diff --git a/arch/arm/kernel/calls.S b/arch/arm/kernel/calls.S
index 2ad4aa2a1536c1ca22e1843352cdd9d2ac9cfbb7..55076a75e5bf3794de942c7f7abe3e1993bfb18e 100644
--- a/arch/arm/kernel/calls.S
+++ b/arch/arm/kernel/calls.S
@@ -131,7 +131,7 @@ __syscall_start:
 		.long	sys_wait4
 /* 115 */	.long	sys_swapoff
 		.long	sys_sysinfo
-		.long	sys_ipc_wrapper
+		.long	sys_ipc
 		.long	sys_fsync
 		.long	sys_sigreturn_wrapper
 /* 120 */	.long	sys_clone_wrapper
@@ -254,7 +254,7 @@ __syscall_start:
 		.long	sys_fremovexattr
 		.long	sys_tkill
 		.long	sys_sendfile64
-/* 240 */	.long	sys_futex_wrapper
+/* 240 */	.long	sys_futex
 		.long	sys_sched_setaffinity
 		.long	sys_sched_getaffinity
 		.long	sys_io_setup
@@ -284,7 +284,7 @@ __syscall_start:
 		.long	sys_fstatfs64
 		.long	sys_tgkill
 		.long	sys_utimes
-/* 270 */	.long	sys_arm_fadvise64_64_wrapper
+/* 270 */	.long	sys_arm_fadvise64_64
 		.long	sys_pciconfig_iobase
 		.long	sys_pciconfig_read
 		.long	sys_pciconfig_write
@@ -333,7 +333,7 @@ __syscall_start:
 		.long	sys_inotify_init
 		.long	sys_inotify_add_watch
 		.long	sys_inotify_rm_watch
-		.long	sys_mbind_wrapper
+		.long	sys_mbind
 /* 320 */	.long	sys_get_mempolicy
 		.long	sys_set_mempolicy
 __syscall_end:
diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index d9fb819bf7cc3960aedb4b9836451371458a4893..2a8d27e18fa7765c13229cbb91f31a113c1b113d 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -611,6 +611,47 @@ ENTRY(__switch_to)
 	.globl	__kuser_helper_start
 __kuser_helper_start:
 
+/*
+ * Reference prototype:
+ *
+ *	void __kernel_memory_barrier(void)
+ *
+ * Input:
+ *
+ *	lr = return address
+ *
+ * Output:
+ *
+ *	none
+ *
+ * Clobbered:
+ *
+ *	the Z flag might be lost
+ *
+ * Definition and user space usage example:
+ *
+ *	typedef void (__kernel_dmb_t)(void);
+ *	#define __kernel_dmb (*(__kernel_dmb_t *)0xffff0fa0)
+ *
+ * Apply any needed memory barrier to preserve consistency with data modified
+ * manually and __kuser_cmpxchg usage.
+ *
+ * This could be used as follows:
+ *
+ * #define __kernel_dmb() \
+ *         asm volatile ( "mov r0, #0xffff0fff; mov lr, pc; sub pc, r0, #95" \
+ *	        : : : "lr","cc" )
+ */
+
+__kuser_memory_barrier:				@ 0xffff0fa0
+
+#if __LINUX_ARM_ARCH__ >= 6 && defined(CONFIG_SMP)
+	mcr	p15, 0, r0, c7, c10, 5	@ dmb
+#endif
+	mov	pc, lr
+
+	.align	5
+
 /*
  * Reference prototype:
  *
@@ -642,6 +683,8 @@ __kuser_helper_start:
  * The C flag is also set if *ptr was changed to allow for assembly
  * optimization in the calling code.
  *
+ * Note: this routine already includes memory barriers as needed.
+ *
  * For example, a user space atomic_add implementation could look like this:
  *
  * #define atomic_add(ptr, val) \
@@ -698,10 +741,16 @@ __kuser_cmpxchg:				@ 0xffff0fc0
 
 #else
 
+#ifdef CONFIG_SMP
+	mcr	p15, 0, r0, c7, c10, 5	@ dmb
+#endif
 	ldrex	r3, [r2]
 	subs	r3, r3, r0
 	strexeq	r3, r1, [r2]
 	rsbs	r0, r3, #0
+#ifdef CONFIG_SMP
+	mcr	p15, 0, r0, c7, c10, 5	@ dmb
+#endif
 	mov	pc, lr
 
 #endif
diff --git a/arch/arm/kernel/entry-common.S b/arch/arm/kernel/entry-common.S
index f7f18307523754ba6fda73990ac5033e08cf1f70..e2b42997ad33ee0f1da6ec4a75440d16ce3c3b06 100644
--- a/arch/arm/kernel/entry-common.S
+++ b/arch/arm/kernel/entry-common.S
@@ -145,7 +145,7 @@ ENTRY(vector_swi)
 #endif
 	enable_irq
 
-	str	r4, [sp, #-S_OFF]!		@ push fifth arg
+	stmdb	sp!, {r4, r5}			@ push fifth and sixth args
 
 	get_thread_info tsk
 	ldr	ip, [tsk, #TI_FLAGS]		@ check for syscall tracing
@@ -204,7 +204,7 @@ ENTRY(sys_call_table)
  * Special system call wrappers
  */
 @ r0 = syscall number
-@ r5 = syscall table
+@ r8 = syscall table
 		.type	sys_syscall, #function
 sys_syscall:
 		eor	scno, r0, #__NR_SYSCALL_BASE
@@ -255,22 +255,6 @@ sys_sigaltstack_wrapper:
 		ldr	r2, [sp, #S_OFF + S_SP]
 		b	do_sigaltstack
 
-sys_futex_wrapper:
-		str	r5, [sp, #4]		@ push sixth arg
-		b	sys_futex
-
-sys_arm_fadvise64_64_wrapper:
-		str	r5, [sp, #4]		@ push r5 to stack
-		b	sys_arm_fadvise64_64
-
-sys_mbind_wrapper:
-		str	r5, [sp, #4]
-		b	sys_mbind
-
-sys_ipc_wrapper:
-		str	r5, [sp, #4]		@ push sixth arg
-		b	sys_ipc
-
 /*
  * Note: off_4k (r5) is always units of 4K.  If we can't do the requested
  * offset, we return EINVAL.