Merge branch 'highmem' into devel

Conflicts:

	arch/arm/mach-clps7500/include/mach/memory.h

19 files changed, 692 insertions, 591 deletions

diff --git a/arch/arm/mm/copypage-feroceon.S b/arch/arm/mm/copypage-feroceon.S
deleted file mode 100644
index 7eb0d32..0000000
--- a/arch/arm/mm/copypage-feroceon.S
+++ /dev/null
@@ -1,95 +0,0 @@
-/*
- *  linux/arch/arm/lib/copypage-feroceon.S
- *
- *  Copyright (C) 2008 Marvell Semiconductors
- *
- * 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.
- *
- * This handles copy_user_page and clear_user_page on Feroceon
- * more optimally than the generic implementations.
- */
-#include <linux/linkage.h>
-#include <linux/init.h>
-#include <asm/asm-offsets.h>
-
-	.text
-	.align	5
-
-ENTRY(feroceon_copy_user_page)
-	stmfd	sp!, {r4-r9, lr}
-	mov	ip, #PAGE_SZ
-1:	mov	lr, r1
-	ldmia	r1!, {r2 - r9}
-	pld	[lr, #32]
-	pld	[lr, #64]
-	pld	[lr, #96]
-	pld	[lr, #128]
-	pld	[lr, #160]
-	pld	[lr, #192]
-	pld	[lr, #224]
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	ldmia	r1!, {r2 - r9}
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	stmia	r0, {r2 - r9}
-	subs	ip, ip, #(32 * 8)
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	bne	1b
-	mcr	p15, 0, ip, c7, c10, 4		@ drain WB
-	ldmfd	sp!, {r4-r9, pc}
-
-	.align	5
-
-ENTRY(feroceon_clear_user_page)
-	stmfd	sp!, {r4-r7, lr}
-	mov	r1, #PAGE_SZ/32
-	mov	r2, #0
-	mov	r3, #0
-	mov	r4, #0
-	mov	r5, #0
-	mov	r6, #0
-	mov	r7, #0
-	mov	ip, #0
-	mov	lr, #0
-1:	stmia	r0, {r2-r7, ip, lr}
-	subs	r1, r1, #1
-	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line
-	add	r0, r0, #32
-	bne	1b
-	mcr	p15, 0, r1, c7, c10, 4		@ drain WB
-	ldmfd	sp!, {r4-r7, pc}
-
-	__INITDATA
-
-	.type	feroceon_user_fns, #object
-ENTRY(feroceon_user_fns)
-	.long	feroceon_clear_user_page
-	.long	feroceon_copy_user_page
-	.size	feroceon_user_fns, . - feroceon_user_fns
diff --git a/arch/arm/mm/copypage-feroceon.c b/arch/arm/mm/copypage-feroceon.c
new file mode 100644
index 0000000..c3ba6a9
--- /dev/null
+++ b/arch/arm/mm/copypage-feroceon.c
@@ -0,0 +1,111 @@
+/*
+ *  linux/arch/arm/mm/copypage-feroceon.S
+ *
+ *  Copyright (C) 2008 Marvell Semiconductors
+ *
+ * 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.
+ *
+ * This handles copy_user_highpage and clear_user_page on Feroceon
+ * more optimally than the generic implementations.
+ */
+#include <linux/init.h>
+#include <linux/highmem.h>
+
+static void __attribute__((naked))
+feroceon_copy_user_page(void *kto, const void *kfrom)
+{
+	asm("\
+	stmfd	sp!, {r4-r9, lr}		\n\
+	mov	ip, %0				\n\
+1:	mov	lr, r1				\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	pld	[lr, #32]			\n\
+	pld	[lr, #64]			\n\
+	pld	[lr, #96]			\n\
+	pld	[lr, #128]			\n\
+	pld	[lr, #160]			\n\
+	pld	[lr, #192]			\n\
+	pld	[lr, #224]			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	ldmia	r1!, {r2 - r9}			\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	stmia	r0, {r2 - r9}			\n\
+	subs	ip, ip, #(32 * 8)		\n\
+	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	r0, r0, #32			\n\
+	bne	1b				\n\
+	mcr	p15, 0, ip, c7, c10, 4		@ drain WB\n\
+	ldmfd	sp!, {r4-r9, pc}"
+	:
+	: "I" (PAGE_SIZE));
+}
+
+void feroceon_copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr)
+{
+	void *kto, *kfrom;
+
+	kto = kmap_atomic(to, KM_USER0);
+	kfrom = kmap_atomic(from, KM_USER1);
+	feroceon_copy_user_page(kto, kfrom);
+	kunmap_atomic(kfrom, KM_USER1);
+	kunmap_atomic(kto, KM_USER0);
+}
+
+void feroceon_clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	void *ptr, *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile ("\
+	mov	r1, %2				\n\
+	mov	r2, #0				\n\
+	mov	r3, #0				\n\
+	mov	r4, #0				\n\
+	mov	r5, #0				\n\
+	mov	r6, #0				\n\
+	mov	r7, #0				\n\
+	mov	ip, #0				\n\
+	mov	lr, #0				\n\
+1:	stmia	%0, {r2-r7, ip, lr}		\n\
+	subs	r1, r1, #1			\n\
+	mcr	p15, 0, %0, c7, c14, 1		@ clean and invalidate D line\n\
+	add	%0, %0, #32			\n\
+	bne	1b				\n\
+	mcr	p15, 0, r1, c7, c10, 4		@ drain WB"
+	: "=r" (ptr)
+	: "0" (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_highpage = feroceon_clear_user_highpage,
+	.cpu_copy_user_highpage	= feroceon_copy_user_highpage,
+};
+
diff --git a/arch/arm/mm/copypage-v3.S b/arch/arm/mm/copypage-v3.S
deleted file mode 100644
index 2ee394b..0000000
--- a/arch/arm/mm/copypage-v3.S
+++ /dev/null
@@ -1,67 +0,0 @@
-/*
- *  linux/arch/arm/lib/copypage.S
- *
- *  Copyright (C) 1995-1999 Russell King
- *
- * 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.
- *
- *  ASM optimised string functions
- */
-#include <linux/linkage.h>
-#include <linux/init.h>
-#include <asm/assembler.h>
-#include <asm/asm-offsets.h>
-
-		.text
-		.align	5
-/*
- * ARMv3 optimised copy_user_page
- *
- * FIXME: do we need to handle cache stuff...
- */
-ENTRY(v3_copy_user_page)
-	stmfd	sp!, {r4, lr}			@	2
-	mov	r2, #PAGE_SZ/64			@	1
-	ldmia	r1!, {r3, r4, ip, lr}		@	4+1
-1:	stmia	r0!, {r3, r4, ip, lr}		@	4
-	ldmia	r1!, {r3, r4, ip, lr}		@	4+1
-	stmia	r0!, {r3, r4, ip, lr}		@	4
-	ldmia	r1!, {r3, r4, ip, lr}		@	4+1
-	stmia	r0!, {r3, r4, ip, lr}		@	4
-	ldmia	r1!, {r3, r4, ip, lr}		@	4
-	subs	r2, r2, #1			@	1
-	stmia	r0!, {r3, r4, ip, lr}		@	4
-	ldmneia	r1!, {r3, r4, ip, lr}		@	4
-	bne	1b				@	1
-	ldmfd	sp!, {r4, pc}			@	3
-
-	.align	5
-/*
- * ARMv3 optimised clear_user_page
- *
- * FIXME: do we need to handle cache stuff...
- */
-ENTRY(v3_clear_user_page)
-	str	lr, [sp, #-4]!
-	mov	r1, #PAGE_SZ/64			@ 1
-	mov	r2, #0				@ 1
-	mov	r3, #0				@ 1
-	mov	ip, #0				@ 1
-	mov	lr, #0				@ 1
-1:	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	subs	r1, r1, #1			@ 1
-	bne	1b				@ 1
-	ldr	pc, [sp], #4
-
-	__INITDATA
-
-	.type	v3_user_fns, #object
-ENTRY(v3_user_fns)
-	.long	v3_clear_user_page
-	.long	v3_copy_user_page
-	.size	v3_user_fns, . - v3_user_fns
diff --git a/arch/arm/mm/copypage-v3.c b/arch/arm/mm/copypage-v3.c
new file mode 100644
index 0000000..70ed96c
--- /dev/null
+++ b/arch/arm/mm/copypage-v3.c
@@ -0,0 +1,81 @@
+/*
+ *  linux/arch/arm/mm/copypage-v3.c
+ *
+ *  Copyright (C) 1995-1999 Russell King
+ *
+ * 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/init.h>
+#include <linux/highmem.h>
+
+/*
+ * ARMv3 optimised copy_user_highpage
+ *
+ * FIXME: do we need to handle cache stuff...
+ */
+static void __attribute__((naked))
+v3_copy_user_page(void *kto, const void *kfrom)
+{
+	asm("\n\
+	stmfd	sp!, {r4, lr}			@	2\n\
+	mov	r2, %2				@	1\n\
+	ldmia	%0!, {r3, r4, ip, lr}		@	4+1\n\
+1:	stmia	%1!, {r3, r4, ip, lr}		@	4\n\
+	ldmia	%0!, {r3, r4, ip, lr}		@	4+1\n\
+	stmia	%1!, {r3, r4, ip, lr}		@	4\n\
+	ldmia	%0!, {r3, r4, ip, lr}		@	4+1\n\
+	stmia	%1!, {r3, r4, ip, lr}		@	4\n\
+	ldmia	%0!, {r3, r4, ip, lr}		@	4\n\
+	subs	r2, r2, #1			@	1\n\
+	stmia	%1!, {r3, r4, ip, lr}		@	4\n\
+	ldmneia	%0!, {r3, r4, ip, lr}		@	4\n\
+	bne	1b				@	1\n\
+	ldmfd	sp!, {r4, pc}			@	3"
+	:
+	: "r" (kfrom), "r" (kto), "I" (PAGE_SIZE / 64));
+}
+
+void v3_copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr)
+{
+	void *kto, *kfrom;
+
+	kto = kmap_atomic(to, KM_USER0);
+	kfrom = kmap_atomic(from, KM_USER1);
+	v3_copy_user_page(kto, kfrom);
+	kunmap_atomic(kfrom, KM_USER1);
+	kunmap_atomic(kto, KM_USER0);
+}
+
+/*
+ * ARMv3 optimised clear_user_page
+ *
+ * FIXME: do we need to handle cache stuff...
+ */
+void v3_clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	void *ptr, *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile("\n\
+	mov	r1, %2				@ 1\n\
+	mov	r2, #0				@ 1\n\
+	mov	r3, #0				@ 1\n\
+	mov	ip, #0				@ 1\n\
+	mov	lr, #0				@ 1\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"
+	: "=r" (ptr)
+	: "0" (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_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 8d33e25..bdb5fd9 100644
--- a/arch/arm/mm/copypage-v4mc.c
+++ b/arch/arm/mm/copypage-v4mc.c
@@ -15,8 +15,8 @@
  */
 #include <linux/init.h>
 #include <linux/mm.h>
+#include <linux/highmem.h>
 
-#include <asm/page.h>
 #include <asm/pgtable.h>
 #include <asm/tlbflush.h>
 #include <asm/cacheflush.h>
@@ -33,7 +33,7 @@
 static DEFINE_SPINLOCK(minicache_lock);
 
 /*
- * ARMv4 mini-dcache optimised copy_user_page
+ * ARMv4 mini-dcache optimised copy_user_highpage
  *
  * We flush the destination cache lines just before we write the data into the
  * corresponding address.  Since the Dcache is read-allocate, this removes the
@@ -42,7 +42,7 @@ static DEFINE_SPINLOCK(minicache_lock);
  *
  * Note: We rely on all ARMv4 processors implementing the "invalidate D line"
  * instruction.  If your processor does not supply this, you have to write your
- * own copy_user_page that does the right thing.
+ * own copy_user_highpage that does the right thing.
  */
 static void __attribute__((naked))
 mc_copy_user_page(void *from, void *to)
@@ -68,50 +68,53 @@ mc_copy_user_page(void *from, void *to)
 	: "r" (from), "r" (to), "I" (PAGE_SIZE / 64));
 }
 
-void v4_mc_copy_user_page(void *kto, const void *kfrom, unsigned long vaddr)
+void v4_mc_copy_user_highpage(struct page *from, struct page *to,
+	unsigned long vaddr)
 {
-	struct page *page = virt_to_page(kfrom);
+	void *kto = kmap_atomic(to, KM_USER1);
 
-	if (test_and_clear_bit(PG_dcache_dirty, &page->flags))
-		__flush_dcache_page(page_mapping(page), page);
+	if (test_and_clear_bit(PG_dcache_dirty, &from->flags))
+		__flush_dcache_page(page_mapping(from), from);
 
 	spin_lock(&minicache_lock);
 
-	set_pte_ext(TOP_PTE(0xffff8000), pfn_pte(__pa(kfrom) >> PAGE_SHIFT, minicache_pgprot), 0);
+	set_pte_ext(TOP_PTE(0xffff8000), pfn_pte(page_to_pfn(from), minicache_pgprot), 0);
 	flush_tlb_kernel_page(0xffff8000);
 
 	mc_copy_user_page((void *)0xffff8000, kto);
 
 	spin_unlock(&minicache_lock);
+
+	kunmap_atomic(kto, KM_USER1);
 }
 
 /*
  * 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\
-	mov	r1, %0				@ 1\n\
+	void *ptr, *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile("\
+	mov	r1, %2				@ 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"
-	:
-	: "I" (PAGE_SIZE / 64));
+	bne	1b				@ 1"
+	: "=r" (ptr)
+	: "0" (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_copy_user_page	= v4_mc_copy_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.S b/arch/arm/mm/copypage-v4wb.S
deleted file mode 100644
index 8311735..0000000
--- a/arch/arm/mm/copypage-v4wb.S
+++ /dev/null
@@ -1,79 +0,0 @@
-/*
- *  linux/arch/arm/lib/copypage.S
- *
- *  Copyright (C) 1995-1999 Russell King
- *
- * 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.
- *
- *  ASM optimised string functions
- */
-#include <linux/linkage.h>
-#include <linux/init.h>
-#include <asm/asm-offsets.h>
-
-	.text
-	.align	5
-/*
- * ARMv4 optimised copy_user_page
- *
- * We flush the destination cache lines just before we write the data into the
- * corresponding address.  Since the Dcache is read-allocate, this removes the
- * Dcache aliasing issue.  The writes will be forwarded to the write buffer,
- * and merged as appropriate.
- *
- * Note: We rely on all ARMv4 processors implementing the "invalidate D line"
- * instruction.  If your processor does not supply this, you have to write your
- * own copy_user_page that does the right thing.
- */
-ENTRY(v4wb_copy_user_page)
-	stmfd	sp!, {r4, lr}			@ 2
-	mov	r2, #PAGE_SZ/64			@ 1
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4
-1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4+1
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4
-	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4
-	subs	r2, r2, #1			@ 1
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmneia	r1!, {r3, r4, ip, lr}		@ 4
-	bne	1b				@ 1
-	mcr	p15, 0, r1, c7, c10, 4		@ 1   drain WB
-	ldmfd	 sp!, {r4, pc}			@ 3
-
-	.align	5
-/*
- * ARMv4 optimised clear_user_page
- *
- * Same story as above.
- */
-ENTRY(v4wb_clear_user_page)
-	str	lr, [sp, #-4]!
-	mov	r1, #PAGE_SZ/64			@ 1
-	mov	r2, #0				@ 1
-	mov	r3, #0				@ 1
-	mov	ip, #0				@ 1
-	mov	lr, #0				@ 1
-1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	subs	r1, r1, #1			@ 1
-	bne	1b				@ 1
-	mcr	p15, 0, r1, c7, c10, 4		@ 1   drain WB
-	ldr	pc, [sp], #4
-
-	__INITDATA
-
-	.type	v4wb_user_fns, #object
-ENTRY(v4wb_user_fns)
-	.long	v4wb_clear_user_page
-	.long	v4wb_copy_user_page
-	.size	v4wb_user_fns, . - v4wb_user_fns
diff --git a/arch/arm/mm/copypage-v4wb.c b/arch/arm/mm/copypage-v4wb.c
new file mode 100644
index 0000000..3ec93da
--- /dev/null
+++ b/arch/arm/mm/copypage-v4wb.c
@@ -0,0 +1,94 @@
+/*
+ *  linux/arch/arm/mm/copypage-v4wb.c
+ *
+ *  Copyright (C) 1995-1999 Russell King
+ *
+ * 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/init.h>
+#include <linux/highmem.h>
+
+/*
+ * ARMv4 optimised copy_user_highpage
+ *
+ * We flush the destination cache lines just before we write the data into the
+ * corresponding address.  Since the Dcache is read-allocate, this removes the
+ * Dcache aliasing issue.  The writes will be forwarded to the write buffer,
+ * and merged as appropriate.
+ *
+ * Note: We rely on all ARMv4 processors implementing the "invalidate D line"
+ * instruction.  If your processor does not supply this, you have to write your
+ * own copy_user_highpage that does the right thing.
+ */
+static void __attribute__((naked))
+v4wb_copy_user_page(void *kto, const void *kfrom)
+{
+	asm("\
+	stmfd	sp!, {r4, lr}			@ 2\n\
+	mov	r2, %0				@ 1\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4\n\
+1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4+1\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4\n\
+	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4\n\
+	subs	r2, r2, #1			@ 1\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmneia	r1!, {r3, r4, ip, lr}		@ 4\n\
+	bne	1b				@ 1\n\
+	mcr	p15, 0, r1, c7, c10, 4		@ 1   drain WB\n\
+	ldmfd	 sp!, {r4, pc}			@ 3"
+	:
+	: "I" (PAGE_SIZE / 64));
+}
+
+void v4wb_copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr)
+{
+	void *kto, *kfrom;
+
+	kto = kmap_atomic(to, KM_USER0);
+	kfrom = kmap_atomic(from, KM_USER1);
+	v4wb_copy_user_page(kto, kfrom);
+	kunmap_atomic(kfrom, KM_USER1);
+	kunmap_atomic(kto, KM_USER0);
+}
+
+/*
+ * ARMv4 optimised clear_user_page
+ *
+ * Same story as above.
+ */
+void v4wb_clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	void *ptr, *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile("\
+	mov	r1, %2				@ 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, %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"
+	: "=r" (ptr)
+	: "0" (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_highpage = v4wb_clear_user_highpage,
+	.cpu_copy_user_highpage	= v4wb_copy_user_highpage,
+};
diff --git a/arch/arm/mm/copypage-v4wt.S b/arch/arm/mm/copypage-v4wt.S
deleted file mode 100644
index e1f2af2..0000000
--- a/arch/arm/mm/copypage-v4wt.S
+++ /dev/null
@@ -1,73 +0,0 @@
-/*
- *  linux/arch/arm/lib/copypage-v4.S
- *
- *  Copyright (C) 1995-1999 Russell King
- *
- * 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.
- *
- *  ASM optimised string functions
- *
- *  This is for CPUs with a writethrough cache and 'flush ID cache' is
- *  the only supported cache operation.
- */
-#include <linux/linkage.h>
-#include <linux/init.h>
-#include <asm/asm-offsets.h>
-
-	.text
-	.align	5
-/*
- * ARMv4 optimised copy_user_page
- *
- * Since we have writethrough caches, we don't have to worry about
- * dirty data in the cache.  However, we do have to ensure that
- * subsequent reads are up to date.
- */
-ENTRY(v4wt_copy_user_page)
-	stmfd	sp!, {r4, lr}			@ 2
-	mov	r2, #PAGE_SZ/64			@ 1
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4
-1:	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4+1
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmia	r1!, {r3, r4, ip, lr}		@ 4
-	subs	r2, r2, #1			@ 1
-	stmia	r0!, {r3, r4, ip, lr}		@ 4
-	ldmneia	r1!, {r3, r4, ip, lr}		@ 4
-	bne	1b				@ 1
-	mcr	p15, 0, r2, c7, c7, 0		@ flush ID cache
-	ldmfd	sp!, {r4, pc}			@ 3
-
-	.align	5
-/*
- * ARMv4 optimised clear_user_page
- *
- * Same story as above.
- */
-ENTRY(v4wt_clear_user_page)
-	str	lr, [sp, #-4]!
-	mov	r1, #PAGE_SZ/64			@ 1
-	mov	r2, #0				@ 1
-	mov	r3, #0				@ 1
-	mov	ip, #0				@ 1
-	mov	lr, #0				@ 1
-1:	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	subs	r1, r1, #1			@ 1
-	bne	1b				@ 1
-	mcr	p15, 0, r2, c7, c7, 0		@ flush ID cache
-	ldr	pc, [sp], #4
-
-	__INITDATA
-
-	.type	v4wt_user_fns, #object
-ENTRY(v4wt_user_fns)
-	.long	v4wt_clear_user_page
-	.long	v4wt_copy_user_page
-	.size	v4wt_user_fns, . - v4wt_user_fns
diff --git a/arch/arm/mm/copypage-v4wt.c b/arch/arm/mm/copypage-v4wt.c
new file mode 100644
index 0000000..0f1188e
--- /dev/null
+++ b/arch/arm/mm/copypage-v4wt.c
@@ -0,0 +1,88 @@
+/*
+ *  linux/arch/arm/mm/copypage-v4wt.S
+ *
+ *  Copyright (C) 1995-1999 Russell King
+ *
+ * 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.
+ *
+ *  This is for CPUs with a writethrough cache and 'flush ID cache' is
+ *  the only supported cache operation.
+ */
+#include <linux/init.h>
+#include <linux/highmem.h>
+
+/*
+ * ARMv4 optimised copy_user_highpage
+ *
+ * Since we have writethrough caches, we don't have to worry about
+ * dirty data in the cache.  However, we do have to ensure that
+ * subsequent reads are up to date.
+ */
+static void __attribute__((naked))
+v4wt_copy_user_page(void *kto, const void *kfrom)
+{
+	asm("\
+	stmfd	sp!, {r4, lr}			@ 2\n\
+	mov	r2, %0				@ 1\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4\n\
+1:	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4+1\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmia	r1!, {r3, r4, ip, lr}		@ 4\n\
+	subs	r2, r2, #1			@ 1\n\
+	stmia	r0!, {r3, r4, ip, lr}		@ 4\n\
+	ldmneia	r1!, {r3, r4, ip, lr}		@ 4\n\
+	bne	1b				@ 1\n\
+	mcr	p15, 0, r2, c7, c7, 0		@ flush ID cache\n\
+	ldmfd	sp!, {r4, pc}			@ 3"
+	:
+	: "I" (PAGE_SIZE / 64));
+}
+
+void v4wt_copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr)
+{
+	void *kto, *kfrom;
+
+	kto = kmap_atomic(to, KM_USER0);
+	kfrom = kmap_atomic(from, KM_USER1);
+	v4wt_copy_user_page(kto, kfrom);
+	kunmap_atomic(kfrom, KM_USER1);
+	kunmap_atomic(kto, KM_USER0);
+}
+
+/*
+ * ARMv4 optimised clear_user_page
+ *
+ * Same story as above.
+ */
+void v4wt_clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	void *ptr, *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile("\
+	mov	r1, %2				@ 1\n\
+	mov	r2, #0				@ 1\n\
+	mov	r3, #0				@ 1\n\
+	mov	ip, #0				@ 1\n\
+	mov	lr, #0				@ 1\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"
+	: "=r" (ptr)
+	: "0" (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_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 0e21c07..4127a7b 100644
--- a/arch/arm/mm/copypage-v6.c
+++ b/arch/arm/mm/copypage-v6.c
@@ -10,8 +10,8 @@
 #include <linux/init.h>
 #include <linux/spinlock.h>
 #include <linux/mm.h>
+#include <linux/highmem.h>
 
-#include <asm/page.h>
 #include <asm/pgtable.h>
 #include <asm/shmparam.h>
 #include <asm/tlbflush.h>
@@ -33,41 +33,56 @@ static DEFINE_SPINLOCK(v6_lock);
  * Copy the user page.  No aliasing to deal with so we can just
  * attack the kernel's existing mapping of these pages.
  */
-static void v6_copy_user_page_nonaliasing(void *kto, const void *kfrom, unsigned long vaddr)
+static void v6_copy_user_highpage_nonaliasing(struct page *to,
+	struct page *from, unsigned long vaddr)
 {
+	void *kto, *kfrom;
+
+	kfrom = kmap_atomic(from, KM_USER0);
+	kto = kmap_atomic(to, KM_USER1);
 	copy_page(kto, kfrom);
+	kunmap_atomic(kto, KM_USER1);
+	kunmap_atomic(kfrom, KM_USER0);
 }
 
 /*
  * 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);
 }
 
 /*
- * Copy the page, taking account of the cache colour.
+ * Discard data in the kernel mapping for the new page.
+ * FIXME: needs this MCRR to be supported.
  */
-static void v6_copy_user_page_aliasing(void *kto, const void *kfrom, unsigned long vaddr)
+static void discard_old_kernel_data(void *kto)
 {
-	unsigned int offset = CACHE_COLOUR(vaddr);
-	unsigned long from, to;
-	struct page *page = virt_to_page(kfrom);
-
-	if (test_and_clear_bit(PG_dcache_dirty, &page->flags))
-		__flush_dcache_page(page_mapping(page), page);
-
-	/*
-	 * 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" (kto),
 	     "r" ((unsigned long)kto + PAGE_SIZE - L1_CACHE_BYTES)
 	   : "cc");
+}
+
+/*
+ * Copy the page, taking account of the cache colour.
+ */
+static void v6_copy_user_highpage_aliasing(struct page *to,
+	struct page *from, unsigned long vaddr)
+{
+	unsigned int offset = CACHE_COLOUR(vaddr);
+	unsigned long kfrom, kto;
+
+	if (test_and_clear_bit(PG_dcache_dirty, &from->flags))
+		__flush_dcache_page(page_mapping(from), from);
+
+	/* FIXME: not highmem safe */
+	discard_old_kernel_data(page_address(to));
 
 	/*
 	 * Now copy the page using the same cache colour as the
@@ -75,16 +90,16 @@ static void v6_copy_user_page_aliasing(void *kto, const void *kfrom, unsigned lo
 	 */
 	spin_lock(&v6_lock);
 
-	set_pte_ext(TOP_PTE(from_address) + offset, pfn_pte(__pa(kfrom) >> PAGE_SHIFT, PAGE_KERNEL), 0);
-	set_pte_ext(TOP_PTE(to_address) + offset, pfn_pte(__pa(kto) >> PAGE_SHIFT, PAGE_KERNEL), 0);
+	set_pte_ext(TOP_PTE(from_address) + offset, pfn_pte(page_to_pfn(from), PAGE_KERNEL), 0);
+	set_pte_ext(TOP_PTE(to_address) + offset, pfn_pte(page_to_pfn(to), PAGE_KERNEL), 0);
 
-	from = from_address + (offset << PAGE_SHIFT);
-	to   = to_address + (offset << PAGE_SHIFT);
+	kfrom = from_address + (offset << PAGE_SHIFT);
+	kto   = to_address + (offset << PAGE_SHIFT);
 
-	flush_tlb_kernel_page(from);
-	flush_tlb_kernel_page(to);
+	flush_tlb_kernel_page(kfrom);
+	flush_tlb_kernel_page(kto);
 
-	copy_page((void *)to, (void *)from);
+	copy_page((void *)kto, (void *)kfrom);
 
 	spin_unlock(&v6_lock);
 }
@@ -94,20 +109,13 @@ static void v6_copy_user_page_aliasing(void *kto, const void *kfrom, unsigned lo
  * 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
@@ -115,7 +123,7 @@ static void v6_clear_user_page_aliasing(void *kaddr, unsigned long vaddr)
 	 */
 	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);
 
@@ -123,15 +131,15 @@ static void v6_clear_user_page_aliasing(void *kaddr, unsigned long vaddr)
 }
 
 struct cpu_user_fns v6_user_fns __initdata = {
-	.cpu_clear_user_page	= v6_clear_user_page_nonaliasing,
-	.cpu_copy_user_page	= v6_copy_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_copy_user_page = v6_copy_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;
 	}
 
 	return 0;
diff --git a/arch/arm/mm/copypage-xsc3.S b/arch/arm/mm/copypage-xsc3.S
deleted file mode 100644
index 9a2cb43..0000000
--- a/arch/arm/mm/copypage-xsc3.S
+++ /dev/null
@@ -1,97 +0,0 @@
-/*
- *  linux/arch/arm/lib/copypage-xsc3.S
- *
- *  Copyright (C) 2004 Intel Corp.
- *
- * 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.
- *
- * Adapted for 3rd gen XScale core, no more mini-dcache
- * Author: Matt Gilbert (matthew.m.gilbert@intel.com)
- */
-
-#include <linux/linkage.h>
-#include <linux/init.h>
-#include <asm/asm-offsets.h>
-
-/*
- * General note:
- *  We don't really want write-allocate cache behaviour for these functions
- *  since that will just eat through 8K of the cache.
- */
-
-	.text
-	.align	5
-/*
- * XSC3 optimised copy_user_page
- *  r0 = destination
- *  r1 = source
- *  r2 = virtual user address of ultimate destination page
- *
- * The source page may have some clean entries in the cache already, but we
- * can safely ignore them - break_cow() will flush them out of the cache
- * if we eventually end up using our copied page.
- *
- */
-ENTRY(xsc3_mc_copy_user_page)
-	stmfd	sp!, {r4, r5, lr}
-	mov	lr, #PAGE_SZ/64-1
-
-	pld	[r1, #0]
-	pld	[r1, #32]
-1:	pld	[r1, #64]
-	pld	[r1, #96]
-
-2:	ldrd	r2, [r1], #8
-	mov	ip, r0
-	ldrd	r4, [r1], #8
-	mcr	p15, 0, ip, c7, c6, 1		@ invalidate
-	strd	r2, [r0], #8
-	ldrd	r2, [r1], #8
-	strd	r4, [r0], #8
-	ldrd	r4, [r1], #8
-	strd	r2, [r0], #8
-	strd	r4, [r0], #8
-	ldrd	r2, [r1], #8
-	mov	ip, r0
-	ldrd	r4, [r1], #8
-	mcr	p15, 0, ip, c7, c6, 1		@ invalidate
-	strd	r2, [r0], #8
-	ldrd	r2, [r1], #8
-	subs	lr, lr, #1
-	strd	r4, [r0], #8
-	ldrd	r4, [r1], #8
-	strd	r2, [r0], #8
-	strd	r4, [r0], #8
-	bgt	1b
-	beq	2b
-
-	ldmfd	sp!, {r4, r5, pc}
-
-	.align	5
-/*
- * XScale optimised clear_user_page
- *  r0 = destination
- *  r1 = virtual user address of ultimate destination page
- */
-ENTRY(xsc3_mc_clear_user_page)
-	mov	r1, #PAGE_SZ/32
-	mov	r2, #0
-	mov	r3, #0
-1:	mcr	p15, 0, r0, c7, c6, 1		@ invalidate line
-	strd	r2, [r0], #8
-	strd	r2, [r0], #8
-	strd	r2, [r0], #8
-	strd	r2, [r0], #8
-	subs	r1, r1, #1
-	bne	1b
-	mov	pc, lr
-
-	__INITDATA
-
-	.type	xsc3_mc_user_fns, #object
-ENTRY(xsc3_mc_user_fns)
-	.long	xsc3_mc_clear_user_page
-	.long	xsc3_mc_copy_user_page
-	.size	xsc3_mc_user_fns, . - xsc3_mc_user_fns
diff --git a/arch/arm/mm/copypage-xsc3.c b/arch/arm/mm/copypage-xsc3.c
new file mode 100644
index 0000000..39a9945
--- /dev/null
+++ b/arch/arm/mm/copypage-xsc3.c
@@ -0,0 +1,113 @@
+/*
+ *  linux/arch/arm/mm/copypage-xsc3.S
+ *
+ *  Copyright (C) 2004 Intel Corp.
+ *
+ * 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.
+ *
+ * Adapted for 3rd gen XScale core, no more mini-dcache
+ * Author: Matt Gilbert (matthew.m.gilbert@intel.com)
+ */
+#include <linux/init.h>
+#include <linux/highmem.h>
+
+/*
+ * General note:
+ *  We don't really want write-allocate cache behaviour for these functions
+ *  since that will just eat through 8K of the cache.
+ */
+
+/*
+ * XSC3 optimised copy_user_highpage
+ *  r0 = destination
+ *  r1 = source
+ *
+ * The source page may have some clean entries in the cache already, but we
+ * can safely ignore them - break_cow() will flush them out of the cache
+ * if we eventually end up using our copied page.
+ *
+ */
+static void __attribute__((naked))
+xsc3_mc_copy_user_page(void *kto, const void *kfrom)
+{
+	asm("\
+	stmfd	sp!, {r4, r5, lr}		\n\
+	mov	lr, %0				\n\
+						\n\
+	pld	[r1, #0]			\n\
+	pld	[r1, #32]			\n\
+1:	pld	[r1, #64]			\n\
+	pld	[r1, #96]			\n\
+						\n\
+2:	ldrd	r2, [r1], #8			\n\
+	mov	ip, r0				\n\
+	ldrd	r4, [r1], #8			\n\
+	mcr	p15, 0, ip, c7, c6, 1		@ invalidate\n\
+	strd	r2, [r0], #8			\n\
+	ldrd	r2, [r1], #8			\n\
+	strd	r4, [r0], #8			\n\
+	ldrd	r4, [r1], #8			\n\
+	strd	r2, [r0], #8			\n\
+	strd	r4, [r0], #8			\n\
+	ldrd	r2, [r1], #8			\n\
+	mov	ip, r0				\n\
+	ldrd	r4, [r1], #8			\n\
+	mcr	p15, 0, ip, c7, c6, 1		@ invalidate\n\
+	strd	r2, [r0], #8			\n\
+	ldrd	r2, [r1], #8			\n\
+	subs	lr, lr, #1			\n\
+	strd	r4, [r0], #8			\n\
+	ldrd	r4, [r1], #8			\n\
+	strd	r2, [r0], #8			\n\
+	strd	r4, [r0], #8			\n\
+	bgt	1b				\n\
+	beq	2b				\n\
+						\n\
+	ldmfd	sp!, {r4, r5, pc}"
+	:
+	: "I" (PAGE_SIZE / 64 - 1));
+}
+
+void xsc3_mc_copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr)
+{
+	void *kto, *kfrom;
+
+	kto = kmap_atomic(to, KM_USER0);
+	kfrom = kmap_atomic(from, KM_USER1);
+	xsc3_mc_copy_user_page(kto, kfrom);
+	kunmap_atomic(kfrom, KM_USER1);
+	kunmap_atomic(kto, KM_USER0);
+}
+
+/*
+ * XScale optimised clear_user_page
+ *  r0 = destination
+ *  r1 = virtual user address of ultimate destination page
+ */
+void xsc3_mc_clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	void *ptr, *kaddr = kmap_atomic(page, KM_USER0);
+	asm volatile ("\
+	mov	r1, %2				\n\
+	mov	r2, #0				\n\
+	mov	r3, #0				\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"
+	: "=r" (ptr)
+	: "0" (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_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 bad4933..d18f239 100644
--- a/arch/arm/mm/copypage-xscale.c
+++ b/arch/arm/mm/copypage-xscale.c
@@ -15,8 +15,8 @@
  */
 #include <linux/init.h>
 #include <linux/mm.h>
+#include <linux/highmem.h>
 
-#include <asm/page.h>
 #include <asm/pgtable.h>
 #include <asm/tlbflush.h>
 #include <asm/cacheflush.h>
@@ -35,7 +35,7 @@
 static DEFINE_SPINLOCK(minicache_lock);
 
 /*
- * XScale mini-dcache optimised copy_user_page
+ * XScale mini-dcache optimised copy_user_highpage
  *
  * We flush the destination cache lines just before we write the data into the
  * corresponding address.  Since the Dcache is read-allocate, this removes the
@@ -90,48 +90,53 @@ mc_copy_user_page(void *from, void *to)
 	: "r" (from), "r" (to), "I" (PAGE_SIZE / 64 - 1));
 }
 
-void xscale_mc_copy_user_page(void *kto, const void *kfrom, unsigned long vaddr)
+void xscale_mc_copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr)
 {
-	struct page *page = virt_to_page(kfrom);
+	void *kto = kmap_atomic(to, KM_USER1);
 
-	if (test_and_clear_bit(PG_dcache_dirty, &page->flags))
-		__flush_dcache_page(page_mapping(page), page);
+	if (test_and_clear_bit(PG_dcache_dirty, &from->flags))
+		__flush_dcache_page(page_mapping(from), from);
 
 	spin_lock(&minicache_lock);
 
-	set_pte_ext(TOP_PTE(COPYPAGE_MINICACHE), pfn_pte(__pa(kfrom) >> PAGE_SHIFT, minicache_pgprot), 0);
+	set_pte_ext(TOP_PTE(COPYPAGE_MINICACHE), pfn_pte(page_to_pfn(from), minicache_pgprot), 0);
 	flush_tlb_kernel_page(COPYPAGE_MINICACHE);
 
 	mc_copy_user_page((void *)COPYPAGE_MINICACHE, kto);
 
 	spin_unlock(&minicache_lock);
+
+	kunmap_atomic(kto, KM_USER1);
 }
 
 /*
  * 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 *ptr, *kaddr = kmap_atomic(page, KM_USER0);
 	asm volatile(
-	"mov	r1, %0				\n\
+	"mov	r1, %2				\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"
-	:
-	: "I" (PAGE_SIZE / 32));
+	bne	1b"
+	: "=r" (ptr)
+	: "0" (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_copy_user_page	= xscale_mc_copy_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/fault.c b/arch/arm/mm/fault.c
index 2df8d9f..ffd8b22 100644
--- a/arch/arm/mm/fault.c
+++ b/arch/arm/mm/fault.c
@@ -14,6 +14,7 @@
 #include <linux/init.h>
 #include <linux/kprobes.h>
 #include <linux/uaccess.h>
+#include <linux/page-flags.h>
 
 #include <asm/system.h>
 #include <asm/pgtable.h>
@@ -83,13 +84,14 @@ void show_pte(struct mm_struct *mm, unsigned long addr)
 			break;
 		}
 
-#ifndef CONFIG_HIGHMEM
 		/* We must not map this if we have highmem enabled */
+		if (PageHighMem(pfn_to_page(pmd_val(*pmd) >> PAGE_SHIFT)))
+			break;
+
 		pte = pte_offset_map(pmd, addr);
 		printk(", *pte=%08lx", pte_val(*pte));
 		printk(", *ppte=%08lx", pte_val(pte[-PTRS_PER_PTE]));
 		pte_unmap(pte);
-#endif
 	} while(0);
 
 	printk("\n");
diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
index 82c4b42..ab5c9ab 100644
--- a/arch/arm/mm/init.c
+++ b/arch/arm/mm/init.c
@@ -64,10 +64,11 @@ static int __init parse_tag_initrd2(const struct tag *tag)
 __tagtable(ATAG_INITRD2, parse_tag_initrd2);
 
 /*
- * This is used to pass memory configuration data from paging_init
- * to mem_init, and by show_mem() to skip holes in the memory map.
+ * This keeps memory configuration data used by a couple memory
+ * initialization functions, as well as show_mem() for the skipping
+ * of holes in the memory map.  It is populated by arm_add_memory().
  */
-static struct meminfo meminfo = { 0, };
+struct meminfo meminfo;
 
 void show_mem(void)
 {
@@ -331,13 +332,12 @@ static void __init bootmem_free_node(int node, struct meminfo *mi)
 	free_area_init_node(node, zone_size, start_pfn, zhole_size);
 }
 
-void __init bootmem_init(struct meminfo *mi)
+void __init bootmem_init(void)
 {
+	struct meminfo *mi = &meminfo;
 	unsigned long memend_pfn = 0;
 	int node, initrd_node;
 
-	memcpy(&meminfo, mi, sizeof(meminfo));
-
 	/*
 	 * Locate which node contains the ramdisk image, if any.
 	 */
@@ -394,20 +394,22 @@ void __init bootmem_init(struct meminfo *mi)
 	max_pfn = max_low_pfn = memend_pfn - PHYS_PFN_OFFSET;
 }
 
-static inline void free_area(unsigned long addr, unsigned long end, char *s)
+static inline int free_area(unsigned long pfn, unsigned long end, char *s)
 {
-	unsigned int size = (end - addr) >> 10;
+	unsigned int pages = 0, size = (end - pfn) << (PAGE_SHIFT - 10);
 
-	for (; addr < end; addr += PAGE_SIZE) {
-		struct page *page = virt_to_page(addr);
+	for (; pfn < end; pfn++) {
+		struct page *page = pfn_to_page(pfn);
 		ClearPageReserved(page);
 		init_page_count(page);
-		free_page(addr);
-		totalram_pages++;
+		__free_page(page);
+		pages++;
 	}
 
 	if (size && s)
 		printk(KERN_INFO "Freeing %s memory: %dK\n", s, size);
+
+	return pages;
 }
 
 static inline void
@@ -478,13 +480,9 @@ static void __init free_unused_memmap_node(int node, struct meminfo *mi)
  */
 void __init mem_init(void)
 {
-	unsigned int codepages, datapages, initpages;
+	unsigned int codesize, datasize, initsize;
 	int i, node;
 
-	codepages = &_etext - &_text;
-	datapages = &_end - &__data_start;
-	initpages = &__init_end - &__init_begin;
-
 #ifndef CONFIG_DISCONTIGMEM
 	max_mapnr   = virt_to_page(high_memory) - mem_map;
 #endif
@@ -501,7 +499,8 @@ void __init mem_init(void)
 
 #ifdef CONFIG_SA1111
 	/* now that our DMA memory is actually so designated, we can free it */
-	free_area(PAGE_OFFSET, (unsigned long)swapper_pg_dir, NULL);
+	totalram_pages += free_area(PHYS_PFN_OFFSET,
+				    __phys_to_pfn(__pa(swapper_pg_dir)), NULL);
 #endif
 
 	/*
@@ -509,18 +508,21 @@ void __init mem_init(void)
 	 * real number of pages we have in this system
 	 */
 	printk(KERN_INFO "Memory:");
-
 	num_physpages = 0;
 	for (i = 0; i < meminfo.nr_banks; i++) {
 		num_physpages += bank_pfn_size(&meminfo.bank[i]);
 		printk(" %ldMB", bank_phys_size(&meminfo.bank[i]) >> 20);
 	}
-
 	printk(" = %luMB total\n", num_physpages >> (20 - PAGE_SHIFT));
+
+	codesize = &_etext - &_text;
+	datasize = &_end - &__data_start;
+	initsize = &__init_end - &__init_begin;
+
 	printk(KERN_NOTICE "Memory: %luKB available (%dK code, "
 		"%dK data, %dK init)\n",
 		(unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
-		codepages >> 10, datapages >> 10, initpages >> 10);
+		codesize >> 10, datasize >> 10, initsize >> 10);
 
 	if (PAGE_SIZE >= 16384 && num_physpages <= 128) {
 		extern int sysctl_overcommit_memory;
@@ -535,11 +537,10 @@ void __init mem_init(void)
 
 void free_initmem(void)
 {
-	if (!machine_is_integrator() && !machine_is_cintegrator()) {
-		free_area((unsigned long)(&__init_begin),
-			  (unsigned long)(&__init_end),
-			  "init");
-	}
+	if (!machine_is_integrator() && !machine_is_cintegrator())
+		totalram_pages += free_area(__phys_to_pfn(__pa(&__init_begin)),
+					    __phys_to_pfn(__pa(&__init_end)),
+					    "init");
 }
 
 #ifdef CONFIG_BLK_DEV_INITRD
@@ -549,7 +550,9 @@ static int keep_initrd;
 void free_initrd_mem(unsigned long start, unsigned long end)
 {
 	if (!keep_initrd)
-		free_area(start, end, "initrd");
+		totalram_pages += free_area(__phys_to_pfn(__pa(start)),
+					    __phys_to_pfn(__pa(end)),
+					    "initrd");
 }
 
 static int __init keepinitrd_setup(char *__unused)
diff --git a/arch/arm/mm/mm.h b/arch/arm/mm/mm.h
index 5d9f539..94367bd 100644
--- a/arch/arm/mm/mm.h
+++ b/arch/arm/mm/mm.h
@@ -32,7 +32,7 @@ struct meminfo;
 struct pglist_data;
 
 void __init create_mapping(struct map_desc *md);
-void __init bootmem_init(struct meminfo *mi);
+void __init bootmem_init(void);
 void reserve_node_zero(struct pglist_data *pgdat);
 
 extern void _text, _stext, _etext, __data_start, _end, __init_begin, __init_end;
diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index f24803c..c0b9a78 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -646,61 +646,79 @@ static void __init early_vmalloc(char **arg)
 			"vmalloc area too small, limiting to %luMB\n",
 			vmalloc_reserve >> 20);
 	}
+
+	if (vmalloc_reserve > VMALLOC_END - (PAGE_OFFSET + SZ_32M)) {
+		vmalloc_reserve = VMALLOC_END - (PAGE_OFFSET + SZ_32M);
+		printk(KERN_WARNING
+			"vmalloc area is too big, limiting to %luMB\n",
+			vmalloc_reserve >> 20);
+	}
 }
 __early_param("vmalloc=", early_vmalloc);
 
 #define VMALLOC_MIN	(void *)(VMALLOC_END - vmalloc_reserve)
 
-static int __init check_membank_valid(struct membank *mb)
+static void __init sanity_check_meminfo(void)
 {
-	/*
-	 * Check whether this memory region has non-zero size or
-	 * invalid node number.
-	 */
-	if (mb->size == 0 || mb->node >= MAX_NUMNODES)
-		return 0;
-
-	/*
-	 * Check whether this memory region would entirely overlap
-	 * the vmalloc area.
-	 */
-	if (phys_to_virt(mb->start) >= VMALLOC_MIN) {
-		printk(KERN_NOTICE "Ignoring RAM at %.8lx-%.8lx "
-			"(vmalloc region overlap).\n",
-			mb->start, mb->start + mb->size - 1);
-		return 0;
-	}
-
-	/*
-	 * Check whether this memory region would partially overlap
-	 * the vmalloc area.
-	 */
-	if (phys_to_virt(mb->start + mb->size) < phys_to_virt(mb->start) ||
-	    phys_to_virt(mb->start + mb->size) > VMALLOC_MIN) {
-		unsigned long newsize = VMALLOC_MIN - phys_to_virt(mb->start);
-
-		printk(KERN_NOTICE "Truncating RAM at %.8lx-%.8lx "
-			"to -%.8lx (vmalloc region overlap).\n",
-			mb->start, mb->start + mb->size - 1,
-			mb->start + newsize - 1);
-		mb->size = newsize;
-	}
+	int i, j;
 
-	return 1;
-}
+	for (i = 0, j = 0; i < meminfo.nr_banks; i++) {
+		struct membank *bank = &meminfo.bank[j];
+		*bank = meminfo.bank[i];
 
-static void __init sanity_check_meminfo(struct meminfo *mi)
-{
-	int i, j;
+#ifdef CONFIG_HIGHMEM
+		/*
+		 * Split those memory banks which are partially overlapping
+		 * the vmalloc area greatly simplifying things later.
+		 */
+		if (__va(bank->start) < VMALLOC_MIN &&
+		    bank->size > VMALLOC_MIN - __va(bank->start)) {
+			if (meminfo.nr_banks >= NR_BANKS) {
+				printk(KERN_CRIT "NR_BANKS too low, "
+						 "ignoring high memory\n");
+			} else {
+				memmove(bank + 1, bank,
+					(meminfo.nr_banks - i) * sizeof(*bank));
+				meminfo.nr_banks++;
+				i++;
+				bank[1].size -= VMALLOC_MIN - __va(bank->start);
+				bank[1].start = __pa(VMALLOC_MIN - 1) + 1;
+				j++;
+			}
+			bank->size = VMALLOC_MIN - __va(bank->start);
+		}
+#else
+		/*
+		 * Check whether this memory bank would entirely overlap
+		 * the vmalloc area.
+		 */
+		if (__va(bank->start) >= VMALLOC_MIN) {
+			printk(KERN_NOTICE "Ignoring RAM at %.8lx-%.8lx "
+			       "(vmalloc region overlap).\n",
+			       bank->start, bank->start + bank->size - 1);
+			continue;
+		}
 
-	for (i = 0, j = 0; i < mi->nr_banks; i++) {
-		if (check_membank_valid(&mi->bank[i]))
-			mi->bank[j++] = mi->bank[i];
+		/*
+		 * Check whether this memory bank would partially overlap
+		 * the vmalloc area.
+		 */
+		if (__va(bank->start + bank->size) > VMALLOC_MIN ||
+		    __va(bank->start + bank->size) < __va(bank->start)) {
+			unsigned long newsize = VMALLOC_MIN - __va(bank->start);
+			printk(KERN_NOTICE "Truncating RAM at %.8lx-%.8lx "
+			       "to -%.8lx (vmalloc region overlap).\n",
+			       bank->start, bank->start + bank->size - 1,
+			       bank->start + newsize - 1);
+			bank->size = newsize;
+		}
+#endif
+		j++;
 	}
-	mi->nr_banks = j;
+	meminfo.nr_banks = j;
 }
 
-static inline void prepare_page_table(struct meminfo *mi)
+static inline void prepare_page_table(void)
 {
 	unsigned long addr;
 
@@ -721,7 +739,7 @@ static inline void prepare_page_table(struct meminfo *mi)
 	 * Clear out all the kernel space mappings, except for the first
 	 * memory bank, up to the end of the vmalloc region.
 	 */
-	for (addr = __phys_to_virt(mi->bank[0].start + mi->bank[0].size);
+	for (addr = __phys_to_virt(bank_phys_end(&meminfo.bank[0]));
 	     addr < VMALLOC_END; addr += PGDIR_SIZE)
 		pmd_clear(pmd_off_k(addr));
 }
@@ -880,14 +898,14 @@ static void __init devicemaps_init(struct machine_desc *mdesc)
  * paging_init() sets up the page tables, initialises the zone memory
  * maps, and sets up the zero page, bad page and bad page tables.
  */
-void __init paging_init(struct meminfo *mi, struct machine_desc *mdesc)
+void __init paging_init(struct machine_desc *mdesc)
 {
 	void *zero_page;
 
 	build_mem_type_table();
-	sanity_check_meminfo(mi);
-	prepare_page_table(mi);
-	bootmem_init(mi);
+	sanity_check_meminfo();
+	prepare_page_table();
+	bootmem_init();
 	devicemaps_init(mdesc);
 
 	top_pmd = pmd_off_k(0xffff0000);
diff --git a/arch/arm/mm/nommu.c b/arch/arm/mm/nommu.c
index 07b62b2..c085f4e 100644
--- a/arch/arm/mm/nommu.c
+++ b/arch/arm/mm/nommu.c
@@ -41,27 +41,13 @@ void __init reserve_node_zero(pg_data_t *pgdat)
 			BOOTMEM_DEFAULT);
 }
 
-static void __init sanity_check_meminfo(struct meminfo *mi)
-{
-	int i, j;
-
-	for (i = 0, j = 0; i < mi->nr_banks; i++) {
-		struct membank *mb = &mi->bank[i];
-
-		if (mb->size != 0 && mb->node < MAX_NUMNODES)
-			mi->bank[j++] = mi->bank[i];
-	}
-	mi->nr_banks = j;
-}
-
 /*
  * paging_init() sets up the page tables, initialises the zone memory
  * maps, and sets up the zero page, bad page and bad page tables.
  */
-void __init paging_init(struct meminfo *mi, struct machine_desc *mdesc)
+void __init paging_init(struct machine_desc *mdesc)
 {
-	sanity_check_meminfo(mi);
-	bootmem_init(mi);
+	bootmem_init();
 }
 
 /*
diff --git a/arch/arm/mm/proc-syms.c b/arch/arm/mm/proc-syms.c
index 2b5ba39..4ad3bf2 100644
--- a/arch/arm/mm/proc-syms.c
+++ b/arch/arm/mm/proc-syms.c
@@ -33,8 +33,8 @@ EXPORT_SYMBOL(cpu_cache);
 
 #ifdef CONFIG_MMU
 #ifndef MULTI_USER
-EXPORT_SYMBOL(__cpu_clear_user_page);
-EXPORT_SYMBOL(__cpu_copy_user_page);
+EXPORT_SYMBOL(__cpu_clear_user_highpage);
+EXPORT_SYMBOL(__cpu_copy_user_highpage);
 #else
 EXPORT_SYMBOL(cpu_user);
 #endif