[BUG] 2.6.37-rc3 massive interactivity regression on ARM

[Christoph Lameter]

On Fri, 10 Dec 2010, Eric Dumazet wrote:

> > Yeah, but that kinda defeats the purpose of having it implemented in
> > seqlock.h. Ideally we'd teach gcc about these long pointers and have
> > something like:
> >
> >   write_seqcount_begin(&this_cpu_read(irq_time_seq));
> >
> > do the right thing.
>
> gcc wont be able to do this yet (%fs/%gs selectors)

The kernel can do that using the __percpu annotation.

> But we can provide this_cpu_write_seqcount_{begin|end}()

No we cannot do hat. this_cpu ops are for per cpu data and not for locking
values shared between processors. We have a mechanism for passing per cpu
pointers with a corresponding annotation.

> static inline void this_cpu_write_seqcount_begin(seqcount_t *s)
				^^^ Would have to be seqcount_t __percpu *s

> {
> 	__this_cpu_inc(s->sequence);
> 	smp_wmb();
> }