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
  ...

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)

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)) : \

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);
-}
-
-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);
-}
+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);
 
-#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)		      \

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)

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

arch/sparc/kernel/pci.c

@@ -779,7 +779,7 @@ static int __pci_mmap_make_offset(struct pci_dev *pdev,
 static void __pci_mmap_set_flags(struct pci_dev *dev, struct vm_area_struct *vma,
 					    enum pci_mmap_state mmap_state)
 {
-	vma->vm_flags |= (VM_IO | VM_RESERVED);
+	vma->vm_flags |= VM_IO | VM_DONTEXPAND | VM_DONTDUMP;
 }
 
 /* Set vm_page_prot of VMA, as appropriate for this architecture, for a pci

arch/sparc/kernel/sun4v_tlb_miss.S

@@ -176,7 +176,7 @@ sun4v_tsb_miss_common:
 
 	sub	%g2, TRAP_PER_CPU_FAULT_INFO, %g2
 
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	mov	SCRATCHPAD_UTSBREG2, %g5
 	ldxa	[%g5] ASI_SCRATCHPAD, %g5
 	cmp	%g5, -1

arch/sparc/kernel/tsb.S

@@ -49,7 +49,7 @@ tsb_miss_page_table_walk:
 	/* Before committing to a full page table walk,
 	 * check the huge page TSB.
 	 */
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 
 661:	ldx		[%g7 + TRAP_PER_CPU_TSB_HUGE], %g5
 	nop
@@ -110,12 +110,9 @@ tsb_miss_page_table_walk:
 tsb_miss_page_table_walk_sun4v_fastpath:
 	USER_PGTABLE_WALK_TL1(%g4, %g7, %g5, %g2, tsb_do_fault)
 
-	/* Load and check PTE.  */
-	ldxa		[%g5] ASI_PHYS_USE_EC, %g5
-	brgez,pn	%g5, tsb_do_fault
-	 nop
+	/* Valid PTE is now in %g5.  */
 
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 661:	sethi		%uhi(_PAGE_SZALL_4U), %g7
 	sllx		%g7, 32, %g7
 	.section	.sun4v_2insn_patch, "ax"

arch/sparc/mm/fault_32.c

@@ -265,6 +265,7 @@ asmlinkage void do_sparc_fault(struct pt_regs *regs, int text_fault, int write,
 		}
 		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

arch/sparc/mm/fault_64.c

@@ -452,6 +452,7 @@ asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
 		}
 		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
@@ -464,13 +465,13 @@ asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
 	up_read(&mm->mmap_sem);
 
 	mm_rss = get_mm_rss(mm);
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	mm_rss -= (mm->context.huge_pte_count * (HPAGE_SIZE / PAGE_SIZE));
 #endif
 	if (unlikely(mm_rss >
 		     mm->context.tsb_block[MM_TSB_BASE].tsb_rss_limit))
 		tsb_grow(mm, MM_TSB_BASE, mm_rss);
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	mm_rss = mm->context.huge_pte_count;
 	if (unlikely(mm_rss >
 		     mm->context.tsb_block[MM_TSB_HUGE].tsb_rss_limit))

arch/sparc/mm/hugetlbpage.c

@@ -303,53 +303,3 @@ struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
 {
 	return NULL;
 }
-
-static void context_reload(void *__data)
-{
-	struct mm_struct *mm = __data;
-
-	if (mm == current->mm)
-		load_secondary_context(mm);
-}
-
-void hugetlb_prefault_arch_hook(struct mm_struct *mm)
-{
-	struct tsb_config *tp = &mm->context.tsb_block[MM_TSB_HUGE];
-
-	if (likely(tp->tsb != NULL))
-		return;
-
-	tsb_grow(mm, MM_TSB_HUGE, 0);
-	tsb_context_switch(mm);
-	smp_tsb_sync(mm);
-
-	/* On UltraSPARC-III+ and later, configure the second half of
-	 * the Data-TLB for huge pages.
-	 */
-	if (tlb_type == cheetah_plus) {
-		unsigned long ctx;
-
-		spin_lock(&ctx_alloc_lock);
-		ctx = mm->context.sparc64_ctx_val;
-		ctx &= ~CTX_PGSZ_MASK;
-		ctx |= CTX_PGSZ_BASE << CTX_PGSZ0_SHIFT;
-		ctx |= CTX_PGSZ_HUGE << CTX_PGSZ1_SHIFT;
-
-		if (ctx != mm->context.sparc64_ctx_val) {
-			/* When changing the page size fields, we
-			 * must perform a context flush so that no
-			 * stale entries match.  This flush must
-			 * occur with the original context register
-			 * settings.
-			 */
-			do_flush_tlb_mm(mm);
-
-			/* Reload the context register of all processors
-			 * also executing in this address space.
-			 */
-			mm->context.sparc64_ctx_val = ctx;
-			on_each_cpu(context_reload, mm, 0);
-		}
-		spin_unlock(&ctx_alloc_lock);
-	}
-}

arch/sparc/mm/init_64.c

@@ -276,7 +276,6 @@ static inline void tsb_insert(struct tsb *ent, unsigned long tag, unsigned long
 }
 
 unsigned long _PAGE_ALL_SZ_BITS __read_mostly;
-unsigned long _PAGE_SZBITS __read_mostly;
 
 static void flush_dcache(unsigned long pfn)
 {
@@ -307,12 +306,24 @@ static void flush_dcache(unsigned long pfn)
 	}
 }
 
+/* mm->context.lock must be held */
+static void __update_mmu_tsb_insert(struct mm_struct *mm, unsigned long tsb_index,
+				    unsigned long tsb_hash_shift, unsigned long address,
+				    unsigned long tte)
+{
+	struct tsb *tsb = mm->context.tsb_block[tsb_index].tsb;
+	unsigned long tag;
+
+	tsb += ((address >> tsb_hash_shift) &
+		(mm->context.tsb_block[tsb_index].tsb_nentries - 1UL));
+	tag = (address >> 22UL);
+	tsb_insert(tsb, tag, tte);
+}
+
 void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *ptep)
 {
+	unsigned long tsb_index, tsb_hash_shift, flags;
 	struct mm_struct *mm;
-	struct tsb *tsb;
-	unsigned long tag, flags;
-	unsigned long tsb_index, tsb_hash_shift;
 	pte_t pte = *ptep;
 
 	if (tlb_type != hypervisor) {
@@ -329,7 +340,7 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *
 
 	spin_lock_irqsave(&mm->context.lock, flags);
 
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb != NULL) {
 		if ((tlb_type == hypervisor &&
 		     (pte_val(pte) & _PAGE_SZALL_4V) == _PAGE_SZHUGE_4V) ||
@@ -341,11 +352,8 @@ void update_mmu_cache(struct vm_area_struct *vma, unsigned long address, pte_t *
 	}
 #endif
 
-	tsb = mm->context.tsb_block[tsb_index].tsb;
-	tsb += ((address >> tsb_hash_shift) &
-		(mm->context.tsb_block[tsb_index].tsb_nentries - 1UL));
-	tag = (address >> 22UL);
-	tsb_insert(tsb, tag, pte_val(pte));
+	__update_mmu_tsb_insert(mm, tsb_index, tsb_hash_shift,
+				address, pte_val(pte));
 
 	spin_unlock_irqrestore(&mm->context.lock, flags);
 }
@@ -2275,8 +2283,7 @@ static void __init sun4u_pgprot_init(void)
 		     __ACCESS_BITS_4U | _PAGE_E_4U);
 
 #ifdef CONFIG_DEBUG_PAGEALLOC
-	kern_linear_pte_xor[0] = (_PAGE_VALID | _PAGE_SZBITS_4U) ^
-		0xfffff80000000000UL;
+	kern_linear_pte_xor[0] = _PAGE_VALID ^ 0xfffff80000000000UL;
 #else
 	kern_linear_pte_xor[0] = (_PAGE_VALID | _PAGE_SZ4MB_4U) ^
 		0xfffff80000000000UL;
@@ -2287,7 +2294,6 @@ static void __init sun4u_pgprot_init(void)
 	for (i = 1; i < 4; i++)
 		kern_linear_pte_xor[i] = kern_linear_pte_xor[0];
 
-	_PAGE_SZBITS = _PAGE_SZBITS_4U;
 	_PAGE_ALL_SZ_BITS =  (_PAGE_SZ4MB_4U | _PAGE_SZ512K_4U |
 			      _PAGE_SZ64K_4U | _PAGE_SZ8K_4U |
 			      _PAGE_SZ32MB_4U | _PAGE_SZ256MB_4U);
@@ -2324,8 +2330,7 @@ static void __init sun4v_pgprot_init(void)
 	_PAGE_CACHE = _PAGE_CACHE_4V;
 
 #ifdef CONFIG_DEBUG_PAGEALLOC
-	kern_linear_pte_xor[0] = (_PAGE_VALID | _PAGE_SZBITS_4V) ^
-		0xfffff80000000000UL;
+	kern_linear_pte_xor[0] = _PAGE_VALID ^ 0xfffff80000000000UL;
 #else
 	kern_linear_pte_xor[0] = (_PAGE_VALID | _PAGE_SZ4MB_4V) ^
 		0xfffff80000000000UL;
@@ -2339,7 +2344,6 @@ static void __init sun4v_pgprot_init(void)
 	pg_iobits = (_PAGE_VALID | _PAGE_PRESENT_4V | __DIRTY_BITS_4V |
 		     __ACCESS_BITS_4V | _PAGE_E_4V);
 
-	_PAGE_SZBITS = _PAGE_SZBITS_4V;
 	_PAGE_ALL_SZ_BITS = (_PAGE_SZ16GB_4V | _PAGE_SZ2GB_4V |
 			     _PAGE_SZ256MB_4V | _PAGE_SZ32MB_4V |
 			     _PAGE_SZ4MB_4V | _PAGE_SZ512K_4V |
@@ -2472,3 +2476,281 @@ void __flush_tlb_all(void)
 	__asm__ __volatile__("wrpr	%0, 0, %%pstate"
 			     : : "r" (pstate));
 }
+
+static pte_t *get_from_cache(struct mm_struct *mm)
+{
+	struct page *page;
+	pte_t *ret;
+
+	spin_lock(&mm->page_table_lock);
+	page = mm->context.pgtable_page;
+	ret = NULL;
+	if (page) {
+		void *p = page_address(page);
+
+		mm->context.pgtable_page = NULL;
+
+		ret = (pte_t *) (p + (PAGE_SIZE / 2));
+	}
+	spin_unlock(&mm->page_table_lock);
+
+	return ret;
+}
+
+static struct page *__alloc_for_cache(struct mm_struct *mm)
+{
+	struct page *page = alloc_page(GFP_KERNEL | __GFP_NOTRACK |
+				       __GFP_REPEAT | __GFP_ZERO);
+
+	if (page) {
+		spin_lock(&mm->page_table_lock);
+		if (!mm->context.pgtable_page) {
+			atomic_set(&page->_count, 2);
+			mm->context.pgtable_page = page;
+		}
+		spin_unlock(&mm->page_table_lock);
+	}
+	return page;
+}
+
+pte_t *pte_alloc_one_kernel(struct mm_struct *mm,
+			    unsigned long address)
+{
+	struct page *page;
+	pte_t *pte;
+
+	pte = get_from_cache(mm);
+	if (pte)
+		return pte;
+
+	page = __alloc_for_cache(mm);
+	if (page)
+		pte = (pte_t *) page_address(page);
+
+	return pte;
+}
+
+pgtable_t pte_alloc_one(struct mm_struct *mm,
+			unsigned long address)
+{
+	struct page *page;
+	pte_t *pte;
+
+	pte = get_from_cache(mm);
+	if (pte)
+		return pte;
+
+	page = __alloc_for_cache(mm);
+	if (page) {
+		pgtable_page_ctor(page);
+		pte = (pte_t *) page_address(page);
+	}
+
+	return pte;
+}
+
+void pte_free_kernel(struct mm_struct *mm, pte_t *pte)
+{
+	struct page *page = virt_to_page(pte);
+	if (put_page_testzero(page))
+		free_hot_cold_page(page, 0);
+}
+
+static void __pte_free(pgtable_t pte)
+{
+	struct page *page = virt_to_page(pte);
+	if (put_page_testzero(page)) {
+		pgtable_page_dtor(page);
+		free_hot_cold_page(page, 0);
+	}
+}
+
+void pte_free(struct mm_struct *mm, pgtable_t pte)
+{
+	__pte_free(pte);
+}
+
+void pgtable_free(void *table, bool is_page)
+{
+	if (is_page)
+		__pte_free(table);
+	else
+		kmem_cache_free(pgtable_cache, table);
+}
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+static pmd_t pmd_set_protbits(pmd_t pmd, pgprot_t pgprot, bool for_modify)
+{
+	if (pgprot_val(pgprot) & _PAGE_VALID)
+		pmd_val(pmd) |= PMD_HUGE_PRESENT;
+	if (tlb_type == hypervisor) {
+		if (pgprot_val(pgprot) & _PAGE_WRITE_4V)
+			pmd_val(pmd) |= PMD_HUGE_WRITE;
+		if (pgprot_val(pgprot) & _PAGE_EXEC_4V)
+			pmd_val(pmd) |= PMD_HUGE_EXEC;
+
+		if (!for_modify) {
+			if (pgprot_val(pgprot) & _PAGE_ACCESSED_4V)
+				pmd_val(pmd) |= PMD_HUGE_ACCESSED;
+			if (pgprot_val(pgprot) & _PAGE_MODIFIED_4V)
+				pmd_val(pmd) |= PMD_HUGE_DIRTY;
+		}
+	} else {
+		if (pgprot_val(pgprot) & _PAGE_WRITE_4U)
+			pmd_val(pmd) |= PMD_HUGE_WRITE;
+		if (pgprot_val(pgprot) & _PAGE_EXEC_4U)
+			pmd_val(pmd) |= PMD_HUGE_EXEC;
+
+		if (!for_modify) {
+			if (pgprot_val(pgprot) & _PAGE_ACCESSED_4U)
+				pmd_val(pmd) |= PMD_HUGE_ACCESSED;
+			if (pgprot_val(pgprot) & _PAGE_MODIFIED_4U)
+				pmd_val(pmd) |= PMD_HUGE_DIRTY;
+		}
+	}
+
+	return pmd;
+}
+
+pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
+{
+	pmd_t pmd;
+
+	pmd_val(pmd) = (page_nr << ((PAGE_SHIFT - PMD_PADDR_SHIFT)));
+	pmd_val(pmd) |= PMD_ISHUGE;
+	pmd = pmd_set_protbits(pmd, pgprot, false);
+	return pmd;
+}
+
+pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
+{
+	pmd_val(pmd) &= ~(PMD_HUGE_PRESENT |
+			  PMD_HUGE_WRITE |
+			  PMD_HUGE_EXEC);
+	pmd = pmd_set_protbits(pmd, newprot, true);
+	return pmd;
+}
+
+pgprot_t pmd_pgprot(pmd_t entry)
+{
+	unsigned long pte = 0;
+
+	if (pmd_val(entry) & PMD_HUGE_PRESENT)
+		pte |= _PAGE_VALID;
+
+	if (tlb_type == hypervisor) {
+		if (pmd_val(entry) & PMD_HUGE_PRESENT)
+			pte |= _PAGE_PRESENT_4V;
+		if (pmd_val(entry) & PMD_HUGE_EXEC)
+			pte |= _PAGE_EXEC_4V;
+		if (pmd_val(entry) & PMD_HUGE_WRITE)
+			pte |= _PAGE_W_4V;
+		if (pmd_val(entry) & PMD_HUGE_ACCESSED)
+			pte |= _PAGE_ACCESSED_4V;
+		if (pmd_val(entry) & PMD_HUGE_DIRTY)
+			pte |= _PAGE_MODIFIED_4V;
+		pte |= _PAGE_CP_4V|_PAGE_CV_4V;
+	} else {
+		if (pmd_val(entry) & PMD_HUGE_PRESENT)
+			pte |= _PAGE_PRESENT_4U;
+		if (pmd_val(entry) & PMD_HUGE_EXEC)
+			pte |= _PAGE_EXEC_4U;
+		if (pmd_val(entry) & PMD_HUGE_WRITE)
+			pte |= _PAGE_W_4U;
+		if (pmd_val(entry) & PMD_HUGE_ACCESSED)
+			pte |= _PAGE_ACCESSED_4U;
+		if (pmd_val(entry) & PMD_HUGE_DIRTY)
+			pte |= _PAGE_MODIFIED_4U;
+		pte |= _PAGE_CP_4U|_PAGE_CV_4U;
+	}
+
+	return __pgprot(pte);
+}
+
+void update_mmu_cache_pmd(struct vm_area_struct *vma, unsigned long addr,
+			  pmd_t *pmd)
+{
+	unsigned long pte, flags;
+	struct mm_struct *mm;
+	pmd_t entry = *pmd;
+	pgprot_t prot;
+
+	if (!pmd_large(entry) || !pmd_young(entry))
+		return;
+
+	pte = (pmd_val(entry) & ~PMD_HUGE_PROTBITS);
+	pte <<= PMD_PADDR_SHIFT;
+	pte |= _PAGE_VALID;
+
+	prot = pmd_pgprot(entry);
+
+	if (tlb_type == hypervisor)
+		pgprot_val(prot) |= _PAGE_SZHUGE_4V;
+	else
+		pgprot_val(prot) |= _PAGE_SZHUGE_4U;
+
+	pte |= pgprot_val(prot);
+
+	mm = vma->vm_mm;
+
+	spin_lock_irqsave(&mm->context.lock, flags);
+
+	if (mm->context.tsb_block[MM_TSB_HUGE].tsb != NULL)
+		__update_mmu_tsb_insert(mm, MM_TSB_HUGE, HPAGE_SHIFT,
+					addr, pte);
+
+	spin_unlock_irqrestore(&mm->context.lock, flags);
+}
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
+
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
+static void context_reload(void *__data)
+{
+	struct mm_struct *mm = __data;
+
+	if (mm == current->mm)
+		load_secondary_context(mm);
+}
+
+void hugetlb_setup(struct mm_struct *mm)
+{
+	struct tsb_config *tp = &mm->context.tsb_block[MM_TSB_HUGE];
+
+	if (likely(tp->tsb != NULL))
+		return;
+
+	tsb_grow(mm, MM_TSB_HUGE, 0);
+	tsb_context_switch(mm);
+	smp_tsb_sync(mm);
+
+	/* On UltraSPARC-III+ and later, configure the second half of
+	 * the Data-TLB for huge pages.
+	 */
+	if (tlb_type == cheetah_plus) {
+		unsigned long ctx;
+
+		spin_lock(&ctx_alloc_lock);
+		ctx = mm->context.sparc64_ctx_val;
+		ctx &= ~CTX_PGSZ_MASK;
+		ctx |= CTX_PGSZ_BASE << CTX_PGSZ0_SHIFT;
+		ctx |= CTX_PGSZ_HUGE << CTX_PGSZ1_SHIFT;
+
+		if (ctx != mm->context.sparc64_ctx_val) {
+			/* When changing the page size fields, we
+			 * must perform a context flush so that no
+			 * stale entries match.  This flush must
+			 * occur with the original context register
+			 * settings.
+			 */
+			do_flush_tlb_mm(mm);
+
+			/* Reload the context register of all processors
+			 * also executing in this address space.
+			 */
+			mm->context.sparc64_ctx_val = ctx;
+			on_each_cpu(context_reload, mm, 0);
+		}
+		spin_unlock(&ctx_alloc_lock);
+	}
+}
+#endif

arch/sparc/mm/tlb.c

@@ -43,16 +43,37 @@ void flush_tlb_pending(void)
 	put_cpu_var(tlb_batch);
 }
 
-void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
-		   pte_t *ptep, pte_t orig, int fullmm)
+static void tlb_batch_add_one(struct mm_struct *mm, unsigned long vaddr,
+			      bool exec)
 {
 	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 	unsigned long nr;
 
 	vaddr &= PAGE_MASK;
-	if (pte_exec(orig))
+	if (exec)
 		vaddr |= 0x1UL;
 
+	nr = tb->tlb_nr;
+
+	if (unlikely(nr != 0 && mm != tb->mm)) {
+		flush_tlb_pending();
+		nr = 0;
+	}
+
+	if (nr == 0)
+		tb->mm = mm;
+
+	tb->vaddrs[nr] = vaddr;
+	tb->tlb_nr = ++nr;
+	if (nr >= TLB_BATCH_NR)
+		flush_tlb_pending();
+
+	put_cpu_var(tlb_batch);
+}
+
+void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+		   pte_t *ptep, pte_t orig, int fullmm)
+{
 	if (tlb_type != hypervisor &&
 	    pte_dirty(orig)) {
 		unsigned long paddr, pfn = pte_pfn(orig);
@@ -77,26 +98,91 @@ void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
 	}
 
 no_cache_flush:
+	if (!fullmm)
+		tlb_batch_add_one(mm, vaddr, pte_exec(orig));
+}
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+static void tlb_batch_pmd_scan(struct mm_struct *mm, unsigned long vaddr,
+			       pmd_t pmd, bool exec)
+{
+	unsigned long end;
+	pte_t *pte;
+
+	pte = pte_offset_map(&pmd, vaddr);
+	end = vaddr + HPAGE_SIZE;
+	while (vaddr < end) {
+		if (pte_val(*pte) & _PAGE_VALID)
+			tlb_batch_add_one(mm, vaddr, exec);
+		pte++;
+		vaddr += PAGE_SIZE;
+	}
+	pte_unmap(pte);
+}
 
-	if (fullmm) {
-		put_cpu_var(tlb_batch);
+void set_pmd_at(struct mm_struct *mm, unsigned long addr,
+		pmd_t *pmdp, pmd_t pmd)
+{
+	pmd_t orig = *pmdp;
+
+	*pmdp = pmd;
+
+	if (mm == &init_mm)
 		return;
+
+	if ((pmd_val(pmd) ^ pmd_val(orig)) & PMD_ISHUGE) {
+		if (pmd_val(pmd) & PMD_ISHUGE)
+			mm->context.huge_pte_count++;
+		else
+			mm->context.huge_pte_count--;
+		if (mm->context.huge_pte_count == 1)
+			hugetlb_setup(mm);
 	}
 
-	nr = tb->tlb_nr;
+	if (!pmd_none(orig)) {
+		bool exec = ((pmd_val(orig) & PMD_HUGE_EXEC) != 0);
 
-	if (unlikely(nr != 0 && mm != tb->mm)) {
-		flush_tlb_pending();
-		nr = 0;
+		addr &= HPAGE_MASK;
+		if (pmd_val(orig) & PMD_ISHUGE)
+			tlb_batch_add_one(mm, addr, exec);
+		else
+			tlb_batch_pmd_scan(mm, addr, orig, exec);
 	}
+}
 
-	if (nr == 0)
-		tb->mm = mm;
+void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable)
+{
+	struct list_head *lh = (struct list_head *) pgtable;
 
-	tb->vaddrs[nr] = vaddr;
-	tb->tlb_nr = ++nr;
-	if (nr >= TLB_BATCH_NR)
-		flush_tlb_pending();
+	assert_spin_locked(&mm->page_table_lock);
 
-	put_cpu_var(tlb_batch);
+	/* FIFO */
+	if (!mm->pmd_huge_pte)
+		INIT_LIST_HEAD(lh);
+	else
+		list_add(lh, (struct list_head *) mm->pmd_huge_pte);
+	mm->pmd_huge_pte = pgtable;
+}
+
+pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm)
+{
+	struct list_head *lh;
+	pgtable_t pgtable;
+
+	assert_spin_locked(&mm->page_table_lock);
+
+	/* FIFO */
+	pgtable = mm->pmd_huge_pte;
+	lh = (struct list_head *) pgtable;
+	if (list_empty(lh))
+		mm->pmd_huge_pte = NULL;
+	else {
+		mm->pmd_huge_pte = (pgtable_t) lh->next;
+		list_del(lh);
+	}
+	pte_val(pgtable[0]) = 0;
+	pte_val(pgtable[1]) = 0;
+
+	return pgtable;
 }
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */

arch/sparc/mm/tsb.c

@@ -78,7 +78,7 @@ void flush_tsb_user(struct tlb_batch *tb)
 		base = __pa(base);
 	__flush_tsb_one(tb, PAGE_SHIFT, base, nentries);
 
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb) {
 		base = (unsigned long) mm->context.tsb_block[MM_TSB_HUGE].tsb;
 		nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries;
@@ -90,29 +90,12 @@ void flush_tsb_user(struct tlb_batch *tb)
 	spin_unlock_irqrestore(&mm->context.lock, flags);
 }
 
-#if defined(CONFIG_SPARC64_PAGE_SIZE_8KB)
 #define HV_PGSZ_IDX_BASE	HV_PGSZ_IDX_8K
 #define HV_PGSZ_MASK_BASE	HV_PGSZ_MASK_8K
-#elif defined(CONFIG_SPARC64_PAGE_SIZE_64KB)
-#define HV_PGSZ_IDX_BASE	HV_PGSZ_IDX_64K
-#define HV_PGSZ_MASK_BASE	HV_PGSZ_MASK_64K
-#else
-#error Broken base page size setting...
-#endif
 
-#ifdef CONFIG_HUGETLB_PAGE
-#if defined(CONFIG_HUGETLB_PAGE_SIZE_64K)
-#define HV_PGSZ_IDX_HUGE	HV_PGSZ_IDX_64K
-#define HV_PGSZ_MASK_HUGE	HV_PGSZ_MASK_64K
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_512K)
-#define HV_PGSZ_IDX_HUGE	HV_PGSZ_IDX_512K
-#define HV_PGSZ_MASK_HUGE	HV_PGSZ_MASK_512K
-#elif defined(CONFIG_HUGETLB_PAGE_SIZE_4MB)
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 #define HV_PGSZ_IDX_HUGE	HV_PGSZ_IDX_4MB
 #define HV_PGSZ_MASK_HUGE	HV_PGSZ_MASK_4MB
-#else
-#error Broken huge page size setting...
-#endif
 #endif
 
 static void setup_tsb_params(struct mm_struct *mm, unsigned long tsb_idx, unsigned long tsb_bytes)
@@ -207,7 +190,7 @@ static void setup_tsb_params(struct mm_struct *mm, unsigned long tsb_idx, unsign
 		case MM_TSB_BASE:
 			hp->pgsz_idx = HV_PGSZ_IDX_BASE;
 			break;
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 		case MM_TSB_HUGE:
 			hp->pgsz_idx = HV_PGSZ_IDX_HUGE;
 			break;
@@ -222,7 +205,7 @@ static void setup_tsb_params(struct mm_struct *mm, unsigned long tsb_idx, unsign
 		case MM_TSB_BASE:
 			hp->pgsz_mask = HV_PGSZ_MASK_BASE;
 			break;
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 		case MM_TSB_HUGE:
 			hp->pgsz_mask = HV_PGSZ_MASK_HUGE;
 			break;
@@ -444,7 +427,7 @@ void tsb_grow(struct mm_struct *mm, unsigned long tsb_index, unsigned long rss)
 
 int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 {
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	unsigned long huge_pte_count;
 #endif
 	unsigned int i;
@@ -453,7 +436,7 @@ int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 
 	mm->context.sparc64_ctx_val = 0UL;
 
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	/* We reset it to zero because the fork() page copying
 	 * will re-increment the counters as the parent PTEs are
 	 * copied into the child address space.
@@ -462,6 +445,8 @@ int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 	mm->context.huge_pte_count = 0;
 #endif
 
+	mm->context.pgtable_page = NULL;
+
 	/* copy_mm() copies over the parent's mm_struct before calling
 	 * us, so we need to zero out the TSB pointer or else tsb_grow()
 	 * will be confused and think there is an older TSB to free up.
@@ -474,7 +459,7 @@ int init_new_context(struct task_struct *tsk, struct mm_struct *mm)
 	 */
 	tsb_grow(mm, MM_TSB_BASE, get_mm_rss(mm));
 
-#ifdef CONFIG_HUGETLB_PAGE
+#if defined(CONFIG_HUGETLB_PAGE) || defined(CONFIG_TRANSPARENT_HUGEPAGE)
 	if (unlikely(huge_pte_count))
 		tsb_grow(mm, MM_TSB_HUGE, huge_pte_count);
 #endif
@@ -500,10 +485,17 @@ static void tsb_destroy_one(struct tsb_config *tp)
 void destroy_context(struct mm_struct *mm)
 {
 	unsigned long flags, i;
+	struct page *page;
 
 	for (i = 0; i < MM_NUM_TSBS; i++)
 		tsb_destroy_one(&mm->context.tsb_block[i]);
 
+	page = mm->context.pgtable_page;
+	if (page && put_page_testzero(page)) {
+		pgtable_page_dtor(page);
+		free_hot_cold_page(page, 0);
+	}
+
 	spin_lock_irqsave(&ctx_alloc_lock, flags);
 
 	if (CTX_VALID(mm->context)) {

arch/tile/Kconfig

@@ -7,12 +7,15 @@ config TILE
 	select HAVE_DMA_API_DEBUG
 	select HAVE_KVM if !TILEGX
 	select GENERIC_FIND_FIRST_BIT
+	select SYSCTL_EXCEPTION_TRACE
 	select USE_GENERIC_SMP_HELPERS
 	select CC_OPTIMIZE_FOR_SIZE
+	select HAVE_DEBUG_KMEMLEAK
 	select HAVE_GENERIC_HARDIRQS
 	select GENERIC_IRQ_PROBE
 	select GENERIC_PENDING_IRQ if SMP
 	select GENERIC_IRQ_SHOW
+	select HAVE_DEBUG_BUGVERBOSE
 	select HAVE_SYSCALL_WRAPPERS if TILEGX
 	select SYS_HYPERVISOR
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG

arch/tile/include/asm/hugetlb.h

@@ -106,6 +106,10 @@ static inline void arch_release_hugepage(struct page *page)
 {
 }
 
+static inline void arch_clear_hugepage_flags(struct page *page)
+{
+}
+
 #ifdef CONFIG_HUGETLB_SUPER_PAGES
 static inline pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma,
 				       struct page *page, int writable)

arch/tile/mm/elf.c

@@ -36,19 +36,14 @@ static void sim_notify_exec(const char *binary_name)
 	} while (c);
 }
 
-static int notify_exec(void)
+static int notify_exec(struct mm_struct *mm)
 {
 	int retval = 0;  /* failure */
-	struct vm_area_struct *vma = current->mm->mmap;
-	while (vma) {
-		if ((vma->vm_flags & VM_EXECUTABLE) && vma->vm_file)
-			break;
-		vma = vma->vm_next;
-	}
-	if (vma) {
+
+	if (mm->exe_file) {
 		char *buf = (char *) __get_free_page(GFP_KERNEL);
 		if (buf) {
-			char *path = d_path(&vma->vm_file->f_path,
+			char *path = d_path(&mm->exe_file->f_path,
 					    buf, PAGE_SIZE);
 			if (!IS_ERR(path)) {
 				sim_notify_exec(path);
@@ -106,16 +101,16 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 	unsigned long vdso_base;
 	int retval = 0;
 
+	down_write(&mm->mmap_sem);
+
 	/*
 	 * Notify the simulator that an exec just occurred.
 	 * If we can't find the filename of the mapping, just use
 	 * whatever was passed as the linux_binprm filename.
 	 */
-	if (!notify_exec())
+	if (!notify_exec(mm))
 		sim_notify_exec(bprm->filename);
 
-	down_write(&mm->mmap_sem);
-
 	/*
 	 * MAYWRITE to allow gdb to COW and set breakpoints
 	 */

arch/tile/mm/fault.c

@@ -454,6 +454,7 @@ static int handle_page_fault(struct pt_regs *regs,
 			tsk->min_flt++;
 		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

arch/um/Kconfig.common

@@ -7,6 +7,7 @@ config UML
 	bool
 	default y
 	select HAVE_GENERIC_HARDIRQS
+	select HAVE_UID16
 	select GENERIC_IRQ_SHOW
 	select GENERIC_CPU_DEVICES
 	select GENERIC_IO

arch/um/kernel/trap.c

@@ -89,6 +89,7 @@ int handle_page_fault(unsigned long address, unsigned long ip,
 				current->min_flt++;
 			if (fault & VM_FAULT_RETRY) {
 				flags &= ~FAULT_FLAG_ALLOW_RETRY;
+				flags |= FAULT_FLAG_TRIED;
 
 				goto retry;
 			}