[PATCH 1/5] asm-generic: add smp_vcond_load_relaxed()

[Okanovic, Haris]

On Wed, 2024-11-06 at 11:39 +0000, Will Deacon wrote:
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.
> 
> 
> 
> On Tue, Nov 05, 2024 at 12:30:37PM -0600, Haris Okanovic wrote:
> > Relaxed poll until desired mask/value is observed at the specified
> > address or timeout.
> > 
> > This macro is a specialization of the generic smp_cond_load_relaxed(),
> > which takes a simple mask/value condition (vcond) instead of an
> > arbitrary expression. It allows architectures to better specialize the
> > implementation, e.g. to enable wfe() polling of the address on arm.
> 
> This doesn't make sense to me. The existing smp_cond_load() functions
> already use wfe on arm64 and I don't see why we need a special helper
> just to do a mask.

We can't turn an arbitrary C expression into a wfe()/wfet() exit
condition, which is one of the inputs to the existing smp_cond_load().
This API is therefore more amenable to hardware acceleration.

> 
> > Signed-off-by: Haris Okanovic <harisokn at amazon.com>
> > ---
> >  include/asm-generic/barrier.h | 25 +++++++++++++++++++++++++
> >  1 file changed, 25 insertions(+)
> > 
> > diff --git a/include/asm-generic/barrier.h b/include/asm-generic/barrier.h
> > index d4f581c1e21d..112027eabbfc 100644
> > --- a/include/asm-generic/barrier.h
> > +++ b/include/asm-generic/barrier.h
> > @@ -256,6 +256,31 @@ do {                                                                     \
> >  })
> >  #endif
> > 
> > +/**
> > + * smp_vcond_load_relaxed() - (Spin) wait until an expected value at address
> > + * with no ordering guarantees. Spins until `(*addr & mask) == val` or
> > + * `nsecs` elapse, and returns the last observed `*addr` value.
> > + *
> > + * @nsecs: timeout in nanoseconds
> > + * @addr: pointer to an integer
> > + * @mask: a bit mask applied to read values
> > + * @val: Expected value with mask
> > + */
> > +#ifndef smp_vcond_load_relaxed
> 
> I know naming is hard, but "vcond" is especially terrible.
> Perhaps smp_cond_load_timeout()?

I agree, naming is hard! I was trying to differentiate it from
smp_cond_load() in some meaningful way - that one is an "expression"
condition this one is a "value" condition.

I'll think it over a bit more.

> 
> Will