Re: [PATCH][V4] Add reboot_pid_ns to handle the reboot syscall

[Oleg Nesterov <oleg@redhat.com>]

On 12/13, Andrew Morton wrote:
>
> > --- a/kernel/sys.c
> > +++ b/kernel/sys.c
> > @@ -444,6 +444,9 @@ SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
> >  	                magic2 != LINUX_REBOOT_MAGIC2C))
> >  		return -EINVAL;
> >
> > +	if (task_active_pid_ns(current) != &init_pid_ns)
> > +		return reboot_pid_ns(task_active_pid_ns(current), cmd);
> > +
> >  	/* Instead of trying to make the power_off code look like
> > 	 * halt when pm_power_off is not set do it the easy way.
> > 	 */
>
> I'll repeat my cruelly-ignored review comment for v3:
>
> This adds a bunch of useless code if CONFIG_PID_NS=n.

Agreed.

> It would be
> better to do
>
> #ifdef CONFIG_PID_NS
> extern void pidns_handle_reboot(int cmd);
> #else
> static inline void pidns_handle_reboot(int cmd)
> {
> }
> #endif

Can't resist.

Why the kernel always prefers to do it this way, adding the ugly
do-nothing inlines?

Isn't it better to simply call pidns_handle_reboot(cmd) under
CONFIG_PID_NS in sys_reboot() ?

	#ifdef CONFIG_PID_NS
	if (task_active_pid_ns(current) != &init_pid_ns)
		return reboot_pid_ns(cmd);
	#endif

This way, if you look at sys_reboot() you can immediately see what
happens, no need to inspect the !CONFIG_PID_NS definition. Plus this
doesn't add the "unnecesary" entry into tag file.

OK, </flame> ;)

Otherwise,

Reviewed-by: Oleg Nesterov <oleg@redhat.com>