Why “sys_clone” is defined in “./arch/h8300/kernel/process.c”?

Question

I am hacking linux-4.13.4, and I learn clone system call from book

Linux Kernel Development 3rd Edition

I am curious about why sys_clone is defined in "./arch/h8300/kernel/process.c" ?

This is the only place I can find the function definition.

In my opinion, the folder path of clone system call is very inconsistent. Is it a historical reason that clone is firstly implemented in architecture h8300 , so Linus Torvalds put clone in /arch/h8300/ ?

Reference:

https://www.classes.cs.uchicago.edu/archive/2006/winter/23000-1/docs/h8300.pdf

https://en.wikipedia.org/wiki/H8_Family

Answer 1

It's not for historical reasons, it's because some implementation of clone depends on the architecture. Some CPUs like the h8300 pass more parameters in the register than the generic sys_clone wrapper found in kernel/fork.c ... There it's defined using the SYSCALL_DEFINE* macros:

In 4.13.5 it's around line 2133.

#ifdef __ARCH_WANT_SYS_CLONE
#ifdef CONFIG_CLONE_BACKWARDS
SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
                 int __user *, parent_tidptr,
                 unsigned long, tls,
                 int __user *, child_tidptr)
#elif defined(CONFIG_CLONE_BACKWARDS2)
SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
                 int __user *, parent_tidptr,
                 int __user *, child_tidptr,
                 unsigned long, tls)
#elif defined(CONFIG_CLONE_BACKWARDS3)
SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
                int, stack_size,
                int __user *, parent_tidptr,
                int __user *, child_tidptr,
                unsigned long, tls)
#else
SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
                 int __user *, parent_tidptr,
                 int __user *, child_tidptr,
                 unsigned long, tls)
#endif
{
        return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
}
#endif

In case of h8300 there is an arch specific sys_clone, which is needed because of the way parameters are passed from the calling process to a forking process in case of h8300 all the params have to be passed through the registers (rather than a mix of registers and stack) and because clone will trounce the registers it requires cpu specific handling.