diff --git a/arch/sh/include/asm/system.h b/arch/sh/include/asm/system.h
index ab79e1f..bf7c4cb 100644
--- a/arch/sh/include/asm/system.h
+++ b/arch/sh/include/asm/system.h
@@ -14,18 +14,6 @@
 
 #define AT_VECTOR_SIZE_ARCH 5 /* entries in ARCH_DLINFO */
 
-#if defined(CONFIG_CPU_SH4A) || defined(CONFIG_CPU_SH5)
-#define __icbi()			\
-{					\
-	unsigned long __addr;		\
-	__addr = 0xa8000000;		\
-	__asm__ __volatile__(		\
-		"icbi   %0\n\t"		\
-		: /* no output */	\
-		: "m" (__m(__addr)));	\
-}
-#endif
-
 /*
  * A brief note on ctrl_barrier(), the control register write barrier.
  *
@@ -44,7 +32,7 @@
 #define mb()		__asm__ __volatile__ ("synco": : :"memory")
 #define rmb()		mb()
 #define wmb()		__asm__ __volatile__ ("synco": : :"memory")
-#define ctrl_barrier()	__icbi()
+#define ctrl_barrier()	__icbi(0xa8000000)
 #define read_barrier_depends()	do { } while(0)
 #else
 #define mb()		__asm__ __volatile__ ("": : :"memory")
diff --git a/arch/sh/include/asm/system_32.h b/arch/sh/include/asm/system_32.h
index d7299d6..5ddd235 100644
--- a/arch/sh/include/asm/system_32.h
+++ b/arch/sh/include/asm/system_32.h
@@ -63,6 +63,16 @@
 #define __restore_dsp(tsk)	do { } while (0)
 #endif
 
+#if defined(CONFIG_CPU_SH4A)
+#define __icbi(addr)	__asm__ __volatile__ ( "icbi @%0\n\t" : : "r" (addr))
+#else
+#define __icbi(addr)	mb()
+#endif
+
+#define __ocbp(addr)	__asm__ __volatile__ ( "ocbp @%0\n\t" : : "r" (addr))
+#define __ocbi(addr)	__asm__ __volatile__ ( "ocbi @%0\n\t" : : "r" (addr))
+#define __ocbwb(addr)	__asm__ __volatile__ ( "ocbwb @%0\n\t" : : "r" (addr))
+
 struct task_struct *__switch_to(struct task_struct *prev,
 				struct task_struct *next);
 
diff --git a/arch/sh/include/asm/system_64.h b/arch/sh/include/asm/system_64.h
index 218b54d..8e4a03e 100644
--- a/arch/sh/include/asm/system_64.h
+++ b/arch/sh/include/asm/system_64.h
@@ -37,6 +37,11 @@
 #define jump_to_uncached()	do { } while (0)
 #define back_to_cached()	do { } while (0)
 
+#define __icbi(addr)	__asm__ __volatile__ ( "icbi %0, 0\n\t" : : "r" (addr))
+#define __ocbp(addr)	__asm__ __volatile__ ( "ocbp %0, 0\n\t" : : "r" (addr))
+#define __ocbi(addr)	__asm__ __volatile__ ( "ocbi %0, 0\n\t" : : "r" (addr))
+#define __ocbwb(addr)	__asm__ __volatile__ ( "ocbwb %0, 0\n\t" : : "r" (addr))
+
 static inline reg_size_t register_align(void *val)
 {
 	return (unsigned long long)(signed long long)(signed long)val;
diff --git a/arch/sh/kernel/sh_ksyms_64.c b/arch/sh/kernel/sh_ksyms_64.c
index f96c95c..d008e17 100644
--- a/arch/sh/kernel/sh_ksyms_64.c
+++ b/arch/sh/kernel/sh_ksyms_64.c
@@ -30,14 +30,6 @@
 EXPORT_SYMBOL(dump_fpu);
 EXPORT_SYMBOL(kernel_thread);
 
-#if !defined(CONFIG_CACHE_OFF) && defined(CONFIG_MMU)
-EXPORT_SYMBOL(clear_user_page);
-#endif
-
-#ifndef CONFIG_CACHE_OFF
-EXPORT_SYMBOL(flush_dcache_page);
-#endif
-
 #ifdef CONFIG_VT
 EXPORT_SYMBOL(screen_info);
 #endif
diff --git a/arch/sh/mm/cache-sh5.c b/arch/sh/mm/cache-sh5.c
index a8f5142d..819e0f9 100644
--- a/arch/sh/mm/cache-sh5.c
+++ b/arch/sh/mm/cache-sh5.c
@@ -25,29 +25,6 @@
 /* Wired TLB entry for the D-cache */
 static unsigned long long dtlb_cache_slot;
 
-void __init cpu_cache_init(void)
-{
-	/* Reserve a slot for dcache colouring in the DTLB */
-	dtlb_cache_slot	= sh64_get_wired_dtlb_entry();
-
-	sh4__flush_region_init();
-}
-
-void __init kmap_coherent_init(void)
-{
-	/* XXX ... */
-}
-
-void *kmap_coherent(struct page *page, unsigned long addr)
-{
-	/* XXX ... */
-	return NULL;
-}
-
-void kunmap_coherent(void)
-{
-}
-
 #ifdef CONFIG_DCACHE_DISABLED
 #define sh64_dcache_purge_all()					do { } while (0)
 #define sh64_dcache_purge_coloured_phy_page(paddr, eaddr)	do { } while (0)
@@ -233,52 +210,6 @@
 	}
 }
 
-/*
- * Invalidate a small range of user context I-cache, not necessarily page
- * (or even cache-line) aligned.
- *
- * Since this is used inside ptrace, the ASID in the mm context typically
- * won't match current_asid.  We'll have to switch ASID to do this.  For
- * safety, and given that the range will be small, do all this under cli.
- *
- * Note, there is a hazard that the ASID in mm->context is no longer
- * actually associated with mm, i.e. if the mm->context has started a new
- * cycle since mm was last active.  However, this is just a performance
- * issue: all that happens is that we invalidate lines belonging to
- * another mm, so the owning process has to refill them when that mm goes
- * live again.  mm itself can't have any cache entries because there will
- * have been a flush_cache_all when the new mm->context cycle started.
- */
-static void sh64_icache_inv_user_small_range(struct mm_struct *mm,
-						unsigned long start, int len)
-{
-	unsigned long long eaddr = start;
-	unsigned long long eaddr_end = start + len;
-	unsigned long current_asid, mm_asid;
-	unsigned long flags;
-	unsigned long long epage_start;
-
-	/*
-	 * Align to start of cache line.  Otherwise, suppose len==8 and
-	 * start was at 32N+28 : the last 4 bytes wouldn't get invalidated.
-	 */
-	eaddr = L1_CACHE_ALIGN(start);
-	eaddr_end = start + len;
-
-	mm_asid = cpu_asid(smp_processor_id(), mm);
-	local_irq_save(flags);
-	current_asid = switch_and_save_asid(mm_asid);
-
-	epage_start = eaddr & PAGE_MASK;
-
-	while (eaddr < eaddr_end) {
-		__asm__ __volatile__("icbi %0, 0" : : "r" (eaddr));
-		eaddr += L1_CACHE_BYTES;
-	}
-	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
@@ -564,7 +495,7 @@
  * Invalidate the entire contents of both caches, after writing back to
  * memory any dirty data from the D-cache.
  */
-void flush_cache_all(void)
+static void sh5_flush_cache_all(void)
 {
 	sh64_dcache_purge_all();
 	sh64_icache_inv_all();
@@ -591,7 +522,7 @@
  * I-cache.  This is similar to the lack of action needed in
  * flush_tlb_mm - see fault.c.
  */
-void flush_cache_mm(struct mm_struct *mm)
+static void sh5_flush_cache_mm(struct mm_struct *mm)
 {
 	sh64_dcache_purge_all();
 }
@@ -603,8 +534,8 @@
  *
  * Note, 'end' is 1 byte beyond the end of the range to flush.
  */
-void flush_cache_range(struct vm_area_struct *vma, unsigned long start,
-		       unsigned long end)
+static void sh5_flush_cache_range(struct vm_area_struct *vma,
+		unsigned long start, unsigned long end)
 {
 	struct mm_struct *mm = vma->vm_mm;
 
@@ -621,8 +552,8 @@
  *
  * Note, this is called with pte lock held.
  */
-void flush_cache_page(struct vm_area_struct *vma, unsigned long eaddr,
-		      unsigned long pfn)
+static void sh5_flush_cache_page(struct vm_area_struct *vma,
+		unsigned long eaddr, unsigned long pfn)
 {
 	sh64_dcache_purge_phy_page(pfn << PAGE_SHIFT);
 
@@ -630,7 +561,7 @@
 		sh64_icache_inv_user_page(vma, eaddr);
 }
 
-void flush_dcache_page(struct page *page)
+static void sh5_flush_dcache_page(struct page *page)
 {
 	sh64_dcache_purge_phy_page(page_to_phys(page));
 	wmb();
@@ -644,7 +575,7 @@
  * mapping, therefore it's guaranteed that there no cache entries for
  * the range in cache sets of the wrong colour.
  */
-void flush_icache_range(unsigned long start, unsigned long end)
+static void sh5_flush_icache_range(unsigned long start, unsigned long end)
 {
 	__flush_purge_region((void *)start, end);
 	wmb();
@@ -652,31 +583,12 @@
 }
 
 /*
- * Flush the range of user (defined by vma->vm_mm) address space starting
- * at 'addr' for 'len' bytes from the cache.  The range does not straddle
- * a page boundary, the unique physical page containing the range is
- * 'page'.  This seems to be used mainly for invalidating an address
- * range following a poke into the program text through the ptrace() call
- * from another process (e.g. for BRK instruction insertion).
- */
-static void flush_icache_user_range(struct vm_area_struct *vma,
-			struct page *page, unsigned long addr, int len)
-{
-
-	sh64_dcache_purge_coloured_phy_page(page_to_phys(page), addr);
-	mb();
-
-	if (vma->vm_flags & VM_EXEC)
-		sh64_icache_inv_user_small_range(vma->vm_mm, addr, len);
-}
-
-/*
  * 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.
  */
-void flush_cache_sigtramp(unsigned long vaddr)
+static void sh5_flush_cache_sigtramp(unsigned long vaddr)
 {
 	unsigned long end = vaddr + L1_CACHE_BYTES;
 
@@ -685,138 +597,19 @@
 	sh64_icache_inv_current_user_range(vaddr, end);
 }
 
-#ifdef CONFIG_MMU
-/*
- * These *MUST* lie in an area of virtual address space that's otherwise
- * unused.
- */
-#define UNIQUE_EADDR_START 0xe0000000UL
-#define UNIQUE_EADDR_END   0xe8000000UL
-
-/*
- * Given a physical address paddr, and a user virtual address user_eaddr
- * which will eventually be mapped to it, create a one-off kernel-private
- * eaddr mapped to the same paddr.  This is used for creating special
- * destination pages for copy_user_page and clear_user_page.
- */
-static unsigned long sh64_make_unique_eaddr(unsigned long user_eaddr,
-					    unsigned long paddr)
+void __init sh5_cache_init(void)
 {
-	static unsigned long current_pointer = UNIQUE_EADDR_START;
-	unsigned long coloured_pointer;
+	flush_cache_all		= sh5_flush_cache_all;
+	flush_cache_mm		= sh5_flush_cache_mm;
+	flush_cache_dup_mm	= sh5_flush_cache_mm;
+	flush_cache_page	= sh5_flush_cache_page;
+	flush_cache_range	= sh5_flush_cache_range;
+	flush_dcache_page	= sh5_flush_dcache_page;
+	flush_icache_range	= sh5_flush_icache_range;
+	flush_cache_sigtramp	= sh5_flush_cache_sigtramp;
 
-	if (current_pointer == UNIQUE_EADDR_END) {
-		sh64_dcache_purge_all();
-		current_pointer = UNIQUE_EADDR_START;
-	}
+	/* Reserve a slot for dcache colouring in the DTLB */
+	dtlb_cache_slot	= sh64_get_wired_dtlb_entry();
 
-	coloured_pointer = (current_pointer & ~CACHE_OC_SYN_MASK) |
-				(user_eaddr & CACHE_OC_SYN_MASK);
-	sh64_setup_dtlb_cache_slot(coloured_pointer, get_asid(), paddr);
-
-	current_pointer += (PAGE_SIZE << CACHE_OC_N_SYNBITS);
-
-	return coloured_pointer;
+	sh4__flush_region_init();
 }
-
-static void sh64_copy_user_page_coloured(void *to, void *from,
-					 unsigned long address)
-{
-	void *coloured_to;
-
-	/*
-	 * Discard any existing cache entries of the wrong colour.  These are
-	 * present quite often, if the kernel has recently used the page
-	 * internally, then given it up, then it's been allocated to the user.
-	 */
-	sh64_dcache_purge_coloured_phy_page(__pa(to), (unsigned long)to);
-
-	coloured_to = (void *)sh64_make_unique_eaddr(address, __pa(to));
-	copy_page(from, coloured_to);
-
-	sh64_teardown_dtlb_cache_slot();
-}
-
-static void sh64_clear_user_page_coloured(void *to, unsigned long address)
-{
-	void *coloured_to;
-
-	/*
-	 * Discard any existing kernel-originated lines of the wrong
-	 * colour (as above)
-	 */
-	sh64_dcache_purge_coloured_phy_page(__pa(to), (unsigned long)to);
-
-	coloured_to = (void *)sh64_make_unique_eaddr(address, __pa(to));
-	clear_page(coloured_to);
-
-	sh64_teardown_dtlb_cache_slot();
-}
-
-/*
- * 'from' and 'to' are kernel virtual addresses (within the superpage
- * mapping of the physical RAM).  'address' is the user virtual address
- * where the copy 'to' will be mapped after.  This allows a custom
- * mapping to be used to ensure that the new copy is placed in the
- * right cache sets for the user to see it without having to bounce it
- * out via memory.  Note however : the call to flush_page_to_ram in
- * (generic)/mm/memory.c:(break_cow) undoes all this good work in that one
- * very important case!
- *
- * TBD : can we guarantee that on every call, any cache entries for
- * 'from' are in the same colour sets as 'address' also?  i.e. is this
- * always used just to deal with COW?  (I suspect not).
- *
- * There are two possibilities here for when the page 'from' was last accessed:
- * - by the kernel : this is OK, no purge required.
- * - by the/a user (e.g. for break_COW) : need to purge.
- *
- * If the potential user mapping at 'address' is the same colour as
- * 'from' there is no need to purge any cache lines from the 'from'
- * page mapped into cache sets of colour 'address'.  (The copy will be
- * accessing the page through 'from').
- */
-void copy_user_page(void *to, void *from, unsigned long address,
-		    struct page *page)
-{
-	if (((address ^ (unsigned long) from) & CACHE_OC_SYN_MASK) != 0)
-		sh64_dcache_purge_coloured_phy_page(__pa(from), address);
-
-	if (((address ^ (unsigned long) to) & CACHE_OC_SYN_MASK) == 0)
-		copy_page(to, from);
-	else
-		sh64_copy_user_page_coloured(to, from, address);
-}
-
-/*
- * 'to' is a kernel virtual address (within the superpage mapping of the
- * physical RAM).  'address' is the user virtual address where the 'to'
- * page will be mapped after.  This allows a custom mapping to be used to
- * ensure that the new copy is placed in the right cache sets for the
- * user to see it without having to bounce it out via memory.
- */
-void clear_user_page(void *to, unsigned long address, struct page *page)
-{
-	if (((address ^ (unsigned long) to) & CACHE_OC_SYN_MASK) == 0)
-		clear_page(to);
-	else
-		sh64_clear_user_page_coloured(to, address);
-}
-
-void copy_to_user_page(struct vm_area_struct *vma, struct page *page,
-		       unsigned long vaddr, void *dst, const void *src,
-		       unsigned long len)
-{
-	flush_cache_page(vma, vaddr, page_to_pfn(page));
-	memcpy(dst, src, len);
-	flush_icache_user_range(vma, page, vaddr, len);
-}
-
-void copy_from_user_page(struct vm_area_struct *vma, struct page *page,
-			 unsigned long vaddr, void *dst, const void *src,
-			 unsigned long len)
-{
-	flush_cache_page(vma, vaddr, page_to_pfn(page));
-	memcpy(dst, src, len);
-}
-#endif
diff --git a/arch/sh/mm/flush-sh4.c b/arch/sh/mm/flush-sh4.c
index 99c50dc..cef4026 100644
--- a/arch/sh/mm/flush-sh4.c
+++ b/arch/sh/mm/flush-sh4.c
@@ -19,28 +19,19 @@
 	cnt = (end - v) / L1_CACHE_BYTES;
 
 	while (cnt >= 8) {
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbwb	@%0" : : "r" (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;
+		__ocbwb(v); v += L1_CACHE_BYTES;
 		cnt -= 8;
 	}
 
 	while (cnt) {
-		asm volatile("ocbwb	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
+		__ocbwb(v); v += L1_CACHE_BYTES;
 		cnt--;
 	}
 }
@@ -62,27 +53,18 @@
 	cnt = (end - v) / L1_CACHE_BYTES;
 
 	while (cnt >= 8) {
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbp	@%0" : : "r" (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;
+		__ocbp(v); v += L1_CACHE_BYTES;
 		cnt -= 8;
 	}
 	while (cnt) {
-		asm volatile("ocbp	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
+		__ocbp(v); v += L1_CACHE_BYTES;
 		cnt--;
 	}
 }
@@ -101,28 +83,19 @@
 	cnt = (end - v) / L1_CACHE_BYTES;
 
 	while (cnt >= 8) {
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
-		asm volatile("ocbi	@%0" : : "r" (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;
+		__ocbi(v); v += L1_CACHE_BYTES;
 		cnt -= 8;
 	}
 
 	while (cnt) {
-		asm volatile("ocbi	@%0" : : "r" (v));
-		v += L1_CACHE_BYTES;
+		__ocbi(v); v += L1_CACHE_BYTES;
 		cnt--;
 	}
 }