OS Manuals

nvidia-installer(1)

nvidia-installer - install, upgrade, or uninstall the NVIDIA Accelerated Graphics Driver Set

nvidia-installer(1) General Commands Manual nvidia-installer(1)

nvidia-installer - install, upgrade, or uninstall the NVIDIA Accelerated Graphics Driver Set

nvidia-installer [ options ]

nvidia-installer is a tool for installing, updating, and uninstalling the NVIDIA Linux Graphics Driver. When the driver is installed by running, for example:

sh NVIDIA-Linux-x86_64-580.178.04.run

The .run file unpacks itself and invokes the contained nvidia-installer utility. nvidia-installer then walks you through the installation process. As part of installation, nvidia-installer is installed on the system for later use, such as for uninstalling the currently installed driver or updating to newer drivers.

In the default 'install' mode of operation, nvidia-installer probes your system to determine which driver files should be installed and where. nvidia-installer then determines which files already present on your system will conflict with driver installation; those files are noted and will be later backed up. At this point, nothing will have been changed on the file system, though a "command list" will have been generated. If running in expert mode (requested with the --expert commandline option), this command list will be presented so that you can review all operations that nvidia-installer intends to perform. Finally, nvidia-installer executes the commandlist, backing up conflicting files already present on the file system, installing the new driver files on the file system, creating symlinks, and running system utilities like depmod(8) and ldconfig(8).

To later uninstall the NVIDIA Linux graphics driver, you can run nvidia-installer --uninstall. In the 'uninstall' mode of operation, all driver files that were installed on the system are deleted, and all files that were backed up during installation are restored to their original locations. The uninstall process should restore your filesystem to its state prior to installation. If you install one NVIDIA Linux graphics driver while another is already installed, this uninstall step is automatically performed on the old driver at the beginning of installation of the new driver.

nvidia-installer's backend is separate from its user interface; the installer will use an ncurses-based user interface if it can find the correct ncurses library, otherwise, it will fall back to a simple commandline user interface. To disable use of the ncurses user interface, use the option --ui=none. Additional user interfaces, utilizing GTK+ or QT, for example, could be provided in the future.

The source code to nvidia-installer is released under the GPL and available here:


⟨URL: https://download.nvidia.com/XFree86/nvidia-installer/ ⟩

Patches are welcome.

Print the nvidia-installer version and exit.
Print usage information for the common commandline options and exit.
Print usage information for the common commandline options as well as the advanced options, and then exit.

This option is obsolete and ignored by nvidia-installer. It is provided for compatibility with older versions of nvidia-installer, which required this option for explicit license acceptance. Use of the NVIDIA driver implies acceptance of the NVIDIA Software License Agreement, contained in the file 'LICENSE' (in the top level directory of the driver package).
Print information about the currently installed NVIDIA driver version.
Uninstall the currently installed NVIDIA driver.
When uninstalling the driver, skip unloading of the NVIDIA kernel module. This option is ignored when the driver is being installed.
Skip the test load of the NVIDIA kernel modules after the modules are built, and skip loading them after installation is complete.
Perform basic sanity tests on an existing NVIDIA driver installation.
Enable 'expert' installation mode; more detailed questions will be asked, and more verbose output will be printed; intended for expert users. The questions may be suppressed with the '--no-questions' commandline option.
Do not ask any questions; the default (normally 'yes') is assumed for all yes/no questions, and the default string is assumed in any situation where the user is prompted for string input.
Run silently; no questions are asked and no output is printed, except for error messages to stderr. This option implies '--ui=none --no-questions'.
The prefix under which the X components of the NVIDIA driver will be installed; the default is '/usr/X11R6' unless nvidia-installer detects that X.Org >= 7.0 is installed, in which case the default is '/usr'. Only under rare circumstances should this option be used.
This is a deprecated synonym for --x-prefix.
The path under which the NVIDIA X server modules will be installed. If this option is not specified, nvidia-installer uses the following search order and selects the first valid directory it finds: 1) `X -showDefaultModulePath`, 2) `pkg-config --variable=moduledir xorg-server`, or 3) the X library path (see the '--x-library-path' option) plus either 'modules' (for X servers older than X.Org 7.0) or 'xorg/modules' (for X.Org 7.0 or later).
The path under which the NVIDIA X libraries will be installed. If this option is not specified, nvidia-installer uses the following search order and selects the first valid directory it finds: 1) `X -showDefaultLibPath`, 2) `pkg-config --variable=libdir xorg-server`, or 3) the X prefix (see the '--x-prefix' option) plus 'lib' on 32bit systems, and either 'lib64' or 'lib' on 64bit systems, depending on the installed Linux distribution.
The path under which X system configuration files will be installed. If this option is not specified, nvidia-installer uses the following search order and selects the first valid directory it finds: 1) `pkg-config --variable=sysconfigdir xorg-server`, or 2) /usr/share/X11/xorg.conf.d.
The prefix under which the OpenGL components of the NVIDIA driver will be installed; the default is: '/usr'. Only under rare circumstances should this option be used. The Linux OpenGL ABI (http://oss.sgi.com/projects/ogl-sample/ABI/) mandates this default value.
The path relative to the OpenGL library installation prefix under which the NVIDIA OpenGL components will be installed. The default is 'lib' on 32bit systems, and 'lib64' or 'lib' on 64bit systems, depending on the installed Linux distribution. Only under very rare circumstances should this option be used.
The prefix under which Wine components of the NVIDIA driver will be installed. The default is '/usr'.
The path relative to the Wine installation prefix under which Wine components of the NVIDIA driver will be installed. The default is '<opengl-libdir>/nvidia/wine', where <opengl-libdir> is the path determined for --opengl-libdir above.
The prefix under which the installer binary will be installed; the default is: '/usr'. Note: please use the '--utility-prefix' option instead.
The prefix under which the NVIDIA utilities (nvidia-installer, nvidia-settings, nvidia-xconfig, nvidia-bug-report.sh) and the NVIDIA utility libraries will be installed; the default is: '/usr'.
The path relative to the utility installation prefix under which the NVIDIA utility libraries will be installed. The default is 'lib' on 32bit systems, and 'lib64' or 'lib' on 64bit systems, depending on the installed Linux distribution.
The prefix under which XDG data files (such as application .desktop files and icons) will be installed. The default is '/usr/share'.
The path relative to the OpenGL library installation prefix under which the NVIDIA GBM backend library will be installed. The default is '<libdir>/gbm', where '<libdir>' is 'lib64' or 'lib', depending on the installed Linux distribution. Only under very rare circumstances should this option be used.
The path relative to the 32-bit OpenGL library installation prefix under which the 32-bit NVIDIA GBM backend library will be installed. The default is '<libdir>/gbm', where <libdir>' is 'lib' or 'lib32', depending on the installed Linux distribution. Only under very rare circumstances should this option be used.
The prefix under which the documentation files for the NVIDIA driver will be installed. The default is: '/usr'.
The directory under which default application profiles for the NVIDIA driver will be installed. The default is: '/usr/share/nvidia'.
The directory containing the kernel include files that should be used when compiling the NVIDIA kernel module. This option is deprecated; please use '--kernel-source-path' instead.
The directory containing the kernel source files that should be used when compiling the NVIDIA kernel module. When not specified, the installer will use '/lib/modules/`uname -r`/build', if that directory exists. Otherwise, it will use '/usr/src/linux'.
The directory containing any KBUILD output files if either one of the 'KBUILD_OUTPUT' or 'O' parameters were passed to KBUILD when building the kernel image/modules. When not specified, the installer will assume that no separate output directory was used.
The directory in which the NVIDIA kernel module should be installed. The default value is either '/lib/modules/`uname -r`/kernel/drivers/video' (if '/lib/modules/`uname -r`/kernel' exists) or '/lib/modules/`uname -r`/video'.
The mount point for the proc file system; if not specified, then this value defaults to '/proc' (which is normally correct). The mount point of the proc filesystem is needed because the contents of '<proc filesystem>/version' is used when identifying if a precompiled kernel interface is available for the currently running kernel. This option should only be needed in very rare circumstances.
File name of the installation log file (the default is: '/var/log/nvidia-installer.log').
Use the specified directory as a temporary directory when generating transient files used by the installer; if not given, then the following list will be searched, and the first one that exists will be used: $TMPDIR, /tmp, ., $HOME.
Specify what user interface to use, if available. Valid values for UI are 'ncurses' (the default) or 'none'. If the ncurses interface fails to initialize, or 'none' is specified, then a simple printf/scanf interface will be used.
Disable use of color in the ncurses user interface.
Skip installation of 'nvidia-modprobe', a setuid root utility which nvidia-installer installs by default. nvidia-modprobe can be used by user-space NVIDIA driver components to load the NVIDIA kernel module, create the NVIDIA device files, and configure certain runtime settings in the kernel when those components run without sufficient privileges to do so on their own, e.g., the CUDA driver run within the permissions of a non-privileged user. This utility is only needed if other means of performing these privileged operations are unavailable. See the nvidia-modprobe(1) man page for full details on the tasks performed by this utility.
Build and install the NVIDIA kernel module for the non-running kernel specified by KERNEL-NAME ( KERNEL-NAME should be the output of `uname -r` when the target kernel is actually running). This option implies '--no-precompiled-interface'. If the options '--kernel-install-path' and '--kernel-source-path' are not given, then they will be inferred from KERNEL-NAME ; eg: '/lib/modules/ KERNEL-NAME /kernel/drivers/video/' and '/lib/modules/ KERNEL-NAME /build/', respectively.
Disable use of precompiled kernel interfaces.
The NVIDIA OpenGL libraries contain an OS ABI note tag, which identifies the minimum kernel version needed to use the library. This option causes the installer to remove this note from the OpenGL libraries during installation.
Normally, the installer will check for several rpms that conflict with the driver (specifically: NVIDIA_GLX and NVIDIA_kernel), and remove them if present. This option disables this check.
During driver installation, conflicting files are backed up, so that they can be restored when the driver is uninstalled. This option causes the installer to simply delete conflicting files, rather than back them up.
Normally, nvidia-installer will recursively search for potentially conflicting libraries under the default OpenGL and X server installation locations. With this option set, the installer will only search in the top-level directories.
Install the kernel modules only, and do not uninstall the existing driver. This is intended to be used to install kernel modules for additional kernels (in cases where you might boot between several different kernels). To use this option, you must already have a driver installed, and the version of the installed driver must match the version of these kernel modules.
Install everything but the kernel modules, and do not remove any existing, possibly conflicting, kernel modules. This can be useful in some debug environments, or when installing the open kernel modules from source. If you use this option, you must be careful to ensure that NVIDIA kernel modules matching this driver version are installed separately.
Do not abort the installation if nvidia-installer detects that an X server is running. Only under very rare circumstances should this option be used.
Before searching for a precompiled kernel interface in the .run file, search in the specified directory.
Normally, nvidia-installer aborts installation if the nouveau kernel driver is in use. Use this option to disable this check.
nvidia-installer will attempt to disable the nouveau kernel driver by default, if it is in use during installation. Use '--no-disable-nouveau to prevent nvidia-installer from disabling nouveau by default.
nvidia-installer can optionally invoke the nvidia-xconfig utility. This will update the system X configuration file so that the NVIDIA X driver is used. The pre-existing X configuration file will be backed up. At the end of installation, nvidia-installer will ask the user if they wish to run nvidia-xconfig; the default response is 'no'. Use this option to make the default response 'yes'. This is useful with the '--no-questions' or '--silent' options, which assume the default values for all questions.
Linux installations using SELinux (Security-Enhanced Linux) require that the security type of all shared libraries be set to 'shlib_t' or 'textrel_shlib_t', depending on the distribution. nvidia-installer will detect when to set the security type, and set it using chcon(1) on the shared libraries it installs. If the execstack(8) system utility is present, nvidia-installer will use it to also clear the executable stack flag of the libraries. Use this option to override nvidia-installer's detection of when to set the security type. Valid values for FORCE-SELINUX are 'yes' (force setting of the security type), 'no' (prevent setting of the security type), and 'default' (let nvidia-installer decide when to set the security type).
When SELinux support is enabled, nvidia-installer will try to determine which chcon argument to use by first trying 'textrel_shlib_t', then 'texrel_shlib_t', then 'shlib_t'. Use this option to override this detection logic.
Normally, nvidia-installer ignores the SIGWINCH signal before it forks to execute commands, e.g. to build the kernel module, and restores the SIGWINCH signal handler after the child process has terminated. This option disables this behavior.
Normally, nvidia-installer will run scripts from /usr/lib/nvidia before and after installing or uninstalling the driver. Use this option to disable execution of these scripts.
Do not install any of the OpenGL-related driver files.
Do not install any of the Wine driver files.
Specify a path where the source directory for the kernel module will be installed. Default: install source directory at /usr/src
Specify the name of the directory where the kernel module sources will be installed. Default: directory name is "nvidia-VERSION"
Skip installation of the kernel module source.
nvidia-installer can optionally register the NVIDIA kernel module sources, if installed, with DKMS, then build and install a kernel module using the DKMS-registered sources. This will allow the DKMS infrastructure to automatically build a new kernel module when changing kernels. During installation, if DKMS is detected, nvidia-installer will ask the user if they wish to register the module with DKMS; the default response is 'yes' unless the --no-dkms option is set, in which case the default response is 'no'.
Specify a path to a private key to use for signing the NVIDIA kernel module. The corresponding public key must also be provided.
Specify a path to a public key to use for verifying the signature of the NVIDIA kernel module. The corresponding private key must also be provided.
Specify a path to a program to use for signing the NVIDIA kernel module. The program will be called with the arguments: program-name <HASH> <PRIVATEKEY> <PUBLICKEY> <MODULE>; if the program returns an error status, it will be called again with the arguments: program-name <PRIVATEKEY> <PUBLICKEY> <MODULE>. Default: use the "sign-file" script in the kernel source directory.
Specify a path where signing keys generated by nvidia-installer will be installed. Default: install keys to '/usr/share/nvidia'.
Specify a cryptographic hash algorithm to use for signing kernel modules. This requires a module signing tool that allows explicit selection of the hash algorithm, and the hash algorithm name must be recognizable by the module signing tool. Default: select a hash algorithm automatically, based on the kernel's configuration.
Specify a cryptographic hash algorithm to use for signing X.509 certificates generated by nvidia-installer. The hash algorithm name must be one of the message digest algorithms recognized by the x509(1) command.
Maintainers of alternate driver installation methods can report the presence and/or availability of an alternate driver installation to nvidia-installer. Setting this option skips the check for alternate driver installations.
Do not install the NVIDIA Unified Memory kernel module. This kernel module is required for CUDA on 64-bit systems, and if it is not installed, the CUDA driver and CUDA applications will not be able to run. The '--no-unified-memory' option should only be used to work around failures to build or install the Unified Memory kernel module on systems that do not need to run CUDA.
Do not install the nvidia-drm kernel module. This kernel module provides several features, including X11 autoconfiguration, support for PRIME, and DRM-KMS. The latter is used to support modesetting on windowing systems that run independently of X11. The '--no-drm' option should only be used to work around failures to build or install the nvidia-drm kernel module on systems that do not need these features.
Do not install the nvidia-peermem kernel module. This kernel module provides support for peer-to-peer memory sharing with Mellanox HCAs (Host Channel Adapters) via GPUDirect RDMA (Remote Direct Memory Access).
Set the concurrency level for operations such as building the kernel module which may be parallelized on SMP systems. By default, this will be set to the number of detected CPUs, or to '1', if nvidia-installer fails to detect the number of CPUs. Systems with a large number of CPUs will have the default concurrency level limited to 32; setting a higher level on the command line will override this limit.
Always install a libGLX_indirect.so.0 symlink, overwriting one if it exists.
Do not install a libGLX_indirect.so.0 symlink.
The installer will try to determine whether the libglvnd libraries are already available, and will install them if they're not. Use --install-libglvnd to always install the libglvnd libraries, overwriting any that already exist. Use --no-install-libglvnd to exclude the libglvnd libraries, even if they appear to be missing.
Install the EGL vendor library config file to this directory. If the libglvnd libraries are already present, then by default the installer will try to determine the path by running `pkg-config --variable=datadir libglvnd`. If that fails, then it will default to /usr/share/glvnd/egl_vendor.d.
If the package includes an EGL external platform library, then install the EGL external platform library config file to this directory. Defaults to /usr/share/egl/egl_external_platform.d.
Override the default destination for a file type. This option takes an argument in the form of '<FILE_TYPE>:<destination>', where <FILE_TYPE> is a file type from the installer .manifest file, and <destination> is an absolute path to the directory where files of that type should be installed. This option may be given multiple times in order to override the destinations for multiple file types. Use of this option takes precedence over any other options that might otherwise influence the destination of the specified file type.
Don't run the depmod(1) utility after modifying kernel modules. This should only be used in cases where depmod(1) will be run separately after running nvidia-installer.
By default, the installer will install systemd unit files if systemctl is detected. Specifying --no-systemd will disable installation of systemd units.
The path to which systemd unit files should be installed. By default, the installer uses `pkg-config --variable=systemdsystemunitdir systemd` to determine the location to install systemd unit files, or '/usr/lib/systemd/system' if pkg-config is not available. Ignored if --no-systemd is specified.
The path to which systemd-sleep script files should be installed. By default, the installer uses `pkg-config --variable=systemdsleepdir systemd` to determine the location to install systemd-sleep script files, or '/usr/lib/systemd/system-sleep' if pkg-config is not available. Ignored if --no-systemd is specified.
The path to which systemd unit enable symlinks should be installed. By default, the installer uses `pkg-config --variable=systemdsystemconfdir systemd` to determine the location to install these symlinks, or '/etc/systemd/system' if pkg-config is not available. Ignored if --no-systemd is specified.
The type of kernel modules to build and install. Valid values are "open" and "proprietary".
Directly set the directory within the package from which to build the kernel modules. This option is deprecated; use "--kernel-module-type" instead.
Print if "open" or "proprietary" modules should be used based on currently attached GPUs and exit. NOTE: A kernel module type may be recommended even if it is not included in this driver package.
Proceed with installation even if an NVIDIA driver is already installed and running.
Rebuild the initramfs after installation is complete, regardless of the default action recommended by nvidia-installer. --no-rebuild-initramfs skips rebuilding the initramfs after installation is complete. These options are useful for non-interactive installations when a specific behavior is desired, regardless of what nvidia-installer would recommend by default in an interactive installation.

Because the NVIDIA installer may interact badly with distribution packages that contain the NVIDIA driver, nvidia-installer provides a mechanism for the distribution to handle manual installation of the driver. If they exist, nvidia-installer will run the following scripts:

• /usr/lib/nvidia/pre-install
• /usr/lib/nvidia/pre-uninstall
• /usr/lib/nvidia/post-uninstall
• /usr/lib/nvidia/post-install
• /usr/lib/nvidia/failed-install
Note that if installation of a new driver requires uninstallation of a previously installed driver, the pre- and post-uninstall scripts will be called after the pre-install script. If the install fails, the installer will execute /usr/lib/nvidia/failed-install instead of /usr/lib/nvidia/post-install. These scripts should not require user interaction.

Use the --no-distro-scripts option to disable execution of these scripts.

In addition to the distribution hook scripts, distributors or other producers of driver packages may report the presence an already installed driver, or the availability of driver packages. Installing a text file to /usr/lib/nvidia/alternate-install-present will alert nvidia-installer that an existing driver is already installed. Installing a text file to /usr/lib/nvidia/alternate-install-available will alert nvidia-installer that an alternate installation option is available. The contents of /usr/lib/nvidia/alternate-install-present or /usr/lib/nvidia/alternate-install-available will be printed in a message informing the user of the presence/availability of the alternate driver installation, and asking the user whether to continue with the installation.

Remove the NVIDIA driver and restore files that were overwritten during the install process.

Aaron Plattner
NVIDIA Corporation

nvidia-xconfig(1), nvidia-settings(1), /usr/share/doc/NVIDIA_GLX-1.0/README.txt

Copyright © 2005-2018 NVIDIA Corporation.

2026-07-07 nvidia-installer 580.178.04

nvidia-tesla-580 580.178.04-r0 · extra-packages · ingested 2026-10-02