Re: [PATCH] seqlock: Use while instead of if+goto in __read_seqcount_begin

[Will Deacon <will@kernel.org>]

On Tue, Apr 14, 2020 at 01:05:16PM +0200, Peter Zijlstra wrote:
> On Fri, Apr 10, 2020 at 12:56:58PM +0100, Will Deacon wrote:
> > On Thu, Apr 09, 2020 at 09:45:58PM +0800, Muchun Song wrote:
> > > The creators of the C language gave us the while keyword. Let's use
> > > that instead of synthesizing it from if+goto.
> > > 
> > > Signed-off-by: Muchun Song <songmuchun@bytedance.com>
> > > ---
> > >  include/linux/seqlock.h | 6 +-----
> > >  1 file changed, 1 insertion(+), 5 deletions(-)
> > > 
> > > diff --git a/include/linux/seqlock.h b/include/linux/seqlock.h
> > > index 8b97204f35a77..7bdea019814ce 100644
> > > --- a/include/linux/seqlock.h
> > > +++ b/include/linux/seqlock.h
> > > @@ -125,12 +125,8 @@ static inline unsigned __read_seqcount_begin(const seqcount_t *s)
> > >  {
> > >  	unsigned ret;
> > >  
> > > -repeat:
> > > -	ret = READ_ONCE(s->sequence);
> > > -	if (unlikely(ret & 1)) {
> > > +	while (unlikely((ret = READ_ONCE(s->sequence)) & 1))
> > >  		cpu_relax();
> > > -		goto repeat;
> > > -	}
> > >  	kcsan_atomic_next(KCSAN_SEQLOCK_REGION_MAX);
> > >  	return ret;
> > 
> > Patch looks fine to me, but I'll leave it to Peter as I don't have a
> > preference either way.
> 
> Linus sometimes prefers the goto variant as that better expresses the
> exception model. But like Will, I don't particularly care. That said,
> Will, would it make sense to use smp_cond_load_relaxed() here ?

Oh yeah, good thinking. Didn't spot that one, but it should work well as
long as smp_cond_load_relaxed() always implies a control dependency (surely
it has to?)

Will