`                         | Default driver branch to install                                                                                      |
 
 ### Red Hat specific variables
 
 
-| Variable | Default value | Description |
+| Variable                               | Default value                                                                                                     | Description                       |
-| -------- | ------------- | ----------- |
+|----------------------------------------|-------------------------------------------------------------------------------------------------------------------|-----------------------------------|
-| `epel_package` | `"https://dl.fedoraproject.org/pub/epel/epel-release-latest-{{ ansible_distribution_major_version }}.noarch.rpm"` | Package to install to enable EPEL |
+| `epel_package`                         | `"https://dl.fedoraproject.org/pub/epel/epel-release-latest-{{ ansible_distribution_major_version }}.noarch.rpm"` | Package to install to enable EPEL |
-| `nvidia_driver_rhel_cuda_repo_baseurl` | `"https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/"` | Base URL to use for CUDA repo |
+| `nvidia_driver_rhel_cuda_repo_baseurl` | `"https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/"`                                | Base URL to use for CUDA repo     |
-| `nvidia_driver_rhel_cuda_repo_gpgkey` | `"https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/7fa2af80.pub"` | GPG key for the CUDA repo |
+| `nvidia_driver_rhel_cuda_repo_gpgkey`  | `"https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/7fa2af80.pub"`                    | GPG key for the CUDA repo         |
 
 ### Ubuntu specific variables
 
@@ -48,13 +48,12 @@ For Ubuntu installs, you have the choice of installing from the Canonical reposi
 
 By default, the Canonical repositories will be used, and the driver installed will be the headless server driver.
 
-| Variable | Default value | Description |
+| Variable                                      | Default value                                                                      | Description                                          |
-| -------- | ------------- | ----------- |
+|-----------------------------------------------|------------------------------------------------------------------------------------|------------------------------------------------------|
-| `nvidia_driver_ubuntu_install_from_cuda_repo` | `no` | Flag whether to use the CUDA repo |
+| `nvidia_driver_ubuntu_install_from_cuda_repo` | `no`                                                                               | Flag whether to use the CUDA repo                    |
-| `nvidia_driver_ubuntu_cuda_repo_baseurl` | `"http://developer.download.nvidia.com/compute/cuda/repos/{{ _ubuntu_repo_dir }}"` | Base URL to use for CUDA repo |
+| `nvidia_driver_ubuntu_cuda_repo_baseurl`      | `"http://developer.download.nvidia.com/compute/cuda/repos/{{ _ubuntu_repo_dir }}"` | Base URL to use for CUDA repo                        |
-| `nvidia_driver_ubuntu_cuda_repo_gpgkey_url` | `"https://developer.download.nvidia.com/compute/cuda/repos/{{ _ubuntu_repo_dir }}/7fa2af80.pub"` | GPG key for the CUDA repo |
+| `nvidia_driver_ubuntu_cuda_package`           | `"cuda-drivers"`                                                                   | Package name to install from CUDA repo               |
-| `nvidia_driver_ubuntu_cuda_repo_gpgkey_id` | `"7fa2af80"` | GPG key ID for the CUDA repo |
+| `nvidia_driver_ubuntu_packages_suffix`        | `"-server"`                                                                        | The suffix added to the apt packages when installing |
-| `nvidia_driver_ubuntu_cuda_package` | `"cuda-drivers"` | Package name to install from CUDA repo |
 
 ## Example playbook
 

defaults/main.yml

@@ -5,7 +5,7 @@ nvidia_driver_skip_reboot: no
 nvidia_driver_module_file: /etc/modprobe.d/nvidia.conf
 nvidia_driver_module_params: ''
 nvidia_driver_add_repos: yes
-nvidia_driver_branch: "510"
+nvidia_driver_branch: "515"
 
 
 ##############################################################################
@@ -14,11 +14,38 @@ nvidia_driver_branch: "510"
 epel_package: "https://dl.fedoraproject.org/pub/epel/epel-release-latest-{{ ansible_distribution_major_version }}.noarch.rpm"
 epel_repo_key: "https://dl.fedoraproject.org/pub/epel/RPM-GPG-KEY-EPEL-{{ ansible_distribution_major_version }}"
 nvidia_driver_rhel_cuda_repo_baseurl: "https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/"
-nvidia_driver_rhel_cuda_repo_gpgkey: "https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/7fa2af80.pub"
+nvidia_driver_rhel_cuda_repo_gpgkey: "https://developer.download.nvidia.com/compute/cuda/repos/{{ _rhel_repo_dir }}/D42D0685.pub"
 
 nvidia_driver_rhel_branch: "{{ nvidia_driver_branch }}"
 
 
+##############################################################################
+# Debian
+##############################################################################
+
+# Driver branch to install with Debian
+nvidia_driver_debian_branch: "{{ nvidia_driver_branch }}"
+
+# Determine if we should install from CUDA repo instead of Canonical repos
+nvidia_driver_debian_install_from_cuda_repo: no
+
+# Installing with Debian repositories
+nvidia_driver_debian_packages:
+- "nvidia-driver"
+- "nvidia-cuda-dev"
+- "nvidia-cuda-toolkit"
+
+nvidia_driver_debian_install_tesla_driver: no
+nvidia_driver_debian_tesla_package: "nvidia-tesla-470-driver"
+
+# Installing with CUDA repositories
+old_nvidia_driver_debian_cuda_repo_gpgkey_id: "7fa2af80"
+nvidia_driver_debian_cuda_repo_baseurl: "https://developer.download.nvidia.com/compute/cuda/repos/{{ _debian_repo_dir }}"
+nvidia_driver_debian_cuda_keyring_package: "cuda-keyring_1.1-1_all.deb"
+nvidia_driver_debian_cuda_keyring_url: "{{ nvidia_driver_debian_cuda_repo_baseurl }}/{{ nvidia_driver_debian_cuda_keyring_package }}"
+nvidia_driver_debian_cuda_package: "cuda-drivers-{{ nvidia_driver_debian_branch }}"
+
+
 ##############################################################################
 # Ubuntu                                                                     #
 ##############################################################################
@@ -29,15 +56,18 @@ nvidia_driver_ubuntu_branch: "{{ nvidia_driver_branch }}"
 # Determine if we should install from CUDA repo instead of Canonical repos
 nvidia_driver_ubuntu_install_from_cuda_repo: no
 
+nvidia_driver_ubuntu_packages_suffix: "-server"
+
 # Installing with Canonical repositories
 nvidia_driver_ubuntu_packages:
-- "nvidia-headless-{{ nvidia_driver_ubuntu_branch }}-server"
+- "nvidia-headless-{{ nvidia_driver_ubuntu_branch }}{{ nvidia_driver_ubuntu_packages_suffix }}"
-- "nvidia-utils-{{ nvidia_driver_ubuntu_branch }}-server"
+- "nvidia-utils-{{ nvidia_driver_ubuntu_branch }}{{ nvidia_driver_ubuntu_packages_suffix }}"
-- "nvidia-headless-no-dkms-{{ nvidia_driver_ubuntu_branch }}-server"
+- "nvidia-headless-no-dkms-{{ nvidia_driver_ubuntu_branch }}{{ nvidia_driver_ubuntu_packages_suffix }}"
-- "nvidia-kernel-source-{{ nvidia_driver_ubuntu_branch }}-server"
+- "nvidia-kernel-source-{{ nvidia_driver_ubuntu_branch }}{{ nvidia_driver_ubuntu_packages_suffix }}"
 
 # Installing with CUDA repositories
-nvidia_driver_ubuntu_cuda_repo_gpgkey_url: "https://developer.download.nvidia.com/compute/cuda/repos/{{ _ubuntu_repo_dir }}/7fa2af80.pub"
+old_nvidia_driver_ubuntu_cuda_repo_gpgkey_id: "7fa2af80"
-nvidia_driver_ubuntu_cuda_repo_gpgkey_id: "7fa2af80"
 nvidia_driver_ubuntu_cuda_repo_baseurl: "https://developer.download.nvidia.com/compute/cuda/repos/{{ _ubuntu_repo_dir }}"
+nvidia_driver_ubuntu_cuda_keyring_package: "cuda-keyring_1.0-1_all.deb"
+nvidia_driver_ubuntu_cuda_keyring_url: "{{ nvidia_driver_ubuntu_cuda_repo_baseurl }}/{{ nvidia_driver_ubuntu_cuda_keyring_package }}"
 nvidia_driver_ubuntu_cuda_package: "cuda-drivers-{{ nvidia_driver_ubuntu_branch }}"

files/cuda-ubuntu.pin

@@ -1,3 +0,0 @@
-Package: *
-Pin: release l=NVIDIA CUDA
-Pin-Priority: 600

molecule/default/molecule.yml

@@ -27,7 +27,7 @@ platforms:
     - cuda_repo
     - ubuntu
 
-  - name: ubuntu-2004-canonical
+  - name: ubuntu-2004-canonical-server
     image: geerlingguy/docker-ubuntu2004-ansible
     volumes:
       - /sys/fs/cgroup:/sys/fs/cgroup:ro
@@ -38,6 +38,18 @@ platforms:
     - canonical_repo
     - ubuntu
 
+  - name: ubuntu-2004-canonical-noserver
+    image: geerlingguy/docker-ubuntu2004-ansible
+    volumes:
+      - /sys/fs/cgroup:/sys/fs/cgroup:ro
+    command: /sbin/init
+    pre_build_image: true
+    privileged: true
+    groups:
+    - canonical_repo
+    - canonical_repo_noserver
+    - ubuntu
+
   - name: ubuntu-2004-cuda
     image: geerlingguy/docker-ubuntu2004-ansible
     volumes:
@@ -49,6 +61,28 @@ platforms:
     - cuda_repo
     - ubuntu
 
+  - name: ubuntu-2204-canonical
+    image: geerlingguy/docker-ubuntu2204-ansible
+    volumes:
+      - /sys/fs/cgroup:/sys/fs/cgroup:ro
+    command: /sbin/init
+    pre_build_image: true
+    privileged: true
+    groups:
+    - canonical_repo
+    - ubuntu
+
+  - name: ubuntu-2204-cuda
+    image: geerlingguy/docker-ubuntu2204-ansible
+    volumes:
+      - /sys/fs/cgroup:/sys/fs/cgroup:ro
+    command: /sbin/init
+    pre_build_image: true
+    privileged: true
+    groups:
+    - cuda_repo
+    - ubuntu
+
   - name: centos-7
     image: geerlingguy/docker-centos7-ansible
     volumes:
@@ -57,13 +91,13 @@ platforms:
     pre_build_image: true
     privileged: true
 
-  - name: centos-8
+#  - name: centos-8
-    image: geerlingguy/docker-centos8-ansible
+#    image: geerlingguy/docker-centos8-ansible
-    volumes:
+#    volumes:
-      - /sys/fs/cgroup:/sys/fs/cgroup:ro
+#      - /sys/fs/cgroup:/sys/fs/cgroup:ro
-    command: /sbin/init
+#    command: /sbin/init
-    pre_build_image: true
+#    pre_build_image: true
-    privileged: true
+#    privileged: true
 
 provisioner:
   name: ansible
@@ -75,6 +109,8 @@ provisioner:
         nvidia_driver_skip_reboot: true
       canonical_repo:
         nvidia_driver_ubuntu_install_from_cuda_repo: false
+      canonical_repo_noserver:
+        nvidia_driver_ubuntu_packages_suffix: ""
       cuda_repo:
         nvidia_driver_ubuntu_install_from_cuda_repo: true
 verifier:

tasks/install-debian-cuda-repo.yml

@@ -0,0 +1,50 @@
+- name: add contrib & non-free repository
+  replace:
+    dest: /etc/apt/sources.list
+    regexp: '^(deb(?!.* contrib).*)'
+    replace: '\1 contrib non-free'
+
+- name: remove old signing key
+  apt_key:
+    id: "{{ old_nvidia_driver_debian_cuda_repo_gpgkey_id }}"
+    state: absent
+  environment: "{{proxy_env if proxy_env is defined else {}}}"
+  when: nvidia_driver_add_repos | bool
+
+- name: add CUDA keyring
+  apt:
+    deb: "{{ nvidia_driver_debian_cuda_keyring_url }}"
+    state: "present"
+  environment: "{{proxy_env if proxy_env is defined else {}}}"
+  when: nvidia_driver_add_repos | bool
+
+- name: force an apt update
+  apt:
+    update_cache: true
+  changed_when: false
+
+- name: ensure kmod is installed
+  apt:
+    name: "kmod"
+    state: "present"
+
+- name: blacklist nouveau
+  kernel_blacklist:
+    name: nouveau
+    state: present
+
+- name: install Linux headers and non-free firmware
+  apt:
+    name: 
+      - linux-headers-{{ ansible_kernel }}
+      - firmware-misc-nonfree
+    state: present
+
+- name: install driver packages
+  apt:
+    name: "{{ nvidia_driver_package_version | ternary(nvidia_driver_debian_cuda_package+'='+nvidia_driver_package_version, nvidia_driver_debian_cuda_package) }}"
+    state: "{{ nvidia_driver_package_state }}"
+    autoremove: "{{ nvidia_driver_package_state == 'absent' }}"
+    purge: "{{ nvidia_driver_package_state == 'absent' }}"
+  register: install_driver
+  environment: "{{proxy_env if proxy_env is defined else {}}}"

tasks/install-debian.yml

@@ -0,0 +1,35 @@
+---
+- name: add contrib & non-free repository
+  replace:
+    dest: /etc/apt/sources.list
+    regexp: '^(deb(?!.* contrib).*)'
+    replace: '\1 contrib non-free'
+
+- name: update apt
+  become: yes
+  apt:
+    update_cache: yes
+
+- name: install Linux headers and non-free firmware
+  apt:
+    name: 
+      - linux-headers-{{ ansible_kernel }}
+      - firmware-misc-nonfree
+    state: present
+
+- name: install driver packages
+  apt:
+    name: "{{ nvidia_driver_package_version | ternary(item+'='+nvidia_driver_package_version, item) }}"
+    state: "{{ nvidia_driver_package_state }}"
+    autoremove: "{{ nvidia_driver_package_state == 'absent' }}"
+    purge: "{{ nvidia_driver_package_state == 'absent' }}"
+  loop: "{{ nvidia_driver_debian_packages }}"
+  register: install_driver
+  environment: "{{proxy_env if proxy_env is defined else {}}}"
+
+- name: install tesla drivers
+  apt:
+    state: present
+    name: "{{ nvidia_driver_debian_tesla_package }}"
+  register: install_driver
+  when: nvidia_driver_debian_install_tesla_driver == true

tasks/install-ubuntu-cuda-repo.yml

@@ -4,34 +4,35 @@
     repo: ppa:graphics-drivers/ppa
     state: absent
 
-- name: add pin file
+- name: remove old signing key
-  copy:
-    src: "cuda-ubuntu.pin"
-    dest: "/etc/apt/preferences.d/cuda-repository-pin-600"
-    owner: "root"
-    group: "root"
-    mode: "0644"
-  when: nvidia_driver_add_repos | bool
-
-- name: add key
   apt_key:
-    url: "{{ nvidia_driver_ubuntu_cuda_repo_gpgkey_url }}"
+    id: "{{ old_nvidia_driver_ubuntu_cuda_repo_gpgkey_id }}"
-    id: "{{ nvidia_driver_ubuntu_cuda_repo_gpgkey_id }}"
+    state: absent
   environment: "{{proxy_env if proxy_env is defined else {}}}"
   when: nvidia_driver_add_repos | bool
 
+- name: add CUDA keyring
+  apt:
+    deb: "{{ nvidia_driver_ubuntu_cuda_keyring_url }}"
+    state: "present"
+  environment: "{{proxy_env if proxy_env is defined else {}}}"
+  when: nvidia_driver_add_repos | bool
+
+- name: force an apt update
+  apt:
+    update_cache: true
+  changed_when: false
+
+- name: ensure kmod is installed
+  apt:
+    name: "kmod"
+    state: "present"
+
 - name: blacklist nouveau
   kernel_blacklist:
     name: nouveau
     state: present
 
-- name: add repo
-  apt_repository:
-    repo: "deb {{ nvidia_driver_ubuntu_cuda_repo_baseurl }} /"
-    update_cache: yes
-  environment: "{{proxy_env if proxy_env is defined else {}}}"
-  when: nvidia_driver_add_repos | bool
-
 - name: install driver packages
   apt:
     name: "{{ nvidia_driver_package_version | ternary(nvidia_driver_ubuntu_cuda_package+'='+nvidia_driver_package_version, nvidia_driver_ubuntu_cuda_package) }}"

tasks/install-ubuntu.yml

@@ -10,6 +10,6 @@
     state: "{{ nvidia_driver_package_state }}"
     autoremove: "{{ nvidia_driver_package_state == 'absent' }}"
     purge: "{{ nvidia_driver_package_state == 'absent' }}"
-  with_items: "{{ nvidia_driver_ubuntu_packages }}"
+  loop: "{{ nvidia_driver_ubuntu_packages }}"
   register: install_driver
   environment: "{{proxy_env if proxy_env is defined else {}}}"

tasks/main.yml

@@ -5,6 +5,14 @@
     state: absent
   ignore_errors: true
 
+- name: debian install tasks (debian repos)
+  include_tasks: install-debian.yml
+  when: ansible_distribution == "Debian" and (not nvidia_driver_debian_install_from_cuda_repo)
+
+- name: debian install tasks (CUDA repos)
+  include_tasks: install-debian.yml
+  when: ansible_distribution == "Debian" and nvidia_driver_debian_install_from_cuda_repo
+
 - name: ubuntu install tasks (canonical repos)
   include_tasks: install-ubuntu.yml
   when: ansible_distribution == 'Ubuntu' and (not nvidia_driver_ubuntu_install_from_cuda_repo)

vars/main.yml

@@ -1,2 +1,3 @@
+_debian_repo_dir: "{{ ansible_distribution | lower }}{{ ansible_distribution_version | replace('.', '') }}/{{ ansible_architecture }}"
 _ubuntu_repo_dir: "{{ ansible_distribution | lower }}{{ ansible_distribution_version | replace('.', '') }}/{{ ansible_architecture }}"
 _rhel_repo_dir: "rhel{{ ansible_distribution_major_version }}/{{ ansible_architecture }}"