Re: [RFC][v6][PATCH 8/9]: Define clone_with_pids() syscall

[Arnd Bergmann <arnd@arndb.de>]

On Thursday 10 September 2009, Sukadev Bhattiprolu wrote:
> 
> +int sys_clone_with_pids(struct pt_regs *regs)
> +{
> +       unsigned long clone_flags;
> +       unsigned long newsp;
> +       int __user *parent_tidptr;
> +       int __user *child_tidptr;
> +       struct pid_set __user *upid_setp;
> +
> +       clone_flags = regs->bx;
> +       newsp = regs->cx;
> +       parent_tidptr = (int __user *)regs->dx;
> +       child_tidptr = (int __user *)regs->di;
> +       upid_setp = (struct pid_set __user *)regs->bp;
> +
> +       if (!newsp)
> +               newsp = regs->sp;
> +
> +       return do_fork_with_pids(clone_flags, newsp, regs, 0, parent_tidptr,
> +                       child_tidptr, upid_setp);
> +}

I wonder if we can avoid spreading copies of this function across
all architectures.

Would it be possible to define it like this?

asmlinkage long
clone_with_pids(int flags, unsigned long child_stack_base, int __user *parent_tid_ptr,
                       int __user *child_tid_ptr, struct pid_set __user *pid_setp)
{
	if (!child_stack_base) {
		struct pt_regs *regs;

		regs = task_pt_regs(current);
		child_stack_base = user_stack_pointer(regs);
	}

	return do_fork_with_pids(clone_flags, child_stack_base, 0,
		parent_tid_ptr, child_tid_ptr, pid_setp);
}

	Arnd <><