Lan Barnes wrote:
It's been a long day.
My vague memory of kernel compiles was:
- add something to extraversion
- make config
- (first compile - no make clean/mrproper)
- make (vmlinuz?)
= make modules
- make install
- make modules_install
- or, make all
It's crapping out
[EMAIL PROTECTED] 2.6.15-1.1831_FC4-i686]# make all
CHK include/linux/version.h
HOSTCC scripts/mod/file2alias.o
HOSTCC scripts/mod/modpost.o
HOSTCC scripts/mod/sumversion.o
HOSTLD scripts/mod/modpost
make[1]: *** No rule to make target `init/main.o', needed by
`init/built-in.o'. Stop.
make: *** [init] Error 2
[EMAIL PROTECTED] 2.6.15-1.1831_FC4-i686]#
I am _so_ tired ... and I can't find my old Red Hat notebook.
I'll see your answers in the morning.
TIA,
Why do you need to compile a kernel in the first place? Perhaps there is
another way for you do get whatever it is you are working on to work.
However, since you are running Fedora Core 4, there are going to be
multiple approaches to building a kernel, depending on your ultimate goal.
The first method is to grab a plain vanilla kernel from kernel.org in
the form of a bzip tarball. Everything up to the point of installation
can be done as a regular user. Create a build directory to put
everything in. Then expand your tarball, change to the directory, and
from there you have a choice of configuration options. If you do the
configuration under X, you can choose to use gconfig, which uses the
gtk+ toolkit for a GUI configuration tool, and tends to be associated
with GNOME desktops. You invoke it with "make gconfig". Or you can use
xconfig, which uses the QT widget set, and is invoked with "make
xconfig" and tends to be associated with the KDE desktop. Or you can use
the ncurses method by doing "make menuconfig". If all else fails you can
do a "make config" which will be long, tedious amd error-prone, because
you can't go back and fix mistakes and you probably won't have a clue as
to what some of the options are (I know I don't).
Once you get done doing the configuration, you will save the config file
(it usually gets saved for you as .config in the base directory). To
build the kernel and modules issue a "make all". To see all your
possible options you can issue a "make help". You'll notice that there
is an option to build RPM packages. When its done building the kernel,
run a "make install" to put everything in its place.
Another way to build a kernel is with an RPM. I build all my RPMs as a
regular user by first overriding the default source install and build
locations for rpm. In my home directory I have a little file called
.rpmmacros with the following contents:
%_topdir %(echo ${HOME}/rpm)
%_tmppath %{_topdir}/tmp
%packager Gus Wirth
# %_enable_debug_packages 0
This means that I have a directory called rpm under my home directory
that is the top level directory for doing rpm stuff, and that I have a
temporary directory under the topdir for doing, well, temporary stuff. I
should probably change this back to my /tmp directory but I needed to do
some troubleshooting and didn't want to risk losing info to automatic
clean-up.
Under the ~/rpm directory I create the directory tree that you would
normally find under /usr/src/redhat/. With that in place, I can install
RPM source packages to my home directory as a regular user and then
build them from there without being root.
To build a Redhat/Fedora kernel (as opposed to a plain vanilla kernel)
will require a few more steps. The tricky part will be to get YOUR
configuration to work as part of the deal. The first thing to do will be
to grab the source RPM for the kernel and install it. You will wind up
with a spec file, a compressed tarball of the original plain vanilla
source, and a whole bunch of patches, plus the configuration files used
by Redhat to create the final kernel. In order for you to do your own
configuration you will probably want to employ a partial build option
for the rpm. To start with, try:
$ rpmbuild -bp kernel.spec
from the ~/rpm/SPECS directory (if you did the .rpmmacros thing). This
will do the prep phase of creating the RPM, which uncompresses the
source and applies all the patches. Getting the patches applied is very
important because they might change your configuration options. If you
have your own patches that need to be applied to the source, you should
hack the spec file and have those included as part of the rpm build
process. If you apply the patches without using the spec file, you won't
be able to create an rpm the usual way.
From here you have a couple options. You can go to the build directory
and configure and build the kernel like in the plain vanilla scenario
above. This is the easiest way to do this. The other option is to do
your configuration, save the config file as something like Lan.config
and then hack up the spec file to use your config file rather than the
ones provided by Redhat (note that I'm getting lazy here and no longer
providing step-by-step instructions). Once you have that done, you can
build the new kernel by issuing:
$ rpmbuild -ba kernel-hacked-by-lan.spec
That will then build your new kernel and package it up for you. Install
the RPM and you're done.
Plan on spending a few hours with this if you haven't done it before.
Gus
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
