diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile
index 7cffbaef064b..f0c0cdb1c183 100644
--- a/arch/arm/kernel/Makefile
+++ b/arch/arm/kernel/Makefile
@@ -25,7 +25,7 @@ obj-$(CONFIG_OABI_COMPAT)	+= sys_oabi-compat.o
 obj-$(CONFIG_CRUNCH)		+= crunch.o crunch-bits.o
 AFLAGS_crunch-bits.o		:= -Wa,-mcpu=ep9312
 
-obj-$(CONFIG_IWMMXT)		+= iwmmxt.o
+obj-$(CONFIG_IWMMXT)		+= iwmmxt.o iwmmxt-notifier.o
 AFLAGS_iwmmxt.o			:= -Wa,-mcpu=iwmmxt
 
 ifneq ($(CONFIG_ARCH_EBSA110),y)
diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 6423a38839b8..0e8aeaf9ff1d 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -590,9 +590,7 @@ ENTRY(__switch_to)
 #ifdef CONFIG_MMU
 	mcr	p15, 0, r6, c3, c0, 0		@ Set domain register
 #endif
-#if defined(CONFIG_IWMMXT)
-	bl	iwmmxt_task_switch
-#elif defined(CONFIG_CPU_XSCALE)
+#if defined(CONFIG_CPU_XSCALE) && !defined(CONFIG_IWMMXT)
 	add	r4, r2, #TI_CPU_DOMAIN + 40	@ cpu_context_save->extra
 	ldmib	r4, {r4, r5}
 	mar	acc0, r4, r5
diff --git a/arch/arm/kernel/iwmmxt-notifier.c b/arch/arm/kernel/iwmmxt-notifier.c
new file mode 100644
index 000000000000..44a86c33796e
--- /dev/null
+++ b/arch/arm/kernel/iwmmxt-notifier.c
@@ -0,0 +1,64 @@
+/*
+ *  linux/arch/arm/kernel/iwmmxt-notifier.c
+ *
+ *  XScale iWMMXt (Concan) context switching and handling
+ *
+ *  Initial code:
+ *  Copyright (c) 2003, Intel Corporation
+ *
+ *  Full lazy switching support, optimizations and more, by Nicolas Pitre
+ *  Copyright (c) 2003-2004, MontaVista Software, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/config.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <asm/thread_notify.h>
+#include <asm/io.h>
+
+static int iwmmxt_do(struct notifier_block *self, unsigned long cmd, void *t)
+{
+	struct thread_info *thread = t;
+
+	switch (cmd) {
+	case THREAD_NOTIFY_FLUSH:
+		/*
+		 * flush_thread() zeroes thread->fpstate, so no need
+		 * to do anything here.
+		 *
+		 * FALLTHROUGH: Ensure we don't try to overwrite our newly
+		 * initialised state information on the first fault.
+		 */
+
+	case THREAD_NOTIFY_RELEASE:
+		iwmmxt_task_release(thread);
+		break;
+
+	case THREAD_NOTIFY_SWITCH:
+		iwmmxt_task_switch(thread);
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block iwmmxt_notifier_block = {
+	.notifier_call	= iwmmxt_do,
+};
+
+static int __init iwmmxt_init(void)
+{
+	thread_register_notifier(&iwmmxt_notifier_block);
+
+	return 0;
+}
+
+late_initcall(iwmmxt_init);
diff --git a/arch/arm/kernel/iwmmxt.S b/arch/arm/kernel/iwmmxt.S
index a3bae95e536c..b63b528f22a6 100644
--- a/arch/arm/kernel/iwmmxt.S
+++ b/arch/arm/kernel/iwmmxt.S
@@ -271,30 +271,27 @@ ENTRY(iwmmxt_task_restore)
 /*
  * Concan handling on task switch
  *
- * r0 = previous task_struct pointer (must be preserved)
- * r1 = previous thread_info pointer
- * r2 = next thread_info pointer (must be preserved)
+ * r0 = next thread_info pointer
  *
- * Called only from __switch_to with task preemption disabled.
- * No need to care about preserving r4 and above.
+ * Called only from the iwmmxt notifier with task preemption disabled.
  */
 ENTRY(iwmmxt_task_switch)
 
-	mrc	p15, 0, r4, c15, c1, 0
-	tst	r4, #0x3			@ CP0 and CP1 accessible?
+	mrc	p15, 0, r1, c15, c1, 0
+	tst	r1, #0x3			@ CP0 and CP1 accessible?
 	bne	1f				@ yes: block them for next task
 
-	ldr	r5, =concan_owner
-	add	r6, r2, #TI_IWMMXT_STATE	@ get next task Concan save area
-	ldr	r5, [r5]			@ get current Concan owner
-	teq	r5, r6				@ next task owns it?
+	ldr	r2, =concan_owner
+	add	r3, r0, #TI_IWMMXT_STATE	@ get next task Concan save area
+	ldr	r2, [r2]			@ get current Concan owner
+	teq	r2, r3				@ next task owns it?
 	movne	pc, lr				@ no: leave Concan disabled
 
-1:	eor	r4, r4, #3			@ flip Concan access
-	mcr	p15, 0, r4, c15, c1, 0
+1:	eor	r1, r1, #3			@ flip Concan access
+	mcr	p15, 0, r1, c15, c1, 0
 
-	mrc	p15, 0, r4, c2, c0, 0
-	sub	pc, lr, r4, lsr #32		@ cpwait and return
+	mrc	p15, 0, r1, c2, c0, 0
+	sub	pc, lr, r1, lsr #32		@ cpwait and return
 
 /*
  * Remove Concan ownership of given task
diff --git a/arch/arm/kernel/process.c b/arch/arm/kernel/process.c
index e1c77ee885a7..b5773a2549ee 100644
--- a/arch/arm/kernel/process.c
+++ b/arch/arm/kernel/process.c
@@ -353,9 +353,6 @@ void flush_thread(void)
 	memset(&thread->fpstate, 0, sizeof(union fp_state));
 
 	thread_notify(THREAD_NOTIFY_FLUSH, thread);
-#if defined(CONFIG_IWMMXT)
-	iwmmxt_task_release(thread);
-#endif
 }
 
 void release_thread(struct task_struct *dead_task)
@@ -363,9 +360,6 @@ void release_thread(struct task_struct *dead_task)
 	struct thread_info *thread = task_thread_info(dead_task);
 
 	thread_notify(THREAD_NOTIFY_RELEASE, thread);
-#if defined(CONFIG_IWMMXT)
-	iwmmxt_task_release(thread);
-#endif
 }
 
 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
diff --git a/include/asm-arm/thread_info.h b/include/asm-arm/thread_info.h
index c46b5c84275f..c52e0bf75353 100644
--- a/include/asm-arm/thread_info.h
+++ b/include/asm-arm/thread_info.h
@@ -111,6 +111,7 @@ extern void iwmmxt_task_disable(struct thread_info *);
 extern void iwmmxt_task_copy(struct thread_info *, void *);
 extern void iwmmxt_task_restore(struct thread_info *, void *);
 extern void iwmmxt_task_release(struct thread_info *);
+extern void iwmmxt_task_switch(struct thread_info *);
 
 #endif