File-copy from v4.4.100

This is the result of 'cp' from a linux-stable tree with the 'v4.4.100'
tag checked out (commit 26d6298789e695c9f627ce49a7bbd2286405798a) on
git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git

Please refer to that tree for all history prior to this point.

Change-Id: I8a9ee2aea93cd29c52c847d0ce33091a73ae6afe
diff --git a/arch/sh/mm/Kconfig b/arch/sh/mm/Kconfig
new file mode 100644
index 0000000..dba285e
--- /dev/null
+++ b/arch/sh/mm/Kconfig
@@ -0,0 +1,272 @@
+menu "Memory management options"
+
+config QUICKLIST
+	def_bool y
+
+config MMU
+        bool "Support for memory management hardware"
+	depends on !CPU_SH2
+	default y
+	help
+	  Some SH processors (such as SH-2/SH-2A) lack an MMU. In order to
+	  boot on these systems, this option must not be set.
+
+	  On other systems (such as the SH-3 and 4) where an MMU exists,
+	  turning this off will boot the kernel on these machines with the
+	  MMU implicitly switched off.
+
+config PAGE_OFFSET
+	hex
+	default "0x80000000" if MMU && SUPERH32
+	default "0x20000000" if MMU && SUPERH64
+	default "0x00000000"
+
+config FORCE_MAX_ZONEORDER
+	int "Maximum zone order"
+	range 9 64 if PAGE_SIZE_16KB
+	default "9" if PAGE_SIZE_16KB
+	range 7 64 if PAGE_SIZE_64KB
+	default "7" if PAGE_SIZE_64KB
+	range 11 64
+	default "14" if !MMU
+	default "11"
+	help
+	  The kernel memory allocator divides physically contiguous memory
+	  blocks into "zones", where each zone is a power of two number of
+	  pages.  This option selects the largest power of two that the kernel
+	  keeps in the memory allocator.  If you need to allocate very large
+	  blocks of physically contiguous memory, then you may need to
+	  increase this value.
+
+	  This config option is actually maximum order plus one. For example,
+	  a value of 11 means that the largest free memory block is 2^10 pages.
+
+	  The page size is not necessarily 4KB. Keep this in mind when
+	  choosing a value for this option.
+
+config MEMORY_START
+	hex "Physical memory start address"
+	default "0x08000000"
+	---help---
+	  Computers built with Hitachi SuperH processors always
+	  map the ROM starting at address zero.  But the processor
+	  does not specify the range that RAM takes.
+
+	  The physical memory (RAM) start address will be automatically
+	  set to 08000000. Other platforms, such as the Solution Engine
+	  boards typically map RAM at 0C000000.
+
+	  Tweak this only when porting to a new machine which does not
+	  already have a defconfig. Changing it from the known correct
+	  value on any of the known systems will only lead to disaster.
+
+config MEMORY_SIZE
+	hex "Physical memory size"
+	default "0x04000000"
+	help
+	  This sets the default memory size assumed by your SH kernel. It can
+	  be overridden as normal by the 'mem=' argument on the kernel command
+	  line. If unsure, consult your board specifications or just leave it
+	  as 0x04000000 which was the default value before this became
+	  configurable.
+
+# Physical addressing modes
+
+config 29BIT
+	def_bool !32BIT
+	depends on SUPERH32
+	select UNCACHED_MAPPING
+
+config 32BIT
+	bool
+	default y if CPU_SH5 || !MMU
+
+config PMB
+	bool "Support 32-bit physical addressing through PMB"
+	depends on MMU && CPU_SH4A && !CPU_SH4AL_DSP
+	select 32BIT
+	select UNCACHED_MAPPING
+	help
+	  If you say Y here, physical addressing will be extended to
+	  32-bits through the SH-4A PMB. If this is not set, legacy
+	  29-bit physical addressing will be used.
+
+config X2TLB
+	def_bool y
+	depends on (CPU_SHX2 || CPU_SHX3) && MMU
+
+config VSYSCALL
+	bool "Support vsyscall page"
+	depends on MMU && (CPU_SH3 || CPU_SH4)
+	default y
+	help
+	  This will enable support for the kernel mapping a vDSO page
+	  in process space, and subsequently handing down the entry point
+	  to the libc through the ELF auxiliary vector.
+
+	  From the kernel side this is used for the signal trampoline.
+	  For systems with an MMU that can afford to give up a page,
+	  (the default value) say Y.
+
+config NUMA
+	bool "Non Uniform Memory Access (NUMA) Support"
+	depends on MMU && SYS_SUPPORTS_NUMA
+	select ARCH_WANT_NUMA_VARIABLE_LOCALITY
+	default n
+	help
+	  Some SH systems have many various memories scattered around
+	  the address space, each with varying latencies. This enables
+	  support for these blocks by binding them to nodes and allowing
+	  memory policies to be used for prioritizing and controlling
+	  allocation behaviour.
+
+config NODES_SHIFT
+	int
+	default "3" if CPU_SUBTYPE_SHX3
+	default "1"
+	depends on NEED_MULTIPLE_NODES
+
+config ARCH_FLATMEM_ENABLE
+	def_bool y
+	depends on !NUMA
+
+config ARCH_SPARSEMEM_ENABLE
+	def_bool y
+	select SPARSEMEM_STATIC
+
+config ARCH_SPARSEMEM_DEFAULT
+	def_bool y
+
+config ARCH_SELECT_MEMORY_MODEL
+	def_bool y
+
+config ARCH_ENABLE_MEMORY_HOTPLUG
+	def_bool y
+	depends on SPARSEMEM && MMU
+
+config ARCH_ENABLE_MEMORY_HOTREMOVE
+	def_bool y
+	depends on SPARSEMEM && MMU
+
+config ARCH_MEMORY_PROBE
+	def_bool y
+	depends on MEMORY_HOTPLUG
+
+config IOREMAP_FIXED
+       def_bool y
+       depends on X2TLB || SUPERH64
+
+config UNCACHED_MAPPING
+	bool
+
+config HAVE_SRAM_POOL
+	bool
+	select GENERIC_ALLOCATOR
+
+choice
+	prompt "Kernel page size"
+	default PAGE_SIZE_4KB
+
+config PAGE_SIZE_4KB
+	bool "4kB"
+	help
+	  This is the default page size used by all SuperH CPUs.
+
+config PAGE_SIZE_8KB
+	bool "8kB"
+	depends on !MMU || X2TLB
+	help
+	  This enables 8kB pages as supported by SH-X2 and later MMUs.
+
+config PAGE_SIZE_16KB
+	bool "16kB"
+	depends on !MMU
+	help
+	  This enables 16kB pages on MMU-less SH systems.
+
+config PAGE_SIZE_64KB
+	bool "64kB"
+	depends on !MMU || CPU_SH4 || CPU_SH5
+	help
+	  This enables support for 64kB pages, possible on all SH-4
+	  CPUs and later.
+
+endchoice
+
+choice
+	prompt "HugeTLB page size"
+	depends on HUGETLB_PAGE
+	default HUGETLB_PAGE_SIZE_1MB if PAGE_SIZE_64KB
+	default HUGETLB_PAGE_SIZE_64K
+
+config HUGETLB_PAGE_SIZE_64K
+	bool "64kB"
+	depends on !PAGE_SIZE_64KB
+
+config HUGETLB_PAGE_SIZE_256K
+	bool "256kB"
+	depends on X2TLB
+
+config HUGETLB_PAGE_SIZE_1MB
+	bool "1MB"
+
+config HUGETLB_PAGE_SIZE_4MB
+	bool "4MB"
+	depends on X2TLB
+
+config HUGETLB_PAGE_SIZE_64MB
+	bool "64MB"
+	depends on X2TLB
+
+config HUGETLB_PAGE_SIZE_512MB
+	bool "512MB"
+	depends on CPU_SH5
+
+endchoice
+
+source "mm/Kconfig"
+
+config SCHED_MC
+	bool "Multi-core scheduler support"
+	depends on SMP
+	default y
+	help
+	  Multi-core scheduler support improves the CPU scheduler's decision
+	  making when dealing with multi-core CPU chips at a cost of slightly
+	  increased overhead in some places. If unsure say N here.
+
+endmenu
+
+menu "Cache configuration"
+
+config SH7705_CACHE_32KB
+	bool "Enable 32KB cache size for SH7705"
+	depends on CPU_SUBTYPE_SH7705
+	default y
+
+choice
+	prompt "Cache mode"
+	default CACHE_WRITEBACK if CPU_SH2A || CPU_SH3 || CPU_SH4 || CPU_SH5
+	default CACHE_WRITETHROUGH if (CPU_SH2 && !CPU_SH2A)
+
+config CACHE_WRITEBACK
+	bool "Write-back"
+
+config CACHE_WRITETHROUGH
+	bool "Write-through"
+	help
+	  Selecting this option will configure the caches in write-through
+	  mode, as opposed to the default write-back configuration.
+
+	  Since there's sill some aliasing issues on SH-4, this option will
+	  unfortunately still require the majority of flushing functions to
+	  be implemented to deal with aliasing.
+
+	  If unsure, say N.
+
+config CACHE_OFF
+	bool "Off"
+
+endchoice
+
+endmenu
diff --git a/arch/sh/mm/Makefile b/arch/sh/mm/Makefile
new file mode 100644
index 0000000..cee6b99
--- /dev/null
+++ b/arch/sh/mm/Makefile
@@ -0,0 +1,72 @@
+#
+# Makefile for the Linux SuperH-specific parts of the memory manager.
+#
+
+obj-y			:= alignment.o cache.o init.o consistent.o mmap.o
+
+cacheops-$(CONFIG_CPU_SH2)		:= cache-sh2.o
+cacheops-$(CONFIG_CPU_SH2A)		:= cache-sh2a.o
+cacheops-$(CONFIG_CPU_SH3)		:= cache-sh3.o
+cacheops-$(CONFIG_CPU_SH4)		:= cache-sh4.o flush-sh4.o
+cacheops-$(CONFIG_CPU_SH5)		:= cache-sh5.o flush-sh4.o
+cacheops-$(CONFIG_SH7705_CACHE_32KB)	+= cache-sh7705.o
+cacheops-$(CONFIG_CPU_SHX3)		+= cache-shx3.o
+
+obj-y			+= $(cacheops-y)
+
+mmu-y			:= nommu.o extable_32.o
+mmu-$(CONFIG_MMU)	:= extable_$(BITS).o fault.o gup.o ioremap.o kmap.o \
+			   pgtable.o tlbex_$(BITS).o tlbflush_$(BITS).o
+
+obj-y			+= $(mmu-y)
+
+debugfs-y			:= asids-debugfs.o
+ifndef CONFIG_CACHE_OFF
+debugfs-$(CONFIG_CPU_SH4)	+= cache-debugfs.o
+endif
+
+ifdef CONFIG_MMU
+debugfs-$(CONFIG_CPU_SH4)	+= tlb-debugfs.o
+tlb-$(CONFIG_CPU_SH3)		:= tlb-sh3.o
+tlb-$(CONFIG_CPU_SH4)		:= tlb-sh4.o tlb-urb.o
+tlb-$(CONFIG_CPU_SH5)		:= tlb-sh5.o
+tlb-$(CONFIG_CPU_HAS_PTEAEX)	:= tlb-pteaex.o tlb-urb.o
+obj-y				+= $(tlb-y)
+endif
+
+obj-$(CONFIG_DEBUG_FS)		+= $(debugfs-y)
+obj-$(CONFIG_HUGETLB_PAGE)	+= hugetlbpage.o
+obj-$(CONFIG_PMB)		+= pmb.o
+obj-$(CONFIG_NUMA)		+= numa.o
+obj-$(CONFIG_IOREMAP_FIXED)	+= ioremap_fixed.o
+obj-$(CONFIG_UNCACHED_MAPPING)	+= uncached.o
+obj-$(CONFIG_HAVE_SRAM_POOL)	+= sram.o
+
+GCOV_PROFILE_pmb.o := n
+
+# Special flags for tlbex_64.o.  This puts restrictions on the number of
+# caller-save registers that the compiler can target when building this file.
+# This is required because the code is called from a context in entry.S where
+# very few registers have been saved in the exception handler (for speed
+# reasons).
+# The caller save registers that have been saved and which can be used are
+# r2,r3,r4,r5 : argument passing
+# r15, r18 : SP and LINK
+# tr0-4 : allow all caller-save TR's.  The compiler seems to be able to make
+#         use of them, so it's probably beneficial to performance to save them
+#         and have them available for it.
+#
+# The resources not listed below are callee save, i.e. the compiler is free to
+# use any of them and will spill them to the stack itself.
+
+CFLAGS_tlbex_64.o += -ffixed-r7 \
+	-ffixed-r8 -ffixed-r9 -ffixed-r10 -ffixed-r11 -ffixed-r12 \
+	-ffixed-r13 -ffixed-r14 -ffixed-r16 -ffixed-r17 -ffixed-r19 \
+	-ffixed-r20 -ffixed-r21 -ffixed-r22 -ffixed-r23 \
+	-ffixed-r24 -ffixed-r25 -ffixed-r26 -ffixed-r27 \
+	-ffixed-r36 -ffixed-r37 -ffixed-r38 -ffixed-r39 -ffixed-r40 \
+	-ffixed-r41 -ffixed-r42 -ffixed-r43  \
+	-ffixed-r60 -ffixed-r61 -ffixed-r62 \
+	-fomit-frame-pointer
+
+ccflags-y := -Werror
diff --git a/arch/sh/mm/alignment.c b/arch/sh/mm/alignment.c
new file mode 100644
index 0000000..ec2b253
--- /dev/null
+++ b/arch/sh/mm/alignment.c
@@ -0,0 +1,190 @@
+/*
+ * Alignment access counters and corresponding user-space interfaces.
+ *
+ * Copyright (C) 2009 ST Microelectronics
+ * Copyright (C) 2009 - 2010 Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/seq_file.h>
+#include <linux/proc_fs.h>
+#include <linux/uaccess.h>
+#include <linux/ratelimit.h>
+#include <asm/alignment.h>
+#include <asm/processor.h>
+
+static unsigned long se_user;
+static unsigned long se_sys;
+static unsigned long se_half;
+static unsigned long se_word;
+static unsigned long se_dword;
+static unsigned long se_multi;
+/* bitfield: 1: warn 2: fixup 4: signal -> combinations 2|4 && 1|2|4 are not
+   valid! */
+static int se_usermode = UM_WARN | UM_FIXUP;
+/* 0: no warning 1: print a warning message, disabled by default */
+static int se_kernmode_warn;
+
+core_param(alignment, se_usermode, int, 0600);
+
+void inc_unaligned_byte_access(void)
+{
+	se_half++;
+}
+
+void inc_unaligned_word_access(void)
+{
+	se_word++;
+}
+
+void inc_unaligned_dword_access(void)
+{
+	se_dword++;
+}
+
+void inc_unaligned_multi_access(void)
+{
+	se_multi++;
+}
+
+void inc_unaligned_user_access(void)
+{
+	se_user++;
+}
+
+void inc_unaligned_kernel_access(void)
+{
+	se_sys++;
+}
+
+/*
+ * This defaults to the global policy which can be set from the command
+ * line, while processes can overload their preferences via prctl().
+ */
+unsigned int unaligned_user_action(void)
+{
+	unsigned int action = se_usermode;
+
+	if (current->thread.flags & SH_THREAD_UAC_SIGBUS) {
+		action &= ~UM_FIXUP;
+		action |= UM_SIGNAL;
+	}
+
+	if (current->thread.flags & SH_THREAD_UAC_NOPRINT)
+		action &= ~UM_WARN;
+
+	return action;
+}
+
+int get_unalign_ctl(struct task_struct *tsk, unsigned long addr)
+{
+	return put_user(tsk->thread.flags & SH_THREAD_UAC_MASK,
+			(unsigned int __user *)addr);
+}
+
+int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
+{
+	tsk->thread.flags = (tsk->thread.flags & ~SH_THREAD_UAC_MASK) |
+			    (val & SH_THREAD_UAC_MASK);
+	return 0;
+}
+
+void unaligned_fixups_notify(struct task_struct *tsk, insn_size_t insn,
+			     struct pt_regs *regs)
+{
+	if (user_mode(regs) && (se_usermode & UM_WARN))
+		pr_notice_ratelimited("Fixing up unaligned userspace access "
+			  "in \"%s\" pid=%d pc=0x%p ins=0x%04hx\n",
+			  tsk->comm, task_pid_nr(tsk),
+			  (void *)instruction_pointer(regs), insn);
+	else if (se_kernmode_warn)
+		pr_notice_ratelimited("Fixing up unaligned kernel access "
+			  "in \"%s\" pid=%d pc=0x%p ins=0x%04hx\n",
+			  tsk->comm, task_pid_nr(tsk),
+			  (void *)instruction_pointer(regs), insn);
+}
+
+static const char *se_usermode_action[] = {
+	"ignored",
+	"warn",
+	"fixup",
+	"fixup+warn",
+	"signal",
+	"signal+warn"
+};
+
+static int alignment_proc_show(struct seq_file *m, void *v)
+{
+	seq_printf(m, "User:\t\t%lu\n", se_user);
+	seq_printf(m, "System:\t\t%lu\n", se_sys);
+	seq_printf(m, "Half:\t\t%lu\n", se_half);
+	seq_printf(m, "Word:\t\t%lu\n", se_word);
+	seq_printf(m, "DWord:\t\t%lu\n", se_dword);
+	seq_printf(m, "Multi:\t\t%lu\n", se_multi);
+	seq_printf(m, "User faults:\t%i (%s)\n", se_usermode,
+			se_usermode_action[se_usermode]);
+	seq_printf(m, "Kernel faults:\t%i (fixup%s)\n", se_kernmode_warn,
+			se_kernmode_warn ? "+warn" : "");
+	return 0;
+}
+
+static int alignment_proc_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, alignment_proc_show, NULL);
+}
+
+static ssize_t alignment_proc_write(struct file *file,
+		const char __user *buffer, size_t count, loff_t *pos)
+{
+	int *data = PDE_DATA(file_inode(file));
+	char mode;
+
+	if (count > 0) {
+		if (get_user(mode, buffer))
+			return -EFAULT;
+		if (mode >= '0' && mode <= '5')
+			*data = mode - '0';
+	}
+	return count;
+}
+
+static const struct file_operations alignment_proc_fops = {
+	.owner		= THIS_MODULE,
+	.open		= alignment_proc_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+	.write		= alignment_proc_write,
+};
+
+/*
+ * This needs to be done after sysctl_init, otherwise sys/ will be
+ * overwritten.  Actually, this shouldn't be in sys/ at all since
+ * it isn't a sysctl, and it doesn't contain sysctl information.
+ * We now locate it in /proc/cpu/alignment instead.
+ */
+static int __init alignment_init(void)
+{
+	struct proc_dir_entry *dir, *res;
+
+	dir = proc_mkdir("cpu", NULL);
+	if (!dir)
+		return -ENOMEM;
+
+	res = proc_create_data("alignment", S_IWUSR | S_IRUGO, dir,
+			       &alignment_proc_fops, &se_usermode);
+	if (!res)
+		return -ENOMEM;
+
+        res = proc_create_data("kernel_alignment", S_IWUSR | S_IRUGO, dir,
+			       &alignment_proc_fops, &se_kernmode_warn);
+        if (!res)
+                return -ENOMEM;
+
+	return 0;
+}
+fs_initcall(alignment_init);
diff --git a/arch/sh/mm/asids-debugfs.c b/arch/sh/mm/asids-debugfs.c
new file mode 100644
index 0000000..ecfc6b0
--- /dev/null
+++ b/arch/sh/mm/asids-debugfs.c
@@ -0,0 +1,75 @@
+/*
+ * debugfs ops for process ASIDs
+ *
+ *  Copyright (C) 2000, 2001  Paolo Alberelli
+ *  Copyright (C) 2003 - 2008  Paul Mundt
+ *  Copyright (C) 2003, 2004  Richard Curnow
+ *
+ * Provides a debugfs file that lists out the ASIDs currently associated
+ * with the processes.
+ *
+ * In the SH-5 case, if the DM.PC register is examined through the debug
+ * link, this shows ASID + PC. To make use of this, the PID->ASID
+ * relationship needs to be known. This is primarily for debugging.
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <linux/spinlock.h>
+#include <asm/processor.h>
+#include <asm/mmu_context.h>
+
+static int asids_seq_show(struct seq_file *file, void *iter)
+{
+	struct task_struct *p;
+
+	read_lock(&tasklist_lock);
+
+	for_each_process(p) {
+		int pid = p->pid;
+
+		if (unlikely(!pid))
+			continue;
+
+		if (p->mm)
+			seq_printf(file, "%5d : %04lx\n", pid,
+				   cpu_asid(smp_processor_id(), p->mm));
+	}
+
+	read_unlock(&tasklist_lock);
+
+	return 0;
+}
+
+static int asids_debugfs_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, asids_seq_show, inode->i_private);
+}
+
+static const struct file_operations asids_debugfs_fops = {
+	.owner		= THIS_MODULE,
+	.open		= asids_debugfs_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int __init asids_debugfs_init(void)
+{
+	struct dentry *asids_dentry;
+
+	asids_dentry = debugfs_create_file("asids", S_IRUSR, arch_debugfs_dir,
+					   NULL, &asids_debugfs_fops);
+	if (!asids_dentry)
+		return -ENOMEM;
+
+	return PTR_ERR_OR_ZERO(asids_dentry);
+}
+module_init(asids_debugfs_init);
+
+MODULE_LICENSE("GPL v2");
diff --git a/arch/sh/mm/cache-debugfs.c b/arch/sh/mm/cache-debugfs.c
new file mode 100644
index 0000000..777e50f
--- /dev/null
+++ b/arch/sh/mm/cache-debugfs.c
@@ -0,0 +1,132 @@
+/*
+ * debugfs ops for the L1 cache
+ *
+ *  Copyright (C) 2006  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <asm/processor.h>
+#include <asm/uaccess.h>
+#include <asm/cache.h>
+#include <asm/io.h>
+
+enum cache_type {
+	CACHE_TYPE_ICACHE,
+	CACHE_TYPE_DCACHE,
+	CACHE_TYPE_UNIFIED,
+};
+
+static int cache_seq_show(struct seq_file *file, void *iter)
+{
+	unsigned int cache_type = (unsigned int)file->private;
+	struct cache_info *cache;
+	unsigned int waysize, way;
+	unsigned long ccr;
+	unsigned long addrstart = 0;
+
+	/*
+	 * Go uncached immediately so we don't skew the results any
+	 * more than we already are..
+	 */
+	jump_to_uncached();
+
+	ccr = __raw_readl(SH_CCR);
+	if ((ccr & CCR_CACHE_ENABLE) == 0) {
+		back_to_cached();
+
+		seq_printf(file, "disabled\n");
+		return 0;
+	}
+
+	if (cache_type == CACHE_TYPE_DCACHE) {
+		addrstart = CACHE_OC_ADDRESS_ARRAY;
+		cache = &current_cpu_data.dcache;
+	} else {
+		addrstart = CACHE_IC_ADDRESS_ARRAY;
+		cache = &current_cpu_data.icache;
+	}
+
+	waysize = cache->sets;
+
+	/*
+	 * If the OC is already in RAM mode, we only have
+	 * half of the entries to consider..
+	 */
+	if ((ccr & CCR_CACHE_ORA) && cache_type == CACHE_TYPE_DCACHE)
+		waysize >>= 1;
+
+	waysize <<= cache->entry_shift;
+
+	for (way = 0; way < cache->ways; way++) {
+		unsigned long addr;
+		unsigned int line;
+
+		seq_printf(file, "-----------------------------------------\n");
+		seq_printf(file, "Way %d\n", way);
+		seq_printf(file, "-----------------------------------------\n");
+
+		for (addr = addrstart, line = 0;
+		     addr < addrstart + waysize;
+		     addr += cache->linesz, line++) {
+			unsigned long data = __raw_readl(addr);
+
+			/* Check the V bit, ignore invalid cachelines */
+			if ((data & 1) == 0)
+				continue;
+
+			/* U: Dirty, cache tag is 10 bits up */
+			seq_printf(file, "%3d: %c 0x%lx\n",
+				   line, data & 2 ? 'U' : ' ',
+				   data & 0x1ffffc00);
+		}
+
+		addrstart += cache->way_incr;
+	}
+
+	back_to_cached();
+
+	return 0;
+}
+
+static int cache_debugfs_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, cache_seq_show, inode->i_private);
+}
+
+static const struct file_operations cache_debugfs_fops = {
+	.owner		= THIS_MODULE,
+	.open		= cache_debugfs_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int __init cache_debugfs_init(void)
+{
+	struct dentry *dcache_dentry, *icache_dentry;
+
+	dcache_dentry = debugfs_create_file("dcache", S_IRUSR, arch_debugfs_dir,
+					    (unsigned int *)CACHE_TYPE_DCACHE,
+					    &cache_debugfs_fops);
+	if (!dcache_dentry)
+		return -ENOMEM;
+
+	icache_dentry = debugfs_create_file("icache", S_IRUSR, arch_debugfs_dir,
+					    (unsigned int *)CACHE_TYPE_ICACHE,
+					    &cache_debugfs_fops);
+	if (!icache_dentry) {
+		debugfs_remove(dcache_dentry);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+module_init(cache_debugfs_init);
+
+MODULE_LICENSE("GPL v2");
diff --git a/arch/sh/mm/cache-sh2.c b/arch/sh/mm/cache-sh2.c
new file mode 100644
index 0000000..a74259f
--- /dev/null
+++ b/arch/sh/mm/cache-sh2.c
@@ -0,0 +1,91 @@
+/*
+ * arch/sh/mm/cache-sh2.c
+ *
+ * Copyright (C) 2002 Paul Mundt
+ * Copyright (C) 2008 Yoshinori Sato
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+
+#include <linux/init.h>
+#include <linux/mm.h>
+
+#include <asm/cache.h>
+#include <asm/addrspace.h>
+#include <asm/processor.h>
+#include <asm/cacheflush.h>
+#include <asm/io.h>
+
+static void sh2__flush_wback_region(void *start, int size)
+{
+	unsigned long v;
+	unsigned long begin, end;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+	for (v = begin; v < end; v+=L1_CACHE_BYTES) {
+		unsigned long addr = CACHE_OC_ADDRESS_ARRAY | (v & 0x00000ff0);
+		int way;
+		for (way = 0; way < 4; way++) {
+			unsigned long data =  __raw_readl(addr | (way << 12));
+			if ((data & CACHE_PHYSADDR_MASK) == (v & CACHE_PHYSADDR_MASK)) {
+				data &= ~SH_CACHE_UPDATED;
+				__raw_writel(data, addr | (way << 12));
+			}
+		}
+	}
+}
+
+static void sh2__flush_purge_region(void *start, int size)
+{
+	unsigned long v;
+	unsigned long begin, end;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+
+	for (v = begin; v < end; v+=L1_CACHE_BYTES)
+		__raw_writel((v & CACHE_PHYSADDR_MASK),
+			  CACHE_OC_ADDRESS_ARRAY | (v & 0x00000ff0) | 0x00000008);
+}
+
+static void sh2__flush_invalidate_region(void *start, int size)
+{
+#ifdef CONFIG_CACHE_WRITEBACK
+	/*
+	 * SH-2 does not support individual line invalidation, only a
+	 * global invalidate.
+	 */
+	unsigned long ccr;
+	unsigned long flags;
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	ccr = __raw_readl(SH_CCR);
+	ccr |= CCR_CACHE_INVALIDATE;
+	__raw_writel(ccr, SH_CCR);
+
+	back_to_cached();
+	local_irq_restore(flags);
+#else
+	unsigned long v;
+	unsigned long begin, end;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+
+	for (v = begin; v < end; v+=L1_CACHE_BYTES)
+		__raw_writel((v & CACHE_PHYSADDR_MASK),
+			  CACHE_OC_ADDRESS_ARRAY | (v & 0x00000ff0) | 0x00000008);
+#endif
+}
+
+void __init sh2_cache_init(void)
+{
+	__flush_wback_region		= sh2__flush_wback_region;
+	__flush_purge_region		= sh2__flush_purge_region;
+	__flush_invalidate_region	= sh2__flush_invalidate_region;
+}
diff --git a/arch/sh/mm/cache-sh2a.c b/arch/sh/mm/cache-sh2a.c
new file mode 100644
index 0000000..ee87d08
--- /dev/null
+++ b/arch/sh/mm/cache-sh2a.c
@@ -0,0 +1,189 @@
+/*
+ * arch/sh/mm/cache-sh2a.c
+ *
+ * Copyright (C) 2008 Yoshinori Sato
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+
+#include <linux/init.h>
+#include <linux/mm.h>
+
+#include <asm/cache.h>
+#include <asm/addrspace.h>
+#include <asm/processor.h>
+#include <asm/cacheflush.h>
+#include <asm/io.h>
+
+/*
+ * The maximum number of pages we support up to when doing ranged dcache
+ * flushing. Anything exceeding this will simply flush the dcache in its
+ * entirety.
+ */
+#define MAX_OCACHE_PAGES	32
+#define MAX_ICACHE_PAGES	32
+
+#ifdef CONFIG_CACHE_WRITEBACK
+static void sh2a_flush_oc_line(unsigned long v, int way)
+{
+	unsigned long addr = (v & 0x000007f0) | (way << 11);
+	unsigned long data;
+
+	data = __raw_readl(CACHE_OC_ADDRESS_ARRAY | addr);
+	if ((data & CACHE_PHYSADDR_MASK) == (v & CACHE_PHYSADDR_MASK)) {
+		data &= ~SH_CACHE_UPDATED;
+		__raw_writel(data, CACHE_OC_ADDRESS_ARRAY | addr);
+	}
+}
+#endif
+
+static void sh2a_invalidate_line(unsigned long cache_addr, unsigned long v)
+{
+	/* Set associative bit to hit all ways */
+	unsigned long addr = (v & 0x000007f0) | SH_CACHE_ASSOC;
+	__raw_writel((addr & CACHE_PHYSADDR_MASK), cache_addr | addr);
+}
+
+/*
+ * Write back the dirty D-caches, but not invalidate them.
+ */
+static void sh2a__flush_wback_region(void *start, int size)
+{
+#ifdef CONFIG_CACHE_WRITEBACK
+	unsigned long v;
+	unsigned long begin, end;
+	unsigned long flags;
+	int nr_ways;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+	nr_ways = current_cpu_data.dcache.ways;
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	/* If there are too many pages then flush the entire cache */
+	if (((end - begin) >> PAGE_SHIFT) >= MAX_OCACHE_PAGES) {
+		begin = CACHE_OC_ADDRESS_ARRAY;
+		end = begin + (nr_ways * current_cpu_data.dcache.way_size);
+
+		for (v = begin; v < end; v += L1_CACHE_BYTES) {
+			unsigned long data = __raw_readl(v);
+			if (data & SH_CACHE_UPDATED)
+				__raw_writel(data & ~SH_CACHE_UPDATED, v);
+		}
+	} else {
+		int way;
+		for (way = 0; way < nr_ways; way++) {
+			for (v = begin; v < end; v += L1_CACHE_BYTES)
+				sh2a_flush_oc_line(v, way);
+		}
+	}
+
+	back_to_cached();
+	local_irq_restore(flags);
+#endif
+}
+
+/*
+ * Write back the dirty D-caches and invalidate them.
+ */
+static void sh2a__flush_purge_region(void *start, int size)
+{
+	unsigned long v;
+	unsigned long begin, end;
+	unsigned long flags;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	for (v = begin; v < end; v+=L1_CACHE_BYTES) {
+#ifdef CONFIG_CACHE_WRITEBACK
+		int way;
+		int nr_ways = current_cpu_data.dcache.ways;
+		for (way = 0; way < nr_ways; way++)
+			sh2a_flush_oc_line(v, way);
+#endif
+		sh2a_invalidate_line(CACHE_OC_ADDRESS_ARRAY, v);
+	}
+
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+/*
+ * Invalidate the D-caches, but no write back please
+ */
+static void sh2a__flush_invalidate_region(void *start, int size)
+{
+	unsigned long v;
+	unsigned long begin, end;
+	unsigned long flags;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	/* If there are too many pages then just blow the cache */
+	if (((end - begin) >> PAGE_SHIFT) >= MAX_OCACHE_PAGES) {
+		__raw_writel(__raw_readl(SH_CCR) | CCR_OCACHE_INVALIDATE,
+			     SH_CCR);
+	} else {
+		for (v = begin; v < end; v += L1_CACHE_BYTES)
+			sh2a_invalidate_line(CACHE_OC_ADDRESS_ARRAY, v);
+	}
+
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+/*
+ * Write back the range of D-cache, and purge the I-cache.
+ */
+static void sh2a_flush_icache_range(void *args)
+{
+	struct flusher_data *data = args;
+	unsigned long start, end;
+	unsigned long v;
+	unsigned long flags;
+
+	start = data->addr1 & ~(L1_CACHE_BYTES-1);
+	end = (data->addr2 + L1_CACHE_BYTES-1) & ~(L1_CACHE_BYTES-1);
+
+#ifdef CONFIG_CACHE_WRITEBACK
+	sh2a__flush_wback_region((void *)start, end-start);
+#endif
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	/* I-Cache invalidate */
+	/* If there are too many pages then just blow the cache */
+	if (((end - start) >> PAGE_SHIFT) >= MAX_ICACHE_PAGES) {
+		__raw_writel(__raw_readl(SH_CCR) | CCR_ICACHE_INVALIDATE,
+			     SH_CCR);
+	} else {
+		for (v = start; v < end; v += L1_CACHE_BYTES)
+			sh2a_invalidate_line(CACHE_IC_ADDRESS_ARRAY, v);
+	}
+
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+void __init sh2a_cache_init(void)
+{
+	local_flush_icache_range	= sh2a_flush_icache_range;
+
+	__flush_wback_region		= sh2a__flush_wback_region;
+	__flush_purge_region		= sh2a__flush_purge_region;
+	__flush_invalidate_region	= sh2a__flush_invalidate_region;
+}
diff --git a/arch/sh/mm/cache-sh3.c b/arch/sh/mm/cache-sh3.c
new file mode 100644
index 0000000..e37523f
--- /dev/null
+++ b/arch/sh/mm/cache-sh3.c
@@ -0,0 +1,105 @@
+/*
+ * arch/sh/mm/cache-sh3.c
+ *
+ * Copyright (C) 1999, 2000  Niibe Yutaka
+ * Copyright (C) 2002 Paul Mundt
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+
+#include <linux/init.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/threads.h>
+#include <asm/addrspace.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/cache.h>
+#include <asm/io.h>
+#include <asm/uaccess.h>
+#include <asm/pgalloc.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+/*
+ * Write back the dirty D-caches, but not invalidate them.
+ *
+ * Is this really worth it, or should we just alias this routine
+ * to __flush_purge_region too?
+ *
+ * START: Virtual Address (U0, P1, or P3)
+ * SIZE: Size of the region.
+ */
+
+static void sh3__flush_wback_region(void *start, int size)
+{
+	unsigned long v, j;
+	unsigned long begin, end;
+	unsigned long flags;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+
+	for (v = begin; v < end; v+=L1_CACHE_BYTES) {
+		unsigned long addrstart = CACHE_OC_ADDRESS_ARRAY;
+		for (j = 0; j < current_cpu_data.dcache.ways; j++) {
+			unsigned long data, addr, p;
+
+			p = __pa(v);
+			addr = addrstart | (v & current_cpu_data.dcache.entry_mask);
+			local_irq_save(flags);
+			data = __raw_readl(addr);
+
+			if ((data & CACHE_PHYSADDR_MASK) ==
+			    (p & CACHE_PHYSADDR_MASK)) {
+				data &= ~SH_CACHE_UPDATED;
+				__raw_writel(data, addr);
+				local_irq_restore(flags);
+				break;
+			}
+			local_irq_restore(flags);
+			addrstart += current_cpu_data.dcache.way_incr;
+		}
+	}
+}
+
+/*
+ * Write back the dirty D-caches and invalidate them.
+ *
+ * START: Virtual Address (U0, P1, or P3)
+ * SIZE: Size of the region.
+ */
+static void sh3__flush_purge_region(void *start, int size)
+{
+	unsigned long v;
+	unsigned long begin, end;
+
+	begin = (unsigned long)start & ~(L1_CACHE_BYTES-1);
+	end = ((unsigned long)start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+
+	for (v = begin; v < end; v+=L1_CACHE_BYTES) {
+		unsigned long data, addr;
+
+		data = (v & 0xfffffc00); /* _Virtual_ address, ~U, ~V */
+		addr = CACHE_OC_ADDRESS_ARRAY |
+			(v & current_cpu_data.dcache.entry_mask) | SH_CACHE_ASSOC;
+		__raw_writel(data, addr);
+	}
+}
+
+void __init sh3_cache_init(void)
+{
+	__flush_wback_region = sh3__flush_wback_region;
+	__flush_purge_region = sh3__flush_purge_region;
+
+	/*
+	 * No write back please
+	 *
+	 * Except I don't think there's any way to avoid the writeback.
+	 * So we just alias it to sh3__flush_purge_region(). dwmw2.
+	 */
+	__flush_invalidate_region = sh3__flush_purge_region;
+}
diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c
new file mode 100644
index 0000000..51d8f7f
--- /dev/null
+++ b/arch/sh/mm/cache-sh4.c
@@ -0,0 +1,394 @@
+/*
+ * arch/sh/mm/cache-sh4.c
+ *
+ * Copyright (C) 1999, 2000, 2002  Niibe Yutaka
+ * Copyright (C) 2001 - 2009  Paul Mundt
+ * Copyright (C) 2003  Richard Curnow
+ * Copyright (c) 2007 STMicroelectronics (R&D) Ltd.
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/mm.h>
+#include <linux/io.h>
+#include <linux/mutex.h>
+#include <linux/fs.h>
+#include <linux/highmem.h>
+#include <asm/pgtable.h>
+#include <asm/mmu_context.h>
+#include <asm/cache_insns.h>
+#include <asm/cacheflush.h>
+
+/*
+ * The maximum number of pages we support up to when doing ranged dcache
+ * flushing. Anything exceeding this will simply flush the dcache in its
+ * entirety.
+ */
+#define MAX_ICACHE_PAGES	32
+
+static void __flush_cache_one(unsigned long addr, unsigned long phys,
+			       unsigned long exec_offset);
+
+/*
+ * Write back the range of D-cache, and purge the I-cache.
+ *
+ * Called from kernel/module.c:sys_init_module and routine for a.out format,
+ * signal handler code and kprobes code
+ */
+static void sh4_flush_icache_range(void *args)
+{
+	struct flusher_data *data = args;
+	unsigned long start, end;
+	unsigned long flags, v;
+	int i;
+
+	start = data->addr1;
+	end = data->addr2;
+
+	/* If there are too many pages then just blow away the caches */
+	if (((end - start) >> PAGE_SHIFT) >= MAX_ICACHE_PAGES) {
+		local_flush_cache_all(NULL);
+		return;
+	}
+
+	/*
+	 * Selectively flush d-cache then invalidate the i-cache.
+	 * This is inefficient, so only use this for small ranges.
+	 */
+	start &= ~(L1_CACHE_BYTES-1);
+	end += L1_CACHE_BYTES-1;
+	end &= ~(L1_CACHE_BYTES-1);
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	for (v = start; v < end; v += L1_CACHE_BYTES) {
+		unsigned long icacheaddr;
+		int j, n;
+
+		__ocbwb(v);
+
+		icacheaddr = CACHE_IC_ADDRESS_ARRAY | (v &
+				cpu_data->icache.entry_mask);
+
+		/* Clear i-cache line valid-bit */
+		n = boot_cpu_data.icache.n_aliases;
+		for (i = 0; i < cpu_data->icache.ways; i++) {
+			for (j = 0; j < n; j++)
+				__raw_writel(0, icacheaddr + (j * PAGE_SIZE));
+			icacheaddr += cpu_data->icache.way_incr;
+		}
+	}
+
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+static inline void flush_cache_one(unsigned long start, unsigned long phys)
+{
+	unsigned long flags, exec_offset = 0;
+
+	/*
+	 * All types of SH-4 require PC to be uncached to operate on the I-cache.
+	 * Some types of SH-4 require PC to be uncached to operate on the D-cache.
+	 */
+	if ((boot_cpu_data.flags & CPU_HAS_P2_FLUSH_BUG) ||
+	    (start < CACHE_OC_ADDRESS_ARRAY))
+		exec_offset = cached_to_uncached;
+
+	local_irq_save(flags);
+	__flush_cache_one(start, phys, exec_offset);
+	local_irq_restore(flags);
+}
+
+/*
+ * Write back & invalidate the D-cache of the page.
+ * (To avoid "alias" issues)
+ */
+static void sh4_flush_dcache_page(void *arg)
+{
+	struct page *page = arg;
+	unsigned long addr = (unsigned long)page_address(page);
+#ifndef CONFIG_SMP
+	struct address_space *mapping = page_mapping(page);
+
+	if (mapping && !mapping_mapped(mapping))
+		clear_bit(PG_dcache_clean, &page->flags);
+	else
+#endif
+		flush_cache_one(CACHE_OC_ADDRESS_ARRAY |
+				(addr & shm_align_mask), page_to_phys(page));
+
+	wmb();
+}
+
+/* TODO: Selective icache invalidation through IC address array.. */
+static void flush_icache_all(void)
+{
+	unsigned long flags, ccr;
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	/* Flush I-cache */
+	ccr = __raw_readl(SH_CCR);
+	ccr |= CCR_CACHE_ICI;
+	__raw_writel(ccr, SH_CCR);
+
+	/*
+	 * back_to_cached() will take care of the barrier for us, don't add
+	 * another one!
+	 */
+
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+static void flush_dcache_all(void)
+{
+	unsigned long addr, end_addr, entry_offset;
+
+	end_addr = CACHE_OC_ADDRESS_ARRAY +
+		(current_cpu_data.dcache.sets <<
+		 current_cpu_data.dcache.entry_shift) *
+			current_cpu_data.dcache.ways;
+
+	entry_offset = 1 << current_cpu_data.dcache.entry_shift;
+
+	for (addr = CACHE_OC_ADDRESS_ARRAY; addr < end_addr; ) {
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+		__raw_writel(0, addr); addr += entry_offset;
+	}
+}
+
+static void sh4_flush_cache_all(void *unused)
+{
+	flush_dcache_all();
+	flush_icache_all();
+}
+
+/*
+ * Note : (RPC) since the caches are physically tagged, the only point
+ * of flush_cache_mm for SH-4 is to get rid of aliases from the
+ * D-cache.  The assumption elsewhere, e.g. flush_cache_range, is that
+ * lines can stay resident so long as the virtual address they were
+ * accessed with (hence cache set) is in accord with the physical
+ * address (i.e. tag).  It's no different here.
+ *
+ * Caller takes mm->mmap_sem.
+ */
+static void sh4_flush_cache_mm(void *arg)
+{
+	struct mm_struct *mm = arg;
+
+	if (cpu_context(smp_processor_id(), mm) == NO_CONTEXT)
+		return;
+
+	flush_dcache_all();
+}
+
+/*
+ * Write back and invalidate I/D-caches for the page.
+ *
+ * ADDR: Virtual Address (U0 address)
+ * PFN: Physical page number
+ */
+static void sh4_flush_cache_page(void *args)
+{
+	struct flusher_data *data = args;
+	struct vm_area_struct *vma;
+	struct page *page;
+	unsigned long address, pfn, phys;
+	int map_coherent = 0;
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+	void *vaddr;
+
+	vma = data->vma;
+	address = data->addr1 & PAGE_MASK;
+	pfn = data->addr2;
+	phys = pfn << PAGE_SHIFT;
+	page = pfn_to_page(pfn);
+
+	if (cpu_context(smp_processor_id(), vma->vm_mm) == NO_CONTEXT)
+		return;
+
+	pgd = pgd_offset(vma->vm_mm, address);
+	pud = pud_offset(pgd, address);
+	pmd = pmd_offset(pud, address);
+	pte = pte_offset_kernel(pmd, address);
+
+	/* If the page isn't present, there is nothing to do here. */
+	if (!(pte_val(*pte) & _PAGE_PRESENT))
+		return;
+
+	if ((vma->vm_mm == current->active_mm))
+		vaddr = NULL;
+	else {
+		/*
+		 * Use kmap_coherent or kmap_atomic to do flushes for
+		 * another ASID than the current one.
+		 */
+		map_coherent = (current_cpu_data.dcache.n_aliases &&
+			test_bit(PG_dcache_clean, &page->flags) &&
+			page_mapped(page));
+		if (map_coherent)
+			vaddr = kmap_coherent(page, address);
+		else
+			vaddr = kmap_atomic(page);
+
+		address = (unsigned long)vaddr;
+	}
+
+	flush_cache_one(CACHE_OC_ADDRESS_ARRAY |
+			(address & shm_align_mask), phys);
+
+	if (vma->vm_flags & VM_EXEC)
+		flush_icache_all();
+
+	if (vaddr) {
+		if (map_coherent)
+			kunmap_coherent(vaddr);
+		else
+			kunmap_atomic(vaddr);
+	}
+}
+
+/*
+ * Write back and invalidate D-caches.
+ *
+ * START, END: Virtual Address (U0 address)
+ *
+ * NOTE: We need to flush the _physical_ page entry.
+ * Flushing the cache lines for U0 only isn't enough.
+ * We need to flush for P1 too, which may contain aliases.
+ */
+static void sh4_flush_cache_range(void *args)
+{
+	struct flusher_data *data = args;
+	struct vm_area_struct *vma;
+	unsigned long start, end;
+
+	vma = data->vma;
+	start = data->addr1;
+	end = data->addr2;
+
+	if (cpu_context(smp_processor_id(), vma->vm_mm) == NO_CONTEXT)
+		return;
+
+	/*
+	 * If cache is only 4k-per-way, there are never any 'aliases'.  Since
+	 * the cache is physically tagged, the data can just be left in there.
+	 */
+	if (boot_cpu_data.dcache.n_aliases == 0)
+		return;
+
+	flush_dcache_all();
+
+	if (vma->vm_flags & VM_EXEC)
+		flush_icache_all();
+}
+
+/**
+ * __flush_cache_one
+ *
+ * @addr:  address in memory mapped cache array
+ * @phys:  P1 address to flush (has to match tags if addr has 'A' bit
+ *         set i.e. associative write)
+ * @exec_offset: set to 0x20000000 if flush has to be executed from P2
+ *               region else 0x0
+ *
+ * The offset into the cache array implied by 'addr' selects the
+ * 'colour' of the virtual address range that will be flushed.  The
+ * operation (purge/write-back) is selected by the lower 2 bits of
+ * 'phys'.
+ */
+static void __flush_cache_one(unsigned long addr, unsigned long phys,
+			       unsigned long exec_offset)
+{
+	int way_count;
+	unsigned long base_addr = addr;
+	struct cache_info *dcache;
+	unsigned long way_incr;
+	unsigned long a, ea, p;
+	unsigned long temp_pc;
+
+	dcache = &boot_cpu_data.dcache;
+	/* Write this way for better assembly. */
+	way_count = dcache->ways;
+	way_incr = dcache->way_incr;
+
+	/*
+	 * Apply exec_offset (i.e. branch to P2 if required.).
+	 *
+	 * FIXME:
+	 *
+	 *	If I write "=r" for the (temp_pc), it puts this in r6 hence
+	 *	trashing exec_offset before it's been added on - why?  Hence
+	 *	"=&r" as a 'workaround'
+	 */
+	asm volatile("mov.l 1f, %0\n\t"
+		     "add   %1, %0\n\t"
+		     "jmp   @%0\n\t"
+		     "nop\n\t"
+		     ".balign 4\n\t"
+		     "1:  .long 2f\n\t"
+		     "2:\n" : "=&r" (temp_pc) : "r" (exec_offset));
+
+	/*
+	 * We know there will be >=1 iteration, so write as do-while to avoid
+	 * pointless nead-of-loop check for 0 iterations.
+	 */
+	do {
+		ea = base_addr + PAGE_SIZE;
+		a = base_addr;
+		p = phys;
+
+		do {
+			*(volatile unsigned long *)a = p;
+			/*
+			 * Next line: intentionally not p+32, saves an add, p
+			 * will do since only the cache tag bits need to
+			 * match.
+			 */
+			*(volatile unsigned long *)(a+32) = p;
+			a += 64;
+			p += 64;
+		} while (a < ea);
+
+		base_addr += way_incr;
+	} while (--way_count != 0);
+}
+
+extern void __weak sh4__flush_region_init(void);
+
+/*
+ * SH-4 has virtually indexed and physically tagged cache.
+ */
+void __init sh4_cache_init(void)
+{
+	printk("PVR=%08x CVR=%08x PRR=%08x\n",
+		__raw_readl(CCN_PVR),
+		__raw_readl(CCN_CVR),
+		__raw_readl(CCN_PRR));
+
+	local_flush_icache_range	= sh4_flush_icache_range;
+	local_flush_dcache_page		= sh4_flush_dcache_page;
+	local_flush_cache_all		= sh4_flush_cache_all;
+	local_flush_cache_mm		= sh4_flush_cache_mm;
+	local_flush_cache_dup_mm	= sh4_flush_cache_mm;
+	local_flush_cache_page		= sh4_flush_cache_page;
+	local_flush_cache_range		= sh4_flush_cache_range;
+
+	sh4__flush_region_init();
+}
diff --git a/arch/sh/mm/cache-sh5.c b/arch/sh/mm/cache-sh5.c
new file mode 100644
index 0000000..d1bffbc
--- /dev/null
+++ b/arch/sh/mm/cache-sh5.c
@@ -0,0 +1,621 @@
+/*
+ * arch/sh/mm/cache-sh5.c
+ *
+ * Copyright (C) 2000, 2001  Paolo Alberelli
+ * Copyright (C) 2002  Benedict Gaster
+ * Copyright (C) 2003  Richard Curnow
+ * Copyright (C) 2003 - 2008  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <asm/tlb.h>
+#include <asm/processor.h>
+#include <asm/cache.h>
+#include <asm/pgalloc.h>
+#include <asm/uaccess.h>
+#include <asm/mmu_context.h>
+
+extern void __weak sh4__flush_region_init(void);
+
+/* Wired TLB entry for the D-cache */
+static unsigned long long dtlb_cache_slot;
+
+/*
+ * The following group of functions deal with mapping and unmapping a
+ * temporary page into a DTLB slot that has been set aside for exclusive
+ * use.
+ */
+static inline void
+sh64_setup_dtlb_cache_slot(unsigned long eaddr, unsigned long asid,
+			   unsigned long paddr)
+{
+	local_irq_disable();
+	sh64_setup_tlb_slot(dtlb_cache_slot, eaddr, asid, paddr);
+}
+
+static inline void sh64_teardown_dtlb_cache_slot(void)
+{
+	sh64_teardown_tlb_slot(dtlb_cache_slot);
+	local_irq_enable();
+}
+
+static inline void sh64_icache_inv_all(void)
+{
+	unsigned long long addr, flag, data;
+	unsigned long flags;
+
+	addr = ICCR0;
+	flag = ICCR0_ICI;
+	data = 0;
+
+	/* Make this a critical section for safety (probably not strictly necessary.) */
+	local_irq_save(flags);
+
+	/* Without %1 it gets unexplicably wrong */
+	__asm__ __volatile__ (
+		"getcfg	%3, 0, %0\n\t"
+		"or	%0, %2, %0\n\t"
+		"putcfg	%3, 0, %0\n\t"
+		"synci"
+		: "=&r" (data)
+		: "0" (data), "r" (flag), "r" (addr));
+
+	local_irq_restore(flags);
+}
+
+static void sh64_icache_inv_kernel_range(unsigned long start, unsigned long end)
+{
+	/* Invalidate range of addresses [start,end] from the I-cache, where
+	 * the addresses lie in the kernel superpage. */
+
+	unsigned long long ullend, addr, aligned_start;
+	aligned_start = (unsigned long long)(signed long long)(signed long) start;
+	addr = L1_CACHE_ALIGN(aligned_start);
+	ullend = (unsigned long long) (signed long long) (signed long) end;
+
+	while (addr <= ullend) {
+		__asm__ __volatile__ ("icbi %0, 0" : : "r" (addr));
+		addr += L1_CACHE_BYTES;
+	}
+}
+
+static void sh64_icache_inv_user_page(struct vm_area_struct *vma, unsigned long eaddr)
+{
+	/* If we get called, we know that vma->vm_flags contains VM_EXEC.
+	   Also, eaddr is page-aligned. */
+	unsigned int cpu = smp_processor_id();
+	unsigned long long addr, end_addr;
+	unsigned long flags = 0;
+	unsigned long running_asid, vma_asid;
+	addr = eaddr;
+	end_addr = addr + PAGE_SIZE;
+
+	/* Check whether we can use the current ASID for the I-cache
+	   invalidation.  For example, if we're called via
+	   access_process_vm->flush_cache_page->here, (e.g. when reading from
+	   /proc), 'running_asid' will be that of the reader, not of the
+	   victim.
+
+	   Also, note the risk that we might get pre-empted between the ASID
+	   compare and blocking IRQs, and before we regain control, the
+	   pid->ASID mapping changes.  However, the whole cache will get
+	   invalidated when the mapping is renewed, so the worst that can
+	   happen is that the loop below ends up invalidating somebody else's
+	   cache entries.
+	*/
+
+	running_asid = get_asid();
+	vma_asid = cpu_asid(cpu, vma->vm_mm);
+	if (running_asid != vma_asid) {
+		local_irq_save(flags);
+		switch_and_save_asid(vma_asid);
+	}
+	while (addr < end_addr) {
+		/* Worth unrolling a little */
+		__asm__ __volatile__("icbi %0,  0" : : "r" (addr));
+		__asm__ __volatile__("icbi %0, 32" : : "r" (addr));
+		__asm__ __volatile__("icbi %0, 64" : : "r" (addr));
+		__asm__ __volatile__("icbi %0, 96" : : "r" (addr));
+		addr += 128;
+	}
+	if (running_asid != vma_asid) {
+		switch_and_save_asid(running_asid);
+		local_irq_restore(flags);
+	}
+}
+
+static void sh64_icache_inv_user_page_range(struct mm_struct *mm,
+			  unsigned long start, unsigned long end)
+{
+	/* Used for invalidating big chunks of I-cache, i.e. assume the range
+	   is whole pages.  If 'start' or 'end' is not page aligned, the code
+	   is conservative and invalidates to the ends of the enclosing pages.
+	   This is functionally OK, just a performance loss. */
+
+	/* See the comments below in sh64_dcache_purge_user_range() regarding
+	   the choice of algorithm.  However, for the I-cache option (2) isn't
+	   available because there are no physical tags so aliases can't be
+	   resolved.  The icbi instruction has to be used through the user
+	   mapping.   Because icbi is cheaper than ocbp on a cache hit, it
+	   would be cheaper to use the selective code for a large range than is
+	   possible with the D-cache.  Just assume 64 for now as a working
+	   figure.
+	   */
+	int n_pages;
+
+	if (!mm)
+		return;
+
+	n_pages = ((end - start) >> PAGE_SHIFT);
+	if (n_pages >= 64) {
+		sh64_icache_inv_all();
+	} else {
+		unsigned long aligned_start;
+		unsigned long eaddr;
+		unsigned long after_last_page_start;
+		unsigned long mm_asid, current_asid;
+		unsigned long flags = 0;
+
+		mm_asid = cpu_asid(smp_processor_id(), mm);
+		current_asid = get_asid();
+
+		if (mm_asid != current_asid) {
+			/* Switch ASID and run the invalidate loop under cli */
+			local_irq_save(flags);
+			switch_and_save_asid(mm_asid);
+		}
+
+		aligned_start = start & PAGE_MASK;
+		after_last_page_start = PAGE_SIZE + ((end - 1) & PAGE_MASK);
+
+		while (aligned_start < after_last_page_start) {
+			struct vm_area_struct *vma;
+			unsigned long vma_end;
+			vma = find_vma(mm, aligned_start);
+			if (!vma || (aligned_start <= vma->vm_end)) {
+				/* Avoid getting stuck in an error condition */
+				aligned_start += PAGE_SIZE;
+				continue;
+			}
+			vma_end = vma->vm_end;
+			if (vma->vm_flags & VM_EXEC) {
+				/* Executable */
+				eaddr = aligned_start;
+				while (eaddr < vma_end) {
+					sh64_icache_inv_user_page(vma, eaddr);
+					eaddr += PAGE_SIZE;
+				}
+			}
+			aligned_start = vma->vm_end; /* Skip to start of next region */
+		}
+
+		if (mm_asid != current_asid) {
+			switch_and_save_asid(current_asid);
+			local_irq_restore(flags);
+		}
+	}
+}
+
+static void sh64_icache_inv_current_user_range(unsigned long start, unsigned long end)
+{
+	/* The icbi instruction never raises ITLBMISS.  i.e. if there's not a
+	   cache hit on the virtual tag the instruction ends there, without a
+	   TLB lookup. */
+
+	unsigned long long aligned_start;
+	unsigned long long ull_end;
+	unsigned long long addr;
+
+	ull_end = end;
+
+	/* Just invalidate over the range using the natural addresses.  TLB
+	   miss handling will be OK (TBC).  Since it's for the current process,
+	   either we're already in the right ASID context, or the ASIDs have
+	   been recycled since we were last active in which case we might just
+	   invalidate another processes I-cache entries : no worries, just a
+	   performance drop for him. */
+	aligned_start = L1_CACHE_ALIGN(start);
+	addr = aligned_start;
+	while (addr < ull_end) {
+		__asm__ __volatile__ ("icbi %0, 0" : : "r" (addr));
+		__asm__ __volatile__ ("nop");
+		__asm__ __volatile__ ("nop");
+		addr += L1_CACHE_BYTES;
+	}
+}
+
+/* Buffer used as the target of alloco instructions to purge data from cache
+   sets by natural eviction. -- RPC */
+#define DUMMY_ALLOCO_AREA_SIZE ((L1_CACHE_BYTES << 10) + (1024 * 4))
+static unsigned char dummy_alloco_area[DUMMY_ALLOCO_AREA_SIZE] __cacheline_aligned = { 0, };
+
+static void inline sh64_dcache_purge_sets(int sets_to_purge_base, int n_sets)
+{
+	/* Purge all ways in a particular block of sets, specified by the base
+	   set number and number of sets.  Can handle wrap-around, if that's
+	   needed.  */
+
+	int dummy_buffer_base_set;
+	unsigned long long eaddr, eaddr0, eaddr1;
+	int j;
+	int set_offset;
+
+	dummy_buffer_base_set = ((int)&dummy_alloco_area &
+				 cpu_data->dcache.entry_mask) >>
+				 cpu_data->dcache.entry_shift;
+	set_offset = sets_to_purge_base - dummy_buffer_base_set;
+
+	for (j = 0; j < n_sets; j++, set_offset++) {
+		set_offset &= (cpu_data->dcache.sets - 1);
+		eaddr0 = (unsigned long long)dummy_alloco_area +
+			(set_offset << cpu_data->dcache.entry_shift);
+
+		/*
+		 * Do one alloco which hits the required set per cache
+		 * way.  For write-back mode, this will purge the #ways
+		 * resident lines.  There's little point unrolling this
+		 * loop because the allocos stall more if they're too
+		 * close together.
+		 */
+		eaddr1 = eaddr0 + cpu_data->dcache.way_size *
+				  cpu_data->dcache.ways;
+
+		for (eaddr = eaddr0; eaddr < eaddr1;
+		     eaddr += cpu_data->dcache.way_size) {
+			__asm__ __volatile__ ("alloco %0, 0" : : "r" (eaddr));
+			__asm__ __volatile__ ("synco"); /* TAKum03020 */
+		}
+
+		eaddr1 = eaddr0 + cpu_data->dcache.way_size *
+				  cpu_data->dcache.ways;
+
+		for (eaddr = eaddr0; eaddr < eaddr1;
+		     eaddr += cpu_data->dcache.way_size) {
+			/*
+			 * Load from each address.  Required because
+			 * alloco is a NOP if the cache is write-through.
+			 */
+			if (test_bit(SH_CACHE_MODE_WT, &(cpu_data->dcache.flags)))
+				__raw_readb((unsigned long)eaddr);
+		}
+	}
+
+	/*
+	 * Don't use OCBI to invalidate the lines.  That costs cycles
+	 * directly.  If the dummy block is just left resident, it will
+	 * naturally get evicted as required.
+	 */
+}
+
+/*
+ * Purge the entire contents of the dcache.  The most efficient way to
+ * achieve this is to use alloco instructions on a region of unused
+ * memory equal in size to the cache, thereby causing the current
+ * contents to be discarded by natural eviction.  The alternative, namely
+ * reading every tag, setting up a mapping for the corresponding page and
+ * doing an OCBP for the line, would be much more expensive.
+ */
+static void sh64_dcache_purge_all(void)
+{
+
+	sh64_dcache_purge_sets(0, cpu_data->dcache.sets);
+}
+
+
+/* Assumes this address (+ (2**n_synbits) pages up from it) aren't used for
+   anything else in the kernel */
+#define MAGIC_PAGE0_START 0xffffffffec000000ULL
+
+/* Purge the physical page 'paddr' from the cache.  It's known that any
+ * cache lines requiring attention have the same page colour as the the
+ * address 'eaddr'.
+ *
+ * This relies on the fact that the D-cache matches on physical tags when
+ * no virtual tag matches.  So we create an alias for the original page
+ * and purge through that.  (Alternatively, we could have done this by
+ * switching ASID to match the original mapping and purged through that,
+ * but that involves ASID switching cost + probably a TLBMISS + refill
+ * anyway.)
+ */
+static void sh64_dcache_purge_coloured_phy_page(unsigned long paddr,
+					        unsigned long eaddr)
+{
+	unsigned long long magic_page_start;
+	unsigned long long magic_eaddr, magic_eaddr_end;
+
+	magic_page_start = MAGIC_PAGE0_START + (eaddr & CACHE_OC_SYN_MASK);
+
+	/* As long as the kernel is not pre-emptible, this doesn't need to be
+	   under cli/sti. */
+	sh64_setup_dtlb_cache_slot(magic_page_start, get_asid(), paddr);
+
+	magic_eaddr = magic_page_start;
+	magic_eaddr_end = magic_eaddr + PAGE_SIZE;
+
+	while (magic_eaddr < magic_eaddr_end) {
+		/* Little point in unrolling this loop - the OCBPs are blocking
+		   and won't go any quicker (i.e. the loop overhead is parallel
+		   to part of the OCBP execution.) */
+		__asm__ __volatile__ ("ocbp %0, 0" : : "r" (magic_eaddr));
+		magic_eaddr += L1_CACHE_BYTES;
+	}
+
+	sh64_teardown_dtlb_cache_slot();
+}
+
+/*
+ * Purge a page given its physical start address, by creating a temporary
+ * 1 page mapping and purging across that.  Even if we know the virtual
+ * address (& vma or mm) of the page, the method here is more elegant
+ * because it avoids issues of coping with page faults on the purge
+ * instructions (i.e. no special-case code required in the critical path
+ * in the TLB miss handling).
+ */
+static void sh64_dcache_purge_phy_page(unsigned long paddr)
+{
+	unsigned long long eaddr_start, eaddr, eaddr_end;
+	int i;
+
+	/* As long as the kernel is not pre-emptible, this doesn't need to be
+	   under cli/sti. */
+	eaddr_start = MAGIC_PAGE0_START;
+	for (i = 0; i < (1 << CACHE_OC_N_SYNBITS); i++) {
+		sh64_setup_dtlb_cache_slot(eaddr_start, get_asid(), paddr);
+
+		eaddr = eaddr_start;
+		eaddr_end = eaddr + PAGE_SIZE;
+		while (eaddr < eaddr_end) {
+			__asm__ __volatile__ ("ocbp %0, 0" : : "r" (eaddr));
+			eaddr += L1_CACHE_BYTES;
+		}
+
+		sh64_teardown_dtlb_cache_slot();
+		eaddr_start += PAGE_SIZE;
+	}
+}
+
+static void sh64_dcache_purge_user_pages(struct mm_struct *mm,
+				unsigned long addr, unsigned long end)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+	pte_t entry;
+	spinlock_t *ptl;
+	unsigned long paddr;
+
+	if (!mm)
+		return; /* No way to find physical address of page */
+
+	pgd = pgd_offset(mm, addr);
+	if (pgd_bad(*pgd))
+		return;
+
+	pud = pud_offset(pgd, addr);
+	if (pud_none(*pud) || pud_bad(*pud))
+		return;
+
+	pmd = pmd_offset(pud, addr);
+	if (pmd_none(*pmd) || pmd_bad(*pmd))
+		return;
+
+	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
+	do {
+		entry = *pte;
+		if (pte_none(entry) || !pte_present(entry))
+			continue;
+		paddr = pte_val(entry) & PAGE_MASK;
+		sh64_dcache_purge_coloured_phy_page(paddr, addr);
+	} while (pte++, addr += PAGE_SIZE, addr != end);
+	pte_unmap_unlock(pte - 1, ptl);
+}
+
+/*
+ * There are at least 5 choices for the implementation of this, with
+ * pros (+), cons(-), comments(*):
+ *
+ * 1. ocbp each line in the range through the original user's ASID
+ *    + no lines spuriously evicted
+ *    - tlbmiss handling (must either handle faults on demand => extra
+ *	special-case code in tlbmiss critical path), or map the page in
+ *	advance (=> flush_tlb_range in advance to avoid multiple hits)
+ *    - ASID switching
+ *    - expensive for large ranges
+ *
+ * 2. temporarily map each page in the range to a special effective
+ *    address and ocbp through the temporary mapping; relies on the
+ *    fact that SH-5 OCB* always do TLB lookup and match on ptags (they
+ *    never look at the etags)
+ *    + no spurious evictions
+ *    - expensive for large ranges
+ *    * surely cheaper than (1)
+ *
+ * 3. walk all the lines in the cache, check the tags, if a match
+ *    occurs create a page mapping to ocbp the line through
+ *    + no spurious evictions
+ *    - tag inspection overhead
+ *    - (especially for small ranges)
+ *    - potential cost of setting up/tearing down page mapping for
+ *	every line that matches the range
+ *    * cost partly independent of range size
+ *
+ * 4. walk all the lines in the cache, check the tags, if a match
+ *    occurs use 4 * alloco to purge the line (+3 other probably
+ *    innocent victims) by natural eviction
+ *    + no tlb mapping overheads
+ *    - spurious evictions
+ *    - tag inspection overhead
+ *
+ * 5. implement like flush_cache_all
+ *    + no tag inspection overhead
+ *    - spurious evictions
+ *    - bad for small ranges
+ *
+ * (1) can be ruled out as more expensive than (2).  (2) appears best
+ * for small ranges.  The choice between (3), (4) and (5) for large
+ * ranges and the range size for the large/small boundary need
+ * benchmarking to determine.
+ *
+ * For now use approach (2) for small ranges and (5) for large ones.
+ */
+static void sh64_dcache_purge_user_range(struct mm_struct *mm,
+			  unsigned long start, unsigned long end)
+{
+	int n_pages = ((end - start) >> PAGE_SHIFT);
+
+	if (n_pages >= 64 || ((start ^ (end - 1)) & PMD_MASK)) {
+		sh64_dcache_purge_all();
+	} else {
+		/* Small range, covered by a single page table page */
+		start &= PAGE_MASK;	/* should already be so */
+		end = PAGE_ALIGN(end);	/* should already be so */
+		sh64_dcache_purge_user_pages(mm, start, end);
+	}
+}
+
+/*
+ * Invalidate the entire contents of both caches, after writing back to
+ * memory any dirty data from the D-cache.
+ */
+static void sh5_flush_cache_all(void *unused)
+{
+	sh64_dcache_purge_all();
+	sh64_icache_inv_all();
+}
+
+/*
+ * Invalidate an entire user-address space from both caches, after
+ * writing back dirty data (e.g. for shared mmap etc).
+ *
+ * This could be coded selectively by inspecting all the tags then
+ * doing 4*alloco on any set containing a match (as for
+ * flush_cache_range), but fork/exit/execve (where this is called from)
+ * are expensive anyway.
+ *
+ * Have to do a purge here, despite the comments re I-cache below.
+ * There could be odd-coloured dirty data associated with the mm still
+ * in the cache - if this gets written out through natural eviction
+ * after the kernel has reused the page there will be chaos.
+ *
+ * The mm being torn down won't ever be active again, so any Icache
+ * lines tagged with its ASID won't be visible for the rest of the
+ * lifetime of this ASID cycle.  Before the ASID gets reused, there
+ * will be a flush_cache_all.  Hence we don't need to touch the
+ * I-cache.  This is similar to the lack of action needed in
+ * flush_tlb_mm - see fault.c.
+ */
+static void sh5_flush_cache_mm(void *unused)
+{
+	sh64_dcache_purge_all();
+}
+
+/*
+ * Invalidate (from both caches) the range [start,end) of virtual
+ * addresses from the user address space specified by mm, after writing
+ * back any dirty data.
+ *
+ * Note, 'end' is 1 byte beyond the end of the range to flush.
+ */
+static void sh5_flush_cache_range(void *args)
+{
+	struct flusher_data *data = args;
+	struct vm_area_struct *vma;
+	unsigned long start, end;
+
+	vma = data->vma;
+	start = data->addr1;
+	end = data->addr2;
+
+	sh64_dcache_purge_user_range(vma->vm_mm, start, end);
+	sh64_icache_inv_user_page_range(vma->vm_mm, start, end);
+}
+
+/*
+ * Invalidate any entries in either cache for the vma within the user
+ * address space vma->vm_mm for the page starting at virtual address
+ * 'eaddr'.   This seems to be used primarily in breaking COW.  Note,
+ * the I-cache must be searched too in case the page in question is
+ * both writable and being executed from (e.g. stack trampolines.)
+ *
+ * Note, this is called with pte lock held.
+ */
+static void sh5_flush_cache_page(void *args)
+{
+	struct flusher_data *data = args;
+	struct vm_area_struct *vma;
+	unsigned long eaddr, pfn;
+
+	vma = data->vma;
+	eaddr = data->addr1;
+	pfn = data->addr2;
+
+	sh64_dcache_purge_phy_page(pfn << PAGE_SHIFT);
+
+	if (vma->vm_flags & VM_EXEC)
+		sh64_icache_inv_user_page(vma, eaddr);
+}
+
+static void sh5_flush_dcache_page(void *page)
+{
+	sh64_dcache_purge_phy_page(page_to_phys((struct page *)page));
+	wmb();
+}
+
+/*
+ * Flush the range [start,end] of kernel virtual address space from
+ * the I-cache.  The corresponding range must be purged from the
+ * D-cache also because the SH-5 doesn't have cache snooping between
+ * the caches.  The addresses will be visible through the superpage
+ * mapping, therefore it's guaranteed that there no cache entries for
+ * the range in cache sets of the wrong colour.
+ */
+static void sh5_flush_icache_range(void *args)
+{
+	struct flusher_data *data = args;
+	unsigned long start, end;
+
+	start = data->addr1;
+	end = data->addr2;
+
+	__flush_purge_region((void *)start, end);
+	wmb();
+	sh64_icache_inv_kernel_range(start, end);
+}
+
+/*
+ * For the address range [start,end), write back the data from the
+ * D-cache and invalidate the corresponding region of the I-cache for the
+ * current process.  Used to flush signal trampolines on the stack to
+ * make them executable.
+ */
+static void sh5_flush_cache_sigtramp(void *vaddr)
+{
+	unsigned long end = (unsigned long)vaddr + L1_CACHE_BYTES;
+
+	__flush_wback_region(vaddr, L1_CACHE_BYTES);
+	wmb();
+	sh64_icache_inv_current_user_range((unsigned long)vaddr, end);
+}
+
+void __init sh5_cache_init(void)
+{
+	local_flush_cache_all		= sh5_flush_cache_all;
+	local_flush_cache_mm		= sh5_flush_cache_mm;
+	local_flush_cache_dup_mm	= sh5_flush_cache_mm;
+	local_flush_cache_page		= sh5_flush_cache_page;
+	local_flush_cache_range		= sh5_flush_cache_range;
+	local_flush_dcache_page		= sh5_flush_dcache_page;
+	local_flush_icache_range	= sh5_flush_icache_range;
+	local_flush_cache_sigtramp	= sh5_flush_cache_sigtramp;
+
+	/* Reserve a slot for dcache colouring in the DTLB */
+	dtlb_cache_slot	= sh64_get_wired_dtlb_entry();
+
+	sh4__flush_region_init();
+}
diff --git a/arch/sh/mm/cache-sh7705.c b/arch/sh/mm/cache-sh7705.c
new file mode 100644
index 0000000..7729cca
--- /dev/null
+++ b/arch/sh/mm/cache-sh7705.c
@@ -0,0 +1,195 @@
+/*
+ * arch/sh/mm/cache-sh7705.c
+ *
+ * Copyright (C) 1999, 2000  Niibe Yutaka
+ * Copyright (C) 2004  Alex Song
+ *
+ * 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.
+ *
+ */
+#include <linux/init.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/fs.h>
+#include <linux/threads.h>
+#include <asm/addrspace.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/processor.h>
+#include <asm/cache.h>
+#include <asm/io.h>
+#include <asm/uaccess.h>
+#include <asm/pgalloc.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+/*
+ * The 32KB cache on the SH7705 suffers from the same synonym problem
+ * as SH4 CPUs
+ */
+static inline void cache_wback_all(void)
+{
+	unsigned long ways, waysize, addrstart;
+
+	ways = current_cpu_data.dcache.ways;
+	waysize = current_cpu_data.dcache.sets;
+	waysize <<= current_cpu_data.dcache.entry_shift;
+
+	addrstart = CACHE_OC_ADDRESS_ARRAY;
+
+	do {
+		unsigned long addr;
+
+		for (addr = addrstart;
+		     addr < addrstart + waysize;
+		     addr += current_cpu_data.dcache.linesz) {
+			unsigned long data;
+			int v = SH_CACHE_UPDATED | SH_CACHE_VALID;
+
+			data = __raw_readl(addr);
+
+			if ((data & v) == v)
+				__raw_writel(data & ~v, addr);
+
+		}
+
+		addrstart += current_cpu_data.dcache.way_incr;
+	} while (--ways);
+}
+
+/*
+ * Write back the range of D-cache, and purge the I-cache.
+ *
+ * Called from kernel/module.c:sys_init_module and routine for a.out format.
+ */
+static void sh7705_flush_icache_range(void *args)
+{
+	struct flusher_data *data = args;
+	unsigned long start, end;
+
+	start = data->addr1;
+	end = data->addr2;
+
+	__flush_wback_region((void *)start, end - start);
+}
+
+/*
+ * Writeback&Invalidate the D-cache of the page
+ */
+static void __flush_dcache_page(unsigned long phys)
+{
+	unsigned long ways, waysize, addrstart;
+	unsigned long flags;
+
+	phys |= SH_CACHE_VALID;
+
+	/*
+	 * Here, phys is the physical address of the page. We check all the
+	 * tags in the cache for those with the same page number as this page
+	 * (by masking off the lowest 2 bits of the 19-bit tag; these bits are
+	 * derived from the offset within in the 4k page). Matching valid
+	 * entries are invalidated.
+	 *
+	 * Since 2 bits of the cache index are derived from the virtual page
+	 * number, knowing this would reduce the number of cache entries to be
+	 * searched by a factor of 4. However this function exists to deal with
+	 * potential cache aliasing, therefore the optimisation is probably not
+	 * possible.
+	 */
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	ways = current_cpu_data.dcache.ways;
+	waysize = current_cpu_data.dcache.sets;
+	waysize <<= current_cpu_data.dcache.entry_shift;
+
+	addrstart = CACHE_OC_ADDRESS_ARRAY;
+
+	do {
+		unsigned long addr;
+
+		for (addr = addrstart;
+		     addr < addrstart + waysize;
+		     addr += current_cpu_data.dcache.linesz) {
+			unsigned long data;
+
+			data = __raw_readl(addr) & (0x1ffffC00 | SH_CACHE_VALID);
+		        if (data == phys) {
+				data &= ~(SH_CACHE_VALID | SH_CACHE_UPDATED);
+				__raw_writel(data, addr);
+			}
+		}
+
+		addrstart += current_cpu_data.dcache.way_incr;
+	} while (--ways);
+
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+/*
+ * Write back & invalidate the D-cache of the page.
+ * (To avoid "alias" issues)
+ */
+static void sh7705_flush_dcache_page(void *arg)
+{
+	struct page *page = arg;
+	struct address_space *mapping = page_mapping(page);
+
+	if (mapping && !mapping_mapped(mapping))
+		clear_bit(PG_dcache_clean, &page->flags);
+	else
+		__flush_dcache_page(__pa(page_address(page)));
+}
+
+static void sh7705_flush_cache_all(void *args)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	cache_wback_all();
+	back_to_cached();
+	local_irq_restore(flags);
+}
+
+/*
+ * Write back and invalidate I/D-caches for the page.
+ *
+ * ADDRESS: Virtual Address (U0 address)
+ */
+static void sh7705_flush_cache_page(void *args)
+{
+	struct flusher_data *data = args;
+	unsigned long pfn = data->addr2;
+
+	__flush_dcache_page(pfn << PAGE_SHIFT);
+}
+
+/*
+ * This is called when a page-cache page is about to be mapped into a
+ * user process' address space.  It offers an opportunity for a
+ * port to ensure d-cache/i-cache coherency if necessary.
+ *
+ * Not entirely sure why this is necessary on SH3 with 32K cache but
+ * without it we get occasional "Memory fault" when loading a program.
+ */
+static void sh7705_flush_icache_page(void *page)
+{
+	__flush_purge_region(page_address(page), PAGE_SIZE);
+}
+
+void __init sh7705_cache_init(void)
+{
+	local_flush_icache_range	= sh7705_flush_icache_range;
+	local_flush_dcache_page		= sh7705_flush_dcache_page;
+	local_flush_cache_all		= sh7705_flush_cache_all;
+	local_flush_cache_mm		= sh7705_flush_cache_all;
+	local_flush_cache_dup_mm	= sh7705_flush_cache_all;
+	local_flush_cache_range		= sh7705_flush_cache_all;
+	local_flush_cache_page		= sh7705_flush_cache_page;
+	local_flush_icache_page		= sh7705_flush_icache_page;
+}
diff --git a/arch/sh/mm/cache-shx3.c b/arch/sh/mm/cache-shx3.c
new file mode 100644
index 0000000..24c58b7
--- /dev/null
+++ b/arch/sh/mm/cache-shx3.c
@@ -0,0 +1,44 @@
+/*
+ * arch/sh/mm/cache-shx3.c - SH-X3 optimized cache ops
+ *
+ * Copyright (C) 2010  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/io.h>
+#include <asm/cache.h>
+
+#define CCR_CACHE_SNM	0x40000		/* Hardware-assisted synonym avoidance */
+#define CCR_CACHE_IBE	0x1000000	/* ICBI broadcast */
+
+void __init shx3_cache_init(void)
+{
+	unsigned int ccr;
+
+	ccr = __raw_readl(SH_CCR);
+
+	/*
+	 * If we've got cache aliases, resolve them in hardware.
+	 */
+	if (boot_cpu_data.dcache.n_aliases || boot_cpu_data.icache.n_aliases) {
+		ccr |= CCR_CACHE_SNM;
+
+		boot_cpu_data.icache.n_aliases = 0;
+		boot_cpu_data.dcache.n_aliases = 0;
+
+		pr_info("Enabling hardware synonym avoidance\n");
+	}
+
+#ifdef CONFIG_SMP
+	/*
+	 * Broadcast I-cache block invalidations by default.
+	 */
+	ccr |= CCR_CACHE_IBE;
+#endif
+
+	writel_uncached(ccr, SH_CCR);
+}
diff --git a/arch/sh/mm/cache.c b/arch/sh/mm/cache.c
new file mode 100644
index 0000000..f770e39
--- /dev/null
+++ b/arch/sh/mm/cache.c
@@ -0,0 +1,356 @@
+/*
+ * arch/sh/mm/cache.c
+ *
+ * Copyright (C) 1999, 2000, 2002  Niibe Yutaka
+ * Copyright (C) 2002 - 2010  Paul Mundt
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/mutex.h>
+#include <linux/fs.h>
+#include <linux/smp.h>
+#include <linux/highmem.h>
+#include <linux/module.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+void (*local_flush_cache_all)(void *args) = cache_noop;
+void (*local_flush_cache_mm)(void *args) = cache_noop;
+void (*local_flush_cache_dup_mm)(void *args) = cache_noop;
+void (*local_flush_cache_page)(void *args) = cache_noop;
+void (*local_flush_cache_range)(void *args) = cache_noop;
+void (*local_flush_dcache_page)(void *args) = cache_noop;
+void (*local_flush_icache_range)(void *args) = cache_noop;
+void (*local_flush_icache_page)(void *args) = cache_noop;
+void (*local_flush_cache_sigtramp)(void *args) = cache_noop;
+
+void (*__flush_wback_region)(void *start, int size);
+EXPORT_SYMBOL(__flush_wback_region);
+void (*__flush_purge_region)(void *start, int size);
+EXPORT_SYMBOL(__flush_purge_region);
+void (*__flush_invalidate_region)(void *start, int size);
+EXPORT_SYMBOL(__flush_invalidate_region);
+
+static inline void noop__flush_region(void *start, int size)
+{
+}
+
+static inline void cacheop_on_each_cpu(void (*func) (void *info), void *info,
+                                   int wait)
+{
+	preempt_disable();
+
+	/*
+	 * It's possible that this gets called early on when IRQs are
+	 * still disabled due to ioremapping by the boot CPU, so don't
+	 * even attempt IPIs unless there are other CPUs online.
+	 */
+	if (num_online_cpus() > 1)
+		smp_call_function(func, info, wait);
+
+	func(info);
+
+	preempt_enable();
+}
+
+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 (boot_cpu_data.dcache.n_aliases && page_mapped(page) &&
+	    test_bit(PG_dcache_clean, &page->flags)) {
+		void *vto = kmap_coherent(page, vaddr) + (vaddr & ~PAGE_MASK);
+		memcpy(vto, src, len);
+		kunmap_coherent(vto);
+	} else {
+		memcpy(dst, src, len);
+		if (boot_cpu_data.dcache.n_aliases)
+			clear_bit(PG_dcache_clean, &page->flags);
+	}
+
+	if (vma->vm_flags & VM_EXEC)
+		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 (boot_cpu_data.dcache.n_aliases && page_mapped(page) &&
+	    test_bit(PG_dcache_clean, &page->flags)) {
+		void *vfrom = kmap_coherent(page, vaddr) + (vaddr & ~PAGE_MASK);
+		memcpy(dst, vfrom, len);
+		kunmap_coherent(vfrom);
+	} else {
+		memcpy(dst, src, len);
+		if (boot_cpu_data.dcache.n_aliases)
+			clear_bit(PG_dcache_clean, &page->flags);
+	}
+}
+
+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 (boot_cpu_data.dcache.n_aliases && page_mapped(from) &&
+	    test_bit(PG_dcache_clean, &from->flags)) {
+		vfrom = kmap_coherent(from, vaddr);
+		copy_page(vto, vfrom);
+		kunmap_coherent(vfrom);
+	} else {
+		vfrom = kmap_atomic(from);
+		copy_page(vto, vfrom);
+		kunmap_atomic(vfrom);
+	}
+
+	if (pages_do_alias((unsigned long)vto, vaddr & PAGE_MASK) ||
+	    (vma->vm_flags & VM_EXEC))
+		__flush_purge_region(vto, PAGE_SIZE);
+
+	kunmap_atomic(vto);
+	/* Make sure this page is cleared on other CPU's too before using it */
+	smp_wmb();
+}
+EXPORT_SYMBOL(copy_user_highpage);
+
+void clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	void *kaddr = kmap_atomic(page);
+
+	clear_page(kaddr);
+
+	if (pages_do_alias((unsigned long)kaddr, vaddr & PAGE_MASK))
+		__flush_purge_region(kaddr, PAGE_SIZE);
+
+	kunmap_atomic(kaddr);
+}
+EXPORT_SYMBOL(clear_user_highpage);
+
+void __update_cache(struct vm_area_struct *vma,
+		    unsigned long address, pte_t pte)
+{
+	struct page *page;
+	unsigned long pfn = pte_pfn(pte);
+
+	if (!boot_cpu_data.dcache.n_aliases)
+		return;
+
+	page = pfn_to_page(pfn);
+	if (pfn_valid(pfn)) {
+		int dirty = !test_and_set_bit(PG_dcache_clean, &page->flags);
+		if (dirty)
+			__flush_purge_region(page_address(page), PAGE_SIZE);
+	}
+}
+
+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 (boot_cpu_data.dcache.n_aliases && page_mapped(page) &&
+		    test_bit(PG_dcache_clean, &page->flags)) {
+			void *kaddr;
+
+			kaddr = kmap_coherent(page, vmaddr);
+			/* XXX.. For now kunmap_coherent() does a purge */
+			/* __flush_purge_region((void *)kaddr, PAGE_SIZE); */
+			kunmap_coherent(kaddr);
+		} else
+			__flush_purge_region((void *)addr, PAGE_SIZE);
+	}
+}
+
+void flush_cache_all(void)
+{
+	cacheop_on_each_cpu(local_flush_cache_all, NULL, 1);
+}
+EXPORT_SYMBOL(flush_cache_all);
+
+void flush_cache_mm(struct mm_struct *mm)
+{
+	if (boot_cpu_data.dcache.n_aliases == 0)
+		return;
+
+	cacheop_on_each_cpu(local_flush_cache_mm, mm, 1);
+}
+
+void flush_cache_dup_mm(struct mm_struct *mm)
+{
+	if (boot_cpu_data.dcache.n_aliases == 0)
+		return;
+
+	cacheop_on_each_cpu(local_flush_cache_dup_mm, mm, 1);
+}
+
+void flush_cache_page(struct vm_area_struct *vma, unsigned long addr,
+		      unsigned long pfn)
+{
+	struct flusher_data data;
+
+	data.vma = vma;
+	data.addr1 = addr;
+	data.addr2 = pfn;
+
+	cacheop_on_each_cpu(local_flush_cache_page, (void *)&data, 1);
+}
+
+void flush_cache_range(struct vm_area_struct *vma, unsigned long start,
+		       unsigned long end)
+{
+	struct flusher_data data;
+
+	data.vma = vma;
+	data.addr1 = start;
+	data.addr2 = end;
+
+	cacheop_on_each_cpu(local_flush_cache_range, (void *)&data, 1);
+}
+EXPORT_SYMBOL(flush_cache_range);
+
+void flush_dcache_page(struct page *page)
+{
+	cacheop_on_each_cpu(local_flush_dcache_page, page, 1);
+}
+EXPORT_SYMBOL(flush_dcache_page);
+
+void flush_icache_range(unsigned long start, unsigned long end)
+{
+	struct flusher_data data;
+
+	data.vma = NULL;
+	data.addr1 = start;
+	data.addr2 = end;
+
+	cacheop_on_each_cpu(local_flush_icache_range, (void *)&data, 1);
+}
+EXPORT_SYMBOL(flush_icache_range);
+
+void flush_icache_page(struct vm_area_struct *vma, struct page *page)
+{
+	/* Nothing uses the VMA, so just pass the struct page along */
+	cacheop_on_each_cpu(local_flush_icache_page, page, 1);
+}
+
+void flush_cache_sigtramp(unsigned long address)
+{
+	cacheop_on_each_cpu(local_flush_cache_sigtramp, (void *)address, 1);
+}
+
+static void compute_alias(struct cache_info *c)
+{
+	c->alias_mask = ((c->sets - 1) << c->entry_shift) & ~(PAGE_SIZE - 1);
+	c->n_aliases = c->alias_mask ? (c->alias_mask >> PAGE_SHIFT) + 1 : 0;
+}
+
+static void __init emit_cache_params(void)
+{
+	printk(KERN_NOTICE "I-cache : n_ways=%d n_sets=%d way_incr=%d\n",
+		boot_cpu_data.icache.ways,
+		boot_cpu_data.icache.sets,
+		boot_cpu_data.icache.way_incr);
+	printk(KERN_NOTICE "I-cache : entry_mask=0x%08x alias_mask=0x%08x n_aliases=%d\n",
+		boot_cpu_data.icache.entry_mask,
+		boot_cpu_data.icache.alias_mask,
+		boot_cpu_data.icache.n_aliases);
+	printk(KERN_NOTICE "D-cache : n_ways=%d n_sets=%d way_incr=%d\n",
+		boot_cpu_data.dcache.ways,
+		boot_cpu_data.dcache.sets,
+		boot_cpu_data.dcache.way_incr);
+	printk(KERN_NOTICE "D-cache : entry_mask=0x%08x alias_mask=0x%08x n_aliases=%d\n",
+		boot_cpu_data.dcache.entry_mask,
+		boot_cpu_data.dcache.alias_mask,
+		boot_cpu_data.dcache.n_aliases);
+
+	/*
+	 * Emit Secondary Cache parameters if the CPU has a probed L2.
+	 */
+	if (boot_cpu_data.flags & CPU_HAS_L2_CACHE) {
+		printk(KERN_NOTICE "S-cache : n_ways=%d n_sets=%d way_incr=%d\n",
+			boot_cpu_data.scache.ways,
+			boot_cpu_data.scache.sets,
+			boot_cpu_data.scache.way_incr);
+		printk(KERN_NOTICE "S-cache : entry_mask=0x%08x alias_mask=0x%08x n_aliases=%d\n",
+			boot_cpu_data.scache.entry_mask,
+			boot_cpu_data.scache.alias_mask,
+			boot_cpu_data.scache.n_aliases);
+	}
+}
+
+void __init cpu_cache_init(void)
+{
+	unsigned int cache_disabled = 0;
+
+#ifdef SH_CCR
+	cache_disabled = !(__raw_readl(SH_CCR) & CCR_CACHE_ENABLE);
+#endif
+
+	compute_alias(&boot_cpu_data.icache);
+	compute_alias(&boot_cpu_data.dcache);
+	compute_alias(&boot_cpu_data.scache);
+
+	__flush_wback_region		= noop__flush_region;
+	__flush_purge_region		= noop__flush_region;
+	__flush_invalidate_region	= noop__flush_region;
+
+	/*
+	 * No flushing is necessary in the disabled cache case so we can
+	 * just keep the noop functions in local_flush_..() and __flush_..()
+	 */
+	if (unlikely(cache_disabled))
+		goto skip;
+
+	if (boot_cpu_data.family == CPU_FAMILY_SH2) {
+		extern void __weak sh2_cache_init(void);
+
+		sh2_cache_init();
+	}
+
+	if (boot_cpu_data.family == CPU_FAMILY_SH2A) {
+		extern void __weak sh2a_cache_init(void);
+
+		sh2a_cache_init();
+	}
+
+	if (boot_cpu_data.family == CPU_FAMILY_SH3) {
+		extern void __weak sh3_cache_init(void);
+
+		sh3_cache_init();
+
+		if ((boot_cpu_data.type == CPU_SH7705) &&
+		    (boot_cpu_data.dcache.sets == 512)) {
+			extern void __weak sh7705_cache_init(void);
+
+			sh7705_cache_init();
+		}
+	}
+
+	if ((boot_cpu_data.family == CPU_FAMILY_SH4) ||
+	    (boot_cpu_data.family == CPU_FAMILY_SH4A) ||
+	    (boot_cpu_data.family == CPU_FAMILY_SH4AL_DSP)) {
+		extern void __weak sh4_cache_init(void);
+
+		sh4_cache_init();
+
+		if ((boot_cpu_data.type == CPU_SH7786) ||
+		    (boot_cpu_data.type == CPU_SHX3)) {
+			extern void __weak shx3_cache_init(void);
+
+			shx3_cache_init();
+		}
+	}
+
+	if (boot_cpu_data.family == CPU_FAMILY_SH5) {
+		extern void __weak sh5_cache_init(void);
+
+		sh5_cache_init();
+	}
+
+skip:
+	emit_cache_params();
+}
diff --git a/arch/sh/mm/consistent.c b/arch/sh/mm/consistent.c
new file mode 100644
index 0000000..b81d9db
--- /dev/null
+++ b/arch/sh/mm/consistent.c
@@ -0,0 +1,159 @@
+/*
+ * arch/sh/mm/consistent.c
+ *
+ * Copyright (C) 2004 - 2007  Paul Mundt
+ *
+ * Declared coherent memory functions based on arch/x86/kernel/pci-dma_32.c
+ *
+ * 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.
+ */
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/platform_device.h>
+#include <linux/dma-mapping.h>
+#include <linux/dma-debug.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/gfp.h>
+#include <asm/cacheflush.h>
+#include <asm/addrspace.h>
+
+#define PREALLOC_DMA_DEBUG_ENTRIES	4096
+
+struct dma_map_ops *dma_ops;
+EXPORT_SYMBOL(dma_ops);
+
+static int __init dma_init(void)
+{
+	dma_debug_init(PREALLOC_DMA_DEBUG_ENTRIES);
+	return 0;
+}
+fs_initcall(dma_init);
+
+void *dma_generic_alloc_coherent(struct device *dev, size_t size,
+				 dma_addr_t *dma_handle, gfp_t gfp,
+				 struct dma_attrs *attrs)
+{
+	void *ret, *ret_nocache;
+	int order = get_order(size);
+
+	gfp |= __GFP_ZERO;
+
+	ret = (void *)__get_free_pages(gfp, order);
+	if (!ret)
+		return NULL;
+
+	/*
+	 * Pages from the page allocator may have data present in
+	 * cache. So flush the cache before using uncached memory.
+	 */
+	dma_cache_sync(dev, ret, size, DMA_BIDIRECTIONAL);
+
+	ret_nocache = (void __force *)ioremap_nocache(virt_to_phys(ret), size);
+	if (!ret_nocache) {
+		free_pages((unsigned long)ret, order);
+		return NULL;
+	}
+
+	split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
+
+	*dma_handle = virt_to_phys(ret);
+
+	return ret_nocache;
+}
+
+void dma_generic_free_coherent(struct device *dev, size_t size,
+			       void *vaddr, dma_addr_t dma_handle,
+			       struct dma_attrs *attrs)
+{
+	int order = get_order(size);
+	unsigned long pfn = dma_handle >> PAGE_SHIFT;
+	int k;
+
+	for (k = 0; k < (1 << order); k++)
+		__free_pages(pfn_to_page(pfn + k), 0);
+
+	iounmap(vaddr);
+}
+
+void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+		    enum dma_data_direction direction)
+{
+	void *addr;
+
+	addr = __in_29bit_mode() ?
+	       (void *)CAC_ADDR((unsigned long)vaddr) : vaddr;
+
+	switch (direction) {
+	case DMA_FROM_DEVICE:		/* invalidate only */
+		__flush_invalidate_region(addr, size);
+		break;
+	case DMA_TO_DEVICE:		/* writeback only */
+		__flush_wback_region(addr, size);
+		break;
+	case DMA_BIDIRECTIONAL:		/* writeback and invalidate */
+		__flush_purge_region(addr, size);
+		break;
+	default:
+		BUG();
+	}
+}
+EXPORT_SYMBOL(dma_cache_sync);
+
+static int __init memchunk_setup(char *str)
+{
+	return 1; /* accept anything that begins with "memchunk." */
+}
+__setup("memchunk.", memchunk_setup);
+
+static void __init memchunk_cmdline_override(char *name, unsigned long *sizep)
+{
+	char *p = boot_command_line;
+	int k = strlen(name);
+
+	while ((p = strstr(p, "memchunk."))) {
+		p += 9; /* strlen("memchunk.") */
+		if (!strncmp(name, p, k) && p[k] == '=') {
+			p += k + 1;
+			*sizep = memparse(p, NULL);
+			pr_info("%s: forcing memory chunk size to 0x%08lx\n",
+				name, *sizep);
+			break;
+		}
+	}
+}
+
+int __init platform_resource_setup_memory(struct platform_device *pdev,
+					  char *name, unsigned long memsize)
+{
+	struct resource *r;
+	dma_addr_t dma_handle;
+	void *buf;
+
+	r = pdev->resource + pdev->num_resources - 1;
+	if (r->flags) {
+		pr_warning("%s: unable to find empty space for resource\n",
+			name);
+		return -EINVAL;
+	}
+
+	memchunk_cmdline_override(name, &memsize);
+	if (!memsize)
+		return 0;
+
+	buf = dma_alloc_coherent(NULL, memsize, &dma_handle, GFP_KERNEL);
+	if (!buf) {
+		pr_warning("%s: unable to allocate memory\n", name);
+		return -ENOMEM;
+	}
+
+	memset(buf, 0, memsize);
+
+	r->flags = IORESOURCE_MEM;
+	r->start = dma_handle;
+	r->end = r->start + memsize - 1;
+	r->name = name;
+	return 0;
+}
diff --git a/arch/sh/mm/extable_32.c b/arch/sh/mm/extable_32.c
new file mode 100644
index 0000000..c1cf446
--- /dev/null
+++ b/arch/sh/mm/extable_32.c
@@ -0,0 +1,21 @@
+/*
+ * linux/arch/sh/mm/extable.c
+ *  Taken from:
+ *   linux/arch/i386/mm/extable.c
+ */
+
+#include <linux/module.h>
+#include <asm/uaccess.h>
+
+int fixup_exception(struct pt_regs *regs)
+{
+	const struct exception_table_entry *fixup;
+
+	fixup = search_exception_tables(regs->pc);
+	if (fixup) {
+		regs->pc = fixup->fixup;
+		return 1;
+	}
+
+	return 0;
+}
diff --git a/arch/sh/mm/extable_64.c b/arch/sh/mm/extable_64.c
new file mode 100644
index 0000000..f054996
--- /dev/null
+++ b/arch/sh/mm/extable_64.c
@@ -0,0 +1,82 @@
+/*
+ * arch/sh/mm/extable_64.c
+ *
+ * Copyright (C) 2003 Richard Curnow
+ * Copyright (C) 2003, 2004  Paul Mundt
+ *
+ * Cloned from the 2.5 SH version..
+ *
+ * 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.
+ */
+#include <linux/rwsem.h>
+#include <linux/module.h>
+#include <asm/uaccess.h>
+
+extern unsigned long copy_user_memcpy, copy_user_memcpy_end;
+extern void __copy_user_fixup(void);
+
+static const struct exception_table_entry __copy_user_fixup_ex = {
+	.fixup = (unsigned long)&__copy_user_fixup,
+};
+
+/*
+ * Some functions that may trap due to a bad user-mode address have too
+ * many loads and stores in them to make it at all practical to label
+ * each one and put them all in the main exception table.
+ *
+ * In particular, the fast memcpy routine is like this.  It's fix-up is
+ * just to fall back to a slow byte-at-a-time copy, which is handled the
+ * conventional way.  So it's functionally OK to just handle any trap
+ * occurring in the fast memcpy with that fixup.
+ */
+static const struct exception_table_entry *check_exception_ranges(unsigned long addr)
+{
+	if ((addr >= (unsigned long)&copy_user_memcpy) &&
+	    (addr <= (unsigned long)&copy_user_memcpy_end))
+		return &__copy_user_fixup_ex;
+
+	return NULL;
+}
+
+/* Simple binary search */
+const struct exception_table_entry *
+search_extable(const struct exception_table_entry *first,
+		 const struct exception_table_entry *last,
+		 unsigned long value)
+{
+	const struct exception_table_entry *mid;
+
+	mid = check_exception_ranges(value);
+	if (mid)
+		return mid;
+
+        while (first <= last) {
+		long diff;
+
+		mid = (last - first) / 2 + first;
+		diff = mid->insn - value;
+                if (diff == 0)
+                        return mid;
+                else if (diff < 0)
+                        first = mid+1;
+                else
+                        last = mid-1;
+        }
+
+        return NULL;
+}
+
+int fixup_exception(struct pt_regs *regs)
+{
+	const struct exception_table_entry *fixup;
+
+	fixup = search_exception_tables(regs->pc);
+	if (fixup) {
+		regs->pc = fixup->fixup;
+		return 1;
+	}
+
+	return 0;
+}
diff --git a/arch/sh/mm/fault.c b/arch/sh/mm/fault.c
new file mode 100644
index 0000000..79d8276
--- /dev/null
+++ b/arch/sh/mm/fault.c
@@ -0,0 +1,520 @@
+/*
+ * Page fault handler for SH with an MMU.
+ *
+ *  Copyright (C) 1999  Niibe Yutaka
+ *  Copyright (C) 2003 - 2012  Paul Mundt
+ *
+ *  Based on linux/arch/i386/mm/fault.c:
+ *   Copyright (C) 1995  Linus Torvalds
+ *
+ * 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.
+ */
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/hardirq.h>
+#include <linux/kprobes.h>
+#include <linux/perf_event.h>
+#include <linux/kdebug.h>
+#include <linux/uaccess.h>
+#include <asm/io_trapped.h>
+#include <asm/mmu_context.h>
+#include <asm/tlbflush.h>
+#include <asm/traps.h>
+
+static inline int notify_page_fault(struct pt_regs *regs, int trap)
+{
+	int ret = 0;
+
+	if (kprobes_built_in() && !user_mode(regs)) {
+		preempt_disable();
+		if (kprobe_running() && kprobe_fault_handler(regs, trap))
+			ret = 1;
+		preempt_enable();
+	}
+
+	return ret;
+}
+
+static void
+force_sig_info_fault(int si_signo, int si_code, unsigned long address,
+		     struct task_struct *tsk)
+{
+	siginfo_t info;
+
+	info.si_signo	= si_signo;
+	info.si_errno	= 0;
+	info.si_code	= si_code;
+	info.si_addr	= (void __user *)address;
+
+	force_sig_info(si_signo, &info, tsk);
+}
+
+/*
+ * This is useful to dump out the page tables associated with
+ * 'addr' in mm 'mm'.
+ */
+static void show_pte(struct mm_struct *mm, unsigned long addr)
+{
+	pgd_t *pgd;
+
+	if (mm) {
+		pgd = mm->pgd;
+	} else {
+		pgd = get_TTB();
+
+		if (unlikely(!pgd))
+			pgd = swapper_pg_dir;
+	}
+
+	printk(KERN_ALERT "pgd = %p\n", pgd);
+	pgd += pgd_index(addr);
+	printk(KERN_ALERT "[%08lx] *pgd=%0*Lx", addr,
+	       (u32)(sizeof(*pgd) * 2), (u64)pgd_val(*pgd));
+
+	do {
+		pud_t *pud;
+		pmd_t *pmd;
+		pte_t *pte;
+
+		if (pgd_none(*pgd))
+			break;
+
+		if (pgd_bad(*pgd)) {
+			printk("(bad)");
+			break;
+		}
+
+		pud = pud_offset(pgd, addr);
+		if (PTRS_PER_PUD != 1)
+			printk(", *pud=%0*Lx", (u32)(sizeof(*pud) * 2),
+			       (u64)pud_val(*pud));
+
+		if (pud_none(*pud))
+			break;
+
+		if (pud_bad(*pud)) {
+			printk("(bad)");
+			break;
+		}
+
+		pmd = pmd_offset(pud, addr);
+		if (PTRS_PER_PMD != 1)
+			printk(", *pmd=%0*Lx", (u32)(sizeof(*pmd) * 2),
+			       (u64)pmd_val(*pmd));
+
+		if (pmd_none(*pmd))
+			break;
+
+		if (pmd_bad(*pmd)) {
+			printk("(bad)");
+			break;
+		}
+
+		/* We must not map this if we have highmem enabled */
+		if (PageHighMem(pfn_to_page(pmd_val(*pmd) >> PAGE_SHIFT)))
+			break;
+
+		pte = pte_offset_kernel(pmd, addr);
+		printk(", *pte=%0*Lx", (u32)(sizeof(*pte) * 2),
+		       (u64)pte_val(*pte));
+	} while (0);
+
+	printk("\n");
+}
+
+static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
+{
+	unsigned index = pgd_index(address);
+	pgd_t *pgd_k;
+	pud_t *pud, *pud_k;
+	pmd_t *pmd, *pmd_k;
+
+	pgd += index;
+	pgd_k = init_mm.pgd + index;
+
+	if (!pgd_present(*pgd_k))
+		return NULL;
+
+	pud = pud_offset(pgd, address);
+	pud_k = pud_offset(pgd_k, address);
+	if (!pud_present(*pud_k))
+		return NULL;
+
+	if (!pud_present(*pud))
+	    set_pud(pud, *pud_k);
+
+	pmd = pmd_offset(pud, address);
+	pmd_k = pmd_offset(pud_k, address);
+	if (!pmd_present(*pmd_k))
+		return NULL;
+
+	if (!pmd_present(*pmd))
+		set_pmd(pmd, *pmd_k);
+	else {
+		/*
+		 * The page tables are fully synchronised so there must
+		 * be another reason for the fault. Return NULL here to
+		 * signal that we have not taken care of the fault.
+		 */
+		BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));
+		return NULL;
+	}
+
+	return pmd_k;
+}
+
+#ifdef CONFIG_SH_STORE_QUEUES
+#define __FAULT_ADDR_LIMIT	P3_ADDR_MAX
+#else
+#define __FAULT_ADDR_LIMIT	VMALLOC_END
+#endif
+
+/*
+ * Handle a fault on the vmalloc or module mapping area
+ */
+static noinline int vmalloc_fault(unsigned long address)
+{
+	pgd_t *pgd_k;
+	pmd_t *pmd_k;
+	pte_t *pte_k;
+
+	/* Make sure we are in vmalloc/module/P3 area: */
+	if (!(address >= VMALLOC_START && address < __FAULT_ADDR_LIMIT))
+		return -1;
+
+	/*
+	 * Synchronize this task's top level page-table
+	 * with the 'reference' page table.
+	 *
+	 * Do _not_ use "current" here. We might be inside
+	 * an interrupt in the middle of a task switch..
+	 */
+	pgd_k = get_TTB();
+	pmd_k = vmalloc_sync_one(pgd_k, address);
+	if (!pmd_k)
+		return -1;
+
+	pte_k = pte_offset_kernel(pmd_k, address);
+	if (!pte_present(*pte_k))
+		return -1;
+
+	return 0;
+}
+
+static void
+show_fault_oops(struct pt_regs *regs, unsigned long address)
+{
+	if (!oops_may_print())
+		return;
+
+	printk(KERN_ALERT "BUG: unable to handle kernel ");
+	if (address < PAGE_SIZE)
+		printk(KERN_CONT "NULL pointer dereference");
+	else
+		printk(KERN_CONT "paging request");
+
+	printk(KERN_CONT " at %08lx\n", address);
+	printk(KERN_ALERT "PC:");
+	printk_address(regs->pc, 1);
+
+	show_pte(NULL, address);
+}
+
+static noinline void
+no_context(struct pt_regs *regs, unsigned long error_code,
+	   unsigned long address)
+{
+	/* Are we prepared to handle this kernel fault?  */
+	if (fixup_exception(regs))
+		return;
+
+	if (handle_trapped_io(regs, address))
+		return;
+
+	/*
+	 * Oops. The kernel tried to access some bad page. We'll have to
+	 * terminate things with extreme prejudice.
+	 */
+	bust_spinlocks(1);
+
+	show_fault_oops(regs, address);
+
+	die("Oops", regs, error_code);
+	bust_spinlocks(0);
+	do_exit(SIGKILL);
+}
+
+static void
+__bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
+		       unsigned long address, int si_code)
+{
+	struct task_struct *tsk = current;
+
+	/* User mode accesses just cause a SIGSEGV */
+	if (user_mode(regs)) {
+		/*
+		 * It's possible to have interrupts off here:
+		 */
+		local_irq_enable();
+
+		force_sig_info_fault(SIGSEGV, si_code, address, tsk);
+
+		return;
+	}
+
+	no_context(regs, error_code, address);
+}
+
+static noinline void
+bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
+		     unsigned long address)
+{
+	__bad_area_nosemaphore(regs, error_code, address, SEGV_MAPERR);
+}
+
+static void
+__bad_area(struct pt_regs *regs, unsigned long error_code,
+	   unsigned long address, int si_code)
+{
+	struct mm_struct *mm = current->mm;
+
+	/*
+	 * Something tried to access memory that isn't in our memory map..
+	 * Fix it, but check if it's kernel or user first..
+	 */
+	up_read(&mm->mmap_sem);
+
+	__bad_area_nosemaphore(regs, error_code, address, si_code);
+}
+
+static noinline void
+bad_area(struct pt_regs *regs, unsigned long error_code, unsigned long address)
+{
+	__bad_area(regs, error_code, address, SEGV_MAPERR);
+}
+
+static noinline void
+bad_area_access_error(struct pt_regs *regs, unsigned long error_code,
+		      unsigned long address)
+{
+	__bad_area(regs, error_code, address, SEGV_ACCERR);
+}
+
+static void
+do_sigbus(struct pt_regs *regs, unsigned long error_code, unsigned long address)
+{
+	struct task_struct *tsk = current;
+	struct mm_struct *mm = tsk->mm;
+
+	up_read(&mm->mmap_sem);
+
+	/* Kernel mode? Handle exceptions or die: */
+	if (!user_mode(regs))
+		no_context(regs, error_code, address);
+
+	force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
+}
+
+static noinline int
+mm_fault_error(struct pt_regs *regs, unsigned long error_code,
+	       unsigned long address, unsigned int fault)
+{
+	/*
+	 * Pagefault was interrupted by SIGKILL. We have no reason to
+	 * continue pagefault.
+	 */
+	if (fatal_signal_pending(current)) {
+		if (!(fault & VM_FAULT_RETRY))
+			up_read(&current->mm->mmap_sem);
+		if (!user_mode(regs))
+			no_context(regs, error_code, address);
+		return 1;
+	}
+
+	if (!(fault & VM_FAULT_ERROR))
+		return 0;
+
+	if (fault & VM_FAULT_OOM) {
+		/* Kernel mode? Handle exceptions or die: */
+		if (!user_mode(regs)) {
+			up_read(&current->mm->mmap_sem);
+			no_context(regs, error_code, address);
+			return 1;
+		}
+		up_read(&current->mm->mmap_sem);
+
+		/*
+		 * 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):
+		 */
+		pagefault_out_of_memory();
+	} else {
+		if (fault & VM_FAULT_SIGBUS)
+			do_sigbus(regs, error_code, address);
+		else if (fault & VM_FAULT_SIGSEGV)
+			bad_area(regs, error_code, address);
+		else
+			BUG();
+	}
+
+	return 1;
+}
+
+static inline int access_error(int error_code, struct vm_area_struct *vma)
+{
+	if (error_code & FAULT_CODE_WRITE) {
+		/* write, present and write, not present: */
+		if (unlikely(!(vma->vm_flags & VM_WRITE)))
+			return 1;
+		return 0;
+	}
+
+	/* ITLB miss on NX page */
+	if (unlikely((error_code & FAULT_CODE_ITLB) &&
+		     !(vma->vm_flags & VM_EXEC)))
+		return 1;
+
+	/* read, not present: */
+	if (unlikely(!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))))
+		return 1;
+
+	return 0;
+}
+
+static int fault_in_kernel_space(unsigned long address)
+{
+	return address >= TASK_SIZE;
+}
+
+/*
+ * 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 error_code,
+					unsigned long address)
+{
+	unsigned long vec;
+	struct task_struct *tsk;
+	struct mm_struct *mm;
+	struct vm_area_struct * vma;
+	int fault;
+	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
+
+	tsk = current;
+	mm = tsk->mm;
+	vec = lookup_exception_vector();
+
+	/*
+	 * 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.
+	 */
+	if (unlikely(fault_in_kernel_space(address))) {
+		if (vmalloc_fault(address) >= 0)
+			return;
+		if (notify_page_fault(regs, vec))
+			return;
+
+		bad_area_nosemaphore(regs, error_code, address);
+		return;
+	}
+
+	if (unlikely(notify_page_fault(regs, vec)))
+		return;
+
+	/* Only enable interrupts if they were on before the fault */
+	if ((regs->sr & SR_IMASK) != SR_IMASK)
+		local_irq_enable();
+
+	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
+
+	/*
+	 * If we're in an interrupt, have no user context or are running
+	 * with pagefaults disabled then we must not take the fault:
+	 */
+	if (unlikely(faulthandler_disabled() || !mm)) {
+		bad_area_nosemaphore(regs, error_code, address);
+		return;
+	}
+
+retry:
+	down_read(&mm->mmap_sem);
+
+	vma = find_vma(mm, address);
+	if (unlikely(!vma)) {
+		bad_area(regs, error_code, address);
+		return;
+	}
+	if (likely(vma->vm_start <= address))
+		goto good_area;
+	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN))) {
+		bad_area(regs, error_code, address);
+		return;
+	}
+	if (unlikely(expand_stack(vma, address))) {
+		bad_area(regs, error_code, address);
+		return;
+	}
+
+	/*
+	 * Ok, we have a good vm_area for this memory access, so
+	 * we can handle it..
+	 */
+good_area:
+	if (unlikely(access_error(error_code, vma))) {
+		bad_area_access_error(regs, error_code, address);
+		return;
+	}
+
+	set_thread_fault_code(error_code);
+
+	if (user_mode(regs))
+		flags |= FAULT_FLAG_USER;
+	if (error_code & FAULT_CODE_WRITE)
+		flags |= FAULT_FLAG_WRITE;
+
+	/*
+	 * 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 (unlikely(fault & (VM_FAULT_RETRY | VM_FAULT_ERROR)))
+		if (mm_fault_error(regs, error_code, address, fault))
+			return;
+
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
+		if (fault & VM_FAULT_MAJOR) {
+			tsk->maj_flt++;
+			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1,
+				      regs, address);
+		} else {
+			tsk->min_flt++;
+			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1,
+				      regs, address);
+		}
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+			flags |= FAULT_FLAG_TRIED;
+
+			/*
+			 * 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);
+}
diff --git a/arch/sh/mm/flush-sh4.c b/arch/sh/mm/flush-sh4.c
new file mode 100644
index 0000000..0b85dd9
--- /dev/null
+++ b/arch/sh/mm/flush-sh4.c
@@ -0,0 +1,110 @@
+#include <linux/mm.h>
+#include <asm/mmu_context.h>
+#include <asm/cache_insns.h>
+#include <asm/cacheflush.h>
+#include <asm/traps.h>
+
+/*
+ * Write back the dirty D-caches, but not invalidate them.
+ *
+ * START: Virtual Address (U0, P1, or P3)
+ * SIZE: Size of the region.
+ */
+static void sh4__flush_wback_region(void *start, int size)
+{
+	reg_size_t aligned_start, v, cnt, end;
+
+	aligned_start = register_align(start);
+	v = aligned_start & ~(L1_CACHE_BYTES-1);
+	end = (aligned_start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+	cnt = (end - v) / L1_CACHE_BYTES;
+
+	while (cnt >= 8) {
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		cnt -= 8;
+	}
+
+	while (cnt) {
+		__ocbwb(v); v += L1_CACHE_BYTES;
+		cnt--;
+	}
+}
+
+/*
+ * Write back the dirty D-caches and invalidate them.
+ *
+ * START: Virtual Address (U0, P1, or P3)
+ * SIZE: Size of the region.
+ */
+static void sh4__flush_purge_region(void *start, int size)
+{
+	reg_size_t aligned_start, v, cnt, end;
+
+	aligned_start = register_align(start);
+	v = aligned_start & ~(L1_CACHE_BYTES-1);
+	end = (aligned_start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+	cnt = (end - v) / L1_CACHE_BYTES;
+
+	while (cnt >= 8) {
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
+		cnt -= 8;
+	}
+	while (cnt) {
+		__ocbp(v); v += L1_CACHE_BYTES;
+		cnt--;
+	}
+}
+
+/*
+ * No write back please
+ */
+static void sh4__flush_invalidate_region(void *start, int size)
+{
+	reg_size_t aligned_start, v, cnt, end;
+
+	aligned_start = register_align(start);
+	v = aligned_start & ~(L1_CACHE_BYTES-1);
+	end = (aligned_start + size + L1_CACHE_BYTES-1)
+		& ~(L1_CACHE_BYTES-1);
+	cnt = (end - v) / L1_CACHE_BYTES;
+
+	while (cnt >= 8) {
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
+		cnt -= 8;
+	}
+
+	while (cnt) {
+		__ocbi(v); v += L1_CACHE_BYTES;
+		cnt--;
+	}
+}
+
+void __init sh4__flush_region_init(void)
+{
+	__flush_wback_region		= sh4__flush_wback_region;
+	__flush_invalidate_region	= sh4__flush_invalidate_region;
+	__flush_purge_region		= sh4__flush_purge_region;
+}
diff --git a/arch/sh/mm/gup.c b/arch/sh/mm/gup.c
new file mode 100644
index 0000000..e7af6a6
--- /dev/null
+++ b/arch/sh/mm/gup.c
@@ -0,0 +1,273 @@
+/*
+ * Lockless get_user_pages_fast for SuperH
+ *
+ * Copyright (C) 2009 - 2010  Paul Mundt
+ *
+ * Cloned from the x86 and PowerPC versions, by:
+ *
+ *	Copyright (C) 2008 Nick Piggin
+ *	Copyright (C) 2008 Novell Inc.
+ */
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/vmstat.h>
+#include <linux/highmem.h>
+#include <asm/pgtable.h>
+
+static inline pte_t gup_get_pte(pte_t *ptep)
+{
+#ifndef CONFIG_X2TLB
+	return READ_ONCE(*ptep);
+#else
+	/*
+	 * With get_user_pages_fast, we walk down the pagetables without
+	 * taking any locks.  For this we would like to load the pointers
+	 * atomically, but that is not possible with 64-bit PTEs.  What
+	 * we do have is the guarantee that a pte will only either go
+	 * from not present to present, or present to not present or both
+	 * -- it will not switch to a completely different present page
+	 * without a TLB flush in between; something that we are blocking
+	 * by holding interrupts off.
+	 *
+	 * Setting ptes from not present to present goes:
+	 * ptep->pte_high = h;
+	 * smp_wmb();
+	 * ptep->pte_low = l;
+	 *
+	 * And present to not present goes:
+	 * ptep->pte_low = 0;
+	 * smp_wmb();
+	 * ptep->pte_high = 0;
+	 *
+	 * We must ensure here that the load of pte_low sees l iff pte_high
+	 * sees h. We load pte_high *after* loading pte_low, which ensures we
+	 * don't see an older value of pte_high.  *Then* we recheck pte_low,
+	 * which ensures that we haven't picked up a changed pte high. We might
+	 * have got rubbish values from pte_low and pte_high, but we are
+	 * guaranteed that pte_low will not have the present bit set *unless*
+	 * it is 'l'. And get_user_pages_fast only operates on present ptes, so
+	 * we're safe.
+	 *
+	 * gup_get_pte should not be used or copied outside gup.c without being
+	 * very careful -- it does not atomically load the pte or anything that
+	 * is likely to be useful for you.
+	 */
+	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;
+#endif
+}
+
+/*
+ * The performance critical leaf functions are made noinline otherwise gcc
+ * inlines everything into a single function which results in too much
+ * register pressure.
+ */
+static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
+		unsigned long end, int write, struct page **pages, int *nr)
+{
+	u64 mask, result;
+	pte_t *ptep;
+
+#ifdef CONFIG_X2TLB
+	result = _PAGE_PRESENT | _PAGE_EXT(_PAGE_EXT_KERN_READ | _PAGE_EXT_USER_READ);
+	if (write)
+		result |= _PAGE_EXT(_PAGE_EXT_KERN_WRITE | _PAGE_EXT_USER_WRITE);
+#elif defined(CONFIG_SUPERH64)
+	result = _PAGE_PRESENT | _PAGE_USER | _PAGE_READ;
+	if (write)
+		result |= _PAGE_WRITE;
+#else
+	result = _PAGE_PRESENT | _PAGE_USER;
+	if (write)
+		result |= _PAGE_RW;
+#endif
+
+	mask = result | _PAGE_SPECIAL;
+
+	ptep = pte_offset_map(&pmd, addr);
+	do {
+		pte_t pte = gup_get_pte(ptep);
+		struct page *page;
+
+		if ((pte_val(pte) & mask) != result) {
+			pte_unmap(ptep);
+			return 0;
+		}
+		VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
+		page = pte_page(pte);
+		get_page(page);
+		__flush_anon_page(page, addr);
+		flush_dcache_page(page);
+		pages[*nr] = page;
+		(*nr)++;
+
+	} while (ptep++, addr += PAGE_SIZE, addr != end);
+	pte_unmap(ptep - 1);
+
+	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);
+		if (pmd_none(pmd))
+			return 0;
+		if (!gup_pte_range(pmd, addr, next, write, pages, nr))
+			return 0;
+	} while (pmdp++, addr = next, addr != end);
+
+	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 (!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;
+
+	/*
+	 * This doesn't prevent pagetable teardown, but does prevent
+	 * the pagetables and pages from being freed.
+	 */
+	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 nr = 0;
+
+	start &= PAGE_MASK;
+	addr = start;
+	len = (unsigned long) nr_pages << PAGE_SHIFT;
+
+	end = start + len;
+	if (end < start)
+		goto slow_irqon;
+
+	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;
+
+	{
+		int ret;
+
+slow:
+		local_irq_enable();
+slow_irqon:
+		/* Try to get the remaining pages with get_user_pages */
+		start += nr << PAGE_SHIFT;
+		pages += nr;
+
+		ret = get_user_pages_unlocked(current, mm, start,
+			(end - start) >> PAGE_SHIFT, write, 0, pages);
+
+		/* 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/sh/mm/hugetlbpage.c b/arch/sh/mm/hugetlbpage.c
new file mode 100644
index 0000000..6385f60
--- /dev/null
+++ b/arch/sh/mm/hugetlbpage.c
@@ -0,0 +1,73 @@
+/*
+ * arch/sh/mm/hugetlbpage.c
+ *
+ * SuperH HugeTLB page support.
+ *
+ * Cloned from sparc64 by Paul Mundt.
+ *
+ * Copyright (C) 2002, 2003 David S. Miller (davem@redhat.com)
+ */
+
+#include <linux/init.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/hugetlb.h>
+#include <linux/pagemap.h>
+#include <linux/sysctl.h>
+
+#include <asm/mman.h>
+#include <asm/pgalloc.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+#include <asm/cacheflush.h>
+
+pte_t *huge_pte_alloc(struct mm_struct *mm,
+			unsigned long addr, unsigned long sz)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+
+	pgd = pgd_offset(mm, addr);
+	if (pgd) {
+		pud = pud_alloc(mm, pgd, addr);
+		if (pud) {
+			pmd = pmd_alloc(mm, pud, addr);
+			if (pmd)
+				pte = pte_alloc_map(mm, NULL, pmd, addr);
+		}
+	}
+
+	return pte;
+}
+
+pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+
+	pgd = pgd_offset(mm, addr);
+	if (pgd) {
+		pud = pud_offset(pgd, addr);
+		if (pud) {
+			pmd = pmd_offset(pud, addr);
+			if (pmd)
+				pte = pte_offset_map(pmd, addr);
+		}
+	}
+
+	return pte;
+}
+
+int pmd_huge(pmd_t pmd)
+{
+	return 0;
+}
+
+int pud_huge(pud_t pud)
+{
+	return 0;
+}
diff --git a/arch/sh/mm/init.c b/arch/sh/mm/init.c
new file mode 100644
index 0000000..7549186
--- /dev/null
+++ b/arch/sh/mm/init.c
@@ -0,0 +1,535 @@
+/*
+ * linux/arch/sh/mm/init.c
+ *
+ *  Copyright (C) 1999  Niibe Yutaka
+ *  Copyright (C) 2002 - 2011  Paul Mundt
+ *
+ *  Based on linux/arch/i386/mm/init.c:
+ *   Copyright (C) 1995  Linus Torvalds
+ */
+#include <linux/mm.h>
+#include <linux/swap.h>
+#include <linux/init.h>
+#include <linux/gfp.h>
+#include <linux/bootmem.h>
+#include <linux/proc_fs.h>
+#include <linux/pagemap.h>
+#include <linux/percpu.h>
+#include <linux/io.h>
+#include <linux/memblock.h>
+#include <linux/dma-mapping.h>
+#include <linux/export.h>
+#include <asm/mmu_context.h>
+#include <asm/mmzone.h>
+#include <asm/kexec.h>
+#include <asm/tlb.h>
+#include <asm/cacheflush.h>
+#include <asm/sections.h>
+#include <asm/setup.h>
+#include <asm/cache.h>
+#include <asm/sizes.h>
+
+pgd_t swapper_pg_dir[PTRS_PER_PGD];
+
+void __init generic_mem_init(void)
+{
+	memblock_add(__MEMORY_START, __MEMORY_SIZE);
+}
+
+void __init __weak plat_mem_setup(void)
+{
+	/* Nothing to see here, move along. */
+}
+
+#ifdef CONFIG_MMU
+static pte_t *__get_pte_phys(unsigned long addr)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+
+	pgd = pgd_offset_k(addr);
+	if (pgd_none(*pgd)) {
+		pgd_ERROR(*pgd);
+		return NULL;
+	}
+
+	pud = pud_alloc(NULL, pgd, addr);
+	if (unlikely(!pud)) {
+		pud_ERROR(*pud);
+		return NULL;
+	}
+
+	pmd = pmd_alloc(NULL, pud, addr);
+	if (unlikely(!pmd)) {
+		pmd_ERROR(*pmd);
+		return NULL;
+	}
+
+	return pte_offset_kernel(pmd, addr);
+}
+
+static void set_pte_phys(unsigned long addr, unsigned long phys, pgprot_t prot)
+{
+	pte_t *pte;
+
+	pte = __get_pte_phys(addr);
+	if (!pte_none(*pte)) {
+		pte_ERROR(*pte);
+		return;
+	}
+
+	set_pte(pte, pfn_pte(phys >> PAGE_SHIFT, prot));
+	local_flush_tlb_one(get_asid(), addr);
+
+	if (pgprot_val(prot) & _PAGE_WIRED)
+		tlb_wire_entry(NULL, addr, *pte);
+}
+
+static void clear_pte_phys(unsigned long addr, pgprot_t prot)
+{
+	pte_t *pte;
+
+	pte = __get_pte_phys(addr);
+
+	if (pgprot_val(prot) & _PAGE_WIRED)
+		tlb_unwire_entry();
+
+	set_pte(pte, pfn_pte(0, __pgprot(0)));
+	local_flush_tlb_one(get_asid(), addr);
+}
+
+void __set_fixmap(enum fixed_addresses idx, unsigned long phys, pgprot_t prot)
+{
+	unsigned long address = __fix_to_virt(idx);
+
+	if (idx >= __end_of_fixed_addresses) {
+		BUG();
+		return;
+	}
+
+	set_pte_phys(address, phys, prot);
+}
+
+void __clear_fixmap(enum fixed_addresses idx, pgprot_t prot)
+{
+	unsigned long address = __fix_to_virt(idx);
+
+	if (idx >= __end_of_fixed_addresses) {
+		BUG();
+		return;
+	}
+
+	clear_pte_phys(address, prot);
+}
+
+static pmd_t * __init one_md_table_init(pud_t *pud)
+{
+	if (pud_none(*pud)) {
+		pmd_t *pmd;
+
+		pmd = alloc_bootmem_pages(PAGE_SIZE);
+		pud_populate(&init_mm, pud, pmd);
+		BUG_ON(pmd != pmd_offset(pud, 0));
+	}
+
+	return pmd_offset(pud, 0);
+}
+
+static pte_t * __init one_page_table_init(pmd_t *pmd)
+{
+	if (pmd_none(*pmd)) {
+		pte_t *pte;
+
+		pte = alloc_bootmem_pages(PAGE_SIZE);
+		pmd_populate_kernel(&init_mm, pmd, pte);
+		BUG_ON(pte != pte_offset_kernel(pmd, 0));
+	}
+
+	return pte_offset_kernel(pmd, 0);
+}
+
+static pte_t * __init page_table_kmap_check(pte_t *pte, pmd_t *pmd,
+					    unsigned long vaddr, pte_t *lastpte)
+{
+	return pte;
+}
+
+void __init page_table_range_init(unsigned long start, unsigned long end,
+					 pgd_t *pgd_base)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte = NULL;
+	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 = one_md_table_init(pud);
+#ifndef __PAGETABLE_PMD_FOLDED
+			pmd += k;
+#endif
+			for (; (k < PTRS_PER_PMD) && (vaddr != end); pmd++, k++) {
+				pte = page_table_kmap_check(one_page_table_init(pmd),
+							    pmd, vaddr, pte);
+				vaddr += PMD_SIZE;
+			}
+			k = 0;
+		}
+		j = 0;
+	}
+}
+#endif	/* CONFIG_MMU */
+
+void __init allocate_pgdat(unsigned int nid)
+{
+	unsigned long start_pfn, end_pfn;
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+	unsigned long phys;
+#endif
+
+	get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+	phys = __memblock_alloc_base(sizeof(struct pglist_data),
+				SMP_CACHE_BYTES, end_pfn << PAGE_SHIFT);
+	/* Retry with all of system memory */
+	if (!phys)
+		phys = __memblock_alloc_base(sizeof(struct pglist_data),
+					SMP_CACHE_BYTES, memblock_end_of_DRAM());
+	if (!phys)
+		panic("Can't allocate pgdat for node %d\n", nid);
+
+	NODE_DATA(nid) = __va(phys);
+	memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
+
+	NODE_DATA(nid)->bdata = &bootmem_node_data[nid];
+#endif
+
+	NODE_DATA(nid)->node_start_pfn = start_pfn;
+	NODE_DATA(nid)->node_spanned_pages = end_pfn - start_pfn;
+}
+
+static void __init bootmem_init_one_node(unsigned int nid)
+{
+	unsigned long total_pages, paddr;
+	unsigned long end_pfn;
+	struct pglist_data *p;
+
+	p = NODE_DATA(nid);
+
+	/* Nothing to do.. */
+	if (!p->node_spanned_pages)
+		return;
+
+	end_pfn = pgdat_end_pfn(p);
+
+	total_pages = bootmem_bootmap_pages(p->node_spanned_pages);
+
+	paddr = memblock_alloc(total_pages << PAGE_SHIFT, PAGE_SIZE);
+	if (!paddr)
+		panic("Can't allocate bootmap for nid[%d]\n", nid);
+
+	init_bootmem_node(p, paddr >> PAGE_SHIFT, p->node_start_pfn, end_pfn);
+
+	free_bootmem_with_active_regions(nid, end_pfn);
+
+	/*
+	 * XXX Handle initial reservations for the system memory node
+	 * only for the moment, we'll refactor this later for handling
+	 * reservations in other nodes.
+	 */
+	if (nid == 0) {
+		struct memblock_region *reg;
+
+		/* Reserve the sections we're already using. */
+		for_each_memblock(reserved, reg) {
+			reserve_bootmem(reg->base, reg->size, BOOTMEM_DEFAULT);
+		}
+	}
+
+	sparse_memory_present_with_active_regions(nid);
+}
+
+static void __init do_init_bootmem(void)
+{
+	struct memblock_region *reg;
+	int i;
+
+	/* Add active regions with valid PFNs. */
+	for_each_memblock(memory, reg) {
+		unsigned long start_pfn, end_pfn;
+		start_pfn = memblock_region_memory_base_pfn(reg);
+		end_pfn = memblock_region_memory_end_pfn(reg);
+		__add_active_range(0, start_pfn, end_pfn);
+	}
+
+	/* All of system RAM sits in node 0 for the non-NUMA case */
+	allocate_pgdat(0);
+	node_set_online(0);
+
+	plat_mem_setup();
+
+	for_each_online_node(i)
+		bootmem_init_one_node(i);
+
+	sparse_init();
+}
+
+static void __init early_reserve_mem(void)
+{
+	unsigned long start_pfn;
+	u32 zero_base = (u32)__MEMORY_START + (u32)PHYSICAL_OFFSET;
+	u32 start = zero_base + (u32)CONFIG_ZERO_PAGE_OFFSET;
+
+	/*
+	 * Partially used pages are not usable - thus
+	 * we are rounding upwards:
+	 */
+	start_pfn = PFN_UP(__pa(_end));
+
+	/*
+	 * Reserve the kernel text and Reserve the bootmem bitmap. We do
+	 * this in two steps (first step was init_bootmem()), because
+	 * this catches the (definitely buggy) case of us accidentally
+	 * initializing the bootmem allocator with an invalid RAM area.
+	 */
+	memblock_reserve(start, (PFN_PHYS(start_pfn) + PAGE_SIZE - 1) - start);
+
+	/*
+	 * Reserve physical pages below CONFIG_ZERO_PAGE_OFFSET.
+	 */
+	if (CONFIG_ZERO_PAGE_OFFSET != 0)
+		memblock_reserve(zero_base, CONFIG_ZERO_PAGE_OFFSET);
+
+	/*
+	 * Handle additional early reservations
+	 */
+	check_for_initrd();
+	reserve_crashkernel();
+}
+
+void __init paging_init(void)
+{
+	unsigned long max_zone_pfns[MAX_NR_ZONES];
+	unsigned long vaddr, end;
+	int nid;
+
+	sh_mv.mv_mem_init();
+
+	early_reserve_mem();
+
+	/*
+	 * Once the early reservations are out of the way, give the
+	 * platforms a chance to kick out some memory.
+	 */
+	if (sh_mv.mv_mem_reserve)
+		sh_mv.mv_mem_reserve();
+
+	memblock_enforce_memory_limit(memory_limit);
+	memblock_allow_resize();
+
+	memblock_dump_all();
+
+	/*
+	 * Determine low and high memory ranges:
+	 */
+	max_low_pfn = max_pfn = memblock_end_of_DRAM() >> PAGE_SHIFT;
+	min_low_pfn = __MEMORY_START >> PAGE_SHIFT;
+
+	nodes_clear(node_online_map);
+
+	memory_start = (unsigned long)__va(__MEMORY_START);
+	memory_end = memory_start + (memory_limit ?: memblock_phys_mem_size());
+
+	uncached_init();
+	pmb_init();
+	do_init_bootmem();
+	ioremap_fixed_init();
+
+	/* We don't need to map the kernel through the TLB, as
+	 * it is permanatly mapped using P1. So clear the
+	 * entire pgd. */
+	memset(swapper_pg_dir, 0, sizeof(swapper_pg_dir));
+
+	/* Set an initial value for the MMU.TTB so we don't have to
+	 * check for a null value. */
+	set_TTB(swapper_pg_dir);
+
+	/*
+	 * Populate the relevant portions of swapper_pg_dir so that
+	 * we can use the fixmap entries without calling kmalloc.
+	 * pte's will be filled in by __set_fixmap().
+	 */
+	vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK;
+	end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK;
+	page_table_range_init(vaddr, end, swapper_pg_dir);
+
+	kmap_coherent_init();
+
+	memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+
+	for_each_online_node(nid) {
+		pg_data_t *pgdat = NODE_DATA(nid);
+		unsigned long low, start_pfn;
+
+		start_pfn = pgdat->bdata->node_min_pfn;
+		low = pgdat->bdata->node_low_pfn;
+
+		if (max_zone_pfns[ZONE_NORMAL] < low)
+			max_zone_pfns[ZONE_NORMAL] = low;
+
+		printk("Node %u: start_pfn = 0x%lx, low = 0x%lx\n",
+		       nid, start_pfn, low);
+	}
+
+	free_area_init_nodes(max_zone_pfns);
+}
+
+/*
+ * Early initialization for any I/O MMUs we might have.
+ */
+static void __init iommu_init(void)
+{
+	no_iommu_init();
+}
+
+unsigned int mem_init_done = 0;
+
+void __init mem_init(void)
+{
+	pg_data_t *pgdat;
+
+	iommu_init();
+
+	high_memory = NULL;
+	for_each_online_pgdat(pgdat)
+		high_memory = max_t(void *, high_memory,
+				    __va(pgdat_end_pfn(pgdat) << PAGE_SHIFT));
+
+	free_all_bootmem();
+
+	/* Set this up early, so we can take care of the zero page */
+	cpu_cache_init();
+
+	/* clear the zero-page */
+	memset(empty_zero_page, 0, PAGE_SIZE);
+	__flush_wback_region(empty_zero_page, PAGE_SIZE);
+
+	vsyscall_init();
+
+	mem_init_print_info(NULL);
+	pr_info("virtual kernel memory layout:\n"
+		"    fixmap  : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+#ifdef CONFIG_HIGHMEM
+		"    pkmap   : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+#endif
+		"    vmalloc : 0x%08lx - 0x%08lx   (%4ld MB)\n"
+		"    lowmem  : 0x%08lx - 0x%08lx   (%4ld MB) (cached)\n"
+#ifdef CONFIG_UNCACHED_MAPPING
+		"            : 0x%08lx - 0x%08lx   (%4ld MB) (uncached)\n"
+#endif
+		"      .init : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+		"      .data : 0x%08lx - 0x%08lx   (%4ld kB)\n"
+		"      .text : 0x%08lx - 0x%08lx   (%4ld kB)\n",
+		FIXADDR_START, FIXADDR_TOP,
+		(FIXADDR_TOP - FIXADDR_START) >> 10,
+
+#ifdef CONFIG_HIGHMEM
+		PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE,
+		(LAST_PKMAP*PAGE_SIZE) >> 10,
+#endif
+
+		(unsigned long)VMALLOC_START, VMALLOC_END,
+		(VMALLOC_END - VMALLOC_START) >> 20,
+
+		(unsigned long)memory_start, (unsigned long)high_memory,
+		((unsigned long)high_memory - (unsigned long)memory_start) >> 20,
+
+#ifdef CONFIG_UNCACHED_MAPPING
+		uncached_start, uncached_end, uncached_size >> 20,
+#endif
+
+		(unsigned long)&__init_begin, (unsigned long)&__init_end,
+		((unsigned long)&__init_end -
+		 (unsigned long)&__init_begin) >> 10,
+
+		(unsigned long)&_etext, (unsigned long)&_edata,
+		((unsigned long)&_edata - (unsigned long)&_etext) >> 10,
+
+		(unsigned long)&_text, (unsigned long)&_etext,
+		((unsigned long)&_etext - (unsigned long)&_text) >> 10);
+
+	mem_init_done = 1;
+}
+
+void free_initmem(void)
+{
+	free_initmem_default(-1);
+}
+
+#ifdef CONFIG_BLK_DEV_INITRD
+void free_initrd_mem(unsigned long start, unsigned long end)
+{
+	free_reserved_area((void *)start, (void *)end, -1, "initrd");
+}
+#endif
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+int arch_add_memory(int nid, u64 start, u64 size, bool for_device)
+{
+	pg_data_t *pgdat;
+	unsigned long start_pfn = PFN_DOWN(start);
+	unsigned long nr_pages = size >> PAGE_SHIFT;
+	int ret;
+
+	pgdat = NODE_DATA(nid);
+
+	/* We only have ZONE_NORMAL, so this is easy.. */
+	ret = __add_pages(nid, pgdat->node_zones +
+			zone_for_memory(nid, start, size, ZONE_NORMAL,
+			for_device),
+			start_pfn, nr_pages);
+	if (unlikely(ret))
+		printk("%s: Failed, __add_pages() == %d\n", __func__, ret);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(arch_add_memory);
+
+#ifdef CONFIG_NUMA
+int memory_add_physaddr_to_nid(u64 addr)
+{
+	/* Node 0 for now.. */
+	return 0;
+}
+EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
+#endif
+
+#ifdef CONFIG_MEMORY_HOTREMOVE
+int arch_remove_memory(u64 start, u64 size)
+{
+	unsigned long start_pfn = PFN_DOWN(start);
+	unsigned long nr_pages = size >> PAGE_SHIFT;
+	struct zone *zone;
+	int ret;
+
+	zone = page_zone(pfn_to_page(start_pfn));
+	ret = __remove_pages(zone, start_pfn, nr_pages);
+	if (unlikely(ret))
+		pr_warn("%s: Failed, __remove_pages() == %d\n", __func__,
+			ret);
+
+	return ret;
+}
+#endif
+#endif /* CONFIG_MEMORY_HOTPLUG */
diff --git a/arch/sh/mm/ioremap.c b/arch/sh/mm/ioremap.c
new file mode 100644
index 0000000..0c99ec2
--- /dev/null
+++ b/arch/sh/mm/ioremap.c
@@ -0,0 +1,137 @@
+/*
+ * arch/sh/mm/ioremap.c
+ *
+ * (C) Copyright 1995 1996 Linus Torvalds
+ * (C) Copyright 2005 - 2010  Paul Mundt
+ *
+ * Re-map IO memory to kernel address space so that we can access it.
+ * This is needed for high PCI addresses that aren't mapped in the
+ * 640k-1MB IO memory area on PC's
+ *
+ * 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.
+ */
+#include <linux/vmalloc.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/mm.h>
+#include <linux/pci.h>
+#include <linux/io.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/addrspace.h>
+#include <asm/cacheflush.h>
+#include <asm/tlbflush.h>
+#include <asm/mmu.h>
+
+/*
+ * 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.
+ */
+void __iomem * __init_refok
+__ioremap_caller(phys_addr_t phys_addr, unsigned long size,
+		 pgprot_t pgprot, void *caller)
+{
+	struct vm_struct *area;
+	unsigned long offset, last_addr, addr, orig_addr;
+	void __iomem *mapped;
+
+	/* Don't allow wraparound or zero size */
+	last_addr = phys_addr + size - 1;
+	if (!size || last_addr < phys_addr)
+		return NULL;
+
+	/*
+	 * If we can't yet use the regular approach, go the fixmap route.
+	 */
+	if (!mem_init_done)
+		return ioremap_fixed(phys_addr, size, pgprot);
+
+	/*
+	 * First try to remap through the PMB.
+	 * PMB entries are all pre-faulted.
+	 */
+	mapped = pmb_remap_caller(phys_addr, size, pgprot, caller);
+	if (mapped && !IS_ERR(mapped))
+		return mapped;
+
+	/*
+	 * 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_caller(size, VM_IOREMAP, caller);
+	if (!area)
+		return NULL;
+	area->phys_addr = phys_addr;
+	orig_addr = addr = (unsigned long)area->addr;
+
+	if (ioremap_page_range(addr, addr + size, phys_addr, pgprot)) {
+		vunmap((void *)orig_addr);
+		return NULL;
+	}
+
+	return (void __iomem *)(offset + (char *)orig_addr);
+}
+EXPORT_SYMBOL(__ioremap_caller);
+
+/*
+ * Simple checks for non-translatable mappings.
+ */
+static inline int iomapping_nontranslatable(unsigned long offset)
+{
+#ifdef CONFIG_29BIT
+	/*
+	 * In 29-bit mode this includes the fixed P1/P2 areas, as well as
+	 * parts of P3.
+	 */
+	if (PXSEG(offset) < P3SEG || offset >= P3_ADDR_MAX)
+		return 1;
+#endif
+
+	return 0;
+}
+
+void __iounmap(void __iomem *addr)
+{
+	unsigned long vaddr = (unsigned long __force)addr;
+	struct vm_struct *p;
+
+	/*
+	 * Nothing to do if there is no translatable mapping.
+	 */
+	if (iomapping_nontranslatable(vaddr))
+		return;
+
+	/*
+	 * There's no VMA if it's from an early fixed mapping.
+	 */
+	if (iounmap_fixed(addr) == 0)
+		return;
+
+	/*
+	 * If the PMB handled it, there's nothing else to do.
+	 */
+	if (pmb_unmap(addr) == 0)
+		return;
+
+	p = remove_vm_area((void *)(vaddr & PAGE_MASK));
+	if (!p) {
+		printk(KERN_ERR "%s: bad address %p\n", __func__, addr);
+		return;
+	}
+
+	kfree(p);
+}
+EXPORT_SYMBOL(__iounmap);
diff --git a/arch/sh/mm/ioremap_fixed.c b/arch/sh/mm/ioremap_fixed.c
new file mode 100644
index 0000000..efbe84a
--- /dev/null
+++ b/arch/sh/mm/ioremap_fixed.c
@@ -0,0 +1,134 @@
+/*
+ * Re-map IO memory to kernel address space so that we can access it.
+ *
+ * These functions should only be used when it is necessary to map a
+ * physical address space into the kernel address space before ioremap()
+ * can be used, e.g. early in boot before paging_init().
+ *
+ * Copyright (C) 2009  Matt Fleming
+ */
+
+#include <linux/vmalloc.h>
+#include <linux/ioport.h>
+#include <linux/module.h>
+#include <linux/mm.h>
+#include <linux/io.h>
+#include <linux/bootmem.h>
+#include <linux/proc_fs.h>
+#include <asm/fixmap.h>
+#include <asm/page.h>
+#include <asm/pgalloc.h>
+#include <asm/addrspace.h>
+#include <asm/cacheflush.h>
+#include <asm/tlbflush.h>
+#include <asm/mmu.h>
+#include <asm/mmu_context.h>
+
+struct ioremap_map {
+	void __iomem *addr;
+	unsigned long size;
+	unsigned long fixmap_addr;
+};
+
+static struct ioremap_map ioremap_maps[FIX_N_IOREMAPS];
+
+void __init ioremap_fixed_init(void)
+{
+	struct ioremap_map *map;
+	int i;
+
+	for (i = 0; i < FIX_N_IOREMAPS; i++) {
+		map = &ioremap_maps[i];
+		map->fixmap_addr = __fix_to_virt(FIX_IOREMAP_BEGIN + i);
+	}
+}
+
+void __init __iomem *
+ioremap_fixed(phys_addr_t phys_addr, unsigned long size, pgprot_t prot)
+{
+	enum fixed_addresses idx0, idx;
+	struct ioremap_map *map;
+	unsigned int nrpages;
+	unsigned long offset;
+	int i, slot;
+
+	/*
+	 * Mappings have to be page-aligned
+	 */
+	offset = phys_addr & ~PAGE_MASK;
+	phys_addr &= PAGE_MASK;
+	size = PAGE_ALIGN(phys_addr + size) - phys_addr;
+
+	slot = -1;
+	for (i = 0; i < FIX_N_IOREMAPS; i++) {
+		map = &ioremap_maps[i];
+		if (!map->addr) {
+			map->size = size;
+			slot = i;
+			break;
+		}
+	}
+
+	if (slot < 0)
+		return NULL;
+
+	/*
+	 * Mappings have to fit in the FIX_IOREMAP area.
+	 */
+	nrpages = size >> PAGE_SHIFT;
+	if (nrpages > FIX_N_IOREMAPS)
+		return NULL;
+
+	/*
+	 * Ok, go for it..
+	 */
+	idx0 = FIX_IOREMAP_BEGIN + slot;
+	idx = idx0;
+	while (nrpages > 0) {
+		pgprot_val(prot) |= _PAGE_WIRED;
+		__set_fixmap(idx, phys_addr, prot);
+		phys_addr += PAGE_SIZE;
+		idx++;
+		--nrpages;
+	}
+
+	map->addr = (void __iomem *)(offset + map->fixmap_addr);
+	return map->addr;
+}
+
+int iounmap_fixed(void __iomem *addr)
+{
+	enum fixed_addresses idx;
+	struct ioremap_map *map;
+	unsigned int nrpages;
+	int i, slot;
+
+	slot = -1;
+	for (i = 0; i < FIX_N_IOREMAPS; i++) {
+		map = &ioremap_maps[i];
+		if (map->addr == addr) {
+			slot = i;
+			break;
+		}
+	}
+
+	/*
+	 * If we don't match, it's not for us.
+	 */
+	if (slot < 0)
+		return -EINVAL;
+
+	nrpages = map->size >> PAGE_SHIFT;
+
+	idx = FIX_IOREMAP_BEGIN + slot + nrpages - 1;
+	while (nrpages > 0) {
+		__clear_fixmap(idx, __pgprot(_PAGE_WIRED));
+		--idx;
+		--nrpages;
+	}
+
+	map->size = 0;
+	map->addr = NULL;
+
+	return 0;
+}
diff --git a/arch/sh/mm/kmap.c b/arch/sh/mm/kmap.c
new file mode 100644
index 0000000..bf25d7c
--- /dev/null
+++ b/arch/sh/mm/kmap.c
@@ -0,0 +1,69 @@
+/*
+ * arch/sh/mm/kmap.c
+ *
+ * Copyright (C) 1999, 2000, 2002  Niibe Yutaka
+ * Copyright (C) 2002 - 2009  Paul Mundt
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/mutex.h>
+#include <linux/fs.h>
+#include <linux/highmem.h>
+#include <linux/module.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+#define kmap_get_fixmap_pte(vaddr)                                     \
+	pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr), (vaddr)), (vaddr)), (vaddr))
+
+static pte_t *kmap_coherent_pte;
+
+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);
+}
+
+void *kmap_coherent(struct page *page, unsigned long addr)
+{
+	enum fixed_addresses idx;
+	unsigned long vaddr;
+
+	BUG_ON(!test_bit(PG_dcache_clean, &page->flags));
+
+	preempt_disable();
+	pagefault_disable();
+
+	idx = FIX_CMAP_END -
+		(((addr >> PAGE_SHIFT) & (FIX_N_COLOURS - 1)) +
+		 (FIX_N_COLOURS * smp_processor_id()));
+
+	vaddr = __fix_to_virt(idx);
+
+	BUG_ON(!pte_none(*(kmap_coherent_pte - idx)));
+	set_pte(kmap_coherent_pte - idx, mk_pte(page, PAGE_KERNEL));
+
+	return (void *)vaddr;
+}
+
+void kunmap_coherent(void *kvaddr)
+{
+	if (kvaddr >= (void *)FIXADDR_START) {
+		unsigned long vaddr = (unsigned long)kvaddr & PAGE_MASK;
+		enum fixed_addresses idx = __virt_to_fix(vaddr);
+
+		/* XXX.. Kill this later, here for sanity at the moment.. */
+		__flush_purge_region((void *)vaddr, PAGE_SIZE);
+
+		pte_clear(&init_mm, vaddr, kmap_coherent_pte - idx);
+		local_flush_tlb_one(get_asid(), vaddr);
+	}
+
+	pagefault_enable();
+	preempt_enable();
+}
diff --git a/arch/sh/mm/mmap.c b/arch/sh/mm/mmap.c
new file mode 100644
index 0000000..7df7d59
--- /dev/null
+++ b/arch/sh/mm/mmap.c
@@ -0,0 +1,163 @@
+/*
+ * arch/sh/mm/mmap.c
+ *
+ * Copyright (C) 2008 - 2009  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/io.h>
+#include <linux/mm.h>
+#include <linux/mman.h>
+#include <linux/module.h>
+#include <asm/page.h>
+#include <asm/processor.h>
+
+unsigned long shm_align_mask = PAGE_SIZE - 1;	/* Sane caches */
+EXPORT_SYMBOL(shm_align_mask);
+
+#ifdef CONFIG_MMU
+/*
+ * To avoid cache aliases, we map the shared page with same color.
+ */
+static inline unsigned long COLOUR_ALIGN(unsigned long addr,
+					 unsigned long pgoff)
+{
+	unsigned long base = (addr + shm_align_mask) & ~shm_align_mask;
+	unsigned long off = (pgoff << PAGE_SHIFT) & shm_align_mask;
+
+	return base + off;
+}
+
+unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr,
+	unsigned long len, unsigned long pgoff, unsigned long flags)
+{
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+	int do_colour_align;
+	struct vm_unmapped_area_info info;
+
+	if (flags & MAP_FIXED) {
+		/* 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;
+	}
+
+	if (unlikely(len > TASK_SIZE))
+		return -ENOMEM;
+
+	do_colour_align = 0;
+	if (filp || (flags & MAP_SHARED))
+		do_colour_align = 1;
+
+	if (addr) {
+		if (do_colour_align)
+			addr = COLOUR_ALIGN(addr, pgoff);
+		else
+			addr = PAGE_ALIGN(addr);
+
+		vma = find_vma(mm, addr);
+		if (TASK_SIZE - len >= addr &&
+		    (!vma || addr + len <= vm_start_gap(vma)))
+			return addr;
+	}
+
+	info.flags = 0;
+	info.length = len;
+	info.low_limit = TASK_UNMAPPED_BASE;
+	info.high_limit = TASK_SIZE;
+	info.align_mask = do_colour_align ? (PAGE_MASK & shm_align_mask) : 0;
+	info.align_offset = pgoff << PAGE_SHIFT;
+	return vm_unmapped_area(&info);
+}
+
+unsigned long
+arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
+			  const unsigned long len, const unsigned long pgoff,
+			  const unsigned long flags)
+{
+	struct vm_area_struct *vma;
+	struct mm_struct *mm = current->mm;
+	unsigned long addr = addr0;
+	int do_colour_align;
+	struct vm_unmapped_area_info info;
+
+	if (flags & MAP_FIXED) {
+		/* 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;
+	}
+
+	if (unlikely(len > TASK_SIZE))
+		return -ENOMEM;
+
+	do_colour_align = 0;
+	if (filp || (flags & MAP_SHARED))
+		do_colour_align = 1;
+
+	/* requesting a specific address */
+	if (addr) {
+		if (do_colour_align)
+			addr = COLOUR_ALIGN(addr, pgoff);
+		else
+			addr = PAGE_ALIGN(addr);
+
+		vma = find_vma(mm, addr);
+		if (TASK_SIZE - len >= addr &&
+		    (!vma || addr + len <= vm_start_gap(vma)))
+			return addr;
+	}
+
+	info.flags = VM_UNMAPPED_AREA_TOPDOWN;
+	info.length = len;
+	info.low_limit = PAGE_SIZE;
+	info.high_limit = mm->mmap_base;
+	info.align_mask = do_colour_align ? (PAGE_MASK & shm_align_mask) : 0;
+	info.align_offset = pgoff << PAGE_SHIFT;
+	addr = vm_unmapped_area(&info);
+
+	/*
+	 * 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.
+	 */
+	if (addr & ~PAGE_MASK) {
+		VM_BUG_ON(addr != -ENOMEM);
+		info.flags = 0;
+		info.low_limit = TASK_UNMAPPED_BASE;
+		info.high_limit = TASK_SIZE;
+		addr = vm_unmapped_area(&info);
+	}
+
+	return addr;
+}
+#endif /* CONFIG_MMU */
+
+/*
+ * You really shouldn't be using read() or write() on /dev/mem.  This
+ * might go away in the future.
+ */
+int valid_phys_addr_range(phys_addr_t addr, size_t count)
+{
+	if (addr < __MEMORY_START)
+		return 0;
+	if (addr + count > __pa(high_memory))
+		return 0;
+
+	return 1;
+}
+
+int valid_mmap_phys_addr_range(unsigned long pfn, size_t size)
+{
+	return 1;
+}
diff --git a/arch/sh/mm/nommu.c b/arch/sh/mm/nommu.c
new file mode 100644
index 0000000..36312d2
--- /dev/null
+++ b/arch/sh/mm/nommu.c
@@ -0,0 +1,104 @@
+/*
+ * arch/sh/mm/nommu.c
+ *
+ * Various helper routines and stubs for MMUless SH.
+ *
+ * Copyright (C) 2002 - 2009 Paul Mundt
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+#include <asm/page.h>
+#include <asm/uaccess.h>
+
+/*
+ * Nothing too terribly exciting here ..
+ */
+void copy_page(void *to, void *from)
+{
+	memcpy(to, from, PAGE_SIZE);
+}
+
+__kernel_size_t __copy_user(void *to, const void *from, __kernel_size_t n)
+{
+	memcpy(to, from, n);
+	return 0;
+}
+
+__kernel_size_t __clear_user(void *to, __kernel_size_t n)
+{
+	memset(to, 0, n);
+	return 0;
+}
+
+void local_flush_tlb_all(void)
+{
+	BUG();
+}
+
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	BUG();
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+			    unsigned long end)
+{
+	BUG();
+}
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	BUG();
+}
+
+void local_flush_tlb_one(unsigned long asid, unsigned long page)
+{
+	BUG();
+}
+
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+	BUG();
+}
+
+void __flush_tlb_global(void)
+{
+}
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t pte)
+{
+}
+
+void __init kmap_coherent_init(void)
+{
+}
+
+void *kmap_coherent(struct page *page, unsigned long addr)
+{
+	BUG();
+	return NULL;
+}
+
+void kunmap_coherent(void *kvaddr)
+{
+	BUG();
+}
+
+void __init page_table_range_init(unsigned long start, unsigned long end,
+				  pgd_t *pgd_base)
+{
+}
+
+void __set_fixmap(enum fixed_addresses idx, unsigned long phys, pgprot_t prot)
+{
+}
+
+void pgtable_cache_init(void)
+{
+}
diff --git a/arch/sh/mm/numa.c b/arch/sh/mm/numa.c
new file mode 100644
index 0000000..05713d1
--- /dev/null
+++ b/arch/sh/mm/numa.c
@@ -0,0 +1,75 @@
+/*
+ * arch/sh/mm/numa.c - Multiple node support for SH machines
+ *
+ *  Copyright (C) 2007  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/module.h>
+#include <linux/bootmem.h>
+#include <linux/memblock.h>
+#include <linux/mm.h>
+#include <linux/numa.h>
+#include <linux/pfn.h>
+#include <asm/sections.h>
+
+struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
+EXPORT_SYMBOL_GPL(node_data);
+
+/*
+ * On SH machines the conventional approach is to stash system RAM
+ * in node 0, and other memory blocks in to node 1 and up, ordered by
+ * latency. Each node's pgdat is node-local at the beginning of the node,
+ * immediately followed by the node mem map.
+ */
+void __init setup_bootmem_node(int nid, unsigned long start, unsigned long end)
+{
+	unsigned long bootmap_pages;
+	unsigned long start_pfn, end_pfn;
+	unsigned long bootmem_paddr;
+
+	/* Don't allow bogus node assignment */
+	BUG_ON(nid >= MAX_NUMNODES || nid <= 0);
+
+	start_pfn = PFN_DOWN(start);
+	end_pfn = PFN_DOWN(end);
+
+	pmb_bolt_mapping((unsigned long)__va(start), start, end - start,
+			 PAGE_KERNEL);
+
+	memblock_add(start, end - start);
+
+	__add_active_range(nid, start_pfn, end_pfn);
+
+	/* Node-local pgdat */
+	NODE_DATA(nid) = __va(memblock_alloc_base(sizeof(struct pglist_data),
+					     SMP_CACHE_BYTES, end));
+	memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
+
+	NODE_DATA(nid)->bdata = &bootmem_node_data[nid];
+	NODE_DATA(nid)->node_start_pfn = start_pfn;
+	NODE_DATA(nid)->node_spanned_pages = end_pfn - start_pfn;
+
+	/* Node-local bootmap */
+	bootmap_pages = bootmem_bootmap_pages(end_pfn - start_pfn);
+	bootmem_paddr = memblock_alloc_base(bootmap_pages << PAGE_SHIFT,
+				       PAGE_SIZE, end);
+	init_bootmem_node(NODE_DATA(nid), bootmem_paddr >> PAGE_SHIFT,
+			  start_pfn, end_pfn);
+
+	free_bootmem_with_active_regions(nid, end_pfn);
+
+	/* Reserve the pgdat and bootmap space with the bootmem allocator */
+	reserve_bootmem_node(NODE_DATA(nid), start_pfn << PAGE_SHIFT,
+			     sizeof(struct pglist_data), BOOTMEM_DEFAULT);
+	reserve_bootmem_node(NODE_DATA(nid), bootmem_paddr,
+			     bootmap_pages << PAGE_SHIFT, BOOTMEM_DEFAULT);
+
+	/* It's up */
+	node_set_online(nid);
+
+	/* Kick sparsemem */
+	sparse_memory_present_with_active_regions(nid);
+}
diff --git a/arch/sh/mm/pgtable.c b/arch/sh/mm/pgtable.c
new file mode 100644
index 0000000..26e03a1
--- /dev/null
+++ b/arch/sh/mm/pgtable.c
@@ -0,0 +1,57 @@
+#include <linux/mm.h>
+#include <linux/slab.h>
+
+#define PGALLOC_GFP GFP_KERNEL | __GFP_REPEAT | __GFP_ZERO
+
+static struct kmem_cache *pgd_cachep;
+#if PAGETABLE_LEVELS > 2
+static struct kmem_cache *pmd_cachep;
+#endif
+
+void pgd_ctor(void *x)
+{
+	pgd_t *pgd = x;
+
+	memcpy(pgd + USER_PTRS_PER_PGD,
+	       swapper_pg_dir + USER_PTRS_PER_PGD,
+	       (PTRS_PER_PGD - USER_PTRS_PER_PGD) * sizeof(pgd_t));
+}
+
+void pgtable_cache_init(void)
+{
+	pgd_cachep = kmem_cache_create("pgd_cache",
+				       PTRS_PER_PGD * (1<<PTE_MAGNITUDE),
+				       PAGE_SIZE, SLAB_PANIC, pgd_ctor);
+#if PAGETABLE_LEVELS > 2
+	pmd_cachep = kmem_cache_create("pmd_cache",
+				       PTRS_PER_PMD * (1<<PTE_MAGNITUDE),
+				       PAGE_SIZE, SLAB_PANIC, NULL);
+#endif
+}
+
+pgd_t *pgd_alloc(struct mm_struct *mm)
+{
+	return kmem_cache_alloc(pgd_cachep, PGALLOC_GFP);
+}
+
+void pgd_free(struct mm_struct *mm, pgd_t *pgd)
+{
+	kmem_cache_free(pgd_cachep, pgd);
+}
+
+#if PAGETABLE_LEVELS > 2
+void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
+{
+	set_pud(pud, __pud((unsigned long)pmd));
+}
+
+pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long address)
+{
+	return kmem_cache_alloc(pmd_cachep, PGALLOC_GFP);
+}
+
+void pmd_free(struct mm_struct *mm, pmd_t *pmd)
+{
+	kmem_cache_free(pmd_cachep, pmd);
+}
+#endif /* PAGETABLE_LEVELS > 2 */
diff --git a/arch/sh/mm/pmb.c b/arch/sh/mm/pmb.c
new file mode 100644
index 0000000..7160c9f
--- /dev/null
+++ b/arch/sh/mm/pmb.c
@@ -0,0 +1,903 @@
+/*
+ * arch/sh/mm/pmb.c
+ *
+ * Privileged Space Mapping Buffer (PMB) Support.
+ *
+ * Copyright (C) 2005 - 2011  Paul Mundt
+ * Copyright (C) 2010  Matt Fleming
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/syscore_ops.h>
+#include <linux/cpu.h>
+#include <linux/module.h>
+#include <linux/bitops.h>
+#include <linux/debugfs.h>
+#include <linux/fs.h>
+#include <linux/seq_file.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/spinlock.h>
+#include <linux/vmalloc.h>
+#include <asm/cacheflush.h>
+#include <asm/sizes.h>
+#include <asm/uaccess.h>
+#include <asm/pgtable.h>
+#include <asm/page.h>
+#include <asm/mmu.h>
+#include <asm/mmu_context.h>
+
+struct pmb_entry;
+
+struct pmb_entry {
+	unsigned long vpn;
+	unsigned long ppn;
+	unsigned long flags;
+	unsigned long size;
+
+	raw_spinlock_t lock;
+
+	/*
+	 * 0 .. NR_PMB_ENTRIES for specific entry selection, or
+	 * PMB_NO_ENTRY to search for a free one
+	 */
+	int entry;
+
+	/* Adjacent entry link for contiguous multi-entry mappings */
+	struct pmb_entry *link;
+};
+
+static struct {
+	unsigned long size;
+	int flag;
+} pmb_sizes[] = {
+	{ .size	= SZ_512M, .flag = PMB_SZ_512M, },
+	{ .size = SZ_128M, .flag = PMB_SZ_128M, },
+	{ .size = SZ_64M,  .flag = PMB_SZ_64M,  },
+	{ .size = SZ_16M,  .flag = PMB_SZ_16M,  },
+};
+
+static void pmb_unmap_entry(struct pmb_entry *, int depth);
+
+static DEFINE_RWLOCK(pmb_rwlock);
+static struct pmb_entry pmb_entry_list[NR_PMB_ENTRIES];
+static DECLARE_BITMAP(pmb_map, NR_PMB_ENTRIES);
+
+static unsigned int pmb_iomapping_enabled;
+
+static __always_inline unsigned long mk_pmb_entry(unsigned int entry)
+{
+	return (entry & PMB_E_MASK) << PMB_E_SHIFT;
+}
+
+static __always_inline unsigned long mk_pmb_addr(unsigned int entry)
+{
+	return mk_pmb_entry(entry) | PMB_ADDR;
+}
+
+static __always_inline unsigned long mk_pmb_data(unsigned int entry)
+{
+	return mk_pmb_entry(entry) | PMB_DATA;
+}
+
+static __always_inline unsigned int pmb_ppn_in_range(unsigned long ppn)
+{
+	return ppn >= __pa(memory_start) && ppn < __pa(memory_end);
+}
+
+/*
+ * Ensure that the PMB entries match our cache configuration.
+ *
+ * When we are in 32-bit address extended mode, CCR.CB becomes
+ * invalid, so care must be taken to manually adjust cacheable
+ * translations.
+ */
+static __always_inline unsigned long pmb_cache_flags(void)
+{
+	unsigned long flags = 0;
+
+#if defined(CONFIG_CACHE_OFF)
+	flags |= PMB_WT | PMB_UB;
+#elif defined(CONFIG_CACHE_WRITETHROUGH)
+	flags |= PMB_C | PMB_WT | PMB_UB;
+#elif defined(CONFIG_CACHE_WRITEBACK)
+	flags |= PMB_C;
+#endif
+
+	return flags;
+}
+
+/*
+ * Convert typical pgprot value to the PMB equivalent
+ */
+static inline unsigned long pgprot_to_pmb_flags(pgprot_t prot)
+{
+	unsigned long pmb_flags = 0;
+	u64 flags = pgprot_val(prot);
+
+	if (flags & _PAGE_CACHABLE)
+		pmb_flags |= PMB_C;
+	if (flags & _PAGE_WT)
+		pmb_flags |= PMB_WT | PMB_UB;
+
+	return pmb_flags;
+}
+
+static inline bool pmb_can_merge(struct pmb_entry *a, struct pmb_entry *b)
+{
+	return (b->vpn == (a->vpn + a->size)) &&
+	       (b->ppn == (a->ppn + a->size)) &&
+	       (b->flags == a->flags);
+}
+
+static bool pmb_mapping_exists(unsigned long vaddr, phys_addr_t phys,
+			       unsigned long size)
+{
+	int i;
+
+	read_lock(&pmb_rwlock);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_entry_list); i++) {
+		struct pmb_entry *pmbe, *iter;
+		unsigned long span;
+
+		if (!test_bit(i, pmb_map))
+			continue;
+
+		pmbe = &pmb_entry_list[i];
+
+		/*
+		 * See if VPN and PPN are bounded by an existing mapping.
+		 */
+		if ((vaddr < pmbe->vpn) || (vaddr >= (pmbe->vpn + pmbe->size)))
+			continue;
+		if ((phys < pmbe->ppn) || (phys >= (pmbe->ppn + pmbe->size)))
+			continue;
+
+		/*
+		 * Now see if we're in range of a simple mapping.
+		 */
+		if (size <= pmbe->size) {
+			read_unlock(&pmb_rwlock);
+			return true;
+		}
+
+		span = pmbe->size;
+
+		/*
+		 * Finally for sizes that involve compound mappings, walk
+		 * the chain.
+		 */
+		for (iter = pmbe->link; iter; iter = iter->link)
+			span += iter->size;
+
+		/*
+		 * Nothing else to do if the range requirements are met.
+		 */
+		if (size <= span) {
+			read_unlock(&pmb_rwlock);
+			return true;
+		}
+	}
+
+	read_unlock(&pmb_rwlock);
+	return false;
+}
+
+static bool pmb_size_valid(unsigned long size)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(pmb_sizes); i++)
+		if (pmb_sizes[i].size == size)
+			return true;
+
+	return false;
+}
+
+static inline bool pmb_addr_valid(unsigned long addr, unsigned long size)
+{
+	return (addr >= P1SEG && (addr + size - 1) < P3SEG);
+}
+
+static inline bool pmb_prot_valid(pgprot_t prot)
+{
+	return (pgprot_val(prot) & _PAGE_USER) == 0;
+}
+
+static int pmb_size_to_flags(unsigned long size)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(pmb_sizes); i++)
+		if (pmb_sizes[i].size == size)
+			return pmb_sizes[i].flag;
+
+	return 0;
+}
+
+static int pmb_alloc_entry(void)
+{
+	int pos;
+
+	pos = find_first_zero_bit(pmb_map, NR_PMB_ENTRIES);
+	if (pos >= 0 && pos < NR_PMB_ENTRIES)
+		__set_bit(pos, pmb_map);
+	else
+		pos = -ENOSPC;
+
+	return pos;
+}
+
+static struct pmb_entry *pmb_alloc(unsigned long vpn, unsigned long ppn,
+				   unsigned long flags, int entry)
+{
+	struct pmb_entry *pmbe;
+	unsigned long irqflags;
+	void *ret = NULL;
+	int pos;
+
+	write_lock_irqsave(&pmb_rwlock, irqflags);
+
+	if (entry == PMB_NO_ENTRY) {
+		pos = pmb_alloc_entry();
+		if (unlikely(pos < 0)) {
+			ret = ERR_PTR(pos);
+			goto out;
+		}
+	} else {
+		if (__test_and_set_bit(entry, pmb_map)) {
+			ret = ERR_PTR(-ENOSPC);
+			goto out;
+		}
+
+		pos = entry;
+	}
+
+	write_unlock_irqrestore(&pmb_rwlock, irqflags);
+
+	pmbe = &pmb_entry_list[pos];
+
+	memset(pmbe, 0, sizeof(struct pmb_entry));
+
+	raw_spin_lock_init(&pmbe->lock);
+
+	pmbe->vpn	= vpn;
+	pmbe->ppn	= ppn;
+	pmbe->flags	= flags;
+	pmbe->entry	= pos;
+
+	return pmbe;
+
+out:
+	write_unlock_irqrestore(&pmb_rwlock, irqflags);
+	return ret;
+}
+
+static void pmb_free(struct pmb_entry *pmbe)
+{
+	__clear_bit(pmbe->entry, pmb_map);
+
+	pmbe->entry	= PMB_NO_ENTRY;
+	pmbe->link	= NULL;
+}
+
+/*
+ * Must be run uncached.
+ */
+static void __set_pmb_entry(struct pmb_entry *pmbe)
+{
+	unsigned long addr, data;
+
+	addr = mk_pmb_addr(pmbe->entry);
+	data = mk_pmb_data(pmbe->entry);
+
+	jump_to_uncached();
+
+	/* Set V-bit */
+	__raw_writel(pmbe->vpn | PMB_V, addr);
+	__raw_writel(pmbe->ppn | pmbe->flags | PMB_V, data);
+
+	back_to_cached();
+}
+
+static void __clear_pmb_entry(struct pmb_entry *pmbe)
+{
+	unsigned long addr, data;
+	unsigned long addr_val, data_val;
+
+	addr = mk_pmb_addr(pmbe->entry);
+	data = mk_pmb_data(pmbe->entry);
+
+	addr_val = __raw_readl(addr);
+	data_val = __raw_readl(data);
+
+	/* Clear V-bit */
+	writel_uncached(addr_val & ~PMB_V, addr);
+	writel_uncached(data_val & ~PMB_V, data);
+}
+
+#ifdef CONFIG_PM
+static void set_pmb_entry(struct pmb_entry *pmbe)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&pmbe->lock, flags);
+	__set_pmb_entry(pmbe);
+	raw_spin_unlock_irqrestore(&pmbe->lock, flags);
+}
+#endif /* CONFIG_PM */
+
+int pmb_bolt_mapping(unsigned long vaddr, phys_addr_t phys,
+		     unsigned long size, pgprot_t prot)
+{
+	struct pmb_entry *pmbp, *pmbe;
+	unsigned long orig_addr, orig_size;
+	unsigned long flags, pmb_flags;
+	int i, mapped;
+
+	if (size < SZ_16M)
+		return -EINVAL;
+	if (!pmb_addr_valid(vaddr, size))
+		return -EFAULT;
+	if (pmb_mapping_exists(vaddr, phys, size))
+		return 0;
+
+	orig_addr = vaddr;
+	orig_size = size;
+
+	flush_tlb_kernel_range(vaddr, vaddr + size);
+
+	pmb_flags = pgprot_to_pmb_flags(prot);
+	pmbp = NULL;
+
+	do {
+		for (i = mapped = 0; i < ARRAY_SIZE(pmb_sizes); i++) {
+			if (size < pmb_sizes[i].size)
+				continue;
+
+			pmbe = pmb_alloc(vaddr, phys, pmb_flags |
+					 pmb_sizes[i].flag, PMB_NO_ENTRY);
+			if (IS_ERR(pmbe)) {
+				pmb_unmap_entry(pmbp, mapped);
+				return PTR_ERR(pmbe);
+			}
+
+			raw_spin_lock_irqsave(&pmbe->lock, flags);
+
+			pmbe->size = pmb_sizes[i].size;
+
+			__set_pmb_entry(pmbe);
+
+			phys	+= pmbe->size;
+			vaddr	+= pmbe->size;
+			size	-= pmbe->size;
+
+			/*
+			 * Link adjacent entries that span multiple PMB
+			 * entries for easier tear-down.
+			 */
+			if (likely(pmbp)) {
+				raw_spin_lock_nested(&pmbp->lock,
+						     SINGLE_DEPTH_NESTING);
+				pmbp->link = pmbe;
+				raw_spin_unlock(&pmbp->lock);
+			}
+
+			pmbp = pmbe;
+
+			/*
+			 * Instead of trying smaller sizes on every
+			 * iteration (even if we succeed in allocating
+			 * space), try using pmb_sizes[i].size again.
+			 */
+			i--;
+			mapped++;
+
+			raw_spin_unlock_irqrestore(&pmbe->lock, flags);
+		}
+	} while (size >= SZ_16M);
+
+	flush_cache_vmap(orig_addr, orig_addr + orig_size);
+
+	return 0;
+}
+
+void __iomem *pmb_remap_caller(phys_addr_t phys, unsigned long size,
+			       pgprot_t prot, void *caller)
+{
+	unsigned long vaddr;
+	phys_addr_t offset, last_addr;
+	phys_addr_t align_mask;
+	unsigned long aligned;
+	struct vm_struct *area;
+	int i, ret;
+
+	if (!pmb_iomapping_enabled)
+		return NULL;
+
+	/*
+	 * Small mappings need to go through the TLB.
+	 */
+	if (size < SZ_16M)
+		return ERR_PTR(-EINVAL);
+	if (!pmb_prot_valid(prot))
+		return ERR_PTR(-EINVAL);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_sizes); i++)
+		if (size >= pmb_sizes[i].size)
+			break;
+
+	last_addr = phys + size;
+	align_mask = ~(pmb_sizes[i].size - 1);
+	offset = phys & ~align_mask;
+	phys &= align_mask;
+	aligned = ALIGN(last_addr, pmb_sizes[i].size) - phys;
+
+	/*
+	 * XXX: This should really start from uncached_end, but this
+	 * causes the MMU to reset, so for now we restrict it to the
+	 * 0xb000...0xc000 range.
+	 */
+	area = __get_vm_area_caller(aligned, VM_IOREMAP, 0xb0000000,
+				    P3SEG, caller);
+	if (!area)
+		return NULL;
+
+	area->phys_addr = phys;
+	vaddr = (unsigned long)area->addr;
+
+	ret = pmb_bolt_mapping(vaddr, phys, size, prot);
+	if (unlikely(ret != 0))
+		return ERR_PTR(ret);
+
+	return (void __iomem *)(offset + (char *)vaddr);
+}
+
+int pmb_unmap(void __iomem *addr)
+{
+	struct pmb_entry *pmbe = NULL;
+	unsigned long vaddr = (unsigned long __force)addr;
+	int i, found = 0;
+
+	read_lock(&pmb_rwlock);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_entry_list); i++) {
+		if (test_bit(i, pmb_map)) {
+			pmbe = &pmb_entry_list[i];
+			if (pmbe->vpn == vaddr) {
+				found = 1;
+				break;
+			}
+		}
+	}
+
+	read_unlock(&pmb_rwlock);
+
+	if (found) {
+		pmb_unmap_entry(pmbe, NR_PMB_ENTRIES);
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+static void __pmb_unmap_entry(struct pmb_entry *pmbe, int depth)
+{
+	do {
+		struct pmb_entry *pmblink = pmbe;
+
+		/*
+		 * We may be called before this pmb_entry has been
+		 * entered into the PMB table via set_pmb_entry(), but
+		 * that's OK because we've allocated a unique slot for
+		 * this entry in pmb_alloc() (even if we haven't filled
+		 * it yet).
+		 *
+		 * Therefore, calling __clear_pmb_entry() is safe as no
+		 * other mapping can be using that slot.
+		 */
+		__clear_pmb_entry(pmbe);
+
+		flush_cache_vunmap(pmbe->vpn, pmbe->vpn + pmbe->size);
+
+		pmbe = pmblink->link;
+
+		pmb_free(pmblink);
+	} while (pmbe && --depth);
+}
+
+static void pmb_unmap_entry(struct pmb_entry *pmbe, int depth)
+{
+	unsigned long flags;
+
+	if (unlikely(!pmbe))
+		return;
+
+	write_lock_irqsave(&pmb_rwlock, flags);
+	__pmb_unmap_entry(pmbe, depth);
+	write_unlock_irqrestore(&pmb_rwlock, flags);
+}
+
+static void __init pmb_notify(void)
+{
+	int i;
+
+	pr_info("PMB: boot mappings:\n");
+
+	read_lock(&pmb_rwlock);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_entry_list); i++) {
+		struct pmb_entry *pmbe;
+
+		if (!test_bit(i, pmb_map))
+			continue;
+
+		pmbe = &pmb_entry_list[i];
+
+		pr_info("       0x%08lx -> 0x%08lx [ %4ldMB %2scached ]\n",
+			pmbe->vpn >> PAGE_SHIFT, pmbe->ppn >> PAGE_SHIFT,
+			pmbe->size >> 20, (pmbe->flags & PMB_C) ? "" : "un");
+	}
+
+	read_unlock(&pmb_rwlock);
+}
+
+/*
+ * Sync our software copy of the PMB mappings with those in hardware. The
+ * mappings in the hardware PMB were either set up by the bootloader or
+ * very early on by the kernel.
+ */
+static void __init pmb_synchronize(void)
+{
+	struct pmb_entry *pmbp = NULL;
+	int i, j;
+
+	/*
+	 * Run through the initial boot mappings, log the established
+	 * ones, and blow away anything that falls outside of the valid
+	 * PPN range. Specifically, we only care about existing mappings
+	 * that impact the cached/uncached sections.
+	 *
+	 * Note that touching these can be a bit of a minefield; the boot
+	 * loader can establish multi-page mappings with the same caching
+	 * attributes, so we need to ensure that we aren't modifying a
+	 * mapping that we're presently executing from, or may execute
+	 * from in the case of straddling page boundaries.
+	 *
+	 * In the future we will have to tidy up after the boot loader by
+	 * jumping between the cached and uncached mappings and tearing
+	 * down alternating mappings while executing from the other.
+	 */
+	for (i = 0; i < NR_PMB_ENTRIES; i++) {
+		unsigned long addr, data;
+		unsigned long addr_val, data_val;
+		unsigned long ppn, vpn, flags;
+		unsigned long irqflags;
+		unsigned int size;
+		struct pmb_entry *pmbe;
+
+		addr = mk_pmb_addr(i);
+		data = mk_pmb_data(i);
+
+		addr_val = __raw_readl(addr);
+		data_val = __raw_readl(data);
+
+		/*
+		 * Skip over any bogus entries
+		 */
+		if (!(data_val & PMB_V) || !(addr_val & PMB_V))
+			continue;
+
+		ppn = data_val & PMB_PFN_MASK;
+		vpn = addr_val & PMB_PFN_MASK;
+
+		/*
+		 * Only preserve in-range mappings.
+		 */
+		if (!pmb_ppn_in_range(ppn)) {
+			/*
+			 * Invalidate anything out of bounds.
+			 */
+			writel_uncached(addr_val & ~PMB_V, addr);
+			writel_uncached(data_val & ~PMB_V, data);
+			continue;
+		}
+
+		/*
+		 * Update the caching attributes if necessary
+		 */
+		if (data_val & PMB_C) {
+			data_val &= ~PMB_CACHE_MASK;
+			data_val |= pmb_cache_flags();
+
+			writel_uncached(data_val, data);
+		}
+
+		size = data_val & PMB_SZ_MASK;
+		flags = size | (data_val & PMB_CACHE_MASK);
+
+		pmbe = pmb_alloc(vpn, ppn, flags, i);
+		if (IS_ERR(pmbe)) {
+			WARN_ON_ONCE(1);
+			continue;
+		}
+
+		raw_spin_lock_irqsave(&pmbe->lock, irqflags);
+
+		for (j = 0; j < ARRAY_SIZE(pmb_sizes); j++)
+			if (pmb_sizes[j].flag == size)
+				pmbe->size = pmb_sizes[j].size;
+
+		if (pmbp) {
+			raw_spin_lock_nested(&pmbp->lock, SINGLE_DEPTH_NESTING);
+			/*
+			 * Compare the previous entry against the current one to
+			 * see if the entries span a contiguous mapping. If so,
+			 * setup the entry links accordingly. Compound mappings
+			 * are later coalesced.
+			 */
+			if (pmb_can_merge(pmbp, pmbe))
+				pmbp->link = pmbe;
+			raw_spin_unlock(&pmbp->lock);
+		}
+
+		pmbp = pmbe;
+
+		raw_spin_unlock_irqrestore(&pmbe->lock, irqflags);
+	}
+}
+
+static void __init pmb_merge(struct pmb_entry *head)
+{
+	unsigned long span, newsize;
+	struct pmb_entry *tail;
+	int i = 1, depth = 0;
+
+	span = newsize = head->size;
+
+	tail = head->link;
+	while (tail) {
+		span += tail->size;
+
+		if (pmb_size_valid(span)) {
+			newsize = span;
+			depth = i;
+		}
+
+		/* This is the end of the line.. */
+		if (!tail->link)
+			break;
+
+		tail = tail->link;
+		i++;
+	}
+
+	/*
+	 * The merged page size must be valid.
+	 */
+	if (!depth || !pmb_size_valid(newsize))
+		return;
+
+	head->flags &= ~PMB_SZ_MASK;
+	head->flags |= pmb_size_to_flags(newsize);
+
+	head->size = newsize;
+
+	__pmb_unmap_entry(head->link, depth);
+	__set_pmb_entry(head);
+}
+
+static void __init pmb_coalesce(void)
+{
+	unsigned long flags;
+	int i;
+
+	write_lock_irqsave(&pmb_rwlock, flags);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_entry_list); i++) {
+		struct pmb_entry *pmbe;
+
+		if (!test_bit(i, pmb_map))
+			continue;
+
+		pmbe = &pmb_entry_list[i];
+
+		/*
+		 * We're only interested in compound mappings
+		 */
+		if (!pmbe->link)
+			continue;
+
+		/*
+		 * Nothing to do if it already uses the largest possible
+		 * page size.
+		 */
+		if (pmbe->size == SZ_512M)
+			continue;
+
+		pmb_merge(pmbe);
+	}
+
+	write_unlock_irqrestore(&pmb_rwlock, flags);
+}
+
+#ifdef CONFIG_UNCACHED_MAPPING
+static void __init pmb_resize(void)
+{
+	int i;
+
+	/*
+	 * If the uncached mapping was constructed by the kernel, it will
+	 * already be a reasonable size.
+	 */
+	if (uncached_size == SZ_16M)
+		return;
+
+	read_lock(&pmb_rwlock);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_entry_list); i++) {
+		struct pmb_entry *pmbe;
+		unsigned long flags;
+
+		if (!test_bit(i, pmb_map))
+			continue;
+
+		pmbe = &pmb_entry_list[i];
+
+		if (pmbe->vpn != uncached_start)
+			continue;
+
+		/*
+		 * Found it, now resize it.
+		 */
+		raw_spin_lock_irqsave(&pmbe->lock, flags);
+
+		pmbe->size = SZ_16M;
+		pmbe->flags &= ~PMB_SZ_MASK;
+		pmbe->flags |= pmb_size_to_flags(pmbe->size);
+
+		uncached_resize(pmbe->size);
+
+		__set_pmb_entry(pmbe);
+
+		raw_spin_unlock_irqrestore(&pmbe->lock, flags);
+	}
+
+	read_unlock(&pmb_rwlock);
+}
+#endif
+
+static int __init early_pmb(char *p)
+{
+	if (!p)
+		return 0;
+
+	if (strstr(p, "iomap"))
+		pmb_iomapping_enabled = 1;
+
+	return 0;
+}
+early_param("pmb", early_pmb);
+
+void __init pmb_init(void)
+{
+	/* Synchronize software state */
+	pmb_synchronize();
+
+	/* Attempt to combine compound mappings */
+	pmb_coalesce();
+
+#ifdef CONFIG_UNCACHED_MAPPING
+	/* Resize initial mappings, if necessary */
+	pmb_resize();
+#endif
+
+	/* Log them */
+	pmb_notify();
+
+	writel_uncached(0, PMB_IRMCR);
+
+	/* Flush out the TLB */
+	local_flush_tlb_all();
+	ctrl_barrier();
+}
+
+bool __in_29bit_mode(void)
+{
+        return (__raw_readl(PMB_PASCR) & PASCR_SE) == 0;
+}
+
+static int pmb_seq_show(struct seq_file *file, void *iter)
+{
+	int i;
+
+	seq_printf(file, "V: Valid, C: Cacheable, WT: Write-Through\n"
+			 "CB: Copy-Back, B: Buffered, UB: Unbuffered\n");
+	seq_printf(file, "ety   vpn  ppn  size   flags\n");
+
+	for (i = 0; i < NR_PMB_ENTRIES; i++) {
+		unsigned long addr, data;
+		unsigned int size;
+		char *sz_str = NULL;
+
+		addr = __raw_readl(mk_pmb_addr(i));
+		data = __raw_readl(mk_pmb_data(i));
+
+		size = data & PMB_SZ_MASK;
+		sz_str = (size == PMB_SZ_16M)  ? " 16MB":
+			 (size == PMB_SZ_64M)  ? " 64MB":
+			 (size == PMB_SZ_128M) ? "128MB":
+					         "512MB";
+
+		/* 02: V 0x88 0x08 128MB C CB  B */
+		seq_printf(file, "%02d: %c 0x%02lx 0x%02lx %s %c %s %s\n",
+			   i, ((addr & PMB_V) && (data & PMB_V)) ? 'V' : ' ',
+			   (addr >> 24) & 0xff, (data >> 24) & 0xff,
+			   sz_str, (data & PMB_C) ? 'C' : ' ',
+			   (data & PMB_WT) ? "WT" : "CB",
+			   (data & PMB_UB) ? "UB" : " B");
+	}
+
+	return 0;
+}
+
+static int pmb_debugfs_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, pmb_seq_show, NULL);
+}
+
+static const struct file_operations pmb_debugfs_fops = {
+	.owner		= THIS_MODULE,
+	.open		= pmb_debugfs_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int __init pmb_debugfs_init(void)
+{
+	struct dentry *dentry;
+
+	dentry = debugfs_create_file("pmb", S_IFREG | S_IRUGO,
+				     arch_debugfs_dir, NULL, &pmb_debugfs_fops);
+	if (!dentry)
+		return -ENOMEM;
+
+	return 0;
+}
+subsys_initcall(pmb_debugfs_init);
+
+#ifdef CONFIG_PM
+static void pmb_syscore_resume(void)
+{
+	struct pmb_entry *pmbe;
+	int i;
+
+	read_lock(&pmb_rwlock);
+
+	for (i = 0; i < ARRAY_SIZE(pmb_entry_list); i++) {
+		if (test_bit(i, pmb_map)) {
+			pmbe = &pmb_entry_list[i];
+			set_pmb_entry(pmbe);
+		}
+	}
+
+	read_unlock(&pmb_rwlock);
+}
+
+static struct syscore_ops pmb_syscore_ops = {
+	.resume = pmb_syscore_resume,
+};
+
+static int __init pmb_sysdev_init(void)
+{
+	register_syscore_ops(&pmb_syscore_ops);
+	return 0;
+}
+subsys_initcall(pmb_sysdev_init);
+#endif
diff --git a/arch/sh/mm/sram.c b/arch/sh/mm/sram.c
new file mode 100644
index 0000000..2d8fa71
--- /dev/null
+++ b/arch/sh/mm/sram.c
@@ -0,0 +1,35 @@
+/*
+ * SRAM pool for tiny memories not otherwise managed.
+ *
+ * Copyright (C) 2010  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <asm/sram.h>
+
+/*
+ * This provides a standard SRAM pool for tiny memories that can be
+ * added either by the CPU or the platform code. Typical SRAM sizes
+ * to be inserted in to the pool will generally be less than the page
+ * size, with anything more reasonably sized handled as a NUMA memory
+ * node.
+ */
+struct gen_pool *sram_pool;
+
+static int __init sram_pool_init(void)
+{
+	/*
+	 * This is a global pool, we don't care about node locality.
+	 */
+	sram_pool = gen_pool_create(1, -1);
+	if (unlikely(!sram_pool))
+		return -ENOMEM;
+
+	return 0;
+}
+core_initcall(sram_pool_init);
diff --git a/arch/sh/mm/tlb-debugfs.c b/arch/sh/mm/tlb-debugfs.c
new file mode 100644
index 0000000..dea637a
--- /dev/null
+++ b/arch/sh/mm/tlb-debugfs.c
@@ -0,0 +1,172 @@
+/*
+ * arch/sh/mm/tlb-debugfs.c
+ *
+ * debugfs ops for SH-4 ITLB/UTLBs.
+ *
+ * Copyright (C) 2010  Matt Fleming
+ *
+ * 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.
+ */
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+#include <asm/processor.h>
+#include <asm/mmu_context.h>
+#include <asm/tlbflush.h>
+
+enum tlb_type {
+	TLB_TYPE_ITLB,
+	TLB_TYPE_UTLB,
+};
+
+static struct {
+	int bits;
+	const char *size;
+} tlb_sizes[] = {
+	{ 0x0, "  1KB" },
+	{ 0x1, "  4KB" },
+	{ 0x2, "  8KB" },
+	{ 0x4, " 64KB" },
+	{ 0x5, "256KB" },
+	{ 0x7, "  1MB" },
+	{ 0x8, "  4MB" },
+	{ 0xc, " 64MB" },
+};
+
+static int tlb_seq_show(struct seq_file *file, void *iter)
+{
+	unsigned int tlb_type = (unsigned int)file->private;
+	unsigned long addr1, addr2, data1, data2;
+	unsigned long flags;
+	unsigned long mmucr;
+	unsigned int nentries, entry;
+	unsigned int urb;
+
+	mmucr = __raw_readl(MMUCR);
+	if ((mmucr & 0x1) == 0) {
+		seq_printf(file, "address translation disabled\n");
+		return 0;
+	}
+
+	if (tlb_type == TLB_TYPE_ITLB) {
+		addr1 = MMU_ITLB_ADDRESS_ARRAY;
+		addr2 = MMU_ITLB_ADDRESS_ARRAY2;
+		data1 = MMU_ITLB_DATA_ARRAY;
+		data2 = MMU_ITLB_DATA_ARRAY2;
+		nentries = 4;
+	} else {
+		addr1 = MMU_UTLB_ADDRESS_ARRAY;
+		addr2 = MMU_UTLB_ADDRESS_ARRAY2;
+		data1 = MMU_UTLB_DATA_ARRAY;
+		data2 = MMU_UTLB_DATA_ARRAY2;
+		nentries = 64;
+	}
+
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	urb = (mmucr & MMUCR_URB) >> MMUCR_URB_SHIFT;
+
+	/* Make the "entry >= urb" test fail. */
+	if (urb == 0)
+		urb = MMUCR_URB_NENTRIES + 1;
+
+	if (tlb_type == TLB_TYPE_ITLB) {
+		addr1 = MMU_ITLB_ADDRESS_ARRAY;
+		addr2 = MMU_ITLB_ADDRESS_ARRAY2;
+		data1 = MMU_ITLB_DATA_ARRAY;
+		data2 = MMU_ITLB_DATA_ARRAY2;
+		nentries = 4;
+	} else {
+		addr1 = MMU_UTLB_ADDRESS_ARRAY;
+		addr2 = MMU_UTLB_ADDRESS_ARRAY2;
+		data1 = MMU_UTLB_DATA_ARRAY;
+		data2 = MMU_UTLB_DATA_ARRAY2;
+		nentries = 64;
+	}
+
+	seq_printf(file, "entry:     vpn        ppn     asid  size valid wired\n");
+
+	for (entry = 0; entry < nentries; entry++) {
+		unsigned long vpn, ppn, asid, size;
+		unsigned long valid;
+		unsigned long val;
+		const char *sz = "    ?";
+		int i;
+
+		val = __raw_readl(addr1 | (entry << MMU_TLB_ENTRY_SHIFT));
+		ctrl_barrier();
+		vpn = val & 0xfffffc00;
+		valid = val & 0x100;
+
+		val = __raw_readl(addr2 | (entry << MMU_TLB_ENTRY_SHIFT));
+		ctrl_barrier();
+		asid = val & MMU_CONTEXT_ASID_MASK;
+
+		val = __raw_readl(data1 | (entry << MMU_TLB_ENTRY_SHIFT));
+		ctrl_barrier();
+		ppn = (val & 0x0ffffc00) << 4;
+
+		val = __raw_readl(data2 | (entry << MMU_TLB_ENTRY_SHIFT));
+		ctrl_barrier();
+		size = (val & 0xf0) >> 4;
+
+		for (i = 0; i < ARRAY_SIZE(tlb_sizes); i++) {
+			if (tlb_sizes[i].bits == size)
+				break;
+		}
+
+		if (i != ARRAY_SIZE(tlb_sizes))
+			sz = tlb_sizes[i].size;
+
+		seq_printf(file, "%2d:    0x%08lx 0x%08lx %5lu %s   %s     %s\n",
+			   entry, vpn, ppn, asid,
+			   sz, valid ? "V" : "-",
+			   (urb <= entry) ? "W" : "-");
+	}
+
+	back_to_cached();
+	local_irq_restore(flags);
+
+	return 0;
+}
+
+static int tlb_debugfs_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, tlb_seq_show, inode->i_private);
+}
+
+static const struct file_operations tlb_debugfs_fops = {
+	.owner		= THIS_MODULE,
+	.open		= tlb_debugfs_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int __init tlb_debugfs_init(void)
+{
+	struct dentry *itlb, *utlb;
+
+	itlb = debugfs_create_file("itlb", S_IRUSR, arch_debugfs_dir,
+				   (unsigned int *)TLB_TYPE_ITLB,
+				   &tlb_debugfs_fops);
+	if (unlikely(!itlb))
+		return -ENOMEM;
+
+	utlb = debugfs_create_file("utlb", S_IRUSR, arch_debugfs_dir,
+				   (unsigned int *)TLB_TYPE_UTLB,
+				   &tlb_debugfs_fops);
+	if (unlikely(!utlb)) {
+		debugfs_remove(itlb);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+module_init(tlb_debugfs_init);
+
+MODULE_LICENSE("GPL v2");
diff --git a/arch/sh/mm/tlb-pteaex.c b/arch/sh/mm/tlb-pteaex.c
new file mode 100644
index 0000000..4db21ad
--- /dev/null
+++ b/arch/sh/mm/tlb-pteaex.c
@@ -0,0 +1,106 @@
+/*
+ * arch/sh/mm/tlb-pteaex.c
+ *
+ * TLB operations for SH-X3 CPUs featuring PTE ASID Extensions.
+ *
+ * Copyright (C) 2009 Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/io.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t pte)
+{
+	unsigned long flags, pteval, vpn;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (vma && current->active_mm != vma->vm_mm)
+		return;
+
+	local_irq_save(flags);
+
+	/* Set PTEH register */
+	vpn = address & MMU_VPN_MASK;
+	__raw_writel(vpn, MMU_PTEH);
+
+	/* Set PTEAEX */
+	__raw_writel(get_asid(), MMU_PTEAEX);
+
+	pteval = pte.pte_low;
+
+	/* Set PTEA register */
+#ifdef CONFIG_X2TLB
+	/*
+	 * For the extended mode TLB this is trivial, only the ESZ and
+	 * EPR bits need to be written out to PTEA, with the remainder of
+	 * the protection bits (with the exception of the compat-mode SZ
+	 * and PR bits, which are cleared) being written out in PTEL.
+	 */
+	__raw_writel(pte.pte_high, MMU_PTEA);
+#endif
+
+	/* Set PTEL register */
+	pteval &= _PAGE_FLAGS_HARDWARE_MASK; /* drop software flags */
+#ifdef CONFIG_CACHE_WRITETHROUGH
+	pteval |= _PAGE_WT;
+#endif
+	/* conveniently, we want all the software flags to be 0 anyway */
+	__raw_writel(pteval, MMU_PTEL);
+
+	/* Load the TLB */
+	asm volatile("ldtlb": /* no output */ : /* no input */ : "memory");
+	local_irq_restore(flags);
+}
+
+/*
+ * While SH-X2 extended TLB mode splits out the memory-mapped I/UTLB
+ * data arrays, SH-X3 cores with PTEAEX split out the memory-mapped
+ * address arrays. In compat mode the second array is inaccessible, while
+ * in extended mode, the legacy 8-bit ASID field in address array 1 has
+ * undefined behaviour.
+ */
+void local_flush_tlb_one(unsigned long asid, unsigned long page)
+{
+	jump_to_uncached();
+	__raw_writel(page, MMU_UTLB_ADDRESS_ARRAY | MMU_PAGE_ASSOC_BIT);
+	__raw_writel(asid, MMU_UTLB_ADDRESS_ARRAY2 | MMU_PAGE_ASSOC_BIT);
+	__raw_writel(page, MMU_ITLB_ADDRESS_ARRAY | MMU_PAGE_ASSOC_BIT);
+	__raw_writel(asid, MMU_ITLB_ADDRESS_ARRAY2 | MMU_PAGE_ASSOC_BIT);
+	back_to_cached();
+}
+
+void local_flush_tlb_all(void)
+{
+	unsigned long flags, status;
+	int i;
+
+	/*
+	 * Flush all the TLB.
+	 */
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	status = __raw_readl(MMUCR);
+	status = ((status & MMUCR_URB) >> MMUCR_URB_SHIFT);
+
+	if (status == 0)
+		status = MMUCR_URB_NENTRIES;
+
+	for (i = 0; i < status; i++)
+		__raw_writel(0x0, MMU_UTLB_ADDRESS_ARRAY | (i << 8));
+
+	for (i = 0; i < 4; i++)
+		__raw_writel(0x0, MMU_ITLB_ADDRESS_ARRAY | (i << 8));
+
+	back_to_cached();
+	ctrl_barrier();
+	local_irq_restore(flags);
+}
diff --git a/arch/sh/mm/tlb-sh3.c b/arch/sh/mm/tlb-sh3.c
new file mode 100644
index 0000000..6554fb4
--- /dev/null
+++ b/arch/sh/mm/tlb-sh3.c
@@ -0,0 +1,97 @@
+/*
+ * arch/sh/mm/tlb-sh3.c
+ *
+ * SH-3 specific TLB operations
+ *
+ * Copyright (C) 1999  Niibe Yutaka
+ * Copyright (C) 2002  Paul Mundt
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+#include <linux/signal.h>
+#include <linux/sched.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/interrupt.h>
+
+#include <asm/io.h>
+#include <asm/uaccess.h>
+#include <asm/pgalloc.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t pte)
+{
+	unsigned long flags, pteval, vpn;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (vma && current->active_mm != vma->vm_mm)
+		return;
+
+	local_irq_save(flags);
+
+	/* Set PTEH register */
+	vpn = (address & MMU_VPN_MASK) | get_asid();
+	__raw_writel(vpn, MMU_PTEH);
+
+	pteval = pte_val(pte);
+
+	/* Set PTEL register */
+	pteval &= _PAGE_FLAGS_HARDWARE_MASK; /* drop software flags */
+	/* conveniently, we want all the software flags to be 0 anyway */
+	__raw_writel(pteval, MMU_PTEL);
+
+	/* Load the TLB */
+	asm volatile("ldtlb": /* no output */ : /* no input */ : "memory");
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_one(unsigned long asid, unsigned long page)
+{
+	unsigned long addr, data;
+	int i, ways = MMU_NTLB_WAYS;
+
+	/*
+	 * NOTE: PTEH.ASID should be set to this MM
+	 *       _AND_ we need to write ASID to the array.
+	 *
+	 * It would be simple if we didn't need to set PTEH.ASID...
+	 */
+	addr = MMU_TLB_ADDRESS_ARRAY | (page & 0x1F000);
+	data = (page & 0xfffe0000) | asid; /* VALID bit is off */
+
+	if ((current_cpu_data.flags & CPU_HAS_MMU_PAGE_ASSOC)) {
+		addr |= MMU_PAGE_ASSOC_BIT;
+		ways = 1;	/* we already know the way .. */
+	}
+
+	for (i = 0; i < ways; i++)
+		__raw_writel(data, addr + (i << 8));
+}
+
+void local_flush_tlb_all(void)
+{
+	unsigned long flags, status;
+
+	/*
+	 * Flush all the TLB.
+	 *
+	 * Write to the MMU control register's bit:
+	 *	TF-bit for SH-3, TI-bit for SH-4.
+	 *      It's same position, bit #2.
+	 */
+	local_irq_save(flags);
+	status = __raw_readl(MMUCR);
+	status |= 0x04;
+	__raw_writel(status, MMUCR);
+	ctrl_barrier();
+	local_irq_restore(flags);
+}
diff --git a/arch/sh/mm/tlb-sh4.c b/arch/sh/mm/tlb-sh4.c
new file mode 100644
index 0000000..d42dd7e
--- /dev/null
+++ b/arch/sh/mm/tlb-sh4.c
@@ -0,0 +1,109 @@
+/*
+ * arch/sh/mm/tlb-sh4.c
+ *
+ * SH-4 specific TLB operations
+ *
+ * Copyright (C) 1999  Niibe Yutaka
+ * Copyright (C) 2002 - 2007 Paul Mundt
+ *
+ * Released under the terms of the GNU GPL v2.0.
+ */
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/io.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t pte)
+{
+	unsigned long flags, pteval, vpn;
+
+	/*
+	 * Handle debugger faulting in for debugee.
+	 */
+	if (vma && current->active_mm != vma->vm_mm)
+		return;
+
+	local_irq_save(flags);
+
+	/* Set PTEH register */
+	vpn = (address & MMU_VPN_MASK) | get_asid();
+	__raw_writel(vpn, MMU_PTEH);
+
+	pteval = pte.pte_low;
+
+	/* Set PTEA register */
+#ifdef CONFIG_X2TLB
+	/*
+	 * For the extended mode TLB this is trivial, only the ESZ and
+	 * EPR bits need to be written out to PTEA, with the remainder of
+	 * the protection bits (with the exception of the compat-mode SZ
+	 * and PR bits, which are cleared) being written out in PTEL.
+	 */
+	__raw_writel(pte.pte_high, MMU_PTEA);
+#else
+	if (cpu_data->flags & CPU_HAS_PTEA) {
+		/* The last 3 bits and the first one of pteval contains
+		 * the PTEA timing control and space attribute bits
+		 */
+		__raw_writel(copy_ptea_attributes(pteval), MMU_PTEA);
+	}
+#endif
+
+	/* Set PTEL register */
+	pteval &= _PAGE_FLAGS_HARDWARE_MASK; /* drop software flags */
+#ifdef CONFIG_CACHE_WRITETHROUGH
+	pteval |= _PAGE_WT;
+#endif
+	/* conveniently, we want all the software flags to be 0 anyway */
+	__raw_writel(pteval, MMU_PTEL);
+
+	/* Load the TLB */
+	asm volatile("ldtlb": /* no output */ : /* no input */ : "memory");
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_one(unsigned long asid, unsigned long page)
+{
+	unsigned long addr, data;
+
+	/*
+	 * NOTE: PTEH.ASID should be set to this MM
+	 *       _AND_ we need to write ASID to the array.
+	 *
+	 * It would be simple if we didn't need to set PTEH.ASID...
+	 */
+	addr = MMU_UTLB_ADDRESS_ARRAY | MMU_PAGE_ASSOC_BIT;
+	data = page | asid; /* VALID bit is off */
+	jump_to_uncached();
+	__raw_writel(data, addr);
+	back_to_cached();
+}
+
+void local_flush_tlb_all(void)
+{
+	unsigned long flags, status;
+	int i;
+
+	/*
+	 * Flush all the TLB.
+	 */
+	local_irq_save(flags);
+	jump_to_uncached();
+
+	status = __raw_readl(MMUCR);
+	status = ((status & MMUCR_URB) >> MMUCR_URB_SHIFT);
+
+	if (status == 0)
+		status = MMUCR_URB_NENTRIES;
+
+	for (i = 0; i < status; i++)
+		__raw_writel(0x0, MMU_UTLB_ADDRESS_ARRAY | (i << 8));
+
+	for (i = 0; i < 4; i++)
+		__raw_writel(0x0, MMU_ITLB_ADDRESS_ARRAY | (i << 8));
+
+	back_to_cached();
+	ctrl_barrier();
+	local_irq_restore(flags);
+}
diff --git a/arch/sh/mm/tlb-sh5.c b/arch/sh/mm/tlb-sh5.c
new file mode 100644
index 0000000..e4bb2a8
--- /dev/null
+++ b/arch/sh/mm/tlb-sh5.c
@@ -0,0 +1,224 @@
+/*
+ * arch/sh/mm/tlb-sh5.c
+ *
+ * Copyright (C) 2003  Paul Mundt <lethal@linux-sh.org>
+ * Copyright (C) 2003  Richard Curnow <richard.curnow@superh.com>
+ *
+ * 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.
+ */
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <asm/page.h>
+#include <asm/tlb.h>
+#include <asm/mmu_context.h>
+
+/**
+ * sh64_tlb_init - Perform initial setup for the DTLB and ITLB.
+ */
+int sh64_tlb_init(void)
+{
+	/* Assign some sane DTLB defaults */
+	cpu_data->dtlb.entries	= 64;
+	cpu_data->dtlb.step	= 0x10;
+
+	cpu_data->dtlb.first	= DTLB_FIXED | cpu_data->dtlb.step;
+	cpu_data->dtlb.next	= cpu_data->dtlb.first;
+
+	cpu_data->dtlb.last	= DTLB_FIXED |
+				  ((cpu_data->dtlb.entries - 1) *
+				   cpu_data->dtlb.step);
+
+	/* And again for the ITLB */
+	cpu_data->itlb.entries	= 64;
+	cpu_data->itlb.step	= 0x10;
+
+	cpu_data->itlb.first	= ITLB_FIXED | cpu_data->itlb.step;
+	cpu_data->itlb.next	= cpu_data->itlb.first;
+	cpu_data->itlb.last	= ITLB_FIXED |
+				  ((cpu_data->itlb.entries - 1) *
+				   cpu_data->itlb.step);
+
+	return 0;
+}
+
+/**
+ * sh64_next_free_dtlb_entry - Find the next available DTLB entry
+ */
+unsigned long long sh64_next_free_dtlb_entry(void)
+{
+	return cpu_data->dtlb.next;
+}
+
+/**
+ * sh64_get_wired_dtlb_entry - Allocate a wired (locked-in) entry in the DTLB
+ */
+unsigned long long sh64_get_wired_dtlb_entry(void)
+{
+	unsigned long long entry = sh64_next_free_dtlb_entry();
+
+	cpu_data->dtlb.first += cpu_data->dtlb.step;
+	cpu_data->dtlb.next  += cpu_data->dtlb.step;
+
+	return entry;
+}
+
+/**
+ * sh64_put_wired_dtlb_entry - Free a wired (locked-in) entry in the DTLB.
+ *
+ * @entry:	Address of TLB slot.
+ *
+ * Works like a stack, last one to allocate must be first one to free.
+ */
+int sh64_put_wired_dtlb_entry(unsigned long long entry)
+{
+	__flush_tlb_slot(entry);
+
+	/*
+	 * We don't do any particularly useful tracking of wired entries,
+	 * so this approach works like a stack .. last one to be allocated
+	 * has to be the first one to be freed.
+	 *
+	 * We could potentially load wired entries into a list and work on
+	 * rebalancing the list periodically (which also entails moving the
+	 * contents of a TLB entry) .. though I have a feeling that this is
+	 * more trouble than it's worth.
+	 */
+
+	/*
+	 * Entry must be valid .. we don't want any ITLB addresses!
+	 */
+	if (entry <= DTLB_FIXED)
+		return -EINVAL;
+
+	/*
+	 * Next, check if we're within range to be freed. (ie, must be the
+	 * entry beneath the first 'free' entry!
+	 */
+	if (entry < (cpu_data->dtlb.first - cpu_data->dtlb.step))
+		return -EINVAL;
+
+	/* If we are, then bring this entry back into the list */
+	cpu_data->dtlb.first	-= cpu_data->dtlb.step;
+	cpu_data->dtlb.next	= entry;
+
+	return 0;
+}
+
+/**
+ * sh64_setup_tlb_slot - Load up a translation in a wired slot.
+ *
+ * @config_addr:	Address of TLB slot.
+ * @eaddr:		Virtual address.
+ * @asid:		Address Space Identifier.
+ * @paddr:		Physical address.
+ *
+ * Load up a virtual<->physical translation for @eaddr<->@paddr in the
+ * pre-allocated TLB slot @config_addr (see sh64_get_wired_dtlb_entry).
+ */
+void sh64_setup_tlb_slot(unsigned long long config_addr, unsigned long eaddr,
+			 unsigned long asid, unsigned long paddr)
+{
+	unsigned long long pteh, ptel;
+
+	pteh = neff_sign_extend(eaddr);
+	pteh &= PAGE_MASK;
+	pteh |= (asid << PTEH_ASID_SHIFT) | PTEH_VALID;
+	ptel = neff_sign_extend(paddr);
+	ptel &= PAGE_MASK;
+	ptel |= (_PAGE_CACHABLE | _PAGE_READ | _PAGE_WRITE);
+
+	asm volatile("putcfg %0, 1, %1\n\t"
+			"putcfg %0, 0, %2\n"
+			: : "r" (config_addr), "r" (ptel), "r" (pteh));
+}
+
+/**
+ * sh64_teardown_tlb_slot - Teardown a translation.
+ *
+ * @config_addr:	Address of TLB slot.
+ *
+ * Teardown any existing mapping in the TLB slot @config_addr.
+ */
+void sh64_teardown_tlb_slot(unsigned long long config_addr)
+	__attribute__ ((alias("__flush_tlb_slot")));
+
+static int dtlb_entry;
+static unsigned long long dtlb_entries[64];
+
+void tlb_wire_entry(struct vm_area_struct *vma, unsigned long addr, pte_t pte)
+{
+	unsigned long long entry;
+	unsigned long paddr, flags;
+
+	BUG_ON(dtlb_entry == ARRAY_SIZE(dtlb_entries));
+
+	local_irq_save(flags);
+
+	entry = sh64_get_wired_dtlb_entry();
+	dtlb_entries[dtlb_entry++] = entry;
+
+	paddr = pte_val(pte) & _PAGE_FLAGS_HARDWARE_MASK;
+	paddr &= ~PAGE_MASK;
+
+	sh64_setup_tlb_slot(entry, addr, get_asid(), paddr);
+
+	local_irq_restore(flags);
+}
+
+void tlb_unwire_entry(void)
+{
+	unsigned long long entry;
+	unsigned long flags;
+
+	BUG_ON(!dtlb_entry);
+
+	local_irq_save(flags);
+	entry = dtlb_entries[dtlb_entry--];
+
+	sh64_teardown_tlb_slot(entry);
+	sh64_put_wired_dtlb_entry(entry);
+
+	local_irq_restore(flags);
+}
+
+void __update_tlb(struct vm_area_struct *vma, unsigned long address, pte_t pte)
+{
+	unsigned long long ptel;
+	unsigned long long pteh=0;
+	struct tlb_info *tlbp;
+	unsigned long long next;
+	unsigned int fault_code = get_thread_fault_code();
+
+	/* Get PTEL first */
+	ptel = pte.pte_low;
+
+	/*
+	 * Set PTEH register
+	 */
+	pteh = neff_sign_extend(address & MMU_VPN_MASK);
+
+	/* Set the ASID. */
+	pteh |= get_asid() << PTEH_ASID_SHIFT;
+	pteh |= PTEH_VALID;
+
+	/* Set PTEL register, set_pte has performed the sign extension */
+	ptel &= _PAGE_FLAGS_HARDWARE_MASK; /* drop software flags */
+
+	if (fault_code & FAULT_CODE_ITLB)
+		tlbp = &cpu_data->itlb;
+	else
+		tlbp = &cpu_data->dtlb;
+
+	next = tlbp->next;
+	__flush_tlb_slot(next);
+	asm volatile ("putcfg %0,1,%2\n\n\t"
+		      "putcfg %0,0,%1\n"
+		      :  : "r" (next), "r" (pteh), "r" (ptel) );
+
+	next += TLB_STEP;
+	if (next > tlbp->last)
+		next = tlbp->first;
+	tlbp->next = next;
+}
diff --git a/arch/sh/mm/tlb-urb.c b/arch/sh/mm/tlb-urb.c
new file mode 100644
index 0000000..c92ce20
--- /dev/null
+++ b/arch/sh/mm/tlb-urb.c
@@ -0,0 +1,93 @@
+/*
+ * arch/sh/mm/tlb-urb.c
+ *
+ * TLB entry wiring helpers for URB-equipped parts.
+ *
+ * Copyright (C) 2010  Matt Fleming
+ *
+ * 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.
+ */
+#include <linux/mm.h>
+#include <linux/io.h>
+#include <asm/tlb.h>
+#include <asm/mmu_context.h>
+
+/*
+ * Load the entry for 'addr' into the TLB and wire the entry.
+ */
+void tlb_wire_entry(struct vm_area_struct *vma, unsigned long addr, pte_t pte)
+{
+	unsigned long status, flags;
+	int urb;
+
+	local_irq_save(flags);
+
+	status = __raw_readl(MMUCR);
+	urb = (status & MMUCR_URB) >> MMUCR_URB_SHIFT;
+	status &= ~MMUCR_URC;
+
+	/*
+	 * Make sure we're not trying to wire the last TLB entry slot.
+	 */
+	BUG_ON(!--urb);
+
+	urb = urb % MMUCR_URB_NENTRIES;
+
+	/*
+	 * Insert this entry into the highest non-wired TLB slot (via
+	 * the URC field).
+	 */
+	status |= (urb << MMUCR_URC_SHIFT);
+	__raw_writel(status, MMUCR);
+	ctrl_barrier();
+
+	/* Load the entry into the TLB */
+	__update_tlb(vma, addr, pte);
+
+	/* ... and wire it up. */
+	status = __raw_readl(MMUCR);
+
+	status &= ~MMUCR_URB;
+	status |= (urb << MMUCR_URB_SHIFT);
+
+	__raw_writel(status, MMUCR);
+	ctrl_barrier();
+
+	local_irq_restore(flags);
+}
+
+/*
+ * Unwire the last wired TLB entry.
+ *
+ * It should also be noted that it is not possible to wire and unwire
+ * TLB entries in an arbitrary order. If you wire TLB entry N, followed
+ * by entry N+1, you must unwire entry N+1 first, then entry N. In this
+ * respect, it works like a stack or LIFO queue.
+ */
+void tlb_unwire_entry(void)
+{
+	unsigned long status, flags;
+	int urb;
+
+	local_irq_save(flags);
+
+	status = __raw_readl(MMUCR);
+	urb = (status & MMUCR_URB) >> MMUCR_URB_SHIFT;
+	status &= ~MMUCR_URB;
+
+	/*
+	 * Make sure we're not trying to unwire a TLB entry when none
+	 * have been wired.
+	 */
+	BUG_ON(urb++ == MMUCR_URB_NENTRIES);
+
+	urb = urb % MMUCR_URB_NENTRIES;
+
+	status |= (urb << MMUCR_URB_SHIFT);
+	__raw_writel(status, MMUCR);
+	ctrl_barrier();
+
+	local_irq_restore(flags);
+}
diff --git a/arch/sh/mm/tlbex_32.c b/arch/sh/mm/tlbex_32.c
new file mode 100644
index 0000000..382262d
--- /dev/null
+++ b/arch/sh/mm/tlbex_32.c
@@ -0,0 +1,78 @@
+/*
+ * TLB miss handler for SH with an MMU.
+ *
+ *  Copyright (C) 1999  Niibe Yutaka
+ *  Copyright (C) 2003 - 2012  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/kprobes.h>
+#include <linux/kdebug.h>
+#include <asm/mmu_context.h>
+#include <asm/thread_info.h>
+
+/*
+ * Called with interrupts disabled.
+ */
+asmlinkage int __kprobes
+handle_tlbmiss(struct pt_regs *regs, unsigned long error_code,
+	       unsigned long address)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+	pte_t entry;
+
+	/*
+	 * We don't take page faults for P1, P2, and parts of P4, these
+	 * are always mapped, whether it be due to legacy behaviour in
+	 * 29-bit mode, or due to PMB configuration in 32-bit mode.
+	 */
+	if (address >= P3SEG && address < P3_ADDR_MAX) {
+		pgd = pgd_offset_k(address);
+	} else {
+		if (unlikely(address >= TASK_SIZE || !current->mm))
+			return 1;
+
+		pgd = pgd_offset(current->mm, address);
+	}
+
+	pud = pud_offset(pgd, address);
+	if (pud_none_or_clear_bad(pud))
+		return 1;
+	pmd = pmd_offset(pud, address);
+	if (pmd_none_or_clear_bad(pmd))
+		return 1;
+	pte = pte_offset_kernel(pmd, address);
+	entry = *pte;
+	if (unlikely(pte_none(entry) || pte_not_present(entry)))
+		return 1;
+	if (unlikely(error_code && !pte_write(entry)))
+		return 1;
+
+	if (error_code)
+		entry = pte_mkdirty(entry);
+	entry = pte_mkyoung(entry);
+
+	set_pte(pte, entry);
+
+#if defined(CONFIG_CPU_SH4) && !defined(CONFIG_SMP)
+	/*
+	 * SH-4 does not set MMUCR.RC to the corresponding TLB entry in
+	 * the case of an initial page write exception, so we need to
+	 * flush it in order to avoid potential TLB entry duplication.
+	 */
+	if (error_code == FAULT_CODE_INITIAL)
+		local_flush_tlb_one(get_asid(), address & PAGE_MASK);
+#endif
+
+	set_thread_fault_code(error_code);
+	update_mmu_cache(NULL, address, pte);
+
+	return 0;
+}
diff --git a/arch/sh/mm/tlbex_64.c b/arch/sh/mm/tlbex_64.c
new file mode 100644
index 0000000..8557548
--- /dev/null
+++ b/arch/sh/mm/tlbex_64.c
@@ -0,0 +1,166 @@
+/*
+ * The SH64 TLB miss.
+ *
+ * Original code from fault.c
+ * Copyright (C) 2000, 2001  Paolo Alberelli
+ *
+ * Fast PTE->TLB refill path
+ * Copyright (C) 2003 Richard.Curnow@superh.com
+ *
+ * IMPORTANT NOTES :
+ * The do_fast_page_fault function is called from a context in entry.S
+ * where very few registers have been saved.  In particular, the code in
+ * this file must be compiled not to use ANY caller-save registers that
+ * are not part of the restricted save set.  Also, it means that code in
+ * this file must not make calls to functions elsewhere in the kernel, or
+ * else the excepting context will see corruption in its caller-save
+ * registers.  Plus, the entry.S save area is non-reentrant, so this code
+ * has to run with SR.BL==1, i.e. no interrupts taken inside it and panic
+ * on any exception.
+ *
+ * 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.
+ */
+#include <linux/signal.h>
+#include <linux/sched.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/interrupt.h>
+#include <linux/kprobes.h>
+#include <asm/tlb.h>
+#include <asm/io.h>
+#include <asm/uaccess.h>
+#include <asm/pgalloc.h>
+#include <asm/mmu_context.h>
+
+static int handle_tlbmiss(unsigned long long protection_flags,
+			  unsigned long address)
+{
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+	pte_t entry;
+
+	if (is_vmalloc_addr((void *)address)) {
+		pgd = pgd_offset_k(address);
+	} else {
+		if (unlikely(address >= TASK_SIZE || !current->mm))
+			return 1;
+
+		pgd = pgd_offset(current->mm, address);
+	}
+
+	pud = pud_offset(pgd, address);
+	if (pud_none(*pud) || !pud_present(*pud))
+		return 1;
+
+	pmd = pmd_offset(pud, address);
+	if (pmd_none(*pmd) || !pmd_present(*pmd))
+		return 1;
+
+	pte = pte_offset_kernel(pmd, address);
+	entry = *pte;
+	if (pte_none(entry) || !pte_present(entry))
+		return 1;
+
+	/*
+	 * If the page doesn't have sufficient protection bits set to
+	 * service the kind of fault being handled, there's not much
+	 * point doing the TLB refill.  Punt the fault to the general
+	 * handler.
+	 */
+	if ((pte_val(entry) & protection_flags) != protection_flags)
+		return 1;
+
+	update_mmu_cache(NULL, address, pte);
+
+	return 0;
+}
+
+/*
+ * Put all this information into one structure so that everything is just
+ * arithmetic relative to a single base address.  This reduces the number
+ * of movi/shori pairs needed just to load addresses of static data.
+ */
+struct expevt_lookup {
+	unsigned short protection_flags[8];
+	unsigned char  is_text_access[8];
+	unsigned char  is_write_access[8];
+};
+
+#define PRU (1<<9)
+#define PRW (1<<8)
+#define PRX (1<<7)
+#define PRR (1<<6)
+
+/* Sized as 8 rather than 4 to allow checking the PTE's PRU bit against whether
+   the fault happened in user mode or privileged mode. */
+static struct expevt_lookup expevt_lookup_table = {
+	.protection_flags = {PRX, PRX, 0, 0, PRR, PRR, PRW, PRW},
+	.is_text_access   = {1,   1,   0, 0, 0,   0,   0,   0}
+};
+
+static inline unsigned int
+expevt_to_fault_code(unsigned long expevt)
+{
+	if (expevt == 0xa40)
+		return FAULT_CODE_ITLB;
+	else if (expevt == 0x060)
+		return FAULT_CODE_WRITE;
+
+	return 0;
+}
+
+/*
+   This routine handles page faults that can be serviced just by refilling a
+   TLB entry from an existing page table entry.  (This case represents a very
+   large majority of page faults.) Return 1 if the fault was successfully
+   handled.  Return 0 if the fault could not be handled.  (This leads into the
+   general fault handling in fault.c which deals with mapping file-backed
+   pages, stack growth, segmentation faults, swapping etc etc)
+ */
+asmlinkage int __kprobes
+do_fast_page_fault(unsigned long long ssr_md, unsigned long long expevt,
+		   unsigned long address)
+{
+	unsigned long long protection_flags;
+	unsigned long long index;
+	unsigned long long expevt4;
+	unsigned int fault_code;
+
+	/* The next few lines implement a way of hashing EXPEVT into a
+	 * small array index which can be used to lookup parameters
+	 * specific to the type of TLBMISS being handled.
+	 *
+	 * Note:
+	 *	ITLBMISS has EXPEVT==0xa40
+	 *	RTLBMISS has EXPEVT==0x040
+	 *	WTLBMISS has EXPEVT==0x060
+	 */
+	expevt4 = (expevt >> 4);
+	/* TODO : xor ssr_md into this expression too. Then we can check
+	 * that PRU is set when it needs to be. */
+	index = expevt4 ^ (expevt4 >> 5);
+	index &= 7;
+
+	fault_code = expevt_to_fault_code(expevt);
+
+	protection_flags = expevt_lookup_table.protection_flags[index];
+
+	if (expevt_lookup_table.is_text_access[index])
+		fault_code |= FAULT_CODE_ITLB;
+	if (!ssr_md)
+		fault_code |= FAULT_CODE_USER;
+
+	set_thread_fault_code(fault_code);
+
+	return handle_tlbmiss(protection_flags, address);
+}
diff --git a/arch/sh/mm/tlbflush_32.c b/arch/sh/mm/tlbflush_32.c
new file mode 100644
index 0000000..a6a20d6
--- /dev/null
+++ b/arch/sh/mm/tlbflush_32.c
@@ -0,0 +1,137 @@
+/*
+ * TLB flushing operations for SH with an MMU.
+ *
+ *  Copyright (C) 1999  Niibe Yutaka
+ *  Copyright (C) 2003  Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/mm.h>
+#include <asm/mmu_context.h>
+#include <asm/tlbflush.h>
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	unsigned int cpu = smp_processor_id();
+
+	if (vma->vm_mm && cpu_context(cpu, vma->vm_mm) != NO_CONTEXT) {
+		unsigned long flags;
+		unsigned long asid;
+		unsigned long saved_asid = MMU_NO_ASID;
+
+		asid = cpu_asid(cpu, vma->vm_mm);
+		page &= PAGE_MASK;
+
+		local_irq_save(flags);
+		if (vma->vm_mm != current->mm) {
+			saved_asid = get_asid();
+			set_asid(asid);
+		}
+		local_flush_tlb_one(asid, page);
+		if (saved_asid != MMU_NO_ASID)
+			set_asid(saved_asid);
+		local_irq_restore(flags);
+	}
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+			   unsigned long end)
+{
+	struct mm_struct *mm = vma->vm_mm;
+	unsigned int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) != NO_CONTEXT) {
+		unsigned long flags;
+		int size;
+
+		local_irq_save(flags);
+		size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+		if (size > (MMU_NTLB_ENTRIES/4)) { /* Too many TLB to flush */
+			cpu_context(cpu, mm) = NO_CONTEXT;
+			if (mm == current->mm)
+				activate_context(mm, cpu);
+		} else {
+			unsigned long asid;
+			unsigned long saved_asid = MMU_NO_ASID;
+
+			asid = cpu_asid(cpu, mm);
+			start &= PAGE_MASK;
+			end += (PAGE_SIZE - 1);
+			end &= PAGE_MASK;
+			if (mm != current->mm) {
+				saved_asid = get_asid();
+				set_asid(asid);
+			}
+			while (start < end) {
+				local_flush_tlb_one(asid, start);
+				start += PAGE_SIZE;
+			}
+			if (saved_asid != MMU_NO_ASID)
+				set_asid(saved_asid);
+		}
+		local_irq_restore(flags);
+	}
+}
+
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+	unsigned int cpu = smp_processor_id();
+	unsigned long flags;
+	int size;
+
+	local_irq_save(flags);
+	size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+	if (size > (MMU_NTLB_ENTRIES/4)) { /* Too many TLB to flush */
+		local_flush_tlb_all();
+	} else {
+		unsigned long asid;
+		unsigned long saved_asid = get_asid();
+
+		asid = cpu_asid(cpu, &init_mm);
+		start &= PAGE_MASK;
+		end += (PAGE_SIZE - 1);
+		end &= PAGE_MASK;
+		set_asid(asid);
+		while (start < end) {
+			local_flush_tlb_one(asid, start);
+			start += PAGE_SIZE;
+		}
+		set_asid(saved_asid);
+	}
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	unsigned int cpu = smp_processor_id();
+
+	/* Invalidate all TLB of this process. */
+	/* Instead of invalidating each TLB, we get new MMU context. */
+	if (cpu_context(cpu, mm) != NO_CONTEXT) {
+		unsigned long flags;
+
+		local_irq_save(flags);
+		cpu_context(cpu, mm) = NO_CONTEXT;
+		if (mm == current->mm)
+			activate_context(mm, cpu);
+		local_irq_restore(flags);
+	}
+}
+
+void __flush_tlb_global(void)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	/*
+	 * This is the most destructive of the TLB flushing options,
+	 * and will tear down all of the UTLB/ITLB mappings, including
+	 * wired entries.
+	 */
+	__raw_writel(__raw_readl(MMUCR) | MMUCR_TI, MMUCR);
+
+	local_irq_restore(flags);
+}
diff --git a/arch/sh/mm/tlbflush_64.c b/arch/sh/mm/tlbflush_64.c
new file mode 100644
index 0000000..f33fdd2
--- /dev/null
+++ b/arch/sh/mm/tlbflush_64.c
@@ -0,0 +1,172 @@
+/*
+ * arch/sh/mm/tlb-flush_64.c
+ *
+ * Copyright (C) 2000, 2001  Paolo Alberelli
+ * Copyright (C) 2003  Richard Curnow (/proc/tlb, bug fixes)
+ * Copyright (C) 2003 - 2012 Paul Mundt
+ *
+ * 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.
+ */
+#include <linux/signal.h>
+#include <linux/rwsem.h>
+#include <linux/sched.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/perf_event.h>
+#include <linux/interrupt.h>
+#include <asm/io.h>
+#include <asm/tlb.h>
+#include <asm/uaccess.h>
+#include <asm/pgalloc.h>
+#include <asm/mmu_context.h>
+
+void local_flush_tlb_one(unsigned long asid, unsigned long page)
+{
+	unsigned long long match, pteh=0, lpage;
+	unsigned long tlb;
+
+	/*
+	 * Sign-extend based on neff.
+	 */
+	lpage = neff_sign_extend(page);
+	match = (asid << PTEH_ASID_SHIFT) | PTEH_VALID;
+	match |= lpage;
+
+	for_each_itlb_entry(tlb) {
+		asm volatile ("getcfg	%1, 0, %0"
+			      : "=r" (pteh)
+			      : "r" (tlb) );
+
+		if (pteh == match) {
+			__flush_tlb_slot(tlb);
+			break;
+		}
+	}
+
+	for_each_dtlb_entry(tlb) {
+		asm volatile ("getcfg	%1, 0, %0"
+			      : "=r" (pteh)
+			      : "r" (tlb) );
+
+		if (pteh == match) {
+			__flush_tlb_slot(tlb);
+			break;
+		}
+
+	}
+}
+
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+	unsigned long flags;
+
+	if (vma->vm_mm) {
+		page &= PAGE_MASK;
+		local_irq_save(flags);
+		local_flush_tlb_one(get_asid(), page);
+		local_irq_restore(flags);
+	}
+}
+
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+			   unsigned long end)
+{
+	unsigned long flags;
+	unsigned long long match, pteh=0, pteh_epn, pteh_low;
+	unsigned long tlb;
+	unsigned int cpu = smp_processor_id();
+	struct mm_struct *mm;
+
+	mm = vma->vm_mm;
+	if (cpu_context(cpu, mm) == NO_CONTEXT)
+		return;
+
+	local_irq_save(flags);
+
+	start &= PAGE_MASK;
+	end &= PAGE_MASK;
+
+	match = (cpu_asid(cpu, mm) << PTEH_ASID_SHIFT) | PTEH_VALID;
+
+	/* Flush ITLB */
+	for_each_itlb_entry(tlb) {
+		asm volatile ("getcfg	%1, 0, %0"
+			      : "=r" (pteh)
+			      : "r" (tlb) );
+
+		pteh_epn = pteh & PAGE_MASK;
+		pteh_low = pteh & ~PAGE_MASK;
+
+		if (pteh_low == match && pteh_epn >= start && pteh_epn <= end)
+			__flush_tlb_slot(tlb);
+	}
+
+	/* Flush DTLB */
+	for_each_dtlb_entry(tlb) {
+		asm volatile ("getcfg	%1, 0, %0"
+			      : "=r" (pteh)
+			      : "r" (tlb) );
+
+		pteh_epn = pteh & PAGE_MASK;
+		pteh_low = pteh & ~PAGE_MASK;
+
+		if (pteh_low == match && pteh_epn >= start && pteh_epn <= end)
+			__flush_tlb_slot(tlb);
+	}
+
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+	unsigned long flags;
+	unsigned int cpu = smp_processor_id();
+
+	if (cpu_context(cpu, mm) == NO_CONTEXT)
+		return;
+
+	local_irq_save(flags);
+
+	cpu_context(cpu, mm) = NO_CONTEXT;
+	if (mm == current->mm)
+		activate_context(mm, cpu);
+
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_all(void)
+{
+	/* Invalidate all, including shared pages, excluding fixed TLBs */
+	unsigned long flags, tlb;
+
+	local_irq_save(flags);
+
+	/* Flush each ITLB entry */
+	for_each_itlb_entry(tlb)
+		__flush_tlb_slot(tlb);
+
+	/* Flush each DTLB entry */
+	for_each_dtlb_entry(tlb)
+		__flush_tlb_slot(tlb);
+
+	local_irq_restore(flags);
+}
+
+void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
+{
+        /* FIXME: Optimize this later.. */
+        flush_tlb_all();
+}
+
+void __flush_tlb_global(void)
+{
+	flush_tlb_all();
+}
diff --git a/arch/sh/mm/uncached.c b/arch/sh/mm/uncached.c
new file mode 100644
index 0000000..a7767da
--- /dev/null
+++ b/arch/sh/mm/uncached.c
@@ -0,0 +1,43 @@
+#include <linux/init.h>
+#include <linux/module.h>
+#include <asm/sizes.h>
+#include <asm/page.h>
+#include <asm/addrspace.h>
+
+/*
+ * This is the offset of the uncached section from its cached alias.
+ *
+ * Legacy platforms handle trivial transitions between cached and
+ * uncached segments by making use of the 1:1 mapping relationship in
+ * 512MB lowmem, others via a special uncached mapping.
+ *
+ * Default value only valid in 29 bit mode, in 32bit mode this will be
+ * updated by the early PMB initialization code.
+ */
+unsigned long cached_to_uncached = SZ_512M;
+unsigned long uncached_size = SZ_512M;
+unsigned long uncached_start, uncached_end;
+EXPORT_SYMBOL(uncached_start);
+EXPORT_SYMBOL(uncached_end);
+
+int virt_addr_uncached(unsigned long kaddr)
+{
+	return (kaddr >= uncached_start) && (kaddr < uncached_end);
+}
+EXPORT_SYMBOL(virt_addr_uncached);
+
+void __init uncached_init(void)
+{
+#if defined(CONFIG_29BIT) || !defined(CONFIG_MMU)
+	uncached_start = P2SEG;
+#else
+	uncached_start = memory_end;
+#endif
+	uncached_end = uncached_start + uncached_size;
+}
+
+void __init uncached_resize(unsigned long size)
+{
+	uncached_size = size;
+	uncached_end = uncached_start + uncached_size;
+}