diff --git a/arch/arm/mm/mm-armv.c b/arch/arm/mm/mm-armv.c
index fb5b40289de2..9e50127be635 100644
--- a/arch/arm/mm/mm-armv.c
+++ b/arch/arm/mm/mm-armv.c
@@ -354,7 +354,7 @@ void __init build_mem_type_table(void)
 {
 	struct cachepolicy *cp;
 	unsigned int cr = get_cr();
-	unsigned int user_pgprot;
+	unsigned int user_pgprot, kern_pgprot;
 	int cpu_arch = cpu_architecture();
 	int i;
 
@@ -381,7 +381,7 @@ void __init build_mem_type_table(void)
 	}
 
 	cp = &cache_policies[cachepolicy];
-	user_pgprot = cp->pte;
+	kern_pgprot = user_pgprot = cp->pte;
 
 	/*
 	 * ARMv6 and above have extended page tables.
@@ -393,6 +393,7 @@ void __init build_mem_type_table(void)
 		 */
 		mem_types[MT_MEMORY].prot_sect &= ~PMD_BIT4;
 		mem_types[MT_ROM].prot_sect &= ~PMD_BIT4;
+
 		/*
 		 * Mark cache clean areas and XIP ROM read only
 		 * from SVC mode and no access from userspace.
@@ -412,32 +413,47 @@ void __init build_mem_type_table(void)
 		 * (iow, non-global)
 		 */
 		user_pgprot |= L_PTE_ASID;
+
+#ifdef CONFIG_SMP
+		/*
+		 * Mark memory with the "shared" attribute for SMP systems
+		 */
+		user_pgprot |= L_PTE_SHARED;
+		kern_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);
+	}
+
+	mem_types[MT_LOW_VECTORS].prot_pte |= kern_pgprot;
+	mem_types[MT_HIGH_VECTORS].prot_pte |= kern_pgprot;
+
 	if (cpu_arch >= CPU_ARCH_ARMv5) {
-		mem_types[MT_LOW_VECTORS].prot_pte |= cp->pte & PTE_CACHEABLE;
-		mem_types[MT_HIGH_VECTORS].prot_pte |= cp->pte & PTE_CACHEABLE;
+#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 {
-		mem_types[MT_LOW_VECTORS].prot_pte |= cp->pte;
-		mem_types[MT_HIGH_VECTORS].prot_pte |= cp->pte;
 		mem_types[MT_MINICLEAN].prot_sect &= ~PMD_SECT_TEX(1);
 	}
 
+	pgprot_kernel = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG |
+				 L_PTE_DIRTY | L_PTE_WRITE |
+				 L_PTE_EXEC | kern_pgprot);
+
 	mem_types[MT_LOW_VECTORS].prot_l1 |= ecc_mask;
 	mem_types[MT_HIGH_VECTORS].prot_l1 |= ecc_mask;
 	mem_types[MT_MEMORY].prot_sect |= ecc_mask | cp->pmd;
 	mem_types[MT_ROM].prot_sect |= cp->pmd;
 
-	for (i = 0; i < 16; i++) {
-		unsigned long v = pgprot_val(protection_map[i]);
-		v = (v & ~(PTE_BUFFERABLE|PTE_CACHEABLE)) | user_pgprot;
-		protection_map[i] = __pgprot(v);
-	}
-
-	pgprot_kernel = __pgprot(L_PTE_PRESENT | L_PTE_YOUNG |
-				 L_PTE_DIRTY | L_PTE_WRITE |
-				 L_PTE_EXEC | cp->pte);
-
 	switch (cp->pmd) {
 	case PMD_SECT_WT:
 		mem_types[MT_CACHECLEAN].prot_sect |= PMD_SECT_WT;
diff --git a/arch/arm/mm/proc-v6.S b/arch/arm/mm/proc-v6.S
index 9bb5fff406fb..a39d8fa2ede5 100644
--- a/arch/arm/mm/proc-v6.S
+++ b/arch/arm/mm/proc-v6.S
@@ -112,6 +112,9 @@ ENTRY(cpu_v6_dcache_clean_area)
 ENTRY(cpu_v6_switch_mm)
 	mov	r2, #0
 	ldr	r1, [r1, #MM_CONTEXT_ID]	@ get mm->context.id
+#ifdef CONFIG_SMP
+	orr	r0, r0, #2			@ set shared pgtable
+#endif
 	mcr	p15, 0, r2, c7, c5, 6		@ flush BTAC/BTB
 	mcr	p15, 0, r2, c7, c10, 4		@ drain write buffer
 	mcr	p15, 0, r0, c2, c0, 0		@ set TTB 0
@@ -140,7 +143,7 @@ ENTRY(cpu_v6_switch_mm)
 ENTRY(cpu_v6_set_pte)
 	str	r1, [r0], #-2048		@ linux version
 
-	bic	r2, r1, #0x000007f0
+	bic	r2, r1, #0x000003f0
 	bic	r2, r2, #0x00000003
 	orr	r2, r2, #PTE_EXT_AP0 | 2
 
@@ -198,6 +201,9 @@ __v6_setup:
 	mcr	p15, 0, r0, c7, c10, 4		@ drain write buffer
 	mcr	p15, 0, r0, c8, c7, 0		@ invalidate I + D TLBs
 	mcr	p15, 0, r0, c2, c0, 2		@ TTB control register
+#ifdef CONFIG_SMP
+	orr	r4, r4, #2			@ set shared pgtable
+#endif
 	mcr	p15, 0, r4, c2, c0, 1		@ load TTB1
 #ifdef CONFIG_VFP
 	mrc	p15, 0, r0, c1, c0, 2