Merge branch 'akpm' (Andrew's patch-bomb)

Merge patches from Andrew Morton: "A few misc things and very nearly all of the MM tree. A tremendous amount of stuff (again), including a significant rbtree library rework." * emailed patches from Andrew Morton <akpm@linux-foundation.org>: (160 commits) sparc64: Support transparent huge pages. mm: thp: Use more portable PMD clearing sequenece in zap_huge_pmd(). mm: Add and use update_mmu_cache_pmd() in transparent huge page code. sparc64: Document PGD and PMD layout. sparc64: Eliminate PTE table memory wastage. sparc64: Halve the size of PTE tables sparc64: Only support 4MB huge pages and 8KB base pages. memory-hotplug: suppress "Trying to free nonexistent resource <XXXXXXXXXXXXXXXX-YYYYYYYYYYYYYYYY>" warning mm: memcg: clean up mm_match_cgroup() signature mm: document PageHuge somewhat mm: use %pK for /proc/vmallocinfo mm, thp: fix mlock statistics mm, thp: fix mapped pages avoiding unevictable list on mlock memory-hotplug: update memory block's state and notify userspace memory-hotplug: preparation to notify memory block's state at memory hot remove mm: avoid section mismatch warning for memblock_type_name make GFP_NOTRACK definition unconditional cma: decrease cc.nr_migratepages after reclaiming pagelist CMA: migrate mlocked pages kpageflags: fix wrong KPF_THP on non-huge compound pages ...
2012-10-09 16:23:15 +09:00
parent 1ea4f4f840 9e695d2ecc
commit 9e2d8656f5
255 changed files with 5013 additions and 3577 deletions
--- a/arch/sparc/include/asm/hugetlb.h
+++ b/arch/sparc/include/asm/hugetlb.h
@@ -10,7 +10,10 @@ void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
 pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 			      pte_t *ptep);

-void hugetlb_prefault_arch_hook(struct mm_struct *mm);
+static inline void hugetlb_prefault_arch_hook(struct mm_struct *mm)
+{
+	hugetlb_setup(mm);
+}

 static inline int is_hugepage_only_range(struct mm_struct *mm,
 					 unsigned long addr,
@@ -82,4 +85,8 @@ static inline void arch_release_hugepage(struct page *page)
 {
 }

+static inline void arch_clear_hugepage_flags(struct page *page)
+{
+}
+
 #endif /* _ASM_SPARC64_HUGETLB_H */
--- a/arch/sparc/include/asm/mmu_64.h
+++ b/arch/sparc/include/asm/mmu_64.h
@@ -30,22 +30,8 @@
 #define CTX_PGSZ_MASK		((CTX_PGSZ_BITS << CTX_PGSZ0_SHIFT) | \
 				 (CTX_PGSZ_BITS << CTX_PGSZ1_SHIFT))

-#if defined(CONFIG_SPARC64_PAGE_SIZE_8KB)
 #define CTX_PGSZ_BASE	CTX_PGSZ_8KB
-#elif defined(CONFIG_SPARC64_PAGE_SIZE_64KB)
-#define CTX_PGSZ_BASE	CTX_PGSZ_64KB
-#else
-#error No page size specified in kernel configuration
-#endif
-
-#if defined(CONFIG_HUGETLB_PAGE_SIZE_4MB)
-#define CTX_PGSZ_HUGE		CTX_PGSZ_4MB
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_512K)
-#define CTX_PGSZ_HUGE		CTX_PGSZ_512KB
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_64K)
-#define CTX_PGSZ_HUGE		CTX_PGSZ_64KB
-#endif
-
+#define CTX_PGSZ_HUGE	CTX_PGSZ_4MB
 #define CTX_PGSZ_KERN	CTX_PGSZ_4MB

 /* Thus, when running on UltraSPARC-III+ and later, we use the following
@@ -96,7 +82,7 @@ struct tsb_config {

 #define MM_TSB_BASE	0

-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 #define MM_TSB_HUGE	1
 #define MM_NUM_TSBS	2
 #else
@@ -107,6 +93,7 @@ typedef struct {
 	spinlock_t		lock;
 	unsigned long		sparc64_ctx_val;
 	unsigned long		huge_pte_count;
+	struct page		*pgtable_page;
 	struct tsb_config	tsb_block[MM_NUM_TSBS];
 	struct hv_tsb_descr	tsb_descr[MM_NUM_TSBS];
 } mm_context_t;
--- a/arch/sparc/include/asm/mmu_context_64.h
+++ b/arch/sparc/include/asm/mmu_context_64.h
@@ -36,7 +36,7 @@ static inline void tsb_context_switch(struct mm_struct *mm)
 {
 	__tsb_context_switch(__pa(mm->pgd),
 			     &mm->context.tsb_block[0],
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 			     (mm->context.tsb_block[1].tsb ?
 			      &mm->context.tsb_block[1] :
 			      NULL)
--- a/arch/sparc/include/asm/page_64.h
+++ b/arch/sparc/include/asm/page_64.h
@@ -3,13 +3,7 @@

 #include <linux/const.h>

-#if defined(CONFIG_SPARC64_PAGE_SIZE_8KB)
 #define PAGE_SHIFT   13
-#elif defined(CONFIG_SPARC64_PAGE_SIZE_64KB)
-#define PAGE_SHIFT   16
-#else
-#error No page size specified in kernel configuration
-#endif

 #define PAGE_SIZE    (_AC(1,UL) << PAGE_SHIFT)
 #define PAGE_MASK    (~(PAGE_SIZE-1))
@@ -21,15 +15,9 @@
 #define DCACHE_ALIASING_POSSIBLE
 #endif

-#if defined(CONFIG_HUGETLB_PAGE_SIZE_4MB)
 #define HPAGE_SHIFT		22
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_512K)
-#define HPAGE_SHIFT		19
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_64K)
-#define HPAGE_SHIFT		16
-#endif

-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 #define HPAGE_SIZE		(_AC(1,UL) << HPAGE_SHIFT)
 #define HPAGE_MASK		(~(HPAGE_SIZE - 1UL))
 #define HUGETLB_PAGE_ORDER	(HPAGE_SHIFT - PAGE_SHIFT)
@@ -38,6 +26,11 @@

 #ifndef __ASSEMBLY__

+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
+struct mm_struct;
+extern void hugetlb_setup(struct mm_struct *mm);
+#endif
+
 #define WANT_PAGE_VIRTUAL

 extern void _clear_page(void *page);
@@ -98,7 +91,7 @@ typedef unsigned long pgprot_t;

 #endif /* (STRICT_MM_TYPECHECKS) */

-typedef struct page *pgtable_t;
+typedef pte_t *pgtable_t;

 #define TASK_UNMAPPED_BASE	(test_thread_flag(TIF_32BIT) ? \
 				 (_AC(0x0000000070000000,UL)) : \
--- a/arch/sparc/include/asm/pgalloc_64.h
+++ b/arch/sparc/include/asm/pgalloc_64.h
@@ -38,51 +38,20 @@ static inline void pmd_free(struct mm_struct *mm, pmd_t *pmd)
 	kmem_cache_free(pgtable_cache, pmd);
 }

-static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
-					  unsigned long address)
-{
-	return (pte_t *)__get_free_page(GFP_KERNEL | __GFP_REPEAT | __GFP_ZERO);
-}
+extern pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
+				   unsigned long address);
+extern pgtable_t pte_alloc_one(struct mm_struct *mm,
+			       unsigned long address);
+extern void pte_free_kernel(struct mm_struct *mm, pte_t *pte);
+extern void pte_free(struct mm_struct *mm, pgtable_t ptepage);

-static inline pgtable_t pte_alloc_one(struct mm_struct *mm,
-					unsigned long address)
-{
-	struct page *page;
-	pte_t *pte;
-
-	pte = pte_alloc_one_kernel(mm, address);
-	if (!pte)
-		return NULL;
-	page = virt_to_page(pte);
-	pgtable_page_ctor(page);
-	return page;
-}
-
-static inline void pte_free_kernel(struct mm_struct *mm, pte_t *pte)
-{
-	free_page((unsigned long)pte);
-}
-
-static inline void pte_free(struct mm_struct *mm, pgtable_t ptepage)
-{
-	pgtable_page_dtor(ptepage);
-	__free_page(ptepage);
-}
-
-#define pmd_populate_kernel(MM, PMD, PTE)	pmd_set(PMD, PTE)
-#define pmd_populate(MM,PMD,PTE_PAGE)		\
-	pmd_populate_kernel(MM,PMD,page_address(PTE_PAGE))
-#define pmd_pgtable(pmd) pmd_page(pmd)
+#define pmd_populate_kernel(MM, PMD, PTE)	pmd_set(MM, PMD, PTE)
+#define pmd_populate(MM, PMD, PTE)		pmd_set(MM, PMD, PTE)
+#define pmd_pgtable(PMD)			((pte_t *)__pmd_page(PMD))

 #define check_pgt_cache()	do { } while (0)

-static inline void pgtable_free(void *table, bool is_page)
-{
-	if (is_page)
-		free_page((unsigned long)table);
-	else
-		kmem_cache_free(pgtable_cache, table);
-}
+extern void pgtable_free(void *table, bool is_page);

 #ifdef CONFIG_SMP

@@ -113,11 +82,10 @@ static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, bool is
 }
 #endif /* !CONFIG_SMP */

-static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
+static inline void __pte_free_tlb(struct mmu_gather *tlb, pte_t *pte,
 				  unsigned long address)
 {
-	pgtable_page_dtor(ptepage);
-	pgtable_free_tlb(tlb, page_address(ptepage), true);
+	pgtable_free_tlb(tlb, pte, true);
 }

 #define __pmd_free_tlb(tlb, pmd, addr)		      \
--- a/arch/sparc/include/asm/pgtable_64.h
+++ b/arch/sparc/include/asm/pgtable_64.h
@@ -45,40 +45,59 @@

 #define vmemmap			((struct page *)VMEMMAP_BASE)

-/* XXX All of this needs to be rethought so we can take advantage
- * XXX cheetah's full 64-bit virtual address space, ie. no more hole
- * XXX in the middle like on spitfire. -DaveM
- */
-/*
- * Given a virtual address, the lowest PAGE_SHIFT bits determine offset
- * into the page; the next higher PAGE_SHIFT-3 bits determine the pte#
- * in the proper pagetable (the -3 is from the 8 byte ptes, and each page
- * table is a single page long). The next higher PMD_BITS determine pmd#
- * in the proper pmdtable (where we must have PMD_BITS <= (PAGE_SHIFT-2)
- * since the pmd entries are 4 bytes, and each pmd page is a single page
- * long). Finally, the higher few bits determine pgde#.
- */
-
 /* PMD_SHIFT determines the size of the area a second-level page
 * table can map
 */
-#define PMD_SHIFT	(PAGE_SHIFT + (PAGE_SHIFT-3))
+#define PMD_SHIFT	(PAGE_SHIFT + (PAGE_SHIFT-4))
 #define PMD_SIZE	(_AC(1,UL) << PMD_SHIFT)
 #define PMD_MASK	(~(PMD_SIZE-1))
 #define PMD_BITS	(PAGE_SHIFT - 2)

 /* PGDIR_SHIFT determines what a third-level page table entry can map */
-#define PGDIR_SHIFT	(PAGE_SHIFT + (PAGE_SHIFT-3) + PMD_BITS)
+#define PGDIR_SHIFT	(PAGE_SHIFT + (PAGE_SHIFT-4) + PMD_BITS)
 #define PGDIR_SIZE	(_AC(1,UL) << PGDIR_SHIFT)
 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
 #define PGDIR_BITS	(PAGE_SHIFT - 2)

+#if (PGDIR_SHIFT + PGDIR_BITS) != 44
+#error Page table parameters do not cover virtual address space properly.
+#endif
+
+#if (PMD_SHIFT != HPAGE_SHIFT)
+#error PMD_SHIFT must equal HPAGE_SHIFT for transparent huge pages.
+#endif
+
+/* PMDs point to PTE tables which are 4K aligned.  */
+#define PMD_PADDR	_AC(0xfffffffe,UL)
+#define PMD_PADDR_SHIFT	_AC(11,UL)
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+#define PMD_ISHUGE	_AC(0x00000001,UL)
+
+/* This is the PMD layout when PMD_ISHUGE is set.  With 4MB huge
+ * pages, this frees up a bunch of bits in the layout that we can
+ * use for the protection settings and software metadata.
+ */
+#define PMD_HUGE_PADDR		_AC(0xfffff800,UL)
+#define PMD_HUGE_PROTBITS	_AC(0x000007ff,UL)
+#define PMD_HUGE_PRESENT	_AC(0x00000400,UL)
+#define PMD_HUGE_WRITE		_AC(0x00000200,UL)
+#define PMD_HUGE_DIRTY		_AC(0x00000100,UL)
+#define PMD_HUGE_ACCESSED	_AC(0x00000080,UL)
+#define PMD_HUGE_EXEC		_AC(0x00000040,UL)
+#define PMD_HUGE_SPLITTING	_AC(0x00000020,UL)
+#endif
+
+/* PGDs point to PMD tables which are 8K aligned.  */
+#define PGD_PADDR	_AC(0xfffffffc,UL)
+#define PGD_PADDR_SHIFT	_AC(11,UL)
+
 #ifndef __ASSEMBLY__

 #include <linux/sched.h>

 /* Entries per page directory level. */
-#define PTRS_PER_PTE	(1UL << (PAGE_SHIFT-3))
+#define PTRS_PER_PTE	(1UL << (PAGE_SHIFT-4))
 #define PTRS_PER_PMD	(1UL << PMD_BITS)
 #define PTRS_PER_PGD	(1UL << PGDIR_BITS)

@@ -160,26 +179,11 @@
 #define _PAGE_SZ8K_4V	  _AC(0x0000000000000000,UL) /* 8K Page              */
 #define _PAGE_SZALL_4V	  _AC(0x0000000000000007,UL) /* All pgsz bits        */

-#if PAGE_SHIFT == 13
 #define _PAGE_SZBITS_4U	_PAGE_SZ8K_4U
 #define _PAGE_SZBITS_4V	_PAGE_SZ8K_4V
-#elif PAGE_SHIFT == 16
-#define _PAGE_SZBITS_4U	_PAGE_SZ64K_4U
-#define _PAGE_SZBITS_4V	_PAGE_SZ64K_4V
-#else
-#error Wrong PAGE_SHIFT specified
-#endif

-#if defined(CONFIG_HUGETLB_PAGE_SIZE_4MB)
 #define _PAGE_SZHUGE_4U	_PAGE_SZ4MB_4U
 #define _PAGE_SZHUGE_4V	_PAGE_SZ4MB_4V
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_512K)
-#define _PAGE_SZHUGE_4U	_PAGE_SZ512K_4U
-#define _PAGE_SZHUGE_4V	_PAGE_SZ512K_4V
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_64K)
-#define _PAGE_SZHUGE_4U	_PAGE_SZ64K_4U
-#define _PAGE_SZHUGE_4V	_PAGE_SZ64K_4V
-#endif

 /* These are actually filled in at boot time by sun4{u,v}_pgprot_init() */
 #define __P000	__pgprot(0)
@@ -218,7 +222,6 @@ extern unsigned long _PAGE_CACHE;

 extern unsigned long pg_iobits;
 extern unsigned long _PAGE_ALL_SZ_BITS;
-extern unsigned long _PAGE_SZBITS;

 extern struct page *mem_map_zero;
 #define ZERO_PAGE(vaddr)	(mem_map_zero)
@@ -231,25 +234,25 @@ extern struct page *mem_map_zero;
 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t prot)
 {
 	unsigned long paddr = pfn << PAGE_SHIFT;
-	unsigned long sz_bits;

-	sz_bits = 0UL;
-	if (_PAGE_SZBITS_4U != 0UL || _PAGE_SZBITS_4V != 0UL) {
-		__asm__ __volatile__(
-		"\n661:	sethi		%%uhi(%1), %0\n"
-		"	sllx		%0, 32, %0\n"
-		"	.section	.sun4v_2insn_patch, \"ax\"\n"
-		"	.word		661b\n"
-		"	mov		%2, %0\n"
-		"	nop\n"
-		"	.previous\n"
-		: "=r" (sz_bits)
-		: "i" (_PAGE_SZBITS_4U), "i" (_PAGE_SZBITS_4V));
-	}
-	return __pte(paddr | sz_bits | pgprot_val(prot));
+	BUILD_BUG_ON(_PAGE_SZBITS_4U != 0UL || _PAGE_SZBITS_4V != 0UL);
+	return __pte(paddr | pgprot_val(prot));
 }
 #define mk_pte(page, pgprot)	pfn_pte(page_to_pfn(page), (pgprot))

+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+extern pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot);
+#define mk_pmd(page, pgprot)	pfn_pmd(page_to_pfn(page), (pgprot))
+
+extern pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot);
+
+static inline pmd_t pmd_mkhuge(pmd_t pmd)
+{
+	/* Do nothing, mk_pmd() does this part.  */
+	return pmd;
+}
+#endif
+
 /* This one can be done with two shifts.  */
 static inline unsigned long pte_pfn(pte_t pte)
 {
@@ -286,6 +289,7 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t prot)
 	 * Note: We encode this into 3 sun4v 2-insn patch sequences.
 	 */

+	BUILD_BUG_ON(_PAGE_SZBITS_4U != 0UL || _PAGE_SZBITS_4V != 0UL);
 	__asm__ __volatile__(
 	"\n661:	sethi		%%uhi(%2), %1\n"
 	"	sethi		%%hi(%2), %0\n"
@@ -307,10 +311,10 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t prot)
 	: "=r" (mask), "=r" (tmp)
 	: "i" (_PAGE_PADDR_4U | _PAGE_MODIFIED_4U | _PAGE_ACCESSED_4U |
 	       _PAGE_CP_4U | _PAGE_CV_4U | _PAGE_E_4U | _PAGE_PRESENT_4U |
-	       _PAGE_SZBITS_4U | _PAGE_SPECIAL),
+	       _PAGE_SPECIAL),
 	  "i" (_PAGE_PADDR_4V | _PAGE_MODIFIED_4V | _PAGE_ACCESSED_4V |
 	       _PAGE_CP_4V | _PAGE_CV_4V | _PAGE_E_4V | _PAGE_PRESENT_4V |
-	       _PAGE_SZBITS_4V | _PAGE_SPECIAL));
+	       _PAGE_SPECIAL));

 	return __pte((pte_val(pte) & mask) | (pgprot_val(prot) & ~mask));
 }
@@ -618,19 +622,130 @@ static inline unsigned long pte_special(pte_t pte)
 	return pte_val(pte) & _PAGE_SPECIAL;
 }

-#define pmd_set(pmdp, ptep)	\
-	(pmd_val(*(pmdp)) = (__pa((unsigned long) (ptep)) >> 11UL))
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+static inline int pmd_young(pmd_t pmd)
+{
+	return pmd_val(pmd) & PMD_HUGE_ACCESSED;
+}
+
+static inline int pmd_write(pmd_t pmd)
+{
+	return pmd_val(pmd) & PMD_HUGE_WRITE;
+}
+
+static inline unsigned long pmd_pfn(pmd_t pmd)
+{
+	unsigned long val = pmd_val(pmd) & PMD_HUGE_PADDR;
+
+	return val >> (PAGE_SHIFT - PMD_PADDR_SHIFT);
+}
+
+static inline int pmd_large(pmd_t pmd)
+{
+	return (pmd_val(pmd) & (PMD_ISHUGE | PMD_HUGE_PRESENT)) ==
+		(PMD_ISHUGE | PMD_HUGE_PRESENT);
+}
+
+static inline int pmd_trans_splitting(pmd_t pmd)
+{
+	return (pmd_val(pmd) & (PMD_ISHUGE|PMD_HUGE_SPLITTING)) ==
+		(PMD_ISHUGE|PMD_HUGE_SPLITTING);
+}
+
+static inline int pmd_trans_huge(pmd_t pmd)
+{
+	return pmd_val(pmd) & PMD_ISHUGE;
+}
+
+#define has_transparent_hugepage() 1
+
+static inline pmd_t pmd_mkold(pmd_t pmd)
+{
+	pmd_val(pmd) &= ~PMD_HUGE_ACCESSED;
+	return pmd;
+}
+
+static inline pmd_t pmd_wrprotect(pmd_t pmd)
+{
+	pmd_val(pmd) &= ~PMD_HUGE_WRITE;
+	return pmd;
+}
+
+static inline pmd_t pmd_mkdirty(pmd_t pmd)
+{
+	pmd_val(pmd) |= PMD_HUGE_DIRTY;
+	return pmd;
+}
+
+static inline pmd_t pmd_mkyoung(pmd_t pmd)
+{
+	pmd_val(pmd) |= PMD_HUGE_ACCESSED;
+	return pmd;
+}
+
+static inline pmd_t pmd_mkwrite(pmd_t pmd)
+{
+	pmd_val(pmd) |= PMD_HUGE_WRITE;
+	return pmd;
+}
+
+static inline pmd_t pmd_mknotpresent(pmd_t pmd)
+{
+	pmd_val(pmd) &= ~PMD_HUGE_PRESENT;
+	return pmd;
+}
+
+static inline pmd_t pmd_mksplitting(pmd_t pmd)
+{
+	pmd_val(pmd) |= PMD_HUGE_SPLITTING;
+	return pmd;
+}
+
+extern pgprot_t pmd_pgprot(pmd_t entry);
+#endif
+
+static inline int pmd_present(pmd_t pmd)
+{
+	return pmd_val(pmd) != 0U;
+}
+
+#define pmd_none(pmd)			(!pmd_val(pmd))
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+extern void set_pmd_at(struct mm_struct *mm, unsigned long addr,
+		       pmd_t *pmdp, pmd_t pmd);
+#else
+static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
+			      pmd_t *pmdp, pmd_t pmd)
+{
+	*pmdp = pmd;
+}
+#endif
+
+static inline void pmd_set(struct mm_struct *mm, pmd_t *pmdp, pte_t *ptep)
+{
+	unsigned long val = __pa((unsigned long) (ptep)) >> PMD_PADDR_SHIFT;
+
+	pmd_val(*pmdp) = val;
+}
+
 #define pud_set(pudp, pmdp)	\
-	(pud_val(*(pudp)) = (__pa((unsigned long) (pmdp)) >> 11UL))
-#define __pmd_page(pmd)		\
-	((unsigned long) __va((((unsigned long)pmd_val(pmd))<<11UL)))
+	(pud_val(*(pudp)) = (__pa((unsigned long) (pmdp)) >> PGD_PADDR_SHIFT))
+static inline unsigned long __pmd_page(pmd_t pmd)
+{
+	unsigned long paddr = (unsigned long) pmd_val(pmd);
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	if (pmd_val(pmd) & PMD_ISHUGE)
+		paddr &= PMD_HUGE_PADDR;
+#endif
+	paddr <<= PMD_PADDR_SHIFT;
+	return ((unsigned long) __va(paddr));
+}
 #define pmd_page(pmd) 			virt_to_page((void *)__pmd_page(pmd))
 #define pud_page_vaddr(pud)		\
-	((unsigned long) __va((((unsigned long)pud_val(pud))<<11UL)))
+	((unsigned long) __va((((unsigned long)pud_val(pud))<<PGD_PADDR_SHIFT)))
 #define pud_page(pud) 			virt_to_page((void *)pud_page_vaddr(pud))
-#define pmd_none(pmd)			(!pmd_val(pmd))
 #define pmd_bad(pmd)			(0)
-#define pmd_present(pmd)		(pmd_val(pmd) != 0U)
 #define pmd_clear(pmdp)			(pmd_val(*(pmdp)) = 0U)
 #define pud_none(pud)			(!pud_val(pud))
 #define pud_bad(pud)			(0)
@@ -664,6 +779,16 @@ static inline unsigned long pte_special(pte_t pte)
 extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
 			  pte_t *ptep, pte_t orig, int fullmm);

+#define __HAVE_ARCH_PMDP_GET_AND_CLEAR
+static inline pmd_t pmdp_get_and_clear(struct mm_struct *mm,
+				       unsigned long addr,
+				       pmd_t *pmdp)
+{
+	pmd_t pmd = *pmdp;
+	set_pmd_at(mm, addr, pmdp, __pmd(0U));
+	return pmd;
+}
+
 static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
 			     pte_t *ptep, pte_t pte, int fullmm)
 {
@@ -719,6 +844,16 @@ extern void mmu_info(struct seq_file *);

 struct vm_area_struct;
 extern void update_mmu_cache(struct vm_area_struct *, unsigned long, pte_t *);
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+extern void update_mmu_cache_pmd(struct vm_area_struct *vma, unsigned long addr,
+				 pmd_t *pmd);
+
+#define __HAVE_ARCH_PGTABLE_DEPOSIT
+extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
+
+#define __HAVE_ARCH_PGTABLE_WITHDRAW
+extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm);
+#endif

 /* Encode and de-code a swap entry */
 #define __swp_type(entry)	(((entry).val >> PAGE_SHIFT) & 0xffUL)
--- a/arch/sparc/include/asm/tsb.h
+++ b/arch/sparc/include/asm/tsb.h
@@ -147,20 +147,96 @@ extern struct tsb_phys_patch_entry __tsb_phys_patch, __tsb_phys_patch_end;
 	brz,pn		REG1, FAIL_LABEL; \
 	 sllx		VADDR, 64 - (PMD_SHIFT + PMD_BITS), REG2; \
 	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
-	sllx		REG1, 11, REG1; \
+	sllx		REG1, PGD_PADDR_SHIFT, REG1; \
 	andn		REG2, 0x3, REG2; \
 	lduwa		[REG1 + REG2] ASI_PHYS_USE_EC, REG1; \
 	brz,pn		REG1, FAIL_LABEL; \
 	 sllx		VADDR, 64 - PMD_SHIFT, REG2; \
-	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
-	sllx		REG1, 11, REG1; \
+	srlx		REG2, 64 - (PAGE_SHIFT - 1), REG2; \
+	sllx		REG1, PMD_PADDR_SHIFT, REG1; \
 	andn		REG2, 0x7, REG2; \
 	add		REG1, REG2, REG1;

-	/* Do a user page table walk in MMU globals.  Leaves physical PTE
-	 * pointer in REG1.  Jumps to FAIL_LABEL on early page table walk
-	 * termination.  Physical base of page tables is in PHYS_PGD which
-	 * will not be modified.
+	/* This macro exists only to make the PMD translator below easier
+	 * to read.  It hides the ELF section switch for the sun4v code
+	 * patching.
+	 */
+#define OR_PTE_BIT(REG, NAME)				\
+661:	or		REG, _PAGE_##NAME##_4U, REG;	\
+	.section	.sun4v_1insn_patch, "ax";	\
+	.word		661b;				\
+	or		REG, _PAGE_##NAME##_4V, REG;	\
+	.previous;
+
+	/* Load into REG the PTE value for VALID, CACHE, and SZHUGE.  */
+#define BUILD_PTE_VALID_SZHUGE_CACHE(REG)				   \
+661:	sethi		%uhi(_PAGE_VALID|_PAGE_SZHUGE_4U), REG;		   \
+	.section	.sun4v_1insn_patch, "ax";			   \
+	.word		661b;						   \
+	sethi		%uhi(_PAGE_VALID), REG;				   \
+	.previous;							   \
+	sllx		REG, 32, REG;					   \
+661:	or		REG, _PAGE_CP_4U|_PAGE_CV_4U, REG;		   \
+	.section	.sun4v_1insn_patch, "ax";			   \
+	.word		661b;						   \
+	or		REG, _PAGE_CP_4V|_PAGE_CV_4V|_PAGE_SZHUGE_4V, REG; \
+	.previous;
+
+	/* PMD has been loaded into REG1, interpret the value, seeing
+	 * if it is a HUGE PMD or a normal one.  If it is not valid
+	 * then jump to FAIL_LABEL.  If it is a HUGE PMD, and it
+	 * translates to a valid PTE, branch to PTE_LABEL.
+	 *
+	 * We translate the PMD by hand, one bit at a time,
+	 * constructing the huge PTE.
+	 *
+	 * So we construct the PTE in REG2 as follows:
+	 *
+	 * 1) Extract the PMD PFN from REG1 and place it into REG2.
+	 *
+	 * 2) Translate PMD protection bits in REG1 into REG2, one bit
+	 *    at a time using andcc tests on REG1 and OR's into REG2.
+	 *
+	 *    Only two bits to be concerned with here, EXEC and WRITE.
+	 *    Now REG1 is freed up and we can use it as a temporary.
+	 *
+	 * 3) Construct the VALID, CACHE, and page size PTE bits in
+	 *    REG1, OR with REG2 to form final PTE.
+	 */
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+#define USER_PGTABLE_CHECK_PMD_HUGE(VADDR, REG1, REG2, FAIL_LABEL, PTE_LABEL) \
+	brz,pn		REG1, FAIL_LABEL;				      \
+	 andcc		REG1, PMD_ISHUGE, %g0;				      \
+	be,pt		%xcc, 700f;					      \
+	 and		REG1, PMD_HUGE_PRESENT|PMD_HUGE_ACCESSED, REG2;	      \
+	cmp		REG2, PMD_HUGE_PRESENT|PMD_HUGE_ACCESSED;	      \
+	bne,pn		%xcc, FAIL_LABEL;				      \
+	 andn		REG1, PMD_HUGE_PROTBITS, REG2;			      \
+	sllx		REG2, PMD_PADDR_SHIFT, REG2;			      \
+	/* REG2 now holds PFN << PAGE_SHIFT */				      \
+	andcc		REG1, PMD_HUGE_EXEC, %g0;			      \
+	bne,a,pt	%xcc, 1f;					      \
+	 OR_PTE_BIT(REG2, EXEC);					      \
+1:	andcc		REG1, PMD_HUGE_WRITE, %g0;			      \
+	bne,a,pt	%xcc, 1f;					      \
+	 OR_PTE_BIT(REG2, W);						      \
+	/* REG1 can now be clobbered, build final PTE */		      \
+1:	BUILD_PTE_VALID_SZHUGE_CACHE(REG1);				      \
+	ba,pt		%xcc, PTE_LABEL;				      \
+	 or		REG1, REG2, REG1;				      \
+700:
+#else
+#define USER_PGTABLE_CHECK_PMD_HUGE(VADDR, REG1, REG2, FAIL_LABEL, PTE_LABEL) \
+	brz,pn		REG1, FAIL_LABEL; \
+	 nop;
+#endif
+
+	/* Do a user page table walk in MMU globals.  Leaves final,
+	 * valid, PTE value in REG1.  Jumps to FAIL_LABEL on early
+	 * page table walk termination or if the PTE is not valid.
+	 *
+	 * Physical base of page tables is in PHYS_PGD which will not
+	 * be modified.
 	 *
 	 * VADDR will not be clobbered, but REG1 and REG2 will.
 	 */
@@ -172,15 +248,19 @@ extern struct tsb_phys_patch_entry __tsb_phys_patch, __tsb_phys_patch_end;
 	brz,pn		REG1, FAIL_LABEL; \
 	 sllx		VADDR, 64 - (PMD_SHIFT + PMD_BITS), REG2; \
 	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
-	sllx		REG1, 11, REG1; \
+	sllx		REG1, PGD_PADDR_SHIFT, REG1; \
 	andn		REG2, 0x3, REG2; \
 	lduwa		[REG1 + REG2] ASI_PHYS_USE_EC, REG1; \
-	brz,pn		REG1, FAIL_LABEL; \
-	 sllx		VADDR, 64 - PMD_SHIFT, REG2; \
-	srlx		REG2, 64 - PAGE_SHIFT, REG2; \
-	sllx		REG1, 11, REG1; \
+	USER_PGTABLE_CHECK_PMD_HUGE(VADDR, REG1, REG2, FAIL_LABEL, 800f) \
+	sllx		VADDR, 64 - PMD_SHIFT, REG2; \
+	srlx		REG2, 64 - (PAGE_SHIFT - 1), REG2; \
+	sllx		REG1, PMD_PADDR_SHIFT, REG1; \
 	andn		REG2, 0x7, REG2; \
-	add		REG1, REG2, REG1;
+	add		REG1, REG2, REG1; \
+	ldxa		[REG1] ASI_PHYS_USE_EC, REG1; \
+	brgez,pn	REG1, FAIL_LABEL; \
+	 nop; \
+800:

 /* Lookup a OBP mapping on VADDR in the prom_trans[] table at TL>0.
 * If no entry is found, FAIL_LABEL will be branched to.  On success