Re: [PATCH 1/1] sbitmap: Replace cmpxchg with xchg
From: Omar Sandoval
Date: Mon Jul 01 2019 - 13:35:38 EST
On Sat, Jun 29, 2019 at 08:42:23AM -0700, Pavel Begunkov wrote:
> Ping?
>
> On 23/05/2019 08:39, Pavel Begunkov (Silence) wrote:
> > From: Pavel Begunkov <asml.silence@xxxxxxxxx>
> >
> > cmpxchg() with an immediate value could be replaced with less expensive
> > xchg(). The same true if new value don't _depend_ on the old one.
> >
> > In the second block, atomic_cmpxchg() return value isn't checked, so
> > after atomic_cmpxchg() -> atomic_xchg() conversion it could be replaced
> > with atomic_set(). Comparison with atomic_read() in the second chunk was
> > left as an optimisation (if that was the initial intention).
> >
> > Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
> > ---
> > lib/sbitmap.c | 10 +++-------
> > 1 file changed, 3 insertions(+), 7 deletions(-)
> >
> > diff --git a/lib/sbitmap.c b/lib/sbitmap.c
> > index 155fe38756ec..7d7e0e278523 100644
> > --- a/lib/sbitmap.c
> > +++ b/lib/sbitmap.c
> > @@ -37,9 +37,7 @@ static inline bool sbitmap_deferred_clear(struct sbitmap *sb, int index)
> > /*
> > * First get a stable cleared mask, setting the old mask to 0.
> > */
> > - do {
> > - mask = sb->map[index].cleared;
> > - } while (cmpxchg(&sb->map[index].cleared, mask, 0) != mask);
> > + mask = xchg(&sb->map[index].cleared, 0);
This hunk is clearly correct.
> > /*
> > * Now clear the masked bits in our free word
> > @@ -527,10 +525,8 @@ static struct sbq_wait_state *sbq_wake_ptr(struct sbitmap_queue *sbq)
> > struct sbq_wait_state *ws = &sbq->ws[wake_index];
> >
> > if (waitqueue_active(&ws->wait)) {
> > - int o = atomic_read(&sbq->wake_index);
> > -
> > - if (wake_index != o)
> > - atomic_cmpxchg(&sbq->wake_index, o, wake_index);
> > + if (wake_index != atomic_read(&sbq->wake_index))
> > + atomic_set(&sbq->wake_index, wake_index);
This hunk used to imply a memory barrier and no longer does. I don't
think that's a problem, though.
Reviewed-by: Omar Sandoval <osandov@xxxxxx>