Merge branch 'ptebits' into devel

Conflicts:

	arch/arm/Kconfig

29 files changed, 333 insertions, 468 deletions

diff --git a/arch/arm/mm/copypage-v4mc.c b/arch/arm/mm/copypage-v4mc.c
index ded0e96d069..8d33e254934 100644
--- a/arch/arm/mm/copypage-v4mc.c
+++ b/arch/arm/mm/copypage-v4mc.c
@@ -28,7 +28,7 @@
  * specific hacks for copying pages efficiently.
  */
 #define minicache_pgprot __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | \
-				  L_PTE_CACHEABLE)
+				  L_PTE_MT_MINICACHE)
 
 static DEFINE_SPINLOCK(minicache_lock);
 
diff --git a/arch/arm/mm/copypage-xscale.c b/arch/arm/mm/copypage-xscale.c
index 2e455f82a4d..bad49331bbf 100644
--- a/arch/arm/mm/copypage-xscale.c
+++ b/arch/arm/mm/copypage-xscale.c
@@ -30,7 +30,7 @@
 #define COPYPAGE_MINICACHE	0xffff8000
 
 #define minicache_pgprot __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | \
-				  L_PTE_CACHEABLE)
+				  L_PTE_MT_MINICACHE)
 
 static DEFINE_SPINLOCK(minicache_lock);
 
diff --git a/arch/arm/mm/fault-armv.c b/arch/arm/mm/fault-armv.c
index af6ed6ef9a8..81d0b8772de 100644
--- a/arch/arm/mm/fault-armv.c
+++ b/arch/arm/mm/fault-armv.c
@@ -23,7 +23,7 @@
 #include <asm/pgtable.h>
 #include <asm/tlbflush.h>
 
-static unsigned long shared_pte_mask = L_PTE_CACHEABLE;
+static unsigned long shared_pte_mask = L_PTE_MT_BUFFERABLE;
 
 /*
  * We take the easy way out of this problem - we make the
@@ -65,9 +65,10 @@ static int adjust_pte(struct vm_area_struct *vma, unsigned long address)
 	 * If this page isn't present, or is already setup to
 	 * fault (ie, is old), we can safely ignore any issues.
 	 */
-	if (ret && pte_val(entry) & shared_pte_mask) {
+	if (ret && (pte_val(entry) & L_PTE_MT_MASK) != shared_pte_mask) {
 		flush_cache_page(vma, address, pte_pfn(entry));
-		pte_val(entry) &= ~shared_pte_mask;
+		pte_val(entry) &= ~L_PTE_MT_MASK;
+		pte_val(entry) |= shared_pte_mask;
 		set_pte_at(vma->vm_mm, address, pte, entry);
 		flush_tlb_page(vma, address);
 	}
@@ -199,7 +200,7 @@ void __init check_writebuffer_bugs(void)
 		unsigned long *p1, *p2;
 		pgprot_t prot = __pgprot(L_PTE_PRESENT|L_PTE_YOUNG|
 					 L_PTE_DIRTY|L_PTE_WRITE|
-					 L_PTE_BUFFERABLE);
+					 L_PTE_MT_BUFFERABLE);
 
 		p1 = vmap(&page, 1, VM_IOREMAP, prot);
 		p2 = vmap(&page, 1, VM_IOREMAP, prot);
@@ -220,7 +221,7 @@ void __init check_writebuffer_bugs(void)
 
 	if (v) {
 		printk("failed, %s\n", reason);
-		shared_pte_mask |= L_PTE_BUFFERABLE;
+		shared_pte_mask = L_PTE_MT_UNCACHED;
 	} else {
 		printk("ok\n");
 	}
diff --git a/arch/arm/mm/ioremap.c b/arch/arm/mm/ioremap.c
index 8a41912ec7c..18373f73f2f 100644
--- a/arch/arm/mm/ioremap.c
+++ b/arch/arm/mm/ioremap.c
@@ -56,8 +56,7 @@ static int remap_area_pte(pmd_t *pmd, unsigned long addr, unsigned long end,
 		if (!pte_none(*pte))
 			goto bad;
 
-		set_pte_ext(pte, pfn_pte(phys_addr >> PAGE_SHIFT, prot),
-			    type->prot_pte_ext);
+		set_pte_ext(pte, pfn_pte(phys_addr >> PAGE_SHIFT, prot), 0);
 		phys_addr += PAGE_SIZE;
 	} while (pte++, addr += PAGE_SIZE, addr != end);
 	return 0;
diff --git a/arch/arm/mm/mm.h b/arch/arm/mm/mm.h
index 96590104ba0..5d9f53907b4 100644
--- a/arch/arm/mm/mm.h
+++ b/arch/arm/mm/mm.h
@@ -18,7 +18,6 @@ static inline pmd_t *pmd_off_k(unsigned long virt)
 
 struct mem_type {
 	unsigned int prot_pte;
-	unsigned int prot_pte_ext;
 	unsigned int prot_l1;
 	unsigned int prot_sect;
 	unsigned int domain;
diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index e7af83e569d..8ba75406455 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -66,27 +66,27 @@ static struct cachepolicy cache_policies[] __initdata = {
 		.policy		= "uncached",
 		.cr_mask	= CR_W|CR_C,
 		.pmd		= PMD_SECT_UNCACHED,
-		.pte		= 0,
+		.pte		= L_PTE_MT_UNCACHED,
 	}, {
 		.policy		= "buffered",
 		.cr_mask	= CR_C,
 		.pmd		= PMD_SECT_BUFFERED,
-		.pte		= PTE_BUFFERABLE,
+		.pte		= L_PTE_MT_BUFFERABLE,
 	}, {
 		.policy		= "writethrough",
 		.cr_mask	= 0,
 		.pmd		= PMD_SECT_WT,
-		.pte		= PTE_CACHEABLE,
+		.pte		= L_PTE_MT_WRITETHROUGH,
 	}, {
 		.policy		= "writeback",
 		.cr_mask	= 0,
 		.pmd		= PMD_SECT_WB,
-		.pte		= PTE_BUFFERABLE|PTE_CACHEABLE,
+		.pte		= L_PTE_MT_WRITEBACK,
 	}, {
 		.policy		= "writealloc",
 		.cr_mask	= 0,
 		.pmd		= PMD_SECT_WBWA,
-		.pte		= PTE_BUFFERABLE|PTE_CACHEABLE,
+		.pte		= L_PTE_MT_WRITEALLOC,
 	}
 };
 
@@ -184,29 +184,28 @@ void adjust_cr(unsigned long mask, unsigned long set)
 
 static struct mem_type mem_types[] = {
 	[MT_DEVICE] = {		  /* Strongly ordered / ARMv6 shared device */
-		.prot_pte	= PROT_PTE_DEVICE,
+		.prot_pte	= PROT_PTE_DEVICE | L_PTE_MT_DEV_SHARED |
+				  L_PTE_SHARED,
 		.prot_l1	= PMD_TYPE_TABLE,
 		.prot_sect	= PROT_SECT_DEVICE | PMD_SECT_UNCACHED,
 		.domain		= DOMAIN_IO,
 	},
 	[MT_DEVICE_NONSHARED] = { /* ARMv6 non-shared device */
-		.prot_pte	= PROT_PTE_DEVICE,
-		.prot_pte_ext	= PTE_EXT_TEX(2),
+		.prot_pte	= PROT_PTE_DEVICE | L_PTE_MT_DEV_NONSHARED,
 		.prot_l1	= PMD_TYPE_TABLE,
 		.prot_sect	= PROT_SECT_DEVICE | PMD_SECT_TEX(2),
 		.domain		= DOMAIN_IO,
 	},
 	[MT_DEVICE_CACHED] = {	  /* ioremap_cached */
-		.prot_pte	= PROT_PTE_DEVICE | L_PTE_CACHEABLE | L_PTE_BUFFERABLE,
+		.prot_pte	= PROT_PTE_DEVICE | L_PTE_MT_DEV_CACHED,
 		.prot_l1	= PMD_TYPE_TABLE,
 		.prot_sect	= PROT_SECT_DEVICE | PMD_SECT_WB,
 		.domain		= DOMAIN_IO,
 	},	
-	[MT_DEVICE_IXP2000] = {	  /* IXP2400 requires XCB=101 for on-chip I/O */
-		.prot_pte	= PROT_PTE_DEVICE,
+	[MT_DEVICE_WC] = {	/* ioremap_wc */
+		.prot_pte	= PROT_PTE_DEVICE | L_PTE_MT_DEV_WC,
 		.prot_l1	= PMD_TYPE_TABLE,
-		.prot_sect	= PROT_SECT_DEVICE | PMD_SECT_BUFFERABLE |
-				  PMD_SECT_TEX(1),
+		.prot_sect	= PROT_SECT_DEVICE | PMD_SECT_BUFFERABLE,
 		.domain		= DOMAIN_IO,
 	},
 	[MT_CACHECLEAN] = {
@@ -251,7 +250,7 @@ static void __init build_mem_type_table(void)
 {
 	struct cachepolicy *cp;
 	unsigned int cr = get_cr();
-	unsigned int user_pgprot, kern_pgprot;
+	unsigned int user_pgprot, kern_pgprot, vecs_pgprot;
 	int cpu_arch = cpu_architecture();
 	int i;
 
@@ -269,6 +268,20 @@ static void __init build_mem_type_table(void)
 			cachepolicy = CPOLICY_WRITEBACK;
 		ecc_mask = 0;
 	}
+#ifdef CONFIG_SMP
+	cachepolicy = CPOLICY_WRITEALLOC;
+#endif
+
+	/*
+	 * On non-Xscale3 ARMv5-and-older systems, use CB=01
+	 * (Uncached/Buffered) for ioremap_wc() mappings.  On XScale3
+	 * and ARMv6+, use TEXCB=00100 mappings (Inner/Outer Uncacheable
+	 * in xsc3 parlance, Uncached Normal in ARMv6 parlance).
+	 */
+	if (cpu_is_xsc3() || cpu_arch >= CPU_ARCH_ARMv6) {
+		mem_types[MT_DEVICE_WC].prot_sect |= PMD_SECT_TEX(1);
+		mem_types[MT_DEVICE_WC].prot_sect &= ~PMD_SECT_BUFFERABLE;
+	}
 
 	/*
 	 * ARMv5 and lower, bit 4 must be set for page tables.
@@ -290,7 +303,15 @@ static void __init build_mem_type_table(void)
 	}
 
 	cp = &cache_policies[cachepolicy];
-	kern_pgprot = user_pgprot = cp->pte;
+	vecs_pgprot = kern_pgprot = user_pgprot = cp->pte;
+
+#ifndef CONFIG_SMP
+	/*
+	 * Only use write-through for non-SMP systems
+	 */
+	if (cpu_arch >= CPU_ARCH_ARMv5 && cachepolicy > CPOLICY_WRITETHROUGH)
+		vecs_pgprot = cache_policies[CPOLICY_WRITETHROUGH].pte;
+#endif
 
 	/*
 	 * Enable CPU-specific coherency if supported.
@@ -318,7 +339,6 @@ static void __init build_mem_type_table(void)
 		/*
 		 * Mark the device area as "shared device"
 		 */
-		mem_types[MT_DEVICE].prot_pte |= L_PTE_BUFFERABLE;
 		mem_types[MT_DEVICE].prot_sect |= PMD_SECT_BUFFERED;
 
 #ifdef CONFIG_SMP
@@ -327,30 +347,21 @@ static void __init build_mem_type_table(void)
 		 */
 		user_pgprot |= L_PTE_SHARED;
 		kern_pgprot |= L_PTE_SHARED;
+		vecs_pgprot |= L_PTE_SHARED;
 		mem_types[MT_MEMORY].prot_sect |= PMD_SECT_S;
 #endif
 	}
 
 	for (i = 0; i < 16; i++) {
 		unsigned long v = pgprot_val(protection_map[i]);
-		v = (v & ~(L_PTE_BUFFERABLE|L_PTE_CACHEABLE)) | user_pgprot;
-		protection_map[i] = __pgprot(v);
+		protection_map[i] = __pgprot(v | user_pgprot);
 	}
 
-	mem_types[MT_LOW_VECTORS].prot_pte |= kern_pgprot;
-	mem_types[MT_HIGH_VECTORS].prot_pte |= kern_pgprot;
+	mem_types[MT_LOW_VECTORS].prot_pte |= vecs_pgprot;
+	mem_types[MT_HIGH_VECTORS].prot_pte |= vecs_pgprot;
 
-	if (cpu_arch >= CPU_ARCH_ARMv5) {
-#ifndef CONFIG_SMP
-		/*
-		 * Only use write-through for non-SMP systems
-		 */
-		mem_types[MT_LOW_VECTORS].prot_pte &= ~L_PTE_BUFFERABLE;
-		mem_types[MT_HIGH_VECTORS].prot_pte &= ~L_PTE_BUFFERABLE;
-#endif
-	} else {
+	if (cpu_arch < CPU_ARCH_ARMv5)
 		mem_types[MT_MINICLEAN].prot_sect &= ~PMD_SECT_TEX(1);
-	}
 
 	pgprot_user   = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | user_pgprot);
 	pgprot_kernel = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG |
@@ -398,8 +409,7 @@ static void __init alloc_init_pte(pmd_t *pmd, unsigned long addr,
 
 	pte = pte_offset_kernel(pmd, addr);
 	do {
-		set_pte_ext(pte, pfn_pte(pfn, __pgprot(type->prot_pte)),
-			    type->prot_pte_ext);
+		set_pte_ext(pte, pfn_pte(pfn, __pgprot(type->prot_pte)), 0);
 		pfn++;
 	} while (pte++, addr += PAGE_SIZE, addr != end);
 }
diff --git a/arch/arm/mm/proc-arm1020.S b/arch/arm/mm/proc-arm1020.S
index 5673f4d6113..b5551bf010a 100644
--- a/arch/arm/mm/proc-arm1020.S
+++ b/arch/arm/mm/proc-arm1020.S
@@ -29,7 +29,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
@@ -399,29 +399,7 @@ ENTRY(cpu_arm1020_switch_mm)
 	.align	5
 ENTRY(cpu_arm1020_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r1, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 #ifndef CONFIG_CPU_DCACHE_DISABLE
 	mcr	p15, 0, r0, c7, c10, 4
diff --git a/arch/arm/mm/proc-arm1020e.S b/arch/arm/mm/proc-arm1020e.S
index 4343fdb0e9e..8bc6740c29e 100644
--- a/arch/arm/mm/proc-arm1020e.S
+++ b/arch/arm/mm/proc-arm1020e.S
@@ -29,7 +29,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
@@ -383,29 +383,7 @@ ENTRY(cpu_arm1020e_switch_mm)
 	.align	5
 ENTRY(cpu_arm1020e_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r1, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 #ifndef CONFIG_CPU_DCACHE_DISABLE
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
diff --git a/arch/arm/mm/proc-arm1022.S b/arch/arm/mm/proc-arm1022.S
index 2a4ea1659e9..2cd03e66c0a 100644
--- a/arch/arm/mm/proc-arm1022.S
+++ b/arch/arm/mm/proc-arm1022.S
@@ -18,7 +18,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
@@ -365,29 +365,7 @@ ENTRY(cpu_arm1022_switch_mm)
 	.align	5
 ENTRY(cpu_arm1022_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r1, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 #ifndef CONFIG_CPU_DCACHE_DISABLE
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
diff --git a/arch/arm/mm/proc-arm1026.S b/arch/arm/mm/proc-arm1026.S
index 77a1babd421..ad961a897f6 100644
--- a/arch/arm/mm/proc-arm1026.S
+++ b/arch/arm/mm/proc-arm1026.S
@@ -18,7 +18,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
@@ -354,29 +354,7 @@ ENTRY(cpu_arm1026_switch_mm)
 	.align	5
 ENTRY(cpu_arm1026_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r1, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 #ifndef CONFIG_CPU_DCACHE_DISABLE
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index c371fc87776..80d6e1de069 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -15,11 +15,13 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
 
+#include "proc-macros.S"
+
 ENTRY(cpu_arm6_dcache_clean_area)
 ENTRY(cpu_arm7_dcache_clean_area)
 		mov	pc, lr
@@ -214,30 +216,13 @@ ENTRY(cpu_arm7_switch_mm)
  *	   : r1 = value to set
  * Purpose : Set a PTE and flush it out of any WB cache
  */
-		.align	5
+	.align	5
 ENTRY(cpu_arm6_set_pte_ext)
 ENTRY(cpu_arm7_set_pte_ext)
 #ifdef CONFIG_MMU
-		str	r1, [r0], #-2048		@ linux version
-
-		eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-		bic	r2, r1, #PTE_SMALL_AP_MASK
-		bic	r2, r2, #PTE_TYPE_MASK
-		orr	r2, r2, #PTE_TYPE_SMALL
-
-		tst	r1, #L_PTE_USER			@ User?
-		orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-		tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-		orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-		tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young
-		movne	r2, #0
-
-		str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext wc_disable=0
 #endif /* CONFIG_MMU */
-		mov	pc, lr
+	mov	pc, lr
 
 /*
  * Function: _arm6_7_reset
diff --git a/arch/arm/mm/proc-arm720.S b/arch/arm/mm/proc-arm720.S
index eda733d3045..85ae18695f1 100644
--- a/arch/arm/mm/proc-arm720.S
+++ b/arch/arm/mm/proc-arm720.S
@@ -36,7 +36,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
@@ -93,29 +93,12 @@ ENTRY(cpu_arm720_switch_mm)
  *	   : r1 = value to set
  * Purpose : Set a PTE and flush it out of any WB cache
  */
-		.align	5
+	.align	5
 ENTRY(cpu_arm720_set_pte_ext)
 #ifdef CONFIG_MMU
-		str	r1, [r0], #-2048		@ linux version
-
-		eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-		bic	r2, r1, #PTE_SMALL_AP_MASK
-		bic	r2, r2, #PTE_TYPE_MASK
-		orr	r2, r2, #PTE_TYPE_SMALL
-
-		tst	r1, #L_PTE_USER			@ User?
-		orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-		tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-		orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-		tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young
-		movne	r2, #0
-
-		str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext wc_disable=0
 #endif
-		mov	pc, lr
+	mov	pc, lr
 
 /*
  * Function: arm720_reset
diff --git a/arch/arm/mm/proc-arm740.S b/arch/arm/mm/proc-arm740.S
index 3a57376c8bc..4f95bee63e9 100644
--- a/arch/arm/mm/proc-arm740.S
+++ b/arch/arm/mm/proc-arm740.S
@@ -12,7 +12,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
diff --git a/arch/arm/mm/proc-arm7tdmi.S b/arch/arm/mm/proc-arm7tdmi.S
index 7b3ecdeb537..93e05fa7bed 100644
--- a/arch/arm/mm/proc-arm7tdmi.S
+++ b/arch/arm/mm/proc-arm7tdmi.S
@@ -12,7 +12,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
diff --git a/arch/arm/mm/proc-arm920.S b/arch/arm/mm/proc-arm920.S
index 28cdb060df4..914d688394f 100644
--- a/arch/arm/mm/proc-arm920.S
+++ b/arch/arm/mm/proc-arm920.S
@@ -28,7 +28,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -351,33 +351,11 @@ ENTRY(cpu_arm920_switch_mm)
 	.align	5
 ENTRY(cpu_arm920_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r2, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
 	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
-#endif /* CONFIG_MMU */
+#endif
 	mov	pc, lr
 
 	__INIT
diff --git a/arch/arm/mm/proc-arm922.S b/arch/arm/mm/proc-arm922.S
index 94ddcb4a4b7..51c9c9859e5 100644
--- a/arch/arm/mm/proc-arm922.S
+++ b/arch/arm/mm/proc-arm922.S
@@ -29,7 +29,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -355,29 +355,7 @@ ENTRY(cpu_arm922_switch_mm)
 	.align	5
 ENTRY(cpu_arm922_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r2, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
 	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
diff --git a/arch/arm/mm/proc-arm925.S b/arch/arm/mm/proc-arm925.S
index d045812f339..2724526d89c 100644
--- a/arch/arm/mm/proc-arm925.S
+++ b/arch/arm/mm/proc-arm925.S
@@ -52,7 +52,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -398,29 +398,7 @@ ENTRY(cpu_arm925_switch_mm)
 	.align	5
 ENTRY(cpu_arm925_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r2, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 #ifndef CONFIG_CPU_DCACHE_WRITETHROUGH
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
diff --git a/arch/arm/mm/proc-arm926.S b/arch/arm/mm/proc-arm926.S
index 4cd33169a7c..54466937bff 100644
--- a/arch/arm/mm/proc-arm926.S
+++ b/arch/arm/mm/proc-arm926.S
@@ -28,7 +28,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -359,29 +359,7 @@ ENTRY(cpu_arm926_switch_mm)
 	.align	5
 ENTRY(cpu_arm926_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
-	eor	r3, r2, #0x0a			@ C & small page?
-	tst	r3, #0x0b
-	biceq	r2, r2, #4
-#endif
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext
 	mov	r0, r0
 #ifndef CONFIG_CPU_DCACHE_WRITETHROUGH
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
diff --git a/arch/arm/mm/proc-arm940.S b/arch/arm/mm/proc-arm940.S
index 551244d5ca1..f595117caf5 100644
--- a/arch/arm/mm/proc-arm940.S
+++ b/arch/arm/mm/proc-arm940.S
@@ -11,7 +11,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
diff --git a/arch/arm/mm/proc-arm946.S b/arch/arm/mm/proc-arm946.S
index 6168c6160de..e03f6ff1fb2 100644
--- a/arch/arm/mm/proc-arm946.S
+++ b/arch/arm/mm/proc-arm946.S
@@ -13,7 +13,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
diff --git a/arch/arm/mm/proc-arm9tdmi.S b/arch/arm/mm/proc-arm9tdmi.S
index c85c1f50e39..be6c11d2b3f 100644
--- a/arch/arm/mm/proc-arm9tdmi.S
+++ b/arch/arm/mm/proc-arm9tdmi.S
@@ -12,7 +12,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/ptrace.h>
diff --git a/arch/arm/mm/proc-feroceon.S b/arch/arm/mm/proc-feroceon.S
index f2e5884c513..2b8bb383755 100644
--- a/arch/arm/mm/proc-feroceon.S
+++ b/arch/arm/mm/proc-feroceon.S
@@ -22,7 +22,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 #include <asm/page.h>
@@ -446,24 +446,7 @@ ENTRY(cpu_feroceon_switch_mm)
 	.align	5
 ENTRY(cpu_feroceon_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext wc_disable=0
 	mov	r0, r0
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
 #if defined(CONFIG_CACHE_FEROCEON_L2) && !defined(CONFIG_L2_CACHE_WRITETHROUGH)
diff --git a/arch/arm/mm/proc-macros.S b/arch/arm/mm/proc-macros.S
index b13150052a7..54b1f721dec 100644
--- a/arch/arm/mm/proc-macros.S
+++ b/arch/arm/mm/proc-macros.S
@@ -71,3 +71,173 @@
 	mov	\reg, #16			@ size offset
 	mov	\reg, \reg, lsl \tmp		@ actual cache line size
 	.endm
+
+
+/*
+ * Sanity check the PTE configuration for the code below - which makes
+ * certain assumptions about how these bits are layed out.
+ */
+#if L_PTE_SHARED != PTE_EXT_SHARED
+#error PTE shared bit mismatch
+#endif
+#if L_PTE_BUFFERABLE != PTE_BUFFERABLE
+#error PTE bufferable bit mismatch
+#endif
+#if L_PTE_CACHEABLE != PTE_CACHEABLE
+#error PTE cacheable bit mismatch
+#endif
+#if (L_PTE_EXEC+L_PTE_USER+L_PTE_WRITE+L_PTE_DIRTY+L_PTE_YOUNG+\
+     L_PTE_FILE+L_PTE_PRESENT) > L_PTE_SHARED
+#error Invalid Linux PTE bit settings
+#endif
+
+/*
+ * The ARMv6 and ARMv7 set_pte_ext translation function.
+ *
+ * Permission translation:
+ *  YUWD  APX AP1 AP0	SVC	User
+ *  0xxx   0   0   0	no acc	no acc
+ *  100x   1   0   1	r/o	no acc
+ *  10x0   1   0   1	r/o	no acc
+ *  1011   0   0   1	r/w	no acc
+ *  110x   0   1   0	r/w	r/o
+ *  11x0   0   1   0	r/w	r/o
+ *  1111   0   1   1	r/w	r/w
+ */
+	.macro	armv6_mt_table pfx
+\pfx\()_mt_table:
+	.long	0x00						@ L_PTE_MT_UNCACHED
+	.long	PTE_EXT_TEX(1)					@ L_PTE_MT_BUFFERABLE
+	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
+	.long	PTE_BUFFERABLE					@ L_PTE_MT_DEV_SHARED
+	.long	0x00						@ unused
+	.long	0x00						@ L_PTE_MT_MINICACHE (not present)
+	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC
+	.long	0x00						@ unused
+	.long	PTE_EXT_TEX(1)					@ L_PTE_MT_DEV_WC
+	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_DEV_CACHED
+	.long	PTE_EXT_TEX(2)					@ L_PTE_MT_DEV_NONSHARED
+	.long	0x00						@ unused
+	.long	0x00						@ unused
+	.long	0x00						@ unused
+	.endm
+
+	.macro	armv6_set_pte_ext pfx
+	str	r1, [r0], #-2048		@ linux version
+
+	bic	r3, r1, #0x000003fc
+	bic	r3, r3, #PTE_TYPE_MASK
+	orr	r3, r3, r2
+	orr	r3, r3, #PTE_EXT_AP0 | 2
+
+	adr	ip, \pfx\()_mt_table
+	and	r2, r1, #L_PTE_MT_MASK
+	ldr	r2, [ip, r2]
+
+	tst	r1, #L_PTE_WRITE
+	tstne	r1, #L_PTE_DIRTY
+	orreq	r3, r3, #PTE_EXT_APX
+
+	tst	r1, #L_PTE_USER
+	orrne	r3, r3, #PTE_EXT_AP1
+	tstne	r3, #PTE_EXT_APX
+	bicne	r3, r3, #PTE_EXT_APX | PTE_EXT_AP0
+
+	tst	r1, #L_PTE_EXEC
+	orreq	r3, r3, #PTE_EXT_XN
+
+	orr	r3, r3, r2
+
+	tst	r1, #L_PTE_YOUNG
+	tstne	r1, #L_PTE_PRESENT
+	moveq	r3, #0
+
+	str	r3, [r0]
+	mcr	p15, 0, r0, c7, c10, 1		@ flush_pte
+	.endm
+
+
+/*
+ * The ARMv3, ARMv4 and ARMv5 set_pte_ext translation function,
+ * covering most CPUs except Xscale and Xscale 3.
+ *
+ * Permission translation:
+ *  YUWD   AP	SVC	User
+ *  0xxx  0x00	no acc	no acc
+ *  100x  0x00	r/o	no acc
+ *  10x0  0x00	r/o	no acc
+ *  1011  0x55	r/w	no acc
+ *  110x  0xaa	r/w	r/o
+ *  11x0  0xaa	r/w	r/o
+ *  1111  0xff	r/w	r/w
+ */
+	.macro	armv3_set_pte_ext wc_disable=1
+	str	r1, [r0], #-2048		@ linux version
+
+	eor	r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
+
+	bic	r2, r1, #PTE_SMALL_AP_MASK	@ keep C, B bits
+	bic	r2, r2, #PTE_TYPE_MASK
+	orr	r2, r2, #PTE_TYPE_SMALL
+
+	tst	r3, #L_PTE_USER			@ user?
+	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
+
+	tst	r3, #L_PTE_WRITE | L_PTE_DIRTY	@ write and dirty?
+	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
+
+	tst	r3, #L_PTE_PRESENT | L_PTE_YOUNG	@ present and young?
+	movne	r2, #0
+
+	.if	\wc_disable
+#ifdef CONFIG_CPU_DCACHE_WRITETHROUGH
+	tst	r2, #PTE_CACHEABLE
+	bicne	r2, r2, #PTE_BUFFERABLE
+#endif
+	.endif
+	str	r2, [r0]			@ hardware version
+	.endm
+
+
+/*
+ * Xscale set_pte_ext translation, split into two halves to cope
+ * with work-arounds.  r3 must be preserved by code between these
+ * two macros.
+ *
+ * Permission translation:
+ *  YUWD  AP	SVC	User
+ *  0xxx  00	no acc	no acc
+ *  100x  00	r/o	no acc
+ *  10x0  00	r/o	no acc
+ *  1011  01	r/w	no acc
+ *  110x  10	r/w	r/o
+ *  11x0  10	r/w	r/o
+ *  1111  11	r/w	r/w
+ */
+	.macro	xscale_set_pte_ext_prologue
+	str	r1, [r0], #-2048		@ linux version
+
+	eor	r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
+
+	bic	r2, r1, #PTE_SMALL_AP_MASK	@ keep C, B bits
+	orr	r2, r2, #PTE_TYPE_EXT		@ extended page
+
+	tst	r3, #L_PTE_USER			@ user?
+	orrne	r2, r2, #PTE_EXT_AP_URO_SRW	@ yes -> user r/o, system r/w
+
+	tst	r3, #L_PTE_WRITE | L_PTE_DIRTY	@ write and dirty?
+	orreq	r2, r2, #PTE_EXT_AP_UNO_SRW	@ yes -> user n/a, system r/w
+						@ combined with user -> user r/w
+	.endm
+
+	.macro	xscale_set_pte_ext_epilogue
+	tst	r3, #L_PTE_PRESENT | L_PTE_YOUNG	@ present and young?
+	movne	r2, #0				@ no -> fault
+
+	str	r2, [r0]			@ hardware version
+	mov	ip, #0
+	mcr	p15, 0, r0, c7, c10, 1		@ clean L1 D line
+	mcr	p15, 0, ip, c7, c10, 4		@ data write barrier
+	.endm
diff --git a/arch/arm/mm/proc-sa110.S b/arch/arm/mm/proc-sa110.S
index bbe10576c86..90a7e5279f2 100644
--- a/arch/arm/mm/proc-sa110.S
+++ b/arch/arm/mm/proc-sa110.S
@@ -17,7 +17,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <mach/hardware.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
@@ -153,24 +153,7 @@ ENTRY(cpu_sa110_switch_mm)
 	.align	5
 ENTRY(cpu_sa110_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext wc_disable=0
 	mov	r0, r0
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
 	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
diff --git a/arch/arm/mm/proc-sa1100.S b/arch/arm/mm/proc-sa1100.S
index 871ba018252..451e2d953e2 100644
--- a/arch/arm/mm/proc-sa1100.S
+++ b/arch/arm/mm/proc-sa1100.S
@@ -22,7 +22,7 @@
 #include <linux/init.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <mach/hardware.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
@@ -166,24 +166,7 @@ ENTRY(cpu_sa1100_switch_mm)
 	.align	5
 ENTRY(cpu_sa1100_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	eor	r1, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	bic	r2, r1, #PTE_SMALL_AP_MASK
-	bic	r2, r2, #PTE_TYPE_MASK
-	orr	r2, r2, #PTE_TYPE_SMALL
-
-	tst	r1, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_SMALL_AP_URO_SRW
-
-	tst	r1, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_SMALL_AP_UNO_SRW
-
-	tst	r1, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0
-
-	str	r2, [r0]			@ hardware version
+	armv3_set_pte_ext wc_disable=0
 	mov	r0, r0
 	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
 	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
diff --git a/arch/arm/mm/proc-v6.S b/arch/arm/mm/proc-v6.S
index 5702ec58b2a..294943b8597 100644
--- a/arch/arm/mm/proc-v6.S
+++ b/arch/arm/mm/proc-v6.S
@@ -13,7 +13,7 @@
 #include <linux/linkage.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 
@@ -114,46 +114,12 @@ ENTRY(cpu_v6_switch_mm)
  *		  (hardware version is stored at -1024 bytes)
  *	- pte   - PTE value to store
  *	- ext	- value for extended PTE bits
- *
- *	Permissions:
- *	  YUWD  APX AP1 AP0	SVC	User
- *	  0xxx   0   0   0	no acc	no acc
- *	  100x   1   0   1	r/o	no acc
- *	  10x0   1   0   1	r/o	no acc
- *	  1011   0   0   1	r/w	no acc
- *	  110x   0   1   0	r/w	r/o
- *	  11x0   0   1   0	r/w	r/o
- *	  1111   0   1   1	r/w	r/w
  */
+	armv6_mt_table cpu_v6
+
 ENTRY(cpu_v6_set_pte_ext)
 #ifdef CONFIG_MMU
-	str	r1, [r0], #-2048		@ linux version
-
-	bic	r3, r1, #0x000003f0
-	bic	r3, r3, #0x00000003
-	orr	r3, r3, r2
-	orr	r3, r3, #PTE_EXT_AP0 | 2
-
-	tst	r1, #L_PTE_WRITE
-	tstne	r1, #L_PTE_DIRTY
-	orreq	r3, r3, #PTE_EXT_APX
-
-	tst	r1, #L_PTE_USER
-	orrne	r3, r3, #PTE_EXT_AP1
-	tstne	r3, #PTE_EXT_APX
-	bicne	r3, r3, #PTE_EXT_APX | PTE_EXT_AP0
-
-	tst	r1, #L_PTE_YOUNG
-	biceq	r3, r3, #PTE_EXT_APX | PTE_EXT_AP_MASK
-
-	tst	r1, #L_PTE_EXEC
-	orreq	r3, r3, #PTE_EXT_XN
-
-	tst	r1, #L_PTE_PRESENT
-	moveq	r3, #0
-
-	str	r3, [r0]
-	mcr	p15, 0, r0, c7, c10, 1 @ flush_pte
+	armv6_set_pte_ext cpu_v6
 #endif
 	mov	pc, lr
 
diff --git a/arch/arm/mm/proc-v7.S b/arch/arm/mm/proc-v7.S
index a67e26f3dce..34e42404192 100644
--- a/arch/arm/mm/proc-v7.S
+++ b/arch/arm/mm/proc-v7.S
@@ -12,7 +12,7 @@
 #include <linux/linkage.h>
 #include <asm/assembler.h>
 #include <asm/asm-offsets.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/pgtable.h>
 
@@ -105,26 +105,19 @@ ENDPROC(cpu_v7_switch_mm)
  *		  (hardware version is stored at -1024 bytes)
  *	- pte   - PTE value to store
  *	- ext	- value for extended PTE bits
- *
- *	Permissions:
- *	  YUWD  APX AP1 AP0	SVC	User
- *	  0xxx   0   0   0	no acc	no acc
- *	  100x   1   0   1	r/o	no acc
- *	  10x0   1   0   1	r/o	no acc
- *	  1011   0   0   1	r/w	no acc
- *	  110x   0   1   0	r/w	r/o
- *	  11x0   0   1   0	r/w	r/o
- *	  1111   0   1   1	r/w	r/w
  */
 ENTRY(cpu_v7_set_pte_ext)
 #ifdef CONFIG_MMU
 	str	r1, [r0], #-2048		@ linux version
 
 	bic	r3, r1, #0x000003f0
-	bic	r3, r3, #0x00000003
+	bic	r3, r3, #PTE_TYPE_MASK
 	orr	r3, r3, r2
 	orr	r3, r3, #PTE_EXT_AP0 | 2
 
+	tst	r2, #1 << 4
+	orrne	r3, r3, #PTE_EXT_TEX(1)
+
 	tst	r1, #L_PTE_WRITE
 	tstne	r1, #L_PTE_DIRTY
 	orreq	r3, r3, #PTE_EXT_APX
@@ -134,13 +127,11 @@ ENTRY(cpu_v7_set_pte_ext)
 	tstne	r3, #PTE_EXT_APX
 	bicne	r3, r3, #PTE_EXT_APX | PTE_EXT_AP0
 
-	tst	r1, #L_PTE_YOUNG
-	biceq	r3, r3, #PTE_EXT_APX | PTE_EXT_AP_MASK
-
 	tst	r1, #L_PTE_EXEC
 	orreq	r3, r3, #PTE_EXT_XN
 
-	tst	r1, #L_PTE_PRESENT
+	tst	r1, #L_PTE_YOUNG
+	tstne	r1, #L_PTE_PRESENT
 	moveq	r3, #0
 
 	str	r3, [r0]
@@ -189,6 +180,10 @@ __v7_setup:
 	mov	r10, #0x1f			@ domains 0, 1 = manager
 	mcr	p15, 0, r10, c3, c0, 0		@ load domain access register
 #endif
+	ldr	r5, =0x40e040e0
+	ldr	r6, =0xff0aa1a8
+	mcr	p15, 0, r5, c10, c2, 0		@ write PRRR
+	mcr	p15, 0, r6, c10, c2, 1		@ write NMRR
 	adr	r5, v7_crval
 	ldmia	r5, {r5, r6}
    	mrc	p15, 0, r0, c1, c0, 0		@ read control register
@@ -205,7 +200,7 @@ ENDPROC(__v7_setup)
 	 */
 	.type	v7_crval, #object
 v7_crval:
-	crval	clear=0x0120c302, mmuset=0x00c0387d, ucset=0x00c0187c
+	crval	clear=0x0120c302, mmuset=0x10c0387d, ucset=0x00c0187c
 
 __v7_setup_stack:
 	.space	4 * 11				@ 11 registers
diff --git a/arch/arm/mm/proc-xsc3.S b/arch/arm/mm/proc-xsc3.S
index 7bd9e7197f6..04dc8b65401 100644
--- a/arch/arm/mm/proc-xsc3.S
+++ b/arch/arm/mm/proc-xsc3.S
@@ -27,7 +27,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <mach/hardware.h>
 #include <asm/pgtable.h>
 #include <asm/pgtable-hwdef.h>
@@ -345,38 +345,38 @@ ENTRY(cpu_xsc3_switch_mm)
  * cpu_xsc3_set_pte_ext(ptep, pte, ext)
  *
  * Set a PTE and flush it out
- *
  */
+cpu_xsc3_mt_table:
+	.long	0x00						@ L_PTE_MT_UNCACHED
+	.long	PTE_EXT_TEX(1)					@ L_PTE_MT_BUFFERABLE
+	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
+	.long	PTE_EXT_TEX(1) | PTE_BUFFERABLE			@ L_PTE_MT_DEV_SHARED
+	.long	0x00						@ unused
+	.long	0x00						@ L_PTE_MT_MINICACHE (not present)
+	.long	PTE_EXT_TEX(5) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC (not present?)
+	.long	0x00						@ unused
+	.long	PTE_EXT_TEX(1)					@ L_PTE_MT_DEV_WC
+	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_DEV_CACHED
+	.long	PTE_EXT_TEX(2)					@ L_PTE_MT_DEV_NONSHARED
+	.long	0x00						@ unused
+	.long	0x00						@ unused
+	.long	0x00						@ unused
+
 	.align	5
 ENTRY(cpu_xsc3_set_pte_ext)
-	str	r1, [r0], #-2048		@ linux version
+	xscale_set_pte_ext_prologue
 
-	bic	r2, r1, #0xff0			@ keep C, B bits
-	orr	r2, r2, #PTE_TYPE_EXT		@ extended page
 	tst	r1, #L_PTE_SHARED		@ shared?
-	orrne	r2, r2, #0x200
-
-	eor	r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	tst	r3, #L_PTE_USER			@ user?
-	orrne	r2, r2, #PTE_EXT_AP_URO_SRW	@ yes -> user r/o, system r/w
-
-	tst	r3, #L_PTE_WRITE | L_PTE_DIRTY	@ write and dirty?
-	orreq	r2, r2, #PTE_EXT_AP_UNO_SRW	@ yes -> user n/a, system r/w
-						@ combined with user -> user r/w
-
-	@ If it's cacheable, it needs to be in L2 also.
-	eor	ip, r1, #L_PTE_CACHEABLE
-	tst	ip, #L_PTE_CACHEABLE
-	orreq	r2, r2, #PTE_EXT_TEX(0x5)
-
-	tst	r3, #L_PTE_PRESENT | L_PTE_YOUNG	@ present and young?
-	movne	r2, #0				@ no -> fault
-
-	str	r2, [r0]			@ hardware version
-	mov	ip, #0
-	mcr	p15, 0, r0, c7, c10, 1		@ clean L1 D line
-	mcr	p15, 0, ip, c7, c10, 4		@ data write barrier
+	and	r1, r1, #L_PTE_MT_MASK
+	adr	ip, cpu_xsc3_mt_table
+	ldr	ip, [ip, r1]
+	orrne	r2, r2, #PTE_EXT_COHERENT	@ interlock: mask in coherent bit
+	bic	r2, r2, #0x0c			@ clear old C,B bits
+	orr	r2, r2, ip
+
+	xscale_set_pte_ext_epilogue
 	mov	pc, lr
 
 	.ltorg
diff --git a/arch/arm/mm/proc-xscale.S b/arch/arm/mm/proc-xscale.S
index 2dd85273976..0cce37b9393 100644
--- a/arch/arm/mm/proc-xscale.S
+++ b/arch/arm/mm/proc-xscale.S
@@ -23,7 +23,7 @@
 #include <linux/linkage.h>
 #include <linux/init.h>
 #include <asm/assembler.h>
-#include <asm/elf.h>
+#include <asm/hwcap.h>
 #include <asm/pgtable.h>
 #include <asm/pgtable-hwdef.h>
 #include <asm/page.h>
@@ -406,8 +406,6 @@ ENTRY(cpu_xscale_dcache_clean_area)
 
 /* =============================== PageTable ============================== */
 
-#define PTE_CACHE_WRITE_ALLOCATE 0
-
 /*
  * cpu_xscale_switch_mm(pgd)
  *
@@ -431,56 +429,42 @@ ENTRY(cpu_xscale_switch_mm)
  *
  * Errata 40: must set memory to write-through for user read-only pages.
  */
+cpu_xscale_mt_table:
+	.long	0x00						@ L_PTE_MT_UNCACHED
+	.long	PTE_BUFFERABLE					@ L_PTE_MT_BUFFERABLE
+	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
+	.long	PTE_EXT_TEX(1) | PTE_BUFFERABLE			@ L_PTE_MT_DEV_SHARED
+	.long	0x00						@ unused
+	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE			@ L_PTE_MT_MINICACHE
+	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC
+	.long	0x00						@ unused
+	.long	PTE_BUFFERABLE					@ L_PTE_MT_DEV_WC
+	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_DEV_CACHED
+	.long	0x00						@ L_PTE_MT_DEV_NONSHARED
+	.long	0x00						@ unused
+	.long	0x00						@ unused
+	.long	0x00						@ unused
+
 	.align	5
 ENTRY(cpu_xscale_set_pte_ext)
-	str	r1, [r0], #-2048		@ linux version
-
-	bic	r2, r1, #0xff0
-	orr	r2, r2, #PTE_TYPE_EXT		@ extended page
-
-	eor	r3, r1, #L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_WRITE | L_PTE_DIRTY
-
-	tst	r3, #L_PTE_USER			@ User?
-	orrne	r2, r2, #PTE_EXT_AP_URO_SRW	@ yes -> user r/o, system r/w
-
-	tst	r3, #L_PTE_WRITE | L_PTE_DIRTY	@ Write and Dirty?
-	orreq	r2, r2, #PTE_EXT_AP_UNO_SRW	@ yes -> user n/a, system r/w
-						@ combined with user -> user r/w
-
-	@
-	@ Handle the X bit.  We want to set this bit for the minicache
-	@ (U = E = B = W = 0, C = 1) or when write allocate is enabled,
-	@ and we have a writeable, cacheable region.  If we ignore the
-	@ U and E bits, we can allow user space to use the minicache as
-	@ well.
-	@
-	@  X = (C & ~W & ~B) | (C & W & B & write_allocate)
-	@
-	eor	ip, r1, #L_PTE_CACHEABLE
-	tst	ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
-#if PTE_CACHE_WRITE_ALLOCATE
-	eorne	ip, r1, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
-	tstne	ip, #L_PTE_CACHEABLE | L_PTE_WRITE | L_PTE_BUFFERABLE
-#endif
-	orreq	r2, r2, #PTE_EXT_TEX(1)
+	xscale_set_pte_ext_prologue
 
 	@
-	@ Erratum 40: The B bit must be cleared for a user read-only
-	@ cacheable page.
-	@
-	@  B = B & ~(U & C & ~W)
+	@ Erratum 40: must set memory to write-through for user read-only pages
 	@
-	and	ip, r1, #L_PTE_USER | L_PTE_WRITE | L_PTE_CACHEABLE
-	teq	ip, #L_PTE_USER | L_PTE_CACHEABLE
-	biceq	r2, r2, #PTE_BUFFERABLE
+	and	ip, r1, #(L_PTE_MT_MASK | L_PTE_USER | L_PTE_WRITE) & ~(4 << 2)
+	teq	ip, #L_PTE_MT_WRITEBACK | L_PTE_USER
 
-	tst	r3, #L_PTE_PRESENT | L_PTE_YOUNG	@ Present and Young?
-	movne	r2, #0				@ no -> fault
+	moveq	r1, #L_PTE_MT_WRITETHROUGH
+	and	r1, r1, #L_PTE_MT_MASK
+	adr	ip, cpu_xscale_mt_table
+	ldr	ip, [ip, r1]
+	bic	r2, r2, #0x0c
+	orr	r2, r2, ip
 
-	str	r2, [r0]			@ hardware version
-	mov	ip, #0
-	mcr	p15, 0, r0, c7, c10, 1		@ Clean D cache line
-	mcr	p15, 0, ip, c7, c10, 4		@ Drain Write (& Fill) Buffer
+	xscale_set_pte_ext_epilogue
 	mov	pc, lr