nvidia-graphics-drivers#LEGACY# for Debian
----------------------------------


INTRODUCTION:
-----------------------------------------------------------------------

Welcome to Debian NVIDIA packages. This document contains notes on the
kernel module interface for the NVIDIA driver offering.


KERNEL MODULE INSTRUCTIONS:
-----------------------------------------------------------------------

 METHOD #1: automatically, via dkms - Dynamic Kernel Module Support

  * Easiest and recommended method.
  * DKMS will take care of building a new module whenever the driver
    or your kernel headers are updated.
  * Install the linux-headers-* package matching your kernel.
  * Install the package nvidia-kernel#LEGACY#-dkms instead of
    nvidia-kernel#LEGACY#-source.
  * STOP reading this file.

 METHOD #2: manually, using module-assistant

  * Requires a Debian supplied kernel or a kernel where you built a
    corresponding linux-headers-KVERS package.
  * Run this command every time your kernel or the driver was updated:

        module-assistant auto-install nvidia#LEGACY#

  * For more information on module-assistant, see its man page and package
    documentation.
  * Additional debug information can be printed by running
        export KBUILD_VERBOSE=1
    before starting the build.
  * STOP reading this file.

There are TWO more ways to build the nvidia-kernel package manually / via
make-kpkg.
Which one depends on your situation with your kernel.

 METHOD #3: You are running a Debian supplied kernel or built a
    linux-headers package along with your own self-built kernel.

 METHOD #4: You are running your own self-built kernel built from kernel
    source.

Which method you choose really depends on what kernel headers you wish to
use, those from a kernel-headers package or those from kernel
source (from which you built your own kernel).

As mentioned above, if you are running a Debian supplied kernel you will
probably want to choose METHOD #3. If you compile your own kernels, METHOD #4.



PRELIMINARY:
- Decide where you want to build your module. By default it
will build under /usr/src like other kernel modules but you may choose
to build it under your home directory as some people prefer and writing to
/usr violates the FHS.



METHOD #3 Using a linux-headers package
***********************************************************************

As root (or using fakeroot)
1. cd /usr/src
   tar xjvf nvidia-kernel#LEGACY#.tar.bz2 -C <YOUR BUILD LOCATION>
   (It will install in <YOUR BUILD LOCATION>/modules)
   - or -
   tar xjvf nvidia-kernel#LEGACY#.tar.bz2  (if building in /usr/src)

   The standard build location is /usr/src

2. Find out your kernel version:

   uname -r   For example: 2.6.14-2-k7

3. Download and install package: linux-headers-2.6.14-2-k7
   It will install in /usr/src/

   Make sure your kernel image and headers have matching release numbers to
   avoid possible problems in packages built from different sources.

4. Set an environment variable (if bash is your shell):

    export KSRC=/usr/src/linux-headers-2.6.14-2-k7

   If you are building for your currently running kernel, and if you have
   the kernel build tree in /lib/modules/<version>/build (or a symlink
   there to the appropriate place), you don't need to do this.  The
   standard Debian linux-headers-* packages will set up that link.

5. Then build nvidia-kernel package:

   cd <YOUR BUILD LOCATION>/modules/nvidia-kernel#LEGACY#
   debian/rules binary_modules

(You can also combine step 4 and 5 into one line:
KSRC=/usr/src/linux-headers-2.6.14-k7 debian/rules binary_modules)

6. Install nvidia-kernel-common on the machine where the module will be
deployed:

   If not installed already
   apt-get install nvidia-kernel-common

7. Install the nvidia-kernel package:

   dpkg -i ../nvidia-kernel#LEGACY#-2.6.14-2-k7_#VERSION#-1+_.Custom_i386.deb
   ( or similar filename )

Now see GENERAL NOTES below method #4

METHOD #4: Using your own kernel source headers
***********************************************************************

To build the nvidia-kernel#LEGACY# deb you need to first make sure you have
kernel-package installed, then do the following:

As root
1. cd /usr/src
   tar xjvf nvidia-kernel#LEGACY#.tar.bz2 -C <YOUR BUILD LOCATION>
   (It will install in <YOUR BUILD LOCATION>/modules)
   - or -
   tar xjvf nvidia-kernel#LEGACY#.tar.bz2  (if building in /usr/src)

2. If you are NOT using the default modules location /usr/src/modules
   then you must set an environment variable that points to your modules
   location.

     export MODULE_LOC=<YOUR BUILD LOCATION>/modules

   This is needed by make-kpkg which is used later. If
   <YOUR BUILD LOCATION> is /usr/src there is no need to set this.

3. Build the modules under MODULE_LOC i.e. <YOU BUILD LOCATION>/modules

     cd linux (or your kernel source directory)
     make-kpkg modules_image

4. Install nvidia-kernel-common on the machine where the module will be
installed:

   apt-get install nvidia-kernel-common

5. Install the nvidia-kernel package:

     cd <YOUR BUILD LOCATION>    (e.g. /usr/src)
     dpkg -i nvidia-kernel#LEGACY#-KVER*.deb

Notes for method #2:

- It is advised not to clean the kernel source tree between
"make-kpkg kernel_image" and "make-kpkg modules_image".

- The Riva framebuffer is known to conflict with the nvidia X driver. If you
are using X it would be wise not to compile it in. The vesa framebuffer is
known to work in some cases, and not in others. (people have recently been
having problems with the vesa driver as well)

GENERAL NOTES:

Also you must add any users who wish to use OpenGL applications to the group
video. You can do this with:

        adduser username video

Also note for AGP issues and further information please see the NVIDIA
README file (README.gz) in the nvidia-glx#LEGACY# package

An old nvidia.ko might be already loaded (run 'lsmod' to check) so
do 'rmmod nvidia' to remove the module from memory.


For any news on this package check
http://bugs.debian.org/nvidia-kernel#LEGACY#-source


 -- Russ Allbery <rra@debian.org>, Sat, 25 Sep 2010 23:30:28 -0700