[PATCH 0/3] Fix atomic64_add_unless return value convention

[PATCH 0/3] Fix atomic64_add_unless return value convention

[Luca Barbieri]

Return values 0 and 1 were swapped both in the generic implementation and
my x86-32 one.

The x86-32 and atomic64_test.c patches should probably be squashed in the
appropriate commits in tip/x86/atomic

Luca Barbieri (3):
  lib: fix atomic64_add_unless test
  x86: fix atomic64_add_unless return value convention
  lib: fix atomic64_add_unless return value convention

 arch/x86/lib/atomic64_386_32.S |    4 ++--
 arch/x86/lib/atomic64_cx8_32.S |    4 ++--
 lib/atomic64.c                 |    4 ++--
 lib/atomic64_test.c            |    4 ++--
 4 files changed, 8 insertions(+), 8 deletions(-)

[PATCH 1/3] lib: fix atomic64_add_unless test

[Luca Barbieri]

Was buggy as it replicated the bug in lib/atomic64.c

Reported-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Luca Barbieri <luca@luca-barbieri.com>
---
 lib/atomic64_test.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/atomic64_test.c b/lib/atomic64_test.c
index 58efdab..ee8e6de 100644
--- a/lib/atomic64_test.c
+++ b/lib/atomic64_test.c
@@ -104,11 +104,11 @@ static __init int test_atomic64(void)
 	BUG_ON(v.counter != r);
 
 	INIT(v0);
-	BUG_ON(!atomic64_add_unless(&v, one, v0));
+	BUG_ON(atomic64_add_unless(&v, one, v0));
 	BUG_ON(v.counter != r);
 
 	INIT(v0);
-	BUG_ON(atomic64_add_unless(&v, one, v1));
+	BUG_ON(!atomic64_add_unless(&v, one, v1));
 	r += one;
 	BUG_ON(v.counter != r);
 
-- 
1.6.6.1.476.g01ddb

[PATCH 2/3] x86: fix atomic64_add_unless return value convention

[Luca Barbieri]

Reported-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Luca Barbieri <luca@luca-barbieri.com>
---
 arch/x86/lib/atomic64_386_32.S |    4 ++--
 arch/x86/lib/atomic64_cx8_32.S |    4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/lib/atomic64_386_32.S b/arch/x86/lib/atomic64_386_32.S
index 5db07fe..a2f847c 100644
--- a/arch/x86/lib/atomic64_386_32.S
+++ b/arch/x86/lib/atomic64_386_32.S
@@ -133,13 +133,13 @@ BEGIN add_unless %ecx
 1:
 	movl %eax,  ($v)
 	movl %edx, 4($v)
-	xorl %eax, %eax
+	movl $1, %eax
 2:
 RETURN
 3:
 	cmpl %edx, %edi
 	jne 1b
-	movl $1, %eax
+	xorl %eax, %eax
 	jmp 2b
 END_
 
diff --git a/arch/x86/lib/atomic64_cx8_32.S b/arch/x86/lib/atomic64_cx8_32.S
index e49c4eb..d0e37b1 100644
--- a/arch/x86/lib/atomic64_cx8_32.S
+++ b/arch/x86/lib/atomic64_cx8_32.S
@@ -180,7 +180,7 @@ ENTRY(atomic64_add_unless_cx8)
 	cmpxchg8b (%ebp)
 	jne 1b
 
-	xorl %eax, %eax
+	movl $1, %eax
 3:
 	addl $8, %esp
 	CFI_ADJUST_CFA_OFFSET -8
@@ -190,7 +190,7 @@ ENTRY(atomic64_add_unless_cx8)
 4:
 	cmpl %edx, 4(%esp)
 	jne 2b
-	movl $1, %eax
+	xorl %eax, %eax
 	jmp 3b
 	CFI_ENDPROC
 ENDPROC(atomic64_add_unless_cx8)
-- 
1.6.6.1.476.g01ddb

[PATCH 3/3] lib: fix atomic64_add_unless return value convention

[Luca Barbieri]

The 0 and 1 return values were incorrectly swapped.

Reported-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Luca Barbieri <luca@luca-barbieri.com>
---
 lib/atomic64.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/atomic64.c b/lib/atomic64.c
index 8bee16e..a21c12b 100644
--- a/lib/atomic64.c
+++ b/lib/atomic64.c
@@ -162,12 +162,12 @@ int atomic64_add_unless(atomic64_t *v, long long a, long long u)
 {
 	unsigned long flags;
 	spinlock_t *lock = lock_addr(v);
-	int ret = 1;
+	int ret = 0;
 
 	spin_lock_irqsave(lock, flags);
 	if (v->counter != u) {
 		v->counter += a;
-		ret = 0;
+		ret = 1;
 	}
 	spin_unlock_irqrestore(lock, flags);
 	return ret;
-- 
1.6.6.1.476.g01ddb

Re: [PATCH 0/3] Fix atomic64_add_unless return value convention

[H. Peter Anvin]

On 03/01/2010 09:14 AM, Luca Barbieri wrote:
> Return values 0 and 1 were swapped both in the generic implementation and
> my x86-32 one.
> 
> The x86-32 and atomic64_test.c patches should probably be squashed in the
> appropriate commits in tip/x86/atomic

We tend to be very reluctant to do that.

Anyway, please resend your patches with actual patch descriptions
(commit messages).

	-hpa

Re: [PATCH 0/3] Fix atomic64_add_unless return value convention

[Luca Barbieri]

>> The x86-32 and atomic64_test.c patches should probably be squashed in the
>> appropriate commits in tip/x86/atomic
>
> We tend to be very reluctant to do that.
OK: I suggested that to avoid introducing broken commits that could
cause annoyance in bisections.

> Anyway, please resend your patches with actual patch descriptions
> (commit messages).
Resent with more verbose descriptions.

Re: [PATCH 3/3] lib: fix atomic64_add_unless return value convention

[Paul Mackerras]

On Mon, Mar 01, 2010 at 06:15:00PM +0100, Luca Barbieri wrote:
> The 0 and 1 return values were incorrectly swapped.
> 
> Reported-by: H. Peter Anvin <hpa@zytor.com>
> Signed-off-by: Luca Barbieri <luca@luca-barbieri.com>

Acked-by: Paul Mackerras <paulus@samba.org>