Re: linux-next: manual merge of the rr tree

[Rusty Russell <rusty@rustcorp.com.au>]

On Wednesday 07 January 2009 00:51:20 Mike Travis wrote:
> Ingo Molnar wrote:
> > I think the complete elimination of cpumask_t should be the primary 
> > priority - before jumping to any other aspect. If we dont get rid of it it 
> > will stick around forever, like the BKL. It was a nice migration helper 
> > but now it's time to wave goodbye? :)
> > 
> > 	Ingo
> 
> I think that's possible for 2.6.30 especially with Rusty's "big hammer"
> patch that removes the definition of cpumask_t.

I have two config option patches.  One removes the old deprecated ops.  This
almost compiles now.  The other removes the struct cpumask and cpumask_t
definitions.  This breaks horribly.  Wading through that is on my TODO.

cpus_allowed in task_struct is fairly nasty.  We'll introduce an accessor
macro for that one I think since it's widespread (a "big hammer" is going to
kill us all if we try it!).

But I agree that we should get to that goal as fast as possible; it's the
only real way to stop people introducing onstack cpumasks, copying them, etc.

Cheers,
Rusty.