Building-linux
How to build a patched Linux kernel for the x86isa model.
The x86isa model is capable of booting a slightly
modified Linux kernel, which includes modifications to add support for
our timer and tty and removes Linux's dependence on
unimplemented peripherals. Here we describe how to obtain a copy of
the Linux source, patch it, and build the kernel. We assume you're
using a Linux system with a "Linux-style" file system.
First, we obtain the Linux source via git. Execute the
following in a terminal:
git clone https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
cd linux
This will clone the Linux source tree into a folder called
linux in your current working directory and then cd into it.
Our patch was tested on version v6.12, so we'll checkout that
version:
git checkout v6.12
Now, we need to patch the kernel. We include the patch at
books/projects/x86isa/linux/x86isa-linux.patch in the ACL2 source
tree. To apply it run
git am <path to patch>
in your copy of the linux source, substituting <path to patch>
for the path to the given x86isa-linux.patch file.
Normally, at this point you'd run make menuconfig to configure
the kernel for your machine, but we included a .config file in
the patch that configures the kernel to work with the x86isa
model, so you don't need to worry too much about the config. However,
run
make olddefconfig
to update the config with details dependent on your compiler
Finally, we can compile the kernel. Run:
make
Using multiple cores can increase build speed. Use
make -j
or
make -j<nprocs>
to use as many threads as you have logical processors or
<nprocs> threads to build respectively.
This will build a bzImage in arch/x86/boot/bzImage, which
is the kernel in Linux's bzImage format. This is the file we will load
into the model. It also produces a vmlinux file in the root of
the Linux source tree, which contains the kernel in ELF format. While
this isn't used to boot, it can be useful when debugging.
