[PATCH] resurrect CONFIG_HIGHPTE

Bill Irwin's patch to fix up pte's in highmem.

With CONFIG_HIGHPTE, the direct pte pointer in struct page becomes the
64-bit physical address of the single pte which is mapping this page.

If the page is not PageDirect then page->pte.chain points at a list of
pte_chains, which each now contain an array of 64-bit physical
addresses of the pte's which are mapping the page.

The functions rmap_ptep_map() and rmap_ptep_unmap() are used for
mapping and unmapping the page which backs the target pte.

The patch touches all architectures (adding do-nothing compatibility
macros and inlines).  It generally mangles lots of header files and may
break non-ia32 compiles.  I've had it in testing since 2.5.31.

arch/i386/config.in

@@ -210,6 +210,10 @@ if [ "$CONFIG_HIGHMEM64G" = "y" ]; then
    define_bool CONFIG_X86_PAE y
 fi
 
+if [ "$CONFIG_HIGHMEM4G" = "y" -o "$CONFIG_HIGHMEM64G" = "y" ]; then
+   bool 'Allocate 3rd-level pagetables from highmem' CONFIG_HIGHPTE
+fi
+
 bool 'Math emulation' CONFIG_MATH_EMULATION
 bool 'MTRR (Memory Type Range Register) support' CONFIG_MTRR
 

include/asm-alpha/pgtable.h

@@ -370,4 +370,6 @@ extern void paging_init(void);
 /* We have our own get_unmapped_area to cope with ADDR_LIMIT_32BIT.  */
 #define HAVE_ARCH_UNMAPPED_AREA
 
+typedef pte_t *pte_addr_t;
+
 #endif /* _ALPHA_PGTABLE_H */

include/asm-arm/pgtable.h

@@ -161,6 +161,8 @@ extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
 #define io_remap_page_range(vma,from,phys,size,prot) \
 		remap_page_range(vma,from,phys,size,prot)
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 #endif /* _ASMARM_PGTABLE_H */

include/asm-cris/pgtable.h

@@ -515,4 +515,6 @@ static inline void update_mmu_cache(struct vm_area_struct * vma,
  */
 #define pgtable_cache_init()   do { } while (0)
 
+typedef pte_t *pte_addr_t;
+
 #endif /* _CRIS_PGTABLE_H */

include/asm-generic/rmap.h

@@ -13,6 +13,16 @@
  * - page->index has the high bits of the address
  * - the lower bits of the address are calculated from the
  *   offset of the page table entry within the page table page
+ *
+ * For CONFIG_HIGHPTE, we need to represent the address of a pte in a
+ * scalar pte_addr_t.  The pfn of the pte's page is shifted left by PAGE_SIZE
+ * bits and is then ORed with the byte offset of the pte within its page.
+ *
+ * For CONFIG_HIGHMEM4G, the pte_addr_t is 32 bits.  20 for the pfn, 12 for
+ * the offset.
+ *
+ * For CONFIG_HIGHMEM64G, the pte_addr_t is 64 bits.  52 for the pfn, 12 for
+ * the offset.
  */
 #include <linux/mm.h>
 
@@ -39,16 +49,42 @@ static inline void pgtable_remove_rmap(struct page * page)
 
 static inline struct mm_struct * ptep_to_mm(pte_t * ptep)
 {
-	struct page * page = virt_to_page(ptep);
+	struct page * page = kmap_atomic_to_page(ptep);
 	return (struct mm_struct *) page->mapping;
 }
 
 static inline unsigned long ptep_to_address(pte_t * ptep)
 {
-	struct page * page = virt_to_page(ptep);
+	struct page * page = kmap_atomic_to_page(ptep);
 	unsigned long low_bits;
 	low_bits = ((unsigned long)ptep & ~PAGE_MASK) * PTRS_PER_PTE;
 	return page->index + low_bits;
 }
 
+#if CONFIG_HIGHPTE
+static inline pte_addr_t ptep_to_paddr(pte_t *ptep)
+{
+	pte_addr_t paddr;
+	paddr = ((pte_addr_t)page_to_pfn(kmap_atomic_to_page(ptep))) << PAGE_SHIFT;
+	return paddr + (pte_addr_t)((unsigned long)ptep & ~PAGE_MASK);
+}
+#else
+static inline pte_addr_t ptep_to_paddr(pte_t *ptep)
+{
+	return (pte_addr_t)ptep;
+}
+#endif
+
+#ifndef CONFIG_HIGHPTE
+static inline pte_t *rmap_ptep_map(pte_addr_t pte_paddr)
+{
+	return (pte_t *)pte_paddr;
+}
+
+static inline void rmap_ptep_unmap(pte_t *pte)
+{
+	return;
+}
+#endif
+
 #endif /* _GENERIC_RMAP_H */

include/asm-i386/fixmap.h

@@ -103,6 +103,7 @@ extern void __set_fixmap (enum fixed_addresses idx,
 #define FIXADDR_START	(FIXADDR_TOP - __FIXADDR_SIZE)
 
 #define __fix_to_virt(x)	(FIXADDR_TOP - ((x) << PAGE_SHIFT))
+#define __virt_to_fix(x)	((FIXADDR_TOP - ((x)&PAGE_MASK)) >> PAGE_SHIFT)
 
 extern void __this_fixmap_does_not_exist(void);
 
@@ -128,4 +129,10 @@ static inline unsigned long fix_to_virt(const unsigned int idx)
         return __fix_to_virt(idx);
 }
 
+static inline unsigned long virt_to_fix(const unsigned long vaddr)
+{
+	BUG_ON(vaddr >= FIXADDR_TOP || vaddr < FIXADDR_START);
+	return __virt_to_fix(vaddr);
+}
+
 #endif

include/asm-i386/highmem.h

@@ -122,6 +122,19 @@ static inline void kunmap_atomic(void *kvaddr, enum km_type type)
 	dec_preempt_count();
 }
 
+static inline struct page *kmap_atomic_to_page(void *ptr)
+{
+	unsigned long idx, vaddr = (unsigned long)ptr;
+	pte_t *pte;
+
+	if (vaddr < FIXADDR_START)
+		return virt_to_page(ptr);
+
+	idx = virt_to_fix(vaddr);
+	pte = kmap_pte - (idx - FIX_KMAP_BEGIN);
+	return pte_page(*pte);
+}
+
 #endif /* __KERNEL__ */
 
 #endif /* _ASM_HIGHMEM_H */

include/asm-i386/kmap_types.h

@@ -19,9 +19,10 @@ D(5)	KM_BIO_SRC_IRQ,
 D(6)	KM_BIO_DST_IRQ,
 D(7)	KM_PTE0,
 D(8)	KM_PTE1,
-D(9)	KM_IRQ0,
-D(10)	KM_IRQ1,
-D(11)	KM_TYPE_NR
+D(9)	KM_PTE2,
+D(10)	KM_IRQ0,
+D(11)	KM_IRQ1,
+D(12)	KM_TYPE_NR
 };
 
 #undef D

include/asm-i386/pgtable.h

@@ -266,6 +266,18 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 #define pte_unmap(pte) kunmap_atomic(pte, KM_PTE0)
 #define pte_unmap_nested(pte) kunmap_atomic(pte, KM_PTE1)
 
+#if defined(CONFIG_HIGHPTE) && defined(CONFIG_HIGHMEM4G)
+typedef u32 pte_addr_t;
+#endif
+
+#if defined(CONFIG_HIGHPTE) && defined(CONFIG_HIGHMEM64G)
+typedef u64 pte_addr_t;
+#endif
+
+#if !defined(CONFIG_HIGHPTE)
+typedef pte_t *pte_addr_t;
+#endif
+
 /*
  * The i386 doesn't have any external MMU info: the kernel page
  * tables contain all the necessary information.

include/asm-i386/rmap.h

@@ -4,4 +4,18 @@
 /* nothing to see, move along */
 #include <asm-generic/rmap.h>
 
+#ifdef CONFIG_HIGHPTE
+static inline pte_t *rmap_ptep_map(pte_addr_t pte_paddr)
+{
+	unsigned long pfn = (unsigned long)(pte_paddr >> PAGE_SHIFT);
+	unsigned long off = ((unsigned long)pte_paddr) & ~PAGE_MASK;
+	return (pte_t *)((char *)kmap_atomic(pfn_to_page(pfn), KM_PTE2) + off);
+}
+
+static inline void rmap_ptep_unmap(pte_t *pte)
+{
+	kunmap_atomic(pte, KM_PTE2);
+}
+#endif
+
 #endif

include/asm-ia64/pgtable.h

@@ -420,6 +420,8 @@ extern unsigned long empty_zero_page[PAGE_SIZE/sizeof(unsigned long)];
 /* We provide our own get_unmapped_area to cope with VA holes for userland */
 #define HAVE_ARCH_UNMAPPED_AREA
 
+typedef pte_t *pte_addr_t;
+
 # endif /* !__ASSEMBLY__ */
 
 /*

include/asm-m68k/pgtable.h

@@ -171,6 +171,9 @@ extern inline void update_mmu_cache(struct vm_area_struct * vma,
 
 #ifndef __ASSEMBLY__
 #include <asm-generic/pgtable.h>
+
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 /*

include/asm-mips/pgtable.h

@@ -771,6 +771,8 @@ extern inline void set_context(unsigned long val)
 
 #include <asm-generic/pgtable.h>
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !defined (_LANGUAGE_ASSEMBLY) */
 
 #define io_remap_page_range remap_page_range

include/asm-mips64/pgtable.h

@@ -811,6 +811,8 @@ extern inline void set_context(unsigned long val)
 
 #include <asm-generic/pgtable.h>
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !defined (_LANGUAGE_ASSEMBLY) */
 
 /*

include/asm-parisc/pgtable.h

@@ -326,6 +326,8 @@ extern inline void update_mmu_cache(struct vm_area_struct * vma,
 
 #include <asm-generic/pgtable.h>
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 #define io_remap_page_range remap_page_range

include/asm-ppc/pgtable.h

@@ -543,6 +543,8 @@ extern void kernel_set_cachemode (unsigned long address, unsigned long size,
  */
 #define pgtable_cache_init()	do { } while (0)
 
-#endif /* __ASSEMBLY__ */
+typedef pte_t *pte_addr_t;
+
+#endif /* !__ASSEMBLY__ */
 #endif /* _PPC_PGTABLE_H */
 #endif /* __KERNEL__ */

include/asm-ppc64/pgtable.h

@@ -377,5 +377,7 @@ extern void update_mmu_cache(struct vm_area_struct *, unsigned long, pte_t);
 extern void hpte_init_pSeries(void);
 extern void hpte_init_iSeries(void);
 
+typedef pte_t *pte_addr_t;
+
 #endif /* __ASSEMBLY__ */
 #endif /* _PPC64_PGTABLE_H */

include/asm-s390/pgtable.h

@@ -505,6 +505,8 @@ extern inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 #define kern_addr_valid(addr)   (1)

include/asm-s390x/pgtable.h

@@ -531,6 +531,8 @@ extern inline pte_t mk_swap_pte(unsigned long type, unsigned long offset)
 #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })
 #define __swp_entry_to_pte(x)	((pte_t) { (x).val })
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 #define kern_addr_valid(addr)   (1)

include/asm-sh/pgtable.h

@@ -310,6 +310,8 @@ extern void update_mmu_cache(struct vm_area_struct * vma,
 
 #define pte_same(A,B)	(pte_val(A) == pte_val(B))
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 #define kern_addr_valid(addr)	(1)

include/asm-sparc/pgtable.h

@@ -442,6 +442,8 @@ extern int io_remap_page_range(struct vm_area_struct *vma, unsigned long from, u
 
 #include <asm-generic/pgtable.h>
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !(__ASSEMBLY__) */
 
 /* We provide our own get_unmapped_area to cope with VA holes for userland */

include/asm-sparc64/pgtable.h

@@ -369,6 +369,8 @@ extern unsigned long get_fb_unmapped_area(struct file *filp, unsigned long, unsi
 
 extern void check_pgt_cache(void);
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !(__ASSEMBLY__) */
 
 #endif /* !(_SPARC64_PGTABLE_H) */

include/asm-x86_64/pgtable.h

@@ -336,6 +336,8 @@ extern inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 #define __pte_to_swp_entry(pte)		((swp_entry_t) { pte_val(pte) })
 #define __swp_entry_to_pte(x)		((pte_t) { (x).val })
 
+typedef pte_t *pte_addr_t;
+
 #endif /* !__ASSEMBLY__ */
 
 #define kern_addr_valid(addr)	(1)

include/linux/highmem.h

@@ -26,6 +26,7 @@ static inline void *kmap(struct page *page) { return page_address(page); }
 
 #define kmap_atomic(page, idx)		page_address(page)
 #define kunmap_atomic(addr, idx)	do { } while (0)
+#define kmap_atomic_to_page(ptr)	virt_to_page(ptr)
 
 #endif /* CONFIG_HIGHMEM */
 

include/linux/mm.h

@@ -159,9 +159,9 @@ struct page {
 	struct list_head lru;		/* Pageout list, eg. active_list;
 					   protected by zone->lru_lock !! */
 	union {
-		struct pte_chain * chain;	/* Reverse pte mapping pointer.
+		struct pte_chain *chain;/* Reverse pte mapping pointer.
 					 * protected by PG_chainlock */
-		pte_t		 * direct;
+		pte_addr_t direct;
 	} pte;
 	unsigned long private;		/* mapping-private opaque data */
 
@@ -321,6 +321,16 @@ static inline void set_page_zone(struct page *page, unsigned long zone_num)
 
 #endif /* CONFIG_HIGHMEM || WANT_PAGE_VIRTUAL */
 
+/*
+ * Return true if this page is mapped into pagetables.  Subtle: test pte.direct
+ * rather than pte.chain.  Because sometimes pte.direct is 64-bit, and .chain
+ * is only 32-bit.
+ */
+static inline int page_mapped(struct page *page)
+{
+	return page->pte.direct != 0;
+}
+
 /*
  * Error return values for the *_nopage functions
  */

mm/page_alloc.c

@@ -93,7 +93,7 @@ void __free_pages_ok (struct page *page, unsigned int order)
 	BUG_ON(PageLocked(page));
 	BUG_ON(PageActive(page));
 	BUG_ON(PageWriteback(page));
-	BUG_ON(page->pte.chain != NULL);
+	BUG_ON(page->pte.direct != 0);
 	if (PageDirty(page))
 		ClearPageDirty(page);
 	BUG_ON(page_count(page) != 0);
@@ -184,7 +184,7 @@ static inline void prep_new_page(struct page *page)
 	BUG_ON(PageActive(page));
 	BUG_ON(PageDirty(page));
 	BUG_ON(PageWriteback(page));
-	BUG_ON(page->pte.chain != NULL);
+	BUG_ON(page->pte.direct != 0);
 	page->flags &= ~(1 << PG_uptodate | 1 << PG_error |
 			1 << PG_referenced | 1 << PG_arch_1 |
 			1 << PG_checked);

mm/rmap.c

@@ -43,11 +43,11 @@
  * We use an array of pte pointers in this structure to minimise cache misses
  * while traversing reverse maps.
  */
-#define NRPTE (L1_CACHE_BYTES/sizeof(void *) - 1)
+#define NRPTE ((L1_CACHE_BYTES - sizeof(void *))/sizeof(pte_addr_t))
 
 struct pte_chain {
 	struct pte_chain *next;
-	pte_t *ptes[NRPTE];
+	pte_addr_t ptes[NRPTE];
 };
 
 static kmem_cache_t	*pte_chain_cache;
@@ -126,8 +126,10 @@ int page_referenced(struct page * page)
 		referenced++;
 
 	if (PageDirect(page)) {
-		if (ptep_test_and_clear_young(page->pte.direct))
+		pte_t *pte = rmap_ptep_map(page->pte.direct);
+		if (ptep_test_and_clear_young(pte))
 			referenced++;
+		rmap_ptep_unmap(pte);
 	} else {
 		int nr_chains = 0;
 
@@ -136,11 +138,15 @@ int page_referenced(struct page * page)
 			int i;
 
 			for (i = NRPTE-1; i >= 0; i--) {
-				pte_t *p = pc->ptes[i];
-				if (!p)
+				pte_addr_t pte_paddr = pc->ptes[i];
+				pte_t *p;
+
+				if (!pte_paddr)
 					break;
+				p = rmap_ptep_map(pte_paddr);
 				if (ptep_test_and_clear_young(p))
 					referenced++;
+				rmap_ptep_unmap(p);
 				nr_chains++;
 			}
 		}
@@ -150,7 +156,6 @@ int page_referenced(struct page * page)
 			SetPageDirect(page);
 			pc->ptes[NRPTE-1] = 0;
 			pte_chain_free(pc);
-			dec_page_state(nr_reverse_maps);
 		}
 	}
 	return referenced;
@@ -166,8 +171,8 @@ int page_referenced(struct page * page)
  */
 void page_add_rmap(struct page * page, pte_t * ptep)
 {
-	struct pte_chain * pte_chain;
-	unsigned long pfn = pte_pfn(*ptep);
+	pte_addr_t pte_paddr = ptep_to_paddr(ptep);
+	struct pte_chain *pte_chain;
 	int i;
 
 #ifdef DEBUG_RMAP
@@ -179,23 +184,26 @@ void page_add_rmap(struct page * page, pte_t * ptep)
 		BUG();
 #endif
 
-	if (!pfn_valid(pfn) || PageReserved(page))
+	if (!pfn_valid(page_to_pfn(page)) || PageReserved(page))
 		return;
 
 	pte_chain_lock(page);
 
 #ifdef DEBUG_RMAP
+	/*
+	 * This stuff needs help to get up to highmem speed.
+	 */
 	{
 		struct pte_chain * pc;
 		if (PageDirect(page)) {
-			if (page->pte.direct == ptep)
+			if (page->pte.direct == pte_paddr)
 				BUG();
 		} else {
 			for (pc = page->pte.chain; pc; pc = pc->next) {
 				for (i = 0; i < NRPTE; i++) {
-					pte_t *p = pc->ptes[i];
+					pte_addr_t p = pc->ptes[i];
 
-					if (p && p == ptep)
+					if (p && p == pte_paddr)
 						BUG();
 				}
 			}
@@ -203,8 +211,8 @@ void page_add_rmap(struct page * page, pte_t * ptep)
 	}
 #endif
 
-	if (page->pte.chain == NULL) {
-		page->pte.direct = ptep;
+	if (page->pte.direct == 0) {
+		page->pte.direct = pte_paddr;
 		SetPageDirect(page);
 		goto out;
 	}
@@ -214,8 +222,8 @@ void page_add_rmap(struct page * page, pte_t * ptep)
 		ClearPageDirect(page);
 		pte_chain = pte_chain_alloc();
 		pte_chain->ptes[NRPTE-1] = page->pte.direct;
-		pte_chain->ptes[NRPTE-2] = ptep;
-		mod_page_state(nr_reverse_maps, 2);
+		pte_chain->ptes[NRPTE-2] = pte_paddr;
+		page->pte.direct = 0;
 		page->pte.chain = pte_chain;
 		goto out;
 	}
@@ -227,23 +235,22 @@ void page_add_rmap(struct page * page, pte_t * ptep)
 		new = pte_chain_alloc();
 		new->next = pte_chain;
 		page->pte.chain = new;
-		new->ptes[NRPTE-1] = ptep;
-		inc_page_state(nr_reverse_maps);
+		new->ptes[NRPTE-1] = pte_paddr;
 		goto out;
 	}
 
-	BUG_ON(pte_chain->ptes[NRPTE-1] == NULL);
+	BUG_ON(!pte_chain->ptes[NRPTE-1]);
 
 	for (i = NRPTE-2; i >= 0; i--) {
-		if (pte_chain->ptes[i] == NULL) {
-			pte_chain->ptes[i] = ptep;
-			inc_page_state(nr_reverse_maps);
+		if (!pte_chain->ptes[i]) {
+			pte_chain->ptes[i] = pte_paddr;
 			goto out;
 		}
 	}
 	BUG();
 out:
 	pte_chain_unlock(page);
+	inc_page_state(nr_reverse_maps);
 	return;
 }
 
@@ -259,19 +266,22 @@ void page_add_rmap(struct page * page, pte_t * ptep)
  */
 void page_remove_rmap(struct page * page, pte_t * ptep)
 {
+	pte_addr_t pte_paddr = ptep_to_paddr(ptep);
 	struct pte_chain *pc;
-	unsigned long pfn = page_to_pfn(page);
 
 	if (!page || !ptep)
 		BUG();
-	if (!pfn_valid(pfn) || PageReserved(page))
+	if (!pfn_valid(page_to_pfn(page)) || PageReserved(page))
 		return;
 
 	pte_chain_lock(page);
 
+	BUG_ON(page->pte.direct == 0);
+ 
 	if (PageDirect(page)) {
-		if (page->pte.direct == ptep) {
-			page->pte.direct = NULL;
+		if (page->pte.direct == pte_paddr) {
+			page->pte.direct = 0;
+			dec_page_state(nr_reverse_maps);
 			ClearPageDirect(page);
 			goto out;
 		}
@@ -285,17 +295,17 @@ void page_remove_rmap(struct page * page, pte_t * ptep)
 			if (pc->next)
 				prefetch(pc->next);
 			for (i = 0; i < NRPTE; i++) {
-				pte_t *p = pc->ptes[i];
+				pte_addr_t pa = pc->ptes[i];
 
-				if (!p)
+				if (!pa)
 					continue;
 				if (victim_i == -1)
 					victim_i = i;
-				if (p != ptep)
+				if (pa != pte_paddr)
 					continue;
 				pc->ptes[i] = start->ptes[victim_i];
-				start->ptes[victim_i] = NULL;
 				dec_page_state(nr_reverse_maps);
+				start->ptes[victim_i] = 0;
 				if (victim_i == NRPTE-1) {
 					/* Emptied a pte_chain */
 					page->pte.chain = start->next;
@@ -343,9 +353,10 @@ void page_remove_rmap(struct page * page, pte_t * ptep)
  *		pte_chain_lock		page_launder()
  *		    mm->page_table_lock	try_to_unmap_one(), trylock
  */
-static int FASTCALL(try_to_unmap_one(struct page *, pte_t *));
-static int try_to_unmap_one(struct page * page, pte_t * ptep)
+static int FASTCALL(try_to_unmap_one(struct page *, pte_addr_t));
+static int try_to_unmap_one(struct page * page, pte_addr_t paddr)
 {
+	pte_t *ptep = rmap_ptep_map(paddr);
 	unsigned long address = ptep_to_address(ptep);
 	struct mm_struct * mm = ptep_to_mm(ptep);
 	struct vm_area_struct * vma;
@@ -359,8 +370,11 @@ static int try_to_unmap_one(struct page * page, pte_t * ptep)
 	 * We need the page_table_lock to protect us from page faults,
 	 * munmap, fork, etc...
 	 */
-	if (!spin_trylock(&mm->page_table_lock))
+	if (!spin_trylock(&mm->page_table_lock)) {
+		rmap_ptep_unmap(ptep);
 		return SWAP_AGAIN;
+	}
+
 
 	/* During mremap, it's possible pages are not in a VMA. */
 	vma = find_vma(mm, address);
@@ -382,8 +396,7 @@ static int try_to_unmap_one(struct page * page, pte_t * ptep)
 
 	/* Store the swap location in the pte. See handle_pte_fault() ... */
 	if (PageSwapCache(page)) {
-		swp_entry_t entry;
-		entry.val = page->index;
+		swp_entry_t entry = { .val = page->index };
 		swap_duplicate(entry);
 		set_pte(ptep, swp_entry_to_pte(entry));
 	}
@@ -397,6 +410,7 @@ static int try_to_unmap_one(struct page * page, pte_t * ptep)
 	ret = SWAP_SUCCESS;
 
 out_unlock:
+	rmap_ptep_unmap(ptep);
 	spin_unlock(&mm->page_table_lock);
 	return ret;
 }
@@ -432,7 +446,7 @@ int try_to_unmap(struct page * page)
 	if (PageDirect(page)) {
 		ret = try_to_unmap_one(page, page->pte.direct);
 		if (ret == SWAP_SUCCESS) {
-			page->pte.direct = NULL;
+			page->pte.direct = 0;
 			ClearPageDirect(page);
 		}
 		goto out;
@@ -446,14 +460,14 @@ int try_to_unmap(struct page * page)
 		if (next_pc)
 			prefetch(next_pc);
 		for (i = 0; i < NRPTE; i++) {
-			pte_t *p = pc->ptes[i];
+			pte_addr_t pte_paddr = pc->ptes[i];
 
-			if (!p)
+			if (!pte_paddr)
 				continue;
 			if (victim_i == -1) 
 				victim_i = i;
 
-			switch (try_to_unmap_one(page, p)) {
+			switch (try_to_unmap_one(page, pte_paddr)) {
 			case SWAP_SUCCESS:
 				/*
 				 * Release a slot.  If we're releasing the
@@ -462,7 +476,7 @@ int try_to_unmap(struct page * page)
 				 * refer to the same thing.  It works out.
 				 */
 				pc->ptes[i] = start->ptes[victim_i];
-				start->ptes[victim_i] = NULL;
+				start->ptes[victim_i] = 0;
 				dec_page_state(nr_reverse_maps);
 				victim_i++;
 				if (victim_i == NRPTE) {

mm/vmscan.c

@@ -75,7 +75,7 @@ static inline int page_mapping_inuse(struct page * page)
 	struct address_space *mapping = page->mapping;
 
 	/* Page is in somebody's page tables. */
-	if (page->pte.chain)
+	if (page_mapped(page))
 		return 1;
 
 	/* XXX: does this happen ? */
@@ -140,7 +140,7 @@ shrink_list(struct list_head *page_list, int nr_pages,
 		 *
 		 * XXX: implement swap clustering ?
 		 */
-		if (page->pte.chain && !mapping && !PagePrivate(page)) {
+		if (page_mapped(page) && !mapping && !PagePrivate(page)) {
 			pte_chain_unlock(page);
 			if (!add_to_swap(page))
 				goto activate_locked;
@@ -151,7 +151,7 @@ shrink_list(struct list_head *page_list, int nr_pages,
 		 * The page is mapped into the page tables of one or more
 		 * processes. Try to unmap it here.
 		 */
-		if (page->pte.chain && mapping) {
+		if (page_mapped(page) && mapping) {
 			switch (try_to_unmap(page)) {
 			case SWAP_ERROR:
 			case SWAP_FAIL:
@@ -408,9 +408,9 @@ refill_inactive_zone(struct zone *zone, const int nr_pages_in)
 	while (!list_empty(&l_hold)) {
 		page = list_entry(l_hold.prev, struct page, lru);
 		list_del(&page->lru);
-		if (page->pte.chain) {
+		if (page_mapped(page)) {
 			pte_chain_lock(page);
-			if (page->pte.chain && page_referenced(page)) {
+			if (page_mapped(page) && page_referenced(page)) {
 				pte_chain_unlock(page);
 				list_add(&page->lru, &l_active);
 				continue;