The summary below is helpful for building custom kernels from kernel.org sources. This method of compiling kernels is the traditional method common to all distros; however, an excellent method of cleanly installing the custom kernel with makepkg and pacman is also included.
Alternatively, you can use ABS to build and install your kernel; see: Kernel Compilation. Using the existing PKGBUILD will automate most of the process and will result in a package. However, some Arch users prefer the
New users who are not fully familiar with compiling software should have a look at Kernel Compilation without ABS for New Users.
- Fetch the kernel source from
ftp.xx.kernel.org/pub/linux/kernel/, where xx is your country key (f.e. 'us', 'uk', 'de', ... - Check  for a complete list of mirrors). If you have no ftp gui, you can use
wget. For this example, we will fetch and compile 22.214.171.124; you should need to change only the version to get a different kernel.
- It is always a good idea to verify the signature for any downloaded tarball. See http://kernel.org/signature.html for how this works. Basically:
$ gpg --keyserver pgp.keyserver.domain --recv-keys 0x12345678 $ gpg --verify linux-126.96.36.199.tar.sign linux-188.8.131.52.tar
- Copy the kernel source to your build directory, e.g.:
$ cp linux-184.108.40.206.tar.bz2 ~/kernelbuild/
- Unpack it and enter the source directory:
$ cd ~/kernelbuild $ tar -xvjf linux-220.127.116.11.tar.bz2 $ cd linux-18.104.22.168
Prepare for compilation by running the following command:
This ensures that the kernel tree is absolutely clean. The kernel team recommends that this command be issued prior to each kernel compilation. Do not rely on the source tree being clean after un-tarring.
This is the most crucial step in customizing the kernel to reflect your computer's precise specifications. By setting the configurations in 'menuconfig' properly, your kernel and computer will function most efficiently.
- (Optional) Copy the .config file from the running kernel, if you want to modify default Arch settings.
$ zcat /proc/config.gz > .config
- Configure your kernel.
The Traditional Way:
$ make menuconfig (Will start with a fresh '.config'. Option dependencies are usually automatically selected.)
Other build targets (Since kernel 2.6.32):
$ make oldconfig (Only works with the old '.config' file, copied into the build directory. Also marks previously unused options as 'NEW'.) $ make localmodconfig (Tries to extract /proc/config.gz from running kernel. Pre-selecting options/modules in use.) $ make localyesconfig (Same as above, except that as many modules as possible compiled into the kernel.) $ make xconfig (Depends on Qt. A nicer interface. Dependency checking not verified.) $ make gconfig (Depends on GTK. Otherwise same as xconfig.) $ make help (Lists ALL targets available.)
Note: for more information about the build target "localmodconfig" refer to the 2.6.32 release notes.
- Make your changes to the kernel and save your config file. It is a good idea to make a backup copy, since you will likely be doing this multiple times until you get all the options right.
- If you are compiling a kernel using your current config file, do not forget to rename your kernel version, or you may replace your existing one by mistake.
$ make menuconfig General setup ---> (-ARCH) Local version - append to kernel release '3.n.n-RCn'
What about /usr/src/ ?
Using the /usr/src/ directory for kernel compilation as root, along with the creation of the corresponding symlink, is considered poor practice by some kernel hackers. They consider the cleanest method to simply use your home directory. If you subscribe to this point of view, build and configure your kernel as normal user, and install as root, or with makepkg and pacman (covered below).
However, this concept has been the target of debate, and other very experienced hackers consider the practice of compiling as root under /usr/src/ to be completely safe, acceptable and even preferable.
Use whichever method you feel more comfortable with.
Compilation and installation
Choose one of the following:
1. Manual, Traditional method
make allif you use GRUB and still have LILO installed; it will configure LILO in the end, and you may no longer be able to boot your machine! Remove LILO (pacman -R lilo) before running
make allif you use GRUB!
- Compile it:
(Same as make vmlinux && make modules && make bzImage - see make help for more information on this.)
- If you have a multi-core computer (i.e Dual Core, Quad Core, etc), you can compile the kernel faster by adding the -j flag. This will use all processors at 100%
$ make -j[# of processors + 1]
- Install modules: (This needs to be done as root.)
# make modules_install
This copies the compiled modules into a directory in /lib/modules named by the kernel version and appended string you set in menuconfig. This way, modules are kept separate from those used by other kernels on your machine.
- Copy kernel to /boot directory:
# cp -v arch/x86/boot/bzImage /boot/vmlinuz-YourKernelName
- initial RAM disk
The initial RAM disk (initrd option in the GRUB menu, or, the file "initramfs-YourKernelName.img") is an initial root file system that is mounted prior to when the real root file system is available. The initrd is bound to the kernel and loaded as part of the kernel boot procedure. The kernel then mounts this initrd as part of the two-stage boot process to load the modules to make the real file systems available and get at the real root file system. The initrd contains a minimal set of directories and executables to achieve this, such as the insmod tool to install kernel modules into the kernel. In the case of desktop or server Linux systems, the initrd is a transient file system. Its lifetime is short, only serving as a bridge to the real root file system. In embedded systems with no mutable storage, the initrd is the permanent root file system.
If you need any modules loaded in order to mount the root filesystem, build a ramdisk (most users need this). The -k parameter accepts the kernel version and appended string you set in menuconfig and is used to locate the modules in /lib/modules:
# mkinitcpio -k FullKernelName -g /boot/initramfs-YourKernelName.img
You are free to name the /boot files anything you want. However, using the [kernel-major-minor-revision] naming scheme helps to keep order if you: Keep multiple kernels/ Use mkinitcpio often/ Build third-party modules.
If you are using LILO and it cannot communicate with the kernel device-mapper driver, you have to run
modprobe dm-mod first.
The System.map file is not required for booting Linux. It is a type of "phone directory" list of functions in a particular build of a kernel. The System.map contains a list of kernel symbols (i.e function names, variable names etc) and their corresponding addresses. This "symbol-name to address mapping" is used by:
- Some processes like klogd, ksymoops etc
- By OOPS handler when information has to be dumped to the screen during a kernel crash (i.e info like in which function it has crashed).
Copy System.map to /boot and create symlink
# cp System.map /boot/System.map-YourKernelName
After completing all steps above, you should have the following 3 files and 1 soft symlink in your /boot directory along with any other previously existing files:
vmlinuz-YourKernelName (Kernel) initramfs-YourKernelName.img (Ramdisk) System.map-YourKernelName (System Map)
2. With makepkg and pacman (Recommended)
See Kernel Compilation.
Add an entry for your amazing new kernel in your bootloader's configuration file - see GRUB or LILO for examples. Note that if you use LILO, the kernel sources include a script to automate the process:
If you use LILO, remember to type
lilo as root at the prompt to update it.
Using the NVIDIA video driver with your custom kernel
To use the NVIDIA driver with your new custom kernel, see: Installing the driver for a custom kernel. You can also install nvidia drivers from AUR.