Re: [PATCH] sh: convert to asm-generic/hardirq.h

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

On Thu, Dec 17, 2009 at 09:19:16AM +0100, Christoph Hellwig wrote:
> On Thu, Dec 17, 2009 at 05:15:54PM +0900, Paul Mundt wrote:
> > On Thu, Dec 17, 2009 at 09:10:48AM +0100, Christoph Hellwig wrote:
> > > On Wed, Aug 05, 2009 at 04:18:28PM +0900, Paul Mundt wrote:
> > > > On Tue, Aug 04, 2009 at 04:55:29PM +0200, Christoph Hellwig wrote:
> > > > > 
> > > > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > > > 
> > > > Applied, thanks.
> > > 
> > > Did this cause problems in the end?  Still doesn't seem to be in Linus'
> > > tree.
> > 
> > Nope, it was in Linus' tree for awhile, but then I needed to add an
> > nmi_count and had to switch back to an arch-specific one.
> 
> Is there a good reason to overload the irq_cpustat with that?  It seems
> like a quite horrible way to duplicate per-cpu data in a less efficient
> way.

I suppose not, but there didn't seem like a lot of good alternatives
either. Shoving it in irq_cpustat is what is presently done by at least
mn10300, x86, and xtensa. sparc chose stashing it in its cpuinfo struct,
but then also has a meaningful per-cpu storage area. If you have any
better ideas I'll certainly consider moving it.