Re: [cpuops cmpxchg double V1 1/4] Generic support for this_cpu_cmpxchg_double

[Tejun Heo <tj@kernel.org>]

Hello,

On Tue, Dec 21, 2010 at 03:24:45PM -0800, H. Peter Anvin wrote:
> On 12/21/2010 02:36 PM, Christoph Lameter wrote:
> > On Sat, 18 Dec 2010, Tejun Heo wrote:
> > 
> >> pointers from the beginning.  Anyways, that's too late, so is it
> >> completely impossible to make cmpxchg_double's take a scalar value?
> >> It can take the pointer all the same, no?
> > 
> > It could take a scalar value like the others but we are then not operating
> > on the scalar alone but also on the following field.

Yes, it's weird but the operation itself is weird enough and named
accordingly, so to me it seems like a much lesser problem than
breaking interface consistency with other this_cpu_ ops.

> I'm a bit confused on this one.  The standard cmpxchg() takes a scalar
> and a pointer, and returns a scalar.  The equivalent for the "double"
> variety would be to return a compound object, basically:
> 
> struct double_ulong {
> 	unsigned long v[2];
> };
> 
> ... which can be returned in registers on both i386 and x86-64.
> 
> It's a bit clumsy from a type perspective, but I'm not sure that that is
> a bad thing.  Doing too much type genericity has caused us problems in
> the past.

Yeah, the above might be better too.  Is there any reason to use
cmpxchg_double on anything smaller?

Thanks.

-- 
tejun