[BUG] 2.6.37-rc3 massive interactivity regression on ARM

[Peter Zijlstra]

On Fri, 2010-12-10 at 13:51 -0600, Christoph Lameter wrote:
> On Fri, 10 Dec 2010, Peter Zijlstra wrote:
> 
> > > > gcc wont be able to do this yet (%fs/%gs selectors)
> > >
> > > The kernel can do that using the __percpu annotation.
> >
> > That's not true:
> >
> > # define __percpu
> >
> > Its a complete NOP.
> 
> The annotation serves for sparse checking. .... If you do not care about
> those checks then you can simply pass a percpu pointer in the same form as
> a regular pointer.

Its not about passing per-cpu pointers, its about passing long pointers.

When I write:

void foo(u64 *bla)
{
	*bla++;
}

DEFINE_PER_CPU(u64, plop);

void bar(void)
{
	foo(__this_cpu_ptr(plop));
}

I want gcc to emit the equivalent to:

__this_cpu_inc(plop); /* incq %fs:(%0) */

Now I guess the C type system will get in the way of this ever working,
since a long pointer would have a distinct type from a regular
pointer :/

The idea is to use 'regular' functions with the per-cpu data in a
transparent manner so as not to have to replicate all logic.

> > > > 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.
> >
> > -enoparse, its not locking anything, is a per-cpu sequence count.
> 
> seqlocks are for synchronization of objects on different processors.
> 
> Seems that you do not have that use case in mind. So a seqlock restricted
> to a single processor? If so then you wont need any of those smp write
> barriers mentioned earlier. A simple compiler barrier() is sufficient.

The seqcount is sometimes read by different CPUs, but I don't see why we
couldn't do what Eric suggested.