Merge branch 'core-futexes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'core-futexes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  arm: Remove bogus comment in futex_atomic_cmpxchg_inatomic()
  futex: Deobfuscate handle_futex_death()
  plist: Add priority list test
  plist: Shrink struct plist_head
  futex,plist: Remove debug lock assignment from plist_node
  futex,plist: Pass the real head of the priority list to plist_del()
  futex: Sanitize futex ops argument types
  futex: Sanitize cmpxchg_futex_value_locked API
  futex: Remove redundant pagefault_disable in futex_atomic_cmpxchg_inatomic()
  futex: Avoid redudant evaluation of task_pid_vnr()
  futex: Update futex_wait_setup comments about locking

19 files changed, 197 insertions, 169 deletions

diff --git a/arch/alpha/include/asm/futex.h b/arch/alpha/include/asm/futex.h
index 945de222ab91..e8a761aee088 100644
--- a/arch/alpha/include/asm/futex.h
+++ b/arch/alpha/include/asm/futex.h
@@ -29,7 +29,7 @@
 	:	"r" (uaddr), "r"(oparg)				\
 	:	"memory")
 
-static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -39,7 +39,7 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -81,21 +81,23 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	int prev, cmp;
+	int ret = 0, cmp;
+	u32 prev;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	__asm__ __volatile__ (
 		__ASM_SMP_MB
-	"1:	ldl_l	%0,0(%2)\n"
-	"	cmpeq	%0,%3,%1\n"
-	"	beq	%1,3f\n"
-	"	mov	%4,%1\n"
-	"2:	stl_c	%1,0(%2)\n"
-	"	beq	%1,4f\n"
+	"1:	ldl_l	%1,0(%3)\n"
+	"	cmpeq	%1,%4,%2\n"
+	"	beq	%2,3f\n"
+	"	mov	%5,%2\n"
+	"2:	stl_c	%2,0(%3)\n"
+	"	beq	%2,4f\n"
 	"3:	.subsection 2\n"
 	"4:	br	1b\n"
 	"	.previous\n"
@@ -105,11 +107,12 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
 	"	.long	2b-.\n"
 	"	lda	$31,3b-2b(%0)\n"
 	"	.previous\n"
-	:	"=&r"(prev), "=&r"(cmp)
+	:	"+r"(ret), "=&r"(prev), "=&r"(cmp)
 	:	"r"(uaddr), "r"((long)oldval), "r"(newval)
 	:	"memory");
 
-	return prev;
+	*uval = prev;
+	return ret;
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/arm/include/asm/futex.h b/arch/arm/include/asm/futex.h
index b33fe7065b38..199a6b6de7f4 100644
--- a/arch/arm/include/asm/futex.h
+++ b/arch/arm/include/asm/futex.h
@@ -35,7 +35,7 @@
 	: "cc", "memory")
 
 static inline int
-futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -46,7 +46,7 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();	/* implies preempt_disable() */
@@ -88,36 +88,35 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	int val;
+	int ret = 0;
+	u32 val;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
-	pagefault_disable();	/* implies preempt_disable() */
-
 	__asm__ __volatile__("@futex_atomic_cmpxchg_inatomic\n"
-	"1:	" T(ldr) "	%0, [%3]\n"
-	"	teq	%0, %1\n"
+	"1:	" T(ldr) "	%1, [%4]\n"
+	"	teq	%1, %2\n"
 	"	it	eq	@ explicit IT needed for the 2b label\n"
-	"2:	" T(streq) "	%2, [%3]\n"
+	"2:	" T(streq) "	%3, [%4]\n"
 	"3:\n"
 	"	.pushsection __ex_table,\"a\"\n"
 	"	.align	3\n"
 	"	.long	1b, 4f, 2b, 4f\n"
 	"	.popsection\n"
 	"	.pushsection .fixup,\"ax\"\n"
-	"4:	mov	%0, %4\n"
+	"4:	mov	%0, %5\n"
 	"	b	3b\n"
 	"	.popsection"
-	: "=&r" (val)
+	: "+r" (ret), "=&r" (val)
 	: "r" (oldval), "r" (newval), "r" (uaddr), "Ir" (-EFAULT)
 	: "cc", "memory");
 
-	pagefault_enable();	/* subsumes preempt_enable() */
-
-	return val;
+	*uval = val;
+	return ret;
 }
 
 #endif /* !SMP */
diff --git a/arch/frv/include/asm/futex.h b/arch/frv/include/asm/futex.h
index 08b3d1da3583..4bea27f50a7a 100644
--- a/arch/frv/include/asm/futex.h
+++ b/arch/frv/include/asm/futex.h
@@ -7,10 +7,11 @@
 #include <asm/errno.h>
 #include <asm/uaccess.h>
 
-extern int futex_atomic_op_inuser(int encoded_op, int __user *uaddr);
+extern int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr);
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
 	return -ENOSYS;
 }
diff --git a/arch/frv/kernel/futex.c b/arch/frv/kernel/futex.c
index 14f64b054c7e..d155ca9e5098 100644
--- a/arch/frv/kernel/futex.c
+++ b/arch/frv/kernel/futex.c
@@ -18,7 +18,7 @@
  * the various futex operations; MMU fault checking is ignored under no-MMU
  * conditions
  */
-static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr, int *_oldval)
+static inline int atomic_futex_op_xchg_set(int oparg, u32 __user *uaddr, int *_oldval)
 {
 	int oldval, ret;
 
@@ -50,7 +50,7 @@ static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr, int *_o
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr, int *_oldval)
+static inline int atomic_futex_op_xchg_add(int oparg, u32 __user *uaddr, int *_oldval)
 {
 	int oldval, ret;
 
@@ -83,7 +83,7 @@ static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr, int *_o
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr, int *_oldval)
+static inline int atomic_futex_op_xchg_or(int oparg, u32 __user *uaddr, int *_oldval)
 {
 	int oldval, ret;
 
@@ -116,7 +116,7 @@ static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr, int *_ol
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr, int *_oldval)
+static inline int atomic_futex_op_xchg_and(int oparg, u32 __user *uaddr, int *_oldval)
 {
 	int oldval, ret;
 
@@ -149,7 +149,7 @@ static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr, int *_o
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr, int *_oldval)
+static inline int atomic_futex_op_xchg_xor(int oparg, u32 __user *uaddr, int *_oldval)
 {
 	int oldval, ret;
 
@@ -186,7 +186,7 @@ static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr, int *_o
 /*
  * do the futex operations
  */
-int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -197,7 +197,7 @@ int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
diff --git a/arch/ia64/include/asm/futex.h b/arch/ia64/include/asm/futex.h
index c7f0f062239c..8428525ddb22 100644
--- a/arch/ia64/include/asm/futex.h
+++ b/arch/ia64/include/asm/futex.h
@@ -46,7 +46,7 @@ do {									\
 } while (0)
 
 static inline int
-futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -56,7 +56,7 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -100,23 +100,26 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	{
-		register unsigned long r8 __asm ("r8");
+		register unsigned long r8 __asm ("r8") = 0;
+		unsigned long prev;
 		__asm__ __volatile__(
 			"	mf;;					\n"
 			"	mov ar.ccv=%3;;				\n"
 			"[1:]	cmpxchg4.acq %0=[%1],%2,ar.ccv		\n"
 			"	.xdata4 \"__ex_table\", 1b-., 2f-.	\n"
 			"[2:]"
-			: "=r" (r8)
+			: "=r" (prev)
 			: "r" (uaddr), "r" (newval),
 			  "rO" ((long) (unsigned) oldval)
 			: "memory");
+		*uval = prev;
 		return r8;
 	}
 }
diff --git a/arch/microblaze/include/asm/futex.h b/arch/microblaze/include/asm/futex.h
index ad3fd61b2fe7..b0526d2716fa 100644
--- a/arch/microblaze/include/asm/futex.h
+++ b/arch/microblaze/include/asm/futex.h
@@ -29,7 +29,7 @@
 })
 
 static inline int
-futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -39,7 +39,7 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -94,31 +94,34 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	int prev, cmp;
+	int ret = 0, cmp;
+	u32 prev;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
-	__asm__ __volatile__ ("1:	lwx	%0, %2, r0;		\
-					cmp	%1, %0, %3;		\
-					beqi	%1, 3f;			\
-				2:	swx	%4, %2, r0;		\
-					addic	%1, r0, 0;		\
-					bnei	%1, 1b;			\
+	__asm__ __volatile__ ("1:	lwx	%1, %3, r0;		\
+					cmp	%2, %1, %4;		\
+					beqi	%2, 3f;			\
+				2:	swx	%5, %3, r0;		\
+					addic	%2, r0, 0;		\
+					bnei	%2, 1b;			\
 				3:					\
 				.section .fixup,\"ax\";			\
 				4:	brid	3b;			\
-					addik	%0, r0, %5;		\
+					addik	%0, r0, %6;		\
 				.previous;				\
 				.section __ex_table,\"a\";		\
 				.word	1b,4b,2b,4b;			\
 				.previous;"				\
-		: "=&r" (prev), "=&r"(cmp)				\
+		: "+r" (ret), "=&r" (prev), "=&r"(cmp)	\
 		: "r" (uaddr), "r" (oldval), "r" (newval), "i" (-EFAULT));
 
-	return prev;
+	*uval = prev;
+	return ret;
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/mips/include/asm/futex.h b/arch/mips/include/asm/futex.h
index b9cce90346cf..6ebf1734b411 100644
--- a/arch/mips/include/asm/futex.h
+++ b/arch/mips/include/asm/futex.h
@@ -75,7 +75,7 @@
 }
 
 static inline int
-futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -85,7 +85,7 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -132,11 +132,13 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	int retval;
+	int ret = 0;
+	u32 val;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	if (cpu_has_llsc && R10000_LLSC_WAR) {
@@ -145,25 +147,25 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
 		"	.set	push					\n"
 		"	.set	noat					\n"
 		"	.set	mips3					\n"
-		"1:	ll	%0, %2					\n"
-		"	bne	%0, %z3, 3f				\n"
+		"1:	ll	%1, %3					\n"
+		"	bne	%1, %z4, 3f				\n"
 		"	.set	mips0					\n"
-		"	move	$1, %z4					\n"
+		"	move	$1, %z5					\n"
 		"	.set	mips3					\n"
-		"2:	sc	$1, %1					\n"
+		"2:	sc	$1, %2					\n"
 		"	beqzl	$1, 1b					\n"
 		__WEAK_LLSC_MB
 		"3:							\n"
 		"	.set	pop					\n"
 		"	.section .fixup,\"ax\"				\n"
-		"4:	li	%0, %5					\n"
+		"4:	li	%0, %6					\n"
 		"	j	3b					\n"
 		"	.previous					\n"
 		"	.section __ex_table,\"a\"			\n"
 		"	"__UA_ADDR "\t1b, 4b				\n"
 		"	"__UA_ADDR "\t2b, 4b				\n"
 		"	.previous					\n"
-		: "=&r" (retval), "=R" (*uaddr)
+		: "+r" (ret), "=&r" (val), "=R" (*uaddr)
 		: "R" (*uaddr), "Jr" (oldval), "Jr" (newval), "i" (-EFAULT)
 		: "memory");
 	} else if (cpu_has_llsc) {
@@ -172,31 +174,32 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
 		"	.set	push					\n"
 		"	.set	noat					\n"
 		"	.set	mips3					\n"
-		"1:	ll	%0, %2					\n"
-		"	bne	%0, %z3, 3f				\n"
+		"1:	ll	%1, %3					\n"
+		"	bne	%1, %z4, 3f				\n"
 		"	.set	mips0					\n"
-		"	move	$1, %z4					\n"
+		"	move	$1, %z5					\n"
 		"	.set	mips3					\n"
-		"2:	sc	$1, %1					\n"
+		"2:	sc	$1, %2					\n"
 		"	beqz	$1, 1b					\n"
 		__WEAK_LLSC_MB
 		"3:							\n"
 		"	.set	pop					\n"
 		"	.section .fixup,\"ax\"				\n"
-		"4:	li	%0, %5					\n"
+		"4:	li	%0, %6					\n"
 		"	j	3b					\n"
 		"	.previous					\n"
 		"	.section __ex_table,\"a\"			\n"
 		"	"__UA_ADDR "\t1b, 4b				\n"
 		"	"__UA_ADDR "\t2b, 4b				\n"
 		"	.previous					\n"
-		: "=&r" (retval), "=R" (*uaddr)
+		: "+r" (ret), "=&r" (val), "=R" (*uaddr)
 		: "R" (*uaddr), "Jr" (oldval), "Jr" (newval), "i" (-EFAULT)
 		: "memory");
 	} else
 		return -ENOSYS;
 
-	return retval;
+	*uval = val;
+	return ret;
 }
 
 #endif
diff --git a/arch/parisc/include/asm/futex.h b/arch/parisc/include/asm/futex.h
index 0c705c3a55ef..67a33cc27ef2 100644
--- a/arch/parisc/include/asm/futex.h
+++ b/arch/parisc/include/asm/futex.h
@@ -8,7 +8,7 @@
 #include <asm/errno.h>
 
 static inline int
-futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -18,7 +18,7 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -51,10 +51,10 @@ futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 
 /* Non-atomic version */
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	int err = 0;
-	int uval;
+	u32 val;
 
 	/* futex.c wants to do a cmpxchg_inatomic on kernel NULL, which is
 	 * our gateway page, and causes no end of trouble...
@@ -62,15 +62,15 @@ futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
 	if (segment_eq(KERNEL_DS, get_fs()) && !uaddr)
 		return -EFAULT;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
-	err = get_user(uval, uaddr);
-	if (err) return -EFAULT;
-	if (uval == oldval)
-		err = put_user(newval, uaddr);
-	if (err) return -EFAULT;
-	return uval;
+	if (get_user(val, uaddr))
+		return -EFAULT;
+	if (val == oldval && put_user(newval, uaddr))
+		return -EFAULT;
+	*uval = val;
+	return 0;
 }
 
 #endif /*__KERNEL__*/
diff --git a/arch/powerpc/include/asm/futex.h b/arch/powerpc/include/asm/futex.h
index 7c589ef81fb0..c94e4a3fe2ef 100644
--- a/arch/powerpc/include/asm/futex.h
+++ b/arch/powerpc/include/asm/futex.h
@@ -30,7 +30,7 @@
 	: "b" (uaddr), "i" (-EFAULT), "r" (oparg) \
 	: "cr0", "memory")
 
-static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -40,7 +40,7 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -82,35 +82,38 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	int prev;
+	int ret = 0;
+	u32 prev;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
         __asm__ __volatile__ (
         PPC_RELEASE_BARRIER
-"1:     lwarx   %0,0,%2         # futex_atomic_cmpxchg_inatomic\n\
-        cmpw    0,%0,%3\n\
+"1:     lwarx   %1,0,%3         # futex_atomic_cmpxchg_inatomic\n\
+        cmpw    0,%1,%4\n\
         bne-    3f\n"
-        PPC405_ERR77(0,%2)
-"2:     stwcx.  %4,0,%2\n\
+        PPC405_ERR77(0,%3)
+"2:     stwcx.  %5,0,%3\n\
         bne-    1b\n"
         PPC_ACQUIRE_BARRIER
 "3:	.section .fixup,\"ax\"\n\
-4:	li	%0,%5\n\
+4:	li	%0,%6\n\
 	b	3b\n\
 	.previous\n\
 	.section __ex_table,\"a\"\n\
 	.align 3\n\
 	" PPC_LONG "1b,4b,2b,4b\n\
 	.previous" \
-        : "=&r" (prev), "+m" (*uaddr)
+        : "+r" (ret), "=&r" (prev), "+m" (*uaddr)
         : "r" (uaddr), "r" (oldval), "r" (newval), "i" (-EFAULT)
         : "cc", "memory");
 
-        return prev;
+	*uval = prev;
+        return ret;
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/s390/include/asm/futex.h b/arch/s390/include/asm/futex.h
index 5c5d02de49e9..81cf36b691f1 100644
--- a/arch/s390/include/asm/futex.h
+++ b/arch/s390/include/asm/futex.h
@@ -7,7 +7,7 @@
 #include <linux/uaccess.h>
 #include <asm/errno.h>
 
-static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser (int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -18,7 +18,7 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -39,13 +39,13 @@ static inline int futex_atomic_op_inuser (int encoded_op, int __user *uaddr)
 	return ret;
 }
 
-static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr,
-						int oldval, int newval)
+static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+						u32 oldval, u32 newval)
 {
-	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(int)))
+	if (! access_ok (VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
-	return uaccess.futex_atomic_cmpxchg(uaddr, oldval, newval);
+	return uaccess.futex_atomic_cmpxchg(uval, uaddr, oldval, newval);
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/s390/include/asm/uaccess.h b/arch/s390/include/asm/uaccess.h
index d6b1ed0ec52b..2d9ea11f919a 100644
--- a/arch/s390/include/asm/uaccess.h
+++ b/arch/s390/include/asm/uaccess.h
@@ -83,8 +83,8 @@ struct uaccess_ops {
 	size_t (*clear_user)(size_t, void __user *);
 	size_t (*strnlen_user)(size_t, const char __user *);
 	size_t (*strncpy_from_user)(size_t, const char __user *, char *);
-	int (*futex_atomic_op)(int op, int __user *, int oparg, int *old);
-	int (*futex_atomic_cmpxchg)(int __user *, int old, int new);
+	int (*futex_atomic_op)(int op, u32 __user *, int oparg, int *old);
+	int (*futex_atomic_cmpxchg)(u32 *, u32 __user *, u32 old, u32 new);
 };
 
 extern struct uaccess_ops uaccess;
diff --git a/arch/s390/lib/uaccess.h b/arch/s390/lib/uaccess.h
index 126011df14f1..1d2536cb630b 100644
--- a/arch/s390/lib/uaccess.h
+++ b/arch/s390/lib/uaccess.h
@@ -12,12 +12,12 @@ extern size_t copy_from_user_std(size_t, const void __user *, void *);
 extern size_t copy_to_user_std(size_t, void __user *, const void *);
 extern size_t strnlen_user_std(size_t, const char __user *);
 extern size_t strncpy_from_user_std(size_t, const char __user *, char *);
-extern int futex_atomic_cmpxchg_std(int __user *, int, int);
-extern int futex_atomic_op_std(int, int __user *, int, int *);
+extern int futex_atomic_cmpxchg_std(u32 *, u32 __user *, u32, u32);
+extern int futex_atomic_op_std(int, u32 __user *, int, int *);
 
 extern size_t copy_from_user_pt(size_t, const void __user *, void *);
 extern size_t copy_to_user_pt(size_t, void __user *, const void *);
-extern int futex_atomic_op_pt(int, int __user *, int, int *);
-extern int futex_atomic_cmpxchg_pt(int __user *, int, int);
+extern int futex_atomic_op_pt(int, u32 __user *, int, int *);
+extern int futex_atomic_cmpxchg_pt(u32 *, u32 __user *, u32, u32);
 
 #endif /* __ARCH_S390_LIB_UACCESS_H */
diff --git a/arch/s390/lib/uaccess_pt.c b/arch/s390/lib/uaccess_pt.c
index 404f2de296dc..74833831417f 100644
--- a/arch/s390/lib/uaccess_pt.c
+++ b/arch/s390/lib/uaccess_pt.c
@@ -302,7 +302,7 @@ fault:
 		     : "0" (-EFAULT), "d" (oparg), "a" (uaddr),		\
 		       "m" (*uaddr) : "cc" );
 
-static int __futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
+static int __futex_atomic_op_pt(int op, u32 __user *uaddr, int oparg, int *old)
 {
 	int oldval = 0, newval, ret;
 
@@ -335,7 +335,7 @@ static int __futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
 	return ret;
 }
 
-int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
+int futex_atomic_op_pt(int op, u32 __user *uaddr, int oparg, int *old)
 {
 	int ret;
 
@@ -354,26 +354,29 @@ int futex_atomic_op_pt(int op, int __user *uaddr, int oparg, int *old)
 	return ret;
 }
 
-static int __futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
+static int __futex_atomic_cmpxchg_pt(u32 *uval, u32 __user *uaddr,
+				     u32 oldval, u32 newval)
 {
 	int ret;
 
 	asm volatile("0: cs   %1,%4,0(%5)\n"
-		     "1: lr   %0,%1\n"
+		     "1: la   %0,0\n"
 		     "2:\n"
 		     EX_TABLE(0b,2b) EX_TABLE(1b,2b)
 		     : "=d" (ret), "+d" (oldval), "=m" (*uaddr)
 		     : "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
 		     : "cc", "memory" );
+	*uval = oldval;
 	return ret;
 }
 
-int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
+int futex_atomic_cmpxchg_pt(u32 *uval, u32 __user *uaddr,
+			    u32 oldval, u32 newval)
 {
 	int ret;
 
 	if (segment_eq(get_fs(), KERNEL_DS))
-		return __futex_atomic_cmpxchg_pt(uaddr, oldval, newval);
+		return __futex_atomic_cmpxchg_pt(uval, uaddr, oldval, newval);
 	spin_lock(&current->mm->page_table_lock);
 	uaddr = (int __user *) __dat_user_addr((unsigned long) uaddr);
 	if (!uaddr) {
@@ -382,7 +385,7 @@ int futex_atomic_cmpxchg_pt(int __user *uaddr, int oldval, int newval)
 	}
 	get_page(virt_to_page(uaddr));
 	spin_unlock(&current->mm->page_table_lock);
-	ret = __futex_atomic_cmpxchg_pt(uaddr, oldval, newval);
+	ret = __futex_atomic_cmpxchg_pt(uval, uaddr, oldval, newval);
 	put_page(virt_to_page(uaddr));
 	return ret;
 }
diff --git a/arch/s390/lib/uaccess_std.c b/arch/s390/lib/uaccess_std.c
index a6c4f7ed24a4..bb1a7eed42ce 100644
--- a/arch/s390/lib/uaccess_std.c
+++ b/arch/s390/lib/uaccess_std.c
@@ -255,7 +255,7 @@ size_t strncpy_from_user_std(size_t size, const char __user *src, char *dst)
 		: "0" (-EFAULT), "d" (oparg), "a" (uaddr),		\
 		  "m" (*uaddr) : "cc");
 
-int futex_atomic_op_std(int op, int __user *uaddr, int oparg, int *old)
+int futex_atomic_op_std(int op, u32 __user *uaddr, int oparg, int *old)
 {
 	int oldval = 0, newval, ret;
 
@@ -287,19 +287,21 @@ int futex_atomic_op_std(int op, int __user *uaddr, int oparg, int *old)
 	return ret;
 }
 
-int futex_atomic_cmpxchg_std(int __user *uaddr, int oldval, int newval)
+int futex_atomic_cmpxchg_std(u32 *uval, u32 __user *uaddr,
+			     u32 oldval, u32 newval)
 {
 	int ret;
 
 	asm volatile(
 		"   sacf 256\n"
 		"0: cs   %1,%4,0(%5)\n"
-		"1: lr   %0,%1\n"
+		"1: la   %0,0\n"
 		"2: sacf 0\n"
 		EX_TABLE(0b,2b) EX_TABLE(1b,2b)
 		: "=d" (ret), "+d" (oldval), "=m" (*uaddr)
 		: "0" (-EFAULT), "d" (newval), "a" (uaddr), "m" (*uaddr)
 		: "cc", "memory" );
+	*uval = oldval;
 	return ret;
 }
 
diff --git a/arch/sh/include/asm/futex-irq.h b/arch/sh/include/asm/futex-irq.h
index a9f16a7f9aea..6cb9f193a95e 100644
--- a/arch/sh/include/asm/futex-irq.h
+++ b/arch/sh/include/asm/futex-irq.h
@@ -3,7 +3,7 @@
 
 #include <asm/system.h>
 
-static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr,
+static inline int atomic_futex_op_xchg_set(int oparg, u32 __user *uaddr,
 					   int *oldval)
 {
 	unsigned long flags;
@@ -20,7 +20,7 @@ static inline int atomic_futex_op_xchg_set(int oparg, int __user *uaddr,
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr,
+static inline int atomic_futex_op_xchg_add(int oparg, u32 __user *uaddr,
 					   int *oldval)
 {
 	unsigned long flags;
@@ -37,7 +37,7 @@ static inline int atomic_futex_op_xchg_add(int oparg, int __user *uaddr,
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr,
+static inline int atomic_futex_op_xchg_or(int oparg, u32 __user *uaddr,
 					  int *oldval)
 {
 	unsigned long flags;
@@ -54,7 +54,7 @@ static inline int atomic_futex_op_xchg_or(int oparg, int __user *uaddr,
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr,
+static inline int atomic_futex_op_xchg_and(int oparg, u32 __user *uaddr,
 					   int *oldval)
 {
 	unsigned long flags;
@@ -71,7 +71,7 @@ static inline int atomic_futex_op_xchg_and(int oparg, int __user *uaddr,
 	return ret;
 }
 
-static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr,
+static inline int atomic_futex_op_xchg_xor(int oparg, u32 __user *uaddr,
 					   int *oldval)
 {
 	unsigned long flags;
@@ -88,11 +88,13 @@ static inline int atomic_futex_op_xchg_xor(int oparg, int __user *uaddr,
 	return ret;
 }
 
-static inline int atomic_futex_op_cmpxchg_inatomic(int __user *uaddr,
-						   int oldval, int newval)
+static inline int atomic_futex_op_cmpxchg_inatomic(u32 *uval,
+						   u32 __user *uaddr,
+						   u32 oldval, u32 newval)
 {
 	unsigned long flags;
-	int ret, prev = 0;
+	int ret;
+	u32 prev = 0;
 
 	local_irq_save(flags);
 
@@ -102,10 +104,8 @@ static inline int atomic_futex_op_cmpxchg_inatomic(int __user *uaddr,
 
 	local_irq_restore(flags);
 
-	if (ret)
-		return ret;
-
-	return prev;
+	*uval = prev;
+	return ret;
 }
 
 #endif /* __ASM_SH_FUTEX_IRQ_H */
diff --git a/arch/sh/include/asm/futex.h b/arch/sh/include/asm/futex.h
index 68256ec5fa35..7be39a646fbd 100644
--- a/arch/sh/include/asm/futex.h
+++ b/arch/sh/include/asm/futex.h
@@ -10,7 +10,7 @@
 /* XXX: UP variants, fix for SH-4A and SMP.. */
 #include <asm/futex-irq.h>
 
-static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -21,7 +21,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -65,12 +65,13 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
-	return atomic_futex_op_cmpxchg_inatomic(uaddr, oldval, newval);
+	return atomic_futex_op_cmpxchg_inatomic(uval, uaddr, oldval, newval);
 }
 
 #endif /* __KERNEL__ */
diff --git a/arch/sparc/include/asm/futex_64.h b/arch/sparc/include/asm/futex_64.h
index 47f95839dc69..444e7bea23bc 100644
--- a/arch/sparc/include/asm/futex_64.h
+++ b/arch/sparc/include/asm/futex_64.h
@@ -30,7 +30,7 @@
 	: "r" (uaddr), "r" (oparg), "i" (-EFAULT)	\
 	: "memory")
 
-static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -38,7 +38,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	int cmparg = (encoded_op << 20) >> 20;
 	int oldval = 0, ret, tem;
 
-	if (unlikely(!access_ok(VERIFY_WRITE, uaddr, sizeof(int))))
+	if (unlikely(!access_ok(VERIFY_WRITE, uaddr, sizeof(u32))))
 		return -EFAULT;
 	if (unlikely((((unsigned long) uaddr) & 0x3UL)))
 		return -EINVAL;
@@ -85,26 +85,30 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 }
 
 static inline int
-futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval, int newval)
+futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+			      u32 oldval, u32 newval)
 {
+	int ret = 0;
+
 	__asm__ __volatile__(
-	"\n1:	casa	[%3] %%asi, %2, %0\n"
+	"\n1:	casa	[%4] %%asi, %3, %1\n"
 	"2:\n"
 	"	.section .fixup,#alloc,#execinstr\n"
 	"	.align	4\n"
 	"3:	sethi	%%hi(2b), %0\n"
 	"	jmpl	%0 + %%lo(2b), %%g0\n"
-	"	 mov	%4, %0\n"
+	"	mov	%5, %0\n"
 	"	.previous\n"
 	"	.section __ex_table,\"a\"\n"
 	"	.align	4\n"
 	"	.word	1b, 3b\n"
 	"	.previous\n"
-	: "=r" (newval)
-	: "0" (newval), "r" (oldval), "r" (uaddr), "i" (-EFAULT)
+	: "+r" (ret), "=r" (newval)
+	: "1" (newval), "r" (oldval), "r" (uaddr), "i" (-EFAULT)
 	: "memory");
 
-	return newval;
+	*uval = newval;
+	return ret;
 }
 
 #endif /* !(_SPARC64_FUTEX_H) */
diff --git a/arch/tile/include/asm/futex.h b/arch/tile/include/asm/futex.h
index fe0d10dcae57..d03ec124a598 100644
--- a/arch/tile/include/asm/futex.h
+++ b/arch/tile/include/asm/futex.h
@@ -29,16 +29,16 @@
 #include <linux/uaccess.h>
 #include <linux/errno.h>
 
-extern struct __get_user futex_set(int __user *v, int i);
-extern struct __get_user futex_add(int __user *v, int n);
-extern struct __get_user futex_or(int __user *v, int n);
-extern struct __get_user futex_andn(int __user *v, int n);
-extern struct __get_user futex_cmpxchg(int __user *v, int o, int n);
+extern struct __get_user futex_set(u32 __user *v, int i);
+extern struct __get_user futex_add(u32 __user *v, int n);
+extern struct __get_user futex_or(u32 __user *v, int n);
+extern struct __get_user futex_andn(u32 __user *v, int n);
+extern struct __get_user futex_cmpxchg(u32 __user *v, int o, int n);
 
 #ifndef __tilegx__
-extern struct __get_user futex_xor(int __user *v, int n);
+extern struct __get_user futex_xor(u32 __user *v, int n);
 #else
-static inline struct __get_user futex_xor(int __user *uaddr, int n)
+static inline struct __get_user futex_xor(u32 __user *uaddr, int n)
 {
 	struct __get_user asm_ret = __get_user_4(uaddr);
 	if (!asm_ret.err) {
@@ -53,7 +53,7 @@ static inline struct __get_user futex_xor(int __user *uaddr, int n)
 }
 #endif
 
-static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -65,7 +65,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	pagefault_disable();
@@ -119,16 +119,17 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	return ret;
 }
 
-static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
-						int newval)
+static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+						u32 oldval, u32 newval)
 {
 	struct __get_user asm_ret;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 	asm_ret = futex_cmpxchg(uaddr, oldval, newval);
-	return asm_ret.err ? asm_ret.err : asm_ret.val;
+	*uval = asm_ret.val;
+	return asm_ret.err;
 }
 
 #ifndef __tilegx__
diff --git a/arch/x86/include/asm/futex.h b/arch/x86/include/asm/futex.h
index 1f11ce44e956..d09bb03653f0 100644
--- a/arch/x86/include/asm/futex.h
+++ b/arch/x86/include/asm/futex.h
@@ -37,7 +37,7 @@
 		       "+m" (*uaddr), "=&r" (tem)		\
 		     : "r" (oparg), "i" (-EFAULT), "1" (0))
 
-static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
+static inline int futex_atomic_op_inuser(int encoded_op, u32 __user *uaddr)
 {
 	int op = (encoded_op >> 28) & 7;
 	int cmp = (encoded_op >> 24) & 15;
@@ -48,7 +48,7 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	if (encoded_op & (FUTEX_OP_OPARG_SHIFT << 28))
 		oparg = 1 << oparg;
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_BSWAP)
@@ -109,9 +109,10 @@ static inline int futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
 	return ret;
 }
 
-static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
-						int newval)
+static inline int futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
+						u32 oldval, u32 newval)
 {
+	int ret = 0;
 
 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_BSWAP)
 	/* Real i386 machines have no cmpxchg instruction */
@@ -119,21 +120,22 @@ static inline int futex_atomic_cmpxchg_inatomic(int __user *uaddr, int oldval,
 		return -ENOSYS;
 #endif
 
-	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(int)))
+	if (!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))
 		return -EFAULT;
 
-	asm volatile("1:\t" LOCK_PREFIX "cmpxchgl %3, %1\n"
+	asm volatile("1:\t" LOCK_PREFIX "cmpxchgl %4, %2\n"
 		     "2:\t.section .fixup, \"ax\"\n"
-		     "3:\tmov     %2, %0\n"
+		     "3:\tmov     %3, %0\n"
 		     "\tjmp     2b\n"
 		     "\t.previous\n"
 		     _ASM_EXTABLE(1b, 3b)
-		     : "=a" (oldval), "+m" (*uaddr)
-		     : "i" (-EFAULT), "r" (newval), "0" (oldval)
+		     : "+r" (ret), "=a" (oldval), "+m" (*uaddr)
+		     : "i" (-EFAULT), "r" (newval), "1" (oldval)
 		     : "memory"
 	);
 
-	return oldval;
+	*uval = oldval;
+	return ret;
 }
 
 #endif