20 files changed, 492 insertions, 70 deletions

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index cb4486ad0f72..e62b37a15a1d 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -939,6 +939,23 @@ config NODES_SHIFT
         default "2"
         depends on NEED_MULTIPLE_NODES
 
+config HIGHMEM
+        bool "High Memory Support (EXPERIMENTAL)"
+        depends on MMU && EXPERIMENTAL
+        help
+          The address space of ARM processors is only 4 Gigabytes large
+          and it has to accommodate user address space, kernel address
+          space as well as some memory mapped IO. That means that, if you
+          have a large amount of physical memory and/or IO, not all of the
+          memory can be "permanently mapped" by the kernel. The physical
+          memory that is not permanently mapped is called "high memory".
+
+          Depending on the selected kernel/user memory split, minimum
+          vmalloc space and actual amount of RAM, you may not need this
+          option which should result in a slightly faster kernel.
+
+          If unsure, say n.
+
 source "mm/Kconfig"
 
 config LEDS
diff --git a/arch/arm/common/dmabounce.c b/arch/arm/common/dmabounce.c
index f030f0775be7..734ac9135998 100644
--- a/arch/arm/common/dmabounce.c
+++ b/arch/arm/common/dmabounce.c
@@ -25,6 +25,7 @@
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/slab.h>
+#include <linux/page-flags.h>
 #include <linux/device.h>
 #include <linux/dma-mapping.h>
 #include <linux/dmapool.h>
@@ -349,6 +350,12 @@ dma_addr_t dma_map_page(struct device *dev, struct page *page,
 
         BUG_ON(!valid_dma_direction(dir));
 
+        if (PageHighMem(page)) {
+                dev_err(dev, "DMA buffer bouncing of HIGHMEM pages "
+                             "is not supported\n");
+                return ~0;
+        }
+
         return map_single(dev, page_address(page) + offset, size, dir);
 }
 EXPORT_SYMBOL(dma_map_page);
diff --git a/arch/arm/include/asm/dma-mapping.h b/arch/arm/include/asm/dma-mapping.h
index 22cb14ec3438..ff46dfa68a97 100644
--- a/arch/arm/include/asm/dma-mapping.h
+++ b/arch/arm/include/asm/dma-mapping.h
@@ -15,10 +15,20 @@
  * must not be used by drivers.
  */
 #ifndef __arch_page_to_dma
+
+#if !defined(CONFIG_HIGHMEM)
 static inline dma_addr_t page_to_dma(struct device *dev, struct page *page)
 {
         return (dma_addr_t)__virt_to_bus((unsigned long)page_address(page));
 }
+#elif defined(__pfn_to_bus)
+static inline dma_addr_t page_to_dma(struct device *dev, struct page *page)
+{
+        return (dma_addr_t)__pfn_to_bus(page_to_pfn(page));
+}
+#else
+#error "this machine class needs to define __arch_page_to_dma to use HIGHMEM"
+#endif
 
 static inline void *dma_to_virt(struct device *dev, dma_addr_t addr)
 {
@@ -57,6 +67,8 @@ static inline dma_addr_t virt_to_dma(struct device *dev, void *addr)
  * Use the driver DMA support - see dma-mapping.h (dma_sync_*)
  */
 extern void dma_cache_maint(const void *kaddr, size_t size, int rw);
+extern void dma_cache_maint_page(struct page *page, unsigned long offset,
+                                 size_t size, int rw);
 
 /*
  * Return whether the given device DMA address mask can be supported
@@ -316,7 +328,7 @@ static inline dma_addr_t dma_map_page(struct device *dev, struct page *page,
         BUG_ON(!valid_dma_direction(dir));
 
         if (!arch_is_coherent())
-                dma_cache_maint(page_address(page) + offset, size, dir);
+                dma_cache_maint_page(page, offset, size, dir);
 
         return page_to_dma(dev, page) + offset;
 }
diff --git a/arch/arm/include/asm/fixmap.h b/arch/arm/include/asm/fixmap.h
new file mode 100644
index 000000000000..bbae919bceb4
--- /dev/null
+++ b/arch/arm/include/asm/fixmap.h
@@ -0,0 +1,41 @@
+#ifndef _ASM_FIXMAP_H
+#define _ASM_FIXMAP_H
+
+/*
+ * Nothing too fancy for now.
+ *
+ * On ARM we already have well known fixed virtual addresses imposed by
+ * the architecture such as the vector page which is located at 0xffff0000,
+ * therefore a second level page table is already allocated covering
+ * 0xfff00000 upwards.
+ *
+ * The cache flushing code in proc-xscale.S uses the virtual area between
+ * 0xfffe0000 and 0xfffeffff.
+ */
+
+#define FIXADDR_START           0xfff00000UL
+#define FIXADDR_TOP             0xfffe0000UL
+#define FIXADDR_SIZE            (FIXADDR_TOP - FIXADDR_START)
+
+#define FIX_KMAP_BEGIN          0
+#define FIX_KMAP_END            (FIXADDR_SIZE >> PAGE_SHIFT)
+
+#define __fix_to_virt(x)        (FIXADDR_START + ((x) << PAGE_SHIFT))
+#define __virt_to_fix(x)        (((x) - FIXADDR_START) >> PAGE_SHIFT)
+
+extern void __this_fixmap_does_not_exist(void);
+
+static inline unsigned long fix_to_virt(const unsigned int idx)
+{
+        if (idx >= FIX_KMAP_END)
+                __this_fixmap_does_not_exist();
+        return __fix_to_virt(idx);
+}
+
+static inline unsigned int virt_to_fix(const unsigned long vaddr)
+{
+        BUG_ON(vaddr >= FIXADDR_TOP || vaddr < FIXADDR_START);
+        return __virt_to_fix(vaddr);
+}
+
+#endif
diff --git a/arch/arm/include/asm/highmem.h b/arch/arm/include/asm/highmem.h
new file mode 100644
index 000000000000..7f36d00600b4
--- /dev/null
+++ b/arch/arm/include/asm/highmem.h
@@ -0,0 +1,31 @@
+#ifndef _ASM_HIGHMEM_H
+#define _ASM_HIGHMEM_H
+
+#include <asm/kmap_types.h>
+
+#define PKMAP_BASE              (PAGE_OFFSET - PMD_SIZE)
+#define LAST_PKMAP              PTRS_PER_PTE
+#define LAST_PKMAP_MASK         (LAST_PKMAP - 1)
+#define PKMAP_NR(virt)          (((virt) - PKMAP_BASE) >> PAGE_SHIFT)
+#define PKMAP_ADDR(nr)          (PKMAP_BASE + ((nr) << PAGE_SHIFT))
+
+#define kmap_prot               PAGE_KERNEL
+
+#define flush_cache_kmaps()     flush_cache_all()
+
+extern pte_t *pkmap_page_table;
+
+#define ARCH_NEEDS_KMAP_HIGH_GET
+
+extern void *kmap_high(struct page *page);
+extern void *kmap_high_get(struct page *page);
+extern void kunmap_high(struct page *page);
+
+extern void *kmap(struct page *page);
+extern void kunmap(struct page *page);
+extern void *kmap_atomic(struct page *page, enum km_type type);
+extern void kunmap_atomic(void *kvaddr, enum km_type type);
+extern void *kmap_atomic_pfn(unsigned long pfn, enum km_type type);
+extern struct page *kmap_atomic_to_page(const void *ptr);
+
+#endif
diff --git a/arch/arm/include/asm/kmap_types.h b/arch/arm/include/asm/kmap_types.h
index 45def13ee17a..d16ec97ec9a9 100644
--- a/arch/arm/include/asm/kmap_types.h
+++ b/arch/arm/include/asm/kmap_types.h
@@ -18,6 +18,7 @@ enum km_type {
         KM_IRQ1,
         KM_SOFTIRQ0,
         KM_SOFTIRQ1,
+        KM_L2_CACHE,
         KM_TYPE_NR
 };
 
diff --git a/arch/arm/include/asm/memory.h b/arch/arm/include/asm/memory.h
index 0202a7c20e62..85763db87449 100644
--- a/arch/arm/include/asm/memory.h
+++ b/arch/arm/include/asm/memory.h
@@ -44,14 +44,21 @@
  * The module space lives between the addresses given by TASK_SIZE
  * and PAGE_OFFSET - it must be within 32MB of the kernel text.
  */
-#define MODULES_END             (PAGE_OFFSET)
-#define MODULES_VADDR           (MODULES_END - 16*1048576)
-
+#define MODULES_VADDR           (PAGE_OFFSET - 16*1024*1024)
 #if TASK_SIZE > MODULES_VADDR
 #error Top of user space clashes with start of module space
 #endif
 
 /*
+ * The highmem pkmap virtual space shares the end of the module area.
+ */
+#ifdef CONFIG_HIGHMEM
+#define MODULES_END             (PAGE_OFFSET - PMD_SIZE)
+#else
+#define MODULES_END             (PAGE_OFFSET)
+#endif
+
+/*
  * The XIP kernel gets mapped at the bottom of the module vm area.
  * Since we use sections to map it, this macro replaces the physical address
  * with its virtual address while keeping offset from the base section.
@@ -181,6 +188,7 @@ static inline void *phys_to_virt(unsigned long x)
 #ifndef __virt_to_bus
 #define __virt_to_bus   __virt_to_phys
 #define __bus_to_virt   __phys_to_virt
+#define __pfn_to_bus(x) ((x) << PAGE_SHIFT)
 #endif
 
 static inline __deprecated unsigned long virt_to_bus(void *x)
diff --git a/arch/arm/mach-iop13xx/include/mach/memory.h b/arch/arm/mach-iop13xx/include/mach/memory.h
index e012bf13c955..42ae29b288a1 100644
--- a/arch/arm/mach-iop13xx/include/mach/memory.h
+++ b/arch/arm/mach-iop13xx/include/mach/memory.h
@@ -59,7 +59,10 @@ static inline unsigned long __lbus_to_virt(dma_addr_t x)
         })
 
 #define __arch_page_to_dma(dev, page)                                   \
-        __arch_virt_to_dma(dev, page_address(page))
+        ({                                                              \
+                /* __is_lbus_virt() can never be true for RAM pages */  \
+                (dma_addr_t)page_to_phys(page);                         \
+        })
 
 #endif /* CONFIG_ARCH_IOP13XX */
 #endif /* !ASSEMBLY */
diff --git a/arch/arm/mach-ks8695/include/mach/memory.h b/arch/arm/mach-ks8695/include/mach/memory.h
index 6d5887cf5742..76e5308685a4 100644
--- a/arch/arm/mach-ks8695/include/mach/memory.h
+++ b/arch/arm/mach-ks8695/include/mach/memory.h
@@ -35,7 +35,11 @@ extern struct bus_type platform_bus_type;
                                         __phys_to_virt(x) : __bus_to_virt(x)); })
 #define __arch_virt_to_dma(dev, x)      ({ is_lbus_device(dev) ? \
                                         (dma_addr_t)__virt_to_phys(x) : (dma_addr_t)__virt_to_bus(x); })
-#define __arch_page_to_dma(dev, x)      __arch_virt_to_dma(dev, page_address(x))
+#define __arch_page_to_dma(dev, x)      \
+        ({ dma_addr_t __dma = page_to_phys(page); \
+           if (!is_lbus_device(dev)) \
+                __dma = __dma - PHYS_OFFSET + KS8695_PCIMEM_PA; \
+           __dma; })
 
 #endif
 
diff --git a/arch/arm/mm/Makefile b/arch/arm/mm/Makefile
index 64149d9e55a5..c264683538bc 100644
--- a/arch/arm/mm/Makefile
+++ b/arch/arm/mm/Makefile
@@ -16,6 +16,7 @@ obj-$(CONFIG_MODULES)		+= proc-syms.o
 
 obj-$(CONFIG_ALIGNMENT_TRAP)    += alignment.o
 obj-$(CONFIG_DISCONTIGMEM)      += discontig.o
+obj-$(CONFIG_HIGHMEM)           += highmem.o
 
 obj-$(CONFIG_CPU_ABRT_NOMMU)    += abort-nommu.o
 obj-$(CONFIG_CPU_ABRT_EV4)      += abort-ev4.o
diff --git a/arch/arm/mm/cache-feroceon-l2.c b/arch/arm/mm/cache-feroceon-l2.c
index 80cd207cbaea..d6dd83826f8a 100644
--- a/arch/arm/mm/cache-feroceon-l2.c
+++ b/arch/arm/mm/cache-feroceon-l2.c
@@ -14,8 +14,12 @@
 
 #include <linux/init.h>
 #include <asm/cacheflush.h>
+#include <asm/kmap_types.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
 #include <plat/cache-feroceon-l2.h>
-
+#include "mm.h"
 
 /*
  * Low-level cache maintenance operations.
@@ -34,14 +38,36 @@
  * The range operations require two successive cp15 writes, in
  * between which we don't want to be preempted.
  */
+
+static inline unsigned long l2_start_va(unsigned long paddr)
+{
+#ifdef CONFIG_HIGHMEM
+        /*
+         * Let's do our own fixmap stuff in a minimal way here.
+         * Because range ops can't be done on physical addresses,
+         * we simply install a virtual mapping for it only for the
+         * TLB lookup to occur, hence no need to flush the untouched
+         * memory mapping.  This is protected with the disabling of
+         * interrupts by the caller.
+         */
+        unsigned long idx = KM_L2_CACHE + KM_TYPE_NR * smp_processor_id();
+        unsigned long vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+        set_pte_ext(TOP_PTE(vaddr), pfn_pte(paddr >> PAGE_SHIFT, PAGE_KERNEL), 0);
+        local_flush_tlb_kernel_page(vaddr);
+        return vaddr + (paddr & ~PAGE_MASK);
+#else
+        return __phys_to_virt(paddr);
+#endif
+}
+
 static inline void l2_clean_pa(unsigned long addr)
 {
         __asm__("mcr p15, 1, %0, c15, c9, 3" : : "r" (addr));
 }
 
-static inline void l2_clean_mva_range(unsigned long start, unsigned long end)
+static inline void l2_clean_pa_range(unsigned long start, unsigned long end)
 {
-        unsigned long flags;
+        unsigned long va_start, va_end, flags;
 
         /*
          * Make sure 'start' and 'end' reference the same page, as
@@ -51,17 +77,14 @@ static inline void l2_clean_mva_range(unsigned long start, unsigned long end)
         BUG_ON((start ^ end) >> PAGE_SHIFT);
 
         raw_local_irq_save(flags);
+        va_start = l2_start_va(start);
+        va_end = va_start + (end - start);
         __asm__("mcr p15, 1, %0, c15, c9, 4\n\t"
                 "mcr p15, 1, %1, c15, c9, 5"
-                : : "r" (start), "r" (end));
+                : : "r" (va_start), "r" (va_end));
         raw_local_irq_restore(flags);
 }
 
-static inline void l2_clean_pa_range(unsigned long start, unsigned long end)
-{
-        l2_clean_mva_range(__phys_to_virt(start), __phys_to_virt(end));
-}
-
 static inline void l2_clean_inv_pa(unsigned long addr)
 {
         __asm__("mcr p15, 1, %0, c15, c10, 3" : : "r" (addr));
@@ -72,9 +95,9 @@ static inline void l2_inv_pa(unsigned long addr)
         __asm__("mcr p15, 1, %0, c15, c11, 3" : : "r" (addr));
 }
 
-static inline void l2_inv_mva_range(unsigned long start, unsigned long end)
+static inline void l2_inv_pa_range(unsigned long start, unsigned long end)
 {
-        unsigned long flags;
+        unsigned long va_start, va_end, flags;
 
         /*
          * Make sure 'start' and 'end' reference the same page, as
@@ -84,17 +107,14 @@ static inline void l2_inv_mva_range(unsigned long start, unsigned long end)
         BUG_ON((start ^ end) >> PAGE_SHIFT);
 
         raw_local_irq_save(flags);
+        va_start = l2_start_va(start);
+        va_end = va_start + (end - start);
         __asm__("mcr p15, 1, %0, c15, c11, 4\n\t"
                 "mcr p15, 1, %1, c15, c11, 5"
-                : : "r" (start), "r" (end));
+                : : "r" (va_start), "r" (va_end));
         raw_local_irq_restore(flags);
 }
 
-static inline void l2_inv_pa_range(unsigned long start, unsigned long end)
-{
-        l2_inv_mva_range(__phys_to_virt(start), __phys_to_virt(end));
-}
-
 
 /*
  * Linux primitives.
diff --git a/arch/arm/mm/cache-xsc3l2.c b/arch/arm/mm/cache-xsc3l2.c
index 464de893a988..5d180cb0bd94 100644
--- a/arch/arm/mm/cache-xsc3l2.c
+++ b/arch/arm/mm/cache-xsc3l2.c
@@ -17,12 +17,14 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  */
 #include <linux/init.h>
-#include <linux/spinlock.h>
-#include <linux/io.h>
-
 #include <asm/system.h>
 #include <asm/cputype.h>
 #include <asm/cacheflush.h>
+#include <asm/kmap_types.h>
+#include <asm/fixmap.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+#include "mm.h"
 
 #define CR_L2   (1 << 26)
 
@@ -47,21 +49,11 @@ static inline void xsc3_l2_clean_mva(unsigned long addr)
         __asm__("mcr p15, 1, %0, c7, c11, 1" : : "r" (addr));
 }
 
-static inline void xsc3_l2_clean_pa(unsigned long addr)
-{
-        xsc3_l2_clean_mva(__phys_to_virt(addr));
-}
-
 static inline void xsc3_l2_inv_mva(unsigned long addr)
 {
         __asm__("mcr p15, 1, %0, c7, c7, 1" : : "r" (addr));
 }
 
-static inline void xsc3_l2_inv_pa(unsigned long addr)
-{
-        xsc3_l2_inv_mva(__phys_to_virt(addr));
-}
-
 static inline void xsc3_l2_inv_all(void)
 {
         unsigned long l2ctype, set_way;
@@ -79,50 +71,103 @@ static inline void xsc3_l2_inv_all(void)
         dsb();
 }
 
+#ifdef CONFIG_HIGHMEM
+#define l2_map_save_flags(x)            raw_local_save_flags(x)
+#define l2_map_restore_flags(x)         raw_local_irq_restore(x)
+#else
+#define l2_map_save_flags(x)            ((x) = 0)
+#define l2_map_restore_flags(x)         ((void)(x))
+#endif
+
+static inline unsigned long l2_map_va(unsigned long pa, unsigned long prev_va,
+                                      unsigned long flags)
+{
+#ifdef CONFIG_HIGHMEM
+        unsigned long va = prev_va & PAGE_MASK;
+        unsigned long pa_offset = pa << (32 - PAGE_SHIFT);
+        if (unlikely(pa_offset < (prev_va << (32 - PAGE_SHIFT)))) {
+                /*
+                 * Switching to a new page.  Because cache ops are
+                 * using virtual addresses only, we must put a mapping
+                 * in place for it.  We also enable interrupts for a
+                 * short while and disable them again to protect this
+                 * mapping.
+                 */
+                unsigned long idx;
+                raw_local_irq_restore(flags);
+                idx = KM_L2_CACHE + KM_TYPE_NR * smp_processor_id();
+                va = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+                raw_local_irq_restore(flags | PSR_I_BIT);
+                set_pte_ext(TOP_PTE(va), pfn_pte(pa >> PAGE_SHIFT, PAGE_KERNEL), 0);
+                local_flush_tlb_kernel_page(va);
+        }
+        return va + (pa_offset >> (32 - PAGE_SHIFT));
+#else
+        return __phys_to_virt(pa);
+#endif
+}
+
 static void xsc3_l2_inv_range(unsigned long start, unsigned long end)
 {
+        unsigned long vaddr, flags;
+
         if (start == 0 && end == -1ul) {
                 xsc3_l2_inv_all();
                 return;
         }
 
+        vaddr = -1;  /* to force the first mapping */
+        l2_map_save_flags(flags);
+
         /*
          * Clean and invalidate partial first cache line.
          */
         if (start & (CACHE_LINE_SIZE - 1)) {
-                xsc3_l2_clean_pa(start & ~(CACHE_LINE_SIZE - 1));
-                xsc3_l2_inv_pa(start & ~(CACHE_LINE_SIZE - 1));
+                vaddr = l2_map_va(start & ~(CACHE_LINE_SIZE - 1), vaddr, flags);
+                xsc3_l2_clean_mva(vaddr);
+                xsc3_l2_inv_mva(vaddr);
                 start = (start | (CACHE_LINE_SIZE - 1)) + 1;
         }
 
         /*
-         * Clean and invalidate partial last cache line.
+         * Invalidate all full cache lines between 'start' and 'end'.
          */
-        if (start < end && (end & (CACHE_LINE_SIZE - 1))) {
-                xsc3_l2_clean_pa(end & ~(CACHE_LINE_SIZE - 1));
-                xsc3_l2_inv_pa(end & ~(CACHE_LINE_SIZE - 1));
-                end &= ~(CACHE_LINE_SIZE - 1);
+        while (start < (end & ~(CACHE_LINE_SIZE - 1))) {
+                vaddr = l2_map_va(start, vaddr, flags);
+                xsc3_l2_inv_mva(vaddr);
+                start += CACHE_LINE_SIZE;
         }
 
         /*
-         * Invalidate all full cache lines between 'start' and 'end'.
+         * Clean and invalidate partial last cache line.
          */
-        while (start < end) {
-                xsc3_l2_inv_pa(start);
-                start += CACHE_LINE_SIZE;
+        if (start < end) {
+                vaddr = l2_map_va(start, vaddr, flags);
+                xsc3_l2_clean_mva(vaddr);
+                xsc3_l2_inv_mva(vaddr);
         }
 
+        l2_map_restore_flags(flags);
+
         dsb();
 }
 
 static void xsc3_l2_clean_range(unsigned long start, unsigned long end)
 {
+        unsigned long vaddr, flags;
+
+        vaddr = -1;  /* to force the first mapping */
+        l2_map_save_flags(flags);
+
         start &= ~(CACHE_LINE_SIZE - 1);
         while (start < end) {
-                xsc3_l2_clean_pa(start);
+                vaddr = l2_map_va(start, vaddr, flags);
+                xsc3_l2_clean_mva(vaddr);
                 start += CACHE_LINE_SIZE;
         }
 
+        l2_map_restore_flags(flags);
+
         dsb();
 }
 
@@ -148,18 +193,26 @@ static inline void xsc3_l2_flush_all(void)
 
 static void xsc3_l2_flush_range(unsigned long start, unsigned long end)
 {
+        unsigned long vaddr, flags;
+
         if (start == 0 && end == -1ul) {
                 xsc3_l2_flush_all();
                 return;
         }
 
+        vaddr = -1;  /* to force the first mapping */
+        l2_map_save_flags(flags);
+
         start &= ~(CACHE_LINE_SIZE - 1);
         while (start < end) {
-                xsc3_l2_clean_pa(start);
-                xsc3_l2_inv_pa(start);
+                vaddr = l2_map_va(start, vaddr, flags);
+                xsc3_l2_clean_mva(vaddr);
+                xsc3_l2_inv_mva(vaddr);
                 start += CACHE_LINE_SIZE;
         }
 
+        l2_map_restore_flags(flags);
+
         dsb();
 }
 
diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index 310e479309ef..510c179b0ac8 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -19,6 +19,7 @@
 #include <linux/dma-mapping.h>
 
 #include <asm/memory.h>
+#include <asm/highmem.h>
 #include <asm/cacheflush.h>
 #include <asm/tlbflush.h>
 #include <asm/sizes.h>
@@ -490,29 +491,101 @@ core_initcall(consistent_init);
  */
 void dma_cache_maint(const void *start, size_t size, int direction)
 {
-        const void *end = start + size;
+        void (*inner_op)(const void *, const void *);
+        void (*outer_op)(unsigned long, unsigned long);
 
-        BUG_ON(!virt_addr_valid(start) || !virt_addr_valid(end - 1));
+        BUG_ON(!virt_addr_valid(start) || !virt_addr_valid(start + size - 1));
 
         switch (direction) {
         case DMA_FROM_DEVICE:           /* invalidate only */
-                dmac_inv_range(start, end);
-                outer_inv_range(__pa(start), __pa(end));
+                inner_op = dmac_inv_range;
+                outer_op = outer_inv_range;
                 break;
         case DMA_TO_DEVICE:             /* writeback only */
-                dmac_clean_range(start, end);
-                outer_clean_range(__pa(start), __pa(end));
+                inner_op = dmac_clean_range;
+                outer_op = outer_clean_range;
                 break;
         case DMA_BIDIRECTIONAL:         /* writeback and invalidate */
-                dmac_flush_range(start, end);
-                outer_flush_range(__pa(start), __pa(end));
+                inner_op = dmac_flush_range;
+                outer_op = outer_flush_range;
                 break;
         default:
                 BUG();
         }
+
+        inner_op(start, start + size);
+        outer_op(__pa(start), __pa(start) + size);
 }
 EXPORT_SYMBOL(dma_cache_maint);
 
+static void dma_cache_maint_contiguous(struct page *page, unsigned long offset,
+                                       size_t size, int direction)
+{
+        void *vaddr;
+        unsigned long paddr;
+        void (*inner_op)(const void *, const void *);
+        void (*outer_op)(unsigned long, unsigned long);
+
+        switch (direction) {
+        case DMA_FROM_DEVICE:           /* invalidate only */
+                inner_op = dmac_inv_range;
+                outer_op = outer_inv_range;
+                break;
+        case DMA_TO_DEVICE:             /* writeback only */
+                inner_op = dmac_clean_range;
+                outer_op = outer_clean_range;
+                break;
+        case DMA_BIDIRECTIONAL:         /* writeback and invalidate */
+                inner_op = dmac_flush_range;
+                outer_op = outer_flush_range;
+                break;
+        default:
+                BUG();
+        }
+
+        if (!PageHighMem(page)) {
+                vaddr = page_address(page) + offset;
+                inner_op(vaddr, vaddr + size);
+        } else {
+                vaddr = kmap_high_get(page);
+                if (vaddr) {
+                        vaddr += offset;
+                        inner_op(vaddr, vaddr + size);
+                        kunmap_high(page);
+                }
+        }
+
+        paddr = page_to_phys(page) + offset;
+        outer_op(paddr, paddr + size);
+}
+
+void dma_cache_maint_page(struct page *page, unsigned long offset,
+                          size_t size, int dir)
+{
+        /*
+         * A single sg entry may refer to multiple physically contiguous
+         * pages.  But we still need to process highmem pages individually.
+         * If highmem is not configured then the bulk of this loop gets
+         * optimized out.
+         */
+        size_t left = size;
+        do {
+                size_t len = left;
+                if (PageHighMem(page) && len + offset > PAGE_SIZE) {
+                        if (offset >= PAGE_SIZE) {
+                                page += offset / PAGE_SIZE;
+                                offset %= PAGE_SIZE;
+                        }
+                        len = PAGE_SIZE - offset;
+                }
+                dma_cache_maint_contiguous(page, offset, len, dir);
+                offset = 0;
+                page++;
+                left -= len;
+        } while (left);
+}
+EXPORT_SYMBOL(dma_cache_maint_page);
+
 /**
  * dma_map_sg - map a set of SG buffers for streaming mode DMA
  * @dev: valid struct device pointer, or NULL for ISA and EISA-like devices
@@ -610,7 +683,8 @@ void dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
                         continue;
 
                 if (!arch_is_coherent())
-                        dma_cache_maint(sg_virt(s), s->length, dir);
+                        dma_cache_maint_page(sg_page(s), s->offset,
+                                             s->length, dir);
         }
 }
 EXPORT_SYMBOL(dma_sync_sg_for_device);
diff --git a/arch/arm/mm/flush.c b/arch/arm/mm/flush.c
index 0fa9bf388f0b..4e283481cee1 100644
--- a/arch/arm/mm/flush.c
+++ b/arch/arm/mm/flush.c
@@ -192,7 +192,7 @@ void flush_dcache_page(struct page *page)
         struct address_space *mapping = page_mapping(page);
 
 #ifndef CONFIG_SMP
-        if (mapping && !mapping_mapped(mapping))
+        if (!PageHighMem(page) && mapping && !mapping_mapped(mapping))
                 set_bit(PG_dcache_dirty, &page->flags);
         else
 #endif
diff --git a/arch/arm/mm/highmem.c b/arch/arm/mm/highmem.c
new file mode 100644
index 000000000000..a34954d9df7d
--- /dev/null
+++ b/arch/arm/mm/highmem.c
@@ -0,0 +1,116 @@
+/*
+ * arch/arm/mm/highmem.c -- ARM highmem support
+ *
+ * Author:      Nicolas Pitre
+ * Created:     september 8, 2008
+ * Copyright:   Marvell Semiconductors Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/highmem.h>
+#include <linux/interrupt.h>
+#include <asm/fixmap.h>
+#include <asm/cacheflush.h>
+#include <asm/tlbflush.h>
+#include "mm.h"
+
+void *kmap(struct page *page)
+{
+        might_sleep();
+        if (!PageHighMem(page))
+                return page_address(page);
+        return kmap_high(page);
+}
+EXPORT_SYMBOL(kmap);
+
+void kunmap(struct page *page)
+{
+        BUG_ON(in_interrupt());
+        if (!PageHighMem(page))
+                return;
+        kunmap_high(page);
+}
+EXPORT_SYMBOL(kunmap);
+
+void *kmap_atomic(struct page *page, enum km_type type)
+{
+        unsigned int idx;
+        unsigned long vaddr;
+
+        pagefault_disable();
+        if (!PageHighMem(page))
+                return page_address(page);
+
+        idx = type + KM_TYPE_NR * smp_processor_id();
+        vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+#ifdef CONFIG_DEBUG_HIGHMEM
+        /*
+         * With debugging enabled, kunmap_atomic forces that entry to 0.
+         * Make sure it was indeed properly unmapped.
+         */
+        BUG_ON(!pte_none(*(TOP_PTE(vaddr))));
+#endif
+        set_pte_ext(TOP_PTE(vaddr), mk_pte(page, kmap_prot), 0);
+        /*
+         * When debugging is off, kunmap_atomic leaves the previous mapping
+         * in place, so this TLB flush ensures the TLB is updated with the
+         * new mapping.
+         */
+        local_flush_tlb_kernel_page(vaddr);
+
+        return (void *)vaddr;
+}
+EXPORT_SYMBOL(kmap_atomic);
+
+void kunmap_atomic(void *kvaddr, enum km_type type)
+{
+        unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK;
+        unsigned int idx = type + KM_TYPE_NR * smp_processor_id();
+
+        if (kvaddr >= (void *)FIXADDR_START) {
+                __cpuc_flush_dcache_page((void *)vaddr);
+#ifdef CONFIG_DEBUG_HIGHMEM
+                BUG_ON(vaddr != __fix_to_virt(FIX_KMAP_BEGIN + idx));
+                set_pte_ext(TOP_PTE(vaddr), __pte(0), 0);
+                local_flush_tlb_kernel_page(vaddr);
+#else
+                (void) idx;  /* to kill a warning */
+#endif
+        }
+        pagefault_enable();
+}
+EXPORT_SYMBOL(kunmap_atomic);
+
+void *kmap_atomic_pfn(unsigned long pfn, enum km_type type)
+{
+        unsigned int idx;
+        unsigned long vaddr;
+
+        pagefault_disable();
+
+        idx = type + KM_TYPE_NR * smp_processor_id();
+        vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
+#ifdef CONFIG_DEBUG_HIGHMEM
+        BUG_ON(!pte_none(*(TOP_PTE(vaddr))));
+#endif
+        set_pte_ext(TOP_PTE(vaddr), pfn_pte(pfn, kmap_prot), 0);
+        local_flush_tlb_kernel_page(vaddr);
+
+        return (void *)vaddr;
+}
+
+struct page *kmap_atomic_to_page(const void *ptr)
+{
+        unsigned long vaddr = (unsigned long)ptr;
+        pte_t *pte;
+
+        if (vaddr < FIXADDR_START)
+                return virt_to_page(ptr);
+
+        pte = TOP_PTE(vaddr);
+        return pte_page(*pte);
+}
diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
index 34df4d9d03a6..8277802ec859 100644
--- a/arch/arm/mm/init.c
+++ b/arch/arm/mm/init.c
@@ -15,6 +15,7 @@
 #include <linux/mman.h>
 #include <linux/nodemask.h>
 #include <linux/initrd.h>
+#include <linux/highmem.h>
 
 #include <asm/mach-types.h>
 #include <asm/sections.h>
@@ -382,7 +383,7 @@ void __init bootmem_init(void)
         for_each_node(node)
                 bootmem_free_node(node, mi);
 
-        high_memory = __va(memend_pfn << PAGE_SHIFT);
+        high_memory = __va((memend_pfn << PAGE_SHIFT) - 1) + 1;
 
         /*
          * This doesn't seem to be used by the Linux memory manager any
@@ -485,7 +486,7 @@ void __init mem_init(void)
         int i, node;
 
 #ifndef CONFIG_DISCONTIGMEM
-        max_mapnr   = virt_to_page(high_memory) - mem_map;
+        max_mapnr   = pfn_to_page(max_pfn + PHYS_PFN_OFFSET) - mem_map;
 #endif
 
         /* this will put all unused low memory onto the freelists */
@@ -504,6 +505,19 @@ void __init mem_init(void)
                                     __phys_to_pfn(__pa(swapper_pg_dir)), NULL);
 #endif
 
+#ifdef CONFIG_HIGHMEM
+        /* set highmem page free */
+        for_each_online_node(node) {
+                for_each_nodebank (i, &meminfo, node) {
+                        unsigned long start = bank_pfn_start(&meminfo.bank[i]);
+                        unsigned long end = bank_pfn_end(&meminfo.bank[i]);
+                        if (start >= max_low_pfn + PHYS_PFN_OFFSET)
+                                totalhigh_pages += free_area(start, end, NULL);
+                }
+        }
+        totalram_pages += totalhigh_pages;
+#endif
+
         /*
          * Since our memory may not be contiguous, calculate the
          * real number of pages we have in this system
@@ -521,9 +535,10 @@ void __init mem_init(void)
         initsize = __init_end - __init_begin;
 
         printk(KERN_NOTICE "Memory: %luKB available (%dK code, "
-                "%dK data, %dK init)\n",
+                "%dK data, %dK init, %luK highmem)\n",
                 (unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
-                codesize >> 10, datasize >> 10, initsize >> 10);
+                codesize >> 10, datasize >> 10, initsize >> 10,
+                (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)));
 
         if (PAGE_SIZE >= 16384 && num_physpages <= 128) {
                 extern int sysctl_overcommit_memory;
diff --git a/arch/arm/mm/mm.h b/arch/arm/mm/mm.h
index 95bbe112965e..c4f6f05198e0 100644
--- a/arch/arm/mm/mm.h
+++ b/arch/arm/mm/mm.h
@@ -1,7 +1,6 @@
-/* the upper-most page table pointer */
-
 #ifdef CONFIG_MMU
 
+/* the upper-most page table pointer */
 extern pmd_t *top_pmd;
 
 #define TOP_PTE(x)      pte_offset_kernel(top_pmd, x)
diff --git a/arch/arm/mm/mmap.c b/arch/arm/mm/mmap.c
index 5358fcc7f61e..f7457fea6de8 100644
--- a/arch/arm/mm/mmap.c
+++ b/arch/arm/mm/mmap.c
@@ -124,7 +124,7 @@ int valid_phys_addr_range(unsigned long addr, size_t size)
 {
         if (addr < PHYS_OFFSET)
                 return 0;
-        if (addr + size > __pa(high_memory))
+        if (addr + size >= __pa(high_memory - 1))
                 return 0;
 
         return 1;
diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index 8c6fc5a6237e..1585814f8414 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -18,9 +18,11 @@
 #include <asm/cputype.h>
 #include <asm/mach-types.h>
 #include <asm/sections.h>
+#include <asm/cachetype.h>
 #include <asm/setup.h>
 #include <asm/sizes.h>
 #include <asm/tlb.h>
+#include <asm/highmem.h>
 
 #include <asm/mach/arch.h>
 #include <asm/mach/map.h>
@@ -700,6 +702,10 @@ static void __init sanity_check_meminfo(void)
                         if (meminfo.nr_banks >= NR_BANKS) {
                                 printk(KERN_CRIT "NR_BANKS too low, "
                                                  "ignoring high memory\n");
+                        } else if (cache_is_vipt_aliasing()) {
+                                printk(KERN_CRIT "HIGHMEM is not yet supported "
+                                                 "with VIPT aliasing cache, "
+                                                 "ignoring high memory\n");
                         } else {
                                 memmove(bank + 1, bank,
                                         (meminfo.nr_banks - i) * sizeof(*bank));
@@ -918,6 +924,17 @@ static void __init devicemaps_init(struct machine_desc *mdesc)
         flush_cache_all();
 }
 
+static void __init kmap_init(void)
+{
+#ifdef CONFIG_HIGHMEM
+        pmd_t *pmd = pmd_off_k(PKMAP_BASE);
+        pte_t *pte = alloc_bootmem_low_pages(2 * PTRS_PER_PTE * sizeof(pte_t));
+        BUG_ON(!pmd_none(*pmd) || !pte);
+        __pmd_populate(pmd, __pa(pte) | _PAGE_KERNEL_TABLE);
+        pkmap_page_table = pte + PTRS_PER_PTE;
+#endif
+}
+
 /*
  * paging_init() sets up the page tables, initialises the zone memory
  * maps, and sets up the zero page, bad page and bad page tables.
@@ -931,6 +948,7 @@ void __init paging_init(struct machine_desc *mdesc)
         prepare_page_table();
         bootmem_init();
         devicemaps_init(mdesc);
+        kmap_init();
 
         top_pmd = pmd_off_k(0xffff0000);
 
diff --git a/arch/arm/plat-omap/include/mach/memory.h b/arch/arm/plat-omap/include/mach/memory.h
index d6b5ca6c7da2..99ed564d9277 100644
--- a/arch/arm/plat-omap/include/mach/memory.h
+++ b/arch/arm/plat-omap/include/mach/memory.h
@@ -61,9 +61,11 @@
 #define lbus_to_virt(x)         ((x) - OMAP1510_LB_OFFSET + PAGE_OFFSET)
 #define is_lbus_device(dev)     (cpu_is_omap15xx() && dev && (strncmp(dev_name(dev), "ohci", 4) == 0))
 
-#define __arch_page_to_dma(dev, page)   ({is_lbus_device(dev) ? \
-                                        (dma_addr_t)virt_to_lbus(page_address(page)) : \
-                                        (dma_addr_t)__virt_to_phys(page_address(page));})
+#define __arch_page_to_dma(dev, page)   \
+        ({ dma_addr_t __dma = page_to_phys(page); \
+           if (is_lbus_device(dev)) \
+                __dma = __dma - PHYS_OFFSET + OMAP1510_LB_OFFSET; \
+           __dma; })
 
 #define __arch_dma_to_virt(dev, addr)   ({ (void *) (is_lbus_device(dev) ? \
                                                 lbus_to_virt(addr) : \