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 7eb0d320d24..00000000000
--- 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 00000000000..c3ba6a94da0
--- /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 2ee394b11bc..00000000000
--- 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 00000000000..70ed96c8af8
--- /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 8d33e254934..bdb5fd983b1 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 83117354b1c..00000000000
--- 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 00000000000..3ec93dab765
--- /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 e1f2af28d54..00000000000
--- 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 00000000000..0f1188efae4
--- /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 0e21c076758..4127a7bddfe 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");
+	/*