Moved, renamed, and deleted files

The original directory structure was scattered and unorganized.
Changes are basically to make it look like kernel structure.

32 files changed, 11299 insertions, 0 deletions

diff --git a/arch/mips/mm/Makefile b/arch/mips/mm/Makefile
new file mode 100644
index 00000000..4aa20280
--- /dev/null
+++ b/arch/mips/mm/Makefile
@@ -0,0 +1,39 @@
+#
+# Makefile for the Linux/MIPS-specific parts of the memory manager.
+#
+
+obj-y				+= cache.o dma-default.o extable.o fault.o \
+				   gup.o init.o mmap.o page.o tlbex.o \
+				   tlbex-fault.o uasm.o
+
+obj-$(CONFIG_32BIT)		+= ioremap.o pgtable-32.o
+obj-$(CONFIG_64BIT)		+= pgtable-64.o
+obj-$(CONFIG_HIGHMEM)		+= highmem.o
+obj-$(CONFIG_HUGETLB_PAGE)	+= hugetlbpage.o
+
+obj-$(CONFIG_CPU_LOONGSON2)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_MIPS32)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_MIPS64)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_NEVADA)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R10000)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R3000)		+= c-r3k.o tlb-r3k.o
+obj-$(CONFIG_CPU_R4300)		+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R4X00)		+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R5000)		+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R5432)		+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R5500)		+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_R8000)		+= c-r4k.o cex-gen.o tlb-r8k.o
+obj-$(CONFIG_CPU_RM7000)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_RM9000)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_SB1)		+= c-r4k.o cerr-sb1.o cex-sb1.o tlb-r4k.o
+obj-$(CONFIG_CPU_TX39XX)	+= c-tx39.o tlb-r3k.o
+obj-$(CONFIG_CPU_TX49XX)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_VR41XX)	+= c-r4k.o cex-gen.o tlb-r4k.o
+obj-$(CONFIG_CPU_CAVIUM_OCTEON)	+= c-octeon.o cex-oct.o tlb-r4k.o
+obj-$(CONFIG_CPU_XLR)		+= c-r4k.o tlb-r4k.o cex-gen.o
+obj-$(CONFIG_CPU_XLP)		+= c-r4k.o tlb-r4k.o cex-gen.o
+
+obj-$(CONFIG_IP22_CPU_SCACHE)	+= sc-ip22.o
+obj-$(CONFIG_R5000_CPU_SCACHE)  += sc-r5k.o
+obj-$(CONFIG_RM7000_CPU_SCACHE)	+= sc-rm7k.o
+obj-$(CONFIG_MIPS_CPU_SCACHE)	+= sc-mips.o
diff --git a/arch/mips/mm/c-octeon.c b/arch/mips/mm/c-octeon.c
new file mode 100644
index 00000000..47037ec5
--- /dev/null
+++ b/arch/mips/mm/c-octeon.c
@@ -0,0 +1,327 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2005-2007 Cavium Networks
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+#include <linux/bitops.h>
+#include <linux/cpu.h>
+#include <linux/io.h>
+
+#include <asm/bcache.h>
+#include <asm/bootinfo.h>
+#include <asm/cacheops.h>
+#include <asm/cpu-features.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/r4kcache.h>
+#include <asm/mmu_context.h>
+#include <asm/war.h>
+
+#include <asm/octeon/octeon.h>
+
+unsigned long long cache_err_dcache[NR_CPUS];
+
+/**
+ * Octeon automatically flushes the dcache on tlb changes, so
+ * from Linux's viewpoint it acts much like a physically
+ * tagged cache. No flushing is needed
+ *
+ */
+static void octeon_flush_data_cache_page(unsigned long addr)
+{
+    /* Nothing to do */
+}
+
+static inline void octeon_local_flush_icache(void)
+{
+	asm volatile ("synci 0($0)");
+}
+
+/*
+ * Flush local I-cache for the specified range.
+ */
+static void local_octeon_flush_icache_range(unsigned long start,
+					    unsigned long end)
+{
+	octeon_local_flush_icache();
+}
+
+/**
+ * Flush caches as necessary for all cores affected by a
+ * vma. If no vma is supplied, all cores are flushed.
+ *
+ * @vma:    VMA to flush or NULL to flush all icaches.
+ */
+static void octeon_flush_icache_all_cores(struct vm_area_struct *vma)
+{
+	extern void octeon_send_ipi_single(int cpu, unsigned int action);
+#ifdef CONFIG_SMP
+	int cpu;
+	cpumask_t mask;
+#endif
+
+	mb();
+	octeon_local_flush_icache();
+#ifdef CONFIG_SMP
+	preempt_disable();
+	cpu = smp_processor_id();
+
+	/*
+	 * If we have a vma structure, we only need to worry about
+	 * cores it has been used on
+	 */
+	if (vma)
+		mask = *mm_cpumask(vma->vm_mm);
+	else
+		mask = *cpu_online_mask;
+	cpumask_clear_cpu(cpu, &mask);
+	for_each_cpu(cpu, &mask)
+		octeon_send_ipi_single(cpu, SMP_ICACHE_FLUSH);
+
+	preempt_enable();
+#endif
+}
+
+
+/**
+ * Called to flush the icache on all cores
+ */
+static void octeon_flush_icache_all(void)
+{
+	octeon_flush_icache_all_cores(NULL);
+}
+
+
+/**
+ * Called to flush all memory associated with a memory
+ * context.
+ *
+ * @mm:     Memory context to flush
+ */
+static void octeon_flush_cache_mm(struct mm_struct *mm)
+{
+	/*
+	 * According to the R4K version of this file, CPUs without
+	 * dcache aliases don't need to do anything here
+	 */
+}
+
+
+/**
+ * Flush a range of kernel addresses out of the icache
+ *
+ */
+static void octeon_flush_icache_range(unsigned long start, unsigned long end)
+{
+	octeon_flush_icache_all_cores(NULL);
+}
+
+
+/**
+ * Flush the icache for a trampoline. These are used for interrupt
+ * and exception hooking.
+ *
+ * @addr:   Address to flush
+ */
+static void octeon_flush_cache_sigtramp(unsigned long addr)
+{
+	struct vm_area_struct *vma;
+
+	vma = find_vma(current->mm, addr);
+	octeon_flush_icache_all_cores(vma);
+}
+
+
+/**
+ * Flush a range out of a vma
+ *
+ * @vma:    VMA to flush
+ * @start:
+ * @end:
+ */
+static void octeon_flush_cache_range(struct vm_area_struct *vma,
+				     unsigned long start, unsigned long end)
+{
+	if (vma->vm_flags & VM_EXEC)
+		octeon_flush_icache_all_cores(vma);
+}
+
+
+/**
+ * Flush a specific page of a vma
+ *
+ * @vma:    VMA to flush page for
+ * @page:   Page to flush
+ * @pfn:
+ */
+static void octeon_flush_cache_page(struct vm_area_struct *vma,
+				    unsigned long page, unsigned long pfn)
+{
+	if (vma->vm_flags & VM_EXEC)
+		octeon_flush_icache_all_cores(vma);
+}
+
+static void octeon_flush_kernel_vmap_range(unsigned long vaddr, int size)
+{
+	BUG();
+}
+
+/**
+ * Probe Octeon's caches
+ *
+ */
+static void __cpuinit probe_octeon(void)
+{
+	unsigned long icache_size;
+	unsigned long dcache_size;
+	unsigned int config1;
+	struct cpuinfo_mips *c = &current_cpu_data;
+
+	config1 = read_c0_config1();
+	switch (c->cputype) {
+	case CPU_CAVIUM_OCTEON:
+	case CPU_CAVIUM_OCTEON_PLUS:
+		c->icache.linesz = 2 << ((config1 >> 19) & 7);
+		c->icache.sets = 64 << ((config1 >> 22) & 7);
+		c->icache.ways = 1 + ((config1 >> 16) & 7);
+		c->icache.flags |= MIPS_CACHE_VTAG;
+		icache_size =
+			c->icache.sets * c->icache.ways * c->icache.linesz;
+		c->icache.waybit = ffs(icache_size / c->icache.ways) - 1;
+		c->dcache.linesz = 128;
+		if (c->cputype == CPU_CAVIUM_OCTEON_PLUS)
+			c->dcache.sets = 2; /* CN5XXX has two Dcache sets */
+		else
+			c->dcache.sets = 1; /* CN3XXX has one Dcache set */
+		c->dcache.ways = 64;
+		dcache_size =
+			c->dcache.sets * c->dcache.ways * c->dcache.linesz;
+		c->dcache.waybit = ffs(dcache_size / c->dcache.ways) - 1;
+		c->options |= MIPS_CPU_PREFETCH;
+		break;
+
+	case CPU_CAVIUM_OCTEON2:
+		c->icache.linesz = 2 << ((config1 >> 19) & 7);
+		c->icache.sets = 8;
+		c->icache.ways = 37;
+		c->icache.flags |= MIPS_CACHE_VTAG;
+		icache_size = c->icache.sets * c->icache.ways * c->icache.linesz;
+
+		c->dcache.linesz = 128;
+		c->dcache.ways = 32;
+		c->dcache.sets = 8;
+		dcache_size = c->dcache.sets * c->dcache.ways * c->dcache.linesz;
+		c->options |= MIPS_CPU_PREFETCH;
+		break;
+
+	default:
+		panic("Unsupported Cavium Networks CPU type");
+		break;
+	}
+
+	/* compute a couple of other cache variables */
+	c->icache.waysize = icache_size / c->icache.ways;
+	c->dcache.waysize = dcache_size / c->dcache.ways;
+
+	c->icache.sets = icache_size / (c->icache.linesz * c->icache.ways);
+	c->dcache.sets = dcache_size / (c->dcache.linesz * c->dcache.ways);
+
+	if (smp_processor_id() == 0) {
+		pr_notice("Primary instruction cache %ldkB, %s, %d way, "
+			  "%d sets, linesize %d bytes.\n",
+			  icache_size >> 10,
+			  cpu_has_vtag_icache ?
+				"virtually tagged" : "physically tagged",
+			  c->icache.ways, c->icache.sets, c->icache.linesz);
+
+		pr_notice("Primary data cache %ldkB, %d-way, %d sets, "
+			  "linesize %d bytes.\n",
+			  dcache_size >> 10, c->dcache.ways,
+			  c->dcache.sets, c->dcache.linesz);
+	}
+}
+
+
+/**
+ * Setup the Octeon cache flush routines
+ *
+ */
+void __cpuinit octeon_cache_init(void)
+{
+	extern unsigned long ebase;
+	extern char except_vec2_octeon;
+
+	memcpy((void *)(ebase + 0x100), &except_vec2_octeon, 0x80);
+	octeon_flush_cache_sigtramp(ebase + 0x100);
+
+	probe_octeon();
+
+	shm_align_mask = PAGE_SIZE - 1;
+
+	flush_cache_all			= octeon_flush_icache_all;
+	__flush_cache_all		= octeon_flush_icache_all;
+	flush_cache_mm			= octeon_flush_cache_mm;
+	flush_cache_page		= octeon_flush_cache_page;
+	flush_cache_range		= octeon_flush_cache_range;
+	flush_cache_sigtramp		= octeon_flush_cache_sigtramp;
+	flush_icache_all		= octeon_flush_icache_all;
+	flush_data_cache_page		= octeon_flush_data_cache_page;
+	flush_icache_range		= octeon_flush_icache_range;
+	local_flush_icache_range	= local_octeon_flush_icache_range;
+
+	__flush_kernel_vmap_range	= octeon_flush_kernel_vmap_range;
+
+	build_clear_page();
+	build_copy_page();
+}
+
+/**
+ * Handle a cache error exception
+ */
+
+static void  cache_parity_error_octeon(int non_recoverable)
+{
+	unsigned long coreid = cvmx_get_core_num();
+	uint64_t icache_err = read_octeon_c0_icacheerr();
+
+	pr_err("Cache error exception:\n");
+	pr_err("cp0_errorepc == %lx\n", read_c0_errorepc());
+	if (icache_err & 1) {
+		pr_err("CacheErr (Icache) == %llx\n",
+		       (unsigned long long)icache_err);
+		write_octeon_c0_icacheerr(0);
+	}
+	if (cache_err_dcache[coreid] & 1) {
+		pr_err("CacheErr (Dcache) == %llx\n",
+		       (unsigned long long)cache_err_dcache[coreid]);
+		cache_err_dcache[coreid] = 0;
+	}
+
+	if (non_recoverable)
+		panic("Can't handle cache error: nested exception");
+}
+
+/**
+ * Called when the the exception is recoverable
+ */
+
+asmlinkage void cache_parity_error_octeon_recoverable(void)
+{
+	cache_parity_error_octeon(0);
+}
+
+/**
+ * Called when the the exception is not recoverable
+ */
+
+asmlinkage void cache_parity_error_octeon_non_recoverable(void)
+{
+	cache_parity_error_octeon(1);
+}
diff --git a/arch/mips/mm/c-r3k.c b/arch/mips/mm/c-r3k.c
new file mode 100644
index 00000000..031c4c2c
--- /dev/null
+++ b/arch/mips/mm/c-r3k.c
@@ -0,0 +1,347 @@
+/*
+ * r2300.c: R2000 and R3000 specific mmu/cache code.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ *
+ * with a lot of changes to make this thing work for R3000s
+ * Tx39XX R4k style caches added. HK
+ * Copyright (C) 1998, 1999, 2000 Harald Koerfgen
+ * Copyright (C) 1998 Gleb Raiko & Vladimir Roganov
+ * Copyright (C) 2001, 2004, 2007  Maciej W. Rozycki
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/isadep.h>
+#include <asm/io.h>
+#include <asm/bootinfo.h>
+#include <asm/cpu.h>
+
+static unsigned long icache_size, dcache_size;		/* Size in bytes */
+static unsigned long icache_lsize, dcache_lsize;	/* Size in bytes */
+
+unsigned long __cpuinit r3k_cache_size(unsigned long ca_flags)
+{
+	unsigned long flags, status, dummy, size;
+	volatile unsigned long *p;
+
+	p = (volatile unsigned long *) KSEG0;
+
+	flags = read_c0_status();
+
+	/* isolate cache space */
+	write_c0_status((ca_flags|flags)&~ST0_IEC);
+
+	*p = 0xa5a55a5a;
+	dummy = *p;
+	status = read_c0_status();
+
+	if (dummy != 0xa5a55a5a || (status & ST0_CM)) {
+		size = 0;
+	} else {
+		for (size = 128; size <= 0x40000; size <<= 1)
+			*(p + size) = 0;
+		*p = -1;
+		for (size = 128;
+		     (size <= 0x40000) && (*(p + size) == 0);
+		     size <<= 1)
+			;
+		if (size > 0x40000)
+			size = 0;
+	}
+
+	write_c0_status(flags);
+
+	return size * sizeof(*p);
+}
+
+unsigned long __cpuinit r3k_cache_lsize(unsigned long ca_flags)
+{
+	unsigned long flags, status, lsize, i;
+	volatile unsigned long *p;
+
+	p = (volatile unsigned long *) KSEG0;
+
+	flags = read_c0_status();
+
+	/* isolate cache space */
+	write_c0_status((ca_flags|flags)&~ST0_IEC);
+
+	for (i = 0; i < 128; i++)
+		*(p + i) = 0;
+	*(volatile unsigned char *)p = 0;
+	for (lsize = 1; lsize < 128; lsize <<= 1) {
+		*(p + lsize);
+		status = read_c0_status();
+		if (!(status & ST0_CM))
+			break;
+	}
+	for (i = 0; i < 128; i += lsize)
+		*(volatile unsigned char *)(p + i) = 0;
+
+	write_c0_status(flags);
+
+	return lsize * sizeof(*p);
+}
+
+static void __cpuinit r3k_probe_cache(void)
+{
+	dcache_size = r3k_cache_size(ST0_ISC);
+	if (dcache_size)
+		dcache_lsize = r3k_cache_lsize(ST0_ISC);
+
+	icache_size = r3k_cache_size(ST0_ISC|ST0_SWC);
+	if (icache_size)
+		icache_lsize = r3k_cache_lsize(ST0_ISC|ST0_SWC);
+}
+
+static void r3k_flush_icache_range(unsigned long start, unsigned long end)
+{
+	unsigned long size, i, flags;
+	volatile unsigned char *p;
+
+	size = end - start;
+	if (size > icache_size || KSEGX(start) != KSEG0) {
+		start = KSEG0;
+		size = icache_size;
+	}
+	p = (char *)start;
+
+	flags = read_c0_status();
+
+	/* isolate cache space */
+	write_c0_status((ST0_ISC|ST0_SWC|flags)&~ST0_IEC);
+
+	for (i = 0; i < size; i += 0x080) {
+		asm( 	"sb\t$0, 0x000(%0)\n\t"
+			"sb\t$0, 0x004(%0)\n\t"
+			"sb\t$0, 0x008(%0)\n\t"
+			"sb\t$0, 0x00c(%0)\n\t"
+			"sb\t$0, 0x010(%0)\n\t"
+			"sb\t$0, 0x014(%0)\n\t"
+			"sb\t$0, 0x018(%0)\n\t"
+			"sb\t$0, 0x01c(%0)\n\t"
+			"sb\t$0, 0x020(%0)\n\t"
+			"sb\t$0, 0x024(%0)\n\t"
+			"sb\t$0, 0x028(%0)\n\t"
+			"sb\t$0, 0x02c(%0)\n\t"
+			"sb\t$0, 0x030(%0)\n\t"
+			"sb\t$0, 0x034(%0)\n\t"
+			"sb\t$0, 0x038(%0)\n\t"
+			"sb\t$0, 0x03c(%0)\n\t"
+			"sb\t$0, 0x040(%0)\n\t"
+			"sb\t$0, 0x044(%0)\n\t"
+			"sb\t$0, 0x048(%0)\n\t"
+			"sb\t$0, 0x04c(%0)\n\t"
+			"sb\t$0, 0x050(%0)\n\t"
+			"sb\t$0, 0x054(%0)\n\t"
+			"sb\t$0, 0x058(%0)\n\t"
+			"sb\t$0, 0x05c(%0)\n\t"
+			"sb\t$0, 0x060(%0)\n\t"
+			"sb\t$0, 0x064(%0)\n\t"
+			"sb\t$0, 0x068(%0)\n\t"
+			"sb\t$0, 0x06c(%0)\n\t"
+			"sb\t$0, 0x070(%0)\n\t"
+			"sb\t$0, 0x074(%0)\n\t"
+			"sb\t$0, 0x078(%0)\n\t"
+			"sb\t$0, 0x07c(%0)\n\t"
+			: : "r" (p) );
+		p += 0x080;
+	}
+
+	write_c0_status(flags);
+}
+
+static void r3k_flush_dcache_range(unsigned long start, unsigned long end)
+{
+	unsigned long size, i, flags;
+	volatile unsigned char *p;
+
+	size = end - start;
+	if (size > dcache_size || KSEGX(start) != KSEG0) {
+		start = KSEG0;
+		size = dcache_size;
+	}
+	p = (char *)start;
+
+	flags = read_c0_status();
+
+	/* isolate cache space */
+	write_c0_status((ST0_ISC|flags)&~ST0_IEC);
+
+	for (i = 0; i < size; i += 0x080) {
+		asm( 	"sb\t$0, 0x000(%0)\n\t"
+			"sb\t$0, 0x004(%0)\n\t"
+			"sb\t$0, 0x008(%0)\n\t"
+			"sb\t$0, 0x00c(%0)\n\t"
+			"sb\t$0, 0x010(%0)\n\t"
+			"sb\t$0, 0x014(%0)\n\t"
+			"sb\t$0, 0x018(%0)\n\t"
+			"sb\t$0, 0x01c(%0)\n\t"
+			"sb\t$0, 0x020(%0)\n\t"
+			"sb\t$0, 0x024(%0)\n\t"
+			"sb\t$0, 0x028(%0)\n\t"
+			"sb\t$0, 0x02c(%0)\n\t"
+			"sb\t$0, 0x030(%0)\n\t"
+			"sb\t$0, 0x034(%0)\n\t"
+			"sb\t$0, 0x038(%0)\n\t"
+			"sb\t$0, 0x03c(%0)\n\t"
+			"sb\t$0, 0x040(%0)\n\t"
+			"sb\t$0, 0x044(%0)\n\t"
+			"sb\t$0, 0x048(%0)\n\t"
+			"sb\t$0, 0x04c(%0)\n\t"
+			"sb\t$0, 0x050(%0)\n\t"
+			"sb\t$0, 0x054(%0)\n\t"
+			"sb\t$0, 0x058(%0)\n\t"
+			"sb\t$0, 0x05c(%0)\n\t"
+			"sb\t$0, 0x060(%0)\n\t"
+			"sb\t$0, 0x064(%0)\n\t"
+			"sb\t$0, 0x068(%0)\n\t"
+			"sb\t$0, 0x06c(%0)\n\t"
+			"sb\t$0, 0x070(%0)\n\t"
+			"sb\t$0, 0x074(%0)\n\t"
+			"sb\t$0, 0x078(%0)\n\t"
+			"sb\t$0, 0x07c(%0)\n\t"
+			: : "r" (p) );
+		p += 0x080;
+	}
+
+	write_c0_status(flags);
+}
+
+static inline void r3k_flush_cache_all(void)
+{
+}
+
+static inline void r3k___flush_cache_all(void)
+{
+	r3k_flush_dcache_range(KSEG0, KSEG0 + dcache_size);
+	r3k_flush_icache_range(KSEG0, KSEG0 + icache_size);
+}
+
+static void r3k_flush_cache_mm(struct mm_struct *mm)
+{
+}
+
+static void r3k_flush_cache_range(struct vm_area_struct *vma,
+				  unsigned long start, unsigned long end)
+{
+}
+
+static void r3k_flush_cache_page(struct vm_area_struct *vma,
+				 unsigned long addr, unsigned long pfn)
+{
+	unsigned long kaddr = KSEG0ADDR(pfn << PAGE_SHIFT);
+	int exec = vma->vm_flags & VM_EXEC;
+	struct mm_struct *mm = vma->vm_mm;
+	pgd_t *pgdp;
+	pud_t *pudp;
+	pmd_t *pmdp;
+	pte_t *ptep;
+
+	pr_debug("cpage[%08lx,%08lx]\n",
+		 cpu_context(smp_processor_id(), mm), addr);
+
+	/* No ASID => no such page in the cache.  */
+	if (cpu_context(smp_processor_id(), mm) == 0)
+		return;
+
+	pgdp = pgd_offset(mm, addr);
+	pudp = pud_offset(pgdp, addr);
+	pmdp = pmd_offset(pudp, addr);
+	ptep = pte_offset(pmdp, addr);
+
+	/* Invalid => no such page in the cache.  */
+	if (!(pte_val(*ptep) & _PAGE_PRESENT))
+		return;
+
+	r3k_flush_dcache_range(kaddr, kaddr + PAGE_SIZE);
+	if (exec)
+		r3k_flush_icache_range(kaddr, kaddr + PAGE_SIZE);
+}
+
+static void local_r3k_flush_data_cache_page(void *addr)
+{
+}
+
+static void r3k_flush_data_cache_page(unsigned long addr)
+{
+}
+
+static void r3k_flush_cache_sigtramp(unsigned long addr)
+{
+	unsigned long flags;
+
+	pr_debug("csigtramp[%08lx]\n", addr);
+
+	flags = read_c0_status();
+
+	write_c0_status(flags&~ST0_IEC);
+
+	/* Fill the TLB to avoid an exception with caches isolated. */
+	asm( 	"lw\t$0, 0x000(%0)\n\t"
+		"lw\t$0, 0x004(%0)\n\t"
+		: : "r" (addr) );
+
+	write_c0_status((ST0_ISC|ST0_SWC|flags)&~ST0_IEC);
+
+	asm( 	"sb\t$0, 0x000(%0)\n\t"
+		"sb\t$0, 0x004(%0)\n\t"
+		: : "r" (addr) );
+
+	write_c0_status(flags);
+}
+
+static void r3k_flush_kernel_vmap_range(unsigned long vaddr, int size)
+{
+	BUG();
+}
+
+static void r3k_dma_cache_wback_inv(unsigned long start, unsigned long size)
+{
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	iob();
+	r3k_flush_dcache_range(start, start + size);
+}
+
+void __cpuinit r3k_cache_init(void)
+{
+	extern void build_clear_page(void);
+	extern void build_copy_page(void);
+
+	r3k_probe_cache();
+
+	flush_cache_all = r3k_flush_cache_all;
+	__flush_cache_all = r3k___flush_cache_all;
+	flush_cache_mm = r3k_flush_cache_mm;
+	flush_cache_range = r3k_flush_cache_range;
+	flush_cache_page = r3k_flush_cache_page;
+	flush_icache_range = r3k_flush_icache_range;
+	local_flush_icache_range = r3k_flush_icache_range;
+
+	__flush_kernel_vmap_range = r3k_flush_kernel_vmap_range;
+
+	flush_cache_sigtramp = r3k_flush_cache_sigtramp;
+	local_flush_data_cache_page = local_r3k_flush_data_cache_page;
+	flush_data_cache_page = r3k_flush_data_cache_page;
+
+	_dma_cache_wback_inv = r3k_dma_cache_wback_inv;
+	_dma_cache_wback = r3k_dma_cache_wback_inv;
+	_dma_cache_inv = r3k_dma_cache_wback_inv;
+
+	printk("Primary instruction cache %ldkB, linesize %ld bytes.\n",
+		icache_size >> 10, icache_lsize);
+	printk("Primary data cache %ldkB, linesize %ld bytes.\n",
+		dcache_size >> 10, dcache_lsize);
+
+	build_clear_page();
+	build_copy_page();
+}
diff --git a/arch/mips/mm/c-r4k.c b/arch/mips/mm/c-r4k.c
new file mode 100644
index 00000000..bda8eb26
--- /dev/null
+++ b/arch/mips/mm/c-r4k.c
@@ -0,0 +1,1468 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ * Copyright (C) 1997, 1998, 1999, 2000, 2001, 2002 Ralf Baechle (ralf@gnu.org)
+ * Copyright (C) 1999, 2000 Silicon Graphics, Inc.
+ */
+#include <linux/hardirq.h>
+#include <linux/init.h>
+#include <linux/highmem.h>
+#include <linux/kernel.h>
+#include <linux/linkage.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/bitops.h>
+
+#include <asm/bcache.h>
+#include <asm/bootinfo.h>
+#include <asm/cache.h>
+#include <asm/cacheops.h>
+#include <asm/cpu.h>
+#include <asm/cpu-features.h>
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/r4kcache.h>
+#include <asm/sections.h>
+#include <asm/mmu_context.h>
+#include <asm/war.h>
+#include <asm/cacheflush.h> /* for run_uncached() */
+
+
+/*
+ * Special Variant of smp_call_function for use by cache functions:
+ *
+ *  o No return value
+ *  o collapses to normal function call on UP kernels
+ *  o collapses to normal function call on systems with a single shared
+ *    primary cache.
+ *  o doesn't disable interrupts on the local CPU
+ */
+static inline void r4k_on_each_cpu(void (*func) (void *info), void *info)
+{
+	preempt_disable();
+
+#if !defined(CONFIG_MIPS_MT_SMP) && !defined(CONFIG_MIPS_MT_SMTC)
+	smp_call_function(func, info, 1);
+#endif
+	func(info);
+	preempt_enable();
+}
+
+#if defined(CONFIG_MIPS_CMP)
+#define cpu_has_safe_index_cacheops 0
+#else
+#define cpu_has_safe_index_cacheops 1
+#endif
+
+/*
+ * Must die.
+ */
+static unsigned long icache_size __read_mostly;
+static unsigned long dcache_size __read_mostly;
+static unsigned long scache_size __read_mostly;
+
+/*
+ * Dummy cache handling routines for machines without boardcaches
+ */
+static void cache_noop(void) {}
+
+static struct bcache_ops no_sc_ops = {
+	.bc_enable = (void *)cache_noop,
+	.bc_disable = (void *)cache_noop,
+	.bc_wback_inv = (void *)cache_noop,
+	.bc_inv = (void *)cache_noop
+};
+
+struct bcache_ops *bcops = &no_sc_ops;
+
+#define cpu_is_r4600_v1_x()	((read_c0_prid() & 0xfffffff0) == 0x00002010)
+#define cpu_is_r4600_v2_x()	((read_c0_prid() & 0xfffffff0) == 0x00002020)
+
+#define R4600_HIT_CACHEOP_WAR_IMPL					\
+do {									\
+	if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())		\
+		*(volatile unsigned long *)CKSEG1;			\
+	if (R4600_V1_HIT_CACHEOP_WAR)					\
+		__asm__ __volatile__("nop;nop;nop;nop");		\
+} while (0)
+
+static void (*r4k_blast_dcache_page)(unsigned long addr);
+
+static inline void r4k_blast_dcache_page_dc32(unsigned long addr)
+{
+	R4600_HIT_CACHEOP_WAR_IMPL;
+	blast_dcache32_page(addr);
+}
+
+static inline void r4k_blast_dcache_page_dc64(unsigned long addr)
+{
+	R4600_HIT_CACHEOP_WAR_IMPL;
+	blast_dcache64_page(addr);
+}
+
+static void __cpuinit r4k_blast_dcache_page_setup(void)
+{
+	unsigned long  dc_lsize = cpu_dcache_line_size();
+
+	if (dc_lsize == 0)
+		r4k_blast_dcache_page = (void *)cache_noop;
+	else if (dc_lsize == 16)
+		r4k_blast_dcache_page = blast_dcache16_page;
+	else if (dc_lsize == 32)
+		r4k_blast_dcache_page = r4k_blast_dcache_page_dc32;
+	else if (dc_lsize == 64)
+		r4k_blast_dcache_page = r4k_blast_dcache_page_dc64;
+}
+
+static void (* r4k_blast_dcache_page_indexed)(unsigned long addr);
+
+static void __cpuinit r4k_blast_dcache_page_indexed_setup(void)
+{
+	unsigned long dc_lsize = cpu_dcache_line_size();
+
+	if (dc_lsize == 0)
+		r4k_blast_dcache_page_indexed = (void *)cache_noop;
+	else if (dc_lsize == 16)
+		r4k_blast_dcache_page_indexed = blast_dcache16_page_indexed;
+	else if (dc_lsize == 32)
+		r4k_blast_dcache_page_indexed = blast_dcache32_page_indexed;
+	else if (dc_lsize == 64)
+		r4k_blast_dcache_page_indexed = blast_dcache64_page_indexed;
+}
+
+static void (* r4k_blast_dcache)(void);
+
+static void __cpuinit r4k_blast_dcache_setup(void)
+{
+	unsigned long dc_lsize = cpu_dcache_line_size();
+
+	if (dc_lsize == 0)
+		r4k_blast_dcache = (void *)cache_noop;
+	else if (dc_lsize == 16)
+		r4k_blast_dcache = blast_dcache16;
+	else if (dc_lsize == 32)
+		r4k_blast_dcache = blast_dcache32;
+	else if (dc_lsize == 64)
+		r4k_blast_dcache = blast_dcache64;
+}
+
+/* force code alignment (used for TX49XX_ICACHE_INDEX_INV_WAR) */
+#define JUMP_TO_ALIGN(order) \
+	__asm__ __volatile__( \
+		"b\t1f\n\t" \
+		".align\t" #order "\n\t" \
+		"1:\n\t" \
+		)
+#define CACHE32_UNROLL32_ALIGN	JUMP_TO_ALIGN(10) /* 32 * 32 = 1024 */
+#define CACHE32_UNROLL32_ALIGN2	JUMP_TO_ALIGN(11)
+
+static inline void blast_r4600_v1_icache32(void)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	blast_icache32();
+	local_irq_restore(flags);
+}
+
+static inline void tx49_blast_icache32(void)
+{
+	unsigned long start = INDEX_BASE;
+	unsigned long end = start + current_cpu_data.icache.waysize;
+	unsigned long ws_inc = 1UL << current_cpu_data.icache.waybit;
+	unsigned long ws_end = current_cpu_data.icache.ways <<
+	                       current_cpu_data.icache.waybit;
+	unsigned long ws, addr;
+
+	CACHE32_UNROLL32_ALIGN2;
+	/* I'm in even chunk.  blast odd chunks */
+	for (ws = 0; ws < ws_end; ws += ws_inc)
+		for (addr = start + 0x400; addr < end; addr += 0x400 * 2)
+			cache32_unroll32(addr|ws, Index_Invalidate_I);
+	CACHE32_UNROLL32_ALIGN;
+	/* I'm in odd chunk.  blast even chunks */
+	for (ws = 0; ws < ws_end; ws += ws_inc)
+		for (addr = start; addr < end; addr += 0x400 * 2)
+			cache32_unroll32(addr|ws, Index_Invalidate_I);
+}
+
+static inline void blast_icache32_r4600_v1_page_indexed(unsigned long page)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	blast_icache32_page_indexed(page);
+	local_irq_restore(flags);
+}
+
+static inline void tx49_blast_icache32_page_indexed(unsigned long page)
+{
+	unsigned long indexmask = current_cpu_data.icache.waysize - 1;
+	unsigned long start = INDEX_BASE + (page & indexmask);
+	unsigned long end = start + PAGE_SIZE;
+	unsigned long ws_inc = 1UL << current_cpu_data.icache.waybit;
+	unsigned long ws_end = current_cpu_data.icache.ways <<
+	                       current_cpu_data.icache.waybit;
+	unsigned long ws, addr;
+
+	CACHE32_UNROLL32_ALIGN2;
+	/* I'm in even chunk.  blast odd chunks */
+	for (ws = 0; ws < ws_end; ws += ws_inc)
+		for (addr = start + 0x400; addr < end; addr += 0x400 * 2)
+			cache32_unroll32(addr|ws, Index_Invalidate_I);
+	CACHE32_UNROLL32_ALIGN;
+	/* I'm in odd chunk.  blast even chunks */
+	for (ws = 0; ws < ws_end; ws += ws_inc)
+		for (addr = start; addr < end; addr += 0x400 * 2)
+			cache32_unroll32(addr|ws, Index_Invalidate_I);
+}
+
+static void (* r4k_blast_icache_page)(unsigned long addr);
+
+static void __cpuinit r4k_blast_icache_page_setup(void)
+{
+	unsigned long ic_lsize = cpu_icache_line_size();
+
+	if (ic_lsize == 0)
+		r4k_blast_icache_page = (void *)cache_noop;
+	else if (ic_lsize == 16)
+		r4k_blast_icache_page = blast_icache16_page;
+	else if (ic_lsize == 32)
+		r4k_blast_icache_page = blast_icache32_page;
+	else if (ic_lsize == 64)
+		r4k_blast_icache_page = blast_icache64_page;
+}
+
+
+static void (* r4k_blast_icache_page_indexed)(unsigned long addr);
+
+static void __cpuinit r4k_blast_icache_page_indexed_setup(void)
+{
+	unsigned long ic_lsize = cpu_icache_line_size();
+
+	if (ic_lsize == 0)
+		r4k_blast_icache_page_indexed = (void *)cache_noop;
+	else if (ic_lsize == 16)
+		r4k_blast_icache_page_indexed = blast_icache16_page_indexed;
+	else if (ic_lsize == 32) {
+		if (R4600_V1_INDEX_ICACHEOP_WAR && cpu_is_r4600_v1_x())
+			r4k_blast_icache_page_indexed =
+				blast_icache32_r4600_v1_page_indexed;
+		else if (TX49XX_ICACHE_INDEX_INV_WAR)
+			r4k_blast_icache_page_indexed =
+				tx49_blast_icache32_page_indexed;
+		else
+			r4k_blast_icache_page_indexed =
+				blast_icache32_page_indexed;
+	} else if (ic_lsize == 64)
+		r4k_blast_icache_page_indexed = blast_icache64_page_indexed;
+}
+
+static void (* r4k_blast_icache)(void);
+
+static void __cpuinit r4k_blast_icache_setup(void)
+{
+	unsigned long ic_lsize = cpu_icache_line_size();
+
+	if (ic_lsize == 0)
+		r4k_blast_icache = (void *)cache_noop;
+	else if (ic_lsize == 16)
+		r4k_blast_icache = blast_icache16;
+	else if (ic_lsize == 32) {
+		if (R4600_V1_INDEX_ICACHEOP_WAR && cpu_is_r4600_v1_x())
+			r4k_blast_icache = blast_r4600_v1_icache32;
+		else if (TX49XX_ICACHE_INDEX_INV_WAR)
+			r4k_blast_icache = tx49_blast_icache32;
+		else
+			r4k_blast_icache = blast_icache32;
+	} else if (ic_lsize == 64)
+		r4k_blast_icache = blast_icache64;
+}
+
+static void (* r4k_blast_scache_page)(unsigned long addr);
+
+static void __cpuinit r4k_blast_scache_page_setup(void)
+{
+	unsigned long sc_lsize = cpu_scache_line_size();
+
+	if (scache_size == 0)
+		r4k_blast_scache_page = (void *)cache_noop;
+	else if (sc_lsize == 16)
+		r4k_blast_scache_page = blast_scache16_page;
+	else if (sc_lsize == 32)
+		r4k_blast_scache_page = blast_scache32_page;
+	else if (sc_lsize == 64)
+		r4k_blast_scache_page = blast_scache64_page;
+	else if (sc_lsize == 128)
+		r4k_blast_scache_page = blast_scache128_page;
+}
+
+static void (* r4k_blast_scache_page_indexed)(unsigned long addr);
+
+static void __cpuinit r4k_blast_scache_page_indexed_setup(void)
+{
+	unsigned long sc_lsize = cpu_scache_line_size();
+
+	if (scache_size == 0)
+		r4k_blast_scache_page_indexed = (void *)cache_noop;
+	else if (sc_lsize == 16)
+		r4k_blast_scache_page_indexed = blast_scache16_page_indexed;
+	else if (sc_lsize == 32)
+		r4k_blast_scache_page_indexed = blast_scache32_page_indexed;
+	else if (sc_lsize == 64)
+		r4k_blast_scache_page_indexed = blast_scache64_page_indexed;
+	else if (sc_lsize == 128)
+		r4k_blast_scache_page_indexed = blast_scache128_page_indexed;
+}
+
+static void (* r4k_blast_scache)(void);
+
+static void __cpuinit r4k_blast_scache_setup(void)
+{
+	unsigned long sc_lsize = cpu_scache_line_size();
+
+	if (scache_size == 0)
+		r4k_blast_scache = (void *)cache_noop;
+	else if (sc_lsize == 16)
+		r4k_blast_scache = blast_scache16;
+	else if (sc_lsize == 32)
+		r4k_blast_scache = blast_scache32;
+	else if (sc_lsize == 64)
+		r4k_blast_scache = blast_scache64;
+	else if (sc_lsize == 128)
+		r4k_blast_scache = blast_scache128;
+}
+
+static inline void local_r4k___flush_cache_all(void * args)
+{
+#if defined(CONFIG_CPU_LOONGSON2)
+	r4k_blast_scache();
+	return;
+#endif
+	r4k_blast_dcache();
+	r4k_blast_icache();
+
+	switch (current_cpu_type()) {
+	case CPU_R4000SC:
+	case CPU_R4000MC:
+	case CPU_R4400SC:
+	case CPU_R4400MC:
+	case CPU_R10000:
+	case CPU_R12000:
+	case CPU_R14000:
+		r4k_blast_scache();
+	}
+}
+
+static void r4k___flush_cache_all(void)
+{
+	r4k_on_each_cpu(local_r4k___flush_cache_all, NULL);
+}
+
+static inline int has_valid_asid(const struct mm_struct *mm)
+{
+#if defined(CONFIG_MIPS_MT_SMP) || defined(CONFIG_MIPS_MT_SMTC)
+	int i;
+
+	for_each_online_cpu(i)
+		if (cpu_context(i, mm))
+			return 1;
+
+	return 0;
+#else
+	return cpu_context(smp_processor_id(), mm);
+#endif
+}
+
+static void r4k__flush_cache_vmap(void)
+{
+	r4k_blast_dcache();
+}
+
+static void r4k__flush_cache_vunmap(void)
+{
+	r4k_blast_dcache();
+}
+
+static inline void local_r4k_flush_cache_range(void * args)
+{
+	struct vm_area_struct *vma = args;
+	int exec = vma->vm_flags & VM_EXEC;
+
+	if (!(has_valid_asid(vma->vm_mm)))
+		return;
+
+	r4k_blast_dcache();
+	if (exec)
+		r4k_blast_icache();
+}
+
+static void r4k_flush_cache_range(struct vm_area_struct *vma,
+	unsigned long start, unsigned long end)
+{
+	int exec = vma->vm_flags & VM_EXEC;
+
+	if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc))
+		r4k_on_each_cpu(local_r4k_flush_cache_range, vma);
+}
+
+static inline void local_r4k_flush_cache_mm(void * args)
+{
+	struct mm_struct *mm = args;
+
+	if (!has_valid_asid(mm))
+		return;
+
+	/*
+	 * Kludge alert.  For obscure reasons R4000SC and R4400SC go nuts if we
+	 * only flush the primary caches but R10000 and R12000 behave sane ...
+	 * R4000SC and R4400SC indexed S-cache ops also invalidate primary
+	 * caches, so we can bail out early.
+	 */
+	if (current_cpu_type() == CPU_R4000SC ||
+	    current_cpu_type() == CPU_R4000MC ||
+	    current_cpu_type() == CPU_R4400SC ||
+	    current_cpu_type() == CPU_R4400MC) {
+		r4k_blast_scache();
+		return;
+	}
+
+	r4k_blast_dcache();
+}
+
+static void r4k_flush_cache_mm(struct mm_struct *mm)
+{
+	if (!cpu_has_dc_aliases)
+		return;
+
+	r4k_on_each_cpu(local_r4k_flush_cache_mm, mm);
+}
+
+struct flush_cache_page_args {
+	struct vm_area_struct *vma;
+	unsigned long addr;
+	unsigned long pfn;
+};
+
+static inline void local_r4k_flush_cache_page(void *args)
+{
+	struct flush_cache_page_args *fcp_args = args;
+	struct vm_area_struct *vma = fcp_args->vma;
+	unsigned long addr = fcp_args->addr;
+	struct page *page = pfn_to_page(fcp_args->pfn);
+	int exec = vma->vm_flags & VM_EXEC;
+	struct mm_struct *mm = vma->vm_mm;
+	int map_coherent = 0;
+	pgd_t *pgdp;
+	pud_t *pudp;
+	pmd_t *pmdp;
+	pte_t *ptep;
+	void *vaddr;
+
+	/*
+	 * If ownes no valid ASID yet, cannot possibly have gotten
+	 * this page into the cache.
+	 */
+	if (!has_valid_asid(mm))
+		return;
+
+	addr &= PAGE_MASK;
+	pgdp = pgd_offset(mm, addr);
+	pudp = pud_offset(pgdp, addr);
+	pmdp = pmd_offset(pudp, addr);
+	ptep = pte_offset(pmdp, addr);
+
+	/*
+	 * If the page isn't marked valid, the page cannot possibly be
+	 * in the cache.
+	 */
+	if (!(pte_present(*ptep)))
+		return;
+
+	if ((mm == current->active_mm) && (pte_val(*ptep) & _PAGE_VALID))
+		vaddr = NULL;
+	else {
+		/*
+		 * Use kmap_coherent or kmap_atomic to do flushes for
+		 * another ASID than the current one.
+		 */
+		map_coherent = (cpu_has_dc_aliases &&
+				page_mapped(page) && !Page_dcache_dirty(page));
+		if (map_coherent)
+			vaddr = kmap_coherent(page, addr);
+		else
+			vaddr = kmap_atomic(page);
+		addr = (unsigned long)vaddr;
+	}
+
+	if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc)) {
+		r4k_blast_dcache_page(addr);
+		if (exec && !cpu_icache_snoops_remote_store)
+			r4k_blast_scache_page(addr);
+	}
+	if (exec) {
+		if (vaddr && cpu_has_vtag_icache && mm == current->active_mm) {
+			int cpu = smp_processor_id();
+
+			if (cpu_context(cpu, mm) != 0)
+				drop_mmu_context(mm, cpu);
+		} else
+			r4k_blast_icache_page(addr);
+	}
+
+	if (vaddr) {
+		if (map_coherent)
+			kunmap_coherent();
+		else
+			kunmap_atomic(vaddr);
+	}
+}
+
+static void r4k_flush_cache_page(struct vm_area_struct *vma,
+	unsigned long addr, unsigned long pfn)
+{
+	struct flush_cache_page_args args;
+
+	args.vma = vma;
+	args.addr = addr;
+	args.pfn = pfn;
+
+	r4k_on_each_cpu(local_r4k_flush_cache_page, &args);
+}
+
+static inline void local_r4k_flush_data_cache_page(void * addr)
+{
+	r4k_blast_dcache_page((unsigned long) addr);
+}
+
+static void r4k_flush_data_cache_page(unsigned long addr)
+{
+	if (in_atomic())
+		local_r4k_flush_data_cache_page((void *)addr);
+	else
+		r4k_on_each_cpu(local_r4k_flush_data_cache_page, (void *) addr);
+}
+
+struct flush_icache_range_args {
+	unsigned long start;
+	unsigned long end;
+};
+
+static inline void local_r4k_flush_icache_range(unsigned long start, unsigned long end)
+{
+	if (!cpu_has_ic_fills_f_dc) {
+		if (end - start >= dcache_size) {
+			r4k_blast_dcache();
+		} else {
+			R4600_HIT_CACHEOP_WAR_IMPL;
+			protected_blast_dcache_range(start, end);
+		}
+	}
+
+	if (end - start > icache_size)
+		r4k_blast_icache();
+	else
+		protected_blast_icache_range(start, end);
+}
+
+static inline void local_r4k_flush_icache_range_ipi(void *args)
+{
+	struct flush_icache_range_args *fir_args = args;
+	unsigned long start = fir_args->start;
+	unsigned long end = fir_args->end;
+
+	local_r4k_flush_icache_range(start, end);
+}
+
+static void r4k_flush_icache_range(unsigned long start, unsigned long end)
+{
+	struct flush_icache_range_args args;
+
+	args.start = start;
+	args.end = end;
+
+	r4k_on_each_cpu(local_r4k_flush_icache_range_ipi, &args);
+	instruction_hazard();
+}
+
+#ifdef CONFIG_DMA_NONCOHERENT
+
+static void r4k_dma_cache_wback_inv(unsigned long addr, unsigned long size)
+{
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	if (cpu_has_inclusive_pcaches) {
+		if (size >= scache_size)
+			r4k_blast_scache();
+		else
+			blast_scache_range(addr, addr + size);
+		__sync();
+		return;
+	}
+
+	/*
+	 * Either no secondary cache or the available caches don't have the
+	 * subset property so we have to flush the primary caches
+	 * explicitly
+	 */
+	if (cpu_has_safe_index_cacheops && size >= dcache_size) {
+		r4k_blast_dcache();
+	} else {
+		R4600_HIT_CACHEOP_WAR_IMPL;
+		blast_dcache_range(addr, addr + size);
+	}
+
+	bc_wback_inv(addr, size);
+	__sync();
+}
+
+static void r4k_dma_cache_inv(unsigned long addr, unsigned long size)
+{
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	if (cpu_has_inclusive_pcaches) {
+		if (size >= scache_size)
+			r4k_blast_scache();
+		else {
+			unsigned long lsize = cpu_scache_line_size();
+			unsigned long almask = ~(lsize - 1);
+
+			/*
+			 * There is no clearly documented alignment requirement
+			 * for the cache instruction on MIPS processors and
+			 * some processors, among them the RM5200 and RM7000
+			 * QED processors will throw an address error for cache
+			 * hit ops with insufficient alignment.  Solved by
+			 * aligning the address to cache line size.
+			 */
+			cache_op(Hit_Writeback_Inv_SD, addr & almask);
+			cache_op(Hit_Writeback_Inv_SD,
+				 (addr + size - 1) & almask);
+			blast_inv_scache_range(addr, addr + size);
+		}
+		__sync();
+		return;
+	}
+
+	if (cpu_has_safe_index_cacheops && size >= dcache_size) {
+		r4k_blast_dcache();
+	} else {
+		unsigned long lsize = cpu_dcache_line_size();
+		unsigned long almask = ~(lsize - 1);
+
+		R4600_HIT_CACHEOP_WAR_IMPL;
+		cache_op(Hit_Writeback_Inv_D, addr & almask);
+		cache_op(Hit_Writeback_Inv_D, (addr + size - 1)  & almask);
+		blast_inv_dcache_range(addr, addr + size);
+	}
+
+	bc_inv(addr, size);
+	__sync();
+}
+#endif /* CONFIG_DMA_NONCOHERENT */
+
+/*
+ * While we're protected against bad userland addresses we don't care
+ * very much about what happens in that case.  Usually a segmentation
+ * fault will dump the process later on anyway ...
+ */
+static void local_r4k_flush_cache_sigtramp(void * arg)
+{
+	unsigned long ic_lsize = cpu_icache_line_size();
+	unsigned long dc_lsize = cpu_dcache_line_size();
+	unsigned long sc_lsize = cpu_scache_line_size();
+	unsigned long addr = (unsigned long) arg;
+
+	R4600_HIT_CACHEOP_WAR_IMPL;
+	if (dc_lsize)
+		protected_writeback_dcache_line(addr & ~(dc_lsize - 1));
+	if (!cpu_icache_snoops_remote_store && scache_size)
+		protected_writeback_scache_line(addr & ~(sc_lsize - 1));
+	if (ic_lsize)
+		protected_flush_icache_line(addr & ~(ic_lsize - 1));
+	if (MIPS4K_ICACHE_REFILL_WAR) {
+		__asm__ __volatile__ (
+			".set push\n\t"
+			".set noat\n\t"
+			".set mips3\n\t"
+#ifdef CONFIG_32BIT
+			"la	$at,1f\n\t"
+#endif
+#ifdef CONFIG_64BIT
+			"dla	$at,1f\n\t"
+#endif
+			"cache	%0,($at)\n\t"
+			"nop; nop; nop\n"
+			"1:\n\t"
+			".set pop"
+			:
+			: "i" (Hit_Invalidate_I));
+	}
+	if (MIPS_CACHE_SYNC_WAR)
+		__asm__ __volatile__ ("sync");
+}
+
+static void r4k_flush_cache_sigtramp(unsigned long addr)
+{
+	r4k_on_each_cpu(local_r4k_flush_cache_sigtramp, (void *) addr);
+}
+
+static void r4k_flush_icache_all(void)
+{
+	if (cpu_has_vtag_icache)
+		r4k_blast_icache();
+}
+
+struct flush_kernel_vmap_range_args {
+	unsigned long	vaddr;
+	int		size;
+};
+
+static inline void local_r4k_flush_kernel_vmap_range(void *args)
+{
+	struct flush_kernel_vmap_range_args *vmra = args;
+	unsigned long vaddr = vmra->vaddr;
+	int size = vmra->size;
+
+	/*
+	 * Aliases only affect the primary caches so don't bother with
+	 * S-caches or T-caches.
+	 */
+	if (cpu_has_safe_index_cacheops && size >= dcache_size)
+		r4k_blast_dcache();
+	else {
+		R4600_HIT_CACHEOP_WAR_IMPL;
+		blast_dcache_range(vaddr, vaddr + size);
+	}
+}
+
+static void r4k_flush_kernel_vmap_range(unsigned long vaddr, int size)
+{
+	struct flush_kernel_vmap_range_args args;
+
+	args.vaddr = (unsigned long) vaddr;
+	args.size = size;
+
+	r4k_on_each_cpu(local_r4k_flush_kernel_vmap_range, &args);
+}
+
+static inline void rm7k_erratum31(void)
+{
+	const unsigned long ic_lsize = 32;
+	unsigned long addr;
+
+	/* RM7000 erratum #31. The icache is screwed at startup. */
+	write_c0_taglo(0);
+	write_c0_taghi(0);
+
+	for (addr = INDEX_BASE; addr <= INDEX_BASE + 4096; addr += ic_lsize) {
+		__asm__ __volatile__ (
+			".set push\n\t"
+			".set noreorder\n\t"
+			".set mips3\n\t"
+			"cache\t%1, 0(%0)\n\t"
+			"cache\t%1, 0x1000(%0)\n\t"
+			"cache\t%1, 0x2000(%0)\n\t"
+			"cache\t%1, 0x3000(%0)\n\t"
+			"cache\t%2, 0(%0)\n\t"
+			"cache\t%2, 0x1000(%0)\n\t"
+			"cache\t%2, 0x2000(%0)\n\t"
+			"cache\t%2, 0x3000(%0)\n\t"
+			"cache\t%1, 0(%0)\n\t"
+			"cache\t%1, 0x1000(%0)\n\t"
+			"cache\t%1, 0x2000(%0)\n\t"
+			"cache\t%1, 0x3000(%0)\n\t"
+			".set pop\n"
+			:
+			: "r" (addr), "i" (Index_Store_Tag_I), "i" (Fill));
+	}
+}
+
+static char *way_string[] __cpuinitdata = { NULL, "direct mapped", "2-way",
+	"3-way", "4-way", "5-way", "6-way", "7-way", "8-way"
+};
+
+static void __cpuinit probe_pcache(void)
+{
+	struct cpuinfo_mips *c = &current_cpu_data;
+	unsigned int config = read_c0_config();
+	unsigned int prid = read_c0_prid();
+	unsigned long config1;
+	unsigned int lsize;
+
+	switch (c->cputype) {
+	case CPU_R4600:			/* QED style two way caches? */
+	case CPU_R4700:
+	case CPU_R5000:
+	case CPU_NEVADA:
+		icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 2;
+		c->icache.waybit = __ffs(icache_size/2);
+
+		dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 2;
+		c->dcache.waybit= __ffs(dcache_size/2);
+
+		c->options |= MIPS_CPU_CACHE_CDEX_P;
+		break;
+
+	case CPU_R5432:
+	case CPU_R5500:
+		icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 2;
+		c->icache.waybit= 0;
+
+		dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 2;
+		c->dcache.waybit = 0;
+
+		c->options |= MIPS_CPU_CACHE_CDEX_P | MIPS_CPU_PREFETCH;
+		break;
+
+	case CPU_TX49XX:
+		icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 4;
+		c->icache.waybit= 0;
+
+		dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 4;
+		c->dcache.waybit = 0;
+
+		c->options |= MIPS_CPU_CACHE_CDEX_P;
+		c->options |= MIPS_CPU_PREFETCH;
+		break;
+
+	case CPU_R4000PC:
+	case CPU_R4000SC:
+	case CPU_R4000MC:
+	case CPU_R4400PC:
+	case CPU_R4400SC:
+	case CPU_R4400MC:
+	case CPU_R4300:
+		icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 1;
+		c->icache.waybit = 0; 	/* doesn't matter */
+
+		dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 1;
+		c->dcache.waybit = 0;	/* does not matter */
+
+		c->options |= MIPS_CPU_CACHE_CDEX_P;
+		break;
+
+	case CPU_R10000:
+	case CPU_R12000:
+	case CPU_R14000:
+		icache_size = 1 << (12 + ((config & R10K_CONF_IC) >> 29));
+		c->icache.linesz = 64;
+		c->icache.ways = 2;
+		c->icache.waybit = 0;
+
+		dcache_size = 1 << (12 + ((config & R10K_CONF_DC) >> 26));
+		c->dcache.linesz = 32;
+		c->dcache.ways = 2;
+		c->dcache.waybit = 0;
+
+		c->options |= MIPS_CPU_PREFETCH;
+		break;
+
+	case CPU_VR4133:
+		write_c0_config(config & ~VR41_CONF_P4K);
+	case CPU_VR4131:
+		/* Workaround for cache instruction bug of VR4131 */
+		if (c->processor_id == 0x0c80U || c->processor_id == 0x0c81U ||
+		    c->processor_id == 0x0c82U) {
+			config |= 0x00400000U;
+			if (c->processor_id == 0x0c80U)
+				config |= VR41_CONF_BP;
+			write_c0_config(config);
+		} else
+			c->options |= MIPS_CPU_CACHE_CDEX_P;
+
+		icache_size = 1 << (10 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 2;
+		c->icache.waybit = __ffs(icache_size/2);
+
+		dcache_size = 1 << (10 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 2;
+		c->dcache.waybit = __ffs(dcache_size/2);
+		break;
+
+	case CPU_VR41XX:
+	case CPU_VR4111:
+	case CPU_VR4121:
+	case CPU_VR4122:
+	case CPU_VR4181:
+	case CPU_VR4181A:
+		icache_size = 1 << (10 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 1;
+		c->icache.waybit = 0; 	/* doesn't matter */
+
+		dcache_size = 1 << (10 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 1;
+		c->dcache.waybit = 0;	/* does not matter */
+
+		c->options |= MIPS_CPU_CACHE_CDEX_P;
+		break;
+
+	case CPU_RM7000:
+		rm7k_erratum31();
+
+	case CPU_RM9000:
+		icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		c->icache.ways = 4;
+		c->icache.waybit = __ffs(icache_size / c->icache.ways);
+
+		dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		c->dcache.ways = 4;
+		c->dcache.waybit = __ffs(dcache_size / c->dcache.ways);
+
+#if !defined(CONFIG_SMP) || !defined(RM9000_CDEX_SMP_WAR)
+		c->options |= MIPS_CPU_CACHE_CDEX_P;
+#endif
+		c->options |= MIPS_CPU_PREFETCH;
+		break;
+
+	case CPU_LOONGSON2:
+		icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+		c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+		if (prid & 0x3)
+			c->icache.ways = 4;
+		else
+			c->icache.ways = 2;
+		c->icache.waybit = 0;
+
+		dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+		c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+		if (prid & 0x3)
+			c->dcache.ways = 4;
+		else
+			c->dcache.ways = 2;
+		c->dcache.waybit = 0;
+		break;
+
+	default:
+		if (!(config & MIPS_CONF_M))
+			panic("Don't know how to probe P-caches on this cpu.");
+
+		/*
+		 * So we seem to be a MIPS32 or MIPS64 CPU
+		 * So let's probe the I-cache ...
+		 */
+		config1 = read_c0_config1();
+
+		if ((lsize = ((config1 >> 19) & 7)))
+			c->icache.linesz = 2 << lsize;
+		else
+			c->icache.linesz = lsize;
+		c->icache.sets = 64 << ((config1 >> 22) & 7);
+		c->icache.ways = 1 + ((config1 >> 16) & 7);
+
+		icache_size = c->icache.sets *
+		              c->icache.ways *
+		              c->icache.linesz;
+		c->icache.waybit = __ffs(icache_size/c->icache.ways);
+
+		if (config & 0x8)		/* VI bit */
+			c->icache.flags |= MIPS_CACHE_VTAG;
+
+		/*
+		 * Now probe the MIPS32 / MIPS64 data cache.
+		 */
+		c->dcache.flags = 0;
+
+		if ((lsize = ((config1 >> 10) & 7)))
+			c->dcache.linesz = 2 << lsize;
+		else
+			c->dcache.linesz= lsize;
+		c->dcache.sets = 64 << ((config1 >> 13) & 7);
+		c->dcache.ways = 1 + ((config1 >> 7) & 7);
+
+		dcache_size = c->dcache.sets *
+		              c->dcache.ways *
+		              c->dcache.linesz;
+		c->dcache.waybit = __ffs(dcache_size/c->dcache.ways);
+
+		c->options |= MIPS_CPU_PREFETCH;
+		break;
+	}
+
+	/*
+	 * Processor configuration sanity check for the R4000SC erratum
+	 * #5.  With page sizes larger than 32kB there is no possibility
+	 * to get a VCE exception anymore so we don't care about this
+	 * misconfiguration.  The case is rather theoretical anyway;
+	 * presumably no vendor is shipping his hardware in the "bad"
+	 * configuration.
+	 */
+	if ((prid & 0xff00) == PRID_IMP_R4000 && (prid & 0xff) < 0x40 &&
+	    !(config & CONF_SC) && c->icache.linesz != 16 &&
+	    PAGE_SIZE <= 0x8000)
+		panic("Improper R4000SC processor configuration detected");
+
+	/* compute a couple of other cache variables */
+	c->icache.waysize = icache_size / c->icache.ways;
+	c->dcache.waysize = dcache_size / c->dcache.ways;
+
+	c->icache.sets = c->icache.linesz ?
+		icache_size / (c->icache.linesz * c->icache.ways) : 0;
+	c->dcache.sets = c->dcache.linesz ?
+		dcache_size / (c->dcache.linesz * c->dcache.ways) : 0;
+
+	/*
+	 * R10000 and R12000 P-caches are odd in a positive way.  They're 32kB
+	 * 2-way virtually indexed so normally would suffer from aliases.  So
+	 * normally they'd suffer from aliases but magic in the hardware deals
+	 * with that for us so we don't need to take care ourselves.
+	 */
+	switch (c->cputype) {
+	case CPU_20KC:
+	case CPU_25KF:
+	case CPU_SB1:
+	case CPU_SB1A:
+	case CPU_XLR:
+		c->dcache.flags |= MIPS_CACHE_PINDEX;
+		break;
+
+	case CPU_R10000:
+	case CPU_R12000:
+	case CPU_R14000:
+		break;
+
+	case CPU_24K:
+	case CPU_34K:
+	case CPU_74K:
+	case CPU_1004K:
+		if ((read_c0_config7() & (1 << 16))) {
+			/* effectively physically indexed dcache,
+			   thus no virtual aliases. */
+			c->dcache.flags |= MIPS_CACHE_PINDEX;
+			break;
+		}
+	default:
+		if (c->dcache.waysize > PAGE_SIZE)
+			c->dcache.flags |= MIPS_CACHE_ALIASES;
+	}
+
+	switch (c->cputype) {
+	case CPU_20KC:
+		/*
+		 * Some older 20Kc chips doesn't have the 'VI' bit in
+		 * the config register.
+		 */
+		c->icache.flags |= MIPS_CACHE_VTAG;
+		break;
+
+	case CPU_ALCHEMY:
+		c->icache.flags |= MIPS_CACHE_IC_F_DC;
+		break;
+	}
+
+#ifdef  CONFIG_CPU_LOONGSON2
+	/*
+	 * LOONGSON2 has 4 way icache, but when using indexed cache op,
+	 * one op will act on all 4 ways
+	 */
+	c->icache.ways = 1;
+#endif
+
+	printk("Primary instruction cache %ldkB, %s, %s, linesize %d bytes.\n",
+	       icache_size >> 10,
+	       c->icache.flags & MIPS_CACHE_VTAG ? "VIVT" : "VIPT",
+	       way_string[c->icache.ways], c->icache.linesz);
+
+	printk("Primary data cache %ldkB, %s, %s, %s, linesize %d bytes\n",
+	       dcache_size >> 10, way_string[c->dcache.ways],
+	       (c->dcache.flags & MIPS_CACHE_PINDEX) ? "PIPT" : "VIPT",
+	       (c->dcache.flags & MIPS_CACHE_ALIASES) ?
+			"cache aliases" : "no aliases",
+	       c->dcache.linesz);
+}
+
+/*
+ * If you even _breathe_ on this function, look at the gcc output and make sure
+ * it does not pop things on and off the stack for the cache sizing loop that
+ * executes in KSEG1 space or else you will crash and burn badly.  You have
+ * been warned.
+ */
+static int __cpuinit probe_scache(void)
+{
+	unsigned long flags, addr, begin, end, pow2;
+	unsigned int config = read_c0_config();
+	struct cpuinfo_mips *c = &current_cpu_data;
+
+	if (config & CONF_SC)
+		return 0;
+
+	begin = (unsigned long) &_stext;
+	begin &= ~((4 * 1024 * 1024) - 1);
+	end = begin + (4 * 1024 * 1024);
+
+	/*
+	 * This is such a bitch, you'd think they would make it easy to do
+	 * this.  Away you daemons of stupidity!
+	 */
+	local_irq_save(flags);
+
+	/* Fill each size-multiple cache line with a valid tag. */
+	pow2 = (64 * 1024);
+	for (addr = begin; addr < end; addr = (begin + pow2)) {
+		unsigned long *p = (unsigned long *) addr;
+		__asm__ __volatile__("nop" : : "r" (*p)); /* whee... */
+		pow2 <<= 1;
+	}
+
+	/* Load first line with zero (therefore invalid) tag. */
+	write_c0_taglo(0);
+	write_c0_taghi(0);
+	__asm__ __volatile__("nop; nop; nop; nop;"); /* avoid the hazard */
+	cache_op(Index_Store_Tag_I, begin);
+	cache_op(Index_Store_Tag_D, begin);
+	cache_op(Index_Store_Tag_SD, begin);
+
+	/* Now search for the wrap around point. */
+	pow2 = (128 * 1024);
+	for (addr = begin + (128 * 1024); addr < end; addr = begin + pow2) {
+		cache_op(Index_Load_Tag_SD, addr);
+		__asm__ __volatile__("nop; nop; nop; nop;"); /* hazard... */
+		if (!read_c0_taglo())
+			break;
+		pow2 <<= 1;
+	}
+	local_irq_restore(flags);
+	addr -= begin;
+
+	scache_size = addr;
+	c->scache.linesz = 16 << ((config & R4K_CONF_SB) >> 22);
+	c->scache.ways = 1;
+	c->dcache.waybit = 0;		/* does not matter */
+
+	return 1;
+}
+
+#if defined(CONFIG_CPU_LOONGSON2)
+static void __init loongson2_sc_init(void)
+{
+	struct cpuinfo_mips *c = &current_cpu_data;
+
+	scache_size = 512*1024;
+	c->scache.linesz = 32;
+	c->scache.ways = 4;
+	c->scache.waybit = 0;
+	c->scache.waysize = scache_size / (c->scache.ways);
+	c->scache.sets = scache_size / (c->scache.linesz * c->scache.ways);
+	pr_info("Unified secondary cache %ldkB %s, linesize %d bytes.\n",
+	       scache_size >> 10, way_string[c->scache.ways], c->scache.linesz);
+
+	c->options |= MIPS_CPU_INCLUSIVE_CACHES;
+}
+#endif
+
+extern int r5k_sc_init(void);
+extern int rm7k_sc_init(void);
+extern int mips_sc_init(void);
+
+static void __cpuinit setup_scache(void)
+{
+	struct cpuinfo_mips *c = &current_cpu_data;
+	unsigned int config = read_c0_config();
+	int sc_present = 0;
+
+	/*
+	 * Do the probing thing on R4000SC and R4400SC processors.  Other
+	 * processors don't have a S-cache that would be relevant to the
+	 * Linux memory management.
+	 */
+	switch (c->cputype) {
+	case CPU_R4000SC:
+	case CPU_R4000MC:
+	case CPU_R4400SC:
+	case CPU_R4400MC:
+		sc_present = run_uncached(probe_scache);
+		if (sc_present)
+			c->options |= MIPS_CPU_CACHE_CDEX_S;
+		break;
+
+	case CPU_R10000:
+	case CPU_R12000:
+	case CPU_R14000:
+		scache_size = 0x80000 << ((config & R10K_CONF_SS) >> 16);
+		c->scache.linesz = 64 << ((config >> 13) & 1);
+		c->scache.ways = 2;
+		c->scache.waybit= 0;
+		sc_present = 1;
+		break;
+
+	case CPU_R5000:
+	case CPU_NEVADA:
+#ifdef CONFIG_R5000_CPU_SCACHE
+		r5k_sc_init();
+#endif
+                return;
+
+	case CPU_RM7000:
+	case CPU_RM9000:
+#ifdef CONFIG_RM7000_CPU_SCACHE
+		rm7k_sc_init();
+#endif
+		return;
+
+#if defined(CONFIG_CPU_LOONGSON2)
+	case CPU_LOONGSON2:
+		loongson2_sc_init();
+		return;
+#endif
+	case CPU_XLP:
+		/* don't need to worry about L2, fully coherent */
+		return;
+
+	default:
+		if (c->isa_level == MIPS_CPU_ISA_M32R1 ||
+		    c->isa_level == MIPS_CPU_ISA_M32R2 ||
+		    c->isa_level == MIPS_CPU_ISA_M64R1 ||
+		    c->isa_level == MIPS_CPU_ISA_M64R2) {
+#ifdef CONFIG_MIPS_CPU_SCACHE
+			if (mips_sc_init ()) {
+				scache_size = c->scache.ways * c->scache.sets * c->scache.linesz;
+				printk("MIPS secondary cache %ldkB, %s, linesize %d bytes.\n",
+				       scache_size >> 10,
+				       way_string[c->scache.ways], c->scache.linesz);
+			}
+#else
+			if (!(c->scache.flags & MIPS_CACHE_NOT_PRESENT))
+				panic("Dunno how to handle MIPS32 / MIPS64 second level cache");
+#endif
+			return;
+		}
+		sc_present = 0;
+	}
+
+	if (!sc_present)
+		return;
+
+	/* compute a couple of other cache variables */
+	c->scache.waysize = scache_size / c->scache.ways;
+
+	c->scache.sets = scache_size / (c->scache.linesz * c->scache.ways);
+
+	printk("Unified secondary cache %ldkB %s, linesize %d bytes.\n",
+	       scache_size >> 10, way_string[c->scache.ways], c->scache.linesz);
+
+	c->options |= MIPS_CPU_INCLUSIVE_CACHES;
+}
+
+void au1x00_fixup_config_od(void)
+{
+	/*
+	 * c0_config.od (bit 19) was write only (and read as 0)
+	 * on the early revisions of Alchemy SOCs.  It disables the bus
+	 * transaction overlapping and needs to be set to fix various errata.
+	 */
+	switch (read_c0_prid()) {
+	case 0x00030100: /* Au1000 DA */
+	case 0x00030201: /* Au1000 HA */
+	case 0x00030202: /* Au1000 HB */
+	case 0x01030200: /* Au1500 AB */
+	/*
+	 * Au1100 errata actually keeps silence about this bit, so we set it
+	 * just in case for those revisions that require it to be set according
+	 * to the (now gone) cpu table.
+	 */
+	case 0x02030200: /* Au1100 AB */
+	case 0x02030201: /* Au1100 BA */
+	case 0x02030202: /* Au1100 BC */
+		set_c0_config(1 << 19);
+		break;
+	}
+}
+
+/* CP0 hazard avoidance. */
+#define NXP_BARRIER()							\
+	 __asm__ __volatile__(						\
+	".set noreorder\n\t"						\
+	"nop; nop; nop; nop; nop; nop;\n\t"				\
+	".set reorder\n\t")
+
+static void nxp_pr4450_fixup_config(void)
+{
+	unsigned long config0;
+
+	config0 = read_c0_config();
+
+	/* clear all three cache coherency fields */
+	config0 &= ~(0x7 | (7 << 25) | (7 << 28));
+	config0 |= (((_page_cachable_default >> _CACHE_SHIFT) <<  0) |
+		    ((_page_cachable_default >> _CACHE_SHIFT) << 25) |
+		    ((_page_cachable_default >> _CACHE_SHIFT) << 28));
+	write_c0_config(config0);
+	NXP_BARRIER();
+}
+
+static int __cpuinitdata cca = -1;
+
+static int __init cca_setup(char *str)
+{
+	get_option(&str, &cca);
+
+	return 1;
+}
+
+__setup("cca=", cca_setup);
+
+static void __cpuinit coherency_setup(void)
+{
+	if (cca < 0 || cca > 7)
+		cca = read_c0_config() & CONF_CM_CMASK;
+	_page_cachable_default = cca << _CACHE_SHIFT;
+
+	pr_debug("Using cache attribute %d\n", cca);
+	change_c0_config(CONF_CM_CMASK, cca);
+
+	/*
+	 * c0_status.cu=0 specifies that updates by the sc instruction use
+	 * the coherency mode specified by the TLB; 1 means cachable
+	 * coherent update on write will be used.  Not all processors have
+	 * this bit and; some wire it to zero, others like Toshiba had the
+	 * silly idea of putting something else there ...
+	 */
+	switch (current_cpu_type()) {
+	case CPU_R4000PC:
+	case CPU_R4000SC:
+	case CPU_R4000MC:
+	case CPU_R4400PC:
+	case CPU_R4400SC:
+	case CPU_R4400MC:
+		clear_c0_config(CONF_CU);
+		break;
+	/*
+	 * We need to catch the early Alchemy SOCs with
+	 * the write-only co_config.od bit and set it back to one on:
+	 * Au1000 rev DA, HA, HB;  Au1100 AB, BA, BC, Au1500 AB
+	 */
+	case CPU_ALCHEMY:
+		au1x00_fixup_config_od();
+		break;
+
+	case PRID_IMP_PR4450:
+		nxp_pr4450_fixup_config();
+		break;
+	}
+}
+
+#if defined(CONFIG_DMA_NONCOHERENT)
+
+static int __cpuinitdata coherentio;
+
+static int __init setcoherentio(char *str)
+{
+	coherentio = 1;
+
+	return 1;
+}
+
+__setup("coherentio", setcoherentio);
+#endif
+
+void __cpuinit r4k_cache_init(void)
+{
+	extern void build_clear_page(void);
+	extern void build_copy_page(void);
+	extern char __weak except_vec2_generic;
+	extern char __weak except_vec2_sb1;
+	struct cpuinfo_mips *c = &current_cpu_data;
+
+	switch (c->cputype) {
+	case CPU_SB1:
+	case CPU_SB1A:
+		set_uncached_handler(0x100, &except_vec2_sb1, 0x80);
+		break;
+
+	default:
+		set_uncached_handler(0x100, &except_vec2_generic, 0x80);
+		break;
+	}
+
+	probe_pcache();
+	setup_scache();
+
+	r4k_blast_dcache_page_setup();
+	r4k_blast_dcache_page_indexed_setup();
+	r4k_blast_dcache_setup();
+	r4k_blast_icache_page_setup();
+	r4k_blast_icache_page_indexed_setup();
+	r4k_blast_icache_setup();
+	r4k_blast_scache_page_setup();
+	r4k_blast_scache_page_indexed_setup();
+	r4k_blast_scache_setup();
+
+	/*
+	 * Some MIPS32 and MIPS64 processors have physically indexed caches.
+	 * This code supports virtually indexed processors and will be
+	 * unnecessarily inefficient on physically indexed processors.
+	 */
+	if (c->dcache.linesz)
+		shm_align_mask = max_t( unsigned long,
+					c->dcache.sets * c->dcache.linesz - 1,
+					PAGE_SIZE - 1);
+	else
+		shm_align_mask = PAGE_SIZE-1;
+
+	__flush_cache_vmap	= r4k__flush_cache_vmap;
+	__flush_cache_vunmap	= r4k__flush_cache_vunmap;
+
+	flush_cache_all		= cache_noop;
+	__flush_cache_all	= r4k___flush_cache_all;
+	flush_cache_mm		= r4k_flush_cache_mm;
+	flush_cache_page	= r4k_flush_cache_page;
+	flush_cache_range	= r4k_flush_cache_range;
+
+	__flush_kernel_vmap_range = r4k_flush_kernel_vmap_range;
+
+	flush_cache_sigtramp	= r4k_flush_cache_sigtramp;
+	flush_icache_all	= r4k_flush_icache_all;
+	local_flush_data_cache_page	= local_r4k_flush_data_cache_page;
+	flush_data_cache_page	= r4k_flush_data_cache_page;
+	flush_icache_range	= r4k_flush_icache_range;
+	local_flush_icache_range	= local_r4k_flush_icache_range;
+
+#if defined(CONFIG_DMA_NONCOHERENT)
+	if (coherentio) {
+		_dma_cache_wback_inv	= (void *)cache_noop;
+		_dma_cache_wback	= (void *)cache_noop;
+		_dma_cache_inv		= (void *)cache_noop;
+	} else {
+		_dma_cache_wback_inv	= r4k_dma_cache_wback_inv;
+		_dma_cache_wback	= r4k_dma_cache_wback_inv;
+		_dma_cache_inv		= r4k_dma_cache_inv;
+	}
+#endif
+
+	build_clear_page();
+	build_copy_page();
+#if !defined(CONFIG_MIPS_CMP)
+	local_r4k___flush_cache_all(NULL);
+#endif
+	coherency_setup();
+}
diff --git a/arch/mips/mm/c-tx39.c b/arch/mips/mm/c-tx39.c
new file mode 100644
index 00000000..87d23cad
--- /dev/null
+++ b/arch/mips/mm/c-tx39.c
@@ -0,0 +1,440 @@
+/*
+ * r2300.c: R2000 and R3000 specific mmu/cache code.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ *
+ * with a lot of changes to make this thing work for R3000s
+ * Tx39XX R4k style caches added. HK
+ * Copyright (C) 1998, 1999, 2000 Harald Koerfgen
+ * Copyright (C) 1998 Gleb Raiko & Vladimir Roganov
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+
+#include <asm/cacheops.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/isadep.h>
+#include <asm/io.h>
+#include <asm/bootinfo.h>
+#include <asm/cpu.h>
+
+/* For R3000 cores with R4000 style caches */
+static unsigned long icache_size, dcache_size;		/* Size in bytes */
+
+#include <asm/r4kcache.h>
+
+extern int r3k_have_wired_reg;	/* in r3k-tlb.c */
+
+/* This sequence is required to ensure icache is disabled immediately */
+#define TX39_STOP_STREAMING() \
+__asm__ __volatile__( \
+	".set    push\n\t" \
+	".set    noreorder\n\t" \
+	"b       1f\n\t" \
+	"nop\n\t" \
+	"1:\n\t" \
+	".set pop" \
+	)
+
+/* TX39H-style cache flush routines. */
+static void tx39h_flush_icache_all(void)
+{
+	unsigned long flags, config;
+
+	/* disable icache (set ICE#) */
+	local_irq_save(flags);
+	config = read_c0_conf();
+	write_c0_conf(config & ~TX39_CONF_ICE);
+	TX39_STOP_STREAMING();
+	blast_icache16();
+	write_c0_conf(config);
+	local_irq_restore(flags);
+}
+
+static void tx39h_dma_cache_wback_inv(unsigned long addr, unsigned long size)
+{
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	iob();
+	blast_inv_dcache_range(addr, addr + size);
+}
+
+
+/* TX39H2,TX39H3 */
+static inline void tx39_blast_dcache_page(unsigned long addr)
+{
+	if (current_cpu_type() != CPU_TX3912)
+		blast_dcache16_page(addr);
+}
+
+static inline void tx39_blast_dcache_page_indexed(unsigned long addr)
+{
+	blast_dcache16_page_indexed(addr);
+}
+
+static inline void tx39_blast_dcache(void)
+{
+	blast_dcache16();
+}
+
+static inline void tx39_blast_icache_page(unsigned long addr)
+{
+	unsigned long flags, config;
+	/* disable icache (set ICE#) */
+	local_irq_save(flags);
+	config = read_c0_conf();
+	write_c0_conf(config & ~TX39_CONF_ICE);
+	TX39_STOP_STREAMING();
+	blast_icache16_page(addr);
+	write_c0_conf(config);
+	local_irq_restore(flags);
+}
+
+static inline void tx39_blast_icache_page_indexed(unsigned long addr)
+{
+	unsigned long flags, config;
+	/* disable icache (set ICE#) */
+	local_irq_save(flags);
+	config = read_c0_conf();
+	write_c0_conf(config & ~TX39_CONF_ICE);
+	TX39_STOP_STREAMING();
+	blast_icache16_page_indexed(addr);
+	write_c0_conf(config);
+	local_irq_restore(flags);
+}
+
+static inline void tx39_blast_icache(void)
+{
+	unsigned long flags, config;
+	/* disable icache (set ICE#) */
+	local_irq_save(flags);
+	config = read_c0_conf();
+	write_c0_conf(config & ~TX39_CONF_ICE);
+	TX39_STOP_STREAMING();
+	blast_icache16();
+	write_c0_conf(config);
+	local_irq_restore(flags);
+}
+
+static void tx39__flush_cache_vmap(void)
+{
+	tx39_blast_dcache();
+}
+
+static void tx39__flush_cache_vunmap(void)
+{
+	tx39_blast_dcache();
+}
+
+static inline void tx39_flush_cache_all(void)
+{
+	if (!cpu_has_dc_aliases)
+		return;
+
+	tx39_blast_dcache();
+}
+
+static inline void tx39___flush_cache_all(void)
+{
+	tx39_blast_dcache();
+	tx39_blast_icache();
+}
+
+static void tx39_flush_cache_mm(struct mm_struct *mm)
+{
+	if (!cpu_has_dc_aliases)
+		return;
+
+	if (cpu_context(smp_processor_id(), mm) != 0)
+		tx39_blast_dcache();
+}
+
+static void tx39_flush_cache_range(struct vm_area_struct *vma,
+	unsigned long start, unsigned long end)
+{
+	if (!cpu_has_dc_aliases)
+		return;
+	if (!(cpu_context(smp_processor_id(), vma->vm_mm)))
+		return;
+
+	tx39_blast_dcache();
+}
+
+static void tx39_flush_cache_page(struct vm_area_struct *vma, unsigned long page, unsigned long pfn)
+{
+	int exec = vma->vm_flags & VM_EXEC;
+	struct mm_struct *mm = vma->vm_mm;
+	pgd_t *pgdp;
+	pud_t *pudp;
+	pmd_t *pmdp;
+	pte_t *ptep;
+
+	/*
+	 * If ownes no valid ASID yet, cannot possibly have gotten
+	 * this page into the cache.
+	 */
+	if (cpu_context(smp_processor_id(), mm) == 0)
+		return;
+
+	page &= PAGE_MASK;
+	pgdp = pgd_offset(mm, page);
+	pudp = pud_offset(pgdp, page);
+	pmdp = pmd_offset(pudp, page);
+	ptep = pte_offset(pmdp, page);
+
+	/*
+	 * If the page isn't marked valid, the page cannot possibly be
+	 * in the cache.
+	 */
+	if (!(pte_val(*ptep) & _PAGE_PRESENT))
+		return;
+
+	/*
+	 * Doing flushes for another ASID than the current one is
+	 * too difficult since stupid R4k caches do a TLB translation
+	 * for every cache flush operation.  So we do indexed flushes
+	 * in that case, which doesn't overly flush the cache too much.
+	 */
+	if ((mm == current->active_mm) && (pte_val(*ptep) & _PAGE_VALID)) {
+		if (cpu_has_dc_aliases || exec)
+			tx39_blast_dcache_page(page);
+		if (exec)
+			tx39_blast_icache_page(page);
+
+		return;
+	}
+
+	/*
+	 * Do indexed flush, too much work to get the (possible) TLB refills
+	 * to work correctly.
+	 */
+	if (cpu_has_dc_aliases || exec)
+		tx39_blast_dcache_page_indexed(page);
+	if (exec)
+		tx39_blast_icache_page_indexed(page);
+}
+
+static void local_tx39_flush_data_cache_page(void * addr)
+{
+	tx39_blast_dcache_page((unsigned long)addr);
+}
+
+static void tx39_flush_data_cache_page(unsigned long addr)
+{
+	tx39_blast_dcache_page(addr);
+}
+
+static void tx39_flush_icache_range(unsigned long start, unsigned long end)
+{
+	if (end - start > dcache_size)
+		tx39_blast_dcache();
+	else
+		protected_blast_dcache_range(start, end);
+
+	if (end - start > icache_size)
+		tx39_blast_icache();
+	else {
+		unsigned long flags, config;
+		/* disable icache (set ICE#) */
+		local_irq_save(flags);
+		config = read_c0_conf();
+		write_c0_conf(config & ~TX39_CONF_ICE);
+		TX39_STOP_STREAMING();
+		protected_blast_icache_range(start, end);
+		write_c0_conf(config);
+		local_irq_restore(flags);
+	}
+}
+
+static void tx39_flush_kernel_vmap_range(unsigned long vaddr, int size)
+{
+	BUG();
+}
+
+static void tx39_dma_cache_wback_inv(unsigned long addr, unsigned long size)
+{
+	unsigned long end;
+
+	if (((size | addr) & (PAGE_SIZE - 1)) == 0) {
+		end = addr + size;
+		do {
+			tx39_blast_dcache_page(addr);
+			addr += PAGE_SIZE;
+		} while(addr != end);
+	} else if (size > dcache_size) {
+		tx39_blast_dcache();
+	} else {
+		blast_dcache_range(addr, addr + size);
+	}
+}
+
+static void tx39_dma_cache_inv(unsigned long addr, unsigned long size)
+{
+	unsigned long end;
+
+	if (((size | addr) & (PAGE_SIZE - 1)) == 0) {
+		end = addr + size;
+		do {
+			tx39_blast_dcache_page(addr);
+			addr += PAGE_SIZE;
+		} while(addr != end);
+	} else if (size > dcache_size) {
+		tx39_blast_dcache();
+	} else {
+		blast_inv_dcache_range(addr, addr + size);
+	}
+}
+
+static void tx39_flush_cache_sigtramp(unsigned long addr)
+{
+	unsigned long ic_lsize = current_cpu_data.icache.linesz;
+	unsigned long dc_lsize = current_cpu_data.dcache.linesz;
+	unsigned long config;
+	unsigned long flags;
+
+	protected_writeback_dcache_line(addr & ~(dc_lsize - 1));
+
+	/* disable icache (set ICE#) */
+	local_irq_save(flags);
+	config = read_c0_conf();
+	write_c0_conf(config & ~TX39_CONF_ICE);
+	TX39_STOP_STREAMING();
+	protected_flush_icache_line(addr & ~(ic_lsize - 1));
+	write_c0_conf(config);
+	local_irq_restore(flags);
+}
+
+static __init void tx39_probe_cache(void)
+{
+	unsigned long config;
+
+	config = read_c0_conf();
+
+	icache_size = 1 << (10 + ((config & TX39_CONF_ICS_MASK) >>
+				  TX39_CONF_ICS_SHIFT));
+	dcache_size = 1 << (10 + ((config & TX39_CONF_DCS_MASK) >>
+				  TX39_CONF_DCS_SHIFT));
+
+	current_cpu_data.icache.linesz = 16;
+	switch (current_cpu_type()) {
+	case CPU_TX3912:
+		current_cpu_data.icache.ways = 1;
+		current_cpu_data.dcache.ways = 1;
+		current_cpu_data.dcache.linesz = 4;
+		break;
+
+	case CPU_TX3927:
+		current_cpu_data.icache.ways = 2;
+		current_cpu_data.dcache.ways = 2;
+		current_cpu_data.dcache.linesz = 16;
+		break;
+
+	case CPU_TX3922:
+	default:
+		current_cpu_data.icache.ways = 1;
+		current_cpu_data.dcache.ways = 1;
+		current_cpu_data.dcache.linesz = 16;
+		break;
+	}
+}
+
+void __cpuinit tx39_cache_init(void)
+{
+	extern void build_clear_page(void);
+	extern void build_copy_page(void);
+	unsigned long config;
+
+	config = read_c0_conf();
+	config &= ~TX39_CONF_WBON;
+	write_c0_conf(config);
+
+	tx39_probe_cache();
+
+	switch (current_cpu_type()) {
+	case CPU_TX3912:
+		/* TX39/H core (writethru direct-map cache) */
+		__flush_cache_vmap	= tx39__flush_cache_vmap;
+		__flush_cache_vunmap	= tx39__flush_cache_vunmap;
+		flush_cache_all	= tx39h_flush_icache_all;
+		__flush_cache_all	= tx39h_flush_icache_all;
+		flush_cache_mm		= (void *) tx39h_flush_icache_all;
+		flush_cache_range	= (void *) tx39h_flush_icache_all;
+		flush_cache_page	= (void *) tx39h_flush_icache_all;
+		flush_icache_range	= (void *) tx39h_flush_icache_all;
+		local_flush_icache_range = (void *) tx39h_flush_icache_all;
+
+		flush_cache_sigtramp	= (void *) tx39h_flush_icache_all;
+		local_flush_data_cache_page	= (void *) tx39h_flush_icache_all;
+		flush_data_cache_page	= (void *) tx39h_flush_icache_all;
+
+		_dma_cache_wback_inv	= tx39h_dma_cache_wback_inv;
+
+		shm_align_mask		= PAGE_SIZE - 1;
+
+		break;
+
+	case CPU_TX3922:
+	case CPU_TX3927:
+	default:
+		/* TX39/H2,H3 core (writeback 2way-set-associative cache) */
+		r3k_have_wired_reg = 1;
+		write_c0_wired(0);	/* set 8 on reset... */
+		/* board-dependent init code may set WBON */
+
+		__flush_cache_vmap	= tx39__flush_cache_vmap;
+		__flush_cache_vunmap	= tx39__flush_cache_vunmap;
+
+		flush_cache_all = tx39_flush_cache_all;
+		__flush_cache_all = tx39___flush_cache_all;
+		flush_cache_mm = tx39_flush_cache_mm;
+		flush_cache_range = tx39_flush_cache_range;
+		flush_cache_page = tx39_flush_cache_page;
+		flush_icache_range = tx39_flush_icache_range;
+		local_flush_icache_range = tx39_flush_icache_range;
+
+		__flush_kernel_vmap_range = tx39_flush_kernel_vmap_range;
+
+		flush_cache_sigtramp = tx39_flush_cache_sigtramp;
+		local_flush_data_cache_page = local_tx39_flush_data_cache_page;
+		flush_data_cache_page = tx39_flush_data_cache_page;
+
+		_dma_cache_wback_inv = tx39_dma_cache_wback_inv;
+		_dma_cache_wback = tx39_dma_cache_wback_inv;
+		_dma_cache_inv = tx39_dma_cache_inv;
+
+		shm_align_mask = max_t(unsigned long,
+		                       (dcache_size / current_cpu_data.dcache.ways) - 1,
+		                       PAGE_SIZE - 1);
+
+		break;
+	}
+
+	current_cpu_data.icache.waysize = icache_size / current_cpu_data.icache.ways;
+	current_cpu_data.dcache.waysize = dcache_size / current_cpu_data.dcache.ways;
+
+	current_cpu_data.icache.sets =
+		current_cpu_data.icache.waysize / current_cpu_data.icache.linesz;
+	current_cpu_data.dcache.sets =
+		current_cpu_data.dcache.waysize / current_cpu_data.dcache.linesz;
+
+	if (current_cpu_data.dcache.waysize > PAGE_SIZE)
+		current_cpu_data.dcache.flags |= MIPS_CACHE_ALIASES;
+
+	current_cpu_data.icache.waybit = 0;
+	current_cpu_data.dcache.waybit = 0;
+
+	printk("Primary instruction cache %ldkB, linesize %d bytes\n",
+		icache_size >> 10, current_cpu_data.icache.linesz);
+	printk("Primary data cache %ldkB, linesize %d bytes\n",
+		dcache_size >> 10, current_cpu_data.dcache.linesz);
+
+	build_clear_page();
+	build_copy_page();
+	tx39h_flush_icache_all();
+}
diff --git a/arch/mips/mm/cache.c b/arch/mips/mm/cache.c
new file mode 100644
index 00000000..829320c7
--- /dev/null
+++ b/arch/mips/mm/cache.c
@@ -0,0 +1,227 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1994 - 2003, 06, 07 by Ralf Baechle (ralf@linux-mips.org)
+ * Copyright (C) 2007 MIPS Technologies, Inc.
+ */
+#include <linux/fs.h>
+#include <linux/fcntl.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/syscalls.h>
+#include <linux/mm.h>
+
+#include <asm/cacheflush.h>
+#include <asm/processor.h>
+#include <asm/cpu.h>
+#include <asm/cpu-features.h>
+
+/* Cache operations. */
+void (*flush_cache_all)(void);
+void (*__flush_cache_all)(void);
+void (*flush_cache_mm)(struct mm_struct *mm);
+void (*flush_cache_range)(struct vm_area_struct *vma, unsigned long start,
+	unsigned long end);
+void (*flush_cache_page)(struct vm_area_struct *vma, unsigned long page,
+	unsigned long pfn);
+void (*flush_icache_range)(unsigned long start, unsigned long end);
+void (*local_flush_icache_range)(unsigned long start, unsigned long end);
+
+void (*__flush_cache_vmap)(void);
+void (*__flush_cache_vunmap)(void);
+
+void (*__flush_kernel_vmap_range)(unsigned long vaddr, int size);
+void (*__invalidate_kernel_vmap_range)(unsigned long vaddr, int size);
+
+EXPORT_SYMBOL_GPL(__flush_kernel_vmap_range);
+
+/* MIPS specific cache operations */
+void (*flush_cache_sigtramp)(unsigned long addr);
+void (*local_flush_data_cache_page)(void * addr);
+void (*flush_data_cache_page)(unsigned long addr);
+void (*flush_icache_all)(void);
+
+EXPORT_SYMBOL_GPL(local_flush_data_cache_page);
+EXPORT_SYMBOL(flush_data_cache_page);
+
+#ifdef CONFIG_DMA_NONCOHERENT
+
+/* DMA cache operations. */
+void (*_dma_cache_wback_inv)(unsigned long start, unsigned long size);
+void (*_dma_cache_wback)(unsigned long start, unsigned long size);
+void (*_dma_cache_inv)(unsigned long start, unsigned long size);
+
+EXPORT_SYMBOL(_dma_cache_wback_inv);
+
+#endif /* CONFIG_DMA_NONCOHERENT */
+
+/*
+ * We could optimize the case where the cache argument is not BCACHE but
+ * that seems very atypical use ...
+ */
+SYSCALL_DEFINE3(cacheflush, unsigned long, addr, unsigned long, bytes,
+	unsigned int, cache)
+{
+	if (bytes == 0)
+		return 0;
+	if (!access_ok(VERIFY_WRITE, (void __user *) addr, bytes))
+		return -EFAULT;
+
+	flush_icache_range(addr, addr + bytes);
+
+	return 0;
+}
+
+void __flush_dcache_page(struct page *page)
+{
+	struct address_space *mapping = page_mapping(page);
+	unsigned long addr;
+
+	if (PageHighMem(page))
+		return;
+	if (mapping && !mapping_mapped(mapping)) {
+		SetPageDcacheDirty(page);
+		return;
+	}
+
+	/*
+	 * We could delay the flush for the !page_mapping case too.  But that
+	 * case is for exec env/arg pages and those are %99 certainly going to
+	 * get faulted into the tlb (and thus flushed) anyways.
+	 */
+	addr = (unsigned long) page_address(page);
+	flush_data_cache_page(addr);
+}
+
+EXPORT_SYMBOL(__flush_dcache_page);
+
+void __flush_anon_page(struct page *page, unsigned long vmaddr)
+{
+	unsigned long addr = (unsigned long) page_address(page);
+
+	if (pages_do_alias(addr, vmaddr)) {
+		if (page_mapped(page) && !Page_dcache_dirty(page)) {
+			void *kaddr;
+
+			kaddr = kmap_coherent(page, vmaddr);
+			flush_data_cache_page((unsigned long)kaddr);
+			kunmap_coherent();
+		} else
+			flush_data_cache_page(addr);
+	}
+}
+
+EXPORT_SYMBOL(__flush_anon_page);
+
+void __update_cache(struct vm_area_struct *vma, unsigned long address,
+	pte_t pte)
+{
+	struct page *page;
+	unsigned long pfn, addr;
+	int exec = (vma->vm_flags & VM_EXEC) && !cpu_has_ic_fills_f_dc;
+
+	pfn = pte_pfn(pte);
+	if (unlikely(!pfn_valid(pfn)))
+		return;
+	page = pfn_to_page(pfn);
+	if (page_mapping(page) && Page_dcache_dirty(page)) {
+		addr = (unsigned long) page_address(page);
+		if (exec || pages_do_alias(addr, address & PAGE_MASK))
+			flush_data_cache_page(addr);
+		ClearPageDcacheDirty(page);
+	}
+}
+
+unsigned long _page_cachable_default;
+EXPORT_SYMBOL(_page_cachable_default);
+
+static inline void setup_protection_map(void)
+{
+	if (kernel_uses_smartmips_rixi) {
+		protection_map[0]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_NO_READ);
+		protection_map[1]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC);
+		protection_map[2]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_NO_READ);
+		protection_map[3]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC);
+		protection_map[4]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_READ);
+		protection_map[5]  = __pgprot(_page_cachable_default | _PAGE_PRESENT);
+		protection_map[6]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_READ);
+		protection_map[7]  = __pgprot(_page_cachable_default | _PAGE_PRESENT);
+
+		protection_map[8]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_NO_READ);
+		protection_map[9]  = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC);
+		protection_map[10] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_WRITE | _PAGE_NO_READ);
+		protection_map[11] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_WRITE);
+		protection_map[12] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_READ);
+		protection_map[13] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
+		protection_map[14] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_WRITE  | _PAGE_NO_READ);
+		protection_map[15] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_WRITE);
+
+	} else {
+		protection_map[0] = PAGE_NONE;
+		protection_map[1] = PAGE_READONLY;
+		protection_map[2] = PAGE_COPY;
+		protection_map[3] = PAGE_COPY;
+		protection_map[4] = PAGE_READONLY;
+		protection_map[5] = PAGE_READONLY;
+		protection_map[6] = PAGE_COPY;
+		protection_map[7] = PAGE_COPY;
+		protection_map[8] = PAGE_NONE;
+		protection_map[9] = PAGE_READONLY;
+		protection_map[10] = PAGE_SHARED;
+		protection_map[11] = PAGE_SHARED;
+		protection_map[12] = PAGE_READONLY;
+		protection_map[13] = PAGE_READONLY;
+		protection_map[14] = PAGE_SHARED;
+		protection_map[15] = PAGE_SHARED;
+	}
+}
+
+void __cpuinit cpu_cache_init(void)
+{
+	if (cpu_has_3k_cache) {
+		extern void __weak r3k_cache_init(void);
+
+		r3k_cache_init();
+	}
+	if (cpu_has_6k_cache) {
+		extern void __weak r6k_cache_init(void);
+
+		r6k_cache_init();
+	}
+	if (cpu_has_4k_cache) {
+		extern void __weak r4k_cache_init(void);
+
+		r4k_cache_init();
+	}
+	if (cpu_has_8k_cache) {
+		extern void __weak r8k_cache_init(void);
+
+		r8k_cache_init();
+	}
+	if (cpu_has_tx39_cache) {
+		extern void __weak tx39_cache_init(void);
+
+		tx39_cache_init();
+	}
+
+	if (cpu_has_octeon_cache) {
+		extern void __weak octeon_cache_init(void);
+
+		octeon_cache_init();
+	}
+
+	setup_protection_map();
+}
+
+int __weak __uncached_access(struct file *file, unsigned long addr)
+{
+	if (file->f_flags & O_DSYNC)
+		return 1;
+
+	return addr >= __pa(high_memory);
+}
diff --git a/arch/mips/mm/cerr-sb1.c b/arch/mips/mm/cerr-sb1.c
new file mode 100644
index 00000000..3571090b
--- /dev/null
+++ b/arch/mips/mm/cerr-sb1.c
@@ -0,0 +1,586 @@
+/*
+ * Copyright (C) 2001,2002,2003 Broadcom Corporation
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
+ */
+#include <linux/sched.h>
+#include <asm/mipsregs.h>
+#include <asm/sibyte/sb1250.h>
+#include <asm/sibyte/sb1250_regs.h>
+
+#if !defined(CONFIG_SIBYTE_BUS_WATCHER) || defined(CONFIG_SIBYTE_BW_TRACE)
+#include <asm/io.h>
+#include <asm/sibyte/sb1250_scd.h>
+#endif
+
+/*
+ * We'd like to dump the L2_ECC_TAG register on errors, but errata make
+ * that unsafe... So for now we don't.  (BCM1250/BCM112x erratum SOC-48.)
+ */
+#undef DUMP_L2_ECC_TAG_ON_ERROR
+
+/* SB1 definitions */
+
+/* XXX should come from config1 XXX */
+#define SB1_CACHE_INDEX_MASK   0x1fe0
+
+#define CP0_ERRCTL_RECOVERABLE (1 << 31)
+#define CP0_ERRCTL_DCACHE      (1 << 30)
+#define CP0_ERRCTL_ICACHE      (1 << 29)
+#define CP0_ERRCTL_MULTIBUS    (1 << 23)
+#define CP0_ERRCTL_MC_TLB      (1 << 15)
+#define CP0_ERRCTL_MC_TIMEOUT  (1 << 14)
+
+#define CP0_CERRI_TAG_PARITY   (1 << 29)
+#define CP0_CERRI_DATA_PARITY  (1 << 28)
+#define CP0_CERRI_EXTERNAL     (1 << 26)
+
+#define CP0_CERRI_IDX_VALID(c) (!((c) & CP0_CERRI_EXTERNAL))
+#define CP0_CERRI_DATA         (CP0_CERRI_DATA_PARITY)
+
+#define CP0_CERRD_MULTIPLE     (1 << 31)
+#define CP0_CERRD_TAG_STATE    (1 << 30)
+#define CP0_CERRD_TAG_ADDRESS  (1 << 29)
+#define CP0_CERRD_DATA_SBE     (1 << 28)
+#define CP0_CERRD_DATA_DBE     (1 << 27)
+#define CP0_CERRD_EXTERNAL     (1 << 26)
+#define CP0_CERRD_LOAD         (1 << 25)
+#define CP0_CERRD_STORE        (1 << 24)
+#define CP0_CERRD_FILLWB       (1 << 23)
+#define CP0_CERRD_COHERENCY    (1 << 22)
+#define CP0_CERRD_DUPTAG       (1 << 21)
+
+#define CP0_CERRD_DPA_VALID(c) (!((c) & CP0_CERRD_EXTERNAL))
+#define CP0_CERRD_IDX_VALID(c) \
+   (((c) & (CP0_CERRD_LOAD | CP0_CERRD_STORE)) ? (!((c) & CP0_CERRD_EXTERNAL)) : 0)
+#define CP0_CERRD_CAUSES \
+   (CP0_CERRD_LOAD | CP0_CERRD_STORE | CP0_CERRD_FILLWB | CP0_CERRD_COHERENCY | CP0_CERRD_DUPTAG)
+#define CP0_CERRD_TYPES \
+   (CP0_CERRD_TAG_STATE | CP0_CERRD_TAG_ADDRESS | CP0_CERRD_DATA_SBE | CP0_CERRD_DATA_DBE | CP0_CERRD_EXTERNAL)
+#define CP0_CERRD_DATA         (CP0_CERRD_DATA_SBE | CP0_CERRD_DATA_DBE)
+
+static uint32_t	extract_ic(unsigned short addr, int data);
+static uint32_t	extract_dc(unsigned short addr, int data);
+
+static inline void breakout_errctl(unsigned int val)
+{
+	if (val & CP0_ERRCTL_RECOVERABLE)
+		printk(" recoverable");
+	if (val & CP0_ERRCTL_DCACHE)
+		printk(" dcache");
+	if (val & CP0_ERRCTL_ICACHE)
+		printk(" icache");
+	if (val & CP0_ERRCTL_MULTIBUS)
+		printk(" multiple-buserr");
+	printk("\n");
+}
+
+static inline void breakout_cerri(unsigned int val)
+{
+	if (val & CP0_CERRI_TAG_PARITY)
+		printk(" tag-parity");
+	if (val & CP0_CERRI_DATA_PARITY)
+		printk(" data-parity");
+	if (val & CP0_CERRI_EXTERNAL)
+		printk(" external");
+	printk("\n");
+}
+
+static inline void breakout_cerrd(unsigned int val)
+{
+	switch (val & CP0_CERRD_CAUSES) {
+	case CP0_CERRD_LOAD:
+		printk(" load,");
+		break;
+	case CP0_CERRD_STORE:
+		printk(" store,");
+		break;
+	case CP0_CERRD_FILLWB:
+		printk(" fill/wb,");
+		break;
+	case CP0_CERRD_COHERENCY:
+		printk(" coherency,");
+		break;
+	case CP0_CERRD_DUPTAG:
+		printk(" duptags,");
+		break;
+	default:
+		printk(" NO CAUSE,");
+		break;
+	}
+	if (!(val & CP0_CERRD_TYPES))
+		printk(" NO TYPE");
+	else {
+		if (val & CP0_CERRD_MULTIPLE)
+			printk(" multi-err");
+		if (val & CP0_CERRD_TAG_STATE)
+			printk(" tag-state");
+		if (val & CP0_CERRD_TAG_ADDRESS)
+			printk(" tag-address");
+		if (val & CP0_CERRD_DATA_SBE)
+			printk(" data-SBE");
+		if (val & CP0_CERRD_DATA_DBE)
+			printk(" data-DBE");
+		if (val & CP0_CERRD_EXTERNAL)
+			printk(" external");
+	}
+	printk("\n");
+}
+
+#ifndef CONFIG_SIBYTE_BUS_WATCHER
+
+static void check_bus_watcher(void)
+{
+	uint32_t status, l2_err, memio_err;
+#ifdef DUMP_L2_ECC_TAG_ON_ERROR
+	uint64_t l2_tag;
+#endif
+
+	/* Destructive read, clears register and interrupt */
+	status = csr_in32(IOADDR(A_SCD_BUS_ERR_STATUS));
+	/* Bit 31 is always on, but there's no #define for that */
+	if (status & ~(1UL << 31)) {
+		l2_err = csr_in32(IOADDR(A_BUS_L2_ERRORS));
+#ifdef DUMP_L2_ECC_TAG_ON_ERROR
+		l2_tag = in64(IOADDR(A_L2_ECC_TAG));
+#endif
+		memio_err = csr_in32(IOADDR(A_BUS_MEM_IO_ERRORS));
+		printk("Bus watcher error counters: %08x %08x\n", l2_err, memio_err);
+		printk("\nLast recorded signature:\n");
+		printk("Request %02x from %d, answered by %d with Dcode %d\n",
+		       (unsigned int)(G_SCD_BERR_TID(status) & 0x3f),
+		       (int)(G_SCD_BERR_TID(status) >> 6),
+		       (int)G_SCD_BERR_RID(status),
+		       (int)G_SCD_BERR_DCODE(status));
+#ifdef DUMP_L2_ECC_TAG_ON_ERROR
+		printk("Last L2 tag w/ bad ECC: %016llx\n", l2_tag);
+#endif
+	} else {
+		printk("Bus watcher indicates no error\n");
+	}
+}
+#else
+extern void check_bus_watcher(void);
+#endif
+
+asmlinkage void sb1_cache_error(void)
+{
+	uint32_t errctl, cerr_i, cerr_d, dpalo, dpahi, eepc, res;
+	unsigned long long cerr_dpa;
+
+#ifdef CONFIG_SIBYTE_BW_TRACE
+	/* Freeze the trace buffer now */
+#if defined(CONFIG_SIBYTE_BCM1x55) || defined(CONFIG_SIBYTE_BCM1x80)
+	csr_out32(M_BCM1480_SCD_TRACE_CFG_FREEZE, IOADDR(A_SCD_TRACE_CFG));
+#else
+	csr_out32(M_SCD_TRACE_CFG_FREEZE, IOADDR(A_SCD_TRACE_CFG));
+#endif
+	printk("Trace buffer frozen\n");
+#endif
+
+	printk("Cache error exception on CPU %x:\n",
+	       (read_c0_prid() >> 25) & 0x7);
+
+	__asm__ __volatile__ (
+	"	.set	push\n\t"
+	"	.set	mips64\n\t"
+	"	.set	noat\n\t"
+	"	mfc0	%0, $26\n\t"
+	"	mfc0	%1, $27\n\t"
+	"	mfc0	%2, $27, 1\n\t"
+	"	dmfc0	$1, $27, 3\n\t"
+	"	dsrl32	%3, $1, 0 \n\t"
+	"	sll	%4, $1, 0 \n\t"
+	"	mfc0	%5, $30\n\t"
+	"	.set	pop"
+	: "=r" (errctl), "=r" (cerr_i), "=r" (cerr_d),
+	  "=r" (dpahi), "=r" (dpalo), "=r" (eepc));
+
+	cerr_dpa = (((uint64_t)dpahi) << 32) | dpalo;
+	printk(" c0_errorepc ==   %08x\n", eepc);
+	printk(" c0_errctl   ==   %08x", errctl);
+	breakout_errctl(errctl);
+	if (errctl & CP0_ERRCTL_ICACHE) {
+		printk(" c0_cerr_i   ==   %08x", cerr_i);
+		breakout_cerri(cerr_i);
+		if (CP0_CERRI_IDX_VALID(cerr_i)) {
+			/* Check index of EPC, allowing for delay slot */
+			if (((eepc & SB1_CACHE_INDEX_MASK) != (cerr_i & SB1_CACHE_INDEX_MASK)) &&
+			    ((eepc & SB1_CACHE_INDEX_MASK) != ((cerr_i & SB1_CACHE_INDEX_MASK) - 4)))
+				printk(" cerr_i idx doesn't match eepc\n");
+			else {
+				res = extract_ic(cerr_i & SB1_CACHE_INDEX_MASK,
+						 (cerr_i & CP0_CERRI_DATA) != 0);
+				if (!(res & cerr_i))
+					printk("...didn't see indicated icache problem\n");
+			}
+		}
+	}
+	if (errctl & CP0_ERRCTL_DCACHE) {
+		printk(" c0_cerr_d   ==   %08x", cerr_d);
+		breakout_cerrd(cerr_d);
+		if (CP0_CERRD_DPA_VALID(cerr_d)) {
+			printk(" c0_cerr_dpa == %010llx\n", cerr_dpa);
+			if (!CP0_CERRD_IDX_VALID(cerr_d)) {
+				res = extract_dc(cerr_dpa & SB1_CACHE_INDEX_MASK,
+						 (cerr_d & CP0_CERRD_DATA) != 0);
+				if (!(res & cerr_d))
+					printk("...didn't see indicated dcache problem\n");
+			} else {
+				if ((cerr_dpa & SB1_CACHE_INDEX_MASK) != (cerr_d & SB1_CACHE_INDEX_MASK))
+					printk(" cerr_d idx doesn't match cerr_dpa\n");
+				else {
+					res = extract_dc(cerr_d & SB1_CACHE_INDEX_MASK,
+							 (cerr_d & CP0_CERRD_DATA) != 0);
+					if (!(res & cerr_d))
+						printk("...didn't see indicated problem\n");
+				}
+			}
+		}
+	}
+
+	check_bus_watcher();
+
+	/*
+	 * Calling panic() when a fatal cache error occurs scrambles the
+	 * state of the system (and the cache), making it difficult to
+	 * investigate after the fact.  However, if you just stall the CPU,
+	 * the other CPU may keep on running, which is typically very
+	 * undesirable.
+	 */
+#ifdef CONFIG_SB1_CERR_STALL
+	while (1)
+		;
+#else
+	panic("unhandled cache error");
+#endif
+}
+
+
+/* Parity lookup table. */
+static const uint8_t parity[256] = {
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
+	0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0
+};
+
+/* Masks to select bits for Hamming parity, mask_72_64[i] for bit[i] */
+static const uint64_t mask_72_64[8] = {
+	0x0738C808099264FFULL,
+	0x38C808099264FF07ULL,
+	0xC808099264FF0738ULL,
+	0x08099264FF0738C8ULL,
+	0x099264FF0738C808ULL,
+	0x9264FF0738C80809ULL,
+	0x64FF0738C8080992ULL,
+	0xFF0738C808099264ULL
+};
+
+/* Calculate the parity on a range of bits */
+static char range_parity(uint64_t dword, int max, int min)
+{
+	char parity = 0;
+	int i;
+	dword >>= min;
+	for (i=max-min; i>=0; i--) {
+		if (dword & 0x1)
+			parity = !parity;
+		dword >>= 1;
+	}
+	return parity;
+}
+
+/* Calculate the 4-bit even byte-parity for an instruction */
+static unsigned char inst_parity(uint32_t word)
+{
+	int i, j;
+	char parity = 0;
+	for (j=0; j<4; j++) {
+		char byte_parity = 0;
+		for (i=0; i<8; i++) {
+			if (word & 0x80000000)
+				byte_parity = !byte_parity;
+			word <<= 1;
+		}
+		parity <<= 1;
+		parity |= byte_parity;
+	}
+	return parity;
+}
+
+static uint32_t extract_ic(unsigned short addr, int data)
+{
+	unsigned short way;
+	int valid;
+	uint32_t taghi, taglolo, taglohi;
+	unsigned long long taglo, va;
+	uint64_t tlo_tmp;
+	uint8_t lru;
+	int res = 0;
+
+	printk("Icache index 0x%04x  ", addr);
+	for (way = 0; way < 4; way++) {
+		/* Index-load-tag-I */
+		__asm__ __volatile__ (
+		"	.set	push		\n\t"
+		"	.set	noreorder	\n\t"
+		"	.set	mips64		\n\t"
+		"	.set	noat		\n\t"
+		"	cache	4, 0(%3)	\n\t"
+		"	mfc0	%0, $29		\n\t"
+		"	dmfc0	$1, $28		\n\t"
+		"	dsrl32	%1, $1, 0	\n\t"
+		"	sll	%2, $1, 0	\n\t"
+		"	.set	pop"
+		: "=r" (taghi), "=r" (taglohi), "=r" (taglolo)
+		: "r" ((way << 13) | addr));
+
+		taglo = ((unsigned long long)taglohi << 32) | taglolo;
+		if (way == 0) {
+			lru = (taghi >> 14) & 0xff;
+			printk("[Bank %d Set 0x%02x]  LRU > %d %d %d %d > MRU\n",
+				    ((addr >> 5) & 0x3), /* bank */
+				    ((addr >> 7) & 0x3f), /* index */
+				    (lru & 0x3),
+				    ((lru >> 2) & 0x3),
+				    ((lru >> 4) & 0x3),
+				    ((lru >> 6) & 0x3));
+		}
+		va = (taglo & 0xC0000FFFFFFFE000ULL) | addr;
+		if ((taglo & (1 << 31)) && (((taglo >> 62) & 0x3) == 3))
+			va |= 0x3FFFF00000000000ULL;
+		valid = ((taghi >> 29) & 1);
+		if (valid) {
+			tlo_tmp = taglo & 0xfff3ff;
+			if (((taglo >> 10) & 1) ^ range_parity(tlo_tmp, 23, 0)) {
+				printk("   ** bad parity in VTag0/G/ASID\n");
+				res |= CP0_CERRI_TAG_PARITY;
+			}
+			if (((taglo >> 11) & 1) ^ range_parity(taglo, 63, 24)) {
+				printk("   ** bad parity in R/VTag1\n");
+				res |= CP0_CERRI_TAG_PARITY;
+			}
+		}
+		if (valid ^ ((taghi >> 27) & 1)) {
+			printk("   ** bad parity for valid bit\n");
+			res |= CP0_CERRI_TAG_PARITY;
+		}
+		printk(" %d  [VA %016llx]  [Vld? %d]  raw tags: %08X-%016llX\n",
+			    way, va, valid, taghi, taglo);
+
+		if (data) {
+			uint32_t datahi, insta, instb;
+			uint8_t predecode;
+			int offset;
+
+			/* (hit all banks and ways) */
+			for (offset = 0; offset < 4; offset++) {
+				/* Index-load-data-I */
+				__asm__ __volatile__ (
+				"	.set	push\n\t"
+				"	.set	noreorder\n\t"
+				"	.set	mips64\n\t"
+				"	.set	noat\n\t"
+				"	cache	6, 0(%3)  \n\t"
+				"	mfc0	%0, $29, 1\n\t"
+				"	dmfc0  $1, $28, 1\n\t"
+				"	dsrl32 %1, $1, 0 \n\t"
+				"	sll    %2, $1, 0 \n\t"
+				"	.set	pop         \n"
+				: "=r" (datahi), "=r" (insta), "=r" (instb)
+				: "r" ((way << 13) | addr | (offset << 3)));
+				predecode = (datahi >> 8) & 0xff;
+				if (((datahi >> 16) & 1) != (uint32_t)range_parity(predecode, 7, 0)) {
+					printk("   ** bad parity in predecode\n");
+					res |= CP0_CERRI_DATA_PARITY;
+				}
+				/* XXXKW should/could check predecode bits themselves */
+				if (((datahi >> 4) & 0xf) ^ inst_parity(insta)) {
+					printk("   ** bad parity in instruction a\n");
+					res |= CP0_CERRI_DATA_PARITY;
+				}
+				if ((datahi & 0xf) ^ inst_parity(instb)) {
+					printk("   ** bad parity in instruction b\n");
+					res |= CP0_CERRI_DATA_PARITY;
+				}
+				printk("  %05X-%08X%08X", datahi, insta, instb);
+			}
+			printk("\n");
+		}
+	}
+	return res;
+}
+
+/* Compute the ECC for a data doubleword */
+static uint8_t dc_ecc(uint64_t dword)
+{
+	uint64_t t;
+	uint32_t w;
+	uint8_t  p;
+	int      i;
+
+	p = 0;
+	for (i = 7; i >= 0; i--)
+	{
+		p <<= 1;
+		t = dword & mask_72_64[i];
+		w = (uint32_t)(t >> 32);
+		p ^= (parity[w>>24] ^ parity[(w>>16) & 0xFF]
+		      ^ parity[(w>>8) & 0xFF] ^ parity[w & 0xFF]);
+		w = (uint32_t)(t & 0xFFFFFFFF);
+		p ^= (parity[w>>24] ^ parity[(w>>16) & 0xFF]
+		      ^ parity[(w>>8) & 0xFF] ^ parity[w & 0xFF]);
+	}
+	return p;
+}
+
+struct dc_state {
+	unsigned char val;
+	char *name;
+};
+
+static struct dc_state dc_states[] = {
+	{ 0x00, "INVALID" },
+	{ 0x0f, "COH-SHD" },
+	{ 0x13, "NCO-E-C" },
+	{ 0x19, "NCO-E-D" },
+	{ 0x16, "COH-E-C" },
+	{ 0x1c, "COH-E-D" },
+	{ 0xff, "*ERROR*" }
+};
+
+#define DC_TAG_VALID(state) \
+    (((state) == 0x0) || ((state) == 0xf) || ((state) == 0x13) || \
+     ((state) == 0x19) || ((state) == 0x16) || ((state) == 0x1c))
+
+static char *dc_state_str(unsigned char state)
+{
+	struct dc_state *dsc = dc_states;
+	while (dsc->val != 0xff) {
+		if (dsc->val == state)
+			break;
+		dsc++;
+	}
+	return dsc->name;
+}
+
+static uint32_t extract_dc(unsigned short addr, int data)
+{
+	int valid, way;
+	unsigned char state;
+	uint32_t taghi, taglolo, taglohi;
+	unsigned long long taglo, pa;
+	uint8_t ecc, lru;
+	int res = 0;
+
+	printk("Dcache index 0x%04x  ", addr);
+	for (way = 0; way < 4; way++) {
+		__asm__ __volatile__ (
+		"	.set	push\n\t"
+		"	.set	noreorder\n\t"
+		"	.set	mips64\n\t"
+		"	.set	noat\n\t"
+		"	cache	5, 0(%3)\n\t"	/* Index-load-tag-D */
+		"	mfc0	%0, $29, 2\n\t"
+		"	dmfc0	$1, $28, 2\n\t"
+		"	dsrl32	%1, $1, 0\n\t"
+		"	sll	%2, $1, 0\n\t"
+		"	.set	pop"
+		: "=r" (taghi), "=r" (taglohi), "=r" (taglolo)
+		: "r" ((way << 13) | addr));
+
+		taglo = ((unsigned long long)taglohi << 32) | taglolo;
+		pa = (taglo & 0xFFFFFFE000ULL) | addr;
+		if (way == 0) {
+			lru = (taghi >> 14) & 0xff;
+			printk("[Bank %d Set 0x%02x]  LRU > %d %d %d %d > MRU\n",
+				    ((addr >> 11) & 0x2) | ((addr >> 5) & 1), /* bank */
+				    ((addr >> 6) & 0x3f), /* index */
+				    (lru & 0x3),
+				    ((lru >> 2) & 0x3),
+				    ((lru >> 4) & 0x3),
+				    ((lru >> 6) & 0x3));
+		}
+		state = (taghi >> 25) & 0x1f;
+		valid = DC_TAG_VALID(state);
+		printk(" %d  [PA %010llx]  [state %s (%02x)]  raw tags: %08X-%016llX\n",
+			    way, pa, dc_state_str(state), state, taghi, taglo);
+		if (valid) {
+			if (((taglo >> 11) & 1) ^ range_parity(taglo, 39, 26)) {
+				printk("   ** bad parity in PTag1\n");
+				res |= CP0_CERRD_TAG_ADDRESS;
+			}
+			if (((taglo >> 10) & 1) ^ range_parity(taglo, 25, 13)) {
+				printk("   ** bad parity in PTag0\n");
+				res |= CP0_CERRD_TAG_ADDRESS;
+			}
+		} else {
+			res |= CP0_CERRD_TAG_STATE;
+		}
+
+		if (data) {
+			uint32_t datalohi, datalolo, datahi;
+			unsigned long long datalo;
+			int offset;
+			char bad_ecc = 0;
+
+			for (offset = 0; offset < 4; offset++) {
+				/* Index-load-data-D */
+				__asm__ __volatile__ (
+				"	.set	push\n\t"
+				"	.set	noreorder\n\t"
+				"	.set	mips64\n\t"
+				"	.set	noat\n\t"
+				"	cache	7, 0(%3)\n\t" /* Index-load-data-D */
+				"	mfc0	%0, $29, 3\n\t"
+				"	dmfc0	$1, $28, 3\n\t"
+				"	dsrl32	%1, $1, 0 \n\t"
+				"	sll	%2, $1, 0 \n\t"
+				"	.set	pop"
+				: "=r" (datahi), "=r" (datalohi), "=r" (datalolo)
+				: "r" ((way << 13) | addr | (offset << 3)));
+				datalo = ((unsigned long long)datalohi << 32) | datalolo;
+				ecc = dc_ecc(datalo);
+				if (ecc != datahi) {
+					int bits;
+					bad_ecc |= 1 << (3-offset);
+					ecc ^= datahi;
+					bits = hweight8(ecc);
+					res |= (bits == 1) ? CP0_CERRD_DATA_SBE : CP0_CERRD_DATA_DBE;
+				}
+				printk("  %02X-%016llX", datahi, datalo);
+			}
+			printk("\n");
+			if (bad_ecc)
+				printk("  dwords w/ bad ECC: %d %d %d %d\n",
+				       !!(bad_ecc & 8), !!(bad_ecc & 4),
+				       !!(bad_ecc & 2), !!(bad_ecc & 1));
+		}
+	}
+	return res;
+}
diff --git a/arch/mips/mm/cex-gen.S b/arch/mips/mm/cex-gen.S
new file mode 100644
index 00000000..e743622f
--- /dev/null
+++ b/arch/mips/mm/cex-gen.S
@@ -0,0 +1,42 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1995 - 1999 Ralf Baechle
+ * Copyright (C) 1999 Silicon Graphics, Inc.
+ *
+ * Cache error handler
+ */
+#include <asm/asm.h>
+#include <asm/regdef.h>
+#include <asm/mipsregs.h>
+#include <asm/stackframe.h>
+
+/*
+ * Game over.  Go to the button.  Press gently.  Swear where allowed by
+ * legislation.
+ */
+	LEAF(except_vec2_generic)
+	.set	noreorder
+	.set	noat
+	.set    mips0
+	/*
+	 * This is a very bad place to be.  Our cache error
+	 * detection has triggered.  If we have write-back data
+	 * in the cache, we may not be able to recover.  As a
+	 * first-order desperate measure, turn off KSEG0 cacheing.
+	 */
+	mfc0	k0,CP0_CONFIG
+	li	k1,~CONF_CM_CMASK
+	and	k0,k0,k1
+	ori	k0,k0,CONF_CM_UNCACHED
+	mtc0	k0,CP0_CONFIG
+	/* Give it a few cycles to sink in... */
+	nop
+	nop
+	nop
+
+	j	cache_parity_error
+	nop
+	END(except_vec2_generic)
diff --git a/arch/mips/mm/cex-oct.S b/arch/mips/mm/cex-oct.S
new file mode 100644
index 00000000..3db8553f
--- /dev/null
+++ b/arch/mips/mm/cex-oct.S
@@ -0,0 +1,70 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2006 Cavium Networks
+ * Cache error handler
+ */
+
+#include <asm/asm.h>
+#include <asm/regdef.h>
+#include <asm/mipsregs.h>
+#include <asm/stackframe.h>
+
+/*
+ * Handle cache error. Indicate to the second level handler whether
+ * the exception is recoverable.
+ */
+	LEAF(except_vec2_octeon)
+
+	.set    push
+	.set	mips64r2
+	.set	noreorder
+	.set	noat
+
+
+	/* due to an errata we need to read the COP0 CacheErr (Dcache)
+	 * before any cache/DRAM access	 */
+
+	rdhwr   k0, $0        /* get core_id */
+	PTR_LA  k1, cache_err_dcache
+	sll     k0, k0, 3
+	PTR_ADDU k1, k0, k1    /* k1 = &cache_err_dcache[core_id] */
+
+	dmfc0   k0, CP0_CACHEERR, 1
+	sd      k0, (k1)
+	dmtc0   $0, CP0_CACHEERR, 1
+
+        /* check whether this is a nested exception */
+	mfc0    k1, CP0_STATUS
+	andi    k1, k1, ST0_EXL
+	beqz    k1, 1f
+	 nop
+	j	cache_parity_error_octeon_non_recoverable
+	 nop
+
+	/* exception is recoverable */
+1:	j	handle_cache_err
+	 nop
+
+	.set    pop
+	END(except_vec2_octeon)
+
+ /* We need to jump to handle_cache_err so that the previous handler
+  * can fit within 0x80 bytes. We also move from 0xFFFFFFFFAXXXXXXX
+  * space (uncached) to the 0xFFFFFFFF8XXXXXXX space (cached).  */
+	LEAF(handle_cache_err)
+	.set    push
+        .set    noreorder
+        .set    noat
+
+	SAVE_ALL
+	KMODE
+	jal     cache_parity_error_octeon_recoverable
+	nop
+	j       ret_from_exception
+	nop
+
+	.set pop
+	END(handle_cache_err)
diff --git a/arch/mips/mm/cex-sb1.S b/arch/mips/mm/cex-sb1.S
new file mode 100644
index 00000000..89c412bc
--- /dev/null
+++ b/arch/mips/mm/cex-sb1.S
@@ -0,0 +1,175 @@
+/*
+ * Copyright (C) 2001,2002,2003 Broadcom Corporation
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
+ */
+#include <linux/init.h>
+
+#include <asm/asm.h>
+#include <asm/regdef.h>
+#include <asm/mipsregs.h>
+#include <asm/stackframe.h>
+#include <asm/cacheops.h>
+#include <asm/sibyte/board.h>
+
+#define C0_ERRCTL     $26             /* CP0: Error info */
+#define C0_CERR_I     $27             /* CP0: Icache error */
+#define C0_CERR_D     $27,1           /* CP0: Dcache error */
+
+	/*
+	 * Based on SiByte sample software cache-err/cerr.S
+	 * CVS revision 1.8.  Only the 'unrecoverable' case
+	 * is changed.
+	 */
+
+	.set	mips64
+	.set	noreorder
+	.set	noat
+
+	/*
+	 * sb1_cerr_vec: code to be copied to the Cache Error
+	 * Exception vector.  The code must be pushed out to memory
+	 * (either by copying to Kseg0 and Kseg1 both, or by flushing
+	 * the L1 and L2) since it is fetched as 0xa0000100.
+	 *
+	 * NOTE: Be sure this handler is at most 28 instructions long
+	 * since the final 16 bytes of the exception vector memory
+	 * (0x170-0x17f) are used to preserve k0, k1, and ra.
+	 */
+
+	__CPUINIT
+
+LEAF(except_vec2_sb1)
+	/*
+	 * If this error is recoverable, we need to exit the handler
+	 * without having dirtied any registers.  To do this,
+	 * save/restore k0 and k1 from low memory (Useg is direct
+	 * mapped while ERL=1). Note that we can't save to a
+	 * CPU-specific location without ruining a register in the
+	 * process.  This means we are vulnerable to data corruption
+	 * whenever the handler is reentered by a second CPU.
+	 */
+	sd	k0,0x170($0)
+	sd	k1,0x178($0)
+
+#ifdef CONFIG_SB1_CEX_ALWAYS_FATAL
+	j	handle_vec2_sb1
+	 nop
+#else
+	/*
+	 * M_ERRCTL_RECOVERABLE is bit 31, which makes it easy to tell
+	 * if we can fast-path out of here for a h/w-recovered error.
+	 */
+	mfc0	k1,C0_ERRCTL
+	bgtz	k1,attempt_recovery
+	 sll	k0,k1,1
+
+recovered_dcache:
+	/*
+	 * Unlock CacheErr-D (which in turn unlocks CacheErr-DPA).
+	 * Ought to log the occurrence of this recovered dcache error.
+	 */
+	b	recovered
+	 mtc0	$0,C0_CERR_D
+
+attempt_recovery:
+	/*
+	 * k0 has C0_ERRCTL << 1, which puts 'DC' at bit 31.  Any
+	 * Dcache errors we can recover from will take more extensive
+	 * processing.  For now, they are considered "unrecoverable".
+	 * Note that 'DC' becoming set (outside of ERL mode) will
+	 * cause 'IC' to clear; so if there's an Icache error, we'll
+	 * only find out about it if we recover from this error and
+	 * continue executing.
+	 */
+	bltz	k0,unrecoverable
+	 sll	k0,1
+
+	/*
+	 * k0 has C0_ERRCTL << 2, which puts 'IC' at bit 31.  If an
+	 * Icache error isn't indicated, I'm not sure why we got here.
+	 * Consider that case "unrecoverable" for now.
+	 */
+	bgez	k0,unrecoverable
+
+attempt_icache_recovery:
+	/*
+	 * External icache errors are due to uncorrectable ECC errors
+	 * in the L2 cache or Memory Controller and cannot be
+	 * recovered here.
+	 */
+	 mfc0	k0,C0_CERR_I		/* delay slot */
+	li	k1,1 << 26		/* ICACHE_EXTERNAL */
+	and	k1,k0
+	bnez	k1,unrecoverable
+	 andi	k0,0x1fe0
+
+	/*
+	 * Since the error is internal, the 'IDX' field from
+	 * CacheErr-I is valid and we can just invalidate all blocks
+	 * in that set.
+	 */
+	cache	Index_Invalidate_I,(0<<13)(k0)
+	cache	Index_Invalidate_I,(1<<13)(k0)
+	cache	Index_Invalidate_I,(2<<13)(k0)
+	cache	Index_Invalidate_I,(3<<13)(k0)
+
+	/* Ought to log this recovered icache error */
+
+recovered:
+	/* Restore the saved registers */
+	ld	k0,0x170($0)
+	ld	k1,0x178($0)
+	eret
+
+unrecoverable:
+	/* Unrecoverable Icache or Dcache error; log it and/or fail */
+	j	handle_vec2_sb1
+	 nop
+#endif
+
+END(except_vec2_sb1)
+
+	__FINIT
+
+	LEAF(handle_vec2_sb1)
+	mfc0	k0,CP0_CONFIG
+	li	k1,~CONF_CM_CMASK
+	and	k0,k0,k1
+	ori	k0,k0,CONF_CM_UNCACHED
+	mtc0	k0,CP0_CONFIG
+
+	SSNOP
+	SSNOP
+	SSNOP
+	SSNOP
+	bnezl	$0, 1f
+1:
+	mfc0	k0, CP0_STATUS
+	sll	k0, k0, 3			# check CU0 (kernel?)
+	bltz	k0, 2f
+	 nop
+
+	/* Get a valid Kseg0 stack pointer.  Any task's stack pointer
+	 * will do, although if we ever want to resume execution we
+	 * better not have corrupted any state. */
+	get_saved_sp
+	move	sp, k1
+
+2:
+	j	sb1_cache_error
+	 nop
+
+	END(handle_vec2_sb1)
diff --git a/arch/mips/mm/dma-default.c b/arch/mips/mm/dma-default.c
new file mode 100644
index 00000000..3fab2046
--- /dev/null
+++ b/arch/mips/mm/dma-default.c
@@ -0,0 +1,351 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2000  Ani Joshi <ajoshi@unixbox.com>
+ * Copyright (C) 2000, 2001, 06  Ralf Baechle <ralf@linux-mips.org>
+ * swiped from i386, and cloned for MIPS by Geert, polished by Ralf.
+ */
+
+#include <linux/types.h>
+#include <linux/dma-mapping.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/scatterlist.h>
+#include <linux/string.h>
+#include <linux/gfp.h>
+#include <linux/highmem.h>
+
+#include <asm/cache.h>
+#include <asm/io.h>
+
+#include <dma-coherence.h>
+
+static inline struct page *dma_addr_to_page(struct device *dev,
+	dma_addr_t dma_addr)
+{
+	return pfn_to_page(
+		plat_dma_addr_to_phys(dev, dma_addr) >> PAGE_SHIFT);
+}
+
+/*
+ * Warning on the terminology - Linux calls an uncached area coherent;
+ * MIPS terminology calls memory areas with hardware maintained coherency
+ * coherent.
+ */
+
+static inline int cpu_is_noncoherent_r10000(struct device *dev)
+{
+	return !plat_device_is_coherent(dev) &&
+	       (current_cpu_type() == CPU_R10000 ||
+	       current_cpu_type() == CPU_R12000);
+}
+
+static gfp_t massage_gfp_flags(const struct device *dev, gfp_t gfp)
+{
+	gfp_t dma_flag;
+
+	/* ignore region specifiers */
+	gfp &= ~(__GFP_DMA | __GFP_DMA32 | __GFP_HIGHMEM);
+
+#ifdef CONFIG_ISA
+	if (dev == NULL)
+		dma_flag = __GFP_DMA;
+	else
+#endif
+#if defined(CONFIG_ZONE_DMA32) && defined(CONFIG_ZONE_DMA)
+	     if (dev->coherent_dma_mask < DMA_BIT_MASK(32))
+			dma_flag = __GFP_DMA;
+	else if (dev->coherent_dma_mask < DMA_BIT_MASK(64))
+			dma_flag = __GFP_DMA32;
+	else
+#endif
+#if defined(CONFIG_ZONE_DMA32) && !defined(CONFIG_ZONE_DMA)
+	     if (dev->coherent_dma_mask < DMA_BIT_MASK(64))
+		dma_flag = __GFP_DMA32;
+	else
+#endif
+#if defined(CONFIG_ZONE_DMA) && !defined(CONFIG_ZONE_DMA32)
+	     if (dev->coherent_dma_mask < DMA_BIT_MASK(64))
+		dma_flag = __GFP_DMA;
+	else
+#endif
+		dma_flag = 0;
+
+	/* Don't invoke OOM killer */
+	gfp |= __GFP_NORETRY;
+
+	return gfp | dma_flag;
+}
+
+void *dma_alloc_noncoherent(struct device *dev, size_t size,
+	dma_addr_t * dma_handle, gfp_t gfp)
+{
+	void *ret;
+
+	gfp = massage_gfp_flags(dev, gfp);
+
+	ret = (void *) __get_free_pages(gfp, get_order(size));
+
+	if (ret != NULL) {
+		memset(ret, 0, size);
+		*dma_handle = plat_map_dma_mem(dev, ret, size);
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(dma_alloc_noncoherent);
+
+static void *mips_dma_alloc_coherent(struct device *dev, size_t size,
+	dma_addr_t * dma_handle, gfp_t gfp, struct dma_attrs *attrs)
+{
+	void *ret;
+
+	if (dma_alloc_from_coherent(dev, size, dma_handle, &ret))
+		return ret;
+
+	gfp = massage_gfp_flags(dev, gfp);
+
+	ret = (void *) __get_free_pages(gfp, get_order(size));
+
+	if (ret) {
+		memset(ret, 0, size);
+		*dma_handle = plat_map_dma_mem(dev, ret, size);
+
+		if (!plat_device_is_coherent(dev)) {
+			dma_cache_wback_inv((unsigned long) ret, size);
+			ret = UNCAC_ADDR(ret);
+		}
+	}
+
+	return ret;
+}
+
+
+void dma_free_noncoherent(struct device *dev, size_t size, void *vaddr,
+	dma_addr_t dma_handle)
+{
+	plat_unmap_dma_mem(dev, dma_handle, size, DMA_BIDIRECTIONAL);
+	free_pages((unsigned long) vaddr, get_order(size));
+}
+EXPORT_SYMBOL(dma_free_noncoherent);
+
+static void mips_dma_free_coherent(struct device *dev, size_t size, void *vaddr,
+	dma_addr_t dma_handle, struct dma_attrs *attrs)
+{
+	unsigned long addr = (unsigned long) vaddr;
+	int order = get_order(size);
+
+	if (dma_release_from_coherent(dev, order, vaddr))
+		return;
+
+	plat_unmap_dma_mem(dev, dma_handle, size, DMA_BIDIRECTIONAL);
+
+	if (!plat_device_is_coherent(dev))
+		addr = CAC_ADDR(addr);
+
+	free_pages(addr, get_order(size));
+}
+
+static inline void __dma_sync_virtual(void *addr, size_t size,
+	enum dma_data_direction direction)
+{
+	switch (direction) {
+	case DMA_TO_DEVICE:
+		dma_cache_wback((unsigned long)addr, size);
+		break;
+
+	case DMA_FROM_DEVICE:
+		dma_cache_inv((unsigned long)addr, size);
+		break;
+
+	case DMA_BIDIRECTIONAL:
+		dma_cache_wback_inv((unsigned long)addr, size);
+		break;
+
+	default:
+		BUG();
+	}
+}
+
+/*
+ * A single sg entry may refer to multiple physically contiguous
+ * pages. But we still need to process highmem pages individually.
+ * If highmem is not configured then the bulk of this loop gets
+ * optimized out.
+ */
+static inline void __dma_sync(struct page *page,
+	unsigned long offset, size_t size, enum dma_data_direction direction)
+{
+	size_t left = size;
+
+	do {
+		size_t len = left;
+
+		if (PageHighMem(page)) {
+			void *addr;
+
+			if (offset + len > PAGE_SIZE) {
+				if (offset >= PAGE_SIZE) {
+					page += offset >> PAGE_SHIFT;
+					offset &= ~PAGE_MASK;
+				}
+				len = PAGE_SIZE - offset;
+			}
+
+			addr = kmap_atomic(page);
+			__dma_sync_virtual(addr + offset, len, direction);
+			kunmap_atomic(addr);
+		} else
+			__dma_sync_virtual(page_address(page) + offset,
+					   size, direction);
+		offset = 0;
+		page++;
+		left -= len;
+	} while (left);
+}
+
+static void mips_dma_unmap_page(struct device *dev, dma_addr_t dma_addr,
+	size_t size, enum dma_data_direction direction, struct dma_attrs *attrs)
+{
+	if (cpu_is_noncoherent_r10000(dev))
+		__dma_sync(dma_addr_to_page(dev, dma_addr),
+			   dma_addr & ~PAGE_MASK, size, direction);
+
+	plat_unmap_dma_mem(dev, dma_addr, size, direction);
+}
+
+static int mips_dma_map_sg(struct device *dev, struct scatterlist *sg,
+	int nents, enum dma_data_direction direction, struct dma_attrs *attrs)
+{
+	int i;
+
+	for (i = 0; i < nents; i++, sg++) {
+		if (!plat_device_is_coherent(dev))
+			__dma_sync(sg_page(sg), sg->offset, sg->length,
+				   direction);
+		sg->dma_address = plat_map_dma_mem_page(dev, sg_page(sg)) +
+				  sg->offset;
+	}
+
+	return nents;
+}
+
+static dma_addr_t mips_dma_map_page(struct device *dev, struct page *page,
+	unsigned long offset, size_t size, enum dma_data_direction direction,
+	struct dma_attrs *attrs)
+{
+	if (!plat_device_is_coherent(dev))
+		__dma_sync(page, offset, size, direction);
+
+	return plat_map_dma_mem_page(dev, page) + offset;
+}
+
+static void mips_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
+	int nhwentries, enum dma_data_direction direction,
+	struct dma_attrs *attrs)
+{
+	int i;
+
+	for (i = 0; i < nhwentries; i++, sg++) {
+		if (!plat_device_is_coherent(dev) &&
+		    direction != DMA_TO_DEVICE)
+			__dma_sync(sg_page(sg), sg->offset, sg->length,
+				   direction);
+		plat_unmap_dma_mem(dev, sg->dma_address, sg->length, direction);
+	}
+}
+
+static void mips_dma_sync_single_for_cpu(struct device *dev,
+	dma_addr_t dma_handle, size_t size, enum dma_data_direction direction)
+{
+	if (cpu_is_noncoherent_r10000(dev))
+		__dma_sync(dma_addr_to_page(dev, dma_handle),
+			   dma_handle & ~PAGE_MASK, size, direction);
+}
+
+static void mips_dma_sync_single_for_device(struct device *dev,
+	dma_addr_t dma_handle, size_t size, enum dma_data_direction direction)
+{
+	plat_extra_sync_for_device(dev);
+	if (!plat_device_is_coherent(dev))
+		__dma_sync(dma_addr_to_page(dev, dma_handle),
+			   dma_handle & ~PAGE_MASK, size, direction);
+}
+
+static void mips_dma_sync_sg_for_cpu(struct device *dev,
+	struct scatterlist *sg, int nelems, enum dma_data_direction direction)
+{
+	int i;
+
+	/* Make sure that gcc doesn't leave the empty loop body.  */
+	for (i = 0; i < nelems; i++, sg++) {
+		if (cpu_is_noncoherent_r10000(dev))
+			__dma_sync(sg_page(sg), sg->offset, sg->length,
+				   direction);
+	}
+}
+
+static void mips_dma_sync_sg_for_device(struct device *dev,
+	struct scatterlist *sg, int nelems, enum dma_data_direction direction)
+{
+	int i;
+
+	/* Make sure that gcc doesn't leave the empty loop body.  */
+	for (i = 0; i < nelems; i++, sg++) {
+		if (!plat_device_is_coherent(dev))
+			__dma_sync(sg_page(sg), sg->offset, sg->length,
+				   direction);
+	}
+}
+
+int mips_dma_mapping_error(struct device *dev, dma_addr_t dma_addr)
+{
+	return plat_dma_mapping_error(dev, dma_addr);
+}
+
+int mips_dma_supported(struct device *dev, u64 mask)
+{
+	return plat_dma_supported(dev, mask);
+}
+
+void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+			 enum dma_data_direction direction)
+{
+	BUG_ON(direction == DMA_NONE);
+
+	plat_extra_sync_for_device(dev);
+	if (!plat_device_is_coherent(dev))
+		__dma_sync_virtual(vaddr, size, direction);
+}
+
+EXPORT_SYMBOL(dma_cache_sync);
+
+static struct dma_map_ops mips_default_dma_map_ops = {
+	.alloc = mips_dma_alloc_coherent,
+	.free = mips_dma_free_coherent,
+	.map_page = mips_dma_map_page,
+	.unmap_page = mips_dma_unmap_page,
+	.map_sg = mips_dma_map_sg,
+	.unmap_sg = mips_dma_unmap_sg,
+	.sync_single_for_cpu = mips_dma_sync_single_for_cpu,
+	.sync_single_for_device = mips_dma_sync_single_for_device,
+	.sync_sg_for_cpu = mips_dma_sync_sg_for_cpu,
+	.sync_sg_for_device = mips_dma_sync_sg_for_device,
+	.mapping_error = mips_dma_mapping_error,
+	.dma_supported = mips_dma_supported
+};
+
+struct dma_map_ops *mips_dma_map_ops = &mips_default_dma_map_ops;
+EXPORT_SYMBOL(mips_dma_map_ops);
+
+#define PREALLOC_DMA_DEBUG_ENTRIES (1 << 16)
+
+static int __init mips_dma_init(void)
+{
+	dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
+
+	return 0;
+}
+fs_initcall(mips_dma_init);
diff --git a/arch/mips/mm/extable.c b/arch/mips/mm/extable.c
new file mode 100644
index 00000000..9d25d2ba
--- /dev/null
+++ b/arch/mips/mm/extable.c
@@ -0,0 +1,25 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1997, 99, 2001 - 2004 Ralf Baechle <ralf@linux-mips.org>
+ */
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <asm/branch.h>
+#include <asm/uaccess.h>
+
+int fixup_exception(struct pt_regs *regs)
+{
+	const struct exception_table_entry *fixup;
+
+	fixup = search_exception_tables(exception_epc(regs));
+	if (fixup) {
+		regs->cp0_epc = fixup->nextinsn;
+
+		return 1;
+	}
+
+	return 0;
+}
diff --git a/arch/mips/mm/fault.c b/arch/mips/mm/fault.c
new file mode 100644
index 00000000..c14f6dfe
--- /dev/null
+++ b/arch/mips/mm/fault.c
@@ -0,0 +1,313 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1995 - 2000 by Ralf Baechle
+ */
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/module.h>
+#include <linux/kprobes.h>
+#include <linux/perf_event.h>
+
+#include <asm/branch.h>
+#include <asm/mmu_context.h>
+#include <asm/uaccess.h>
+#include <asm/ptrace.h>
+#include <asm/highmem.h>		/* For VMALLOC_END */
+#include <linux/kdebug.h>
+
+/*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ */
+asmlinkage void __kprobes do_page_fault(struct pt_regs *regs, unsigned long write,
+			      unsigned long address)
+{
+	struct vm_area_struct * vma = NULL;
+	struct task_struct *tsk = current;
+	struct mm_struct *mm = tsk->mm;
+	const int field = sizeof(unsigned long) * 2;
+	siginfo_t info;
+	int fault;
+	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE |
+						 (write ? FAULT_FLAG_WRITE : 0);
+
+#if 0
+	printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", raw_smp_processor_id(),
+	       current->comm, current->pid, field, address, write,
+	       field, regs->cp0_epc);
+#endif
+
+#ifdef CONFIG_KPROBES
+	/*
+	 * This is to notify the fault handler of the kprobes.  The
+	 * exception code is redundant as it is also carried in REGS,
+	 * but we pass it anyhow.
+	 */
+	if (notify_die(DIE_PAGE_FAULT, "page fault", regs, -1,
+		       (regs->cp0_cause >> 2) & 0x1f, SIGSEGV) == NOTIFY_STOP)
+		return;
+#endif
+
+	info.si_code = SEGV_MAPERR;
+
+	/*
+	 * We fault-in kernel-space virtual memory on-demand. The
+	 * 'reference' page table is init_mm.pgd.
+	 *
+	 * NOTE! We MUST NOT take any locks for this case. We may
+	 * be in an interrupt or a critical region, and should
+	 * only copy the information from the master page table,
+	 * nothing more.
+	 */
+#ifdef CONFIG_64BIT
+# define VMALLOC_FAULT_TARGET no_context
+#else
+# define VMALLOC_FAULT_TARGET vmalloc_fault
+#endif
+
+	if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END))
+		goto VMALLOC_FAULT_TARGET;
+#ifdef MODULE_START
+	if (unlikely(address >= MODULE_START && address < MODULE_END))
+		goto VMALLOC_FAULT_TARGET;
+#endif
+
+	/*
+	 * If we're in an interrupt or have no user
+	 * context, we must not take the fault..
+	 */
+	if (in_atomic() || !mm)
+		goto bad_area_nosemaphore;
+
+retry:
+	down_read(&mm->mmap_sem);
+	vma = find_vma(mm, address);
+	if (!vma)
+		goto bad_area;
+	if (vma->vm_start <= address)
+		goto good_area;
+	if (!(vma->vm_flags & VM_GROWSDOWN))
+		goto bad_area;
+	if (expand_stack(vma, address))
+		goto bad_area;
+/*
+ * Ok, we have a good vm_area for this memory access, so
+ * we can handle it..
+ */
+good_area:
+	info.si_code = SEGV_ACCERR;
+
+	if (write) {
+		if (!(vma->vm_flags & VM_WRITE))
+			goto bad_area;
+	} else {
+		if (kernel_uses_smartmips_rixi) {
+			if (address == regs->cp0_epc && !(vma->vm_flags & VM_EXEC)) {
+#if 0
+				pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] XI violation\n",
+					  raw_smp_processor_id(),
+					  current->comm, current->pid,
+					  field, address, write,
+					  field, regs->cp0_epc);
+#endif
+				goto bad_area;
+			}
+			if (!(vma->vm_flags & VM_READ)) {
+#if 0
+				pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] RI violation\n",
+					  raw_smp_processor_id(),
+					  current->comm, current->pid,
+					  field, address, write,
+					  field, regs->cp0_epc);
+#endif
+				goto bad_area;
+			}
+		} else {
+			if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
+				goto bad_area;
+		}
+	}
+
+	/*
+	 * If for any reason at all we couldn't handle the fault,
+	 * make sure we exit gracefully rather than endlessly redo
+	 * the fault.
+	 */
+	fault = handle_mm_fault(mm, vma, address, flags);
+
+	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
+		return;
+
+	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
+	if (unlikely(fault & VM_FAULT_ERROR)) {
+		if (fault & VM_FAULT_OOM)
+			goto out_of_memory;
+		else if (fault & VM_FAULT_SIGBUS)
+			goto do_sigbus;
+		BUG();
+	}
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
+		if (fault & VM_FAULT_MAJOR) {
+			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1,
+						  regs, address);
+			tsk->maj_flt++;
+		} else {
+			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1,
+						  regs, address);
+			tsk->min_flt++;
+		}
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+
+			/*
+			 * No need to up_read(&mm->mmap_sem) as we would
+			 * have already released it in __lock_page_or_retry
+			 * in mm/filemap.c.
+			 */
+
+			goto retry;
+		}
+	}
+
+	up_read(&mm->mmap_sem);
+	return;
+
+/*
+ * Something tried to access memory that isn't in our memory map..
+ * Fix it, but check if it's kernel or user first..
+ */
+bad_area:
+	up_read(&mm->mmap_sem);
+
+bad_area_nosemaphore:
+	/* User mode accesses just cause a SIGSEGV */
+	if (user_mode(regs)) {
+		tsk->thread.cp0_badvaddr = address;
+		tsk->thread.error_code = write;
+#if 0
+		printk("do_page_fault() #2: sending SIGSEGV to %s for "
+		       "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
+		       tsk->comm,
+		       write ? "write access to" : "read access from",
+		       field, address,
+		       field, (unsigned long) regs->cp0_epc,
+		       field, (unsigned long) regs->regs[31]);
+#endif
+		info.si_signo = SIGSEGV;
+		info.si_errno = 0;
+		/* info.si_code has been set above */
+		info.si_addr = (void __user *) address;
+		force_sig_info(SIGSEGV, &info, tsk);
+		return;
+	}
+
+no_context:
+	/* Are we prepared to handle this kernel fault?  */
+	if (fixup_exception(regs)) {
+		current->thread.cp0_baduaddr = address;
+		return;
+	}
+
+	/*
+	 * Oops. The kernel tried to access some bad page. We'll have to
+	 * terminate things with extreme prejudice.
+	 */
+	bust_spinlocks(1);
+
+	printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
+	       "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
+	       raw_smp_processor_id(), field, address, field, regs->cp0_epc,
+	       field,  regs->regs[31]);
+	die("Oops", regs);
+
+out_of_memory:
+	/*
+	 * We ran out of memory, call the OOM killer, and return the userspace
+	 * (which will retry the fault, or kill us if we got oom-killed).
+	 */
+	up_read(&mm->mmap_sem);
+	pagefault_out_of_memory();
+	return;
+
+do_sigbus:
+	up_read(&mm->mmap_sem);
+
+	/* Kernel mode? Handle exceptions or die */
+	if (!user_mode(regs))
+		goto no_context;
+	else
+	/*
+	 * Send a sigbus, regardless of whether we were in kernel
+	 * or user mode.
+	 */
+#if 0
+		printk("do_page_fault() #3: sending SIGBUS to %s for "
+		       "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
+		       tsk->comm,
+		       write ? "write access to" : "read access from",
+		       field, address,
+		       field, (unsigned long) regs->cp0_epc,
+		       field, (unsigned long) regs->regs[31]);
+#endif
+	tsk->thread.cp0_badvaddr = address;
+	info.si_signo = SIGBUS;
+	info.si_errno = 0;
+	info.si_code = BUS_ADRERR;
+	info.si_addr = (void __user *) address;
+	force_sig_info(SIGBUS, &info, tsk);
+
+	return;
+#ifndef CONFIG_64BIT
+vmalloc_fault:
+	{
+		/*
+		 * Synchronize this task's top level page-table
+		 * with the 'reference' page table.
+		 *
+		 * Do _not_ use "tsk" here. We might be inside
+		 * an interrupt in the middle of a task switch..
+		 */
+		int offset = __pgd_offset(address);
+		pgd_t *pgd, *pgd_k;
+		pud_t *pud, *pud_k;
+		pmd_t *pmd, *pmd_k;
+		pte_t *pte_k;
+
+		pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset;
+		pgd_k = init_mm.pgd + offset;
+
+		if (!pgd_present(*pgd_k))
+			goto no_context;
+		set_pgd(pgd, *pgd_k);
+
+		pud = pud_offset(pgd, address);
+		pud_k = pud_offset(pgd_k, address);
+		if (!pud_present(*pud_k))
+			goto no_context;
+
+		pmd = pmd_offset(pud, address);
+		pmd_k = pmd_offset(pud_k, address);
+		if (!pmd_present(*pmd_k))
+			goto no_context;
+		set_pmd(pmd, *pmd_k);
+
+		pte_k = pte_offset_kernel(pmd_k, address);
+		if (!pte_present(*pte_k))
+			goto no_context;
+		return;
+	}
+#endif
+}
diff --git a/arch/mips/mm/gup.c b/arch/mips/mm/gup.c
new file mode 100644
index 00000000..33aadbcf
--- /dev/null
+++ b/arch/mips/mm/gup.c
@@ -0,0 +1,315 @@
+/*
+ * Lockless get_user_pages_fast for MIPS
+ *
+ * Copyright (C) 2008 Nick Piggin
+ * Copyright (C) 2008 Novell Inc.
+ * Copyright (C) 2011 Ralf Baechle
+ */
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/vmstat.h>
+#include <linux/highmem.h>
+#include <linux/swap.h>
+#include <linux/hugetlb.h>
+
+#include <asm/pgtable.h>
+
+static inline pte_t gup_get_pte(pte_t *ptep)
+{
+#if defined(CONFIG_64BIT_PHYS_ADDR) && defined(CONFIG_CPU_MIPS32)
+	pte_t pte;
+
+retry:
+	pte.pte_low = ptep->pte_low;
+	smp_rmb();
+	pte.pte_high = ptep->pte_high;
+	smp_rmb();
+	if (unlikely(pte.pte_low != ptep->pte_low))
+		goto retry;
+
+	return pte;
+#else
+	return ACCESS_ONCE(*ptep);
+#endif
+}
+
+static int gup_pte_range(pmd_t pmd, unsigned long addr, unsigned long end,
+			int write, struct page **pages, int *nr)
+{
+	pte_t *ptep = pte_offset_map(&pmd, addr);
+	do {
+		pte_t pte = gup_get_pte(ptep);
+		struct page *page;
+
+		if (!pte_present(pte) ||
+		    pte_special(pte) || (write && !pte_write(pte))) {
+			pte_unmap(ptep);
+			return 0;
+		}
+		VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
+		page = pte_page(pte);
+		get_page(page);
+		SetPageReferenced(page);
+		pages[*nr] = page;
+		(*nr)++;
+
+	} while (ptep++, addr += PAGE_SIZE, addr != end);
+
+	pte_unmap(ptep - 1);
+	return 1;
+}
+
+static inline void get_head_page_multiple(struct page *page, int nr)
+{
+	VM_BUG_ON(page != compound_head(page));
+	VM_BUG_ON(page_count(page) == 0);
+	atomic_add(nr, &page->_count);
+	SetPageReferenced(page);
+}
+
+static int gup_huge_pmd(pmd_t pmd, unsigned long addr, unsigned long end,
+			int write, struct page **pages, int *nr)
+{
+	pte_t pte = *(pte_t *)&pmd;
+	struct page *head, *page;
+	int refs;
+
+	if (write && !pte_write(pte))
+		return 0;
+	/* hugepages are never "special" */
+	VM_BUG_ON(pte_special(pte));
+	VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
+
+	refs = 0;
+	head = pte_page(pte);
+	page = head + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
+	do {
+		VM_BUG_ON(compound_head(page) != head);
+		pages[*nr] = page;
+		if (PageTail(page))
+			get_huge_page_tail(page);
+		(*nr)++;
+		page++;
+		refs++;
+	} while (addr += PAGE_SIZE, addr != end);
+
+	get_head_page_multiple(head, refs);
+	return 1;
+}
+
+static int gup_pmd_range(pud_t pud, unsigned long addr, unsigned long end,
+			int write, struct page **pages, int *nr)
+{
+	unsigned long next;
+	pmd_t *pmdp;
+
+	pmdp = pmd_offset(&pud, addr);
+	do {
+		pmd_t pmd = *pmdp;
+
+		next = pmd_addr_end(addr, end);
+		/*
+		 * The pmd_trans_splitting() check below explains why
+		 * pmdp_splitting_flush has to flush the tlb, to stop
+		 * this gup-fast code from running while we set the
+		 * splitting bit in the pmd. Returning zero will take
+		 * the slow path that will call wait_split_huge_page()
+		 * if the pmd is still in splitting state. gup-fast
+		 * can't because it has irq disabled and
+		 * wait_split_huge_page() would never return as the
+		 * tlb flush IPI wouldn't run.
+		 */
+		if (pmd_none(pmd) || pmd_trans_splitting(pmd))
+			return 0;
+		if (unlikely(pmd_huge(pmd))) {
+			if (!gup_huge_pmd(pmd, addr, next, write, pages,nr))
+				return 0;
+		} else {
+			if (!gup_pte_range(pmd, addr, next, write, pages,nr))
+				return 0;
+		}
+	} while (pmdp++, addr = next, addr != end);
+
+	return 1;
+}
+
+static int gup_huge_pud(pud_t pud, unsigned long addr, unsigned long end,
+			int write, struct page **pages, int *nr)
+{
+	pte_t pte = *(pte_t *)&pud;
+	struct page *head, *page;
+	int refs;
+
+	if (write && !pte_write(pte))
+		return 0;
+	/* hugepages are never "special" */
+	VM_BUG_ON(pte_special(pte));
+	VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
+
+	refs = 0;
+	head = pte_page(pte);
+	page = head + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
+	do {
+		VM_BUG_ON(compound_head(page) != head);
+		pages[*nr] = page;
+		(*nr)++;
+		page++;
+		refs++;
+	} while (addr += PAGE_SIZE, addr != end);
+
+	get_head_page_multiple(head, refs);
+	return 1;
+}
+
+static int gup_pud_range(pgd_t pgd, unsigned long addr, unsigned long end,
+			int write, struct page **pages, int *nr)
+{
+	unsigned long next;
+	pud_t *pudp;
+
+	pudp = pud_offset(&pgd, addr);
+	do {
+		pud_t pud = *pudp;
+
+		next = pud_addr_end(addr, end);
+		if (pud_none(pud))
+			return 0;
+		if (unlikely(pud_huge(pud))) {
+			if (!gup_huge_pud(pud, addr, next, write, pages,nr))
+				return 0;
+		} else {
+			if (!gup_pmd_range(pud, addr, next, write, pages,nr))
+				return 0;
+		}
+	} while (pudp++, addr = next, addr != end);
+
+	return 1;
+}
+
+/*
+ * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
+ * back to the regular GUP.
+ */
+int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
+			  struct page **pages)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long addr, len, end;
+	unsigned long next;
+	unsigned long flags;
+	pgd_t *pgdp;
+	int nr = 0;
+
+	start &= PAGE_MASK;
+	addr = start;
+	len = (unsigned long) nr_pages << PAGE_SHIFT;
+	end = start + len;
+	if (unlikely(!access_ok(write ? VERIFY_WRITE : VERIFY_READ,
+					(void __user *)start, len)))
+		return 0;
+
+	/*
+	 * XXX: batch / limit 'nr', to avoid large irq off latency
+	 * needs some instrumenting to determine the common sizes used by
+	 * important workloads (eg. DB2), and whether limiting the batch
+	 * size will decrease performance.
+	 *
+	 * It seems like we're in the clear for the moment. Direct-IO is
+	 * the main guy that batches up lots of get_user_pages, and even
+	 * they are limited to 64-at-a-time which is not so many.
+	 */
+	/*
+	 * This doesn't prevent pagetable teardown, but does prevent
+	 * the pagetables and pages from being freed.
+	 *
+	 * So long as we atomically load page table pointers versus teardown,
+	 * we can follow the address down to the page and take a ref on it.
+	 */
+	local_irq_save(flags);
+	pgdp = pgd_offset(mm, addr);
+	do {
+		pgd_t pgd = *pgdp;
+
+		next = pgd_addr_end(addr, end);
+		if (pgd_none(pgd))
+			break;
+		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
+			break;
+	} while (pgdp++, addr = next, addr != end);
+	local_irq_restore(flags);
+
+	return nr;
+}
+
+/**
+ * get_user_pages_fast() - pin user pages in memory
+ * @start:	starting user address
+ * @nr_pages:	number of pages from start to pin
+ * @write:	whether pages will be written to
+ * @pages:	array that receives pointers to the pages pinned.
+ * 		Should be at least nr_pages long.
+ *
+ * Attempt to pin user pages in memory without taking mm->mmap_sem.
+ * If not successful, it will fall back to taking the lock and
+ * calling get_user_pages().
+ *
+ * Returns number of pages pinned. This may be fewer than the number
+ * requested. If nr_pages is 0 or negative, returns 0. If no pages
+ * were pinned, returns -errno.
+ */
+int get_user_pages_fast(unsigned long start, int nr_pages, int write,
+			struct page **pages)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long addr, len, end;
+	unsigned long next;
+	pgd_t *pgdp;
+	int ret, nr = 0;
+
+	start &= PAGE_MASK;
+	addr = start;
+	len = (unsigned long) nr_pages << PAGE_SHIFT;
+
+	end = start + len;
+	if (end < start)
+		goto slow_irqon;
+
+	/* XXX: batch / limit 'nr' */
+	local_irq_disable();
+	pgdp = pgd_offset(mm, addr);
+	do {
+		pgd_t pgd = *pgdp;
+
+		next = pgd_addr_end(addr, end);
+		if (pgd_none(pgd))
+			goto slow;
+		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
+			goto slow;
+	} while (pgdp++, addr = next, addr != end);
+	local_irq_enable();
+
+	VM_BUG_ON(nr != (end - start) >> PAGE_SHIFT);
+	return nr;
+slow:
+	local_irq_enable();
+
+slow_irqon:
+	/* Try to get the remaining pages with get_user_pages */
+	start += nr << PAGE_SHIFT;
+	pages += nr;
+
+	down_read(&mm->mmap_sem);
+	ret = get_user_pages(current, mm, start,
+				(end - start) >> PAGE_SHIFT,
+				write, 0, pages, NULL);
+	up_read(&mm->mmap_sem);
+
+	/* Have to be a bit careful with return values */
+	if (nr > 0) {
+		if (ret < 0)
+			ret = nr;
+		else
+			ret += nr;
+	}
+	return ret;
+}
diff --git a/arch/mips/mm/highmem.c b/arch/mips/mm/highmem.c
new file mode 100644
index 00000000..aff57057
--- /dev/null
+++ b/arch/mips/mm/highmem.c
@@ -0,0 +1,137 @@
+#include <linux/module.h>
+#include <linux/highmem.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <asm/fixmap.h>
+#include <asm/tlbflush.h>
+
+static pte_t *kmap_pte;
+
+unsigned long highstart_pfn, highend_pfn;
+
+void *kmap(struct page *page)
+{
+	void *addr;
+
+	might_sleep();
+	if (!PageHighMem(page))
+		return page_address(page);
+	addr = kmap_high(page);
+	flush_tlb_one((unsigned long)addr);
+
+	return addr;
+}
+EXPORT_SYMBOL(kmap);
+
+void kunmap(struct page *page)
+{
+	BUG_ON(in_interrupt());
+	if (!PageHighMem(page))
+		return;
+	kunmap_high(page);
+}
+EXPORT_SYMBOL(kunmap);
+
+/*
+ * kmap_atomic/kunmap_atomic is significantly faster than kmap/kunmap because
+ * no global lock is needed and because the kmap code must perform a global TLB
+ * invalidation when the kmap pool wraps.
+ *
+ * However when holding an atomic kmap is is not legal to sleep, so atomic
+ * kmaps are appropriate for short, tight code paths only.
+ */
+
+void *kmap_atomic(struct page *page)
+{
+	unsigned long vaddr;
+	int idx, type;
+
+	/* even !CONFIG_PREEMPT needs this, for in_atomic in do_page_fault */
+	pagefault_disable();
+	if (!PageHighMem(page))
+		return page_address(page);
+
+	type = kmap_atomic_idx_push();
+	idx = type + KM_TYPE_NR*smp_processor_id();
+	vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+#ifdef CONFIG_DEBUG_HIGHMEM
+	BUG_ON(!pte_none(*(kmap_pte - idx)));
+#endif
+	set_pte(kmap_pte-idx, mk_pte(page, PAGE_KERNEL));
+	local_flush_tlb_one((unsigned long)vaddr);
+
+	return (void*) vaddr;
+}
+EXPORT_SYMBOL(kmap_atomic);
+
+void __kunmap_atomic(void *kvaddr)
+{
+	unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK;
+	int type;
+
+	if (vaddr < FIXADDR_START) { // FIXME
+		pagefault_enable();
+		return;
+	}
+
+	type = kmap_atomic_idx();
+#ifdef CONFIG_DEBUG_HIGHMEM
+	{
+		int idx = type + KM_TYPE_NR * smp_processor_id();
+
+		BUG_ON(vaddr != __fix_to_virt(FIX_KMAP_BEGIN + idx));
+
+		/*
+		 * force other mappings to Oops if they'll try to access
+		 * this pte without first remap it
+		 */
+		pte_clear(&init_mm, vaddr, kmap_pte-idx);
+		local_flush_tlb_one(vaddr);
+	}
+#endif
+	kmap_atomic_idx_pop();
+	pagefault_enable();
+}
+EXPORT_SYMBOL(__kunmap_atomic);
+
+/*
+ * This is the same as kmap_atomic() but can map memory that doesn't
+ * have a struct page associated with it.
+ */
+void *kmap_atomic_pfn(unsigned long pfn)
+{
+	unsigned long vaddr;
+	int idx, type;
+
+	pagefault_disable();
+
+	type = kmap_atomic_idx_push();
+	idx = type + KM_TYPE_NR*smp_processor_id();
+	vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+	set_pte(kmap_pte-idx, pfn_pte(pfn, PAGE_KERNEL));
+	flush_tlb_one(vaddr);
+
+	return (void*) vaddr;
+}
+
+struct page *kmap_atomic_to_page(void *ptr)
+{
+	unsigned long idx, vaddr = (unsigned long)ptr;
+	pte_t *pte;
+
+	if (vaddr < FIXADDR_START)
+		return virt_to_page(ptr);
+
+	idx = virt_to_fix(vaddr);
+	pte = kmap_pte - (idx - FIX_KMAP_BEGIN);
+	return pte_page(*pte);
+}
+
+void __init kmap_init(void)
+{
+	unsigned long kmap_vstart;
+
+	/* cache the first kmap pte */
+	kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN);
+	kmap_pte = kmap_get_fixmap_pte(kmap_vstart);
+}
diff --git a/arch/mips/mm/hugetlbpage.c b/arch/mips/mm/hugetlbpage.c
new file mode 100644
index 00000000..a7fee0df
--- /dev/null
+++ b/arch/mips/mm/hugetlbpage.c
@@ -0,0 +1,98 @@
+/*
+ * MIPS Huge TLB Page Support for Kernel.
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2002, Rohit Seth <rohit.seth@intel.com>
+ * Copyright 2005, Embedded Alley Solutions, Inc.
+ * Matt Porter <mporter@embeddedalley.com>
+ * Copyright (C) 2008, 2009 Cavium Networks, Inc.
+ */
+
+#include <linux/init.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/pagemap.h>
+#include <linux/err.h>
+#include <linux/sysctl.h>
+#include <asm/mman.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+
+pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr,
+		      unsigned long sz)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pte_t *pte = NULL;
+
+	pgd = pgd_offset(mm, addr);
+	pud = pud_alloc(mm, pgd, addr);
+	if (pud)
+		pte = (pte_t *)pmd_alloc(mm, pud, addr);
+
+	return pte;
+}
+
+pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd = NULL;
+
+	pgd = pgd_offset(mm, addr);
+	if (pgd_present(*pgd)) {
+		pud = pud_offset(pgd, addr);
+		if (pud_present(*pud))
+			pmd = pmd_offset(pud, addr);
+	}
+	return (pte_t *) pmd;
+}
+
+int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
+{
+	return 0;
+}
+
+/*
+ * This function checks for proper alignment of input addr and len parameters.
+ */
+int is_aligned_hugepage_range(unsigned long addr, unsigned long len)
+{
+	if (len & ~HPAGE_MASK)
+		return -EINVAL;
+	if (addr & ~HPAGE_MASK)
+		return -EINVAL;
+	return 0;
+}
+
+struct page *
+follow_huge_addr(struct mm_struct *mm, unsigned long address, int write)
+{
+	return ERR_PTR(-EINVAL);
+}
+
+int pmd_huge(pmd_t pmd)
+{
+	return (pmd_val(pmd) & _PAGE_HUGE) != 0;
+}
+
+int pud_huge(pud_t pud)
+{
+	return (pud_val(pud) & _PAGE_HUGE) != 0;
+}
+
+struct page *
+follow_huge_pmd(struct mm_struct *mm, unsigned long address,
+		pmd_t *pmd, int write)
+{
+	struct page *page;
+
+	page = pte_page(*(pte_t *)pmd);
+	if (page)
+		page += ((address & ~HPAGE_MASK) >> PAGE_SHIFT);
+	return page;
+}
diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
new file mode 100644
index 00000000..1a85ba92
--- /dev/null
+++ b/arch/mips/mm/init.c
@@ -0,0 +1,487 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1994 - 2000 Ralf Baechle
+ * Copyright (C) 1999, 2000 Silicon Graphics, Inc.
+ * Kevin D. Kissell, kevink@mips.com and Carsten Langgaard, carstenl@mips.com
+ * Copyright (C) 2000 MIPS Technologies, Inc.  All rights reserved.
+ */
+#include <linux/bug.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/pagemap.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/bootmem.h>
+#include <linux/highmem.h>
+#include <linux/swap.h>
+#include <linux/proc_fs.h>
+#include <linux/pfn.h>
+#include <linux/hardirq.h>
+#include <linux/gfp.h>
+
+#include <asm/asm-offsets.h>
+#include <asm/bootinfo.h>
+#include <asm/cachectl.h>
+#include <asm/cpu.h>
+#include <asm/dma.h>
+#include <asm/kmap_types.h>
+#include <asm/mmu_context.h>
+#include <asm/sections.h>
+#include <asm/pgtable.h>
+#include <asm/pgalloc.h>
+#include <asm/tlb.h>
+#include <asm/fixmap.h>
+
+/* Atomicity and interruptability */
+#ifdef CONFIG_MIPS_MT_SMTC
+
+#include <asm/mipsmtregs.h>
+
+#define ENTER_CRITICAL(flags) \
+	{ \
+	unsigned int mvpflags; \
+	local_irq_save(flags);\
+	mvpflags = dvpe()
+#define EXIT_CRITICAL(flags) \
+	evpe(mvpflags); \
+	local_irq_restore(flags); \
+	}
+#else
+
+#define ENTER_CRITICAL(flags) local_irq_save(flags)
+#define EXIT_CRITICAL(flags) local_irq_restore(flags)
+
+#endif /* CONFIG_MIPS_MT_SMTC */
+
+/*
+ * We have up to 8 empty zeroed pages so we can map one of the right colour
+ * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
+ * where we have to avoid VCED / VECI exceptions for good performance at
+ * any price.  Since page is never written to after the initialization we
+ * don't have to care about aliases on other CPUs.
+ */
+unsigned long empty_zero_page, zero_page_mask;
+EXPORT_SYMBOL_GPL(empty_zero_page);
+
+/*
+ * Not static inline because used by IP27 special magic initialization code
+ */
+unsigned long setup_zero_pages(void)
+{
+	unsigned int order;
+	unsigned long size;
+	struct page *page;
+
+	if (cpu_has_vce)
+		order = 3;
+	else
+		order = 0;
+
+	empty_zero_page = __get_free_pages(GFP_KERNEL | __GFP_ZERO, order);
+	if (!empty_zero_page)
+		panic("Oh boy, that early out of memory?");
+
+	page = virt_to_page((void *)empty_zero_page);
+	split_page(page, order);
+	while (page < virt_to_page((void *)(empty_zero_page + (PAGE_SIZE << order)))) {
+		SetPageReserved(page);
+		page++;
+	}
+
+	size = PAGE_SIZE << order;
+	zero_page_mask = (size - 1) & PAGE_MASK;
+
+	return 1UL << order;
+}
+
+#ifdef CONFIG_MIPS_MT_SMTC
+static pte_t *kmap_coherent_pte;
+static void __init kmap_coherent_init(void)
+{
+	unsigned long vaddr;
+
+	/* cache the first coherent kmap pte */
+	vaddr = __fix_to_virt(FIX_CMAP_BEGIN);
+	kmap_coherent_pte = kmap_get_fixmap_pte(vaddr);
+}
+#else
+static inline void kmap_coherent_init(void) {}
+#endif
+
+void *kmap_coherent(struct page *page, unsigned long addr)
+{
+	enum fixed_addresses idx;
+	unsigned long vaddr, flags, entrylo;
+	unsigned long old_ctx;
+	pte_t pte;
+	int tlbidx;
+
+	BUG_ON(Page_dcache_dirty(page));
+
+	inc_preempt_count();
+	idx = (addr >> PAGE_SHIFT) & (FIX_N_COLOURS - 1);
+#ifdef CONFIG_MIPS_MT_SMTC
+	idx += FIX_N_COLOURS * smp_processor_id() +
+		(in_interrupt() ? (FIX_N_COLOURS * NR_CPUS) : 0);
+#else
+	idx += in_interrupt() ? FIX_N_COLOURS : 0;
+#endif
+	vaddr = __fix_to_virt(FIX_CMAP_END - idx);
+	pte = mk_pte(page, PAGE_KERNEL);
+#if defined(CONFIG_64BIT_PHYS_ADDR) && defined(CONFIG_CPU_MIPS32)
+	entrylo = pte.pte_high;
+#else
+	entrylo = pte_to_entrylo(pte_val(pte));
+#endif
+
+	ENTER_CRITICAL(flags);
+	old_ctx = read_c0_entryhi();
+	write_c0_entryhi(vaddr & (PAGE_MASK << 1));
+	write_c0_entrylo0(entrylo);
+	write_c0_entrylo1(entrylo);
+#ifdef CONFIG_MIPS_MT_SMTC
+	set_pte(kmap_coherent_pte - (FIX_CMAP_END - idx), pte);
+	/* preload TLB instead of local_flush_tlb_one() */
+	mtc0_tlbw_hazard();
+	tlb_probe();
+	tlb_probe_hazard();
+	tlbidx = read_c0_index();
+	mtc0_tlbw_hazard();
+	if (tlbidx < 0)
+		tlb_write_random();
+	else
+		tlb_write_indexed();
+#else
+	tlbidx = read_c0_wired();
+	write_c0_wired(tlbidx + 1);
+	write_c0_index(tlbidx);
+	mtc0_tlbw_hazard();
+	tlb_write_indexed();
+#endif
+	tlbw_use_hazard();
+	write_c0_entryhi(old_ctx);
+	EXIT_CRITICAL(flags);
+
+	return (void*) vaddr;
+}
+
+#define UNIQUE_ENTRYHI(idx) (CKSEG0 + ((idx) << (PAGE_SHIFT + 1)))
+
+void kunmap_coherent(void)
+{
+#ifndef CONFIG_MIPS_MT_SMTC
+	unsigned int wired;
+	unsigned long flags, old_ctx;
+
+	ENTER_CRITICAL(flags);
+	old_ctx = read_c0_entryhi();
+	wired = read_c0_wired() - 1;
+	write_c0_wired(wired);
+	write_c0_index(wired);
+	write_c0_entryhi(UNIQUE_ENTRYHI(wired));
+	write_c0_entrylo0(0);
+	write_c0_entrylo1(0);
+	mtc0_tlbw_hazard();
+	tlb_write_indexed();
+	tlbw_use_hazard();
+	write_c0_entryhi(old_ctx);
+	EXIT_CRITICAL(flags);
+#endif
+	dec_preempt_count();
+	preempt_check_resched();
+}
+
+void copy_user_highpage(struct page *to, struct page *from,
+	unsigned long vaddr, struct vm_area_struct *vma)
+{
+	void *vfrom, *vto;
+
+	vto = kmap_atomic(to);
+	if (cpu_has_dc_aliases &&
+	    page_mapped(from) && !Page_dcache_dirty(from)) {
+		vfrom = kmap_coherent(from, vaddr);
+		copy_page(vto, vfrom);
+		kunmap_coherent();
+	} else {
+		vfrom = kmap_atomic(from);
+		copy_page(vto, vfrom);
+		kunmap_atomic(vfrom);
+	}
+	if ((!cpu_has_ic_fills_f_dc) ||
+	    pages_do_alias((unsigned long)vto, vaddr & PAGE_MASK))
+		flush_data_cache_page((unsigned long)vto);
+	kunmap_atomic(vto);
+	/* Make sure this page is cleared on other CPU's too before using it */
+	smp_wmb();
+}
+
+void copy_to_user_page(struct vm_area_struct *vma,
+	struct page *page, unsigned long vaddr, void *dst, const void *src,
+	unsigned long len)
+{
+	if (cpu_has_dc_aliases &&
+	    page_mapped(page) && !Page_dcache_dirty(page)) {
+		void *vto = kmap_coherent(page, vaddr) + (vaddr & ~PAGE_MASK);
+		memcpy(vto, src, len);
+		kunmap_coherent();
+	} else {
+		memcpy(dst, src, len);
+		if (cpu_has_dc_aliases)
+			SetPageDcacheDirty(page);
+	}
+	if ((vma->vm_flags & VM_EXEC) && !cpu_has_ic_fills_f_dc)
+		flush_cache_page(vma, vaddr, page_to_pfn(page));
+}
+
+void copy_from_user_page(struct vm_area_struct *vma,
+	struct page *page, unsigned long vaddr, void *dst, const void *src,
+	unsigned long len)
+{
+	if (cpu_has_dc_aliases &&
+	    page_mapped(page) && !Page_dcache_dirty(page)) {
+		void *vfrom = kmap_coherent(page, vaddr) + (vaddr & ~PAGE_MASK);
+		memcpy(dst, vfrom, len);
+		kunmap_coherent();
+	} else {
+		memcpy(dst, src, len);
+		if (cpu_has_dc_aliases)
+			SetPageDcacheDirty(page);
+	}
+}
+
+void __init fixrange_init(unsigned long start, unsigned long end,
+	pgd_t *pgd_base)
+{
+#if defined(CONFIG_HIGHMEM) || defined(CONFIG_MIPS_MT_SMTC)
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+	int i, j, k;
+	unsigned long vaddr;
+
+	vaddr = start;
+	i = __pgd_offset(vaddr);
+	j = __pud_offset(vaddr);
+	k = __pmd_offset(vaddr);
+	pgd = pgd_base + i;
+
+	for ( ; (i < PTRS_PER_PGD) && (vaddr < end); pgd++, i++) {
+		pud = (pud_t *)pgd;
+		for ( ; (j < PTRS_PER_PUD) && (vaddr < end); pud++, j++) {
+			pmd = (pmd_t *)pud;
+			for (; (k < PTRS_PER_PMD) && (vaddr < end); pmd++, k++) {
+				if (pmd_none(*pmd)) {
+					pte = (pte_t *) alloc_bootmem_low_pages(PAGE_SIZE);
+					set_pmd(pmd, __pmd((unsigned long)pte));
+					BUG_ON(pte != pte_offset_kernel(pmd, 0));
+				}
+				vaddr += PMD_SIZE;
+			}
+			k = 0;
+		}
+		j = 0;
+	}
+#endif
+}
+
+#ifndef CONFIG_NEED_MULTIPLE_NODES
+int page_is_ram(unsigned long pagenr)
+{
+	int i;
+
+	for (i = 0; i < boot_mem_map.nr_map; i++) {
+		unsigned long addr, end;
+
+		switch (boot_mem_map.map[i].type) {
+		case BOOT_MEM_RAM:
+		case BOOT_MEM_INIT_RAM:
+			break;
+		default:
+			/* not usable memory */
+			continue;
+		}
+
+		addr = PFN_UP(boot_mem_map.map[i].addr);
+		end = PFN_DOWN(boot_mem_map.map[i].addr +
+			       boot_mem_map.map[i].size);
+
+		if (pagenr >= addr && pagenr < end)
+			return 1;
+	}
+
+	return 0;
+}
+
+void __init paging_init(void)
+{
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
+	unsigned long lastpfn __maybe_unused;
+
+	pagetable_init();
+
+#ifdef CONFIG_HIGHMEM
+	kmap_init();
+#endif
+	kmap_coherent_init();
+
+#ifdef CONFIG_ZONE_DMA
+	max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+#endif
+#ifdef CONFIG_ZONE_DMA32
+	max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
+#endif
+	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
+	lastpfn = max_low_pfn;
+#ifdef CONFIG_HIGHMEM
+	max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
+	lastpfn = highend_pfn;
+
+	if (cpu_has_dc_aliases && max_low_pfn != highend_pfn) {
+		printk(KERN_WARNING "This processor doesn't support highmem."
+		       " %ldk highmem ignored\n",
+		       (highend_pfn - max_low_pfn) << (PAGE_SHIFT - 10));
+		max_zone_pfns[ZONE_HIGHMEM] = max_low_pfn;
+		lastpfn = max_low_pfn;
+	}
+#endif
+
+	free_area_init_nodes(max_zone_pfns);
+}
+
+#ifdef CONFIG_64BIT
+static struct kcore_list kcore_kseg0;
+#endif
+
+void __init mem_init(void)
+{
+	unsigned long codesize, reservedpages, datasize, initsize;
+	unsigned long tmp, ram;
+
+#ifdef CONFIG_HIGHMEM
+#ifdef CONFIG_DISCONTIGMEM
+#error "CONFIG_HIGHMEM and CONFIG_DISCONTIGMEM dont work together yet"
+#endif
+	max_mapnr = highend_pfn ? highend_pfn : max_low_pfn;
+#else
+	max_mapnr = max_low_pfn;
+#endif
+	high_memory = (void *) __va(max_low_pfn << PAGE_SHIFT);
+
+	totalram_pages += free_all_bootmem();
+	totalram_pages -= setup_zero_pages();	/* Setup zeroed pages.  */
+
+	reservedpages = ram = 0;
+	for (tmp = 0; tmp < max_low_pfn; tmp++)
+		if (page_is_ram(tmp) && pfn_valid(tmp)) {
+			ram++;
+			if (PageReserved(pfn_to_page(tmp)))
+				reservedpages++;
+		}
+	num_physpages = ram;
+
+#ifdef CONFIG_HIGHMEM
+	for (tmp = highstart_pfn; tmp < highend_pfn; tmp++) {
+		struct page *page = pfn_to_page(tmp);
+
+		if (!page_is_ram(tmp)) {
+			SetPageReserved(page);
+			continue;
+		}
+		ClearPageReserved(page);
+		init_page_count(page);
+		__free_page(page);
+		totalhigh_pages++;
+	}
+	totalram_pages += totalhigh_pages;
+	num_physpages += totalhigh_pages;
+#endif
+
+	codesize =  (unsigned long) &_etext - (unsigned long) &_text;
+	datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
+	initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
+
+#ifdef CONFIG_64BIT
+	if ((unsigned long) &_text > (unsigned long) CKSEG0)
+		/* The -4 is a hack so that user tools don't have to handle
+		   the overflow.  */
+		kclist_add(&kcore_kseg0, (void *) CKSEG0,
+				0x80000000 - 4, KCORE_TEXT);
+#endif
+
+	printk(KERN_INFO "Memory: %luk/%luk available (%ldk kernel code, "
+	       "%ldk reserved, %ldk data, %ldk init, %ldk highmem)\n",
+	       nr_free_pages() << (PAGE_SHIFT-10),
+	       ram << (PAGE_SHIFT-10),
+	       codesize >> 10,
+	       reservedpages << (PAGE_SHIFT-10),
+	       datasize >> 10,
+	       initsize >> 10,
+	       totalhigh_pages << (PAGE_SHIFT-10));
+}
+#endif /* !CONFIG_NEED_MULTIPLE_NODES */
+
+void free_init_pages(const char *what, unsigned long begin, unsigned long end)
+{
+	unsigned long pfn;
+
+	for (pfn = PFN_UP(begin); pfn < PFN_DOWN(end); pfn++) {
+		struct page *page = pfn_to_page(pfn);
+		void *addr = phys_to_virt(PFN_PHYS(pfn));
+
+		ClearPageReserved(page);
+		init_page_count(page);
+		memset(addr, POISON_FREE_INITMEM, PAGE_SIZE);
+		__free_page(page);
+		totalram_pages++;
+	}
+	printk(KERN_INFO "Freeing %s: %ldk freed\n", what, (end - begin) >> 10);
+}
+
+#ifdef CONFIG_BLK_DEV_INITRD
+void free_initrd_mem(unsigned long start, unsigned long end)
+{
+	free_init_pages("initrd memory",
+			virt_to_phys((void *)start),
+			virt_to_phys((void *)end));
+}
+#endif
+
+void __init_refok free_initmem(void)
+{
+	prom_free_prom_memory();
+	free_init_pages("unused kernel memory",
+			__pa_symbol(&__init_begin),
+			__pa_symbol(&__init_end));
+}
+
+#ifndef CONFIG_MIPS_PGD_C0_CONTEXT
+unsigned long pgd_current[NR_CPUS];
+#endif
+/*
+ * On 64-bit we've got three-level pagetables with a slightly
+ * different layout ...
+ */
+#define __page_aligned(order) __attribute__((__aligned__(PAGE_SIZE<<order)))
+
+/*
+ * gcc 3.3 and older have trouble determining that PTRS_PER_PGD and PGD_ORDER
+ * are constants.  So we use the variants from asm-offset.h until that gcc
+ * will officially be retired.
+ */
+pgd_t swapper_pg_dir[_PTRS_PER_PGD] __page_aligned(_PGD_ORDER);
+#ifndef __PAGETABLE_PMD_FOLDED
+pmd_t invalid_pmd_table[PTRS_PER_PMD] __page_aligned(PMD_ORDER);
+#endif
+pte_t invalid_pte_table[PTRS_PER_PTE] __page_aligned(PTE_ORDER);
diff --git a/arch/mips/mm/ioremap.c b/arch/mips/mm/ioremap.c
new file mode 100644
index 00000000..cacfd31e
--- /dev/null
+++ b/arch/mips/mm/ioremap.c
@@ -0,0 +1,192 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * (C) Copyright 1995 1996 Linus Torvalds
+ * (C) Copyright 2001, 2002 Ralf Baechle
+ */
+#include <linux/module.h>
+#include <asm/addrspace.h>
+#include <asm/byteorder.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <asm/cacheflush.h>
+#include <asm/io.h>
+#include <asm/tlbflush.h>
+
+static inline void remap_area_pte(pte_t * pte, unsigned long address,
+	phys_t size, phys_t phys_addr, unsigned long flags)
+{
+	phys_t end;
+	unsigned long pfn;
+	pgprot_t pgprot = __pgprot(_PAGE_GLOBAL | _PAGE_PRESENT | __READABLE
+	                           | __WRITEABLE | flags);
+
+	address &= ~PMD_MASK;
+	end = address + size;
+	if (end > PMD_SIZE)
+		end = PMD_SIZE;
+	BUG_ON(address >= end);
+	pfn = phys_addr >> PAGE_SHIFT;
+	do {
+		if (!pte_none(*pte)) {
+			printk("remap_area_pte: page already exists\n");
+			BUG();
+		}
+		set_pte(pte, pfn_pte(pfn, pgprot));
+		address += PAGE_SIZE;
+		pfn++;
+		pte++;
+	} while (address && (address < end));
+}
+
+static inline int remap_area_pmd(pmd_t * pmd, unsigned long address,
+	phys_t size, phys_t phys_addr, unsigned long flags)
+{
+	phys_t end;
+
+	address &= ~PGDIR_MASK;
+	end = address + size;
+	if (end > PGDIR_SIZE)
+		end = PGDIR_SIZE;
+	phys_addr -= address;
+	BUG_ON(address >= end);
+	do {
+		pte_t * pte = pte_alloc_kernel(pmd, address);
+		if (!pte)
+			return -ENOMEM;
+		remap_area_pte(pte, address, end - address, address + phys_addr, flags);
+		address = (address + PMD_SIZE) & PMD_MASK;
+		pmd++;
+	} while (address && (address < end));
+	return 0;
+}
+
+static int remap_area_pages(unsigned long address, phys_t phys_addr,
+	phys_t size, unsigned long flags)
+{
+	int error;
+	pgd_t * dir;
+	unsigned long end = address + size;
+
+	phys_addr -= address;
+	dir = pgd_offset(&init_mm, address);
+	flush_cache_all();
+	BUG_ON(address >= end);
+	do {
+		pud_t *pud;
+		pmd_t *pmd;
+
+		error = -ENOMEM;
+		pud = pud_alloc(&init_mm, dir, address);
+		if (!pud)
+			break;
+		pmd = pmd_alloc(&init_mm, pud, address);
+		if (!pmd)
+			break;
+		if (remap_area_pmd(pmd, address, end - address,
+					 phys_addr + address, flags))
+			break;
+		error = 0;
+		address = (address + PGDIR_SIZE) & PGDIR_MASK;
+		dir++;
+	} while (address && (address < end));
+	flush_tlb_all();
+	return error;
+}
+
+/*
+ * Generic mapping function (not visible outside):
+ */
+
+/*
+ * Remap an arbitrary physical address space into the kernel virtual
+ * address space. Needed when the kernel wants to access high addresses
+ * directly.
+ *
+ * NOTE! We need to allow non-page-aligned mappings too: we will obviously
+ * have to convert them into an offset in a page-aligned mapping, but the
+ * caller shouldn't need to know that small detail.
+ */
+
+#define IS_LOW512(addr) (!((phys_t)(addr) & (phys_t) ~0x1fffffffULL))
+
+void __iomem * __ioremap(phys_t phys_addr, phys_t size, unsigned long flags)
+{
+	struct vm_struct * area;
+	unsigned long offset;
+	phys_t last_addr;
+	void * addr;
+
+	phys_addr = fixup_bigphys_addr(phys_addr, size);
+
+	/* Don't allow wraparound or zero size */
+	last_addr = phys_addr + size - 1;
+	if (!size || last_addr < phys_addr)
+		return NULL;
+
+	/*
+	 * Map uncached objects in the low 512mb of address space using KSEG1,
+	 * otherwise map using page tables.
+	 */
+	if (IS_LOW512(phys_addr) && IS_LOW512(last_addr) &&
+	    flags == _CACHE_UNCACHED)
+		return (void __iomem *) CKSEG1ADDR(phys_addr);
+
+	/*
+	 * Don't allow anybody to remap normal RAM that we're using..
+	 */
+	if (phys_addr < virt_to_phys(high_memory)) {
+		char *t_addr, *t_end;
+		struct page *page;
+
+		t_addr = __va(phys_addr);
+		t_end = t_addr + (size - 1);
+
+		for(page = virt_to_page(t_addr); page <= virt_to_page(t_end); page++)
+			if(!PageReserved(page))
+				return NULL;
+	}
+
+	/*
+	 * Mappings have to be page-aligned
+	 */
+	offset = phys_addr & ~PAGE_MASK;
+	phys_addr &= PAGE_MASK;
+	size = PAGE_ALIGN(last_addr + 1) - phys_addr;
+
+	/*
+	 * Ok, go for it..
+	 */
+	area = get_vm_area(size, VM_IOREMAP);
+	if (!area)
+		return NULL;
+	addr = area->addr;
+	if (remap_area_pages((unsigned long) addr, phys_addr, size, flags)) {
+		vunmap(addr);
+		return NULL;
+	}
+
+	return (void __iomem *) (offset + (char *)addr);
+}
+
+#define IS_KSEG1(addr) (((unsigned long)(addr) & ~0x1fffffffUL) == CKSEG1)
+
+void __iounmap(const volatile void __iomem *addr)
+{
+	struct vm_struct *p;
+
+	if (IS_KSEG1(addr))
+		return;
+
+	p = remove_vm_area((void *) (PAGE_MASK & (unsigned long __force) addr));
+	if (!p)
+		printk(KERN_ERR "iounmap: bad address %p\n", addr);
+
+        kfree(p);
+}
+
+EXPORT_SYMBOL(__ioremap);
+EXPORT_SYMBOL(__iounmap);
diff --git a/arch/mips/mm/mmap.c b/arch/mips/mm/mmap.c
new file mode 100644
index 00000000..302d779d
--- /dev/null
+++ b/arch/mips/mm/mmap.c
@@ -0,0 +1,271 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2011 Wind River Systems,
+ *   written by Ralf Baechle <ralf@linux-mips.org>
+ */
+#include <linux/compiler.h>
+#include <linux/errno.h>
+#include <linux/mm.h>
+#include <linux/mman.h>
+#include <linux/module.h>
+#include <linux/personality.h>
+#include <linux/random.h>
+#include <linux/sched.h>
+
+unsigned long shm_align_mask = PAGE_SIZE - 1;	/* Sane caches */
+EXPORT_SYMBOL(shm_align_mask);
+
+/* gap between mmap and stack */
+#define MIN_GAP (128*1024*1024UL)
+#define MAX_GAP ((TASK_SIZE)/6*5)
+
+static int mmap_is_legacy(void)
+{
+	if (current->personality & ADDR_COMPAT_LAYOUT)
+		return 1;
+
+	if (rlimit(RLIMIT_STACK) == RLIM_INFINITY)
+		return 1;
+
+	return sysctl_legacy_va_layout;
+}
+
+static unsigned long mmap_base(unsigned long rnd)
+{
+	unsigned long gap = rlimit(RLIMIT_STACK);
+
+	if (gap < MIN_GAP)
+		gap = MIN_GAP;
+	else if (gap > MAX_GAP)
+		gap = MAX_GAP;
+
+	return PAGE_ALIGN(TASK_SIZE - gap - rnd);
+}
+
+static inline unsigned long COLOUR_ALIGN_DOWN(unsigned long addr,
+					      unsigned long pgoff)
+{
+	unsigned long base = addr & ~shm_align_mask;
+	unsigned long off = (pgoff << PAGE_SHIFT) & shm_align_mask;
+
+	if (base + off <= addr)
+		return base + off;
+
+	return base - off;
+}
+
+#define COLOUR_ALIGN(addr, pgoff)				\
+	((((addr) + shm_align_mask) & ~shm_align_mask) +	\
+	 (((pgoff) << PAGE_SHIFT) & shm_align_mask))
+
+enum mmap_allocation_direction {UP, DOWN};
+
+static unsigned long arch_get_unmapped_area_common(struct file *filp,
+	unsigned long addr0, unsigned long len, unsigned long pgoff,
+	unsigned long flags, enum mmap_allocation_direction dir)
+{
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+	unsigned long addr = addr0;
+	int do_color_align;
+
+	if (unlikely(len > TASK_SIZE))
+		return -ENOMEM;
+
+	if (flags & MAP_FIXED) {
+		/* Even MAP_FIXED mappings must reside within TASK_SIZE */
+		if (TASK_SIZE - len < addr)
+			return -EINVAL;
+
+		/*
+		 * We do not accept a shared mapping if it would violate
+		 * cache aliasing constraints.
+		 */
+		if ((flags & MAP_SHARED) &&
+		    ((addr - (pgoff << PAGE_SHIFT)) & shm_align_mask))
+			return -EINVAL;
+		return addr;
+	}
+
+	do_color_align = 0;
+	if (filp || (flags & MAP_SHARED))
+		do_color_align = 1;
+
+	/* requesting a specific address */
+	if (addr) {
+		if (do_color_align)
+			addr = COLOUR_ALIGN(addr, pgoff);
+		else
+			addr = PAGE_ALIGN(addr);
+
+		vma = find_vma(mm, addr);
+		if (TASK_SIZE - len >= addr &&
+		    (!vma || addr + len <= vma->vm_start))
+			return addr;
+	}
+
+	if (dir == UP) {
+		addr = mm->mmap_base;
+		if (do_color_align)
+			addr = COLOUR_ALIGN(addr, pgoff);
+		else
+			addr = PAGE_ALIGN(addr);
+
+		for (vma = find_vma(current->mm, addr); ; vma = vma->vm_next) {
+			/* At this point:  (!vma || addr < vma->vm_end). */
+			if (TASK_SIZE - len < addr)
+				return -ENOMEM;
+			if (!vma || addr + len <= vma->vm_start)
+				return addr;
+			addr = vma->vm_end;
+			if (do_color_align)
+				addr = COLOUR_ALIGN(addr, pgoff);
+		 }
+	 } else {
+		/* check if free_area_cache is useful for us */
+		if (len <= mm->cached_hole_size) {
+			mm->cached_hole_size = 0;
+			mm->free_area_cache = mm->mmap_base;
+		}
+
+		/*
+		 * either no address requested, or the mapping can't fit into
+		 * the requested address hole
+		 */
+		addr = mm->free_area_cache;
+		if (do_color_align) {
+			unsigned long base =
+				COLOUR_ALIGN_DOWN(addr - len, pgoff);
+			addr = base + len;
+		}
+
+		/* make sure it can fit in the remaining address space */
+		if (likely(addr > len)) {
+			vma = find_vma(mm, addr - len);
+			if (!vma || addr <= vma->vm_start) {
+				/* cache the address as a hint for next time */
+				return mm->free_area_cache = addr - len;
+			}
+		}
+
+		if (unlikely(mm->mmap_base < len))
+			goto bottomup;
+
+		addr = mm->mmap_base - len;
+		if (do_color_align)
+			addr = COLOUR_ALIGN_DOWN(addr, pgoff);
+
+		do {
+			/*
+			 * Lookup failure means no vma is above this address,
+			 * else if new region fits below vma->vm_start,
+			 * return with success:
+			 */
+			vma = find_vma(mm, addr);
+			if (likely(!vma || addr + len <= vma->vm_start)) {
+				/* cache the address as a hint for next time */
+				return mm->free_area_cache = addr;
+			}
+
+			/* remember the largest hole we saw so far */
+			if (addr + mm->cached_hole_size < vma->vm_start)
+				mm->cached_hole_size = vma->vm_start - addr;
+
+			/* try just below the current vma->vm_start */
+			addr = vma->vm_start - len;
+			if (do_color_align)
+				addr = COLOUR_ALIGN_DOWN(addr, pgoff);
+		} while (likely(len < vma->vm_start));
+
+bottomup:
+		/*
+		 * A failed mmap() very likely causes application failure,
+		 * so fall back to the bottom-up function here. This scenario
+		 * can happen with large stack limits and large mmap()
+		 * allocations.
+		 */
+		mm->cached_hole_size = ~0UL;
+		mm->free_area_cache = TASK_UNMAPPED_BASE;
+		addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags);
+		/*
+		 * Restore the topdown base:
+		 */
+		mm->free_area_cache = mm->mmap_base;
+		mm->cached_hole_size = ~0UL;
+
+		return addr;
+	}
+}
+
+unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr0,
+	unsigned long len, unsigned long pgoff, unsigned long flags)
+{
+	return arch_get_unmapped_area_common(filp,
+			addr0, len, pgoff, flags, UP);
+}
+
+/*
+ * There is no need to export this but sched.h declares the function as
+ * extern so making it static here results in an error.
+ */
+unsigned long arch_get_unmapped_area_topdown(struct file *filp,
+	unsigned long addr0, unsigned long len, unsigned long pgoff,
+	unsigned long flags)
+{
+	return arch_get_unmapped_area_common(filp,
+			addr0, len, pgoff, flags, DOWN);
+}
+
+void arch_pick_mmap_layout(struct mm_struct *mm)
+{
+	unsigned long random_factor = 0UL;
+
+	if (current->flags & PF_RANDOMIZE) {
+		random_factor = get_random_int();
+		random_factor = random_factor << PAGE_SHIFT;
+		if (TASK_IS_32BIT_ADDR)
+			random_factor &= 0xfffffful;
+		else
+			random_factor &= 0xffffffful;
+	}
+
+	if (mmap_is_legacy()) {
+		mm->mmap_base = TASK_UNMAPPED_BASE + random_factor;
+		mm->get_unmapped_area = arch_get_unmapped_area;
+		mm->unmap_area = arch_unmap_area;
+	} else {
+		mm->mmap_base = mmap_base(random_factor);
+		mm->get_unmapped_area = arch_get_unmapped_area_topdown;
+		mm->unmap_area = arch_unmap_area_topdown;
+	}
+}
+
+static inline unsigned long brk_rnd(void)
+{
+	unsigned long rnd = get_random_int();
+
+	rnd = rnd << PAGE_SHIFT;
+	/* 8MB for 32bit, 256MB for 64bit */
+	if (TASK_IS_32BIT_ADDR)
+		rnd = rnd & 0x7ffffful;
+	else
+		rnd = rnd & 0xffffffful;
+
+	return rnd;
+}
+
+unsigned long arch_randomize_brk(struct mm_struct *mm)
+{
+	unsigned long base = mm->brk;
+	unsigned long ret;
+
+	ret = PAGE_ALIGN(base + brk_rnd());
+
+	if (ret < mm->brk)
+		return mm->brk;
+
+	return ret;
+}
diff --git a/arch/mips/mm/page.c b/arch/mips/mm/page.c
new file mode 100644
index 00000000..cc0b6268
--- /dev/null
+++ b/arch/mips/mm/page.c
@@ -0,0 +1,690 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2003, 04, 05 Ralf Baechle (ralf@linux-mips.org)
+ * Copyright (C) 2007  Maciej W. Rozycki
+ * Copyright (C) 2008  Thiemo Seufer
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/proc_fs.h>
+
+#include <asm/bugs.h>
+#include <asm/cacheops.h>
+#include <asm/inst.h>
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/prefetch.h>
+#include <asm/bootinfo.h>
+#include <asm/mipsregs.h>
+#include <asm/mmu_context.h>
+#include <asm/cpu.h>
+#include <asm/war.h>
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+#include <asm/sibyte/sb1250.h>
+#include <asm/sibyte/sb1250_regs.h>
+#include <asm/sibyte/sb1250_dma.h>
+#endif
+
+#include <asm/uasm.h>
+
+/* Registers used in the assembled routines. */
+#define ZERO 0
+#define AT 2
+#define A0 4
+#define A1 5
+#define A2 6
+#define T0 8
+#define T1 9
+#define T2 10
+#define T3 11
+#define T9 25
+#define RA 31
+
+/* Handle labels (which must be positive integers). */
+enum label_id {
+	label_clear_nopref = 1,
+	label_clear_pref,
+	label_copy_nopref,
+	label_copy_pref_both,
+	label_copy_pref_store,
+};
+
+UASM_L_LA(_clear_nopref)
+UASM_L_LA(_clear_pref)
+UASM_L_LA(_copy_nopref)
+UASM_L_LA(_copy_pref_both)
+UASM_L_LA(_copy_pref_store)
+
+/* We need one branch and therefore one relocation per target label. */
+static struct uasm_label __cpuinitdata labels[5];
+static struct uasm_reloc __cpuinitdata relocs[5];
+
+#define cpu_is_r4600_v1_x()	((read_c0_prid() & 0xfffffff0) == 0x00002010)
+#define cpu_is_r4600_v2_x()	((read_c0_prid() & 0xfffffff0) == 0x00002020)
+
+/*
+ * Maximum sizes:
+ *
+ * R4000 128 bytes S-cache:		0x058 bytes
+ * R4600 v1.7:				0x05c bytes
+ * R4600 v2.0:				0x060 bytes
+ * With prefetching, 16 word strides	0x120 bytes
+ */
+
+static u32 clear_page_array[0x120 / 4];
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+void clear_page_cpu(void *page) __attribute__((alias("clear_page_array")));
+#else
+void clear_page(void *page) __attribute__((alias("clear_page_array")));
+#endif
+
+EXPORT_SYMBOL(clear_page);
+
+/*
+ * Maximum sizes:
+ *
+ * R4000 128 bytes S-cache:		0x11c bytes
+ * R4600 v1.7:				0x080 bytes
+ * R4600 v2.0:				0x07c bytes
+ * With prefetching, 16 word strides	0x540 bytes
+ */
+static u32 copy_page_array[0x540 / 4];
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+void
+copy_page_cpu(void *to, void *from) __attribute__((alias("copy_page_array")));
+#else
+void copy_page(void *to, void *from) __attribute__((alias("copy_page_array")));
+#endif
+
+EXPORT_SYMBOL(copy_page);
+
+
+static int pref_bias_clear_store __cpuinitdata;
+static int pref_bias_copy_load __cpuinitdata;
+static int pref_bias_copy_store __cpuinitdata;
+
+static u32 pref_src_mode __cpuinitdata;
+static u32 pref_dst_mode __cpuinitdata;
+
+static int clear_word_size __cpuinitdata;
+static int copy_word_size __cpuinitdata;
+
+static int half_clear_loop_size __cpuinitdata;
+static int half_copy_loop_size __cpuinitdata;
+
+static int cache_line_size __cpuinitdata;
+#define cache_line_mask() (cache_line_size - 1)
+
+static inline void __cpuinit
+pg_addiu(u32 **buf, unsigned int reg1, unsigned int reg2, unsigned int off)
+{
+	if (cpu_has_64bit_gp_regs && DADDI_WAR && r4k_daddiu_bug()) {
+		if (off > 0x7fff) {
+			uasm_i_lui(buf, T9, uasm_rel_hi(off));
+			uasm_i_addiu(buf, T9, T9, uasm_rel_lo(off));
+		} else
+			uasm_i_addiu(buf, T9, ZERO, off);
+		uasm_i_daddu(buf, reg1, reg2, T9);
+	} else {
+		if (off > 0x7fff) {
+			uasm_i_lui(buf, T9, uasm_rel_hi(off));
+			uasm_i_addiu(buf, T9, T9, uasm_rel_lo(off));
+			UASM_i_ADDU(buf, reg1, reg2, T9);
+		} else
+			UASM_i_ADDIU(buf, reg1, reg2, off);
+	}
+}
+
+static void __cpuinit set_prefetch_parameters(void)
+{
+	if (cpu_has_64bit_gp_regs || cpu_has_64bit_zero_reg)
+		clear_word_size = 8;
+	else
+		clear_word_size = 4;
+
+	if (cpu_has_64bit_gp_regs)
+		copy_word_size = 8;
+	else
+		copy_word_size = 4;
+
+	/*
+	 * The pref's used here are using "streaming" hints, which cause the
+	 * copied data to be kicked out of the cache sooner.  A page copy often
+	 * ends up copying a lot more data than is commonly used, so this seems
+	 * to make sense in terms of reducing cache pollution, but I've no real
+	 * performance data to back this up.
+	 */
+	if (cpu_has_prefetch) {
+		/*
+		 * XXX: Most prefetch bias values in here are based on
+		 * guesswork.
+		 */
+		cache_line_size = cpu_dcache_line_size();
+		switch (current_cpu_type()) {
+		case CPU_R5500:
+		case CPU_TX49XX:
+			/* These processors only support the Pref_Load. */
+			pref_bias_copy_load = 256;
+			break;
+
+		case CPU_RM9000:
+			/*
+			 * As a workaround for erratum G105 which make the
+			 * PrepareForStore hint unusable we fall back to
+			 * StoreRetained on the RM9000.  Once it is known which
+			 * versions of the RM9000 we'll be able to condition-
+			 * alize this.
+			 */
+
+		case CPU_R10000:
+		case CPU_R12000:
+		case CPU_R14000:
+			/*
+			 * Those values have been experimentally tuned for an
+			 * Origin 200.
+			 */
+			pref_bias_clear_store = 512;
+			pref_bias_copy_load = 256;
+			pref_bias_copy_store = 256;
+			pref_src_mode = Pref_LoadStreamed;
+			pref_dst_mode = Pref_StoreStreamed;
+			break;
+
+		case CPU_SB1:
+		case CPU_SB1A:
+			pref_bias_clear_store = 128;
+			pref_bias_copy_load = 128;
+			pref_bias_copy_store = 128;
+			/*
+			 * SB1 pass1 Pref_LoadStreamed/Pref_StoreStreamed
+			 * hints are broken.
+			 */
+			if (current_cpu_type() == CPU_SB1 &&
+			    (current_cpu_data.processor_id & 0xff) < 0x02) {
+				pref_src_mode = Pref_Load;
+				pref_dst_mode = Pref_Store;
+			} else {
+				pref_src_mode = Pref_LoadStreamed;
+				pref_dst_mode = Pref_StoreStreamed;
+			}
+			break;
+
+		default:
+			pref_bias_clear_store = 128;
+			pref_bias_copy_load = 256;
+			pref_bias_copy_store = 128;
+			pref_src_mode = Pref_LoadStreamed;
+			pref_dst_mode = Pref_PrepareForStore;
+			break;
+		}
+	} else {
+		if (cpu_has_cache_cdex_s)
+			cache_line_size = cpu_scache_line_size();
+		else if (cpu_has_cache_cdex_p)
+			cache_line_size = cpu_dcache_line_size();
+	}
+	/*
+	 * Too much unrolling will overflow the available space in
+	 * clear_space_array / copy_page_array.
+	 */
+	half_clear_loop_size = min(16 * clear_word_size,
+				   max(cache_line_size >> 1,
+				       4 * clear_word_size));
+	half_copy_loop_size = min(16 * copy_word_size,
+				  max(cache_line_size >> 1,
+				      4 * copy_word_size));
+}
+
+static void __cpuinit build_clear_store(u32 **buf, int off)
+{
+	if (cpu_has_64bit_gp_regs || cpu_has_64bit_zero_reg) {
+		uasm_i_sd(buf, ZERO, off, A0);
+	} else {
+		uasm_i_sw(buf, ZERO, off, A0);
+	}
+}
+
+static inline void __cpuinit build_clear_pref(u32 **buf, int off)
+{
+	if (off & cache_line_mask())
+		return;
+
+	if (pref_bias_clear_store) {
+		uasm_i_pref(buf, pref_dst_mode, pref_bias_clear_store + off,
+			    A0);
+	} else if (cache_line_size == (half_clear_loop_size << 1)) {
+		if (cpu_has_cache_cdex_s) {
+			uasm_i_cache(buf, Create_Dirty_Excl_SD, off, A0);
+		} else if (cpu_has_cache_cdex_p) {
+			if (R4600_V1_HIT_CACHEOP_WAR && cpu_is_r4600_v1_x()) {
+				uasm_i_nop(buf);
+				uasm_i_nop(buf);
+				uasm_i_nop(buf);
+				uasm_i_nop(buf);
+			}
+
+			if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+				uasm_i_lw(buf, ZERO, ZERO, AT);
+
+			uasm_i_cache(buf, Create_Dirty_Excl_D, off, A0);
+		}
+		}
+}
+
+void __cpuinit build_clear_page(void)
+{
+	int off;
+	u32 *buf = (u32 *)&clear_page_array;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	int i;
+
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	set_prefetch_parameters();
+
+	/*
+	 * This algorithm makes the following assumptions:
+	 *   - The prefetch bias is a multiple of 2 words.
+	 *   - The prefetch bias is less than one page.
+	 */
+	BUG_ON(pref_bias_clear_store % (2 * clear_word_size));
+	BUG_ON(PAGE_SIZE < pref_bias_clear_store);
+
+	off = PAGE_SIZE - pref_bias_clear_store;
+	if (off > 0xffff || !pref_bias_clear_store)
+		pg_addiu(&buf, A2, A0, off);
+	else
+		uasm_i_ori(&buf, A2, A0, off);
+
+	if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+		uasm_i_lui(&buf, AT, 0xa000);
+
+	off = cache_line_size ? min(8, pref_bias_clear_store / cache_line_size)
+	                        * cache_line_size : 0;
+	while (off) {
+		build_clear_pref(&buf, -off);
+		off -= cache_line_size;
+	}
+	uasm_l_clear_pref(&l, buf);
+	do {
+		build_clear_pref(&buf, off);
+		build_clear_store(&buf, off);
+		off += clear_word_size;
+	} while (off < half_clear_loop_size);
+	pg_addiu(&buf, A0, A0, 2 * off);
+	off = -off;
+	do {
+		build_clear_pref(&buf, off);
+		if (off == -clear_word_size)
+			uasm_il_bne(&buf, &r, A0, A2, label_clear_pref);
+		build_clear_store(&buf, off);
+		off += clear_word_size;
+	} while (off < 0);
+
+	if (pref_bias_clear_store) {
+		pg_addiu(&buf, A2, A0, pref_bias_clear_store);
+		uasm_l_clear_nopref(&l, buf);
+		off = 0;
+		do {
+			build_clear_store(&buf, off);
+			off += clear_word_size;
+		} while (off < half_clear_loop_size);
+		pg_addiu(&buf, A0, A0, 2 * off);
+		off = -off;
+		do {
+			if (off == -clear_word_size)
+				uasm_il_bne(&buf, &r, A0, A2,
+					    label_clear_nopref);
+			build_clear_store(&buf, off);
+			off += clear_word_size;
+		} while (off < 0);
+	}
+
+	uasm_i_jr(&buf, RA);
+	uasm_i_nop(&buf);
+
+	BUG_ON(buf > clear_page_array + ARRAY_SIZE(clear_page_array));
+
+	uasm_resolve_relocs(relocs, labels);
+
+	pr_debug("Synthesized clear page handler (%u instructions).\n",
+		 (u32)(buf - clear_page_array));
+
+	pr_debug("\t.set push\n");
+	pr_debug("\t.set noreorder\n");
+	for (i = 0; i < (buf - clear_page_array); i++)
+		pr_debug("\t.word 0x%08x\n", clear_page_array[i]);
+	pr_debug("\t.set pop\n");
+}
+
+static void __cpuinit build_copy_load(u32 **buf, int reg, int off)
+{
+	if (cpu_has_64bit_gp_regs) {
+		uasm_i_ld(buf, reg, off, A1);
+	} else {
+		uasm_i_lw(buf, reg, off, A1);
+	}
+}
+
+static void __cpuinit build_copy_store(u32 **buf, int reg, int off)
+{
+	if (cpu_has_64bit_gp_regs) {
+		uasm_i_sd(buf, reg, off, A0);
+	} else {
+		uasm_i_sw(buf, reg, off, A0);
+	}
+}
+
+static inline void build_copy_load_pref(u32 **buf, int off)
+{
+	if (off & cache_line_mask())
+		return;
+
+	if (pref_bias_copy_load)
+		uasm_i_pref(buf, pref_src_mode, pref_bias_copy_load + off, A1);
+}
+
+static inline void build_copy_store_pref(u32 **buf, int off)
+{
+	if (off & cache_line_mask())
+		return;
+
+	if (pref_bias_copy_store) {
+		uasm_i_pref(buf, pref_dst_mode, pref_bias_copy_store + off,
+			    A0);
+	} else if (cache_line_size == (half_copy_loop_size << 1)) {
+		if (cpu_has_cache_cdex_s) {
+			uasm_i_cache(buf, Create_Dirty_Excl_SD, off, A0);
+		} else if (cpu_has_cache_cdex_p) {
+			if (R4600_V1_HIT_CACHEOP_WAR && cpu_is_r4600_v1_x()) {
+				uasm_i_nop(buf);
+				uasm_i_nop(buf);
+				uasm_i_nop(buf);
+				uasm_i_nop(buf);
+			}
+
+			if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+				uasm_i_lw(buf, ZERO, ZERO, AT);
+
+			uasm_i_cache(buf, Create_Dirty_Excl_D, off, A0);
+		}
+	}
+}
+
+void __cpuinit build_copy_page(void)
+{
+	int off;
+	u32 *buf = (u32 *)&copy_page_array;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	int i;
+
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	set_prefetch_parameters();
+
+	/*
+	 * This algorithm makes the following assumptions:
+	 *   - All prefetch biases are multiples of 8 words.
+	 *   - The prefetch biases are less than one page.
+	 *   - The store prefetch bias isn't greater than the load
+	 *     prefetch bias.
+	 */
+	BUG_ON(pref_bias_copy_load % (8 * copy_word_size));
+	BUG_ON(pref_bias_copy_store % (8 * copy_word_size));
+	BUG_ON(PAGE_SIZE < pref_bias_copy_load);
+	BUG_ON(pref_bias_copy_store > pref_bias_copy_load);
+
+	off = PAGE_SIZE - pref_bias_copy_load;
+	if (off > 0xffff || !pref_bias_copy_load)
+		pg_addiu(&buf, A2, A0, off);
+	else
+		uasm_i_ori(&buf, A2, A0, off);
+
+	if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+		uasm_i_lui(&buf, AT, 0xa000);
+
+	off = cache_line_size ? min(8, pref_bias_copy_load / cache_line_size) *
+	                        cache_line_size : 0;
+	while (off) {
+		build_copy_load_pref(&buf, -off);
+		off -= cache_line_size;
+	}
+	off = cache_line_size ? min(8, pref_bias_copy_store / cache_line_size) *
+	                        cache_line_size : 0;
+	while (off) {
+		build_copy_store_pref(&buf, -off);
+		off -= cache_line_size;
+	}
+	uasm_l_copy_pref_both(&l, buf);
+	do {
+		build_copy_load_pref(&buf, off);
+		build_copy_load(&buf, T0, off);
+		build_copy_load_pref(&buf, off + copy_word_size);
+		build_copy_load(&buf, T1, off + copy_word_size);
+		build_copy_load_pref(&buf, off + 2 * copy_word_size);
+		build_copy_load(&buf, T2, off + 2 * copy_word_size);
+		build_copy_load_pref(&buf, off + 3 * copy_word_size);
+		build_copy_load(&buf, T3, off + 3 * copy_word_size);
+		build_copy_store_pref(&buf, off);
+		build_copy_store(&buf, T0, off);
+		build_copy_store_pref(&buf, off + copy_word_size);
+		build_copy_store(&buf, T1, off + copy_word_size);
+		build_copy_store_pref(&buf, off + 2 * copy_word_size);
+		build_copy_store(&buf, T2, off + 2 * copy_word_size);
+		build_copy_store_pref(&buf, off + 3 * copy_word_size);
+		build_copy_store(&buf, T3, off + 3 * copy_word_size);
+		off += 4 * copy_word_size;
+	} while (off < half_copy_loop_size);
+	pg_addiu(&buf, A1, A1, 2 * off);
+	pg_addiu(&buf, A0, A0, 2 * off);
+	off = -off;
+	do {
+		build_copy_load_pref(&buf, off);
+		build_copy_load(&buf, T0, off);
+		build_copy_load_pref(&buf, off + copy_word_size);
+		build_copy_load(&buf, T1, off + copy_word_size);
+		build_copy_load_pref(&buf, off + 2 * copy_word_size);
+		build_copy_load(&buf, T2, off + 2 * copy_word_size);
+		build_copy_load_pref(&buf, off + 3 * copy_word_size);
+		build_copy_load(&buf, T3, off + 3 * copy_word_size);
+		build_copy_store_pref(&buf, off);
+		build_copy_store(&buf, T0, off);
+		build_copy_store_pref(&buf, off + copy_word_size);
+		build_copy_store(&buf, T1, off + copy_word_size);
+		build_copy_store_pref(&buf, off + 2 * copy_word_size);
+		build_copy_store(&buf, T2, off + 2 * copy_word_size);
+		build_copy_store_pref(&buf, off + 3 * copy_word_size);
+		if (off == -(4 * copy_word_size))
+			uasm_il_bne(&buf, &r, A2, A0, label_copy_pref_both);
+		build_copy_store(&buf, T3, off + 3 * copy_word_size);
+		off += 4 * copy_word_size;
+	} while (off < 0);
+
+	if (pref_bias_copy_load - pref_bias_copy_store) {
+		pg_addiu(&buf, A2, A0,
+			 pref_bias_copy_load - pref_bias_copy_store);
+		uasm_l_copy_pref_store(&l, buf);
+		off = 0;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store_pref(&buf, off);
+			build_copy_store(&buf, T0, off);
+			build_copy_store_pref(&buf, off + copy_word_size);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store_pref(&buf, off + 2 * copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			build_copy_store_pref(&buf, off + 3 * copy_word_size);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < half_copy_loop_size);
+		pg_addiu(&buf, A1, A1, 2 * off);
+		pg_addiu(&buf, A0, A0, 2 * off);
+		off = -off;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store_pref(&buf, off);
+			build_copy_store(&buf, T0, off);
+			build_copy_store_pref(&buf, off + copy_word_size);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store_pref(&buf, off + 2 * copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			build_copy_store_pref(&buf, off + 3 * copy_word_size);
+			if (off == -(4 * copy_word_size))
+				uasm_il_bne(&buf, &r, A2, A0,
+					    label_copy_pref_store);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < 0);
+	}
+
+	if (pref_bias_copy_store) {
+		pg_addiu(&buf, A2, A0, pref_bias_copy_store);
+		uasm_l_copy_nopref(&l, buf);
+		off = 0;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store(&buf, T0, off);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < half_copy_loop_size);
+		pg_addiu(&buf, A1, A1, 2 * off);
+		pg_addiu(&buf, A0, A0, 2 * off);
+		off = -off;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store(&buf, T0, off);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			if (off == -(4 * copy_word_size))
+				uasm_il_bne(&buf, &r, A2, A0,
+					    label_copy_nopref);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < 0);
+	}
+
+	uasm_i_jr(&buf, RA);
+	uasm_i_nop(&buf);
+
+	BUG_ON(buf > copy_page_array + ARRAY_SIZE(copy_page_array));
+
+	uasm_resolve_relocs(relocs, labels);
+
+	pr_debug("Synthesized copy page handler (%u instructions).\n",
+		 (u32)(buf - copy_page_array));
+
+	pr_debug("\t.set push\n");
+	pr_debug("\t.set noreorder\n");
+	for (i = 0; i < (buf - copy_page_array); i++)
+		pr_debug("\t.word 0x%08x\n", copy_page_array[i]);
+	pr_debug("\t.set pop\n");
+}
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+
+/*
+ * Pad descriptors to cacheline, since each is exclusively owned by a
+ * particular CPU.
+ */
+struct dmadscr {
+	u64 dscr_a;
+	u64 dscr_b;
+	u64 pad_a;
+	u64 pad_b;
+} ____cacheline_aligned_in_smp page_descr[DM_NUM_CHANNELS];
+
+void sb1_dma_init(void)
+{
+	int i;
+
+	for (i = 0; i < DM_NUM_CHANNELS; i++) {
+		const u64 base_val = CPHYSADDR((unsigned long)&page_descr[i]) |
+				     V_DM_DSCR_BASE_RINGSZ(1);
+		void *base_reg = IOADDR(A_DM_REGISTER(i, R_DM_DSCR_BASE));
+
+		__raw_writeq(base_val, base_reg);
+		__raw_writeq(base_val | M_DM_DSCR_BASE_RESET, base_reg);
+		__raw_writeq(base_val | M_DM_DSCR_BASE_ENABL, base_reg);
+	}
+}
+
+void clear_page(void *page)
+{
+	u64 to_phys = CPHYSADDR((unsigned long)page);
+	unsigned int cpu = smp_processor_id();
+
+	/* if the page is not in KSEG0, use old way */
+	if ((long)KSEGX((unsigned long)page) != (long)CKSEG0)
+		return clear_page_cpu(page);
+
+	page_descr[cpu].dscr_a = to_phys | M_DM_DSCRA_ZERO_MEM |
+				 M_DM_DSCRA_L2C_DEST | M_DM_DSCRA_INTERRUPT;
+	page_descr[cpu].dscr_b = V_DM_DSCRB_SRC_LENGTH(PAGE_SIZE);
+	__raw_writeq(1, IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_COUNT)));
+
+	/*
+	 * Don't really want to do it this way, but there's no
+	 * reliable way to delay completion detection.
+	 */
+	while (!(__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE_DEBUG)))
+		 & M_DM_DSCR_BASE_INTERRUPT))
+		;
+	__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE)));
+}
+
+void copy_page(void *to, void *from)
+{
+	u64 from_phys = CPHYSADDR((unsigned long)from);
+	u64 to_phys = CPHYSADDR((unsigned long)to);
+	unsigned int cpu = smp_processor_id();
+
+	/* if any page is not in KSEG0, use old way */
+	if ((long)KSEGX((unsigned long)to) != (long)CKSEG0
+	    || (long)KSEGX((unsigned long)from) != (long)CKSEG0)
+		return copy_page_cpu(to, from);
+
+	page_descr[cpu].dscr_a = to_phys | M_DM_DSCRA_L2C_DEST |
+				 M_DM_DSCRA_INTERRUPT;
+	page_descr[cpu].dscr_b = from_phys | V_DM_DSCRB_SRC_LENGTH(PAGE_SIZE);
+	__raw_writeq(1, IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_COUNT)));
+
+	/*
+	 * Don't really want to do it this way, but there's no
+	 * reliable way to delay completion detection.
+	 */
+	while (!(__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE_DEBUG)))
+		 & M_DM_DSCR_BASE_INTERRUPT))
+		;
+	__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE)));
+}
+
+#endif /* CONFIG_SIBYTE_DMA_PAGEOPS */
diff --git a/arch/mips/mm/pgtable-32.c b/arch/mips/mm/pgtable-32.c
new file mode 100644
index 00000000..adc6911b
--- /dev/null
+++ b/arch/mips/mm/pgtable-32.c
@@ -0,0 +1,70 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2003 by Ralf Baechle
+ */
+#include <linux/init.h>
+#include <linux/mm.h>
+#include <linux/bootmem.h>
+#include <linux/highmem.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <asm/pgalloc.h>
+
+void pgd_init(unsigned long page)
+{
+	unsigned long *p = (unsigned long *) page;
+	int i;
+
+	for (i = 0; i < USER_PTRS_PER_PGD; i+=8) {
+		p[i + 0] = (unsigned long) invalid_pte_table;
+		p[i + 1] = (unsigned long) invalid_pte_table;
+		p[i + 2] = (unsigned long) invalid_pte_table;
+		p[i + 3] = (unsigned long) invalid_pte_table;
+		p[i + 4] = (unsigned long) invalid_pte_table;
+		p[i + 5] = (unsigned long) invalid_pte_table;
+		p[i + 6] = (unsigned long) invalid_pte_table;
+		p[i + 7] = (unsigned long) invalid_pte_table;
+	}
+}
+
+void __init pagetable_init(void)
+{
+	unsigned long vaddr;
+	pgd_t *pgd_base;
+#ifdef CONFIG_HIGHMEM
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+#endif
+
+	/* Initialize the entire pgd.  */
+	pgd_init((unsigned long)swapper_pg_dir);
+	pgd_init((unsigned long)swapper_pg_dir
+		 + sizeof(pgd_t) * USER_PTRS_PER_PGD);
+
+	pgd_base = swapper_pg_dir;
+
+	/*
+	 * Fixed mappings:
+	 */
+	vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK;
+	fixrange_init(vaddr, vaddr + FIXADDR_SIZE, pgd_base);
+
+#ifdef CONFIG_HIGHMEM
+	/*
+	 * Permanent kmaps:
+	 */
+	vaddr = PKMAP_BASE;
+	fixrange_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base);
+
+	pgd = swapper_pg_dir + __pgd_offset(vaddr);
+	pud = pud_offset(pgd, vaddr);
+	pmd = pmd_offset(pud, vaddr);
+	pte = pte_offset_kernel(pmd, vaddr);
+	pkmap_page_table = pte;
+#endif
+}
diff --git a/arch/mips/mm/pgtable-64.c b/arch/mips/mm/pgtable-64.c
new file mode 100644
index 00000000..cda4e300
--- /dev/null
+++ b/arch/mips/mm/pgtable-64.c
@@ -0,0 +1,80 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1999, 2000 by Silicon Graphics
+ * Copyright (C) 2003 by Ralf Baechle
+ */
+#include <linux/init.h>
+#include <linux/mm.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <asm/pgalloc.h>
+
+void pgd_init(unsigned long page)
+{
+	unsigned long *p, *end;
+	unsigned long entry;
+
+#ifdef __PAGETABLE_PMD_FOLDED
+	entry = (unsigned long)invalid_pte_table;
+#else
+	entry = (unsigned long)invalid_pmd_table;
+#endif
+
+ 	p = (unsigned long *) page;
+	end = p + PTRS_PER_PGD;
+
+	while (p < end) {
+		p[0] = entry;
+		p[1] = entry;
+		p[2] = entry;
+		p[3] = entry;
+		p[4] = entry;
+		p[5] = entry;
+		p[6] = entry;
+		p[7] = entry;
+		p += 8;
+	}
+}
+
+#ifndef __PAGETABLE_PMD_FOLDED
+void pmd_init(unsigned long addr, unsigned long pagetable)
+{
+	unsigned long *p, *end;
+
+ 	p = (unsigned long *) addr;
+	end = p + PTRS_PER_PMD;
+
+	while (p < end) {
+		p[0] = pagetable;
+		p[1] = pagetable;
+		p[2] = pagetable;
+		p[3] = pagetable;
+		p[4] = pagetable;
+		p[5] = pagetable;
+		p[6] = pagetable;
+		p[7] = pagetable;
+		p += 8;
+	}
+}
+#endif
+
+void __init pagetable_init(void)
+{
+	unsigned long vaddr;
+	pgd_t *pgd_base;
+
+	/* Initialize the entire pgd.  */
+	pgd_init((unsigned long)swapper_pg_dir);
+#ifndef __PAGETABLE_PMD_FOLDED
+	pmd_init((unsigned long)invalid_pmd_table, (unsigned long)invalid_pte_table);
+#endif
+	pgd_base = swapper_pg_dir;
+	/*
+	 * Fixed mappings:
+	 */
+	vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK;
+	fixrange_init(vaddr, vaddr + FIXADDR_SIZE, pgd_base);
+}
diff --git a/arch/mips/mm/sc-ip22.c b/arch/mips/mm/sc-ip22.c
new file mode 100644
index 00000000..1eb708ef
--- /dev/null
+++ b/arch/mips/mm/sc-ip22.c
@@ -0,0 +1,176 @@
+/*
+ * sc-ip22.c: Indy cache management functions.
+ *
+ * Copyright (C) 1997, 2001 Ralf Baechle (ralf@gnu.org),
+ * derived from r4xx0.c by David S. Miller (davem@davemloft.net).
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+
+#include <asm/bcache.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/bootinfo.h>
+#include <asm/sgi/ip22.h>
+#include <asm/sgi/mc.h>
+
+/* Secondary cache size in bytes, if present.  */
+static unsigned long scache_size;
+
+#undef DEBUG_CACHE
+
+#define SC_SIZE 0x00080000
+#define SC_LINE 32
+#define CI_MASK (SC_SIZE - SC_LINE)
+#define SC_INDEX(n) ((n) & CI_MASK)
+
+static inline void indy_sc_wipe(unsigned long first, unsigned long last)
+{
+	unsigned long tmp;
+
+	__asm__ __volatile__(
+	".set\tpush\t\t\t# indy_sc_wipe\n\t"
+	".set\tnoreorder\n\t"
+	".set\tmips3\n\t"
+	".set\tnoat\n\t"
+	"mfc0\t%2, $12\n\t"
+	"li\t$1, 0x80\t\t\t# Go 64 bit\n\t"
+	"mtc0\t$1, $12\n\t"
+
+	"dli\t$1, 0x9000000080000000\n\t"
+	"or\t%0, $1\t\t\t# first line to flush\n\t"
+	"or\t%1, $1\t\t\t# last line to flush\n\t"
+	".set\tat\n\t"
+
+	"1:\tsw\t$0, 0(%0)\n\t"
+	"bne\t%0, %1, 1b\n\t"
+	" daddu\t%0, 32\n\t"
+
+	"mtc0\t%2, $12\t\t\t# Back to 32 bit\n\t"
+	"nop; nop; nop; nop;\n\t"
+	".set\tpop"
+	: "=r" (first), "=r" (last), "=&r" (tmp)
+	: "0" (first), "1" (last));
+}
+
+static void indy_sc_wback_invalidate(unsigned long addr, unsigned long size)
+{
+	unsigned long first_line, last_line;
+	unsigned long flags;
+
+#ifdef DEBUG_CACHE
+	printk("indy_sc_wback_invalidate[%08lx,%08lx]", addr, size);
+#endif
+
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	/* Which lines to flush?  */
+	first_line = SC_INDEX(addr);
+	last_line = SC_INDEX(addr + size - 1);
+
+	local_irq_save(flags);
+	if (first_line <= last_line) {
+		indy_sc_wipe(first_line, last_line);
+		goto out;
+	}
+
+	indy_sc_wipe(first_line, SC_SIZE - SC_LINE);
+	indy_sc_wipe(0, last_line);
+out:
+	local_irq_restore(flags);
+}
+
+static void indy_sc_enable(void)
+{
+	unsigned long addr, tmp1, tmp2;
+
+	/* This is really cool... */
+#ifdef DEBUG_CACHE
+	printk("Enabling R4600 SCACHE\n");
+#endif
+	__asm__ __volatile__(
+	".set\tpush\n\t"
+	".set\tnoreorder\n\t"
+	".set\tmips3\n\t"
+	"mfc0\t%2, $12\n\t"
+	"nop; nop; nop; nop;\n\t"
+	"li\t%1, 0x80\n\t"
+	"mtc0\t%1, $12\n\t"
+	"nop; nop; nop; nop;\n\t"
+	"li\t%0, 0x1\n\t"
+	"dsll\t%0, 31\n\t"
+	"lui\t%1, 0x9000\n\t"
+	"dsll32\t%1, 0\n\t"
+	"or\t%0, %1, %0\n\t"
+	"sb\t$0, 0(%0)\n\t"
+	"mtc0\t$0, $12\n\t"
+	"nop; nop; nop; nop;\n\t"
+	"mtc0\t%2, $12\n\t"
+	"nop; nop; nop; nop;\n\t"
+	".set\tpop"
+	: "=r" (tmp1), "=r" (tmp2), "=r" (addr));
+}
+
+static void indy_sc_disable(void)
+{
+	unsigned long tmp1, tmp2, tmp3;
+
+#ifdef DEBUG_CACHE
+	printk("Disabling R4600 SCACHE\n");
+#endif
+	__asm__ __volatile__(
+	".set\tpush\n\t"
+	".set\tnoreorder\n\t"
+	".set\tmips3\n\t"
+	"li\t%0, 0x1\n\t"
+	"dsll\t%0, 31\n\t"
+	"lui\t%1, 0x9000\n\t"
+	"dsll32\t%1, 0\n\t"
+	"or\t%0, %1, %0\n\t"
+	"mfc0\t%2, $12\n\t"
+	"nop; nop; nop; nop\n\t"
+	"li\t%1, 0x80\n\t"
+	"mtc0\t%1, $12\n\t"
+	"nop; nop; nop; nop\n\t"
+	"sh\t$0, 0(%0)\n\t"
+	"mtc0\t$0, $12\n\t"
+	"nop; nop; nop; nop\n\t"
+	"mtc0\t%2, $12\n\t"
+	"nop; nop; nop; nop\n\t"
+	".set\tpop"
+	: "=r" (tmp1), "=r" (tmp2), "=r" (tmp3));
+}
+
+static inline int __init indy_sc_probe(void)
+{
+	unsigned int size = ip22_eeprom_read(&sgimc->eeprom, 17);
+	if (size == 0)
+		return 0;
+
+	size <<= PAGE_SHIFT;
+	printk(KERN_INFO "R4600/R5000 SCACHE size %dK, linesize 32 bytes.\n",
+	       size >> 10);
+	scache_size = size;
+
+	return 1;
+}
+
+/* XXX Check with wje if the Indy caches can differenciate between
+   writeback + invalidate and just invalidate.  */
+static struct bcache_ops indy_sc_ops = {
+	.bc_enable = indy_sc_enable,
+	.bc_disable = indy_sc_disable,
+	.bc_wback_inv = indy_sc_wback_invalidate,
+	.bc_inv = indy_sc_wback_invalidate
+};
+
+void __cpuinit indy_sc_init(void)
+{
+	if (indy_sc_probe()) {
+		indy_sc_enable();
+		bcops = &indy_sc_ops;
+	}
+}
diff --git a/arch/mips/mm/sc-mips.c b/arch/mips/mm/sc-mips.c
new file mode 100644
index 00000000..93d937b4
--- /dev/null
+++ b/arch/mips/mm/sc-mips.c
@@ -0,0 +1,145 @@
+/*
+ * Copyright (C) 2006 Chris Dearman (chris@mips.com),
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+
+#include <asm/mipsregs.h>
+#include <asm/bcache.h>
+#include <asm/cacheops.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/r4kcache.h>
+
+/*
+ * MIPS32/MIPS64 L2 cache handling
+ */
+
+/*
+ * Writeback and invalidate the secondary cache before DMA.
+ */
+static void mips_sc_wback_inv(unsigned long addr, unsigned long size)
+{
+	blast_scache_range(addr, addr + size);
+}
+
+/*
+ * Invalidate the secondary cache before DMA.
+ */
+static void mips_sc_inv(unsigned long addr, unsigned long size)
+{
+	unsigned long lsize = cpu_scache_line_size();
+	unsigned long almask = ~(lsize - 1);
+
+	cache_op(Hit_Writeback_Inv_SD, addr & almask);
+	cache_op(Hit_Writeback_Inv_SD, (addr + size - 1) & almask);
+	blast_inv_scache_range(addr, addr + size);
+}
+
+static void mips_sc_enable(void)
+{
+	/* L2 cache is permanently enabled */
+}
+
+static void mips_sc_disable(void)
+{
+	/* L2 cache is permanently enabled */
+}
+
+static struct bcache_ops mips_sc_ops = {
+	.bc_enable = mips_sc_enable,
+	.bc_disable = mips_sc_disable,
+	.bc_wback_inv = mips_sc_wback_inv,
+	.bc_inv = mips_sc_inv
+};
+
+/*
+ * Check if the L2 cache controller is activated on a particular platform.
+ * MTI's L2 controller and the L2 cache controller of Broadcom's BMIPS
+ * cores both use c0_config2's bit 12 as "L2 Bypass" bit, that is the
+ * cache being disabled.  However there is no guarantee for this to be
+ * true on all platforms.  In an act of stupidity the spec defined bits
+ * 12..15 as implementation defined so below function will eventually have
+ * to be replaced by a platform specific probe.
+ */
+static inline int mips_sc_is_activated(struct cpuinfo_mips *c)
+{
+	unsigned int config2 = read_c0_config2();
+	unsigned int tmp;
+
+	/* Check the bypass bit (L2B) */
+	switch (c->cputype) {
+	case CPU_34K:
+	case CPU_74K:
+	case CPU_1004K:
+	case CPU_BMIPS5000:
+		if (config2 & (1 << 12))
+			return 0;
+	}
+
+	tmp = (config2 >> 4) & 0x0f;
+	if (0 < tmp && tmp <= 7)
+		c->scache.linesz = 2 << tmp;
+	else
+		return 0;
+	return 1;
+}
+
+static inline int __init mips_sc_probe(void)
+{
+	struct cpuinfo_mips *c = &current_cpu_data;
+	unsigned int config1, config2;
+	unsigned int tmp;
+
+	/* Mark as not present until probe completed */
+	c->scache.flags |= MIPS_CACHE_NOT_PRESENT;
+
+	/* Ignore anything but MIPSxx processors */
+	if (c->isa_level != MIPS_CPU_ISA_M32R1 &&
+	    c->isa_level != MIPS_CPU_ISA_M32R2 &&
+	    c->isa_level != MIPS_CPU_ISA_M64R1 &&
+	    c->isa_level != MIPS_CPU_ISA_M64R2)
+		return 0;
+
+	/* Does this MIPS32/MIPS64 CPU have a config2 register? */
+	config1 = read_c0_config1();
+	if (!(config1 & MIPS_CONF_M))
+		return 0;
+
+	config2 = read_c0_config2();
+
+	if (!mips_sc_is_activated(c))
+		return 0;
+
+	tmp = (config2 >> 8) & 0x0f;
+	if (0 <= tmp && tmp <= 7)
+		c->scache.sets = 64 << tmp;
+	else
+		return 0;
+
+	tmp = (config2 >> 0) & 0x0f;
+	if (0 <= tmp && tmp <= 7)
+		c->scache.ways = tmp + 1;
+	else
+		return 0;
+
+	c->scache.waysize = c->scache.sets * c->scache.linesz;
+	c->scache.waybit = __ffs(c->scache.waysize);
+
+	c->scache.flags &= ~MIPS_CACHE_NOT_PRESENT;
+
+	return 1;
+}
+
+int __cpuinit mips_sc_init(void)
+{
+	int found = mips_sc_probe();
+	if (found) {
+		mips_sc_enable();
+		bcops = &mips_sc_ops;
+	}
+	return found;
+}
diff --git a/arch/mips/mm/sc-r5k.c b/arch/mips/mm/sc-r5k.c
new file mode 100644
index 00000000..8d90ff25
--- /dev/null
+++ b/arch/mips/mm/sc-r5k.c
@@ -0,0 +1,107 @@
+/*
+ * Copyright (C) 1997, 2001 Ralf Baechle (ralf@gnu.org),
+ * derived from r4xx0.c by David S. Miller (davem@davemloft.net).
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+
+#include <asm/mipsregs.h>
+#include <asm/bcache.h>
+#include <asm/cacheops.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/r4kcache.h>
+
+/* Secondary cache size in bytes, if present.  */
+static unsigned long scache_size;
+
+#define SC_LINE 32
+#define SC_PAGE (128*SC_LINE)
+
+static inline void blast_r5000_scache(void)
+{
+	unsigned long start = INDEX_BASE;
+	unsigned long end = start + scache_size;
+
+	while(start < end) {
+		cache_op(R5K_Page_Invalidate_S, start);
+		start += SC_PAGE;
+	}
+}
+
+static void r5k_dma_cache_inv_sc(unsigned long addr, unsigned long size)
+{
+	unsigned long end, a;
+
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	if (size >= scache_size) {
+		blast_r5000_scache();
+		return;
+	}
+
+	/* On the R5000 secondary cache we cannot
+	 * invalidate less than a page at a time.
+	 * The secondary cache is physically indexed, write-through.
+	 */
+	a = addr & ~(SC_PAGE - 1);
+	end = (addr + size - 1) & ~(SC_PAGE - 1);
+	while (a <= end) {
+		cache_op(R5K_Page_Invalidate_S, a);
+		a += SC_PAGE;
+	}
+}
+
+static void r5k_sc_enable(void)
+{
+        unsigned long flags;
+
+	local_irq_save(flags);
+	set_c0_config(R5K_CONF_SE);
+	blast_r5000_scache();
+	local_irq_restore(flags);
+}
+
+static void r5k_sc_disable(void)
+{
+        unsigned long flags;
+
+	local_irq_save(flags);
+	blast_r5000_scache();
+	clear_c0_config(R5K_CONF_SE);
+	local_irq_restore(flags);
+}
+
+static inline int __init r5k_sc_probe(void)
+{
+	unsigned long config = read_c0_config();
+
+	if (config & CONF_SC)
+		return(0);
+
+	scache_size = (512 * 1024) << ((config & R5K_CONF_SS) >> 20);
+
+	printk("R5000 SCACHE size %ldkB, linesize 32 bytes.\n",
+			scache_size >> 10);
+
+	return 1;
+}
+
+static struct bcache_ops r5k_sc_ops = {
+	.bc_enable = r5k_sc_enable,
+	.bc_disable = r5k_sc_disable,
+	.bc_wback_inv = r5k_dma_cache_inv_sc,
+	.bc_inv = r5k_dma_cache_inv_sc
+};
+
+void __cpuinit r5k_sc_init(void)
+{
+	if (r5k_sc_probe()) {
+		r5k_sc_enable();
+		bcops = &r5k_sc_ops;
+	}
+}
diff --git a/arch/mips/mm/sc-rm7k.c b/arch/mips/mm/sc-rm7k.c
new file mode 100644
index 00000000..274af3be
--- /dev/null
+++ b/arch/mips/mm/sc-rm7k.c
@@ -0,0 +1,270 @@
+/*
+ * sc-rm7k.c: RM7000 cache management functions.
+ *
+ * Copyright (C) 1997, 2001, 2003, 2004 Ralf Baechle (ralf@linux-mips.org)
+ */
+
+#undef DEBUG
+
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/bitops.h>
+
+#include <asm/addrspace.h>
+#include <asm/bcache.h>
+#include <asm/cacheops.h>
+#include <asm/mipsregs.h>
+#include <asm/processor.h>
+#include <asm/sections.h>
+#include <asm/cacheflush.h> /* for run_uncached() */
+
+/* Primary cache parameters. */
+#define sc_lsize	32
+#define tc_pagesize	(32*128)
+
+/* Secondary cache parameters. */
+#define scache_size	(256*1024)	/* Fixed to 256KiB on RM7000 */
+
+/* Tertiary cache parameters */
+#define tc_lsize	32
+
+extern unsigned long icache_way_size, dcache_way_size;
+static unsigned long tcache_size;
+
+#include <asm/r4kcache.h>
+
+static int rm7k_tcache_init;
+
+/*
+ * Writeback and invalidate the primary cache dcache before DMA.
+ * (XXX These need to be fixed ...)
+ */
+static void rm7k_sc_wback_inv(unsigned long addr, unsigned long size)
+{
+	unsigned long end, a;
+
+	pr_debug("rm7k_sc_wback_inv[%08lx,%08lx]", addr, size);
+
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	blast_scache_range(addr, addr + size);
+
+	if (!rm7k_tcache_init)
+		return;
+
+	a = addr & ~(tc_pagesize - 1);
+	end = (addr + size - 1) & ~(tc_pagesize - 1);
+	while(1) {
+		invalidate_tcache_page(a);	/* Page_Invalidate_T */
+		if (a == end)
+			break;
+		a += tc_pagesize;
+	}
+}
+
+static void rm7k_sc_inv(unsigned long addr, unsigned long size)
+{
+	unsigned long end, a;
+
+	pr_debug("rm7k_sc_inv[%08lx,%08lx]", addr, size);
+
+	/* Catch bad driver code */
+	BUG_ON(size == 0);
+
+	blast_inv_scache_range(addr, addr + size);
+
+	if (!rm7k_tcache_init)
+		return;
+
+	a = addr & ~(tc_pagesize - 1);
+	end = (addr + size - 1) & ~(tc_pagesize - 1);
+	while(1) {
+		invalidate_tcache_page(a);	/* Page_Invalidate_T */
+		if (a == end)
+			break;
+		a += tc_pagesize;
+	}
+}
+
+static void blast_rm7k_tcache(void)
+{
+	unsigned long start = CKSEG0ADDR(0);
+	unsigned long end = start + tcache_size;
+
+	write_c0_taglo(0);
+
+	while (start < end) {
+		cache_op(Page_Invalidate_T, start);
+		start += tc_pagesize;
+	}
+}
+
+/*
+ * This function is executed in uncached address space.
+ */
+static __cpuinit void __rm7k_tc_enable(void)
+{
+	int i;
+
+	set_c0_config(RM7K_CONF_TE);
+
+	write_c0_taglo(0);
+	write_c0_taghi(0);
+
+	for (i = 0; i < tcache_size; i += tc_lsize)
+		cache_op(Index_Store_Tag_T, CKSEG0ADDR(i));
+}
+
+static __cpuinit void rm7k_tc_enable(void)
+{
+	if (read_c0_config() & RM7K_CONF_TE)
+		return;
+
+	BUG_ON(tcache_size == 0);
+
+	run_uncached(__rm7k_tc_enable);
+}
+
+/*
+ * This function is executed in uncached address space.
+ */
+static __cpuinit void __rm7k_sc_enable(void)
+{
+	int i;
+
+	set_c0_config(RM7K_CONF_SE);
+
+	write_c0_taglo(0);
+	write_c0_taghi(0);
+
+	for (i = 0; i < scache_size; i += sc_lsize)
+		cache_op(Index_Store_Tag_SD, CKSEG0ADDR(i));
+}
+
+static __cpuinit void rm7k_sc_enable(void)
+{
+	if (read_c0_config() & RM7K_CONF_SE)
+		return;
+
+	pr_info("Enabling secondary cache...\n");
+	run_uncached(__rm7k_sc_enable);
+
+	if (rm7k_tcache_init)
+		rm7k_tc_enable();
+}
+
+static void rm7k_tc_disable(void)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	blast_rm7k_tcache();
+	clear_c0_config(RM7K_CONF_TE);
+	local_irq_save(flags);
+}
+
+static void rm7k_sc_disable(void)
+{
+	clear_c0_config(RM7K_CONF_SE);
+
+	if (rm7k_tcache_init)
+		rm7k_tc_disable();
+}
+
+static struct bcache_ops rm7k_sc_ops = {
+	.bc_enable = rm7k_sc_enable,
+	.bc_disable = rm7k_sc_disable,
+	.bc_wback_inv = rm7k_sc_wback_inv,
+	.bc_inv = rm7k_sc_inv
+};
+
+/*
+ * This is a probing function like the one found in c-r4k.c, we look for the
+ * wrap around point with different addresses.
+ */
+static __cpuinit void __probe_tcache(void)
+{
+	unsigned long flags, addr, begin, end, pow2;
+
+	begin = (unsigned long) &_stext;
+	begin  &= ~((8 * 1024 * 1024) - 1);
+	end = begin + (8 * 1024 * 1024);
+
+	local_irq_save(flags);
+
+	set_c0_config(RM7K_CONF_TE);
+
+	/* Fill size-multiple lines with a valid tag */
+	pow2 = (256 * 1024);
+	for (addr = begin; addr <= end; addr = (begin + pow2)) {
+		unsigned long *p = (unsigned long *) addr;
+		__asm__ __volatile__("nop" : : "r" (*p));
+		pow2 <<= 1;
+	}
+
+	/* Load first line with a 0 tag, to check after */
+	write_c0_taglo(0);
+	write_c0_taghi(0);
+	cache_op(Index_Store_Tag_T, begin);
+
+	/* Look for the wrap-around */
+	pow2 = (512 * 1024);
+	for (addr = begin + (512 * 1024); addr <= end; addr = begin + pow2) {
+		cache_op(Index_Load_Tag_T, addr);
+		if (!read_c0_taglo())
+			break;
+		pow2 <<= 1;
+	}
+
+	addr -= begin;
+	tcache_size = addr;
+
+	clear_c0_config(RM7K_CONF_TE);
+
+	local_irq_restore(flags);
+}
+
+void __cpuinit rm7k_sc_init(void)
+{
+	struct cpuinfo_mips *c = &current_cpu_data;
+	unsigned int config = read_c0_config();
+
+	if ((config & RM7K_CONF_SC))
+		return;
+
+	c->scache.linesz = sc_lsize;
+	c->scache.ways = 4;
+	c->scache.waybit= __ffs(scache_size / c->scache.ways);
+	c->scache.waysize = scache_size / c->scache.ways;
+	c->scache.sets = scache_size / (c->scache.linesz * c->scache.ways);
+	printk(KERN_INFO "Secondary cache size %dK, linesize %d bytes.\n",
+	       (scache_size >> 10), sc_lsize);
+
+	if (!(config & RM7K_CONF_SE))
+		rm7k_sc_enable();
+
+	bcops = &rm7k_sc_ops;
+
+	/*
+	 * While we're at it let's deal with the tertiary cache.
+	 */
+
+	rm7k_tcache_init = 0;
+	tcache_size = 0;
+
+	if (config & RM7K_CONF_TC)
+		return;
+
+	/*
+	 * No efficient way to ask the hardware for the size of the tcache,
+	 * so must probe for it.
+	 */
+	run_uncached(__probe_tcache);
+	rm7k_tc_enable();
+	rm7k_tcache_init = 1;
+	c->tcache.linesz = tc_lsize;
+	c->tcache.ways = 1;
+	pr_info("Tertiary cache size %ldK.\n", (tcache_size >> 10));
+}
diff --git a/arch/mips/mm/tlb-r3k.c b/arch/mips/mm/tlb-r3k.c
new file mode 100644
index 00000000..a63d1ed0
--- /dev/null
+++ b/arch/mips/mm/tlb-r3k.c
@@ -0,0 +1,284 @@
+/*
+ * r2300.c: R2000 and R3000 specific mmu/cache code.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ *
+ * with a lot of changes to make this thing work for R3000s
+ * Tx39XX R4k style caches added. HK
+ * Copyright (C) 1998, 1999, 2000 Harald Koerfgen
+ * Copyright (C) 1998 Gleb Raiko & Vladimir Roganov
+ * Copyright (C) 2002  Ralf Baechle
+ * Copyright (C) 2002  Maciej W. Rozycki
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/tlbmisc.h>
+#include <asm/isadep.h>
+#include <asm/io.h>
+#include <asm/bootinfo.h>
+#include <asm/cpu.h>
+
+#undef DEBUG_TLB
+
+extern void build_tlb_refill_handler(void);
+
+/* CP0 hazard avoidance. */
+#define BARRIER				\
+	__asm__ __volatile__(		\
+		".set	push\n\t"	\
+		".set	noreorder\n\t"	\
+		"nop\n\t"		\
+		".set	pop\n\t")
+
+int r3k_have_wired_reg;		/* should be in cpu_data? */
+
+/* TLB operations. */
+void local_flush_tlb_all(void)
+{
+	unsigned long flags;
+	unsigned long old_ctx;
+	int entry;
+
+#ifdef DEBUG_TLB
+	printk("[tlball]");
+#endif
+
+	local_irq_save(flags);
+	old_ctx = read_c0_entryhi() & ASID_MASK;
+	write_c0_entrylo0(0);
+	entry = r3k_have_wired_reg ? read_c0_wired() : 8;
+	for (; entry < current_cpu_data.tlbsize; entry++) {
+		write_c0_index(entry << 8);
+		write_c0_entryhi((entry | 0x80000) << 12);
+		BARRIER;
+		tlb_write_indexed();
+	}
+	write_c0_entryhi(old_ctx);
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) != 0) {
+#ifdef DEBUG_TLB
+		printk("[tlbmm<%lu>]", (unsigned long)cpu_context(cpu, mm));
+#endif
+		drop_mmu_context(mm, cpu);
+	}
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+			   unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) != 0) {
+		unsigned long size, flags;
+
+#ifdef DEBUG_TLB
+		printk("[tlbrange<%lu,0x%08lx,0x%08lx>]",
+			cpu_context(cpu, mm) & ASID_MASK, start, end);
+#endif
+		local_irq_save(flags);
+		size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+		if (size <= current_cpu_data.tlbsize) {
+			int oldpid = read_c0_entryhi() & ASID_MASK;
+			int newpid = cpu_context(cpu, mm) & ASID_MASK;
+
+			start &= PAGE_MASK;
+			end += PAGE_SIZE - 1;
+			end &= PAGE_MASK;
+			while (start < end) {
+				int idx;
+
+				write_c0_entryhi(start | newpid);
+				start += PAGE_SIZE;	/* BARRIER */
+				tlb_probe();
+				idx = read_c0_index();
+				write_c0_entrylo0(0);
+				write_c0_entryhi(KSEG0);
+				if (idx < 0)		/* BARRIER */
+					continue;
+				tlb_write_indexed();
+			}
+			write_c0_entryhi(oldpid);
+		} else {
+			drop_mmu_context(mm, cpu);
+		}
+		local_irq_restore(flags);
+	}
+}
+
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+	unsigned long size, flags;
+
+#ifdef DEBUG_TLB
+	printk("[tlbrange<%lu,0x%08lx,0x%08lx>]", start, end);
+#endif
+	local_irq_save(flags);
+	size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	if (size <= current_cpu_data.tlbsize) {
+		int pid = read_c0_entryhi();
+
+		start &= PAGE_MASK;
+		end += PAGE_SIZE - 1;
+		end &= PAGE_MASK;
+
+		while (start < end) {
+			int idx;
+
+			write_c0_entryhi(start);
+			start += PAGE_SIZE;		/* BARRIER */
+			tlb_probe();
+			idx = read_c0_index();
+			write_c0_entrylo0(0);
+			write_c0_entryhi(KSEG0);
+			if (idx < 0)			/* BARRIER */
+				continue;
+			tlb_write_indexed();
+		}
+		write_c0_entryhi(pid);
+	} else {
+		local_flush_tlb_all();
+	}
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	int cpu = smp_processor_id();
+
+	if (!vma || cpu_context(cpu, vma->vm_mm) != 0) {
+		unsigned long flags;
+		int oldpid, newpid, idx;
+
+#ifdef DEBUG_TLB
+		printk("[tlbpage<%lu,0x%08lx>]", cpu_context(cpu, vma->vm_mm), page);
+#endif
+		newpid = cpu_context(cpu, vma->vm_mm) & ASID_MASK;
+		page &= PAGE_MASK;
+		local_irq_save(flags);
+		oldpid = read_c0_entryhi() & ASID_MASK;
+		write_c0_entryhi(page | newpid);
+		BARRIER;
+		tlb_probe();
+		idx = read_c0_index();
+		write_c0_entrylo0(0);
+		write_c0_entryhi(KSEG0);
+		if (idx < 0)				/* BARRIER */
+			goto finish;
+		tlb_write_indexed();
+
+finish:
+		write_c0_entryhi(oldpid);
+		local_irq_restore(flags);
+	}
+}
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t pte)
+{
+	unsigned long flags;
+	int idx, pid;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (current->active_mm != vma->vm_mm)
+		return;
+
+	pid = read_c0_entryhi() & ASID_MASK;
+
+#ifdef DEBUG_TLB
+	if ((pid != (cpu_context(cpu, vma->vm_mm) & ASID_MASK)) || (cpu_context(cpu, vma->vm_mm) == 0)) {
+		printk("update_mmu_cache: Wheee, bogus tlbpid mmpid=%lu tlbpid=%d\n",
+		       (cpu_context(cpu, vma->vm_mm)), pid);
+	}
+#endif
+
+	local_irq_save(flags);
+	address &= PAGE_MASK;
+	write_c0_entryhi(address | pid);
+	BARRIER;
+	tlb_probe();
+	idx = read_c0_index();
+	write_c0_entrylo0(pte_val(pte));
+	write_c0_entryhi(address | pid);
+	if (idx < 0) {					/* BARRIER */
+		tlb_write_random();
+	} else {
+		tlb_write_indexed();
+	}
+	write_c0_entryhi(pid);
+	local_irq_restore(flags);
+}
+
+void add_wired_entry(unsigned long entrylo0, unsigned long entrylo1,
+		     unsigned long entryhi, unsigned long pagemask)
+{
+	unsigned long flags;
+	unsigned long old_ctx;
+	static unsigned long wired = 0;
+
+	if (r3k_have_wired_reg) {			/* TX39XX */
+		unsigned long old_pagemask;
+		unsigned long w;
+
+#ifdef DEBUG_TLB
+		printk("[tlbwired<entry lo0 %8x, hi %8x\n, pagemask %8x>]\n",
+		       entrylo0, entryhi, pagemask);
+#endif
+
+		local_irq_save(flags);
+		/* Save old context and create impossible VPN2 value */
+		old_ctx = read_c0_entryhi() & ASID_MASK;
+		old_pagemask = read_c0_pagemask();
+		w = read_c0_wired();
+		write_c0_wired(w + 1);
+		write_c0_index(w << 8);
+		write_c0_pagemask(pagemask);
+		write_c0_entryhi(entryhi);
+		write_c0_entrylo0(entrylo0);
+		BARRIER;
+		tlb_write_indexed();
+
+		write_c0_entryhi(old_ctx);
+		write_c0_pagemask(old_pagemask);
+		local_flush_tlb_all();
+		local_irq_restore(flags);
+
+	} else if (wired < 8) {
+#ifdef DEBUG_TLB
+		printk("[tlbwired<entry lo0 %8x, hi %8x\n>]\n",
+		       entrylo0, entryhi);
+#endif
+
+		local_irq_save(flags);
+		old_ctx = read_c0_entryhi() & ASID_MASK;
+		write_c0_entrylo0(entrylo0);
+		write_c0_entryhi(entryhi);
+		write_c0_index(wired);
+		wired++;				/* BARRIER */
+		tlb_write_indexed();
+		write_c0_entryhi(old_ctx);
+		local_flush_tlb_all();
+		local_irq_restore(flags);
+	}
+}
+
+void __cpuinit tlb_init(void)
+{
+	local_flush_tlb_all();
+
+	build_tlb_refill_handler();
+}
diff --git a/arch/mips/mm/tlb-r4k.c b/arch/mips/mm/tlb-r4k.c
new file mode 100644
index 00000000..d2572cb2
--- /dev/null
+++ b/arch/mips/mm/tlb-r4k.c
@@ -0,0 +1,432 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ * Copyright (C) 1997, 1998, 1999, 2000 Ralf Baechle ralf@gnu.org
+ * Carsten Langgaard, carstenl@mips.com
+ * Copyright (C) 2002 MIPS Technologies, Inc.  All rights reserved.
+ */
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+
+#include <asm/cpu.h>
+#include <asm/bootinfo.h>
+#include <asm/mmu_context.h>
+#include <asm/pgtable.h>
+#include <asm/tlbmisc.h>
+
+extern void build_tlb_refill_handler(void);
+
+/*
+ * Make sure all entries differ.  If they're not different
+ * MIPS32 will take revenge ...
+ */
+#define UNIQUE_ENTRYHI(idx) (CKSEG0 + ((idx) << (PAGE_SHIFT + 1)))
+
+/* Atomicity and interruptability */
+#ifdef CONFIG_MIPS_MT_SMTC
+
+#include <asm/smtc.h>
+#include <asm/mipsmtregs.h>
+
+#define ENTER_CRITICAL(flags) \
+	{ \
+	unsigned int mvpflags; \
+	local_irq_save(flags);\
+	mvpflags = dvpe()
+#define EXIT_CRITICAL(flags) \
+	evpe(mvpflags); \
+	local_irq_restore(flags); \
+	}
+#else
+
+#define ENTER_CRITICAL(flags) local_irq_save(flags)
+#define EXIT_CRITICAL(flags) local_irq_restore(flags)
+
+#endif /* CONFIG_MIPS_MT_SMTC */
+
+#if defined(CONFIG_CPU_LOONGSON2)
+/*
+ * LOONGSON2 has a 4 entry itlb which is a subset of dtlb,
+ * unfortrunately, itlb is not totally transparent to software.
+ */
+#define FLUSH_ITLB write_c0_diag(4);
+
+#define FLUSH_ITLB_VM(vma) { if ((vma)->vm_flags & VM_EXEC)  write_c0_diag(4); }
+
+#else
+
+#define FLUSH_ITLB
+#define FLUSH_ITLB_VM(vma)
+
+#endif
+
+void local_flush_tlb_all(void)
+{
+	unsigned long flags;
+	unsigned long old_ctx;
+	int entry;
+
+	ENTER_CRITICAL(flags);
+	/* Save old context and create impossible VPN2 value */
+	old_ctx = read_c0_entryhi();
+	write_c0_entrylo0(0);
+	write_c0_entrylo1(0);
+
+	entry = read_c0_wired();
+
+	/* Blast 'em all away. */
+	while (entry < current_cpu_data.tlbsize) {
+		/* Make sure all entries differ. */
+		write_c0_entryhi(UNIQUE_ENTRYHI(entry));
+		write_c0_index(entry);
+		mtc0_tlbw_hazard();
+		tlb_write_indexed();
+		entry++;
+	}
+	tlbw_use_hazard();
+	write_c0_entryhi(old_ctx);
+	FLUSH_ITLB;
+	EXIT_CRITICAL(flags);
+}
+
+/* All entries common to a mm share an asid.  To effectively flush
+   these entries, we just bump the asid. */
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	int cpu;
+
+	preempt_disable();
+
+	cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) != 0) {
+		drop_mmu_context(mm, cpu);
+	}
+
+	preempt_enable();
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+	unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) != 0) {
+		unsigned long size, flags;
+		int huge = is_vm_hugetlb_page(vma);
+
+		ENTER_CRITICAL(flags);
+		if (huge) {
+			start = round_down(start, HPAGE_SIZE);
+			end = round_up(end, HPAGE_SIZE);
+			size = (end - start) >> HPAGE_SHIFT;
+		} else {
+			start = round_down(start, PAGE_SIZE << 1);
+			end = round_up(end, PAGE_SIZE << 1);
+			size = (end - start) >> (PAGE_SHIFT + 1);
+		}
+		if (size <= current_cpu_data.tlbsize/2) {
+			int oldpid = read_c0_entryhi();
+			int newpid = cpu_asid(cpu, mm);
+
+			while (start < end) {
+				int idx;
+
+				write_c0_entryhi(start | newpid);
+				if (huge)
+					start += HPAGE_SIZE;
+				else
+					start += (PAGE_SIZE << 1);
+				mtc0_tlbw_hazard();
+				tlb_probe();
+				tlb_probe_hazard();
+				idx = read_c0_index();
+				write_c0_entrylo0(0);
+				write_c0_entrylo1(0);
+				if (idx < 0)
+					continue;
+				/* Make sure all entries differ. */
+				write_c0_entryhi(UNIQUE_ENTRYHI(idx));
+				mtc0_tlbw_hazard();
+				tlb_write_indexed();
+			}
+			tlbw_use_hazard();
+			write_c0_entryhi(oldpid);
+		} else {
+			drop_mmu_context(mm, cpu);
+		}
+		FLUSH_ITLB;
+		EXIT_CRITICAL(flags);
+	}
+}
+
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+	unsigned long size, flags;
+
+	ENTER_CRITICAL(flags);
+	size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	size = (size + 1) >> 1;
+	if (size <= current_cpu_data.tlbsize / 2) {
+		int pid = read_c0_entryhi();
+
+		start &= (PAGE_MASK << 1);
+		end += ((PAGE_SIZE << 1) - 1);
+		end &= (PAGE_MASK << 1);
+
+		while (start < end) {
+			int idx;
+
+			write_c0_entryhi(start);
+			start += (PAGE_SIZE << 1);
+			mtc0_tlbw_hazard();
+			tlb_probe();
+			tlb_probe_hazard();
+			idx = read_c0_index();
+			write_c0_entrylo0(0);
+			write_c0_entrylo1(0);
+			if (idx < 0)
+				continue;
+			/* Make sure all entries differ. */
+			write_c0_entryhi(UNIQUE_ENTRYHI(idx));
+			mtc0_tlbw_hazard();
+			tlb_write_indexed();
+		}
+		tlbw_use_hazard();
+		write_c0_entryhi(pid);
+	} else {
+		local_flush_tlb_all();
+	}
+	FLUSH_ITLB;
+	EXIT_CRITICAL(flags);
+}
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, vma->vm_mm) != 0) {
+		unsigned long flags;
+		int oldpid, newpid, idx;
+
+		newpid = cpu_asid(cpu, vma->vm_mm);
+		page &= (PAGE_MASK << 1);
+		ENTER_CRITICAL(flags);
+		oldpid = read_c0_entryhi();
+		write_c0_entryhi(page | newpid);
+		mtc0_tlbw_hazard();
+		tlb_probe();
+		tlb_probe_hazard();
+		idx = read_c0_index();
+		write_c0_entrylo0(0);
+		write_c0_entrylo1(0);
+		if (idx < 0)
+			goto finish;
+		/* Make sure all entries differ. */
+		write_c0_entryhi(UNIQUE_ENTRYHI(idx));
+		mtc0_tlbw_hazard();
+		tlb_write_indexed();
+		tlbw_use_hazard();
+
+	finish:
+		write_c0_entryhi(oldpid);
+		FLUSH_ITLB_VM(vma);
+		EXIT_CRITICAL(flags);
+	}
+}
+
+/*
+ * This one is only used for pages with the global bit set so we don't care
+ * much about the ASID.
+ */
+void local_flush_tlb_one(unsigned long page)
+{
+	unsigned long flags;
+	int oldpid, idx;
+
+	ENTER_CRITICAL(flags);
+	oldpid = read_c0_entryhi();
+	page &= (PAGE_MASK << 1);
+	write_c0_entryhi(page);
+	mtc0_tlbw_hazard();
+	tlb_probe();
+	tlb_probe_hazard();
+	idx = read_c0_index();
+	write_c0_entrylo0(0);
+	write_c0_entrylo1(0);
+	if (idx >= 0) {
+		/* Make sure all entries differ. */
+		write_c0_entryhi(UNIQUE_ENTRYHI(idx));
+		mtc0_tlbw_hazard();
+		tlb_write_indexed();
+		tlbw_use_hazard();
+	}
+	write_c0_entryhi(oldpid);
+	FLUSH_ITLB;
+	EXIT_CRITICAL(flags);
+}
+
+/*
+ * We will need multiple versions of update_mmu_cache(), one that just
+ * updates the TLB with the new pte(s), and another which also checks
+ * for the R4k "end of page" hardware bug and does the needy.
+ */
+void __update_tlb(struct vm_area_struct * vma, unsigned long address, pte_t pte)
+{
+	unsigned long flags;
+	pgd_t *pgdp;
+	pud_t *pudp;
+	pmd_t *pmdp;
+	pte_t *ptep;
+	int idx, pid;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (current->active_mm != vma->vm_mm)
+		return;
+
+	ENTER_CRITICAL(flags);
+
+	pid = read_c0_entryhi() & ASID_MASK;
+	address &= (PAGE_MASK << 1);
+	write_c0_entryhi(address | pid);
+	pgdp = pgd_offset(vma->vm_mm, address);
+	mtc0_tlbw_hazard();
+	tlb_probe();
+	tlb_probe_hazard();
+	pudp = pud_offset(pgdp, address);
+	pmdp = pmd_offset(pudp, address);
+	idx = read_c0_index();
+#ifdef CONFIG_HUGETLB_PAGE
+	/* this could be a huge page  */
+	if (pmd_huge(*pmdp)) {
+		unsigned long lo;
+		write_c0_pagemask(PM_HUGE_MASK);
+		ptep = (pte_t *)pmdp;
+		lo = pte_to_entrylo(pte_val(*ptep));
+		write_c0_entrylo0(lo);
+		write_c0_entrylo1(lo + (HPAGE_SIZE >> 7));
+
+		mtc0_tlbw_hazard();
+		if (idx < 0)
+			tlb_write_random();
+		else
+			tlb_write_indexed();
+		write_c0_pagemask(PM_DEFAULT_MASK);
+	} else
+#endif
+	{
+		ptep = pte_offset_map(pmdp, address);
+
+#if defined(CONFIG_64BIT_PHYS_ADDR) && defined(CONFIG_CPU_MIPS32)
+		write_c0_entrylo0(ptep->pte_high);
+		ptep++;
+		write_c0_entrylo1(ptep->pte_high);
+#else
+		write_c0_entrylo0(pte_to_entrylo(pte_val(*ptep++)));
+		write_c0_entrylo1(pte_to_entrylo(pte_val(*ptep)));
+#endif
+		mtc0_tlbw_hazard();
+		if (idx < 0)
+			tlb_write_random();
+		else
+			tlb_write_indexed();
+	}
+	tlbw_use_hazard();
+	FLUSH_ITLB_VM(vma);
+	EXIT_CRITICAL(flags);
+}
+
+void add_wired_entry(unsigned long entrylo0, unsigned long entrylo1,
+		     unsigned long entryhi, unsigned long pagemask)
+{
+	unsigned long flags;
+	unsigned long wired;
+	unsigned long old_pagemask;
+	unsigned long old_ctx;
+
+	ENTER_CRITICAL(flags);
+	/* Save old context and create impossible VPN2 value */
+	old_ctx = read_c0_entryhi();
+	old_pagemask = read_c0_pagemask();
+	wired = read_c0_wired();
+	write_c0_wired(wired + 1);
+	write_c0_index(wired);
+	tlbw_use_hazard();	/* What is the hazard here? */
+	write_c0_pagemask(pagemask);
+	write_c0_entryhi(entryhi);
+	write_c0_entrylo0(entrylo0);
+	write_c0_entrylo1(entrylo1);
+	mtc0_tlbw_hazard();
+	tlb_write_indexed();
+	tlbw_use_hazard();
+
+	write_c0_entryhi(old_ctx);
+	tlbw_use_hazard();	/* What is the hazard here? */
+	write_c0_pagemask(old_pagemask);
+	local_flush_tlb_all();
+	EXIT_CRITICAL(flags);
+}
+
+static int __cpuinitdata ntlb;
+static int __init set_ntlb(char *str)
+{
+	get_option(&str, &ntlb);
+	return 1;
+}
+
+__setup("ntlb=", set_ntlb);
+
+void __cpuinit tlb_init(void)
+{
+	/*
+	 * You should never change this register:
+	 *   - On R4600 1.7 the tlbp never hits for pages smaller than
+	 *     the value in the c0_pagemask register.
+	 *   - The entire mm handling assumes the c0_pagemask register to
+	 *     be set to fixed-size pages.
+	 */
+	write_c0_pagemask(PM_DEFAULT_MASK);
+	write_c0_wired(0);
+	if (current_cpu_type() == CPU_R10000 ||
+	    current_cpu_type() == CPU_R12000 ||
+	    current_cpu_type() == CPU_R14000)
+		write_c0_framemask(0);
+
+	if (kernel_uses_smartmips_rixi) {
+		/*
+		 * Enable the no read, no exec bits, and enable large virtual
+		 * address.
+		 */
+		u32 pg = PG_RIE | PG_XIE;
+#ifdef CONFIG_64BIT
+		pg |= PG_ELPA;
+#endif
+		write_c0_pagegrain(pg);
+	}
+
+        /* From this point on the ARC firmware is dead.  */
+	local_flush_tlb_all();
+
+	/* Did I tell you that ARC SUCKS?  */
+
+	if (ntlb) {
+		if (ntlb > 1 && ntlb <= current_cpu_data.tlbsize) {
+			int wired = current_cpu_data.tlbsize - ntlb;
+			write_c0_wired(wired);
+			write_c0_index(wired-1);
+			printk("Restricting TLB to %d entries\n", ntlb);
+		} else
+			printk("Ignoring invalid argument ntlb=%d\n", ntlb);
+	}
+
+	build_tlb_refill_handler();
+}
diff --git a/arch/mips/mm/tlb-r8k.c b/arch/mips/mm/tlb-r8k.c
new file mode 100644
index 00000000..91c2499f
--- /dev/null
+++ b/arch/mips/mm/tlb-r8k.c
@@ -0,0 +1,248 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1996 David S. Miller (davem@davemloft.net)
+ * Copyright (C) 1997, 1998, 1999, 2000 Ralf Baechle ralf@gnu.org
+ * Carsten Langgaard, carstenl@mips.com
+ * Copyright (C) 2002 MIPS Technologies, Inc.  All rights reserved.
+ */
+#include <linux/init.h>
+#include <linux/sched.h>
+#include <linux/smp.h>
+#include <linux/mm.h>
+
+#include <asm/cpu.h>
+#include <asm/bootinfo.h>
+#include <asm/mmu_context.h>
+#include <asm/pgtable.h>
+
+extern void build_tlb_refill_handler(void);
+
+#define TFP_TLB_SIZE		384
+#define TFP_TLB_SET_SHIFT	7
+
+/* CP0 hazard avoidance. */
+#define BARRIER __asm__ __volatile__(".set noreorder\n\t" \
+				     "nop; nop; nop; nop; nop; nop;\n\t" \
+				     ".set reorder\n\t")
+
+void local_flush_tlb_all(void)
+{
+	unsigned long flags;
+	unsigned long old_ctx;
+	int entry;
+
+	local_irq_save(flags);
+	/* Save old context and create impossible VPN2 value */
+	old_ctx = read_c0_entryhi();
+	write_c0_entrylo(0);
+
+	for (entry = 0; entry < TFP_TLB_SIZE; entry++) {
+		write_c0_tlbset(entry >> TFP_TLB_SET_SHIFT);
+		write_c0_vaddr(entry << PAGE_SHIFT);
+		write_c0_entryhi(CKSEG0 + (entry << (PAGE_SHIFT + 1)));
+		mtc0_tlbw_hazard();
+		tlb_write();
+	}
+	tlbw_use_hazard();
+	write_c0_entryhi(old_ctx);
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) != 0)
+		drop_mmu_context(mm, cpu);
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+	unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	int cpu = smp_processor_id();
+	unsigned long flags;
+	int oldpid, newpid, size;
+
+	if (!cpu_context(cpu, mm))
+		return;
+
+	size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	size = (size + 1) >> 1;
+
+	local_irq_save(flags);
+
+	if (size > TFP_TLB_SIZE / 2) {
+		drop_mmu_context(mm, cpu);
+		goto out_restore;
+	}
+
+	oldpid = read_c0_entryhi();
+	newpid = cpu_asid(cpu, mm);
+
+	write_c0_entrylo(0);
+
+	start &= PAGE_MASK;
+	end += (PAGE_SIZE - 1);
+	end &= PAGE_MASK;
+	while (start < end) {
+		signed long idx;
+
+		write_c0_vaddr(start);
+		write_c0_entryhi(start);
+		start += PAGE_SIZE;
+		tlb_probe();
+		idx = read_c0_tlbset();
+		if (idx < 0)
+			continue;
+
+		write_c0_entryhi(CKSEG0 + (idx << (PAGE_SHIFT + 1)));
+		tlb_write();
+	}
+	write_c0_entryhi(oldpid);
+
+out_restore:
+	local_irq_restore(flags);
+}
+
+/* Usable for KV1 addresses only! */
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+	unsigned long size, flags;
+
+	size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	size = (size + 1) >> 1;
+
+	if (size > TFP_TLB_SIZE / 2) {
+		local_flush_tlb_all();
+		return;
+	}
+
+	local_irq_save(flags);
+
+	write_c0_entrylo(0);
+
+	start &= PAGE_MASK;
+	end += (PAGE_SIZE - 1);
+	end &= PAGE_MASK;
+	while (start < end) {
+		signed long idx;
+
+		write_c0_vaddr(start);
+		write_c0_entryhi(start);
+		start += PAGE_SIZE;
+		tlb_probe();
+		idx = read_c0_tlbset();
+		if (idx < 0)
+			continue;
+
+		write_c0_entryhi(CKSEG0 + (idx << (PAGE_SHIFT + 1)));
+		tlb_write();
+	}
+
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	int cpu = smp_processor_id();
+	unsigned long flags;
+	int oldpid, newpid;
+	signed long idx;
+
+	if (!cpu_context(cpu, vma->vm_mm))
+		return;
+
+	newpid = cpu_asid(cpu, vma->vm_mm);
+	page &= PAGE_MASK;
+	local_irq_save(flags);
+	oldpid = read_c0_entryhi();
+	write_c0_vaddr(page);
+	write_c0_entryhi(newpid);
+	tlb_probe();
+	idx = read_c0_tlbset();
+	if (idx < 0)
+		goto finish;
+
+	write_c0_entrylo(0);
+	write_c0_entryhi(CKSEG0 + (idx << (PAGE_SHIFT + 1)));
+	tlb_write();
+
+finish:
+	write_c0_entryhi(oldpid);
+	local_irq_restore(flags);
+}
+
+/*
+ * We will need multiple versions of update_mmu_cache(), one that just
+ * updates the TLB with the new pte(s), and another which also checks
+ * for the R4k "end of page" hardware bug and does the needy.
+ */
+void __update_tlb(struct vm_area_struct * vma, unsigned long address, pte_t pte)
+{
+	unsigned long flags;
+	pgd_t *pgdp;
+	pmd_t *pmdp;
+	pte_t *ptep;
+	int pid;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (current->active_mm != vma->vm_mm)
+		return;
+
+	pid = read_c0_entryhi() & ASID_MASK;
+
+	local_irq_save(flags);
+	address &= PAGE_MASK;
+	write_c0_vaddr(address);
+	write_c0_entryhi(pid);
+	pgdp = pgd_offset(vma->vm_mm, address);
+	pmdp = pmd_offset(pgdp, address);
+	ptep = pte_offset_map(pmdp, address);
+	tlb_probe();
+
+	write_c0_entrylo(pte_val(*ptep++) >> 6);
+	tlb_write();
+
+	write_c0_entryhi(pid);
+	local_irq_restore(flags);
+}
+
+static void __cpuinit probe_tlb(unsigned long config)
+{
+	struct cpuinfo_mips *c = &current_cpu_data;
+
+	c->tlbsize = 3 * 128;		/* 3 sets each 128 entries */
+}
+
+void __cpuinit tlb_init(void)
+{
+	unsigned int config = read_c0_config();
+	unsigned long status;
+
+	probe_tlb(config);
+
+	status = read_c0_status();
+	status &= ~(ST0_UPS | ST0_KPS);
+#ifdef CONFIG_PAGE_SIZE_4KB
+	status |= (TFP_PAGESIZE_4K << 32) | (TFP_PAGESIZE_4K << 36);
+#elif defined(CONFIG_PAGE_SIZE_8KB)
+	status |= (TFP_PAGESIZE_8K << 32) | (TFP_PAGESIZE_8K << 36);
+#elif defined(CONFIG_PAGE_SIZE_16KB)
+	status |= (TFP_PAGESIZE_16K << 32) | (TFP_PAGESIZE_16K << 36);
+#elif defined(CONFIG_PAGE_SIZE_64KB)
+	status |= (TFP_PAGESIZE_64K << 32) | (TFP_PAGESIZE_64K << 36);
+#endif
+	write_c0_status(status);
+
+	write_c0_wired(0);
+
+	local_flush_tlb_all();
+
+	build_tlb_refill_handler();
+}
diff --git a/arch/mips/mm/tlbex-fault.S b/arch/mips/mm/tlbex-fault.S
new file mode 100644
index 00000000..e99eaa1f
--- /dev/null
+++ b/arch/mips/mm/tlbex-fault.S
@@ -0,0 +1,28 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 1999 Ralf Baechle
+ * Copyright (C) 1999 Silicon Graphics, Inc.
+ */
+#include <asm/mipsregs.h>
+#include <asm/page.h>
+#include <asm/regdef.h>
+#include <asm/stackframe.h>
+
+	.macro tlb_do_page_fault, write
+	NESTED(tlb_do_page_fault_\write, PT_SIZE, sp)
+	SAVE_ALL
+	MFC0	a2, CP0_BADVADDR
+	KMODE
+	move	a0, sp
+	REG_S	a2, PT_BVADDR(sp)
+	li	a1, \write
+	PTR_LA	ra, ret_from_exception
+	j	do_page_fault
+	END(tlb_do_page_fault_\write)
+	.endm
+
+	tlb_do_page_fault 0
+	tlb_do_page_fault 1
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c
new file mode 100644
index 00000000..0bc485b3
--- /dev/null
+++ b/arch/mips/mm/tlbex.c
@@ -0,0 +1,2160 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Synthesize TLB refill handlers at runtime.
+ *
+ * Copyright (C) 2004, 2005, 2006, 2008  Thiemo Seufer
+ * Copyright (C) 2005, 2007, 2008, 2009  Maciej W. Rozycki
+ * Copyright (C) 2006  Ralf Baechle (ralf@linux-mips.org)
+ * Copyright (C) 2008, 2009 Cavium Networks, Inc.
+ *
+ * ... and the days got worse and worse and now you see
+ * I've gone completly out of my mind.
+ *
+ * They're coming to take me a away haha
+ * they're coming to take me a away hoho hihi haha
+ * to the funny farm where code is beautiful all the time ...
+ *
+ * (Condolences to Napoleon XIV)
+ */
+
+#include <linux/bug.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/smp.h>
+#include <linux/string.h>
+#include <linux/init.h>
+#include <linux/cache.h>
+
+#include <asm/cacheflush.h>
+#include <asm/pgtable.h>
+#include <asm/war.h>
+#include <asm/uasm.h>
+#include <asm/setup.h>
+
+/*
+ * TLB load/store/modify handlers.
+ *
+ * Only the fastpath gets synthesized at runtime, the slowpath for
+ * do_page_fault remains normal asm.
+ */
+extern void tlb_do_page_fault_0(void);
+extern void tlb_do_page_fault_1(void);
+
+struct work_registers {
+	int r1;
+	int r2;
+	int r3;
+};
+
+struct tlb_reg_save {
+	unsigned long a;
+	unsigned long b;
+} ____cacheline_aligned_in_smp;
+
+static struct tlb_reg_save handler_reg_save[NR_CPUS];
+
+static inline int r45k_bvahwbug(void)
+{
+	/* XXX: We should probe for the presence of this bug, but we don't. */
+	return 0;
+}
+
+static inline int r4k_250MHZhwbug(void)
+{
+	/* XXX: We should probe for the presence of this bug, but we don't. */
+	return 0;
+}
+
+static inline int __maybe_unused bcm1250_m3_war(void)
+{
+	return BCM1250_M3_WAR;
+}
+
+static inline int __maybe_unused r10000_llsc_war(void)
+{
+	return R10000_LLSC_WAR;
+}
+
+static int use_bbit_insns(void)
+{
+	switch (current_cpu_type()) {
+	case CPU_CAVIUM_OCTEON:
+	case CPU_CAVIUM_OCTEON_PLUS:
+	case CPU_CAVIUM_OCTEON2:
+		return 1;
+	default:
+		return 0;
+	}
+}
+
+static int use_lwx_insns(void)
+{
+	switch (current_cpu_type()) {
+	case CPU_CAVIUM_OCTEON2:
+		return 1;
+	default:
+		return 0;
+	}
+}
+#if defined(CONFIG_CAVIUM_OCTEON_CVMSEG_SIZE) && \
+    CONFIG_CAVIUM_OCTEON_CVMSEG_SIZE > 0
+static bool scratchpad_available(void)
+{
+	return true;
+}
+static int scratchpad_offset(int i)
+{
+	/*
+	 * CVMSEG starts at address -32768 and extends for
+	 * CAVIUM_OCTEON_CVMSEG_SIZE 128 byte cache lines.
+	 */
+	i += 1; /* Kernel use starts at the top and works down. */
+	return CONFIG_CAVIUM_OCTEON_CVMSEG_SIZE * 128 - (8 * i) - 32768;
+}
+#else
+static bool scratchpad_available(void)
+{
+	return false;
+}
+static int scratchpad_offset(int i)
+{
+	BUG();
+	/* Really unreachable, but evidently some GCC want this. */
+	return 0;
+}
+#endif
+/*
+ * Found by experiment: At least some revisions of the 4kc throw under
+ * some circumstances a machine check exception, triggered by invalid
+ * values in the index register.  Delaying the tlbp instruction until
+ * after the next branch,  plus adding an additional nop in front of
+ * tlbwi/tlbwr avoids the invalid index register values. Nobody knows
+ * why; it's not an issue caused by the core RTL.
+ *
+ */
+static int __cpuinit m4kc_tlbp_war(void)
+{
+	return (current_cpu_data.processor_id & 0xffff00) ==
+	       (PRID_COMP_MIPS | PRID_IMP_4KC);
+}
+
+/* Handle labels (which must be positive integers). */
+enum label_id {
+	label_second_part = 1,
+	label_leave,
+	label_vmalloc,
+	label_vmalloc_done,
+	label_tlbw_hazard,
+	label_split,
+	label_tlbl_goaround1,
+	label_tlbl_goaround2,
+	label_nopage_tlbl,
+	label_nopage_tlbs,
+	label_nopage_tlbm,
+	label_smp_pgtable_change,
+	label_r3000_write_probe_fail,
+	label_large_segbits_fault,
+#ifdef CONFIG_HUGETLB_PAGE
+	label_tlb_huge_update,
+#endif
+};
+
+UASM_L_LA(_second_part)
+UASM_L_LA(_leave)
+UASM_L_LA(_vmalloc)
+UASM_L_LA(_vmalloc_done)
+UASM_L_LA(_tlbw_hazard)
+UASM_L_LA(_split)
+UASM_L_LA(_tlbl_goaround1)
+UASM_L_LA(_tlbl_goaround2)
+UASM_L_LA(_nopage_tlbl)
+UASM_L_LA(_nopage_tlbs)
+UASM_L_LA(_nopage_tlbm)
+UASM_L_LA(_smp_pgtable_change)
+UASM_L_LA(_r3000_write_probe_fail)
+UASM_L_LA(_large_segbits_fault)
+#ifdef CONFIG_HUGETLB_PAGE
+UASM_L_LA(_tlb_huge_update)
+#endif
+
+/*
+ * For debug purposes.
+ */
+static inline void dump_handler(const u32 *handler, int count)
+{
+	int i;
+
+	pr_debug("\t.set push\n");
+	pr_debug("\t.set noreorder\n");
+
+	for (i = 0; i < count; i++)
+		pr_debug("\t%p\t.word 0x%08x\n", &handler[i], handler[i]);
+
+	pr_debug("\t.set pop\n");
+}
+
+/* The only general purpose registers allowed in TLB handlers. */
+#define K0		26
+#define K1		27
+
+/* Some CP0 registers */
+#define C0_INDEX	0, 0
+#define C0_ENTRYLO0	2, 0
+#define C0_TCBIND	2, 2
+#define C0_ENTRYLO1	3, 0
+#define C0_CONTEXT	4, 0
+#define C0_PAGEMASK	5, 0
+#define C0_BADVADDR	8, 0
+#define C0_ENTRYHI	10, 0
+#define C0_EPC		14, 0
+#define C0_XCONTEXT	20, 0
+
+#ifdef CONFIG_64BIT
+# define GET_CONTEXT(buf, reg) UASM_i_MFC0(buf, reg, C0_XCONTEXT)
+#else
+# define GET_CONTEXT(buf, reg) UASM_i_MFC0(buf, reg, C0_CONTEXT)
+#endif
+
+/* The worst case length of the handler is around 18 instructions for
+ * R3000-style TLBs and up to 63 instructions for R4000-style TLBs.
+ * Maximum space available is 32 instructions for R3000 and 64
+ * instructions for R4000.
+ *
+ * We deliberately chose a buffer size of 128, so we won't scribble
+ * over anything important on overflow before we panic.
+ */
+static u32 tlb_handler[128] __cpuinitdata;
+
+/* simply assume worst case size for labels and relocs */
+static struct uasm_label labels[128] __cpuinitdata;
+static struct uasm_reloc relocs[128] __cpuinitdata;
+
+#ifdef CONFIG_64BIT
+static int check_for_high_segbits __cpuinitdata;
+#endif
+
+static int check_for_high_segbits __cpuinitdata;
+
+static unsigned int kscratch_used_mask __cpuinitdata;
+
+static int __cpuinit allocate_kscratch(void)
+{
+	int r;
+	unsigned int a = cpu_data[0].kscratch_mask & ~kscratch_used_mask;
+
+	r = ffs(a);
+
+	if (r == 0)
+		return -1;
+
+	r--; /* make it zero based */
+
+	kscratch_used_mask |= (1 << r);
+
+	return r;
+}
+
+static int scratch_reg __cpuinitdata;
+static int pgd_reg __cpuinitdata;
+enum vmalloc64_mode {not_refill, refill_scratch, refill_noscratch};
+
+static struct work_registers __cpuinit build_get_work_registers(u32 **p)
+{
+	struct work_registers r;
+
+	int smp_processor_id_reg;
+	int smp_processor_id_sel;
+	int smp_processor_id_shift;
+
+	if (scratch_reg > 0) {
+		/* Save in CPU local C0_KScratch? */
+		UASM_i_MTC0(p, 1, 31, scratch_reg);
+		r.r1 = K0;
+		r.r2 = K1;
+		r.r3 = 1;
+		return r;
+	}
+
+	if (num_possible_cpus() > 1) {
+#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
+		smp_processor_id_shift = 51;
+		smp_processor_id_reg = 20; /* XContext */
+		smp_processor_id_sel = 0;
+#else
+# ifdef CONFIG_32BIT
+		smp_processor_id_shift = 25;
+		smp_processor_id_reg = 4; /* Context */
+		smp_processor_id_sel = 0;
+# endif
+# ifdef CONFIG_64BIT
+		smp_processor_id_shift = 26;
+		smp_processor_id_reg = 4; /* Context */
+		smp_processor_id_sel = 0;
+# endif
+#endif
+		/* Get smp_processor_id */
+		UASM_i_MFC0(p, K0, smp_processor_id_reg, smp_processor_id_sel);
+		UASM_i_SRL_SAFE(p, K0, K0, smp_processor_id_shift);
+
+		/* handler_reg_save index in K0 */
+		UASM_i_SLL(p, K0, K0, ilog2(sizeof(struct tlb_reg_save)));
+
+		UASM_i_LA(p, K1, (long)&handler_reg_save);
+		UASM_i_ADDU(p, K0, K0, K1);
+	} else {
+		UASM_i_LA(p, K0, (long)&handler_reg_save);
+	}
+	/* K0 now points to save area, save $1 and $2  */
+	UASM_i_SW(p, 1, offsetof(struct tlb_reg_save, a), K0);
+	UASM_i_SW(p, 2, offsetof(struct tlb_reg_save, b), K0);
+
+	r.r1 = K1;
+	r.r2 = 1;
+	r.r3 = 2;
+	return r;
+}
+
+static void __cpuinit build_restore_work_registers(u32 **p)
+{
+	if (scratch_reg > 0) {
+		UASM_i_MFC0(p, 1, 31, scratch_reg);
+		return;
+	}
+	/* K0 already points to save area, restore $1 and $2  */
+	UASM_i_LW(p, 1, offsetof(struct tlb_reg_save, a), K0);
+	UASM_i_LW(p, 2, offsetof(struct tlb_reg_save, b), K0);
+}
+
+#ifndef CONFIG_MIPS_PGD_C0_CONTEXT
+
+/*
+ * CONFIG_MIPS_PGD_C0_CONTEXT implies 64 bit and lack of pgd_current,
+ * we cannot do r3000 under these circumstances.
+ *
+ * Declare pgd_current here instead of including mmu_context.h to avoid type
+ * conflicts for tlbmiss_handler_setup_pgd
+ */
+extern unsigned long pgd_current[];
+
+/*
+ * The R3000 TLB handler is simple.
+ */
+static void __cpuinit build_r3000_tlb_refill_handler(void)
+{
+	long pgdc = (long)pgd_current;
+	u32 *p;
+
+	memset(tlb_handler, 0, sizeof(tlb_handler));
+	p = tlb_handler;
+
+	uasm_i_mfc0(&p, K0, C0_BADVADDR);
+	uasm_i_lui(&p, K1, uasm_rel_hi(pgdc)); /* cp0 delay */
+	uasm_i_lw(&p, K1, uasm_rel_lo(pgdc), K1);
+	uasm_i_srl(&p, K0, K0, 22); /* load delay */
+	uasm_i_sll(&p, K0, K0, 2);
+	uasm_i_addu(&p, K1, K1, K0);
+	uasm_i_mfc0(&p, K0, C0_CONTEXT);
+	uasm_i_lw(&p, K1, 0, K1); /* cp0 delay */
+	uasm_i_andi(&p, K0, K0, 0xffc); /* load delay */
+	uasm_i_addu(&p, K1, K1, K0);
+	uasm_i_lw(&p, K0, 0, K1);
+	uasm_i_nop(&p); /* load delay */
+	uasm_i_mtc0(&p, K0, C0_ENTRYLO0);
+	uasm_i_mfc0(&p, K1, C0_EPC); /* cp0 delay */
+	uasm_i_tlbwr(&p); /* cp0 delay */
+	uasm_i_jr(&p, K1);
+	uasm_i_rfe(&p); /* branch delay */
+
+	if (p > tlb_handler + 32)
+		panic("TLB refill handler space exceeded");
+
+	pr_debug("Wrote TLB refill handler (%u instructions).\n",
+		 (unsigned int)(p - tlb_handler));
+
+	memcpy((void *)ebase, tlb_handler, 0x80);
+
+	dump_handler((u32 *)ebase, 32);
+}
+#endif /* CONFIG_MIPS_PGD_C0_CONTEXT */
+
+/*
+ * The R4000 TLB handler is much more complicated. We have two
+ * consecutive handler areas with 32 instructions space each.
+ * Since they aren't used at the same time, we can overflow in the
+ * other one.To keep things simple, we first assume linear space,
+ * then we relocate it to the final handler layout as needed.
+ */
+static u32 final_handler[64] __cpuinitdata;
+
+/*
+ * Hazards
+ *
+ * From the IDT errata for the QED RM5230 (Nevada), processor revision 1.0:
+ * 2. A timing hazard exists for the TLBP instruction.
+ *
+ *      stalling_instruction
+ *      TLBP
+ *
+ * The JTLB is being read for the TLBP throughout the stall generated by the
+ * previous instruction. This is not really correct as the stalling instruction
+ * can modify the address used to access the JTLB.  The failure symptom is that
+ * the TLBP instruction will use an address created for the stalling instruction
+ * and not the address held in C0_ENHI and thus report the wrong results.
+ *
+ * The software work-around is to not allow the instruction preceding the TLBP
+ * to stall - make it an NOP or some other instruction guaranteed not to stall.
+ *
+ * Errata 2 will not be fixed.  This errata is also on the R5000.
+ *
+ * As if we MIPS hackers wouldn't know how to nop pipelines happy ...
+ */
+static void __cpuinit __maybe_unused build_tlb_probe_entry(u32 **p)
+{
+	switch (current_cpu_type()) {
+	/* Found by experiment: R4600 v2.0/R4700 needs this, too.  */
+	case CPU_R4600:
+	case CPU_R4700:
+	case CPU_R5000:
+	case CPU_R5000A:
+	case CPU_NEVADA:
+		uasm_i_nop(p);
+		uasm_i_tlbp(p);
+		break;
+
+	default:
+		uasm_i_tlbp(p);
+		break;
+	}
+}
+
+/*
+ * Write random or indexed TLB entry, and care about the hazards from
+ * the preceding mtc0 and for the following eret.
+ */
+enum tlb_write_entry { tlb_random, tlb_indexed };
+
+static void __cpuinit build_tlb_write_entry(u32 **p, struct uasm_label **l,
+					 struct uasm_reloc **r,
+					 enum tlb_write_entry wmode)
+{
+	void(*tlbw)(u32 **) = NULL;
+
+	switch (wmode) {
+	case tlb_random: tlbw = uasm_i_tlbwr; break;
+	case tlb_indexed: tlbw = uasm_i_tlbwi; break;
+	}
+
+	if (cpu_has_mips_r2) {
+		if (cpu_has_mips_r2_exec_hazard)
+			uasm_i_ehb(p);
+		tlbw(p);
+		return;
+	}
+
+	switch (current_cpu_type()) {
+	case CPU_R4000PC:
+	case CPU_R4000SC:
+	case CPU_R4000MC:
+	case CPU_R4400PC:
+	case CPU_R4400SC:
+	case CPU_R4400MC:
+		/*
+		 * This branch uses up a mtc0 hazard nop slot and saves
+		 * two nops after the tlbw instruction.
+		 */
+		uasm_il_bgezl(p, r, 0, label_tlbw_hazard);
+		tlbw(p);
+		uasm_l_tlbw_hazard(l, *p);
+		uasm_i_nop(p);
+		break;
+
+	case CPU_R4600:
+	case CPU_R4700:
+	case CPU_R5000:
+	case CPU_R5000A:
+		uasm_i_nop(p);
+		tlbw(p);
+		uasm_i_nop(p);
+		break;
+
+	case CPU_R4300:
+	case CPU_5KC:
+	case CPU_TX49XX:
+	case CPU_PR4450:
+	case CPU_XLR:
+		uasm_i_nop(p);
+		tlbw(p);
+		break;
+
+	case CPU_R10000:
+	case CPU_R12000:
+	case CPU_R14000:
+	case CPU_4KC:
+	case CPU_4KEC:
+	case CPU_SB1:
+	case CPU_SB1A:
+	case CPU_4KSC:
+	case CPU_20KC:
+	case CPU_25KF:
+	case CPU_BMIPS32:
+	case CPU_BMIPS3300:
+	case CPU_BMIPS4350:
+	case CPU_BMIPS4380:
+	case CPU_BMIPS5000:
+	case CPU_LOONGSON2:
+	case CPU_R5500:
+		if (m4kc_tlbp_war())
+			uasm_i_nop(p);
+	case CPU_ALCHEMY:
+		tlbw(p);
+		break;
+
+	case CPU_NEVADA:
+		uasm_i_nop(p); /* QED specifies 2 nops hazard */
+		/*
+		 * This branch uses up a mtc0 hazard nop slot and saves
+		 * a nop after the tlbw instruction.
+		 */
+		uasm_il_bgezl(p, r, 0, label_tlbw_hazard);
+		tlbw(p);
+		uasm_l_tlbw_hazard(l, *p);
+		break;
+
+	case CPU_RM7000:
+		uasm_i_nop(p);
+		uasm_i_nop(p);
+		uasm_i_nop(p);
+		uasm_i_nop(p);
+		tlbw(p);
+		break;
+
+	case CPU_RM9000:
+		/*
+		 * When the JTLB is updated by tlbwi or tlbwr, a subsequent
+		 * use of the JTLB for instructions should not occur for 4
+		 * cpu cycles and use for data translations should not occur
+		 * for 3 cpu cycles.
+		 */
+		uasm_i_ssnop(p);
+		uasm_i_ssnop(p);
+		uasm_i_ssnop(p);
+		uasm_i_ssnop(p);
+		tlbw(p);
+		uasm_i_ssnop(p);
+		uasm_i_ssnop(p);
+		uasm_i_ssnop(p);
+		uasm_i_ssnop(p);
+		break;
+
+	case CPU_VR4111:
+	case CPU_VR4121:
+	case CPU_VR4122:
+	case CPU_VR4181:
+	case CPU_VR4181A:
+		uasm_i_nop(p);
+		uasm_i_nop(p);
+		tlbw(p);
+		uasm_i_nop(p);
+		uasm_i_nop(p);
+		break;
+
+	case CPU_VR4131:
+	case CPU_VR4133:
+	case CPU_R5432:
+		uasm_i_nop(p);
+		uasm_i_nop(p);
+		tlbw(p);
+		break;
+
+	case CPU_JZRISC:
+		tlbw(p);
+		uasm_i_nop(p);
+		break;
+
+	default:
+		panic("No TLB refill handler yet (CPU type: %d)",
+		      current_cpu_data.cputype);
+		break;
+	}
+}
+
+static __cpuinit __maybe_unused void build_convert_pte_to_entrylo(u32 **p,
+								  unsigned int reg)
+{
+	if (kernel_uses_smartmips_rixi) {
+		UASM_i_SRL(p, reg, reg, ilog2(_PAGE_NO_EXEC));
+		UASM_i_ROTR(p, reg, reg, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+	} else {
+#ifdef CONFIG_64BIT_PHYS_ADDR
+		uasm_i_dsrl_safe(p, reg, reg, ilog2(_PAGE_GLOBAL));
+#else
+		UASM_i_SRL(p, reg, reg, ilog2(_PAGE_GLOBAL));
+#endif
+	}
+}
+
+#ifdef CONFIG_HUGETLB_PAGE
+
+static __cpuinit void build_restore_pagemask(u32 **p,
+					     struct uasm_reloc **r,
+					     unsigned int tmp,
+					     enum label_id lid,
+					     int restore_scratch)
+{
+	if (restore_scratch) {
+		/* Reset default page size */
+		if (PM_DEFAULT_MASK >> 16) {
+			uasm_i_lui(p, tmp, PM_DEFAULT_MASK >> 16);
+			uasm_i_ori(p, tmp, tmp, PM_DEFAULT_MASK & 0xffff);
+			uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+			uasm_il_b(p, r, lid);
+		} else if (PM_DEFAULT_MASK) {
+			uasm_i_ori(p, tmp, 0, PM_DEFAULT_MASK);
+			uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+			uasm_il_b(p, r, lid);
+		} else {
+			uasm_i_mtc0(p, 0, C0_PAGEMASK);
+			uasm_il_b(p, r, lid);
+		}
+		if (scratch_reg > 0)
+			UASM_i_MFC0(p, 1, 31, scratch_reg);
+		else
+			UASM_i_LW(p, 1, scratchpad_offset(0), 0);
+	} else {
+		/* Reset default page size */
+		if (PM_DEFAULT_MASK >> 16) {
+			uasm_i_lui(p, tmp, PM_DEFAULT_MASK >> 16);
+			uasm_i_ori(p, tmp, tmp, PM_DEFAULT_MASK & 0xffff);
+			uasm_il_b(p, r, lid);
+			uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+		} else if (PM_DEFAULT_MASK) {
+			uasm_i_ori(p, tmp, 0, PM_DEFAULT_MASK);
+			uasm_il_b(p, r, lid);
+			uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+		} else {
+			uasm_il_b(p, r, lid);
+			uasm_i_mtc0(p, 0, C0_PAGEMASK);
+		}
+	}
+}
+
+static __cpuinit void build_huge_tlb_write_entry(u32 **p,
+						 struct uasm_label **l,
+						 struct uasm_reloc **r,
+						 unsigned int tmp,
+						 enum tlb_write_entry wmode,
+						 int restore_scratch)
+{
+	/* Set huge page tlb entry size */
+	uasm_i_lui(p, tmp, PM_HUGE_MASK >> 16);
+	uasm_i_ori(p, tmp, tmp, PM_HUGE_MASK & 0xffff);
+	uasm_i_mtc0(p, tmp, C0_PAGEMASK);
+
+	build_tlb_write_entry(p, l, r, wmode);
+
+	build_restore_pagemask(p, r, tmp, label_leave, restore_scratch);
+}
+
+/*
+ * Check if Huge PTE is present, if so then jump to LABEL.
+ */
+static void __cpuinit
+build_is_huge_pte(u32 **p, struct uasm_reloc **r, unsigned int tmp,
+		unsigned int pmd, int lid)
+{
+	UASM_i_LW(p, tmp, 0, pmd);
+	if (use_bbit_insns()) {
+		uasm_il_bbit1(p, r, tmp, ilog2(_PAGE_HUGE), lid);
+	} else {
+		uasm_i_andi(p, tmp, tmp, _PAGE_HUGE);
+		uasm_il_bnez(p, r, tmp, lid);
+	}
+}
+
+static __cpuinit void build_huge_update_entries(u32 **p,
+						unsigned int pte,
+						unsigned int tmp)
+{
+	int small_sequence;
+
+	/*
+	 * A huge PTE describes an area the size of the
+	 * configured huge page size. This is twice the
+	 * of the large TLB entry size we intend to use.
+	 * A TLB entry half the size of the configured
+	 * huge page size is configured into entrylo0
+	 * and entrylo1 to cover the contiguous huge PTE
+	 * address space.
+	 */
+	small_sequence = (HPAGE_SIZE >> 7) < 0x10000;
+
+	/* We can clobber tmp.  It isn't used after this.*/
+	if (!small_sequence)
+		uasm_i_lui(p, tmp, HPAGE_SIZE >> (7 + 16));
+
+	build_convert_pte_to_entrylo(p, pte);
+	UASM_i_MTC0(p, pte, C0_ENTRYLO0); /* load it */
+	/* convert to entrylo1 */
+	if (small_sequence)
+		UASM_i_ADDIU(p, pte, pte, HPAGE_SIZE >> 7);
+	else
+		UASM_i_ADDU(p, pte, pte, tmp);
+
+	UASM_i_MTC0(p, pte, C0_ENTRYLO1); /* load it */
+}
+
+static __cpuinit void build_huge_handler_tail(u32 **p,
+					      struct uasm_reloc **r,
+					      struct uasm_label **l,
+					      unsigned int pte,
+					      unsigned int ptr)
+{
+#ifdef CONFIG_SMP
+	UASM_i_SC(p, pte, 0, ptr);
+	uasm_il_beqz(p, r, pte, label_tlb_huge_update);
+	UASM_i_LW(p, pte, 0, ptr); /* Needed because SC killed our PTE */
+#else
+	UASM_i_SW(p, pte, 0, ptr);
+#endif
+	build_huge_update_entries(p, pte, ptr);
+	build_huge_tlb_write_entry(p, l, r, pte, tlb_indexed, 0);
+}
+#endif /* CONFIG_HUGETLB_PAGE */
+
+#ifdef CONFIG_64BIT
+/*
+ * TMP and PTR are scratch.
+ * TMP will be clobbered, PTR will hold the pmd entry.
+ */
+static void __cpuinit
+build_get_pmde64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
+		 unsigned int tmp, unsigned int ptr)
+{
+#ifndef CONFIG_MIPS_PGD_C0_CONTEXT
+	long pgdc = (long)pgd_current;
+#endif
+	/*
+	 * The vmalloc handling is not in the hotpath.
+	 */
+	uasm_i_dmfc0(p, tmp, C0_BADVADDR);
+
+	if (check_for_high_segbits) {
+		/*
+		 * The kernel currently implicitely assumes that the
+		 * MIPS SEGBITS parameter for the processor is
+		 * (PGDIR_SHIFT+PGDIR_BITS) or less, and will never
+		 * allocate virtual addresses outside the maximum
+		 * range for SEGBITS = (PGDIR_SHIFT+PGDIR_BITS). But
+		 * that doesn't prevent user code from accessing the
+		 * higher xuseg addresses.  Here, we make sure that
+		 * everything but the lower xuseg addresses goes down
+		 * the module_alloc/vmalloc path.
+		 */
+		uasm_i_dsrl_safe(p, ptr, tmp, PGDIR_SHIFT + PGD_ORDER + PAGE_SHIFT - 3);
+		uasm_il_bnez(p, r, ptr, label_vmalloc);
+	} else {
+		uasm_il_bltz(p, r, tmp, label_vmalloc);
+	}
+	/* No uasm_i_nop needed here, since the next insn doesn't touch TMP. */
+
+#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
+	if (pgd_reg != -1) {
+		/* pgd is in pgd_reg */
+		UASM_i_MFC0(p, ptr, 31, pgd_reg);
+	} else {
+		/*
+		 * &pgd << 11 stored in CONTEXT [23..63].
+		 */
+		UASM_i_MFC0(p, ptr, C0_CONTEXT);
+
+		/* Clear lower 23 bits of context. */
+		uasm_i_dins(p, ptr, 0, 0, 23);
+
+		/* 1 0  1 0 1  << 6  xkphys cached */
+		uasm_i_ori(p, ptr, ptr, 0x540);
+		uasm_i_drotr(p, ptr, ptr, 11);
+	}
+#elif defined(CONFIG_SMP)
+# ifdef  CONFIG_MIPS_MT_SMTC
+	/*
+	 * SMTC uses TCBind value as "CPU" index
+	 */
+	uasm_i_mfc0(p, ptr, C0_TCBIND);
+	uasm_i_dsrl_safe(p, ptr, ptr, 19);
+# else
+	/*
+	 * 64 bit SMP running in XKPHYS has smp_processor_id() << 3
+	 * stored in CONTEXT.
+	 */
+	uasm_i_dmfc0(p, ptr, C0_CONTEXT);
+	uasm_i_dsrl_safe(p, ptr, ptr, 23);
+# endif
+	UASM_i_LA_mostly(p, tmp, pgdc);
+	uasm_i_daddu(p, ptr, ptr, tmp);
+	uasm_i_dmfc0(p, tmp, C0_BADVADDR);
+	uasm_i_ld(p, ptr, uasm_rel_lo(pgdc), ptr);
+#else
+	UASM_i_LA_mostly(p, ptr, pgdc);
+	uasm_i_ld(p, ptr, uasm_rel_lo(pgdc), ptr);
+#endif
+
+	uasm_l_vmalloc_done(l, *p);
+
+	/* get pgd offset in bytes */
+	uasm_i_dsrl_safe(p, tmp, tmp, PGDIR_SHIFT - 3);
+
+	uasm_i_andi(p, tmp, tmp, (PTRS_PER_PGD - 1)<<3);
+	uasm_i_daddu(p, ptr, ptr, tmp); /* add in pgd offset */
+#ifndef __PAGETABLE_PMD_FOLDED
+	uasm_i_dmfc0(p, tmp, C0_BADVADDR); /* get faulting address */
+	uasm_i_ld(p, ptr, 0, ptr); /* get pmd pointer */
+	uasm_i_dsrl_safe(p, tmp, tmp, PMD_SHIFT-3); /* get pmd offset in bytes */
+	uasm_i_andi(p, tmp, tmp, (PTRS_PER_PMD - 1)<<3);
+	uasm_i_daddu(p, ptr, ptr, tmp); /* add in pmd offset */
+#endif
+}
+
+/*
+ * BVADDR is the faulting address, PTR is scratch.
+ * PTR will hold the pgd for vmalloc.
+ */
+static void __cpuinit
+build_get_pgd_vmalloc64(u32 **p, struct uasm_label **l, struct uasm_reloc **r,
+			unsigned int bvaddr, unsigned int ptr,
+			enum vmalloc64_mode mode)
+{
+	long swpd = (long)swapper_pg_dir;
+	int single_insn_swpd;
+	int did_vmalloc_branch = 0;
+
+	single_insn_swpd = uasm_in_compat_space_p(swpd) && !uasm_rel_lo(swpd);
+
+	uasm_l_vmalloc(l, *p);
+
+	if (mode != not_refill && check_for_high_segbits) {
+		if (single_insn_swpd) {
+			uasm_il_bltz(p, r, bvaddr, label_vmalloc_done);
+			uasm_i_lui(p, ptr, uasm_rel_hi(swpd));
+			did_vmalloc_branch = 1;
+			/* fall through */
+		} else {
+			uasm_il_bgez(p, r, bvaddr, label_large_segbits_fault);
+		}
+	}
+	if (!did_vmalloc_branch) {
+		if (uasm_in_compat_space_p(swpd) && !uasm_rel_lo(swpd)) {
+			uasm_il_b(p, r, label_vmalloc_done);
+			uasm_i_lui(p, ptr, uasm_rel_hi(swpd));
+		} else {
+			UASM_i_LA_mostly(p, ptr, swpd);
+			uasm_il_b(p, r, label_vmalloc_done);
+			if (uasm_in_compat_space_p(swpd))
+				uasm_i_addiu(p, ptr, ptr, uasm_rel_lo(swpd));
+			else
+				uasm_i_daddiu(p, ptr, ptr, uasm_rel_lo(swpd));
+		}
+	}
+	if (mode != not_refill && check_for_high_segbits) {
+		uasm_l_large_segbits_fault(l, *p);
+		/*
+		 * We get here if we are an xsseg address, or if we are
+		 * an xuseg address above (PGDIR_SHIFT+PGDIR_BITS) boundary.
+		 *
+		 * Ignoring xsseg (assume disabled so would generate
+		 * (address errors?), the only remaining possibility
+		 * is the upper xuseg addresses.  On processors with
+		 * TLB_SEGBITS <= PGDIR_SHIFT+PGDIR_BITS, these
+		 * addresses would have taken an address error. We try
+		 * to mimic that here by taking a load/istream page
+		 * fault.
+		 */
+		UASM_i_LA(p, ptr, (unsigned long)tlb_do_page_fault_0);
+		uasm_i_jr(p, ptr);
+
+		if (mode == refill_scratch) {
+			if (scratch_reg > 0)
+				UASM_i_MFC0(p, 1, 31, scratch_reg);
+			else
+				UASM_i_LW(p, 1, scratchpad_offset(0), 0);
+		} else {
+			uasm_i_nop(p);
+		}
+	}
+}
+
+#else /* !CONFIG_64BIT */
+
+/*
+ * TMP and PTR are scratch.
+ * TMP will be clobbered, PTR will hold the pgd entry.
+ */
+static void __cpuinit __maybe_unused
+build_get_pgde32(u32 **p, unsigned int tmp, unsigned int ptr)
+{
+	long pgdc = (long)pgd_current;
+
+	/* 32 bit SMP has smp_processor_id() stored in CONTEXT. */
+#ifdef CONFIG_SMP
+#ifdef  CONFIG_MIPS_MT_SMTC
+	/*
+	 * SMTC uses TCBind value as "CPU" index
+	 */
+	uasm_i_mfc0(p, ptr, C0_TCBIND);
+	UASM_i_LA_mostly(p, tmp, pgdc);
+	uasm_i_srl(p, ptr, ptr, 19);
+#else
+	/*
+	 * smp_processor_id() << 3 is stored in CONTEXT.
+         */
+	uasm_i_mfc0(p, ptr, C0_CONTEXT);
+	UASM_i_LA_mostly(p, tmp, pgdc);
+	uasm_i_srl(p, ptr, ptr, 23);
+#endif
+	uasm_i_addu(p, ptr, tmp, ptr);
+#else
+	UASM_i_LA_mostly(p, ptr, pgdc);
+#endif
+	uasm_i_mfc0(p, tmp, C0_BADVADDR); /* get faulting address */
+	uasm_i_lw(p, ptr, uasm_rel_lo(pgdc), ptr);
+	uasm_i_srl(p, tmp, tmp, PGDIR_SHIFT); /* get pgd only bits */
+	uasm_i_sll(p, tmp, tmp, PGD_T_LOG2);
+	uasm_i_addu(p, ptr, ptr, tmp); /* add in pgd offset */
+}
+
+#endif /* !CONFIG_64BIT */
+
+static void __cpuinit build_adjust_context(u32 **p, unsigned int ctx)
+{
+	unsigned int shift = 4 - (PTE_T_LOG2 + 1) + PAGE_SHIFT - 12;
+	unsigned int mask = (PTRS_PER_PTE / 2 - 1) << (PTE_T_LOG2 + 1);
+
+	switch (current_cpu_type()) {
+	case CPU_VR41XX:
+	case CPU_VR4111:
+	case CPU_VR4121:
+	case CPU_VR4122:
+	case CPU_VR4131:
+	case CPU_VR4181:
+	case CPU_VR4181A:
+	case CPU_VR4133:
+		shift += 2;
+		break;
+
+	default:
+		break;
+	}
+
+	if (shift)
+		UASM_i_SRL(p, ctx, ctx, shift);
+	uasm_i_andi(p, ctx, ctx, mask);
+}
+
+static void __cpuinit build_get_ptep(u32 **p, unsigned int tmp, unsigned int ptr)
+{
+	/*
+	 * Bug workaround for the Nevada. It seems as if under certain
+	 * circumstances the move from cp0_context might produce a
+	 * bogus result when the mfc0 instruction and its consumer are
+	 * in a different cacheline or a load instruction, probably any
+	 * memory reference, is between them.
+	 */
+	switch (current_cpu_type()) {
+	case CPU_NEVADA:
+		UASM_i_LW(p, ptr, 0, ptr);
+		GET_CONTEXT(p, tmp); /* get context reg */
+		break;
+
+	default:
+		GET_CONTEXT(p, tmp); /* get context reg */
+		UASM_i_LW(p, ptr, 0, ptr);
+		break;
+	}
+
+	build_adjust_context(p, tmp);
+	UASM_i_ADDU(p, ptr, ptr, tmp); /* add in offset */
+}
+
+static void __cpuinit build_update_entries(u32 **p, unsigned int tmp,
+					unsigned int ptep)
+{
+	/*
+	 * 64bit address support (36bit on a 32bit CPU) in a 32bit
+	 * Kernel is a special case. Only a few CPUs use it.
+	 */
+#ifdef CONFIG_64BIT_PHYS_ADDR
+	if (cpu_has_64bits) {
+		uasm_i_ld(p, tmp, 0, ptep); /* get even pte */
+		uasm_i_ld(p, ptep, sizeof(pte_t), ptep); /* get odd pte */
+		if (kernel_uses_smartmips_rixi) {
+			UASM_i_SRL(p, tmp, tmp, ilog2(_PAGE_NO_EXEC));
+			UASM_i_SRL(p, ptep, ptep, ilog2(_PAGE_NO_EXEC));
+			UASM_i_ROTR(p, tmp, tmp, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+			UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
+			UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+		} else {
+			uasm_i_dsrl_safe(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); /* convert to entrylo0 */
+			UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
+			uasm_i_dsrl_safe(p, ptep, ptep, ilog2(_PAGE_GLOBAL)); /* convert to entrylo1 */
+		}
+		UASM_i_MTC0(p, ptep, C0_ENTRYLO1); /* load it */
+	} else {
+		int pte_off_even = sizeof(pte_t) / 2;
+		int pte_off_odd = pte_off_even + sizeof(pte_t);
+
+		/* The pte entries are pre-shifted */
+		uasm_i_lw(p, tmp, pte_off_even, ptep); /* get even pte */
+		UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
+		uasm_i_lw(p, ptep, pte_off_odd, ptep); /* get odd pte */
+		UASM_i_MTC0(p, ptep, C0_ENTRYLO1); /* load it */
+	}
+#else
+	UASM_i_LW(p, tmp, 0, ptep); /* get even pte */
+	UASM_i_LW(p, ptep, sizeof(pte_t), ptep); /* get odd pte */
+	if (r45k_bvahwbug())
+		build_tlb_probe_entry(p);
+	if (kernel_uses_smartmips_rixi) {
+		UASM_i_SRL(p, tmp, tmp, ilog2(_PAGE_NO_EXEC));
+		UASM_i_SRL(p, ptep, ptep, ilog2(_PAGE_NO_EXEC));
+		UASM_i_ROTR(p, tmp, tmp, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+		if (r4k_250MHZhwbug())
+			UASM_i_MTC0(p, 0, C0_ENTRYLO0);
+		UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
+		UASM_i_ROTR(p, ptep, ptep, ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+	} else {
+		UASM_i_SRL(p, tmp, tmp, ilog2(_PAGE_GLOBAL)); /* convert to entrylo0 */
+		if (r4k_250MHZhwbug())
+			UASM_i_MTC0(p, 0, C0_ENTRYLO0);
+		UASM_i_MTC0(p, tmp, C0_ENTRYLO0); /* load it */
+		UASM_i_SRL(p, ptep, ptep, ilog2(_PAGE_GLOBAL)); /* convert to entrylo1 */
+		if (r45k_bvahwbug())
+			uasm_i_mfc0(p, tmp, C0_INDEX);
+	}
+	if (r4k_250MHZhwbug())
+		UASM_i_MTC0(p, 0, C0_ENTRYLO1);
+	UASM_i_MTC0(p, ptep, C0_ENTRYLO1); /* load it */
+#endif
+}
+
+struct mips_huge_tlb_info {
+	int huge_pte;
+	int restore_scratch;
+};
+
+static struct mips_huge_tlb_info __cpuinit
+build_fast_tlb_refill_handler (u32 **p, struct uasm_label **l,
+			       struct uasm_reloc **r, unsigned int tmp,
+			       unsigned int ptr, int c0_scratch)
+{
+	struct mips_huge_tlb_info rv;
+	unsigned int even, odd;
+	int vmalloc_branch_delay_filled = 0;
+	const int scratch = 1; /* Our extra working register */
+
+	rv.huge_pte = scratch;
+	rv.restore_scratch = 0;
+
+	if (check_for_high_segbits) {
+		UASM_i_MFC0(p, tmp, C0_BADVADDR);
+
+		if (pgd_reg != -1)
+			UASM_i_MFC0(p, ptr, 31, pgd_reg);
+		else
+			UASM_i_MFC0(p, ptr, C0_CONTEXT);
+
+		if (c0_scratch >= 0)
+			UASM_i_MTC0(p, scratch, 31, c0_scratch);
+		else
+			UASM_i_SW(p, scratch, scratchpad_offset(0), 0);
+
+		uasm_i_dsrl_safe(p, scratch, tmp,
+				 PGDIR_SHIFT + PGD_ORDER + PAGE_SHIFT - 3);
+		uasm_il_bnez(p, r, scratch, label_vmalloc);
+
+		if (pgd_reg == -1) {
+			vmalloc_branch_delay_filled = 1;
+			/* Clear lower 23 bits of context. */
+			uasm_i_dins(p, ptr, 0, 0, 23);
+		}
+	} else {
+		if (pgd_reg != -1)
+			UASM_i_MFC0(p, ptr, 31, pgd_reg);
+		else
+			UASM_i_MFC0(p, ptr, C0_CONTEXT);
+
+		UASM_i_MFC0(p, tmp, C0_BADVADDR);
+
+		if (c0_scratch >= 0)
+			UASM_i_MTC0(p, scratch, 31, c0_scratch);
+		else
+			UASM_i_SW(p, scratch, scratchpad_offset(0), 0);
+
+		if (pgd_reg == -1)
+			/* Clear lower 23 bits of context. */
+			uasm_i_dins(p, ptr, 0, 0, 23);
+
+		uasm_il_bltz(p, r, tmp, label_vmalloc);
+	}
+
+	if (pgd_reg == -1) {
+		vmalloc_branch_delay_filled = 1;
+		/* 1 0  1 0 1  << 6  xkphys cached */
+		uasm_i_ori(p, ptr, ptr, 0x540);
+		uasm_i_drotr(p, ptr, ptr, 11);
+	}
+
+#ifdef __PAGETABLE_PMD_FOLDED
+#define LOC_PTEP scratch
+#else
+#define LOC_PTEP ptr
+#endif
+
+	if (!vmalloc_branch_delay_filled)
+		/* get pgd offset in bytes */
+		uasm_i_dsrl_safe(p, scratch, tmp, PGDIR_SHIFT - 3);
+
+	uasm_l_vmalloc_done(l, *p);
+
+	/*
+	 *                         tmp          ptr
+	 * fall-through case =   badvaddr  *pgd_current
+	 * vmalloc case      =   badvaddr  swapper_pg_dir
+	 */
+
+	if (vmalloc_branch_delay_filled)
+		/* get pgd offset in bytes */
+		uasm_i_dsrl_safe(p, scratch, tmp, PGDIR_SHIFT - 3);
+
+#ifdef __PAGETABLE_PMD_FOLDED
+	GET_CONTEXT(p, tmp); /* get context reg */
+#endif
+	uasm_i_andi(p, scratch, scratch, (PTRS_PER_PGD - 1) << 3);
+
+	if (use_lwx_insns()) {
+		UASM_i_LWX(p, LOC_PTEP, scratch, ptr);
+	} else {
+		uasm_i_daddu(p, ptr, ptr, scratch); /* add in pgd offset */
+		uasm_i_ld(p, LOC_PTEP, 0, ptr); /* get pmd pointer */
+	}
+
+#ifndef __PAGETABLE_PMD_FOLDED
+	/* get pmd offset in bytes */
+	uasm_i_dsrl_safe(p, scratch, tmp, PMD_SHIFT - 3);
+	uasm_i_andi(p, scratch, scratch, (PTRS_PER_PMD - 1) << 3);
+	GET_CONTEXT(p, tmp); /* get context reg */
+
+	if (use_lwx_insns()) {
+		UASM_i_LWX(p, scratch, scratch, ptr);
+	} else {
+		uasm_i_daddu(p, ptr, ptr, scratch); /* add in pmd offset */
+		UASM_i_LW(p, scratch, 0, ptr);
+	}
+#endif
+	/* Adjust the context during the load latency. */
+	build_adjust_context(p, tmp);
+
+#ifdef CONFIG_HUGETLB_PAGE
+	uasm_il_bbit1(p, r, scratch, ilog2(_PAGE_HUGE), label_tlb_huge_update);
+	/*
+	 * The in the LWX case we don't want to do the load in the
+	 * delay slot.  It cannot issue in the same cycle and may be
+	 * speculative and unneeded.
+	 */
+	if (use_lwx_insns())
+		uasm_i_nop(p);
+#endif /* CONFIG_HUGETLB_PAGE */
+
+
+	/* build_update_entries */
+	if (use_lwx_insns()) {
+		even = ptr;
+		odd = tmp;
+		UASM_i_LWX(p, even, scratch, tmp);
+		UASM_i_ADDIU(p, tmp, tmp, sizeof(pte_t));
+		UASM_i_LWX(p, odd, scratch, tmp);
+	} else {
+		UASM_i_ADDU(p, ptr, scratch, tmp); /* add in offset */
+		even = tmp;
+		odd = ptr;
+		UASM_i_LW(p, even, 0, ptr); /* get even pte */
+		UASM_i_LW(p, odd, sizeof(pte_t), ptr); /* get odd pte */
+	}
+	if (kernel_uses_smartmips_rixi) {
+		uasm_i_dsrl_safe(p, even, even, ilog2(_PAGE_NO_EXEC));
+		uasm_i_dsrl_safe(p, odd, odd, ilog2(_PAGE_NO_EXEC));
+		uasm_i_drotr(p, even, even,
+			     ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+		UASM_i_MTC0(p, even, C0_ENTRYLO0); /* load it */
+		uasm_i_drotr(p, odd, odd,
+			     ilog2(_PAGE_GLOBAL) - ilog2(_PAGE_NO_EXEC));
+	} else {
+		uasm_i_dsrl_safe(p, even, even, ilog2(_PAGE_GLOBAL));
+		UASM_i_MTC0(p, even, C0_ENTRYLO0); /* load it */
+		uasm_i_dsrl_safe(p, odd, odd, ilog2(_PAGE_GLOBAL));
+	}
+	UASM_i_MTC0(p, odd, C0_ENTRYLO1); /* load it */
+
+	if (c0_scratch >= 0) {
+		UASM_i_MFC0(p, scratch, 31, c0_scratch);
+		build_tlb_write_entry(p, l, r, tlb_random);
+		uasm_l_leave(l, *p);
+		rv.restore_scratch = 1;
+	} else if (PAGE_SHIFT == 14 || PAGE_SHIFT == 13)  {
+		build_tlb_write_entry(p, l, r, tlb_random);
+		uasm_l_leave(l, *p);
+		UASM_i_LW(p, scratch, scratchpad_offset(0), 0);
+	} else {
+		UASM_i_LW(p, scratch, scratchpad_offset(0), 0);
+		build_tlb_write_entry(p, l, r, tlb_random);
+		uasm_l_leave(l, *p);
+		rv.restore_scratch = 1;
+	}
+
+	uasm_i_eret(p); /* return from trap */
+
+	return rv;
+}
+
+/*
+ * For a 64-bit kernel, we are using the 64-bit XTLB refill exception
+ * because EXL == 0.  If we wrap, we can also use the 32 instruction
+ * slots before the XTLB refill exception handler which belong to the
+ * unused TLB refill exception.
+ */
+#define MIPS64_REFILL_INSNS 32
+
+static void __cpuinit build_r4000_tlb_refill_handler(void)
+{
+	u32 *p = tlb_handler;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	u32 *f;
+	unsigned int final_len;
+	struct mips_huge_tlb_info htlb_info __maybe_unused;
+	enum vmalloc64_mode vmalloc_mode __maybe_unused;
+
+	memset(tlb_handler, 0, sizeof(tlb_handler));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+	memset(final_handler, 0, sizeof(final_handler));
+
+	if ((scratch_reg > 0 || scratchpad_available()) && use_bbit_insns()) {
+		htlb_info = build_fast_tlb_refill_handler(&p, &l, &r, K0, K1,
+							  scratch_reg);
+		vmalloc_mode = refill_scratch;
+	} else {
+		htlb_info.huge_pte = K0;
+		htlb_info.restore_scratch = 0;
+		vmalloc_mode = refill_noscratch;
+		/*
+		 * create the plain linear handler
+		 */
+		if (bcm1250_m3_war()) {
+			unsigned int segbits = 44;
+
+			uasm_i_dmfc0(&p, K0, C0_BADVADDR);
+			uasm_i_dmfc0(&p, K1, C0_ENTRYHI);
+			uasm_i_xor(&p, K0, K0, K1);
+			uasm_i_dsrl_safe(&p, K1, K0, 62);
+			uasm_i_dsrl_safe(&p, K0, K0, 12 + 1);
+			uasm_i_dsll_safe(&p, K0, K0, 64 + 12 + 1 - segbits);
+			uasm_i_or(&p, K0, K0, K1);
+			uasm_il_bnez(&p, &r, K0, label_leave);
+			/* No need for uasm_i_nop */
+		}
+
+#ifdef CONFIG_64BIT
+		build_get_pmde64(&p, &l, &r, K0, K1); /* get pmd in K1 */
+#else
+		build_get_pgde32(&p, K0, K1); /* get pgd in K1 */
+#endif
+
+#ifdef CONFIG_HUGETLB_PAGE
+		build_is_huge_pte(&p, &r, K0, K1, label_tlb_huge_update);
+#endif
+
+		build_get_ptep(&p, K0, K1);
+		build_update_entries(&p, K0, K1);
+		build_tlb_write_entry(&p, &l, &r, tlb_random);
+		uasm_l_leave(&l, p);
+		uasm_i_eret(&p); /* return from trap */
+	}
+#ifdef CONFIG_HUGETLB_PAGE
+	uasm_l_tlb_huge_update(&l, p);
+	build_huge_update_entries(&p, htlb_info.huge_pte, K1);
+	build_huge_tlb_write_entry(&p, &l, &r, K0, tlb_random,
+				   htlb_info.restore_scratch);
+#endif
+
+#ifdef CONFIG_64BIT
+	build_get_pgd_vmalloc64(&p, &l, &r, K0, K1, vmalloc_mode);
+#endif
+
+	/*
+	 * Overflow check: For the 64bit handler, we need at least one
+	 * free instruction slot for the wrap-around branch. In worst
+	 * case, if the intended insertion point is a delay slot, we
+	 * need three, with the second nop'ed and the third being
+	 * unused.
+	 */
+	/* Loongson2 ebase is different than r4k, we have more space */
+#if defined(CONFIG_32BIT) || defined(CONFIG_CPU_LOONGSON2)
+	if ((p - tlb_handler) > 64)
+		panic("TLB refill handler space exceeded");
+#else
+	if (((p - tlb_handler) > (MIPS64_REFILL_INSNS * 2) - 1)
+	    || (((p - tlb_handler) > (MIPS64_REFILL_INSNS * 2) - 3)
+		&& uasm_insn_has_bdelay(relocs,
+					tlb_handler + MIPS64_REFILL_INSNS - 3)))
+		panic("TLB refill handler space exceeded");
+#endif
+
+	/*
+	 * Now fold the handler in the TLB refill handler space.
+	 */
+#if defined(CONFIG_32BIT) || defined(CONFIG_CPU_LOONGSON2)
+	f = final_handler;
+	/* Simplest case, just copy the handler. */
+	uasm_copy_handler(relocs, labels, tlb_handler, p, f);
+	final_len = p - tlb_handler;
+#else /* CONFIG_64BIT */
+	f = final_handler + MIPS64_REFILL_INSNS;
+	if ((p - tlb_handler) <= MIPS64_REFILL_INSNS) {
+		/* Just copy the handler. */
+		uasm_copy_handler(relocs, labels, tlb_handler, p, f);
+		final_len = p - tlb_handler;
+	} else {
+#if defined(CONFIG_HUGETLB_PAGE)
+		const enum label_id ls = label_tlb_huge_update;
+#else
+		const enum label_id ls = label_vmalloc;
+#endif
+		u32 *split;
+		int ov = 0;
+		int i;
+
+		for (i = 0; i < ARRAY_SIZE(labels) && labels[i].lab != ls; i++)
+			;
+		BUG_ON(i == ARRAY_SIZE(labels));
+		split = labels[i].addr;
+
+		/*
+		 * See if we have overflown one way or the other.
+		 */
+		if (split > tlb_handler + MIPS64_REFILL_INSNS ||
+		    split < p - MIPS64_REFILL_INSNS)
+			ov = 1;
+
+		if (ov) {
+			/*
+			 * Split two instructions before the end.  One
+			 * for the branch and one for the instruction
+			 * in the delay slot.
+			 */
+			split = tlb_handler + MIPS64_REFILL_INSNS - 2;
+
+			/*
+			 * If the branch would fall in a delay slot,
+			 * we must back up an additional instruction
+			 * so that it is no longer in a delay slot.
+			 */
+			if (uasm_insn_has_bdelay(relocs, split - 1))
+				split--;
+		}
+		/* Copy first part of the handler. */
+		uasm_copy_handler(relocs, labels, tlb_handler, split, f);
+		f += split - tlb_handler;
+
+		if (ov) {
+			/* Insert branch. */
+			uasm_l_split(&l, final_handler);
+			uasm_il_b(&f, &r, label_split);
+			if (uasm_insn_has_bdelay(relocs, split))
+				uasm_i_nop(&f);
+			else {
+				uasm_copy_handler(relocs, labels,
+						  split, split + 1, f);
+				uasm_move_labels(labels, f, f + 1, -1);
+				f++;
+				split++;
+			}
+		}
+
+		/* Copy the rest of the handler. */
+		uasm_copy_handler(relocs, labels, split, p, final_handler);
+		final_len = (f - (final_handler + MIPS64_REFILL_INSNS)) +
+			    (p - split);
+	}
+#endif /* CONFIG_64BIT */
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB refill handler (%u instructions).\n",
+		 final_len);
+
+	memcpy((void *)ebase, final_handler, 0x100);
+
+	dump_handler((u32 *)ebase, 64);
+}
+
+/*
+ * 128 instructions for the fastpath handler is generous and should
+ * never be exceeded.
+ */
+#define FASTPATH_SIZE 128
+
+u32 handle_tlbl[FASTPATH_SIZE] __cacheline_aligned;
+u32 handle_tlbs[FASTPATH_SIZE] __cacheline_aligned;
+u32 handle_tlbm[FASTPATH_SIZE] __cacheline_aligned;
+#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
+u32 tlbmiss_handler_setup_pgd[16] __cacheline_aligned;
+
+static void __cpuinit build_r4000_setup_pgd(void)
+{
+	const int a0 = 4;
+	const int a1 = 5;
+	u32 *p = tlbmiss_handler_setup_pgd;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+
+	memset(tlbmiss_handler_setup_pgd, 0, sizeof(tlbmiss_handler_setup_pgd));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	pgd_reg = allocate_kscratch();
+
+	if (pgd_reg == -1) {
+		/* PGD << 11 in c0_Context */
+		/*
+		 * If it is a ckseg0 address, convert to a physical
+		 * address.  Shifting right by 29 and adding 4 will
+		 * result in zero for these addresses.
+		 *
+		 */
+		UASM_i_SRA(&p, a1, a0, 29);
+		UASM_i_ADDIU(&p, a1, a1, 4);
+		uasm_il_bnez(&p, &r, a1, label_tlbl_goaround1);
+		uasm_i_nop(&p);
+		uasm_i_dinsm(&p, a0, 0, 29, 64 - 29);
+		uasm_l_tlbl_goaround1(&l, p);
+		UASM_i_SLL(&p, a0, a0, 11);
+		uasm_i_jr(&p, 31);
+		UASM_i_MTC0(&p, a0, C0_CONTEXT);
+	} else {
+		/* PGD in c0_KScratch */
+		uasm_i_jr(&p, 31);
+		UASM_i_MTC0(&p, a0, 31, pgd_reg);
+	}
+	if (p - tlbmiss_handler_setup_pgd > ARRAY_SIZE(tlbmiss_handler_setup_pgd))
+		panic("tlbmiss_handler_setup_pgd space exceeded");
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote tlbmiss_handler_setup_pgd (%u instructions).\n",
+		 (unsigned int)(p - tlbmiss_handler_setup_pgd));
+
+	dump_handler(tlbmiss_handler_setup_pgd,
+		     ARRAY_SIZE(tlbmiss_handler_setup_pgd));
+}
+#endif
+
+static void __cpuinit
+iPTE_LW(u32 **p, unsigned int pte, unsigned int ptr)
+{
+#ifdef CONFIG_SMP
+# ifdef CONFIG_64BIT_PHYS_ADDR
+	if (cpu_has_64bits)
+		uasm_i_lld(p, pte, 0, ptr);
+	else
+# endif
+		UASM_i_LL(p, pte, 0, ptr);
+#else
+# ifdef CONFIG_64BIT_PHYS_ADDR
+	if (cpu_has_64bits)
+		uasm_i_ld(p, pte, 0, ptr);
+	else
+# endif
+		UASM_i_LW(p, pte, 0, ptr);
+#endif
+}
+
+static void __cpuinit
+iPTE_SW(u32 **p, struct uasm_reloc **r, unsigned int pte, unsigned int ptr,
+	unsigned int mode)
+{
+#ifdef CONFIG_64BIT_PHYS_ADDR
+	unsigned int hwmode = mode & (_PAGE_VALID | _PAGE_DIRTY);
+#endif
+
+	uasm_i_ori(p, pte, pte, mode);
+#ifdef CONFIG_SMP
+# ifdef CONFIG_64BIT_PHYS_ADDR
+	if (cpu_has_64bits)
+		uasm_i_scd(p, pte, 0, ptr);
+	else
+# endif
+		UASM_i_SC(p, pte, 0, ptr);
+
+	if (r10000_llsc_war())
+		uasm_il_beqzl(p, r, pte, label_smp_pgtable_change);
+	else
+		uasm_il_beqz(p, r, pte, label_smp_pgtable_change);
+
+# ifdef CONFIG_64BIT_PHYS_ADDR
+	if (!cpu_has_64bits) {
+		/* no uasm_i_nop needed */
+		uasm_i_ll(p, pte, sizeof(pte_t) / 2, ptr);
+		uasm_i_ori(p, pte, pte, hwmode);
+		uasm_i_sc(p, pte, sizeof(pte_t) / 2, ptr);
+		uasm_il_beqz(p, r, pte, label_smp_pgtable_change);
+		/* no uasm_i_nop needed */
+		uasm_i_lw(p, pte, 0, ptr);
+	} else
+		uasm_i_nop(p);
+# else
+	uasm_i_nop(p);
+# endif
+#else
+# ifdef CONFIG_64BIT_PHYS_ADDR
+	if (cpu_has_64bits)
+		uasm_i_sd(p, pte, 0, ptr);
+	else
+# endif
+		UASM_i_SW(p, pte, 0, ptr);
+
+# ifdef CONFIG_64BIT_PHYS_ADDR
+	if (!cpu_has_64bits) {
+		uasm_i_lw(p, pte, sizeof(pte_t) / 2, ptr);
+		uasm_i_ori(p, pte, pte, hwmode);
+		uasm_i_sw(p, pte, sizeof(pte_t) / 2, ptr);
+		uasm_i_lw(p, pte, 0, ptr);
+	}
+# endif
+#endif
+}
+
+/*
+ * Check if PTE is present, if not then jump to LABEL. PTR points to
+ * the page table where this PTE is located, PTE will be re-loaded
+ * with it's original value.
+ */
+static void __cpuinit
+build_pte_present(u32 **p, struct uasm_reloc **r,
+		  int pte, int ptr, int scratch, enum label_id lid)
+{
+	int t = scratch >= 0 ? scratch : pte;
+
+	if (kernel_uses_smartmips_rixi) {
+		if (use_bbit_insns()) {
+			uasm_il_bbit0(p, r, pte, ilog2(_PAGE_PRESENT), lid);
+			uasm_i_nop(p);
+		} else {
+			uasm_i_andi(p, t, pte, _PAGE_PRESENT);
+			uasm_il_beqz(p, r, t, lid);
+			if (pte == t)
+				/* You lose the SMP race :-(*/
+				iPTE_LW(p, pte, ptr);
+		}
+	} else {
+		uasm_i_andi(p, t, pte, _PAGE_PRESENT | _PAGE_READ);
+		uasm_i_xori(p, t, t, _PAGE_PRESENT | _PAGE_READ);
+		uasm_il_bnez(p, r, t, lid);
+		if (pte == t)
+			/* You lose the SMP race :-(*/
+			iPTE_LW(p, pte, ptr);
+	}
+}
+
+/* Make PTE valid, store result in PTR. */
+static void __cpuinit
+build_make_valid(u32 **p, struct uasm_reloc **r, unsigned int pte,
+		 unsigned int ptr)
+{
+	unsigned int mode = _PAGE_VALID | _PAGE_ACCESSED;
+
+	iPTE_SW(p, r, pte, ptr, mode);
+}
+
+/*
+ * Check if PTE can be written to, if not branch to LABEL. Regardless
+ * restore PTE with value from PTR when done.
+ */
+static void __cpuinit
+build_pte_writable(u32 **p, struct uasm_reloc **r,
+		   unsigned int pte, unsigned int ptr, int scratch,
+		   enum label_id lid)
+{
+	int t = scratch >= 0 ? scratch : pte;
+
+	uasm_i_andi(p, t, pte, _PAGE_PRESENT | _PAGE_WRITE);
+	uasm_i_xori(p, t, t, _PAGE_PRESENT | _PAGE_WRITE);
+	uasm_il_bnez(p, r, t, lid);
+	if (pte == t)
+		/* You lose the SMP race :-(*/
+		iPTE_LW(p, pte, ptr);
+	else
+		uasm_i_nop(p);
+}
+
+/* Make PTE writable, update software status bits as well, then store
+ * at PTR.
+ */
+static void __cpuinit
+build_make_write(u32 **p, struct uasm_reloc **r, unsigned int pte,
+		 unsigned int ptr)
+{
+	unsigned int mode = (_PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID
+			     | _PAGE_DIRTY);
+
+	iPTE_SW(p, r, pte, ptr, mode);
+}
+
+/*
+ * Check if PTE can be modified, if not branch to LABEL. Regardless
+ * restore PTE with value from PTR when done.
+ */
+static void __cpuinit
+build_pte_modifiable(u32 **p, struct uasm_reloc **r,
+		     unsigned int pte, unsigned int ptr, int scratch,
+		     enum label_id lid)
+{
+	if (use_bbit_insns()) {
+		uasm_il_bbit0(p, r, pte, ilog2(_PAGE_WRITE), lid);
+		uasm_i_nop(p);
+	} else {
+		int t = scratch >= 0 ? scratch : pte;
+		uasm_i_andi(p, t, pte, _PAGE_WRITE);
+		uasm_il_beqz(p, r, t, lid);
+		if (pte == t)
+			/* You lose the SMP race :-(*/
+			iPTE_LW(p, pte, ptr);
+	}
+}
+
+#ifndef CONFIG_MIPS_PGD_C0_CONTEXT
+
+
+/*
+ * R3000 style TLB load/store/modify handlers.
+ */
+
+/*
+ * This places the pte into ENTRYLO0 and writes it with tlbwi.
+ * Then it returns.
+ */
+static void __cpuinit
+build_r3000_pte_reload_tlbwi(u32 **p, unsigned int pte, unsigned int tmp)
+{
+	uasm_i_mtc0(p, pte, C0_ENTRYLO0); /* cp0 delay */
+	uasm_i_mfc0(p, tmp, C0_EPC); /* cp0 delay */
+	uasm_i_tlbwi(p);
+	uasm_i_jr(p, tmp);
+	uasm_i_rfe(p); /* branch delay */
+}
+
+/*
+ * This places the pte into ENTRYLO0 and writes it with tlbwi
+ * or tlbwr as appropriate.  This is because the index register
+ * may have the probe fail bit set as a result of a trap on a
+ * kseg2 access, i.e. without refill.  Then it returns.
+ */
+static void __cpuinit
+build_r3000_tlb_reload_write(u32 **p, struct uasm_label **l,
+			     struct uasm_reloc **r, unsigned int pte,
+			     unsigned int tmp)
+{
+	uasm_i_mfc0(p, tmp, C0_INDEX);
+	uasm_i_mtc0(p, pte, C0_ENTRYLO0); /* cp0 delay */
+	uasm_il_bltz(p, r, tmp, label_r3000_write_probe_fail); /* cp0 delay */
+	uasm_i_mfc0(p, tmp, C0_EPC); /* branch delay */
+	uasm_i_tlbwi(p); /* cp0 delay */
+	uasm_i_jr(p, tmp);
+	uasm_i_rfe(p); /* branch delay */
+	uasm_l_r3000_write_probe_fail(l, *p);
+	uasm_i_tlbwr(p); /* cp0 delay */
+	uasm_i_jr(p, tmp);
+	uasm_i_rfe(p); /* branch delay */
+}
+
+static void __cpuinit
+build_r3000_tlbchange_handler_head(u32 **p, unsigned int pte,
+				   unsigned int ptr)
+{
+	long pgdc = (long)pgd_current;
+
+	uasm_i_mfc0(p, pte, C0_BADVADDR);
+	uasm_i_lui(p, ptr, uasm_rel_hi(pgdc)); /* cp0 delay */
+	uasm_i_lw(p, ptr, uasm_rel_lo(pgdc), ptr);
+	uasm_i_srl(p, pte, pte, 22); /* load delay */
+	uasm_i_sll(p, pte, pte, 2);
+	uasm_i_addu(p, ptr, ptr, pte);
+	uasm_i_mfc0(p, pte, C0_CONTEXT);
+	uasm_i_lw(p, ptr, 0, ptr); /* cp0 delay */
+	uasm_i_andi(p, pte, pte, 0xffc); /* load delay */
+	uasm_i_addu(p, ptr, ptr, pte);
+	uasm_i_lw(p, pte, 0, ptr);
+	uasm_i_tlbp(p); /* load delay */
+}
+
+static void __cpuinit build_r3000_tlb_load_handler(void)
+{
+	u32 *p = handle_tlbl;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+
+	memset(handle_tlbl, 0, sizeof(handle_tlbl));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	build_r3000_tlbchange_handler_head(&p, K0, K1);
+	build_pte_present(&p, &r, K0, K1, -1, label_nopage_tlbl);
+	uasm_i_nop(&p); /* load delay */
+	build_make_valid(&p, &r, K0, K1);
+	build_r3000_tlb_reload_write(&p, &l, &r, K0, K1);
+
+	uasm_l_nopage_tlbl(&l, p);
+	uasm_i_j(&p, (unsigned long)tlb_do_page_fault_0 & 0x0fffffff);
+	uasm_i_nop(&p);
+
+	if ((p - handle_tlbl) > FASTPATH_SIZE)
+		panic("TLB load handler fastpath space exceeded");
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB load handler fastpath (%u instructions).\n",
+		 (unsigned int)(p - handle_tlbl));
+
+	dump_handler(handle_tlbl, ARRAY_SIZE(handle_tlbl));
+}
+
+static void __cpuinit build_r3000_tlb_store_handler(void)
+{
+	u32 *p = handle_tlbs;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+
+	memset(handle_tlbs, 0, sizeof(handle_tlbs));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	build_r3000_tlbchange_handler_head(&p, K0, K1);
+	build_pte_writable(&p, &r, K0, K1, -1, label_nopage_tlbs);
+	uasm_i_nop(&p); /* load delay */
+	build_make_write(&p, &r, K0, K1);
+	build_r3000_tlb_reload_write(&p, &l, &r, K0, K1);
+
+	uasm_l_nopage_tlbs(&l, p);
+	uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
+	uasm_i_nop(&p);
+
+	if ((p - handle_tlbs) > FASTPATH_SIZE)
+		panic("TLB store handler fastpath space exceeded");
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB store handler fastpath (%u instructions).\n",
+		 (unsigned int)(p - handle_tlbs));
+
+	dump_handler(handle_tlbs, ARRAY_SIZE(handle_tlbs));
+}
+
+static void __cpuinit build_r3000_tlb_modify_handler(void)
+{
+	u32 *p = handle_tlbm;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+
+	memset(handle_tlbm, 0, sizeof(handle_tlbm));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	build_r3000_tlbchange_handler_head(&p, K0, K1);
+	build_pte_modifiable(&p, &r, K0, K1,  -1, label_nopage_tlbm);
+	uasm_i_nop(&p); /* load delay */
+	build_make_write(&p, &r, K0, K1);
+	build_r3000_pte_reload_tlbwi(&p, K0, K1);
+
+	uasm_l_nopage_tlbm(&l, p);
+	uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
+	uasm_i_nop(&p);
+
+	if ((p - handle_tlbm) > FASTPATH_SIZE)
+		panic("TLB modify handler fastpath space exceeded");
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB modify handler fastpath (%u instructions).\n",
+		 (unsigned int)(p - handle_tlbm));
+
+	dump_handler(handle_tlbm, ARRAY_SIZE(handle_tlbm));
+}
+#endif /* CONFIG_MIPS_PGD_C0_CONTEXT */
+
+/*
+ * R4000 style TLB load/store/modify handlers.
+ */
+static struct work_registers __cpuinit
+build_r4000_tlbchange_handler_head(u32 **p, struct uasm_label **l,
+				   struct uasm_reloc **r)
+{
+	struct work_registers wr = build_get_work_registers(p);
+
+#ifdef CONFIG_64BIT
+	build_get_pmde64(p, l, r, wr.r1, wr.r2); /* get pmd in ptr */
+#else
+	build_get_pgde32(p, wr.r1, wr.r2); /* get pgd in ptr */
+#endif
+
+#ifdef CONFIG_HUGETLB_PAGE
+	/*
+	 * For huge tlb entries, pmd doesn't contain an address but
+	 * instead contains the tlb pte. Check the PAGE_HUGE bit and
+	 * see if we need to jump to huge tlb processing.
+	 */
+	build_is_huge_pte(p, r, wr.r1, wr.r2, label_tlb_huge_update);
+#endif
+
+	UASM_i_MFC0(p, wr.r1, C0_BADVADDR);
+	UASM_i_LW(p, wr.r2, 0, wr.r2);
+	UASM_i_SRL(p, wr.r1, wr.r1, PAGE_SHIFT + PTE_ORDER - PTE_T_LOG2);
+	uasm_i_andi(p, wr.r1, wr.r1, (PTRS_PER_PTE - 1) << PTE_T_LOG2);
+	UASM_i_ADDU(p, wr.r2, wr.r2, wr.r1);
+
+#ifdef CONFIG_SMP
+	uasm_l_smp_pgtable_change(l, *p);
+#endif
+	iPTE_LW(p, wr.r1, wr.r2); /* get even pte */
+	if (!m4kc_tlbp_war())
+		build_tlb_probe_entry(p);
+	return wr;
+}
+
+static void __cpuinit
+build_r4000_tlbchange_handler_tail(u32 **p, struct uasm_label **l,
+				   struct uasm_reloc **r, unsigned int tmp,
+				   unsigned int ptr)
+{
+	uasm_i_ori(p, ptr, ptr, sizeof(pte_t));
+	uasm_i_xori(p, ptr, ptr, sizeof(pte_t));
+	build_update_entries(p, tmp, ptr);
+	build_tlb_write_entry(p, l, r, tlb_indexed);
+	uasm_l_leave(l, *p);
+	build_restore_work_registers(p);
+	uasm_i_eret(p); /* return from trap */
+
+#ifdef CONFIG_64BIT
+	build_get_pgd_vmalloc64(p, l, r, tmp, ptr, not_refill);
+#endif
+}
+
+static void __cpuinit build_r4000_tlb_load_handler(void)
+{
+	u32 *p = handle_tlbl;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	struct work_registers wr;
+
+	memset(handle_tlbl, 0, sizeof(handle_tlbl));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	if (bcm1250_m3_war()) {
+		unsigned int segbits = 44;
+
+		uasm_i_dmfc0(&p, K0, C0_BADVADDR);
+		uasm_i_dmfc0(&p, K1, C0_ENTRYHI);
+		uasm_i_xor(&p, K0, K0, K1);
+		uasm_i_dsrl_safe(&p, K1, K0, 62);
+		uasm_i_dsrl_safe(&p, K0, K0, 12 + 1);
+		uasm_i_dsll_safe(&p, K0, K0, 64 + 12 + 1 - segbits);
+		uasm_i_or(&p, K0, K0, K1);
+		uasm_il_bnez(&p, &r, K0, label_leave);
+		/* No need for uasm_i_nop */
+	}
+
+	wr = build_r4000_tlbchange_handler_head(&p, &l, &r);
+	build_pte_present(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbl);
+	if (m4kc_tlbp_war())
+		build_tlb_probe_entry(&p);
+
+	if (kernel_uses_smartmips_rixi) {
+		/*
+		 * If the page is not _PAGE_VALID, RI or XI could not
+		 * have triggered it.  Skip the expensive test..
+		 */
+		if (use_bbit_insns()) {
+			uasm_il_bbit0(&p, &r, wr.r1, ilog2(_PAGE_VALID),
+				      label_tlbl_goaround1);
+		} else {
+			uasm_i_andi(&p, wr.r3, wr.r1, _PAGE_VALID);
+			uasm_il_beqz(&p, &r, wr.r3, label_tlbl_goaround1);
+		}
+		uasm_i_nop(&p);
+
+		uasm_i_tlbr(&p);
+		/* Examine  entrylo 0 or 1 based on ptr. */
+		if (use_bbit_insns()) {
+			uasm_i_bbit0(&p, wr.r2, ilog2(sizeof(pte_t)), 8);
+		} else {
+			uasm_i_andi(&p, wr.r3, wr.r2, sizeof(pte_t));
+			uasm_i_beqz(&p, wr.r3, 8);
+		}
+		/* load it in the delay slot*/
+		UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO0);
+		/* load it if ptr is odd */
+		UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO1);
+		/*
+		 * If the entryLo (now in wr.r3) is valid (bit 1), RI or
+		 * XI must have triggered it.
+		 */
+		if (use_bbit_insns()) {
+			uasm_il_bbit1(&p, &r, wr.r3, 1, label_nopage_tlbl);
+			uasm_i_nop(&p);
+			uasm_l_tlbl_goaround1(&l, p);
+		} else {
+			uasm_i_andi(&p, wr.r3, wr.r3, 2);
+			uasm_il_bnez(&p, &r, wr.r3, label_nopage_tlbl);
+			uasm_i_nop(&p);
+		}
+		uasm_l_tlbl_goaround1(&l, p);
+	}
+	build_make_valid(&p, &r, wr.r1, wr.r2);
+	build_r4000_tlbchange_handler_tail(&p, &l, &r, wr.r1, wr.r2);
+
+#ifdef CONFIG_HUGETLB_PAGE
+	/*
+	 * This is the entry point when build_r4000_tlbchange_handler_head
+	 * spots a huge page.
+	 */
+	uasm_l_tlb_huge_update(&l, p);
+	iPTE_LW(&p, wr.r1, wr.r2);
+	build_pte_present(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbl);
+	build_tlb_probe_entry(&p);
+
+	if (kernel_uses_smartmips_rixi) {
+		/*
+		 * If the page is not _PAGE_VALID, RI or XI could not
+		 * have triggered it.  Skip the expensive test..
+		 */
+		if (use_bbit_insns()) {
+			uasm_il_bbit0(&p, &r, wr.r1, ilog2(_PAGE_VALID),
+				      label_tlbl_goaround2);
+		} else {
+			uasm_i_andi(&p, wr.r3, wr.r1, _PAGE_VALID);
+			uasm_il_beqz(&p, &r, wr.r3, label_tlbl_goaround2);
+		}
+		uasm_i_nop(&p);
+
+		uasm_i_tlbr(&p);
+		/* Examine  entrylo 0 or 1 based on ptr. */
+		if (use_bbit_insns()) {
+			uasm_i_bbit0(&p, wr.r2, ilog2(sizeof(pte_t)), 8);
+		} else {
+			uasm_i_andi(&p, wr.r3, wr.r2, sizeof(pte_t));
+			uasm_i_beqz(&p, wr.r3, 8);
+		}
+		/* load it in the delay slot*/
+		UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO0);
+		/* load it if ptr is odd */
+		UASM_i_MFC0(&p, wr.r3, C0_ENTRYLO1);
+		/*
+		 * If the entryLo (now in wr.r3) is valid (bit 1), RI or
+		 * XI must have triggered it.
+		 */
+		if (use_bbit_insns()) {
+			uasm_il_bbit0(&p, &r, wr.r3, 1, label_tlbl_goaround2);
+		} else {
+			uasm_i_andi(&p, wr.r3, wr.r3, 2);
+			uasm_il_beqz(&p, &r, wr.r3, label_tlbl_goaround2);
+		}
+		if (PM_DEFAULT_MASK == 0)
+			uasm_i_nop(&p);
+		/*
+		 * We clobbered C0_PAGEMASK, restore it.  On the other branch
+		 * it is restored in build_huge_tlb_write_entry.
+		 */
+		build_restore_pagemask(&p, &r, wr.r3, label_nopage_tlbl, 0);
+
+		uasm_l_tlbl_goaround2(&l, p);
+	}
+	uasm_i_ori(&p, wr.r1, wr.r1, (_PAGE_ACCESSED | _PAGE_VALID));
+	build_huge_handler_tail(&p, &r, &l, wr.r1, wr.r2);
+#endif
+
+	uasm_l_nopage_tlbl(&l, p);
+	build_restore_work_registers(&p);
+	uasm_i_j(&p, (unsigned long)tlb_do_page_fault_0 & 0x0fffffff);
+	uasm_i_nop(&p);
+
+	if ((p - handle_tlbl) > FASTPATH_SIZE)
+		panic("TLB load handler fastpath space exceeded");
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB load handler fastpath (%u instructions).\n",
+		 (unsigned int)(p - handle_tlbl));
+
+	dump_handler(handle_tlbl, ARRAY_SIZE(handle_tlbl));
+}
+
+static void __cpuinit build_r4000_tlb_store_handler(void)
+{
+	u32 *p = handle_tlbs;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	struct work_registers wr;
+
+	memset(handle_tlbs, 0, sizeof(handle_tlbs));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	wr = build_r4000_tlbchange_handler_head(&p, &l, &r);
+	build_pte_writable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbs);
+	if (m4kc_tlbp_war())
+		build_tlb_probe_entry(&p);
+	build_make_write(&p, &r, wr.r1, wr.r2);
+	build_r4000_tlbchange_handler_tail(&p, &l, &r, wr.r1, wr.r2);
+
+#ifdef CONFIG_HUGETLB_PAGE
+	/*
+	 * This is the entry point when
+	 * build_r4000_tlbchange_handler_head spots a huge page.
+	 */
+	uasm_l_tlb_huge_update(&l, p);
+	iPTE_LW(&p, wr.r1, wr.r2);
+	build_pte_writable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbs);
+	build_tlb_probe_entry(&p);
+	uasm_i_ori(&p, wr.r1, wr.r1,
+		   _PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID | _PAGE_DIRTY);
+	build_huge_handler_tail(&p, &r, &l, wr.r1, wr.r2);
+#endif
+
+	uasm_l_nopage_tlbs(&l, p);
+	build_restore_work_registers(&p);
+	uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
+	uasm_i_nop(&p);
+
+	if ((p - handle_tlbs) > FASTPATH_SIZE)
+		panic("TLB store handler fastpath space exceeded");
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB store handler fastpath (%u instructions).\n",
+		 (unsigned int)(p - handle_tlbs));
+
+	dump_handler(handle_tlbs, ARRAY_SIZE(handle_tlbs));
+}
+
+static void __cpuinit build_r4000_tlb_modify_handler(void)
+{
+	u32 *p = handle_tlbm;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	struct work_registers wr;
+
+	memset(handle_tlbm, 0, sizeof(handle_tlbm));
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	wr = build_r4000_tlbchange_handler_head(&p, &l, &r);
+	build_pte_modifiable(&p, &r, wr.r1, wr.r2, wr.r3, label_nopage_tlbm);
+	if (m4kc_tlbp_war())
+		build_tlb_probe_entry(&p);
+	/* Present and writable bits set, set accessed and dirty bits. */
+	build_make_write(&p, &r, wr.r1, wr.r2);
+	build_r4000_tlbchange_handler_tail(&p, &l, &r, wr.r1, wr.r2);
+
+#ifdef CONFIG_HUGETLB_PAGE
+	/*
+	 * This is the entry point when
+	 * build_r4000_tlbchange_handler_head spots a huge page.
+	 */
+	uasm_l_tlb_huge_update(&l, p);
+	iPTE_LW(&p, wr.r1, wr.r2);
+	build_pte_modifiable(&p, &r, wr.r1, wr.r2,  wr.r3, label_nopage_tlbm);
+	build_tlb_probe_entry(&p);
+	uasm_i_ori(&p, wr.r1, wr.r1,
+		   _PAGE_ACCESSED | _PAGE_MODIFIED | _PAGE_VALID | _PAGE_DIRTY);
+	build_huge_handler_tail(&p, &r, &l, wr.r1, wr.r2);
+#endif
+
+	uasm_l_nopage_tlbm(&l, p);
+	build_restore_work_registers(&p);
+	uasm_i_j(&p, (unsigned long)tlb_do_page_fault_1 & 0x0fffffff);
+	uasm_i_nop(&p);
+
+	if ((p - handle_tlbm) > FASTPATH_SIZE)
+		panic("TLB modify handler fastpath space exceeded");
+
+	uasm_resolve_relocs(relocs, labels);
+	pr_debug("Wrote TLB modify handler fastpath (%u instructions).\n",
+		 (unsigned int)(p - handle_tlbm));
+
+	dump_handler(handle_tlbm, ARRAY_SIZE(handle_tlbm));
+}
+
+void __cpuinit build_tlb_refill_handler(void)
+{
+	/*
+	 * The refill handler is generated per-CPU, multi-node systems
+	 * may have local storage for it. The other handlers are only
+	 * needed once.
+	 */
+	static int run_once = 0;
+
+#ifdef CONFIG_64BIT
+	check_for_high_segbits = current_cpu_data.vmbits > (PGDIR_SHIFT + PGD_ORDER + PAGE_SHIFT - 3);
+#endif
+
+	switch (current_cpu_type()) {
+	case CPU_R2000:
+	case CPU_R3000:
+	case CPU_R3000A:
+	case CPU_R3081E:
+	case CPU_TX3912:
+	case CPU_TX3922:
+	case CPU_TX3927:
+#ifndef CONFIG_MIPS_PGD_C0_CONTEXT
+		build_r3000_tlb_refill_handler();
+		if (!run_once) {
+			build_r3000_tlb_load_handler();
+			build_r3000_tlb_store_handler();
+			build_r3000_tlb_modify_handler();
+			run_once++;
+		}
+#else
+		panic("No R3000 TLB refill handler");
+#endif
+		break;
+
+	case CPU_R6000:
+	case CPU_R6000A:
+		panic("No R6000 TLB refill handler yet");
+		break;
+
+	case CPU_R8000:
+		panic("No R8000 TLB refill handler yet");
+		break;
+
+	default:
+		if (!run_once) {
+			scratch_reg = allocate_kscratch();
+#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
+			build_r4000_setup_pgd();
+#endif
+			build_r4000_tlb_load_handler();
+			build_r4000_tlb_store_handler();
+			build_r4000_tlb_modify_handler();
+			run_once++;
+		}
+		build_r4000_tlb_refill_handler();
+	}
+}
+
+void __cpuinit flush_tlb_handlers(void)
+{
+	local_flush_icache_range((unsigned long)handle_tlbl,
+			   (unsigned long)handle_tlbl + sizeof(handle_tlbl));
+	local_flush_icache_range((unsigned long)handle_tlbs,
+			   (unsigned long)handle_tlbs + sizeof(handle_tlbs));
+	local_flush_icache_range((unsigned long)handle_tlbm,
+			   (unsigned long)handle_tlbm + sizeof(handle_tlbm));
+#ifdef CONFIG_MIPS_PGD_C0_CONTEXT
+	local_flush_icache_range((unsigned long)tlbmiss_handler_setup_pgd,
+			   (unsigned long)tlbmiss_handler_setup_pgd + sizeof(handle_tlbm));
+#endif
+}
diff --git a/arch/mips/mm/uasm.c b/arch/mips/mm/uasm.c
new file mode 100644
index 00000000..5fa18515
--- /dev/null
+++ b/arch/mips/mm/uasm.c
@@ -0,0 +1,699 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * A small micro-assembler. It is intentionally kept simple, does only
+ * support a subset of instructions, and does not try to hide pipeline
+ * effects like branch delay slots.
+ *
+ * Copyright (C) 2004, 2005, 2006, 2008  Thiemo Seufer
+ * Copyright (C) 2005, 2007  Maciej W. Rozycki
+ * Copyright (C) 2006  Ralf Baechle (ralf@linux-mips.org)
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/init.h>
+
+#include <asm/inst.h>
+#include <asm/elf.h>
+#include <asm/bugs.h>
+#include <asm/uasm.h>
+
+enum fields {
+	RS = 0x001,
+	RT = 0x002,
+	RD = 0x004,
+	RE = 0x008,
+	SIMM = 0x010,
+	UIMM = 0x020,
+	BIMM = 0x040,
+	JIMM = 0x080,
+	FUNC = 0x100,
+	SET = 0x200,
+	SCIMM = 0x400
+};
+
+#define OP_MASK		0x3f
+#define OP_SH		26
+#define RS_MASK		0x1f
+#define RS_SH		21
+#define RT_MASK		0x1f
+#define RT_SH		16
+#define RD_MASK		0x1f
+#define RD_SH		11
+#define RE_MASK		0x1f
+#define RE_SH		6
+#define IMM_MASK	0xffff
+#define IMM_SH		0
+#define JIMM_MASK	0x3ffffff
+#define JIMM_SH		0
+#define FUNC_MASK	0x3f
+#define FUNC_SH		0
+#define SET_MASK	0x7
+#define SET_SH		0
+#define SCIMM_MASK	0xfffff
+#define SCIMM_SH	6
+
+enum opcode {
+	insn_invalid,
+	insn_addu, insn_addiu, insn_and, insn_andi, insn_beq,
+	insn_beql, insn_bgez, insn_bgezl, insn_bltz, insn_bltzl,
+	insn_bne, insn_cache, insn_daddu, insn_daddiu, insn_dmfc0,
+	insn_dmtc0, insn_dsll, insn_dsll32, insn_dsra, insn_dsrl,
+	insn_dsrl32, insn_drotr, insn_drotr32, insn_dsubu, insn_eret,
+	insn_j, insn_jal, insn_jr, insn_ld, insn_ll, insn_lld,
+	insn_lui, insn_lw, insn_mfc0, insn_mtc0, insn_or, insn_ori,
+	insn_pref, insn_rfe, insn_sc, insn_scd, insn_sd, insn_sll,
+	insn_sra, insn_srl, insn_rotr, insn_subu, insn_sw, insn_tlbp,
+	insn_tlbr, insn_tlbwi, insn_tlbwr, insn_xor, insn_xori,
+	insn_dins, insn_dinsm, insn_syscall, insn_bbit0, insn_bbit1,
+	insn_lwx, insn_ldx
+};
+
+struct insn {
+	enum opcode opcode;
+	u32 match;
+	enum fields fields;
+};
+
+/* This macro sets the non-variable bits of an instruction. */
+#define M(a, b, c, d, e, f)					\
+	((a) << OP_SH						\
+	 | (b) << RS_SH						\
+	 | (c) << RT_SH						\
+	 | (d) << RD_SH						\
+	 | (e) << RE_SH						\
+	 | (f) << FUNC_SH)
+
+static struct insn insn_table[] __uasminitdata = {
+	{ insn_addiu, M(addiu_op, 0, 0, 0, 0, 0), RS | RT | SIMM },
+	{ insn_addu, M(spec_op, 0, 0, 0, 0, addu_op), RS | RT | RD },
+	{ insn_and, M(spec_op, 0, 0, 0, 0, and_op), RS | RT | RD },
+	{ insn_andi, M(andi_op, 0, 0, 0, 0, 0), RS | RT | UIMM },
+	{ insn_beq, M(beq_op, 0, 0, 0, 0, 0), RS | RT | BIMM },
+	{ insn_beql, M(beql_op, 0, 0, 0, 0, 0), RS | RT | BIMM },
+	{ insn_bgez, M(bcond_op, 0, bgez_op, 0, 0, 0), RS | BIMM },
+	{ insn_bgezl, M(bcond_op, 0, bgezl_op, 0, 0, 0), RS | BIMM },
+	{ insn_bltz, M(bcond_op, 0, bltz_op, 0, 0, 0), RS | BIMM },
+	{ insn_bltzl, M(bcond_op, 0, bltzl_op, 0, 0, 0), RS | BIMM },
+	{ insn_bne, M(bne_op, 0, 0, 0, 0, 0), RS | RT | BIMM },
+	{ insn_cache,  M(cache_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_daddiu, M(daddiu_op, 0, 0, 0, 0, 0), RS | RT | SIMM },
+	{ insn_daddu, M(spec_op, 0, 0, 0, 0, daddu_op), RS | RT | RD },
+	{ insn_dmfc0, M(cop0_op, dmfc_op, 0, 0, 0, 0), RT | RD | SET},
+	{ insn_dmtc0, M(cop0_op, dmtc_op, 0, 0, 0, 0), RT | RD | SET},
+	{ insn_dsll, M(spec_op, 0, 0, 0, 0, dsll_op), RT | RD | RE },
+	{ insn_dsll32, M(spec_op, 0, 0, 0, 0, dsll32_op), RT | RD | RE },
+	{ insn_dsra, M(spec_op, 0, 0, 0, 0, dsra_op), RT | RD | RE },
+	{ insn_dsrl, M(spec_op, 0, 0, 0, 0, dsrl_op), RT | RD | RE },
+	{ insn_dsrl32, M(spec_op, 0, 0, 0, 0, dsrl32_op), RT | RD | RE },
+	{ insn_drotr, M(spec_op, 1, 0, 0, 0, dsrl_op), RT | RD | RE },
+	{ insn_drotr32, M(spec_op, 1, 0, 0, 0, dsrl32_op), RT | RD | RE },
+	{ insn_dsubu, M(spec_op, 0, 0, 0, 0, dsubu_op), RS | RT | RD },
+	{ insn_eret,  M(cop0_op, cop_op, 0, 0, 0, eret_op),  0 },
+	{ insn_j,  M(j_op, 0, 0, 0, 0, 0),  JIMM },
+	{ insn_jal,  M(jal_op, 0, 0, 0, 0, 0),  JIMM },
+	{ insn_jr,  M(spec_op, 0, 0, 0, 0, jr_op),  RS },
+	{ insn_ld,  M(ld_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_ll,  M(ll_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_lld,  M(lld_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_lui,  M(lui_op, 0, 0, 0, 0, 0),  RT | SIMM },
+	{ insn_lw,  M(lw_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_mfc0,  M(cop0_op, mfc_op, 0, 0, 0, 0),  RT | RD | SET},
+	{ insn_mtc0,  M(cop0_op, mtc_op, 0, 0, 0, 0),  RT | RD | SET},
+	{ insn_or,  M(spec_op, 0, 0, 0, 0, or_op),  RS | RT | RD },
+	{ insn_ori,  M(ori_op, 0, 0, 0, 0, 0),  RS | RT | UIMM },
+	{ insn_pref,  M(pref_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_rfe,  M(cop0_op, cop_op, 0, 0, 0, rfe_op),  0 },
+	{ insn_sc,  M(sc_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_scd,  M(scd_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_sd,  M(sd_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_sll,  M(spec_op, 0, 0, 0, 0, sll_op),  RT | RD | RE },
+	{ insn_sra,  M(spec_op, 0, 0, 0, 0, sra_op),  RT | RD | RE },
+	{ insn_srl,  M(spec_op, 0, 0, 0, 0, srl_op),  RT | RD | RE },
+	{ insn_rotr,  M(spec_op, 1, 0, 0, 0, srl_op),  RT | RD | RE },
+	{ insn_subu,  M(spec_op, 0, 0, 0, 0, subu_op),  RS | RT | RD },
+	{ insn_sw,  M(sw_op, 0, 0, 0, 0, 0),  RS | RT | SIMM },
+	{ insn_tlbp,  M(cop0_op, cop_op, 0, 0, 0, tlbp_op),  0 },
+	{ insn_tlbr,  M(cop0_op, cop_op, 0, 0, 0, tlbr_op),  0 },
+	{ insn_tlbwi,  M(cop0_op, cop_op, 0, 0, 0, tlbwi_op),  0 },
+	{ insn_tlbwr,  M(cop0_op, cop_op, 0, 0, 0, tlbwr_op),  0 },
+	{ insn_xor,  M(spec_op, 0, 0, 0, 0, xor_op),  RS | RT | RD },
+	{ insn_xori,  M(xori_op, 0, 0, 0, 0, 0),  RS | RT | UIMM },
+	{ insn_dins, M(spec3_op, 0, 0, 0, 0, dins_op), RS | RT | RD | RE },
+	{ insn_dinsm, M(spec3_op, 0, 0, 0, 0, dinsm_op), RS | RT | RD | RE },
+	{ insn_syscall, M(spec_op, 0, 0, 0, 0, syscall_op), SCIMM},
+	{ insn_bbit0, M(lwc2_op, 0, 0, 0, 0, 0), RS | RT | BIMM },
+	{ insn_bbit1, M(swc2_op, 0, 0, 0, 0, 0), RS | RT | BIMM },
+	{ insn_lwx, M(spec3_op, 0, 0, 0, lwx_op, lx_op), RS | RT | RD },
+	{ insn_ldx, M(spec3_op, 0, 0, 0, ldx_op, lx_op), RS | RT | RD },
+	{ insn_invalid, 0, 0 }
+};
+
+#undef M
+
+static inline __uasminit u32 build_rs(u32 arg)
+{
+	WARN(arg & ~RS_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg & RS_MASK) << RS_SH;
+}
+
+static inline __uasminit u32 build_rt(u32 arg)
+{
+	WARN(arg & ~RT_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg & RT_MASK) << RT_SH;
+}
+
+static inline __uasminit u32 build_rd(u32 arg)
+{
+	WARN(arg & ~RD_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg & RD_MASK) << RD_SH;
+}
+
+static inline __uasminit u32 build_re(u32 arg)
+{
+	WARN(arg & ~RE_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg & RE_MASK) << RE_SH;
+}
+
+static inline __uasminit u32 build_simm(s32 arg)
+{
+	WARN(arg > 0x7fff || arg < -0x8000,
+	     KERN_WARNING "Micro-assembler field overflow\n");
+
+	return arg & 0xffff;
+}
+
+static inline __uasminit u32 build_uimm(u32 arg)
+{
+	WARN(arg & ~IMM_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return arg & IMM_MASK;
+}
+
+static inline __uasminit u32 build_bimm(s32 arg)
+{
+	WARN(arg > 0x1ffff || arg < -0x20000,
+	     KERN_WARNING "Micro-assembler field overflow\n");
+
+	WARN(arg & 0x3, KERN_WARNING "Invalid micro-assembler branch target\n");
+
+	return ((arg < 0) ? (1 << 15) : 0) | ((arg >> 2) & 0x7fff);
+}
+
+static inline __uasminit u32 build_jimm(u32 arg)
+{
+	WARN(arg & ~(JIMM_MASK << 2),
+	     KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg >> 2) & JIMM_MASK;
+}
+
+static inline __uasminit u32 build_scimm(u32 arg)
+{
+	WARN(arg & ~SCIMM_MASK,
+	     KERN_WARNING "Micro-assembler field overflow\n");
+
+	return (arg & SCIMM_MASK) << SCIMM_SH;
+}
+
+static inline __uasminit u32 build_func(u32 arg)
+{
+	WARN(arg & ~FUNC_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return arg & FUNC_MASK;
+}
+
+static inline __uasminit u32 build_set(u32 arg)
+{
+	WARN(arg & ~SET_MASK, KERN_WARNING "Micro-assembler field overflow\n");
+
+	return arg & SET_MASK;
+}
+
+/*
+ * The order of opcode arguments is implicitly left to right,
+ * starting with RS and ending with FUNC or IMM.
+ */
+static void __uasminit build_insn(u32 **buf, enum opcode opc, ...)
+{
+	struct insn *ip = NULL;
+	unsigned int i;
+	va_list ap;
+	u32 op;
+
+	for (i = 0; insn_table[i].opcode != insn_invalid; i++)
+		if (insn_table[i].opcode == opc) {
+			ip = &insn_table[i];
+			break;
+		}
+
+	if (!ip || (opc == insn_daddiu && r4k_daddiu_bug()))
+		panic("Unsupported Micro-assembler instruction %d", opc);
+
+	op = ip->match;
+	va_start(ap, opc);
+	if (ip->fields & RS)
+		op |= build_rs(va_arg(ap, u32));
+	if (ip->fields & RT)
+		op |= build_rt(va_arg(ap, u32));
+	if (ip->fields & RD)
+		op |= build_rd(va_arg(ap, u32));
+	if (ip->fields & RE)
+		op |= build_re(va_arg(ap, u32));
+	if (ip->fields & SIMM)
+		op |= build_simm(va_arg(ap, s32));
+	if (ip->fields & UIMM)
+		op |= build_uimm(va_arg(ap, u32));
+	if (ip->fields & BIMM)
+		op |= build_bimm(va_arg(ap, s32));
+	if (ip->fields & JIMM)
+		op |= build_jimm(va_arg(ap, u32));
+	if (ip->fields & FUNC)
+		op |= build_func(va_arg(ap, u32));
+	if (ip->fields & SET)
+		op |= build_set(va_arg(ap, u32));
+	if (ip->fields & SCIMM)
+		op |= build_scimm(va_arg(ap, u32));
+	va_end(ap);
+
+	**buf = op;
+	(*buf)++;
+}
+
+#define I_u1u2u3(op)					\
+Ip_u1u2u3(op)						\
+{							\
+	build_insn(buf, insn##op, a, b, c);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u2u1u3(op)					\
+Ip_u2u1u3(op)						\
+{							\
+	build_insn(buf, insn##op, b, a, c);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u3u1u2(op)					\
+Ip_u3u1u2(op)						\
+{							\
+	build_insn(buf, insn##op, b, c, a);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u1u2s3(op)					\
+Ip_u1u2s3(op)						\
+{							\
+	build_insn(buf, insn##op, a, b, c);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u2s3u1(op)					\
+Ip_u2s3u1(op)						\
+{							\
+	build_insn(buf, insn##op, c, a, b);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u2u1s3(op)					\
+Ip_u2u1s3(op)						\
+{							\
+	build_insn(buf, insn##op, b, a, c);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u2u1msbu3(op)					\
+Ip_u2u1msbu3(op)					\
+{							\
+	build_insn(buf, insn##op, b, a, c+d-1, c);	\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u2u1msb32u3(op)				\
+Ip_u2u1msbu3(op)					\
+{							\
+	build_insn(buf, insn##op, b, a, c+d-33, c);	\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u1u2(op)					\
+Ip_u1u2(op)						\
+{							\
+	build_insn(buf, insn##op, a, b);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u1s2(op)					\
+Ip_u1s2(op)						\
+{							\
+	build_insn(buf, insn##op, a, b);		\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_u1(op)					\
+Ip_u1(op)						\
+{							\
+	build_insn(buf, insn##op, a);			\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+#define I_0(op)						\
+Ip_0(op)						\
+{							\
+	build_insn(buf, insn##op);			\
+}							\
+UASM_EXPORT_SYMBOL(uasm_i##op);
+
+I_u2u1s3(_addiu)
+I_u3u1u2(_addu)
+I_u2u1u3(_andi)
+I_u3u1u2(_and)
+I_u1u2s3(_beq)
+I_u1u2s3(_beql)
+I_u1s2(_bgez)
+I_u1s2(_bgezl)
+I_u1s2(_bltz)
+I_u1s2(_bltzl)
+I_u1u2s3(_bne)
+I_u2s3u1(_cache)
+I_u1u2u3(_dmfc0)
+I_u1u2u3(_dmtc0)
+I_u2u1s3(_daddiu)
+I_u3u1u2(_daddu)
+I_u2u1u3(_dsll)
+I_u2u1u3(_dsll32)
+I_u2u1u3(_dsra)
+I_u2u1u3(_dsrl)
+I_u2u1u3(_dsrl32)
+I_u2u1u3(_drotr)
+I_u2u1u3(_drotr32)
+I_u3u1u2(_dsubu)
+I_0(_eret)
+I_u1(_j)
+I_u1(_jal)
+I_u1(_jr)
+I_u2s3u1(_ld)
+I_u2s3u1(_ll)
+I_u2s3u1(_lld)
+I_u1s2(_lui)
+I_u2s3u1(_lw)
+I_u1u2u3(_mfc0)
+I_u1u2u3(_mtc0)
+I_u2u1u3(_ori)
+I_u3u1u2(_or)
+I_0(_rfe)
+I_u2s3u1(_sc)
+I_u2s3u1(_scd)
+I_u2s3u1(_sd)
+I_u2u1u3(_sll)
+I_u2u1u3(_sra)
+I_u2u1u3(_srl)
+I_u2u1u3(_rotr)
+I_u3u1u2(_subu)
+I_u2s3u1(_sw)
+I_0(_tlbp)
+I_0(_tlbr)
+I_0(_tlbwi)
+I_0(_tlbwr)
+I_u3u1u2(_xor)
+I_u2u1u3(_xori)
+I_u2u1msbu3(_dins);
+I_u2u1msb32u3(_dinsm);
+I_u1(_syscall);
+I_u1u2s3(_bbit0);
+I_u1u2s3(_bbit1);
+I_u3u1u2(_lwx)
+I_u3u1u2(_ldx)
+
+#ifdef CONFIG_CPU_CAVIUM_OCTEON
+#include <asm/octeon/octeon.h>
+void __uasminit uasm_i_pref(u32 **buf, unsigned int a, signed int b,
+			    unsigned int c)
+{
+	if (OCTEON_IS_MODEL(OCTEON_CN63XX_PASS1_X) && a <= 24 && a != 5)
+		/*
+		 * As per erratum Core-14449, replace prefetches 0-4,
+		 * 6-24 with 'pref 28'.
+		 */
+		build_insn(buf, insn_pref, c, 28, b);
+	else
+		build_insn(buf, insn_pref, c, a, b);
+}
+UASM_EXPORT_SYMBOL(uasm_i_pref);
+#else
+I_u2s3u1(_pref)
+#endif
+
+/* Handle labels. */
+void __uasminit uasm_build_label(struct uasm_label **lab, u32 *addr, int lid)
+{
+	(*lab)->addr = addr;
+	(*lab)->lab = lid;
+	(*lab)++;
+}
+UASM_EXPORT_SYMBOL(uasm_build_label);
+
+int __uasminit uasm_in_compat_space_p(long addr)
+{
+	/* Is this address in 32bit compat space? */
+#ifdef CONFIG_64BIT
+	return (((addr) & 0xffffffff00000000L) == 0xffffffff00000000L);
+#else
+	return 1;
+#endif
+}
+UASM_EXPORT_SYMBOL(uasm_in_compat_space_p);
+
+static int __uasminit uasm_rel_highest(long val)
+{
+#ifdef CONFIG_64BIT
+	return ((((val + 0x800080008000L) >> 48) & 0xffff) ^ 0x8000) - 0x8000;
+#else
+	return 0;
+#endif
+}
+
+static int __uasminit uasm_rel_higher(long val)
+{
+#ifdef CONFIG_64BIT
+	return ((((val + 0x80008000L) >> 32) & 0xffff) ^ 0x8000) - 0x8000;
+#else
+	return 0;
+#endif
+}
+
+int __uasminit uasm_rel_hi(long val)
+{
+	return ((((val + 0x8000L) >> 16) & 0xffff) ^ 0x8000) - 0x8000;
+}
+UASM_EXPORT_SYMBOL(uasm_rel_hi);
+
+int __uasminit uasm_rel_lo(long val)
+{
+	return ((val & 0xffff) ^ 0x8000) - 0x8000;
+}
+UASM_EXPORT_SYMBOL(uasm_rel_lo);
+
+void __uasminit UASM_i_LA_mostly(u32 **buf, unsigned int rs, long addr)
+{
+	if (!uasm_in_compat_space_p(addr)) {
+		uasm_i_lui(buf, rs, uasm_rel_highest(addr));
+		if (uasm_rel_higher(addr))
+			uasm_i_daddiu(buf, rs, rs, uasm_rel_higher(addr));
+		if (uasm_rel_hi(addr)) {
+			uasm_i_dsll(buf, rs, rs, 16);
+			uasm_i_daddiu(buf, rs, rs, uasm_rel_hi(addr));
+			uasm_i_dsll(buf, rs, rs, 16);
+		} else
+			uasm_i_dsll32(buf, rs, rs, 0);
+	} else
+		uasm_i_lui(buf, rs, uasm_rel_hi(addr));
+}
+UASM_EXPORT_SYMBOL(UASM_i_LA_mostly);
+
+void __uasminit UASM_i_LA(u32 **buf, unsigned int rs, long addr)
+{
+	UASM_i_LA_mostly(buf, rs, addr);
+	if (uasm_rel_lo(addr)) {
+		if (!uasm_in_compat_space_p(addr))
+			uasm_i_daddiu(buf, rs, rs, uasm_rel_lo(addr));
+		else
+			uasm_i_addiu(buf, rs, rs, uasm_rel_lo(addr));
+	}
+}
+UASM_EXPORT_SYMBOL(UASM_i_LA);
+
+/* Handle relocations. */
+void __uasminit
+uasm_r_mips_pc16(struct uasm_reloc **rel, u32 *addr, int lid)
+{
+	(*rel)->addr = addr;
+	(*rel)->type = R_MIPS_PC16;
+	(*rel)->lab = lid;
+	(*rel)++;
+}
+UASM_EXPORT_SYMBOL(uasm_r_mips_pc16);
+
+static inline void __uasminit
+__resolve_relocs(struct uasm_reloc *rel, struct uasm_label *lab)
+{
+	long laddr = (long)lab->addr;
+	long raddr = (long)rel->addr;
+
+	switch (rel->type) {
+	case R_MIPS_PC16:
+		*rel->addr |= build_bimm(laddr - (raddr + 4));
+		break;
+
+	default:
+		panic("Unsupported Micro-assembler relocation %d",
+		      rel->type);
+	}
+}
+
+void __uasminit
+uasm_resolve_relocs(struct uasm_reloc *rel, struct uasm_label *lab)
+{
+	struct uasm_label *l;
+
+	for (; rel->lab != UASM_LABEL_INVALID; rel++)
+		for (l = lab; l->lab != UASM_LABEL_INVALID; l++)
+			if (rel->lab == l->lab)
+				__resolve_relocs(rel, l);
+}
+UASM_EXPORT_SYMBOL(uasm_resolve_relocs);
+
+void __uasminit
+uasm_move_relocs(struct uasm_reloc *rel, u32 *first, u32 *end, long off)
+{
+	for (; rel->lab != UASM_LABEL_INVALID; rel++)
+		if (rel->addr >= first && rel->addr < end)
+			rel->addr += off;
+}
+UASM_EXPORT_SYMBOL(uasm_move_relocs);
+
+void __uasminit
+uasm_move_labels(struct uasm_label *lab, u32 *first, u32 *end, long off)
+{
+	for (; lab->lab != UASM_LABEL_INVALID; lab++)
+		if (lab->addr >= first && lab->addr < end)
+			lab->addr += off;
+}
+UASM_EXPORT_SYMBOL(uasm_move_labels);
+
+void __uasminit
+uasm_copy_handler(struct uasm_reloc *rel, struct uasm_label *lab, u32 *first,
+		  u32 *end, u32 *target)
+{
+	long off = (long)(target - first);
+
+	memcpy(target, first, (end - first) * sizeof(u32));
+
+	uasm_move_relocs(rel, first, end, off);
+	uasm_move_labels(lab, first, end, off);
+}
+UASM_EXPORT_SYMBOL(uasm_copy_handler);
+
+int __uasminit uasm_insn_has_bdelay(struct uasm_reloc *rel, u32 *addr)
+{
+	for (; rel->lab != UASM_LABEL_INVALID; rel++) {
+		if (rel->addr == addr
+		    && (rel->type == R_MIPS_PC16
+			|| rel->type == R_MIPS_26))
+			return 1;
+	}
+
+	return 0;
+}
+UASM_EXPORT_SYMBOL(uasm_insn_has_bdelay);
+
+/* Convenience functions for labeled branches. */
+void __uasminit
+uasm_il_bltz(u32 **p, struct uasm_reloc **r, unsigned int reg, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bltz(p, reg, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bltz);
+
+void __uasminit
+uasm_il_b(u32 **p, struct uasm_reloc **r, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_b(p, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_b);
+
+void __uasminit
+uasm_il_beqz(u32 **p, struct uasm_reloc **r, unsigned int reg, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_beqz(p, reg, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_beqz);
+
+void __uasminit
+uasm_il_beqzl(u32 **p, struct uasm_reloc **r, unsigned int reg, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_beqzl(p, reg, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_beqzl);
+
+void __uasminit
+uasm_il_bne(u32 **p, struct uasm_reloc **r, unsigned int reg1,
+	unsigned int reg2, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bne(p, reg1, reg2, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bne);
+
+void __uasminit
+uasm_il_bnez(u32 **p, struct uasm_reloc **r, unsigned int reg, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bnez(p, reg, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bnez);
+
+void __uasminit
+uasm_il_bgezl(u32 **p, struct uasm_reloc **r, unsigned int reg, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bgezl(p, reg, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bgezl);
+
+void __uasminit
+uasm_il_bgez(u32 **p, struct uasm_reloc **r, unsigned int reg, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bgez(p, reg, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bgez);
+
+void __uasminit
+uasm_il_bbit0(u32 **p, struct uasm_reloc **r, unsigned int reg,
+	      unsigned int bit, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bbit0(p, reg, bit, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bbit0);
+
+void __uasminit
+uasm_il_bbit1(u32 **p, struct uasm_reloc **r, unsigned int reg,
+	      unsigned int bit, int lid)
+{
+	uasm_r_mips_pc16(r, *p, lid);
+	uasm_i_bbit1(p, reg, bit, 0);
+}
+UASM_EXPORT_SYMBOL(uasm_il_bbit1);