Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _LINUX_MM_H |
| 2 | #define _LINUX_MM_H |
| 3 | |
| 4 | #include <linux/sched.h> |
| 5 | #include <linux/errno.h> |
Randy.Dunlap | c59ede7 | 2006-01-11 12:17:46 -0800 | [diff] [blame] | 6 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | |
| 8 | #ifdef __KERNEL__ |
| 9 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | #include <linux/gfp.h> |
| 11 | #include <linux/list.h> |
| 12 | #include <linux/mmzone.h> |
| 13 | #include <linux/rbtree.h> |
| 14 | #include <linux/prio_tree.h> |
| 15 | #include <linux/fs.h> |
Ingo Molnar | de5097c | 2006-01-09 15:59:21 -0800 | [diff] [blame] | 16 | #include <linux/mutex.h> |
Ingo Molnar | 9a11b49a | 2006-07-03 00:24:33 -0700 | [diff] [blame] | 17 | #include <linux/debug_locks.h> |
Peter Zijlstra | d08b385 | 2006-09-25 23:30:57 -0700 | [diff] [blame] | 18 | #include <linux/backing-dev.h> |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 19 | #include <linux/mm_types.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 20 | |
| 21 | struct mempolicy; |
| 22 | struct anon_vma; |
| 23 | |
| 24 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ |
| 25 | extern unsigned long max_mapnr; |
| 26 | #endif |
| 27 | |
| 28 | extern unsigned long num_physpages; |
| 29 | extern void * high_memory; |
| 30 | extern unsigned long vmalloc_earlyreserve; |
| 31 | extern int page_cluster; |
| 32 | |
| 33 | #ifdef CONFIG_SYSCTL |
| 34 | extern int sysctl_legacy_va_layout; |
| 35 | #else |
| 36 | #define sysctl_legacy_va_layout 0 |
| 37 | #endif |
| 38 | |
| 39 | #include <asm/page.h> |
| 40 | #include <asm/pgtable.h> |
| 41 | #include <asm/processor.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 42 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 43 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
| 44 | |
| 45 | /* |
| 46 | * Linux kernel virtual memory manager primitives. |
| 47 | * The idea being to have a "virtual" mm in the same way |
| 48 | * we have a virtual fs - giving a cleaner interface to the |
| 49 | * mm details, and allowing different kinds of memory mappings |
| 50 | * (from shared memory to executable loading to arbitrary |
| 51 | * mmap() functions). |
| 52 | */ |
| 53 | |
| 54 | /* |
| 55 | * This struct defines a memory VMM memory area. There is one of these |
| 56 | * per VM-area/task. A VM area is any part of the process virtual memory |
| 57 | * space that has a special rule for the page-fault handlers (ie a shared |
| 58 | * library, the executable area etc). |
| 59 | */ |
| 60 | struct vm_area_struct { |
| 61 | struct mm_struct * vm_mm; /* The address space we belong to. */ |
| 62 | unsigned long vm_start; /* Our start address within vm_mm. */ |
| 63 | unsigned long vm_end; /* The first byte after our end address |
| 64 | within vm_mm. */ |
| 65 | |
| 66 | /* linked list of VM areas per task, sorted by address */ |
| 67 | struct vm_area_struct *vm_next; |
| 68 | |
| 69 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ |
| 70 | unsigned long vm_flags; /* Flags, listed below. */ |
| 71 | |
| 72 | struct rb_node vm_rb; |
| 73 | |
| 74 | /* |
| 75 | * For areas with an address space and backing store, |
| 76 | * linkage into the address_space->i_mmap prio tree, or |
| 77 | * linkage to the list of like vmas hanging off its node, or |
| 78 | * linkage of vma in the address_space->i_mmap_nonlinear list. |
| 79 | */ |
| 80 | union { |
| 81 | struct { |
| 82 | struct list_head list; |
| 83 | void *parent; /* aligns with prio_tree_node parent */ |
| 84 | struct vm_area_struct *head; |
| 85 | } vm_set; |
| 86 | |
| 87 | struct raw_prio_tree_node prio_tree_node; |
| 88 | } shared; |
| 89 | |
| 90 | /* |
| 91 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma |
| 92 | * list, after a COW of one of the file pages. A MAP_SHARED vma |
| 93 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack |
| 94 | * or brk vma (with NULL file) can only be in an anon_vma list. |
| 95 | */ |
| 96 | struct list_head anon_vma_node; /* Serialized by anon_vma->lock */ |
| 97 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
| 98 | |
| 99 | /* Function pointers to deal with this struct. */ |
| 100 | struct vm_operations_struct * vm_ops; |
| 101 | |
| 102 | /* Information about our backing store: */ |
| 103 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE |
| 104 | units, *not* PAGE_CACHE_SIZE */ |
| 105 | struct file * vm_file; /* File we map to (can be NULL). */ |
| 106 | void * vm_private_data; /* was vm_pte (shared mem) */ |
| 107 | unsigned long vm_truncate_count;/* truncate_count or restart_addr */ |
| 108 | |
| 109 | #ifndef CONFIG_MMU |
| 110 | atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */ |
| 111 | #endif |
| 112 | #ifdef CONFIG_NUMA |
| 113 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ |
| 114 | #endif |
| 115 | }; |
| 116 | |
Christoph Lameter | c43692e | 2006-12-06 20:32:48 -0800 | [diff] [blame] | 117 | extern struct kmem_cache *vm_area_cachep; |
| 118 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 119 | /* |
| 120 | * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is |
| 121 | * disabled, then there's a single shared list of VMAs maintained by the |
| 122 | * system, and mm's subscribe to these individually |
| 123 | */ |
| 124 | struct vm_list_struct { |
| 125 | struct vm_list_struct *next; |
| 126 | struct vm_area_struct *vma; |
| 127 | }; |
| 128 | |
| 129 | #ifndef CONFIG_MMU |
| 130 | extern struct rb_root nommu_vma_tree; |
| 131 | extern struct rw_semaphore nommu_vma_sem; |
| 132 | |
| 133 | extern unsigned int kobjsize(const void *objp); |
| 134 | #endif |
| 135 | |
| 136 | /* |
| 137 | * vm_flags.. |
| 138 | */ |
| 139 | #define VM_READ 0x00000001 /* currently active flags */ |
| 140 | #define VM_WRITE 0x00000002 |
| 141 | #define VM_EXEC 0x00000004 |
| 142 | #define VM_SHARED 0x00000008 |
| 143 | |
Paolo 'Blaisorblade' Giarrusso | 7e2cff4 | 2005-09-21 09:55:39 -0700 | [diff] [blame] | 144 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 145 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
| 146 | #define VM_MAYWRITE 0x00000020 |
| 147 | #define VM_MAYEXEC 0x00000040 |
| 148 | #define VM_MAYSHARE 0x00000080 |
| 149 | |
| 150 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ |
| 151 | #define VM_GROWSUP 0x00000200 |
Linus Torvalds | 6aab341 | 2005-11-28 14:34:23 -0800 | [diff] [blame] | 152 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 153 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
| 154 | |
| 155 | #define VM_EXECUTABLE 0x00001000 |
| 156 | #define VM_LOCKED 0x00002000 |
| 157 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ |
| 158 | |
| 159 | /* Used by sys_madvise() */ |
| 160 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ |
| 161 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ |
| 162 | |
| 163 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ |
| 164 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ |
Hugh Dickins | 0b14c17 | 2005-11-21 21:32:15 -0800 | [diff] [blame] | 165 | #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 166 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
| 167 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
| 168 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ |
| 169 | #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ |
Linus Torvalds | 4d7672b | 2005-12-16 10:21:23 -0800 | [diff] [blame] | 170 | #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ |
Roland McGrath | e5b97dd | 2007-01-26 00:56:48 -0800 | [diff] [blame] | 171 | #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 172 | |
| 173 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
| 174 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS |
| 175 | #endif |
| 176 | |
| 177 | #ifdef CONFIG_STACK_GROWSUP |
| 178 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 179 | #else |
| 180 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 181 | #endif |
| 182 | |
| 183 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) |
| 184 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK |
| 185 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) |
| 186 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) |
| 187 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) |
| 188 | |
| 189 | /* |
| 190 | * mapping from the currently active vm_flags protection bits (the |
| 191 | * low four bits) to a page protection mask.. |
| 192 | */ |
| 193 | extern pgprot_t protection_map[16]; |
| 194 | |
| 195 | |
| 196 | /* |
| 197 | * These are the virtual MM functions - opening of an area, closing and |
| 198 | * unmapping it (needed to keep files on disk up-to-date etc), pointer |
| 199 | * to the functions called when a no-page or a wp-page exception occurs. |
| 200 | */ |
| 201 | struct vm_operations_struct { |
| 202 | void (*open)(struct vm_area_struct * area); |
| 203 | void (*close)(struct vm_area_struct * area); |
| 204 | struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type); |
Jes Sorensen | f4b8180 | 2006-09-27 01:50:10 -0700 | [diff] [blame] | 205 | unsigned long (*nopfn)(struct vm_area_struct * area, unsigned long address); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 206 | int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock); |
David Howells | 9637a5e | 2006-06-23 02:03:43 -0700 | [diff] [blame] | 207 | |
| 208 | /* notification that a previously read-only page is about to become |
| 209 | * writable, if an error is returned it will cause a SIGBUS */ |
| 210 | int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 211 | #ifdef CONFIG_NUMA |
| 212 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
| 213 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
| 214 | unsigned long addr); |
Christoph Lameter | 7b2259b | 2006-06-25 05:46:48 -0700 | [diff] [blame] | 215 | int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, |
| 216 | const nodemask_t *to, unsigned long flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 217 | #endif |
| 218 | }; |
| 219 | |
| 220 | struct mmu_gather; |
| 221 | struct inode; |
| 222 | |
Andrew Morton | 349aef0 | 2006-01-08 01:04:36 -0800 | [diff] [blame] | 223 | #define page_private(page) ((page)->private) |
| 224 | #define set_page_private(page, v) ((page)->private = (v)) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 225 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 226 | /* |
| 227 | * FIXME: take this include out, include page-flags.h in |
| 228 | * files which need it (119 of them) |
| 229 | */ |
| 230 | #include <linux/page-flags.h> |
| 231 | |
Nick Piggin | 725d704 | 2006-09-25 23:30:55 -0700 | [diff] [blame] | 232 | #ifdef CONFIG_DEBUG_VM |
| 233 | #define VM_BUG_ON(cond) BUG_ON(cond) |
| 234 | #else |
| 235 | #define VM_BUG_ON(condition) do { } while(0) |
| 236 | #endif |
| 237 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 238 | /* |
| 239 | * Methods to modify the page usage count. |
| 240 | * |
| 241 | * What counts for a page usage: |
| 242 | * - cache mapping (page->mapping) |
| 243 | * - private data (page->private) |
| 244 | * - page mapped in a task's page tables, each mapping |
| 245 | * is counted separately |
| 246 | * |
| 247 | * Also, many kernel routines increase the page count before a critical |
| 248 | * routine so they can be sure the page doesn't go away from under them. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 249 | */ |
| 250 | |
| 251 | /* |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 252 | * Drop a ref, return true if the refcount fell to zero (the page has no users) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 253 | */ |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 254 | static inline int put_page_testzero(struct page *page) |
| 255 | { |
Nick Piggin | 725d704 | 2006-09-25 23:30:55 -0700 | [diff] [blame] | 256 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
Nick Piggin | 8dc04ef | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 257 | return atomic_dec_and_test(&page->_count); |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 258 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | |
| 260 | /* |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 261 | * Try to grab a ref unless the page has a refcount of zero, return false if |
| 262 | * that is the case. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 263 | */ |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 264 | static inline int get_page_unless_zero(struct page *page) |
| 265 | { |
Nick Piggin | 725d704 | 2006-09-25 23:30:55 -0700 | [diff] [blame] | 266 | VM_BUG_ON(PageCompound(page)); |
Nick Piggin | 8dc04ef | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 267 | return atomic_inc_not_zero(&page->_count); |
Nick Piggin | 7c8ee9a | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 268 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 269 | |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 270 | static inline int page_count(struct page *page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 271 | { |
Nick Piggin | 617d221 | 2006-03-22 00:08:43 -0800 | [diff] [blame] | 272 | if (unlikely(PageCompound(page))) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 273 | page = (struct page *)page_private(page); |
Nick Piggin | 8dc04ef | 2006-03-22 00:08:03 -0800 | [diff] [blame] | 274 | return atomic_read(&page->_count); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | } |
| 276 | |
| 277 | static inline void get_page(struct page *page) |
| 278 | { |
| 279 | if (unlikely(PageCompound(page))) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 280 | page = (struct page *)page_private(page); |
Nick Piggin | 725d704 | 2006-09-25 23:30:55 -0700 | [diff] [blame] | 281 | VM_BUG_ON(atomic_read(&page->_count) == 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 282 | atomic_inc(&page->_count); |
| 283 | } |
| 284 | |
Nick Piggin | 7835e98 | 2006-03-22 00:08:40 -0800 | [diff] [blame] | 285 | /* |
| 286 | * Setup the page count before being freed into the page allocator for |
| 287 | * the first time (boot or memory hotplug) |
| 288 | */ |
| 289 | static inline void init_page_count(struct page *page) |
| 290 | { |
| 291 | atomic_set(&page->_count, 1); |
| 292 | } |
| 293 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 294 | void put_page(struct page *page); |
Alexander Zarochentsev | 1d7ea73 | 2006-08-13 23:24:27 -0700 | [diff] [blame] | 295 | void put_pages_list(struct list_head *pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 296 | |
Nick Piggin | 8dfcc9b | 2006-03-22 00:08:05 -0800 | [diff] [blame] | 297 | void split_page(struct page *page, unsigned int order); |
Nick Piggin | 8dfcc9b | 2006-03-22 00:08:05 -0800 | [diff] [blame] | 298 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 299 | /* |
Andy Whitcroft | 33f2ef8 | 2006-12-06 20:33:32 -0800 | [diff] [blame] | 300 | * Compound pages have a destructor function. Provide a |
| 301 | * prototype for that function and accessor functions. |
| 302 | * These are _only_ valid on the head of a PG_compound page. |
| 303 | */ |
| 304 | typedef void compound_page_dtor(struct page *); |
| 305 | |
| 306 | static inline void set_compound_page_dtor(struct page *page, |
| 307 | compound_page_dtor *dtor) |
| 308 | { |
| 309 | page[1].lru.next = (void *)dtor; |
| 310 | } |
| 311 | |
| 312 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) |
| 313 | { |
| 314 | return (compound_page_dtor *)page[1].lru.next; |
| 315 | } |
| 316 | |
| 317 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | * Multiple processes may "see" the same page. E.g. for untouched |
| 319 | * mappings of /dev/null, all processes see the same page full of |
| 320 | * zeroes, and text pages of executables and shared libraries have |
| 321 | * only one copy in memory, at most, normally. |
| 322 | * |
| 323 | * For the non-reserved pages, page_count(page) denotes a reference count. |
Paolo 'Blaisorblade' Giarrusso | 7e871b6 | 2005-09-21 09:55:38 -0700 | [diff] [blame] | 324 | * page_count() == 0 means the page is free. page->lru is then used for |
| 325 | * freelist management in the buddy allocator. |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 326 | * page_count() > 0 means the page has been allocated. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 327 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 328 | * Pages are allocated by the slab allocator in order to provide memory |
| 329 | * to kmalloc and kmem_cache_alloc. In this case, the management of the |
| 330 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c |
| 331 | * unless a particular usage is carefully commented. (the responsibility of |
| 332 | * freeing the kmalloc memory is the caller's, of course). |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 333 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 334 | * A page may be used by anyone else who does a __get_free_page(). |
| 335 | * In this case, page_count still tracks the references, and should only |
| 336 | * be used through the normal accessor functions. The top bits of page->flags |
| 337 | * and page->virtual store page management information, but all other fields |
| 338 | * are unused and could be used privately, carefully. The management of this |
| 339 | * page is the responsibility of the one who allocated it, and those who have |
| 340 | * subsequently been given references to it. |
| 341 | * |
| 342 | * The other pages (we may call them "pagecache pages") are completely |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 343 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
| 344 | * The following discussion applies only to them. |
| 345 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 346 | * A pagecache page contains an opaque `private' member, which belongs to the |
| 347 | * page's address_space. Usually, this is the address of a circular list of |
| 348 | * the page's disk buffers. PG_private must be set to tell the VM to call |
| 349 | * into the filesystem to release these pages. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 350 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 351 | * A page may belong to an inode's memory mapping. In this case, page->mapping |
| 352 | * is the pointer to the inode, and page->index is the file offset of the page, |
| 353 | * in units of PAGE_CACHE_SIZE. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 354 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 355 | * If pagecache pages are not associated with an inode, they are said to be |
| 356 | * anonymous pages. These may become associated with the swapcache, and in that |
| 357 | * case PG_swapcache is set, and page->private is an offset into the swapcache. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 358 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 359 | * In either case (swapcache or inode backed), the pagecache itself holds one |
| 360 | * reference to the page. Setting PG_private should also increment the |
| 361 | * refcount. The each user mapping also has a reference to the page. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 362 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 363 | * The pagecache pages are stored in a per-mapping radix tree, which is |
| 364 | * rooted at mapping->page_tree, and indexed by offset. |
| 365 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space |
| 366 | * lists, we instead now tag pages as dirty/writeback in the radix tree. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 367 | * |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 368 | * All pagecache pages may be subject to I/O: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 369 | * - inode pages may need to be read from disk, |
| 370 | * - inode pages which have been modified and are MAP_SHARED may need |
Nick Piggin | da6052f | 2006-09-25 23:31:35 -0700 | [diff] [blame] | 371 | * to be written back to the inode on disk, |
| 372 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been |
| 373 | * modified may need to be swapped out to swap space and (later) to be read |
| 374 | * back into memory. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 375 | */ |
| 376 | |
| 377 | /* |
| 378 | * The zone field is never updated after free_area_init_core() |
| 379 | * sets it, so none of the operations on it need to be atomic. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 380 | */ |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 381 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 382 | |
| 383 | /* |
| 384 | * page->flags layout: |
| 385 | * |
| 386 | * There are three possibilities for how page->flags get |
| 387 | * laid out. The first is for the normal case, without |
| 388 | * sparsemem. The second is for sparsemem when there is |
| 389 | * plenty of space for node and section. The last is when |
| 390 | * we have run out of space and have to fall back to an |
| 391 | * alternate (slower) way of determining the node. |
| 392 | * |
| 393 | * No sparsemem: | NODE | ZONE | ... | FLAGS | |
| 394 | * with space for node: | SECTION | NODE | ZONE | ... | FLAGS | |
| 395 | * no space for node: | SECTION | ZONE | ... | FLAGS | |
| 396 | */ |
| 397 | #ifdef CONFIG_SPARSEMEM |
| 398 | #define SECTIONS_WIDTH SECTIONS_SHIFT |
| 399 | #else |
| 400 | #define SECTIONS_WIDTH 0 |
| 401 | #endif |
| 402 | |
| 403 | #define ZONES_WIDTH ZONES_SHIFT |
| 404 | |
| 405 | #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED |
| 406 | #define NODES_WIDTH NODES_SHIFT |
| 407 | #else |
| 408 | #define NODES_WIDTH 0 |
| 409 | #endif |
| 410 | |
| 411 | /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ |
Andi Kleen | 07808b7 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 412 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 413 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) |
| 414 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) |
| 415 | |
| 416 | /* |
| 417 | * We are going to use the flags for the page to node mapping if its in |
| 418 | * there. This includes the case where there is no node, so it is implicit. |
| 419 | */ |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 420 | #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) |
| 421 | #define NODE_NOT_IN_PAGE_FLAGS |
| 422 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 423 | |
| 424 | #ifndef PFN_SECTION_SHIFT |
| 425 | #define PFN_SECTION_SHIFT 0 |
| 426 | #endif |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 427 | |
| 428 | /* |
| 429 | * Define the bit shifts to access each section. For non-existant |
| 430 | * sections we define the shift as 0; that plus a 0 mask ensures |
| 431 | * the compiler will optimise away reference to them. |
| 432 | */ |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 433 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) |
| 434 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) |
| 435 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 436 | |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 437 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ |
| 438 | #ifdef NODE_NOT_IN_PAGEFLAGS |
| 439 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) |
Andy Whitcroft | bd8029b | 2007-02-10 01:43:14 -0800 | [diff] [blame] | 440 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ |
| 441 | SECTIONS_PGOFF : ZONES_PGOFF) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 442 | #else |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 443 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) |
Andy Whitcroft | bd8029b | 2007-02-10 01:43:14 -0800 | [diff] [blame] | 444 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ |
| 445 | NODES_PGOFF : ZONES_PGOFF) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 446 | #endif |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 447 | |
Andy Whitcroft | bd8029b | 2007-02-10 01:43:14 -0800 | [diff] [blame] | 448 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 449 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 450 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED |
| 451 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 452 | #endif |
| 453 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 454 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) |
| 455 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) |
| 456 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 457 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 458 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 459 | static inline enum zone_type page_zonenum(struct page *page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 460 | { |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 461 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 462 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 463 | |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 464 | /* |
| 465 | * The identification function is only used by the buddy allocator for |
| 466 | * determining if two pages could be buddies. We are not really |
| 467 | * identifying a zone since we could be using a the section number |
| 468 | * id if we have not node id available in page flags. |
| 469 | * We guarantee only that it will return the same value for two |
| 470 | * combinable pages in a zone. |
| 471 | */ |
Andy Whitcroft | cb2b95e | 2006-06-23 02:03:01 -0700 | [diff] [blame] | 472 | static inline int page_zone_id(struct page *page) |
| 473 | { |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 474 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 475 | } |
| 476 | |
Andy Whitcroft | 25ba77c | 2006-12-06 20:33:03 -0800 | [diff] [blame] | 477 | static inline int zone_to_nid(struct zone *zone) |
Christoph Lameter | 89fa302 | 2006-09-25 23:31:55 -0700 | [diff] [blame] | 478 | { |
Christoph Lameter | d5f541e | 2006-09-27 01:50:08 -0700 | [diff] [blame] | 479 | #ifdef CONFIG_NUMA |
| 480 | return zone->node; |
| 481 | #else |
| 482 | return 0; |
| 483 | #endif |
Christoph Lameter | 89fa302 | 2006-09-25 23:31:55 -0700 | [diff] [blame] | 484 | } |
| 485 | |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 486 | #ifdef NODE_NOT_IN_PAGE_FLAGS |
Andy Whitcroft | 25ba77c | 2006-12-06 20:33:03 -0800 | [diff] [blame] | 487 | extern int page_to_nid(struct page *page); |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 488 | #else |
Andy Whitcroft | 25ba77c | 2006-12-06 20:33:03 -0800 | [diff] [blame] | 489 | static inline int page_to_nid(struct page *page) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 490 | { |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 491 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 492 | } |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 493 | #endif |
| 494 | |
| 495 | static inline struct zone *page_zone(struct page *page) |
| 496 | { |
| 497 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; |
| 498 | } |
| 499 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 500 | static inline unsigned long page_to_section(struct page *page) |
| 501 | { |
| 502 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; |
| 503 | } |
| 504 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 505 | static inline void set_page_zone(struct page *page, enum zone_type zone) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 506 | { |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 507 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); |
| 508 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; |
| 509 | } |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 510 | |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 511 | static inline void set_page_node(struct page *page, unsigned long node) |
| 512 | { |
| 513 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); |
| 514 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; |
| 515 | } |
Christoph Lameter | 89689ae | 2006-12-06 20:31:45 -0800 | [diff] [blame] | 516 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 517 | static inline void set_page_section(struct page *page, unsigned long section) |
| 518 | { |
| 519 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); |
| 520 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; |
| 521 | } |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 522 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 523 | static inline void set_page_links(struct page *page, enum zone_type zone, |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 524 | unsigned long node, unsigned long pfn) |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 525 | { |
| 526 | set_page_zone(page, zone); |
| 527 | set_page_node(page, node); |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 528 | set_page_section(page, pfn_to_section_nr(pfn)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 529 | } |
| 530 | |
Christoph Lameter | f6ac235 | 2006-06-30 01:55:32 -0700 | [diff] [blame] | 531 | /* |
| 532 | * Some inline functions in vmstat.h depend on page_zone() |
| 533 | */ |
| 534 | #include <linux/vmstat.h> |
| 535 | |
Ingo Molnar | 652050a | 2006-01-14 13:21:30 -0800 | [diff] [blame] | 536 | static __always_inline void *lowmem_page_address(struct page *page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 537 | { |
| 538 | return __va(page_to_pfn(page) << PAGE_SHIFT); |
| 539 | } |
| 540 | |
| 541 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) |
| 542 | #define HASHED_PAGE_VIRTUAL |
| 543 | #endif |
| 544 | |
| 545 | #if defined(WANT_PAGE_VIRTUAL) |
| 546 | #define page_address(page) ((page)->virtual) |
| 547 | #define set_page_address(page, address) \ |
| 548 | do { \ |
| 549 | (page)->virtual = (address); \ |
| 550 | } while(0) |
| 551 | #define page_address_init() do { } while(0) |
| 552 | #endif |
| 553 | |
| 554 | #if defined(HASHED_PAGE_VIRTUAL) |
| 555 | void *page_address(struct page *page); |
| 556 | void set_page_address(struct page *page, void *virtual); |
| 557 | void page_address_init(void); |
| 558 | #endif |
| 559 | |
| 560 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) |
| 561 | #define page_address(page) lowmem_page_address(page) |
| 562 | #define set_page_address(page, address) do { } while(0) |
| 563 | #define page_address_init() do { } while(0) |
| 564 | #endif |
| 565 | |
| 566 | /* |
| 567 | * On an anonymous page mapped into a user virtual memory area, |
| 568 | * page->mapping points to its anon_vma, not to a struct address_space; |
| 569 | * with the PAGE_MAPPING_ANON bit set to distinguish it. |
| 570 | * |
| 571 | * Please note that, confusingly, "page_mapping" refers to the inode |
| 572 | * address_space which maps the page from disk; whereas "page_mapped" |
| 573 | * refers to user virtual address space into which the page is mapped. |
| 574 | */ |
| 575 | #define PAGE_MAPPING_ANON 1 |
| 576 | |
| 577 | extern struct address_space swapper_space; |
| 578 | static inline struct address_space *page_mapping(struct page *page) |
| 579 | { |
| 580 | struct address_space *mapping = page->mapping; |
| 581 | |
| 582 | if (unlikely(PageSwapCache(page))) |
| 583 | mapping = &swapper_space; |
| 584 | else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) |
| 585 | mapping = NULL; |
| 586 | return mapping; |
| 587 | } |
| 588 | |
| 589 | static inline int PageAnon(struct page *page) |
| 590 | { |
| 591 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; |
| 592 | } |
| 593 | |
| 594 | /* |
| 595 | * Return the pagecache index of the passed page. Regular pagecache pages |
| 596 | * use ->index whereas swapcache pages use ->private |
| 597 | */ |
| 598 | static inline pgoff_t page_index(struct page *page) |
| 599 | { |
| 600 | if (unlikely(PageSwapCache(page))) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 601 | return page_private(page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 602 | return page->index; |
| 603 | } |
| 604 | |
| 605 | /* |
| 606 | * The atomic page->_mapcount, like _count, starts from -1: |
| 607 | * so that transitions both from it and to it can be tracked, |
| 608 | * using atomic_inc_and_test and atomic_add_negative(-1). |
| 609 | */ |
| 610 | static inline void reset_page_mapcount(struct page *page) |
| 611 | { |
| 612 | atomic_set(&(page)->_mapcount, -1); |
| 613 | } |
| 614 | |
| 615 | static inline int page_mapcount(struct page *page) |
| 616 | { |
| 617 | return atomic_read(&(page)->_mapcount) + 1; |
| 618 | } |
| 619 | |
| 620 | /* |
| 621 | * Return true if this page is mapped into pagetables. |
| 622 | */ |
| 623 | static inline int page_mapped(struct page *page) |
| 624 | { |
| 625 | return atomic_read(&(page)->_mapcount) >= 0; |
| 626 | } |
| 627 | |
| 628 | /* |
| 629 | * Error return values for the *_nopage functions |
| 630 | */ |
| 631 | #define NOPAGE_SIGBUS (NULL) |
| 632 | #define NOPAGE_OOM ((struct page *) (-1)) |
Benjamin Herrenschmidt | 7f7bbbe | 2006-10-06 00:43:53 -0700 | [diff] [blame] | 633 | #define NOPAGE_REFAULT ((struct page *) (-2)) /* Return to userspace, rerun */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 634 | |
| 635 | /* |
Jes Sorensen | f4b8180 | 2006-09-27 01:50:10 -0700 | [diff] [blame] | 636 | * Error return values for the *_nopfn functions |
| 637 | */ |
| 638 | #define NOPFN_SIGBUS ((unsigned long) -1) |
| 639 | #define NOPFN_OOM ((unsigned long) -2) |
Benjamin Herrenschmidt | 22cd25e | 2007-02-12 00:51:38 -0800 | [diff] [blame] | 640 | #define NOPFN_REFAULT ((unsigned long) -3) |
Jes Sorensen | f4b8180 | 2006-09-27 01:50:10 -0700 | [diff] [blame] | 641 | |
| 642 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 643 | * Different kinds of faults, as returned by handle_mm_fault(). |
| 644 | * Used to decide whether a process gets delivered SIGBUS or |
| 645 | * just gets major/minor fault counters bumped up. |
| 646 | */ |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 647 | #define VM_FAULT_OOM 0x00 |
| 648 | #define VM_FAULT_SIGBUS 0x01 |
| 649 | #define VM_FAULT_MINOR 0x02 |
| 650 | #define VM_FAULT_MAJOR 0x03 |
| 651 | |
| 652 | /* |
| 653 | * Special case for get_user_pages. |
| 654 | * Must be in a distinct bit from the above VM_FAULT_ flags. |
| 655 | */ |
| 656 | #define VM_FAULT_WRITE 0x10 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 657 | |
| 658 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
| 659 | |
| 660 | extern void show_free_areas(void); |
| 661 | |
| 662 | #ifdef CONFIG_SHMEM |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 663 | int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new); |
| 664 | struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, |
| 665 | unsigned long addr); |
| 666 | int shmem_lock(struct file *file, int lock, struct user_struct *user); |
| 667 | #else |
Russell King | 03b00eb | 2006-01-06 00:10:52 -0800 | [diff] [blame] | 668 | static inline int shmem_lock(struct file *file, int lock, |
| 669 | struct user_struct *user) |
| 670 | { |
| 671 | return 0; |
| 672 | } |
| 673 | |
| 674 | static inline int shmem_set_policy(struct vm_area_struct *vma, |
| 675 | struct mempolicy *new) |
| 676 | { |
| 677 | return 0; |
| 678 | } |
| 679 | |
| 680 | static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, |
| 681 | unsigned long addr) |
| 682 | { |
| 683 | return NULL; |
| 684 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 685 | #endif |
| 686 | struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags); |
| 687 | |
| 688 | int shmem_zero_setup(struct vm_area_struct *); |
| 689 | |
David Howells | b0e1519 | 2006-01-06 00:11:42 -0800 | [diff] [blame] | 690 | #ifndef CONFIG_MMU |
| 691 | extern unsigned long shmem_get_unmapped_area(struct file *file, |
| 692 | unsigned long addr, |
| 693 | unsigned long len, |
| 694 | unsigned long pgoff, |
| 695 | unsigned long flags); |
| 696 | #endif |
| 697 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 698 | static inline int can_do_mlock(void) |
| 699 | { |
| 700 | if (capable(CAP_IPC_LOCK)) |
| 701 | return 1; |
| 702 | if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0) |
| 703 | return 1; |
| 704 | return 0; |
| 705 | } |
| 706 | extern int user_shm_lock(size_t, struct user_struct *); |
| 707 | extern void user_shm_unlock(size_t, struct user_struct *); |
| 708 | |
| 709 | /* |
| 710 | * Parameter block passed down to zap_pte_range in exceptional cases. |
| 711 | */ |
| 712 | struct zap_details { |
| 713 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ |
| 714 | struct address_space *check_mapping; /* Check page->mapping if set */ |
| 715 | pgoff_t first_index; /* Lowest page->index to unmap */ |
| 716 | pgoff_t last_index; /* Highest page->index to unmap */ |
| 717 | spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 718 | unsigned long truncate_count; /* Compare vm_truncate_count */ |
| 719 | }; |
| 720 | |
Linus Torvalds | 6aab341 | 2005-11-28 14:34:23 -0800 | [diff] [blame] | 721 | struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t); |
Hugh Dickins | ee39b37 | 2005-04-19 13:29:15 -0700 | [diff] [blame] | 722 | unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 723 | unsigned long size, struct zap_details *); |
Hugh Dickins | 508034a | 2005-10-29 18:16:30 -0700 | [diff] [blame] | 724 | unsigned long unmap_vmas(struct mmu_gather **tlb, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 725 | struct vm_area_struct *start_vma, unsigned long start_addr, |
| 726 | unsigned long end_addr, unsigned long *nr_accounted, |
| 727 | struct zap_details *); |
Hugh Dickins | 3bf5ee9 | 2005-04-19 13:29:16 -0700 | [diff] [blame] | 728 | void free_pgd_range(struct mmu_gather **tlb, unsigned long addr, |
| 729 | unsigned long end, unsigned long floor, unsigned long ceiling); |
| 730 | void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma, |
Hugh Dickins | e0da382 | 2005-04-19 13:29:15 -0700 | [diff] [blame] | 731 | unsigned long floor, unsigned long ceiling); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 732 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
| 733 | struct vm_area_struct *vma); |
| 734 | int zeromap_page_range(struct vm_area_struct *vma, unsigned long from, |
| 735 | unsigned long size, pgprot_t prot); |
| 736 | void unmap_mapping_range(struct address_space *mapping, |
| 737 | loff_t const holebegin, loff_t const holelen, int even_cows); |
| 738 | |
| 739 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
| 740 | loff_t const holebegin, loff_t const holelen) |
| 741 | { |
| 742 | unmap_mapping_range(mapping, holebegin, holelen, 0); |
| 743 | } |
| 744 | |
| 745 | extern int vmtruncate(struct inode * inode, loff_t offset); |
Badari Pulavarty | f6b3ec2 | 2006-01-06 00:10:38 -0800 | [diff] [blame] | 746 | extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 747 | extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); |
| 748 | extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot); |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 749 | |
David Howells | 7ee1dd3 | 2006-01-06 00:11:44 -0800 | [diff] [blame] | 750 | #ifdef CONFIG_MMU |
| 751 | extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, |
| 752 | unsigned long address, int write_access); |
| 753 | |
| 754 | static inline int handle_mm_fault(struct mm_struct *mm, |
| 755 | struct vm_area_struct *vma, unsigned long address, |
| 756 | int write_access) |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 757 | { |
David Howells | 7ee1dd3 | 2006-01-06 00:11:44 -0800 | [diff] [blame] | 758 | return __handle_mm_fault(mm, vma, address, write_access) & |
| 759 | (~VM_FAULT_WRITE); |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 760 | } |
David Howells | 7ee1dd3 | 2006-01-06 00:11:44 -0800 | [diff] [blame] | 761 | #else |
| 762 | static inline int handle_mm_fault(struct mm_struct *mm, |
| 763 | struct vm_area_struct *vma, unsigned long address, |
| 764 | int write_access) |
| 765 | { |
| 766 | /* should never happen if there's no MMU */ |
| 767 | BUG(); |
| 768 | return VM_FAULT_SIGBUS; |
| 769 | } |
| 770 | #endif |
Nick Piggin | f33ea7f | 2005-08-03 20:24:01 +1000 | [diff] [blame] | 771 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 772 | extern int make_pages_present(unsigned long addr, unsigned long end); |
| 773 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); |
| 774 | void install_arg_page(struct vm_area_struct *, struct page *, unsigned long); |
| 775 | |
| 776 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, |
| 777 | int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); |
Nick Piggin | b581003 | 2005-10-29 18:16:12 -0700 | [diff] [blame] | 778 | void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 779 | |
David Howells | cf9a2ae | 2006-08-29 19:05:54 +0100 | [diff] [blame] | 780 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); |
| 781 | extern void do_invalidatepage(struct page *page, unsigned long offset); |
| 782 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 783 | int __set_page_dirty_nobuffers(struct page *page); |
Ken Chen | 7671932 | 2007-02-10 01:43:15 -0800 | [diff] [blame] | 784 | int __set_page_dirty_no_writeback(struct page *page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 785 | int redirty_page_for_writepage(struct writeback_control *wbc, |
| 786 | struct page *page); |
| 787 | int FASTCALL(set_page_dirty(struct page *page)); |
| 788 | int set_page_dirty_lock(struct page *page); |
| 789 | int clear_page_dirty_for_io(struct page *page); |
| 790 | |
| 791 | extern unsigned long do_mremap(unsigned long addr, |
| 792 | unsigned long old_len, unsigned long new_len, |
| 793 | unsigned long flags, unsigned long new_addr); |
| 794 | |
| 795 | /* |
| 796 | * Prototype to add a shrinker callback for ageable caches. |
| 797 | * |
| 798 | * These functions are passed a count `nr_to_scan' and a gfpmask. They should |
| 799 | * scan `nr_to_scan' objects, attempting to free them. |
| 800 | * |
Domen Puncer | 845d343 | 2005-05-05 16:16:14 -0700 | [diff] [blame] | 801 | * The callback must return the number of objects which remain in the cache. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 802 | * |
Domen Puncer | 845d343 | 2005-05-05 16:16:14 -0700 | [diff] [blame] | 803 | * The callback will be passed nr_to_scan == 0 when the VM is querying the |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 804 | * cache size, so a fastpath for that case is appropriate. |
| 805 | */ |
Al Viro | 6daa0e2 | 2005-10-21 03:18:50 -0400 | [diff] [blame] | 806 | typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 807 | |
| 808 | /* |
| 809 | * Add an aging callback. The int is the number of 'seeks' it takes |
| 810 | * to recreate one of the objects that these functions age. |
| 811 | */ |
| 812 | |
| 813 | #define DEFAULT_SEEKS 2 |
| 814 | struct shrinker; |
| 815 | extern struct shrinker *set_shrinker(int, shrinker_t); |
| 816 | extern void remove_shrinker(struct shrinker *shrinker); |
| 817 | |
Peter Zijlstra | d08b385 | 2006-09-25 23:30:57 -0700 | [diff] [blame] | 818 | /* |
| 819 | * Some shared mappigns will want the pages marked read-only |
| 820 | * to track write events. If so, we'll downgrade vm_page_prot |
| 821 | * to the private version (using protection_map[] without the |
| 822 | * VM_SHARED bit). |
| 823 | */ |
| 824 | static inline int vma_wants_writenotify(struct vm_area_struct *vma) |
| 825 | { |
| 826 | unsigned int vm_flags = vma->vm_flags; |
| 827 | |
| 828 | /* If it was private or non-writable, the write bit is already clear */ |
| 829 | if ((vm_flags & (VM_WRITE|VM_SHARED)) != ((VM_WRITE|VM_SHARED))) |
| 830 | return 0; |
| 831 | |
| 832 | /* The backer wishes to know when pages are first written to? */ |
| 833 | if (vma->vm_ops && vma->vm_ops->page_mkwrite) |
| 834 | return 1; |
| 835 | |
| 836 | /* The open routine did something to the protections already? */ |
| 837 | if (pgprot_val(vma->vm_page_prot) != |
| 838 | pgprot_val(protection_map[vm_flags & |
| 839 | (VM_READ|VM_WRITE|VM_EXEC|VM_SHARED)])) |
| 840 | return 0; |
| 841 | |
| 842 | /* Specialty mapping? */ |
| 843 | if (vm_flags & (VM_PFNMAP|VM_INSERTPAGE)) |
| 844 | return 0; |
| 845 | |
| 846 | /* Can the mapping track the dirty pages? */ |
| 847 | return vma->vm_file && vma->vm_file->f_mapping && |
| 848 | mapping_cap_account_dirty(vma->vm_file->f_mapping); |
| 849 | } |
| 850 | |
Linus Torvalds | c9cfcdd | 2005-11-29 14:03:14 -0800 | [diff] [blame] | 851 | extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl)); |
| 852 | |
Nick Piggin | 5f22df0 | 2007-05-06 14:49:02 -0700 | [diff] [blame^] | 853 | #ifdef __PAGETABLE_PUD_FOLDED |
| 854 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, |
| 855 | unsigned long address) |
| 856 | { |
| 857 | return 0; |
| 858 | } |
| 859 | #else |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 860 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); |
Nick Piggin | 5f22df0 | 2007-05-06 14:49:02 -0700 | [diff] [blame^] | 861 | #endif |
| 862 | |
| 863 | #ifdef __PAGETABLE_PMD_FOLDED |
| 864 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, |
| 865 | unsigned long address) |
| 866 | { |
| 867 | return 0; |
| 868 | } |
| 869 | #else |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 870 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); |
Nick Piggin | 5f22df0 | 2007-05-06 14:49:02 -0700 | [diff] [blame^] | 871 | #endif |
| 872 | |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 873 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); |
| 874 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); |
| 875 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 876 | /* |
| 877 | * The following ifdef needed to get the 4level-fixup.h header to work. |
| 878 | * Remove it when 4level-fixup.h has been removed. |
| 879 | */ |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 880 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 881 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
| 882 | { |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 883 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? |
| 884 | NULL: pud_offset(pgd, address); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 885 | } |
| 886 | |
| 887 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
| 888 | { |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 889 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? |
| 890 | NULL: pmd_offset(pud, address); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 891 | } |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 892 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ |
| 893 | |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 894 | #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS |
| 895 | /* |
| 896 | * We tuck a spinlock to guard each pagetable page into its struct page, |
| 897 | * at page->private, with BUILD_BUG_ON to make sure that this will not |
| 898 | * overflow into the next struct page (as it might with DEBUG_SPINLOCK). |
| 899 | * When freeing, reset page->mapping so free_pages_check won't complain. |
| 900 | */ |
Andrew Morton | 349aef0 | 2006-01-08 01:04:36 -0800 | [diff] [blame] | 901 | #define __pte_lockptr(page) &((page)->ptl) |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 902 | #define pte_lock_init(_page) do { \ |
| 903 | spin_lock_init(__pte_lockptr(_page)); \ |
| 904 | } while (0) |
| 905 | #define pte_lock_deinit(page) ((page)->mapping = NULL) |
| 906 | #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) |
| 907 | #else |
| 908 | /* |
| 909 | * We use mm->page_table_lock to guard all pagetable pages of the mm. |
| 910 | */ |
| 911 | #define pte_lock_init(page) do {} while (0) |
| 912 | #define pte_lock_deinit(page) do {} while (0) |
| 913 | #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) |
| 914 | #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */ |
| 915 | |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 916 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ |
| 917 | ({ \ |
Hugh Dickins | 4c21e2f | 2005-10-29 18:16:40 -0700 | [diff] [blame] | 918 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 919 | pte_t *__pte = pte_offset_map(pmd, address); \ |
| 920 | *(ptlp) = __ptl; \ |
| 921 | spin_lock(__ptl); \ |
| 922 | __pte; \ |
| 923 | }) |
| 924 | |
| 925 | #define pte_unmap_unlock(pte, ptl) do { \ |
| 926 | spin_unlock(ptl); \ |
| 927 | pte_unmap(pte); \ |
| 928 | } while (0) |
| 929 | |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 930 | #define pte_alloc_map(mm, pmd, address) \ |
| 931 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ |
| 932 | NULL: pte_offset_map(pmd, address)) |
| 933 | |
Hugh Dickins | c74df32 | 2005-10-29 18:16:23 -0700 | [diff] [blame] | 934 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ |
| 935 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ |
| 936 | NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) |
| 937 | |
Hugh Dickins | 1bb3630 | 2005-10-29 18:16:22 -0700 | [diff] [blame] | 938 | #define pte_alloc_kernel(pmd, address) \ |
| 939 | ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ |
| 940 | NULL: pte_offset_kernel(pmd, address)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 941 | |
| 942 | extern void free_area_init(unsigned long * zones_size); |
| 943 | extern void free_area_init_node(int nid, pg_data_t *pgdat, |
| 944 | unsigned long * zones_size, unsigned long zone_start_pfn, |
| 945 | unsigned long *zholes_size); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 946 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
| 947 | /* |
| 948 | * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its |
| 949 | * zones, allocate the backing mem_map and account for memory holes in a more |
| 950 | * architecture independent manner. This is a substitute for creating the |
| 951 | * zone_sizes[] and zholes_size[] arrays and passing them to |
| 952 | * free_area_init_node() |
| 953 | * |
| 954 | * An architecture is expected to register range of page frames backed by |
| 955 | * physical memory with add_active_range() before calling |
| 956 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic |
| 957 | * usage, an architecture is expected to do something like |
| 958 | * |
| 959 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, |
| 960 | * max_highmem_pfn}; |
| 961 | * for_each_valid_physical_page_range() |
| 962 | * add_active_range(node_id, start_pfn, end_pfn) |
| 963 | * free_area_init_nodes(max_zone_pfns); |
| 964 | * |
| 965 | * If the architecture guarantees that there are no holes in the ranges |
| 966 | * registered with add_active_range(), free_bootmem_active_regions() |
| 967 | * will call free_bootmem_node() for each registered physical page range. |
| 968 | * Similarly sparse_memory_present_with_active_regions() calls |
| 969 | * memory_present() for each range when SPARSEMEM is enabled. |
| 970 | * |
| 971 | * See mm/page_alloc.c for more information on each function exposed by |
| 972 | * CONFIG_ARCH_POPULATES_NODE_MAP |
| 973 | */ |
| 974 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); |
| 975 | extern void add_active_range(unsigned int nid, unsigned long start_pfn, |
| 976 | unsigned long end_pfn); |
| 977 | extern void shrink_active_range(unsigned int nid, unsigned long old_end_pfn, |
| 978 | unsigned long new_end_pfn); |
Mel Gorman | fb01439 | 2006-09-27 01:49:59 -0700 | [diff] [blame] | 979 | extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn, |
| 980 | unsigned long end_pfn); |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 981 | extern void remove_all_active_ranges(void); |
| 982 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, |
| 983 | unsigned long end_pfn); |
| 984 | extern void get_pfn_range_for_nid(unsigned int nid, |
| 985 | unsigned long *start_pfn, unsigned long *end_pfn); |
| 986 | extern unsigned long find_min_pfn_with_active_regions(void); |
| 987 | extern unsigned long find_max_pfn_with_active_regions(void); |
| 988 | extern void free_bootmem_with_active_regions(int nid, |
| 989 | unsigned long max_low_pfn); |
| 990 | extern void sparse_memory_present_with_active_regions(int nid); |
| 991 | #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID |
| 992 | extern int early_pfn_to_nid(unsigned long pfn); |
| 993 | #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ |
| 994 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ |
Mel Gorman | 0e0b864 | 2006-09-27 01:49:56 -0700 | [diff] [blame] | 995 | extern void set_dma_reserve(unsigned long new_dma_reserve); |
Dave Hansen | a2f3aa0 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 996 | extern void memmap_init_zone(unsigned long, int, unsigned long, |
| 997 | unsigned long, enum memmap_context); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 998 | extern void setup_per_zone_pages_min(void); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 999 | extern void mem_init(void); |
| 1000 | extern void show_mem(void); |
| 1001 | extern void si_meminfo(struct sysinfo * val); |
| 1002 | extern void si_meminfo_node(struct sysinfo *val, int nid); |
| 1003 | |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 1004 | #ifdef CONFIG_NUMA |
| 1005 | extern void setup_per_cpu_pageset(void); |
| 1006 | #else |
| 1007 | static inline void setup_per_cpu_pageset(void) {} |
| 1008 | #endif |
| 1009 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1010 | /* prio_tree.c */ |
| 1011 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); |
| 1012 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); |
| 1013 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); |
| 1014 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, |
| 1015 | struct prio_tree_iter *iter); |
| 1016 | |
| 1017 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ |
| 1018 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ |
| 1019 | (vma = vma_prio_tree_next(vma, iter)); ) |
| 1020 | |
| 1021 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, |
| 1022 | struct list_head *list) |
| 1023 | { |
| 1024 | vma->shared.vm_set.parent = NULL; |
| 1025 | list_add_tail(&vma->shared.vm_set.list, list); |
| 1026 | } |
| 1027 | |
| 1028 | /* mmap.c */ |
| 1029 | extern int __vm_enough_memory(long pages, int cap_sys_admin); |
| 1030 | extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, |
| 1031 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); |
| 1032 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
| 1033 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, |
| 1034 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, |
| 1035 | struct mempolicy *); |
| 1036 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
| 1037 | extern int split_vma(struct mm_struct *, |
| 1038 | struct vm_area_struct *, unsigned long addr, int new_below); |
| 1039 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
| 1040 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, |
| 1041 | struct rb_node **, struct rb_node *); |
Hugh Dickins | a8fb561 | 2005-10-29 18:15:57 -0700 | [diff] [blame] | 1042 | extern void unlink_file_vma(struct vm_area_struct *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1043 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
| 1044 | unsigned long addr, unsigned long len, pgoff_t pgoff); |
| 1045 | extern void exit_mmap(struct mm_struct *); |
akpm@osdl.org | 119f657 | 2005-05-01 08:58:35 -0700 | [diff] [blame] | 1046 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
Roland McGrath | fa5dc22 | 2007-02-08 14:20:41 -0800 | [diff] [blame] | 1047 | extern int install_special_mapping(struct mm_struct *mm, |
| 1048 | unsigned long addr, unsigned long len, |
| 1049 | unsigned long flags, struct page **pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1050 | |
| 1051 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
| 1052 | |
| 1053 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, |
| 1054 | unsigned long len, unsigned long prot, |
| 1055 | unsigned long flag, unsigned long pgoff); |
| 1056 | |
| 1057 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, |
| 1058 | unsigned long len, unsigned long prot, |
| 1059 | unsigned long flag, unsigned long offset) |
| 1060 | { |
| 1061 | unsigned long ret = -EINVAL; |
| 1062 | if ((offset + PAGE_ALIGN(len)) < offset) |
| 1063 | goto out; |
| 1064 | if (!(offset & ~PAGE_MASK)) |
| 1065 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); |
| 1066 | out: |
| 1067 | return ret; |
| 1068 | } |
| 1069 | |
| 1070 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); |
| 1071 | |
| 1072 | extern unsigned long do_brk(unsigned long, unsigned long); |
| 1073 | |
| 1074 | /* filemap.c */ |
| 1075 | extern unsigned long page_unuse(struct page *); |
| 1076 | extern void truncate_inode_pages(struct address_space *, loff_t); |
Hans Reiser | d733907 | 2006-01-06 00:10:36 -0800 | [diff] [blame] | 1077 | extern void truncate_inode_pages_range(struct address_space *, |
| 1078 | loff_t lstart, loff_t lend); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1079 | |
| 1080 | /* generic vm_area_ops exported for stackable file systems */ |
| 1081 | extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *); |
| 1082 | extern int filemap_populate(struct vm_area_struct *, unsigned long, |
| 1083 | unsigned long, pgprot_t, unsigned long, int); |
| 1084 | |
| 1085 | /* mm/page-writeback.c */ |
| 1086 | int write_one_page(struct page *page, int wait); |
| 1087 | |
| 1088 | /* readahead.c */ |
| 1089 | #define VM_MAX_READAHEAD 128 /* kbytes */ |
| 1090 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ |
| 1091 | #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before |
| 1092 | * turning readahead off */ |
| 1093 | |
| 1094 | int do_page_cache_readahead(struct address_space *mapping, struct file *filp, |
Andrew Morton | 7361f4d | 2005-11-07 00:59:28 -0800 | [diff] [blame] | 1095 | pgoff_t offset, unsigned long nr_to_read); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1096 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
Andrew Morton | 7361f4d | 2005-11-07 00:59:28 -0800 | [diff] [blame] | 1097 | pgoff_t offset, unsigned long nr_to_read); |
| 1098 | unsigned long page_cache_readahead(struct address_space *mapping, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1099 | struct file_ra_state *ra, |
| 1100 | struct file *filp, |
Andrew Morton | 7361f4d | 2005-11-07 00:59:28 -0800 | [diff] [blame] | 1101 | pgoff_t offset, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1102 | unsigned long size); |
| 1103 | void handle_ra_miss(struct address_space *mapping, |
| 1104 | struct file_ra_state *ra, pgoff_t offset); |
| 1105 | unsigned long max_sane_readahead(unsigned long nr); |
| 1106 | |
| 1107 | /* Do stack extension */ |
Hugh Dickins | 46dea3d | 2005-10-29 18:16:20 -0700 | [diff] [blame] | 1108 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); |
Matthew Wilcox | 9ab8851 | 2005-11-18 16:16:42 -0500 | [diff] [blame] | 1109 | #ifdef CONFIG_IA64 |
Hugh Dickins | 46dea3d | 2005-10-29 18:16:20 -0700 | [diff] [blame] | 1110 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); |
Matthew Wilcox | 9ab8851 | 2005-11-18 16:16:42 -0500 | [diff] [blame] | 1111 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1112 | |
| 1113 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ |
| 1114 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); |
| 1115 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, |
| 1116 | struct vm_area_struct **pprev); |
| 1117 | |
| 1118 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, |
| 1119 | NULL if none. Assume start_addr < end_addr. */ |
| 1120 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) |
| 1121 | { |
| 1122 | struct vm_area_struct * vma = find_vma(mm,start_addr); |
| 1123 | |
| 1124 | if (vma && end_addr <= vma->vm_start) |
| 1125 | vma = NULL; |
| 1126 | return vma; |
| 1127 | } |
| 1128 | |
| 1129 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
| 1130 | { |
| 1131 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
| 1132 | } |
| 1133 | |
Hugh Dickins | 804af2c | 2006-07-26 21:39:49 +0100 | [diff] [blame] | 1134 | pgprot_t vm_get_page_prot(unsigned long vm_flags); |
Hugh Dickins | deceb6c | 2005-10-29 18:16:33 -0700 | [diff] [blame] | 1135 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); |
| 1136 | struct page *vmalloc_to_page(void *addr); |
| 1137 | unsigned long vmalloc_to_pfn(void *addr); |
| 1138 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, |
| 1139 | unsigned long pfn, unsigned long size, pgprot_t); |
Linus Torvalds | a145dd4 | 2005-11-30 09:35:19 -0800 | [diff] [blame] | 1140 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); |
Nick Piggin | e0dc0d8 | 2007-02-12 00:51:36 -0800 | [diff] [blame] | 1141 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
| 1142 | unsigned long pfn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1143 | |
Linus Torvalds | 6aab341 | 2005-11-28 14:34:23 -0800 | [diff] [blame] | 1144 | struct page *follow_page(struct vm_area_struct *, unsigned long address, |
Hugh Dickins | deceb6c | 2005-10-29 18:16:33 -0700 | [diff] [blame] | 1145 | unsigned int foll_flags); |
| 1146 | #define FOLL_WRITE 0x01 /* check pte is writable */ |
| 1147 | #define FOLL_TOUCH 0x02 /* mark page accessed */ |
| 1148 | #define FOLL_GET 0x04 /* do get_page on page */ |
| 1149 | #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1150 | |
Jeremy Fitzhardinge | aee16b3 | 2007-05-06 14:48:54 -0700 | [diff] [blame] | 1151 | typedef int (*pte_fn_t)(pte_t *pte, struct page *pmd_page, unsigned long addr, |
| 1152 | void *data); |
| 1153 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, |
| 1154 | unsigned long size, pte_fn_t fn, void *data); |
| 1155 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1156 | #ifdef CONFIG_PROC_FS |
Hugh Dickins | ab50b8e | 2005-10-29 18:15:56 -0700 | [diff] [blame] | 1157 | void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1158 | #else |
Hugh Dickins | ab50b8e | 2005-10-29 18:15:56 -0700 | [diff] [blame] | 1159 | static inline void vm_stat_account(struct mm_struct *mm, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1160 | unsigned long flags, struct file *file, long pages) |
| 1161 | { |
| 1162 | } |
| 1163 | #endif /* CONFIG_PROC_FS */ |
| 1164 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1165 | #ifndef CONFIG_DEBUG_PAGEALLOC |
| 1166 | static inline void |
Nick Piggin | 9858db5 | 2006-10-11 01:21:30 -0700 | [diff] [blame] | 1167 | kernel_map_pages(struct page *page, int numpages, int enable) {} |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1168 | #endif |
| 1169 | |
| 1170 | extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); |
| 1171 | #ifdef __HAVE_ARCH_GATE_AREA |
| 1172 | int in_gate_area_no_task(unsigned long addr); |
| 1173 | int in_gate_area(struct task_struct *task, unsigned long addr); |
| 1174 | #else |
| 1175 | int in_gate_area_no_task(unsigned long addr); |
| 1176 | #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) |
| 1177 | #endif /* __HAVE_ARCH_GATE_AREA */ |
| 1178 | |
Andrew Morton | 9d0243b | 2006-01-08 01:00:39 -0800 | [diff] [blame] | 1179 | int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *, |
| 1180 | void __user *, size_t *, loff_t *); |
Andrew Morton | 69e0594 | 2006-03-22 00:08:19 -0800 | [diff] [blame] | 1181 | unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, |
Andrew Morton | 9d0243b | 2006-01-08 01:00:39 -0800 | [diff] [blame] | 1182 | unsigned long lru_pages); |
| 1183 | void drop_pagecache(void); |
| 1184 | void drop_slab(void); |
| 1185 | |
Luke Yang | 7a9166e | 2006-02-20 18:28:07 -0800 | [diff] [blame] | 1186 | #ifndef CONFIG_MMU |
| 1187 | #define randomize_va_space 0 |
| 1188 | #else |
Andi Kleen | a62eaf1 | 2006-02-16 23:41:58 +0100 | [diff] [blame] | 1189 | extern int randomize_va_space; |
Luke Yang | 7a9166e | 2006-02-20 18:28:07 -0800 | [diff] [blame] | 1190 | #endif |
Andi Kleen | a62eaf1 | 2006-02-16 23:41:58 +0100 | [diff] [blame] | 1191 | |
David Howells | f269fdd | 2006-09-27 01:50:23 -0700 | [diff] [blame] | 1192 | __attribute__((weak)) const char *arch_vma_name(struct vm_area_struct *vma); |
Ingo Molnar | e6e5494 | 2006-06-27 02:53:50 -0700 | [diff] [blame] | 1193 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1194 | #endif /* __KERNEL__ */ |
| 1195 | #endif /* _LINUX_MM_H */ |