kref: prefer atomic_inc_not_zero to atomic_add_unless
authorJason A. Donenfeld <Jason@zx2c4.com>
Thu, 15 Dec 2016 05:01:10 +0000 (06:01 +0100)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Fri, 16 Dec 2016 07:33:40 +0000 (08:33 +0100)
commit56e3d1cd05cc7b24cfcae8714b0661bf607aaba3
tree7cbfed2e37d10b0201057cb4fecf75f3f545e84c
parentc02f39ac00a66a75e55f84acdd7939cb37709085
kref: prefer atomic_inc_not_zero to atomic_add_unless

On most platforms, there exists this ifdef:

 #define atomic_inc_not_zero(v) atomic_add_unless((v), 1, 0)

This makes this patch functionally useless. However, on PPC, there is
actually an explicit definition of atomic_inc_not_zero with its own
assembly that is slightly more optimized than atomic_add_unless. So,
this patch changes kref to use atomic_inc_not_zero instead, for PPC and
any future platforms that might provide an explicit implementation.

This also puts this usage of kref more in line with a verbatim reading
of the examples in Paul McKenney's paper [1] in the section titled "2.4
Atomic Counting With Check and Release Memory Barrier", which uses
atomic_inc_not_zero.

[1] http://open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2167.pdf

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161215050110.3241-1-Jason@zx2c4.com
include/linux/kref.h