Allow disabling nvidia-drm-module, copying runfile from files and ena…

…bling nvidia-persistenced-service

defaults/main.yml

@@ -11,7 +11,9 @@ cuda_use_runfile: False
 cuda_runfile_url: "https://developer.download.nvidia.com/compute/cuda/11.2.2/local_installers/cuda_11.2.2_460.32.03_linux.run"
 cuda_runfile_driver: True
 cuda_runfile_toolkit: True
+cuda_runfile_download: True
 cuda_runfile_remove: True
+cuda_runfile_disable_nvidia_drm: False
 cuda_restart_node_on_install: True
 cuda_init: True
 cuda_init_restart_service: True

files/nvidia-persistenced.service

@@ -0,0 +1,14 @@
+[Unit]
+Description=NVIDIA Persistence Daemon
+After=syslog.target
+
+[Service]
+Type=forking
+PIDFile=/var/run/nvidia-persistenced/nvidia-persistenced.pid
+Restart=always
+ExecStart=/usr/bin/nvidia-persistenced --verbose
+ExecStopPost=/bin/rm -rf /var/run/nvidia-persistenced/*
+TimeoutSec=300
+
+[Install]
+WantedBy=multi-user.target

tasks/install_runfile.yml

@@ -51,14 +51,24 @@
     path: /tmp/cuda_runfile
     state: directory
 
-- name: "Download runfile"
-  get_url:
-    url: "{{ cuda_runfile_url }}"
-    dest: "/tmp/cuda_runfile/{{ cuda_runfile_sh }}"
+- name: "Obtain runfile"
+  block:
+
+  - name: "Copy pre-downloaded runfile"
+    copy:
+      src: "{{ cuda_runfile_sh }}"
+      dest: /tmp/cuda_runfile
+    when: not cuda_runfile_download
+
+  - name: "Download runfile"
+    get_url:
+      url: "{{ cuda_runfile_url }}"
+      dest: "/tmp/cuda_runfile/{{ cuda_runfile_sh }}"
+    when: cuda_runfile_download
+
   when: (cuda_runfile_toolkit and not cuda_toolkit_installed) or
         (cuda_runfile_driver and not cuda_driver_installed)
 
-
 - name: "Run installer for toolkit"
   command: bash /tmp/cuda_runfile/{{ cuda_runfile_sh }} --silent --toolkit
   register: cuda_toolkit_install_out
@@ -85,12 +95,20 @@
       msg: "Building driver {{ cuda_driver_runfile }} for kernel {{ cuda_driver_kernel_version }}"
 
   - name: 'Install driver'
-    command: bash {{ cuda_driver_runfile }} --silent --kernel-name={{ cuda_driver_kernel_version }} --kernel-source-path=/usr/src/kernels/{{ cuda_driver_kernel_version }}
+    command: bash {{ cuda_driver_runfile }} --silent --kernel-name={{ cuda_driver_kernel_version }} --kernel-source-path=/usr/src/kernels/{{ cuda_driver_kernel_version }} {{ '--no-drm' if cuda_runfile_disable_nvidia_drm else '' }}
+
+  when: cuda_runfile_driver and not cuda_driver_installed
+
+  - name: "Install nvidia-persistenced systemd-file"
+    copy:
+      src: files/nvidia-persistenced.service
+      dest: /etc/systemd/system/nvidia-persistenced.service
+    when: cuda_init_persistence_mode | bool
 
   when: cuda_runfile_driver and not cuda_driver_installed
 
-- name: 'Remove installer'
+- name: "Remove installer"
   file:
     path: /tmp/cuda_runfile
     state: absent
-  when: and cuda_runfile_remove
+  when: cuda_runfile_remove