Re: [PATCH 2/3] Full conversion to early_initcall() interface, remove old interface.

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 27 Jun 2008 14:28:00 +0300
Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro> wrote:

> On Fri, 27 Jun 2008 12:54:21 +0200
> Johannes Weiner <hannes@saeurebad.de> wrote:
> 
> > Hi Eduard,
> 
> Hi.
>  
> > You forgot to remove the declaration from linux/interrupt.h.
> 
> The last patch was made against linux-next, as Andrew Morton
> suggested, where spawn_ksoftirqd() was declared in the scope of
> __do_pre_smp_initcall():
> @@ -793,17 +784,6 @@ static void __init __do_pre_smp_initcalls(void)
>  		do_one_initcall(*call);
>  }
>  
> -static void __init do_pre_smp_initcalls(void)
> -{
> -	extern int spawn_ksoftirqd(void);
> -
> -	init_call_single_data();
> -	migration_init();
> -	spawn_ksoftirqd();
> -	if (!nosoftlockup)
> -		spawn_softlockup_task();
> -}
> -
> 
> It may be that some other patch in -mmotm moves that into
> include/linux/interrupt.h. In linux-next with my patch, running
> $ find . -name \*.h | xargs grep "spawn_ksoftirqd"
> shows nothing.

Yes, I dropped that patch as it's no longer relevant.

> Thanks for spotting this difference.
> 
> (Using -mmotm + quilt myself is an additional headache, since git
> already does what I want in terms of patch (read commits) management;
> that's why I prefered linux-next.)

It is a bit of a pain.  I avoid asking people to raise patches against
-mm unless it's really necessary.