On Wed, Jun 01, 2016 at 11:11:38AM +0800, Boqun Feng wrote:

Hi Peter,

On Tue, May 31, 2016 at 12:19:44PM +0200, Peter Zijlstra wrote:
[snip]

quoted

@@ -329,20 +361,53 @@ atomic64_##op##_return_relaxed(long a, a
 	return t;							\
 }
 
+#define ATOMIC64_FETCH_OP_RELAXED(op, asm_op)				\
+static inline long							\
+atomic64_fetch_##op##_relaxed(long a, atomic64_t *v)			\
+{									\
+	long res, t;							\
+									\
+	__asm__ __volatile__(						\
+"1:	ldarx	%0,0,%4		# atomic64_fetch_" #op "_relaxed\n"	\
+	#asm_op " %1,%3,%0\n"						\
+"	stdcx.	%1,0,%4\n"						\
+"	bne-	1b\n"							\
+	: "=&r" (res), "=&r" (t), "+m" (v->counter)			\
+	: "r" (a), "r" (&v->counter)					\
+	: "cc");							\
+									\
+	return t;							\

Looks like I missed this one in v1, it should be
	
	return res;

because the primitives will return the values before modified by the
operations.

FWIW, I tested on ppc with ATOMIC64_SELFTEST=y for the following branch:

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git locking/atomic (afee54ef5b1f2b04)

without this modification, I can hit:

------------[ cut here ]------------
kernel BUG at lib/atomic64_test.c:181!
Oops: Exception in kernel mode, sig: 5 [#1]

with this modification, all the atomic selftests are passed ;-)

Regards,
Boqun