Re: [PATCH 1/2] locking: Use spin primitives for busy loops

2017-09-05 Thread Nicholas Piggin
Hi Peter, Thanks for the very good reviews, sorry been a bit busy and I figure this didn't get time to make 4.14 so I'll take your comments and submit again for next window. On Fri, 1 Sep 2017 14:23:20 +0200 Peter Zijlstra wrote: > This really should have been many patches. Sure, will do. >

Re: [PATCH 1/2] locking: Use spin primitives for busy loops

2017-09-01 Thread Peter Zijlstra
This really should have been many patches. On Sun, Aug 20, 2017 at 07:25:01PM +1000, Nicholas Piggin wrote: > @@ -108,12 +109,10 @@ static inline unsigned __read_seqcount_begin(const > seqcount_t *s) > { > unsigned ret; > > -repeat: > ret = READ_ONCE(s->sequence); > - if (unl

[PATCH 1/2] locking: Use spin primitives for busy loops

2017-08-20 Thread Nicholas Piggin
Commit fd851a3cdc ("spin loop primitives for busy waiting") introduced a begin/relax/end sequence for busy loops, to improve behaviour with some architectures. Convert most of the generic locking primitives over to use these spin primitives. Signed-off-by: Nicholas Piggin --- include/linux/bit_