Re: [RFC v2 5/7] powerpc: atomic: Implement cmpxchg{,64}_* and atomic{,64}_cmpxchg_* variants

On Thu, Oct 01, 2015 at 02:27:15PM +0200, Peter Zijlstra wrote:

On Wed, Sep 16, 2015 at 11:49:33PM +0800, Boqun Feng wrote:

quoted

Unlike other atomic operation variants, cmpxchg{,64}_acquire and
atomic{,64}_cmpxchg_acquire don't have acquire semantics if the cmp part
fails, so we need to implement these using assembly.

I think that is actually expected and documented. That is, a cmpxchg
only implies barriers on success. See:

  ed2de9f74ecb ("locking/Documentation: Clarify failed cmpxchg() memory ordering semantics")

Also:

654672d4ba1a6 (Will Deacon     2015-08-06 17:54:37 +0100  28)  * store portion of the operation. Note that a failed cmpxchg_acquire
654672d4ba1a6 (Will Deacon     2015-08-06 17:54:37 +0100  29)  * does -not- imply any memory ordering constraints.