2.3 The Post-compilation Stage

This is the important stuff.. you will now have to include your new configuration as an option in your bootmanager. The method I will describe is to add a new kernel to LILO. This way, you have the option of booting the new kernel, as well as booting the old kernel, just in case there are any problems.

your newly compiled kernel is in /usr/src/linux/arch/*architecture*/boot/ where *architecture* is the type of computer you are using. If its a plain vanilla pc (like mine), then you can substitiute *architecture* with i386. The kernel's name is bzImage (or if you used make zImage during compile, it will be zImage).

You must also know where you keep your current kernel's. On my system it is /boot but some people use their root directory.

- So.. bearing in mind your specifics..

# cd /boot
# mv bzImage  vmlinuz-2.2.9   - renames the old kernel so you dont overwrite it, and you know what it is
# mv system.map  systemp.map-2.2.9  - moves the old system map
# cp /usr/src/linux/arch/i386/boot/bzImage  /boot/  - copies the new kernel to the boot directory
# cp /usr/src/linux/System.map  /boot/  - copies the new system map to the boot directory

I will take moment here to discuss the System.map file, and the organisation of this directory in general. In the above example I have assumed that you have a static System.map you want to replace, although some systems have a symbolic link named system.map pointing to a different file. I have assumed this on the basis that most distributions install a static system.map file, and so if you don't know whether you have a static system.map file, or just a symbolic link to another file, the odds are you have a static one, and will be perfectly fine when following the instructions given.

It is important to remember the differences and filenames you've just changed, as they need to be reflected in your lilo configuration file.

- Pull up /etc/lilo.conf with your favorite editor

If you have only one kernel in here, and you have no other OS on your system, their should only be one stanza relating to kernel locations and names. Other OS's you use will have identifiable labels. A stanza with a label=windows should be left alone. You only need to be concerned with your linux boot parameters.

- Find your old kernel's stanza.. an example of what it may look like is shown below

image = /boot/bzImage
root = /dev/hda1
label = linux

As you remember I changed my old bzImage to vmlinuzold so I need to edit that stanza to reflect the changes.. so it becomes

image = /boot/vmlinuzold
root = dev/hda1
label = old

- Then you need to add a new stanza for your new kernel.. an example form my /etc/lilo.conf is..

image = /boot/bzImage
root = /dev/hda1
label = new

- Now you need to save your lilo.conf and then run the following to update LILO

# /sbin/lilo

- If lilo gives output of options without any errors.. well.. you've finished. All that remains is to reboot, and at the LILO prompt (attained by pressing Shift when the word LILO appears in the boot sequence) type your new kernel label, which in my case is "new". Obviously I'll leave out the inverted comma's :)

Once you've checked it, by booting (keeping that boot floppy handy :)), and you are happy with it, you can edit /etc/lilo.conf once more, to make your new kernel the default kernel to boot. To do that, just make your new kernel's stanza the first one in the list of boot options. In my case, just switch the vmlinuzold stanza with the bzImage stanza.
 

3. Patching a Kernel

The advantage of patching an existing kernel, over downloading the next kernel release, is that people with little time and/or bandwidth can keep their kernel's up to date, just by applying a 20k (approx.) patch file.

You need to download the patches you require, in sequential order from the patchlevel of your current kernel source, to the patchlevel you wish to upgrade to.

For example, if I had kernel 2.2.6 and I wanted to upgrade to kernel 2.2.11 i would need patches 2.2.7.. 2.2.8.. 2.2.9 through to 2.2.11. Kernel patches follow the same naming conventions as the kernel itself does. i.e. patch-2.2.8 is he patch to upgrade from 2.2.7 to 2.2.8.

You do not have to store your kernel patches in any specific location, rather it is the location where you execute the patch command that is important. To patch your kernel, you must do the following as root..

# cd /usr/src - You must execute your patch command from this location
# gunzip -c patchfile | patch -p0 - Execute this command for every patch you have, where patchfile is the name and location of your patch

Note: You run the patch command above for each patchfile in sequential order. Following the previous example, if i had kernel 2.2.6 and wanted to patch up to 2.2.11 I would first apply patch 2.2.7, followed by 2.2.8, 2.2.9 through to 2.2.11
You cannot use any wildcards such as * either, as this applies the patches in ASCII order, not numerical order.

This is all you should need to do. To make sure all the patches were applied successfully, issue the following as root..

# find /usr/src/linux -follow -name "*.rej" -print
#find /usr/src/linux -follow -name "*#" -print

If either of the above two commands return any files, then some parts of the patch process could not be applied for some reason. If you are in doubt as to what action should be taken to address such problems, I would suggest you download a kernel source package, and start from the beginning

If everything went cleanly, you have successfully patched your kernel source, and are free to compile a new up-to-date kernel.
 

4. Other Sources of Information

The documents in your kernel sources are the most detailed.. You should find them here

the LDP has a good selection of kernel info

For those of you who want to download your kernels/patches, find your local mirror here.

7DS

[-Previous Page-]