Re: mm: memory/cpu hotplug section mismatch.

[Paul Mundt <lethal@linux-sh.org>]

On Mon, Jun 11, 2007 at 08:27:32AM -0700, Randy Dunlap wrote:
> On Mon, 11 Jun 2007 14:09:55 +0900 Paul Mundt wrote:
> > On Mon, Jun 11, 2007 at 02:01:45PM +0900, KAMEZAWA Hiroyuki wrote:
> > > On Mon, 11 Jun 2007 13:35:43 +0900
> > > Paul Mundt <lethal@linux-sh.org> wrote:
> > > > This happens because CONFIG_HOTPLUG_CPU=n sets __cpuinit to __init, but
> > > > CONFIG_MEMORY_HOTPLUG=y unsets __meminit.
> > > 
> > > It seems this zone_batchsize() is called by cpu-hotplug and
> > > memory-hotplug.  So, __init_refok doesn't look good, here.
> > > 
> > > maybe we can use __devinit here. (Because HOTPLUG_CPU and
> > > MEMORY_HOTPLUG are depend on CONFIG_HOTPLUG.)
> > > 
> > Yes, that's probably a more reasonable way to go. The __devinit name is a
> > bit misleading, though..
> 
> __meminit does not fit/work here?
> 
No, for the reasons already noted.

If CONFIG_MEMORY_HOTPLUG=n __meminit == __init, and if
CONFIG_HOTPLUG_CPU=n __cpuinit == __init. However, with one set and the
other disabled, you end up with a reference between __init and a regular
non-init function.

CONFIG_HOTPLUG is the only thing that they both have in common, so only
__devinit will gaurantee the proper behaviour. __init_refok is the
opposite of the behaviour that is desired, as Kamezawa-san was quick to
point out.

Simply switching to __meminit will cause zone_batchlist() to emit a
section mismatch on CONFIG_HOTPLUG_CPU=y and CONFIG_MEMORY_HOTPLUG=n
configurations instead of the other way around.