seqcount: Introduce raw_write_seqcount_barrier()
authorPeter Zijlstra <[email protected]>
Wed, 17 Jun 2015 12:29:24 +0000 (14:29 +0200)
committerThomas Gleixner <[email protected]>
Thu, 18 Jun 2015 22:09:56 +0000 (00:09 +0200)
commitc4bfa3f5f906aee2e084c5b1fb15caf876338ef8
tree0b6892c0cee030274dbb0b4998e1f854762a0fd5
parenta7c6f571ff51cc77d90dd54968f7c5c938c43998
seqcount: Introduce raw_write_seqcount_barrier()

Introduce raw_write_seqcount_barrier(), a new construct that can be
used to provide write barrier semantics in seqcount read loops instead
of the usual consistency guarantee.

raw_write_seqcount_barier() is equivalent to:

raw_write_seqcount_begin();
raw_write_seqcount_end();

But avoids issueing two back-to-back smp_wmb() instructions.

This construct works because the read side will 'stall' when observing
odd values. This means that -- referring to the example in the comment
below -- even though there is no (matching) read barrier between the
loads of X and Y, we cannot observe !x && !y, because:

 - if we observe Y == false we must observe the first sequence
   increment, which makes us loop, until

 - we observe !(seq & 1) -- the second sequence increment -- at which
   time we must also observe T == true.

Suggested-by: Oleg Nesterov <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: Al Viro <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
include/linux/seqlock.h