diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h
index 1581b8c..77747df 100644
--- a/arch/arm/include/asm/page.h
+++ b/arch/arm/include/asm/page.h
@@ -111,7 +111,7 @@
 struct page;
 
 struct cpu_user_fns {
-	void (*cpu_clear_user_page)(void *p, unsigned long user);
+	void (*cpu_clear_user_highpage)(struct page *page, unsigned long vaddr);
 	void (*cpu_copy_user_highpage)(struct page *to, struct page *from,
 			unsigned long vaddr);
 };
@@ -119,20 +119,21 @@
 #ifdef MULTI_USER
 extern struct cpu_user_fns cpu_user;
 
-#define __cpu_clear_user_page		cpu_user.cpu_clear_user_page
+#define __cpu_clear_user_highpage	cpu_user.cpu_clear_user_highpage
 #define __cpu_copy_user_highpage	cpu_user.cpu_copy_user_highpage
 
 #else
 
-#define __cpu_clear_user_page		__glue(_USER,_clear_user_page)
+#define __cpu_clear_user_highpage	__glue(_USER,_clear_user_highpage)
 #define __cpu_copy_user_highpage	__glue(_USER,_copy_user_highpage)
 
-extern void __cpu_clear_user_page(void *p, unsigned long user);
+extern void __cpu_clear_user_highpage(struct page *page, unsigned long vaddr);
 extern void __cpu_copy_user_highpage(struct page *to, struct page *from,
 			unsigned long vaddr);
 #endif
 
-#define clear_user_page(addr,vaddr,pg)	 __cpu_clear_user_page(addr, vaddr)
+#define clear_user_highpage(page,vaddr)		\
+	 __cpu_clear_user_highpage(page, vaddr)
 
 #define __HAVE_ARCH_COPY_USER_HIGHPAGE
 #define copy_user_highpage(to,from,vaddr,vma)	\
diff --git a/arch/arm/mm/copypage-feroceon.c b/arch/arm/mm/copypage-feroceon.c
index edd7168..c3651b2 100644
--- a/arch/arm/mm/copypage-feroceon.c
+++ b/arch/arm/mm/copypage-feroceon.c
@@ -79,12 +79,11 @@
 	kunmap_atomic(kto, KM_USER0);
 }
 
-void __attribute__((naked))
-feroceon_clear_user_page(void *kaddr, unsigned long vaddr)
+void feroceon_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	asm("\
-	stmfd	sp!, {r4-r7, lr}		\n\
-	mov	r1, %0				\n\
+	mov	r1, %1				\n\
 	mov	r2, #0				\n\
 	mov	r3, #0				\n\
 	mov	r4, #0				\n\
@@ -93,19 +92,20 @@
 	mov	r7, #0				\n\
 	mov	ip, #0				\n\
 	mov	lr, #0				\n\
-1:	stmia	r0, {r2-r7, ip, lr}		\n\
+1:	stmia	%0, {r2-r7, ip, lr}		\n\
 	subs	r1, r1, #1			\n\
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	mcr	p15, 0, %0, c7, c14, 1		@ clean and invalidate D line\n\
 	add	r0, r0, #32			\n\
 	bne	1b				\n\
-	mcr	p15, 0, r1, c7, c10, 4		@ drain WB\n\
-	ldmfd	sp!, {r4-r7, pc}"
+	mcr	p15, 0, r1, c7, c10, 4		@ drain WB"
 	:
-	: "I" (PAGE_SIZE / 32));
+	: "r" (kaddr), "I" (PAGE_SIZE / 32)
+	: "r1", "r2", "r3", "r4", "r5", "r6", "r7", "ip", "lr");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns feroceon_user_fns __initdata = {
-	.cpu_clear_user_page	= feroceon_clear_user_page,
+	.cpu_clear_user_highpage = feroceon_clear_user_highpage,
 	.cpu_copy_user_highpage	= feroceon_copy_user_highpage,
 };
 
diff --git a/arch/arm/mm/copypage-v3.c b/arch/arm/mm/copypage-v3.c
index 52df8f0..13ce0ba 100644
--- a/arch/arm/mm/copypage-v3.c
+++ b/arch/arm/mm/copypage-v3.c
@@ -54,10 +54,10 @@
  *
  * FIXME: do we need to handle cache stuff...
  */
-void __attribute__((naked)) v3_clear_user_page(void *kaddr, unsigned long vaddr)
+void v3_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	asm("\n\
-	str	lr, [sp, #-4]!\n\
 	mov	r1, %1				@ 1\n\
 	mov	r2, #0				@ 1\n\
 	mov	r3, #0				@ 1\n\
@@ -68,13 +68,14 @@
 	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
 	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
 	subs	r1, r1, #1			@ 1\n\
-	bne	1b				@ 1\n\
-	ldr	pc, [sp], #4"
+	bne	1b				@ 1"
 	:
-	: "r" (kaddr), "I" (PAGE_SIZE / 64));
+	: "r" (kaddr), "I" (PAGE_SIZE / 64)
+	: "r1", "r2", "r3", "ip", "lr");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns v3_user_fns __initdata = {
-	.cpu_clear_user_page	= v3_clear_user_page,
+	.cpu_clear_user_highpage = v3_clear_user_highpage,
 	.cpu_copy_user_highpage	= v3_copy_user_highpage,
 };
diff --git a/arch/arm/mm/copypage-v4mc.c b/arch/arm/mm/copypage-v4mc.c
index a7dc838..a5eae50 100644
--- a/arch/arm/mm/copypage-v4mc.c
+++ b/arch/arm/mm/copypage-v4mc.c
@@ -91,30 +91,30 @@
 /*
  * ARMv4 optimised clear_user_page
  */
-void __attribute__((naked))
-v4_mc_clear_user_page(void *kaddr, unsigned long vaddr)
+void v4_mc_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
-	asm volatile(
-	"str	lr, [sp, #-4]!\n\
+	void *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile("\
 	mov	r1, %0				@ 1\n\
 	mov	r2, #0				@ 1\n\
 	mov	r3, #0				@ 1\n\
 	mov	ip, #0				@ 1\n\
 	mov	lr, #0				@ 1\n\
-1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+1:	mcr	p15, 0, %0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	mcr	p15, 0, %0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
 	subs	r1, r1, #1			@ 1\n\
-	bne	1b				@ 1\n\
-	ldr	pc, [sp], #4"
+	bne	1b				@ 1"
 	:
-	: "I" (PAGE_SIZE / 64));
+	: "r" (kaddr), "I" (PAGE_SIZE / 64)
+	: "r1", "r2", "r3", "ip", "lr");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns v4_mc_user_fns __initdata = {
-	.cpu_clear_user_page	= v4_mc_clear_user_page, 
+	.cpu_clear_user_highpage = v4_mc_clear_user_highpage,
 	.cpu_copy_user_highpage	= v4_mc_copy_user_highpage,
 };
diff --git a/arch/arm/mm/copypage-v4wb.c b/arch/arm/mm/copypage-v4wb.c
index 186a68a..9144a96 100644
--- a/arch/arm/mm/copypage-v4wb.c
+++ b/arch/arm/mm/copypage-v4wb.c
@@ -64,31 +64,31 @@
  *
  * Same story as above.
  */
-void __attribute__((naked))
-v4wb_clear_user_page(void *kaddr, unsigned long vaddr)
+void v4wb_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	asm("\
-	str	lr, [sp, #-4]!\n\
-	mov	r1, %0				@ 1\n\
+	mov	r1, %1				@ 1\n\
 	mov	r2, #0				@ 1\n\
 	mov	r3, #0				@ 1\n\
 	mov	ip, #0				@ 1\n\
 	mov	lr, #0				@ 1\n\
-1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+1:	mcr	p15, 0, %0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	mcr	p15, 0, %0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
 	subs	r1, r1, #1			@ 1\n\
 	bne	1b				@ 1\n\
-	mcr	p15, 0, r1, c7, c10, 4		@ 1   drain WB\n\
-	ldr	pc, [sp], #4"
+	mcr	p15, 0, r1, c7, c10, 4		@ 1   drain WB"
 	:
-	: "I" (PAGE_SIZE / 64));
+	: "r" (kaddr), "I" (PAGE_SIZE / 64)
+	: "r1", "r2", "r3", "ip", "lr");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns v4wb_user_fns __initdata = {
-	.cpu_clear_user_page	= v4wb_clear_user_page,
+	.cpu_clear_user_highpage = v4wb_clear_user_highpage,
 	.cpu_copy_user_highpage	= v4wb_copy_user_highpage,
 };
diff --git a/arch/arm/mm/copypage-v4wt.c b/arch/arm/mm/copypage-v4wt.c
index 86c2cfd..b8a345d 100644
--- a/arch/arm/mm/copypage-v4wt.c
+++ b/arch/arm/mm/copypage-v4wt.c
@@ -60,29 +60,29 @@
  *
  * Same story as above.
  */
-void __attribute__((naked))
-v4wt_clear_user_page(void *kaddr, unsigned long vaddr)
+void v4wt_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	asm("\
-	str	lr, [sp, #-4]!\n\
-	mov	r1, %0				@ 1\n\
+	mov	r1, %1				@ 1\n\
 	mov	r2, #0				@ 1\n\
 	mov	r3, #0				@ 1\n\
 	mov	ip, #0				@ 1\n\
 	mov	lr, #0				@ 1\n\
-1:	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
-	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+1:	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	%0!, {r2, r3, ip, lr}		@ 4\n\
 	subs	r1, r1, #1			@ 1\n\
 	bne	1b				@ 1\n\
-	mcr	p15, 0, r2, c7, c7, 0		@ flush ID cache\n\
-	ldr	pc, [sp], #4"
+	mcr	p15, 0, r2, c7, c7, 0		@ flush ID cache"
 	:
-	: "I" (PAGE_SIZE / 64));
+	: "r" (kaddr), "I" (PAGE_SIZE / 64)
+	: "r1", "r2", "r3", "ip", "lr");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns v4wt_user_fns __initdata = {
-	.cpu_clear_user_page	= v4wt_clear_user_page,
+	.cpu_clear_user_highpage = v4wt_clear_user_highpage,
 	.cpu_copy_user_highpage	= v4wt_copy_user_highpage,
 };
diff --git a/arch/arm/mm/copypage-v6.c b/arch/arm/mm/copypage-v6.c
index 2ea75d0..4127a7b 100644
--- a/arch/arm/mm/copypage-v6.c
+++ b/arch/arm/mm/copypage-v6.c
@@ -49,9 +49,11 @@
  * Clear the user page.  No aliasing to deal with so we can just
  * attack the kernel's existing mapping of this page.
  */
-static void v6_clear_user_page_nonaliasing(void *kaddr, unsigned long vaddr)
+static void v6_clear_user_highpage_nonaliasing(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	clear_page(kaddr);
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 /*
@@ -107,20 +109,13 @@
  * so remap the kernel page into the same cache colour as the user
  * page.
  */
-static void v6_clear_user_page_aliasing(void *kaddr, unsigned long vaddr)
+static void v6_clear_user_highpage_aliasing(struct page *page, unsigned long vaddr)
 {
 	unsigned int offset = CACHE_COLOUR(vaddr);
 	unsigned long to = to_address + (offset << PAGE_SHIFT);
 
-	/*
-	 * Discard data in the kernel mapping for the new page
-	 * FIXME: needs this MCRR to be supported.
-	 */
-	__asm__("mcrr	p15, 0, %1, %0, c6	@ 0xec401f06"
-	   :
-	   : "r" (kaddr),
-	     "r" ((unsigned long)kaddr + PAGE_SIZE - L1_CACHE_BYTES)
-	   : "cc");
+	/* FIXME: not highmem safe */
+	discard_old_kernel_data(page_address(page));
 
 	/*
 	 * Now clear the page using the same cache colour as
@@ -128,7 +123,7 @@
 	 */
 	spin_lock(&v6_lock);
 
-	set_pte_ext(TOP_PTE(to_address) + offset, pfn_pte(__pa(kaddr) >> PAGE_SHIFT, PAGE_KERNEL), 0);
+	set_pte_ext(TOP_PTE(to_address) + offset, pfn_pte(page_to_pfn(page), PAGE_KERNEL), 0);
 	flush_tlb_kernel_page(to);
 	clear_page((void *)to);
 
@@ -136,14 +131,14 @@
 }
 
 struct cpu_user_fns v6_user_fns __initdata = {
-	.cpu_clear_user_page	= v6_clear_user_page_nonaliasing,
+	.cpu_clear_user_highpage = v6_clear_user_highpage_nonaliasing,
 	.cpu_copy_user_highpage	= v6_copy_user_highpage_nonaliasing,
 };
 
 static int __init v6_userpage_init(void)
 {
 	if (cache_is_vipt_aliasing()) {
-		cpu_user.cpu_clear_user_page = v6_clear_user_page_aliasing;
+		cpu_user.cpu_clear_user_highpage = v6_clear_user_highpage_aliasing;
 		cpu_user.cpu_copy_user_highpage = v6_copy_user_highpage_aliasing;
 	}
 
diff --git a/arch/arm/mm/copypage-xsc3.c b/arch/arm/mm/copypage-xsc3.c
index caa697c..0e7cb32 100644
--- a/arch/arm/mm/copypage-xsc3.c
+++ b/arch/arm/mm/copypage-xsc3.c
@@ -87,26 +87,27 @@
  *  r0 = destination
  *  r1 = virtual user address of ultimate destination page
  */
-void __attribute__((naked))
-xsc3_mc_clear_user_page(void *kaddr, unsigned long vaddr)
+void xsc3_mc_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	asm("\
-	mov	r1, %0				\n\
+	mov	r1, %1				\n\
 	mov	r2, #0				\n\
 	mov	r3, #0				\n\
-1:	mcr	p15, 0, r0, c7, c6, 1		@ invalidate line\n\
-	strd	r2, [r0], #8			\n\
-	strd	r2, [r0], #8			\n\
-	strd	r2, [r0], #8			\n\
-	strd	r2, [r0], #8			\n\
+1:	mcr	p15, 0, %0, c7, c6, 1		@ invalidate line\n\
+	strd	r2, [%0], #8			\n\
+	strd	r2, [%0], #8			\n\
+	strd	r2, [%0], #8			\n\
+	strd	r2, [%0], #8			\n\
 	subs	r1, r1, #1			\n\
-	bne	1b				\n\
-	mov	pc, lr"
+	bne	1b"
 	:
-	: "I" (PAGE_SIZE / 32));
+	: "r" (kaddr), "I" (PAGE_SIZE / 32)
+	: "r1", "r2", "r3");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns xsc3_mc_user_fns __initdata = {
-	.cpu_clear_user_page	= xsc3_mc_clear_user_page,
+	.cpu_clear_user_highpage = xsc3_mc_clear_user_highpage,
 	.cpu_copy_user_highpage	= xsc3_mc_copy_user_highpage,
 };
diff --git a/arch/arm/mm/copypage-xscale.c b/arch/arm/mm/copypage-xscale.c
index 01bafaf..aa9f2ff 100644
--- a/arch/arm/mm/copypage-xscale.c
+++ b/arch/arm/mm/copypage-xscale.c
@@ -113,28 +113,30 @@
 /*
  * XScale optimised clear_user_page
  */
-void __attribute__((naked))
-xscale_mc_clear_user_page(void *kaddr, unsigned long vaddr)
+void
+xscale_mc_clear_user_highpage(struct page *page, unsigned long vaddr)
 {
+	void *kaddr = kmap_atomic(page, KM_USER0);
 	asm volatile(
-	"mov	r1, %0				\n\
+	"mov	r1, %1				\n\
 	mov	r2, #0				\n\
 	mov	r3, #0				\n\
-1:	mov	ip, r0				\n\
-	strd	r2, [r0], #8			\n\
-	strd	r2, [r0], #8			\n\
-	strd	r2, [r0], #8			\n\
-	strd	r2, [r0], #8			\n\
+1:	mov	ip, %0				\n\
+	strd	r2, [%0], #8			\n\
+	strd	r2, [%0], #8			\n\
+	strd	r2, [%0], #8			\n\
+	strd	r2, [%0], #8			\n\
 	mcr	p15, 0, ip, c7, c10, 1		@ clean D line\n\
 	subs	r1, r1, #1			\n\
 	mcr	p15, 0, ip, c7, c6, 1		@ invalidate D line\n\
-	bne	1b				\n\
-	mov	pc, lr"
+	bne	1b"
 	:
-	: "I" (PAGE_SIZE / 32));
+	: "r" (kaddr), "I" (PAGE_SIZE / 32)
+	: "r1", "r2", "r3", "ip");
+	kunmap_atomic(kaddr, KM_USER0);
 }
 
 struct cpu_user_fns xscale_mc_user_fns __initdata = {
-	.cpu_clear_user_page	= xscale_mc_clear_user_page, 
+	.cpu_clear_user_highpage = xscale_mc_clear_user_highpage,
 	.cpu_copy_user_highpage	= xscale_mc_copy_user_highpage,
 };
diff --git a/arch/arm/mm/proc-syms.c b/arch/arm/mm/proc-syms.c
index b9743e6..4ad3bf2 100644
--- a/arch/arm/mm/proc-syms.c
+++ b/arch/arm/mm/proc-syms.c
@@ -33,7 +33,7 @@
 
 #ifdef CONFIG_MMU
 #ifndef MULTI_USER
-EXPORT_SYMBOL(__cpu_clear_user_page);
+EXPORT_SYMBOL(__cpu_clear_user_highpage);
 EXPORT_SYMBOL(__cpu_copy_user_highpage);
 #else
 EXPORT_SYMBOL(cpu_user);