System development requires knowing how to debug kernel know how, in this example will be used Qemu and GDB. Qemu creates the virtual machine that kernel will run on and GDB will connect to it to help us understand how things tick.
$ tar xf linux-4.9.258.tar.xz $ cd linux-4.9.258
Default configuration disable some security configurations that allow us to debug (random memory layout KALSR), CONFIG_COMPAT_BRK don't randomize position of the programs (randomize_va_space) useful when debugging a program (tracing). Configuration flags to enable;
And to disable;
This changes can be achieved by creating a config-fragment and then merge it with the configuration.
$ cat <<EOF >.config-fragment CONFIG_64BIT=y CONFIG_DEBUG_KERNEL=y CONFIG_HAVE_ARCH_KGDB=y CONFIG_COMPAT_BRK=y CONFIG_FTRACE=y CONFIG_PRINTK=y CONFIG_BLK_DEV_INITRD=y CONFIG_BINFMT_ELF=y CONFIG_TTY=y CONFIG_DEBUG_INFO=y CONFIG_DEBUG_INFO_DWARF4=y CONFIG_GDB_SCRIPTS=y CONFIG_READABLE_ASM=y CONFIG_FRAME_POINTER=y CONFIG_KGDB=y CONFIG_KGDB_LOW_LEVEL_TRAP=y CONFIG_EARLY_PRINTK=y CONFIG_CC_OPTIMIZE_FOR_SIZE=n EOF
Create a tiny config;
$ make ARCH=x86_64 tinyconfig
Merge config with the following script;
$ ./scripts/kconfig/merge_config.sh .config .config-fragment
Check or change the configuration according to your needs;
$ make nconfig
Build Kernel and modules;
$ make -j $(nproc)
Now that you have the kernel compiled you can create a simple init program, this program is called when kernel finish to load and setup its internals and is ready to launch first process. Init program should not exit or kernel will panic. Create init.S;
.global _start _start: mov $1, %rax mov $1, %rdi mov $message, %rsi mov $message_len, %rdx syscall jmp . message: .ascii "FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR\n" .equ message_len, . - message
Assemble, link and create simple initial ram disk;
mkdir d as --64 -o init.o init.S # assemble ld -o d/init init.o # link cd d find . | cpio -o -H newc | gzip > ../rootfs.cpio.gz
Clean temporary directory;
cd .. rm -r d/
Can be used C to create init program;
#include <stdio.h> #include <unistd.h> int main() { printf("FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR\n"); sleep(0xFFFFFFFF); return 0; }
$ gcc -static init.c -o init
Test qemu, kernel and simple init program, you should see "FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR FOOBAR";
$ qemu-system-x86_64 -enable-kvm --kernel arch/x86_64/boot/bzImage \ --initrd rootfs.cpio.gz
If everything goes well you can start qemu without starting the CPU (-S) and with gdb server on TCP port 1234 (-s).
$ qemu-system-x86_64 -enable-kvm --kernel arch/x86/boot/bzImage \ --initrd rootfs.cpio.gz \ -S -s
On another terminal start gdb;
gdb \ -ex "add-auto-load-safe-path $(pwd)" \ -ex "file vmlinux" \ -ex 'set arch i386:x86-64:intel' \ -ex 'target remote localhost:1234' \ -ex 'hbreak start_kernel' \ -ex 'continue' \ -ex 'disconnect' \ -ex 'set arch i386:x86-64' \ -ex 'target remote localhost:1234'
(gdb) info thread Id Target Id Frame * 1 Thread 1 (CPU#0 [running]) start_kernel () at init/main.c:480
(gdb) info frame Stack level 0, frame at 0xffffffff81e03f90: rip = 0xffffffff81f4db2d in start_kernel (init/main.c:480); saved rip = 0xffffffff81f4d28e called by frame at 0xffffffff81e03fa0 source language c. Arglist at 0xffffffff81e03f80, args: Locals at 0xffffffff81e03f80, Previous frame's sp is 0xffffffff81e03f90 Saved registers: rip at 0xffffffff81e03f88
(gdb) print $rip $2 = (void (*)()) 0xffffffff81f4db2d <start_kernel> (gdb)
lx-symbols allows to debug kernel modules, after starting the vm and loading the module use lx-symbols to load the symbols from all the modules loaded in the kernel.
(gdb) apropos lx (gdb) lx-symbols
It's useful to set conditional breakpoints or a break point can be trigger by unrelated tasks, example of a break point on do_exit function but only by the process with pid 1;
(gdb) br do_exit if $lx_current()->pid == 1C Index
This is part of the LeetIO System Documentation. Copyright (C) 2021 LeetIO Team. See the file Gnu Free Documentation License for copying conditions.