Re: drivers/base/core.c broken for non-CONFIG_HOTPLUG

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

On Fri, 20 Jul 2007 14:06:19 +0200 Kay Sievers <kay.sievers@vrfy.org> wrote:

> On Fri, 2007-07-20 at 00:59 -0700, Andrew Morton wrote:
> > On Fri, 20 Jul 2007 09:01:06 +0200 Robert Schwebel <r.schwebel@pengutronix.de> wrote:
> > 
> > > You add and use
> > > 
> > > 	extern const char *kobject_actions[];
> > 
> > Added it in a .c file, too.  Even checkpatch.pl knows that
> > was wrong.
> 
> The kernel is doing that all over the place for global symbols that are
> not meant to be used out of a "private" context or where a "private"
> header file of a subsystem doesn't exist.

Whatever the reason, it's still wrong, because the problem remains: lack of
typechecking between the definition and its users.

It's defensible in one situation: where the definition is in assembly
language.  And even then the symbol can be declared in a header, if only
because a second user may be added.

> I can change that, if this isn't the way to do it?

Please.  Let's find a reasonably-relevant header for it.  If there isn't
one, let's add it; there are quite possibly other things which should be in
that header so choose with care, with that possibility in mind.