Installing NVIDIA graphics drivers on recent distributions (Fedora, Ubuntu, ...)

[zander]

Please note: Whenever possible, it is recommended that you use your Linux distribution's NVIDIA Linux graphics driver packages. This will ensure better integration with the distribution's native package management system and reduce the likelihood of problems after system updates, etc..


Fedora Core 6, Fdora 7/8/9 & RHEL5

If you wish to install the NVIDIA Linux graphics driver on a Fedora Core 6, Fedora 7/9 or RHEL5 system, please ensure that your system meets the following requirements:

* the latest update kernel is installed and in use
* a kernel-devel RPM is installed that corresponds to the kernel that is in use. Note, FC6 has a known bug which occasionally installs a kernel-devel RPM that is a different arch than the kernel (i586 & i686 mixed). See: https://bugzilla.redhat.com/bugzilla....cgi?id=211941
* the pkgconfig and xorg-x11-server-sdk RPMs are installed (only applies to FC6)
* Xen kernels are not currently supported

The items above can be addressed with the following commands (they need to be run as root):

# yum install kernel-devel xorg-x11-server-sdk pkgconfig
# yum update
# reboot

If you are using an SMP (multi-CPU and/or multi-core) enabled system in FC5(only), please replace kernel-devel with kernel-smp-devel in the command above.

The SELinux requirement can be addressed via two mechanisms after the NVIDIA Linux graphics driver has been installed:

1) If you do not wish to use SELinux enforcement, you can disable it by running the `setenforce 0` command before starting X, or by adding selinux=0 to the end of the kernel line in /etc/grub.conf and rebooting.

2) If you wish to use SELinux enforcement, you will need to change the security context of the NVIDIA X driver module and of the server-side NVIDIA GLX extension module. To achieve this, please run the commands listed for your platform below:

Linux/x86 (32-bit):

# chcon -t texrel_shlib_t /usr/lib/xorg/modules/drivers/nvidia_drv.so
# chcon -t texrel_shlib_t /usr/lib/xorg/modules/extensions/libglx.so.1.0.9631
# chcon -t texrel_shlib_t /usr/lib/tls/libnvidia-tls.so.1
# chcon -t texrel_shlib_t /usr/lib/libGLcore.so.1.0.9631

Linux/x86-64 (64-bit):

# chcon -t texrel_shlib_t /usr/lib64/xorg/modules/drivers/nvidia_drv.so
# chcon -t texrel_shlib_t /usr/lib64/xorg/modules/extensions/libglx.so.1.0.9631
# chcon -t texrel_shlib_t /usr/lib64/libGLcore.so.1.0.9631
# chcon -t texrel_shlib_t /usr/lib64/tls/libnvidia-tls.so.1

Debian GNU/Linux or [K]Ubuntu with Xorg 7.x

If you wish to install the NVIDIA Linux graphics driver on a Debian GNU/Linux or Ubuntu system that ships with Xorg 7.x, please ensure that your system meets the following requirements:

* development tools like make and gcc are installed
* the linux-headers package matching the installed Linux kernel is installed
* the pkg-config and xserver-xorg-dev packages are installed
* the nvidia-glx package has been uninstalled with the --purge option and the files /etc/init.d/nvidia-glx and /etc/init.d/nvidia-kernel do not exist

If you use Ubuntu, please also ensure that the linux-restricted-modules or linux-restricted-modules-common packages have been uninstalled. Alternatively, you can edit the /etc/default/linux-restricted-modules or /etc/default/linux-restricted-modules-common configuration file and disable the NVIDIA linux-restricted kernel modules (nvidia, nvidia_legacy) via:

DISABLED_MODULES="nv nvidia_new"

Additionally, delete the following file if it exists:

/lib/linux-restricted-modules/.nvidia_new_installed

Please note: unfortunately, it has become difficult to keep track of the pre-/post-installation steps required for [K]Ubuntu, and the above instructions may be incomplete. If in doubt, it is recommended that you use your distributor's NVIDIA Linux graphics driver packages, exclusively.


If you require further assistance after following the instructions above, please see:
http://www.nvnews.net/vbulletin/showthread.php?t=46678.

[AaronP]

32-bit distributions with 64-bit kernels

Some distributions have the option of installing a 64-bit kernel for use with all 32-bit userspace programs. This configuration is not supported by the NVIDIA Linux Graphics Driver. If you try to install the 64-bit driver package on such a system, you will receive an error like the following:

./nvidia-installer: No such file or directory

If you have this configuration, use your distribution's package manager to install a 32-bit kernel and then install the 32-bit version of the NVIDIA Linux Graphics Driver.

[AaronP]

Server crash with X.org 7.3 and Xinerama

X.org 7.3 enables the Composite extension by default, but it fails to initialize when Xinerama is enabled. This causes a bad interaction with the NVIDIA driver resulting in a server crash. To work around this problem run

# nvidia-xconfig --no-composite

to explicitly disable the Composite extension.

Mouse problem with X.org 7.3 and multiple X screens or Xinerama

The version of xf86-input-mouse shipped with X.org 7.3 contains a bug that confines the pointer to one of the screens. You can avoid this problem by using the xf86-input-evdev driver instead.

[Edit: This is fixed in xf86-input-mouse-1.2.3]

[AaronP]

Fedora 9 ships with a prerelease version of xorg-server 1.5. This server has improved autoconfiguration that allows it to function without a configuration file. Because /etc/X11/xorg.conf does not exist, nvidia-xconfig will create an /etc/X11/XF86Config file instead. While this will work, some people may find it confusing.

Unfortunately, xorg-server 1.4.99.901 removed support for the RgbPath option, so X configuration files generated by nvidia-xconfig will not work. For these servers, I recommended that you delete everything but the "Device" section and leave the rest up to the X server's autoconfiguration:

Section "Device"
    Identifier     "NVIDIA Device"
    Driver         "nvidia"
EndSection

Future releases of nvidia-xconfig will be better equipped to handle these minimalist configuration files.

[AaronP]

Fedora 12 (and possibly other Linux distributions) ships with the "nouveau" driver enabled by default. Having this driver loaded prevents the NVIDIA driver from loading, producing an error like the following:

NVRM: The NVIDIA probe routine was not called for 1 device(s).
NVRM: This can occur when a driver such as rivafb, nvidiafb or
NVRM: rivatv was loaded and obtained ownership of the NVIDIA
NVRM: device(s).
NVRM: Try unloading the rivafb, nvidiafb or rivatv kernel module
NVRM: (and/or reconfigure your kernel without rivafb/nvidiafb
NVRM: support), then try loading the NVIDIA kernel module again.
NVRM: No NVIDIA graphics adapter probed!

Please consult your distribution's documentation on how to disable the nouveau driver before installing the NVIDIA driver.