Merge branch 'for-2.6.32' into for-2.6.33

6 files changed, 57 insertions, 24 deletions

diff --git a/arch/sh/mm/Kconfig b/arch/sh/mm/Kconfig
index 64dc1ad59801..7f7b52f9beba 100644
--- a/arch/sh/mm/Kconfig
+++ b/arch/sh/mm/Kconfig
@@ -227,7 +227,7 @@ endchoice
 
 choice
         prompt "HugeTLB page size"
-        depends on HUGETLB_PAGE && (CPU_SH4 || CPU_SH5) && MMU
+        depends on HUGETLB_PAGE
         default HUGETLB_PAGE_SIZE_1MB if PAGE_SIZE_64KB
         default HUGETLB_PAGE_SIZE_64K
 
diff --git a/arch/sh/mm/cache-sh4.c b/arch/sh/mm/cache-sh4.c
index b2453bbef4cd..519e2d16cd06 100644
--- a/arch/sh/mm/cache-sh4.c
+++ b/arch/sh/mm/cache-sh4.c
@@ -26,7 +26,7 @@
 #define MAX_DCACHE_PAGES        64      /* XXX: Tune for ways */
 #define MAX_ICACHE_PAGES        32
 
-static void __flush_cache_4096(unsigned long addr, unsigned long phys,
+static void __flush_cache_one(unsigned long addr, unsigned long phys,
                                unsigned long exec_offset);
 
 /*
@@ -43,7 +43,7 @@ static void (*__flush_dcache_segment_fn)(unsigned long, unsigned long) =
  * 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)
+static void __uses_jump_to_uncached sh4_flush_icache_range(void *args)
 {
         struct flusher_data *data = args;
         unsigned long start, end;
@@ -89,8 +89,7 @@ static void sh4_flush_icache_range(void *args)
         local_irq_restore(flags);
 }
 
-static inline void flush_cache_4096(unsigned long start,
-                                    unsigned long phys)
+static inline void flush_cache_one(unsigned long start, unsigned long phys)
 {
         unsigned long flags, exec_offset = 0;
 
@@ -103,8 +102,7 @@ static inline void flush_cache_4096(unsigned long start,
                 exec_offset = 0x20000000;
 
         local_irq_save(flags);
-        __flush_cache_4096(start | SH_CACHE_ASSOC,
-                           P1SEGADDR(phys), exec_offset);
+        __flush_cache_one(start | SH_CACHE_ASSOC, P1SEGADDR(phys), exec_offset);
         local_irq_restore(flags);
 }
 
@@ -129,8 +127,8 @@ static void sh4_flush_dcache_page(void *arg)
 
                 /* Loop all the D-cache */
                 n = boot_cpu_data.dcache.n_aliases;
-                for (i = 0; i < n; i++, addr += 4096)
-                        flush_cache_4096(addr, phys);
+                for (i = 0; i < n; i++, addr += PAGE_SIZE)
+                        flush_cache_one(addr, phys);
         }
 
         wmb();
@@ -318,11 +316,11 @@ static void sh4_flush_cache_page(void *args)
         /* We only need to flush D-cache when we have alias */
         if ((address^phys) & alias_mask) {
                 /* Loop 4K of the D-cache */
-                flush_cache_4096(
+                flush_cache_one(
                         CACHE_OC_ADDRESS_ARRAY | (address & alias_mask),
                         phys);
                 /* Loop another 4K of the D-cache */
-                flush_cache_4096(
+                flush_cache_one(
                         CACHE_OC_ADDRESS_ARRAY | (phys & alias_mask),
                         phys);
         }
@@ -337,7 +335,7 @@ static void sh4_flush_cache_page(void *args)
                  * kernel has never executed the code through its identity
                  * translation.
                  */
-                flush_cache_4096(
+                flush_cache_one(
                         CACHE_IC_ADDRESS_ARRAY | (address & alias_mask),
                         phys);
         }
@@ -393,7 +391,7 @@ static void sh4_flush_cache_range(void *args)
 }
 
 /**
- * __flush_cache_4096
+ * __flush_cache_one
  *
  * @addr:  address in memory mapped cache array
  * @phys:  P1 address to flush (has to match tags if addr has 'A' bit
@@ -406,7 +404,7 @@ static void sh4_flush_cache_range(void *args)
  * operation (purge/write-back) is selected by the lower 2 bits of
  * 'phys'.
  */
-static void __flush_cache_4096(unsigned long addr, unsigned long phys,
+static void __flush_cache_one(unsigned long addr, unsigned long phys,
                                unsigned long exec_offset)
 {
         int way_count;
diff --git a/arch/sh/mm/cache-sh7705.c b/arch/sh/mm/cache-sh7705.c
index 2cadee2037ac..2601935eb589 100644
--- a/arch/sh/mm/cache-sh7705.c
+++ b/arch/sh/mm/cache-sh7705.c
@@ -78,7 +78,7 @@ static void sh7705_flush_icache_range(void *args)
 /*
  * Writeback&Invalidate the D-cache of the page
  */
-static void __flush_dcache_page(unsigned long phys)
+static void __uses_jump_to_uncached __flush_dcache_page(unsigned long phys)
 {
         unsigned long ways, waysize, addrstart;
         unsigned long flags;
@@ -144,7 +144,7 @@ static void sh7705_flush_dcache_page(void *arg)
                 __flush_dcache_page(PHYSADDR(page_address(page)));
 }
 
-static void sh7705_flush_cache_all(void *args)
+static void __uses_jump_to_uncached sh7705_flush_cache_all(void *args)
 {
         unsigned long flags;
 
diff --git a/arch/sh/mm/cache.c b/arch/sh/mm/cache.c
index 35c37b7f717a..a2dc7f9ecc51 100644
--- a/arch/sh/mm/cache.c
+++ b/arch/sh/mm/cache.c
@@ -128,7 +128,7 @@ void __update_cache(struct vm_area_struct *vma,
                 return;
 
         page = pfn_to_page(pfn);
-        if (pfn_valid(pfn) && page_mapping(page)) {
+        if (pfn_valid(pfn)) {
                 int dirty = test_and_clear_bit(PG_dcache_dirty, &page->flags);
                 if (dirty) {
                         unsigned long addr = (unsigned long)page_address(page);
@@ -265,6 +265,8 @@ static void __init emit_cache_params(void)
 
 void __init cpu_cache_init(void)
 {
+        unsigned int cache_disabled = !(__raw_readl(CCR) & CCR_CACHE_ENABLE);
+
         compute_alias(&boot_cpu_data.icache);
         compute_alias(&boot_cpu_data.dcache);
         compute_alias(&boot_cpu_data.scache);
@@ -273,6 +275,13 @@ void __init cpu_cache_init(void)
         __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);
 
@@ -312,5 +321,6 @@ void __init cpu_cache_init(void)
                 sh5_cache_init();
         }
 
+skip:
         emit_cache_params();
 }
diff --git a/arch/sh/mm/ioremap_32.c b/arch/sh/mm/ioremap_32.c
index c3250614e3ae..a86eaa9d75a5 100644
--- a/arch/sh/mm/ioremap_32.c
+++ b/arch/sh/mm/ioremap_32.c
@@ -83,7 +83,7 @@ void __iomem *__ioremap(unsigned long phys_addr, unsigned long size,
          *
          * PMB entries are all pre-faulted.
          */
-        if (unlikely(size >= 0x1000000)) {
+        if (unlikely(phys_addr >= P1SEG)) {
                 unsigned long mapped = pmb_remap(addr, phys_addr, size, flags);
 
                 if (likely(mapped)) {
diff --git a/arch/sh/mm/pmb.c b/arch/sh/mm/pmb.c
index b1a714a92b14..aade31102112 100644
--- a/arch/sh/mm/pmb.c
+++ b/arch/sh/mm/pmb.c
@@ -33,6 +33,8 @@
 
 #define NR_PMB_ENTRIES  16
 
+static void __pmb_unmap(struct pmb_entry *);
+
 static struct kmem_cache *pmb_cache;
 static unsigned long pmb_map;
 
@@ -218,9 +220,10 @@ static struct {
 long pmb_remap(unsigned long vaddr, unsigned long phys,
                unsigned long size, unsigned long flags)
 {
-        struct pmb_entry *pmbp;
+        struct pmb_entry *pmbp, *pmbe;
         unsigned long wanted;
         int pmb_flags, i;
+        long err;
 
         /* Convert typical pgprot value to the PMB equivalent */
         if (flags & _PAGE_CACHABLE) {
@@ -236,20 +239,22 @@ long pmb_remap(unsigned long vaddr, unsigned long phys,
 
 again:
         for (i = 0; i < ARRAY_SIZE(pmb_sizes); i++) {
-                struct pmb_entry *pmbe;
                 int ret;
 
                 if (size < pmb_sizes[i].size)
                         continue;
 
                 pmbe = pmb_alloc(vaddr, phys, pmb_flags | pmb_sizes[i].flag);
-                if (IS_ERR(pmbe))
-                        return PTR_ERR(pmbe);
+                if (IS_ERR(pmbe)) {
+                        err = PTR_ERR(pmbe);
+                        goto out;
+                }
 
                 ret = set_pmb_entry(pmbe);
                 if (ret != 0) {
                         pmb_free(pmbe);
-                        return -EBUSY;
+                        err = -EBUSY;
+                        goto out;
                 }
 
                 phys    += pmb_sizes[i].size;
@@ -264,12 +269,25 @@ again:
                         pmbp->link = pmbe;
 
                 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--;
         }
 
         if (size >= 0x1000000)
                 goto again;
 
         return wanted - size;
+
+out:
+        if (pmbp)
+                __pmb_unmap(pmbp);
+
+        return err;
 }
 
 void pmb_unmap(unsigned long addr)
@@ -283,12 +301,19 @@ void pmb_unmap(unsigned long addr)
         if (unlikely(!pmbe))
                 return;
 
+        __pmb_unmap(pmbe);
+}
+
+static void __pmb_unmap(struct pmb_entry *pmbe)
+{
         WARN_ON(!test_bit(pmbe->entry, &pmb_map));
 
         do {
                 struct pmb_entry *pmblink = pmbe;
 
-                clear_pmb_entry(pmbe);
+                if (pmbe->entry != PMB_NO_ENTRY)
+                        clear_pmb_entry(pmbe);
+
                 pmbe = pmblink->link;
 
                 pmb_free(pmblink);