Re: [PATCH 7/9] irq-remove: scsi driver trivial

[Arjan van de Ven <arjan@infradead.org>]

On Fri, 26 Oct 2007 20:12:40 -0400
Jeff Garzik <jeff@garzik.org> wrote:

> Arjan van de Ven wrote:
> > On Fri, 26 Oct 2007 17:47:58 -0400
> > Jeff Garzik <jeff@garzik.org> wrote:
> > 
> >> Andrew Morton wrote:
> >>> That was a goofup.  I proposed that we should add a #define
> >>> TWO_ARG_IRQ_HANDLERS (or whatever) and I think I actually wrote
> >>> the patch, but it got lost.
> >>>
> >>> I agree it would be a kind thing to do in this case.
> > 
> >>
> >> Yep, I was thinking that including
> >>
> >> 	#define IRQ_HANDLER_V3
> >>
> >> would be a good idea.
> >>
> > 
> > it sets a certain precedent though.... we don't do this for the 500
> > other API changes we do each release (see stable-api-nonsense)... so
> > this one is mostly arbitrary picked out
> 
> We do for include/linux/netdevice.h, see HAVE_xxx -- and we should do
> it because the last irq handler change was a pain for backports, and
> this makes life easier for the backporters.  irq handling is probably
> far more global than any other kernel API except kmalloc()

to be honest, in a perfect world we turn this around, and have the
older kernels that people want to backport to have the
LEGACY_IRQ_HANDLER defines... not accumulating going forward....
(in fact, with what you're proposing you'll always get #ifndef's..)

the other serious question is.. how is IRQ_HANDLER_V3 different from a
#ifdef VERSION >= 2.6.24 .....
it's not really ;)



-- 
If you want to reach me at my work email, use arjan@linux.intel.com
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org