Re: [cpuops cmpxchg double V2 1/4] Generic support for this_cpu_cmpxchg_double

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

* Christoph Lameter (cl@linux.com) wrote:
> On Fri, 21 Jan 2011, Mathieu Desnoyers wrote:
> 
> > > > With:
> > > >   struct cmpxchg_double *pcp
> > >
> > > That does not conform to the parameter conventions in other this_cpu_ops.
> > > The first parameter is a variable because the notion of a pointer is
> > > problematic given that percpu operations use a segment prefix to relocate
> > > pointers.
> >
> > So the first argument could be along the lines of:
> >
> > struct cmpxchg_double pcp
> >
> > then.
> 
> Ok then you would pass a struct by value? Or use a non-scalar as a
> variable passed to a this_cpu_op? So far per cpu scalars have been the
> only variables allowed to be specified in this_cpu operations.

What I have in mind is that the struct passed would be non-scalar for this
specific operation. I'm not sure about the distinction between "pass a struct by
value" and "use a non-scalar as a variable passed to a this_cpu_op" -- I feel
I'm missing an important detail in what you say, because I see these as being
the same thing.

> 
> > > >         struct cmpxchg_double casdbl;
> > > >         struct {
> > > >                 void *ptr;
> > > >                 unsigned long cpuid_tid;
> > > >         } t;
> > > > }
> > >
> > > There is no need for aliases with the existing implementation.
> > >
> > > How will the macro check the parameters now?
> >
> > Well, my last proposal to check __alignof__ within a __builtin_choose_expr
> > check wouldn't need this union actually, which would be much better I think.
> 
> The existing implementation has a check for alignment. That is not the
> problem.

It's a dynamic check right ? (based on VM_BUG_ON() if I remember well) It adds
code and runtime conditions, which would go away if we let the alignment check
be done at compile-time.

> The typechecking would need to be addressed. I.e. if I pass a
> pointer for old and an ulong for the new value then I'd like to see the
> compiler complain. Or if the first parameter is a long but the type of the
> first word is a pointer etc etc.

Hrm. Then the only solution I see would be to require that the structure
used as percpu_dd parameter have fixed field names (yeah, that's a bit odd, but
could not come up with a more elegant solution at the moment):

struct mycustomdoublestruct {
        sometype word1;
        someothertype word2;
}

So we can access percpu_dd.word1 and percpu_dd.word2 within
this_cpu_cmpxchg_double for the type checking.

Mathieu


-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com