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> |
| 6 | |
| 7 | #ifdef __KERNEL__ |
| 8 | |
| 9 | #include <linux/config.h> |
| 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> |
| 16 | |
| 17 | struct mempolicy; |
| 18 | struct anon_vma; |
| 19 | |
| 20 | #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ |
| 21 | extern unsigned long max_mapnr; |
| 22 | #endif |
| 23 | |
| 24 | extern unsigned long num_physpages; |
| 25 | extern void * high_memory; |
| 26 | extern unsigned long vmalloc_earlyreserve; |
| 27 | extern int page_cluster; |
| 28 | |
| 29 | #ifdef CONFIG_SYSCTL |
| 30 | extern int sysctl_legacy_va_layout; |
| 31 | #else |
| 32 | #define sysctl_legacy_va_layout 0 |
| 33 | #endif |
| 34 | |
| 35 | #include <asm/page.h> |
| 36 | #include <asm/pgtable.h> |
| 37 | #include <asm/processor.h> |
| 38 | #include <asm/atomic.h> |
| 39 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) |
| 41 | |
| 42 | /* |
| 43 | * Linux kernel virtual memory manager primitives. |
| 44 | * The idea being to have a "virtual" mm in the same way |
| 45 | * we have a virtual fs - giving a cleaner interface to the |
| 46 | * mm details, and allowing different kinds of memory mappings |
| 47 | * (from shared memory to executable loading to arbitrary |
| 48 | * mmap() functions). |
| 49 | */ |
| 50 | |
| 51 | /* |
| 52 | * This struct defines a memory VMM memory area. There is one of these |
| 53 | * per VM-area/task. A VM area is any part of the process virtual memory |
| 54 | * space that has a special rule for the page-fault handlers (ie a shared |
| 55 | * library, the executable area etc). |
| 56 | */ |
| 57 | struct vm_area_struct { |
| 58 | struct mm_struct * vm_mm; /* The address space we belong to. */ |
| 59 | unsigned long vm_start; /* Our start address within vm_mm. */ |
| 60 | unsigned long vm_end; /* The first byte after our end address |
| 61 | within vm_mm. */ |
| 62 | |
| 63 | /* linked list of VM areas per task, sorted by address */ |
| 64 | struct vm_area_struct *vm_next; |
| 65 | |
| 66 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ |
| 67 | unsigned long vm_flags; /* Flags, listed below. */ |
| 68 | |
| 69 | struct rb_node vm_rb; |
| 70 | |
| 71 | /* |
| 72 | * For areas with an address space and backing store, |
| 73 | * linkage into the address_space->i_mmap prio tree, or |
| 74 | * linkage to the list of like vmas hanging off its node, or |
| 75 | * linkage of vma in the address_space->i_mmap_nonlinear list. |
| 76 | */ |
| 77 | union { |
| 78 | struct { |
| 79 | struct list_head list; |
| 80 | void *parent; /* aligns with prio_tree_node parent */ |
| 81 | struct vm_area_struct *head; |
| 82 | } vm_set; |
| 83 | |
| 84 | struct raw_prio_tree_node prio_tree_node; |
| 85 | } shared; |
| 86 | |
| 87 | /* |
| 88 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma |
| 89 | * list, after a COW of one of the file pages. A MAP_SHARED vma |
| 90 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack |
| 91 | * or brk vma (with NULL file) can only be in an anon_vma list. |
| 92 | */ |
| 93 | struct list_head anon_vma_node; /* Serialized by anon_vma->lock */ |
| 94 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
| 95 | |
| 96 | /* Function pointers to deal with this struct. */ |
| 97 | struct vm_operations_struct * vm_ops; |
| 98 | |
| 99 | /* Information about our backing store: */ |
| 100 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE |
| 101 | units, *not* PAGE_CACHE_SIZE */ |
| 102 | struct file * vm_file; /* File we map to (can be NULL). */ |
| 103 | void * vm_private_data; /* was vm_pte (shared mem) */ |
| 104 | unsigned long vm_truncate_count;/* truncate_count or restart_addr */ |
| 105 | |
| 106 | #ifndef CONFIG_MMU |
| 107 | atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */ |
| 108 | #endif |
| 109 | #ifdef CONFIG_NUMA |
| 110 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ |
| 111 | #endif |
| 112 | }; |
| 113 | |
| 114 | /* |
| 115 | * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is |
| 116 | * disabled, then there's a single shared list of VMAs maintained by the |
| 117 | * system, and mm's subscribe to these individually |
| 118 | */ |
| 119 | struct vm_list_struct { |
| 120 | struct vm_list_struct *next; |
| 121 | struct vm_area_struct *vma; |
| 122 | }; |
| 123 | |
| 124 | #ifndef CONFIG_MMU |
| 125 | extern struct rb_root nommu_vma_tree; |
| 126 | extern struct rw_semaphore nommu_vma_sem; |
| 127 | |
| 128 | extern unsigned int kobjsize(const void *objp); |
| 129 | #endif |
| 130 | |
| 131 | /* |
| 132 | * vm_flags.. |
| 133 | */ |
| 134 | #define VM_READ 0x00000001 /* currently active flags */ |
| 135 | #define VM_WRITE 0x00000002 |
| 136 | #define VM_EXEC 0x00000004 |
| 137 | #define VM_SHARED 0x00000008 |
| 138 | |
| 139 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ |
| 140 | #define VM_MAYWRITE 0x00000020 |
| 141 | #define VM_MAYEXEC 0x00000040 |
| 142 | #define VM_MAYSHARE 0x00000080 |
| 143 | |
| 144 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ |
| 145 | #define VM_GROWSUP 0x00000200 |
| 146 | #define VM_SHM 0x00000400 /* shared memory area, don't swap out */ |
| 147 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ |
| 148 | |
| 149 | #define VM_EXECUTABLE 0x00001000 |
| 150 | #define VM_LOCKED 0x00002000 |
| 151 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ |
| 152 | |
| 153 | /* Used by sys_madvise() */ |
| 154 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ |
| 155 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ |
| 156 | |
| 157 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ |
| 158 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ |
| 159 | #define VM_RESERVED 0x00080000 /* Don't unmap it from swap_out */ |
| 160 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ |
| 161 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ |
| 162 | #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ |
| 163 | #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ |
| 164 | |
| 165 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ |
| 166 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS |
| 167 | #endif |
| 168 | |
| 169 | #ifdef CONFIG_STACK_GROWSUP |
| 170 | #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 171 | #else |
| 172 | #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) |
| 173 | #endif |
| 174 | |
| 175 | #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) |
| 176 | #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK |
| 177 | #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) |
| 178 | #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) |
| 179 | #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) |
| 180 | |
| 181 | /* |
| 182 | * mapping from the currently active vm_flags protection bits (the |
| 183 | * low four bits) to a page protection mask.. |
| 184 | */ |
| 185 | extern pgprot_t protection_map[16]; |
| 186 | |
| 187 | |
| 188 | /* |
| 189 | * These are the virtual MM functions - opening of an area, closing and |
| 190 | * unmapping it (needed to keep files on disk up-to-date etc), pointer |
| 191 | * to the functions called when a no-page or a wp-page exception occurs. |
| 192 | */ |
| 193 | struct vm_operations_struct { |
| 194 | void (*open)(struct vm_area_struct * area); |
| 195 | void (*close)(struct vm_area_struct * area); |
| 196 | struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type); |
| 197 | int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock); |
| 198 | #ifdef CONFIG_NUMA |
| 199 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); |
| 200 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, |
| 201 | unsigned long addr); |
| 202 | #endif |
| 203 | }; |
| 204 | |
| 205 | struct mmu_gather; |
| 206 | struct inode; |
| 207 | |
| 208 | #ifdef ARCH_HAS_ATOMIC_UNSIGNED |
| 209 | typedef unsigned page_flags_t; |
| 210 | #else |
| 211 | typedef unsigned long page_flags_t; |
| 212 | #endif |
| 213 | |
| 214 | /* |
| 215 | * Each physical page in the system has a struct page associated with |
| 216 | * it to keep track of whatever it is we are using the page for at the |
| 217 | * moment. Note that we have no way to track which tasks are using |
| 218 | * a page. |
| 219 | */ |
| 220 | struct page { |
| 221 | page_flags_t flags; /* Atomic flags, some possibly |
| 222 | * updated asynchronously */ |
| 223 | atomic_t _count; /* Usage count, see below. */ |
| 224 | atomic_t _mapcount; /* Count of ptes mapped in mms, |
| 225 | * to show when page is mapped |
| 226 | * & limit reverse map searches. |
| 227 | */ |
| 228 | unsigned long private; /* Mapping-private opaque data: |
| 229 | * usually used for buffer_heads |
| 230 | * if PagePrivate set; used for |
| 231 | * swp_entry_t if PageSwapCache |
| 232 | * When page is free, this indicates |
| 233 | * order in the buddy system. |
| 234 | */ |
| 235 | struct address_space *mapping; /* If low bit clear, points to |
| 236 | * inode address_space, or NULL. |
| 237 | * If page mapped as anonymous |
| 238 | * memory, low bit is set, and |
| 239 | * it points to anon_vma object: |
| 240 | * see PAGE_MAPPING_ANON below. |
| 241 | */ |
| 242 | pgoff_t index; /* Our offset within mapping. */ |
| 243 | struct list_head lru; /* Pageout list, eg. active_list |
| 244 | * protected by zone->lru_lock ! |
| 245 | */ |
| 246 | /* |
| 247 | * On machines where all RAM is mapped into kernel address space, |
| 248 | * we can simply calculate the virtual address. On machines with |
| 249 | * highmem some memory is mapped into kernel virtual memory |
| 250 | * dynamically, so we need a place to store that address. |
| 251 | * Note that this field could be 16 bits on x86 ... ;) |
| 252 | * |
| 253 | * Architectures with slow multiplication can define |
| 254 | * WANT_PAGE_VIRTUAL in asm/page.h |
| 255 | */ |
| 256 | #if defined(WANT_PAGE_VIRTUAL) |
| 257 | void *virtual; /* Kernel virtual address (NULL if |
| 258 | not kmapped, ie. highmem) */ |
| 259 | #endif /* WANT_PAGE_VIRTUAL */ |
| 260 | }; |
| 261 | |
| 262 | /* |
| 263 | * FIXME: take this include out, include page-flags.h in |
| 264 | * files which need it (119 of them) |
| 265 | */ |
| 266 | #include <linux/page-flags.h> |
| 267 | |
| 268 | /* |
| 269 | * Methods to modify the page usage count. |
| 270 | * |
| 271 | * What counts for a page usage: |
| 272 | * - cache mapping (page->mapping) |
| 273 | * - private data (page->private) |
| 274 | * - page mapped in a task's page tables, each mapping |
| 275 | * is counted separately |
| 276 | * |
| 277 | * Also, many kernel routines increase the page count before a critical |
| 278 | * routine so they can be sure the page doesn't go away from under them. |
| 279 | * |
| 280 | * Since 2.6.6 (approx), a free page has ->_count = -1. This is so that we |
| 281 | * can use atomic_add_negative(-1, page->_count) to detect when the page |
| 282 | * becomes free and so that we can also use atomic_inc_and_test to atomically |
| 283 | * detect when we just tried to grab a ref on a page which some other CPU has |
| 284 | * already deemed to be freeable. |
| 285 | * |
| 286 | * NO code should make assumptions about this internal detail! Use the provided |
| 287 | * macros which retain the old rules: page_count(page) == 0 is a free page. |
| 288 | */ |
| 289 | |
| 290 | /* |
| 291 | * Drop a ref, return true if the logical refcount fell to zero (the page has |
| 292 | * no users) |
| 293 | */ |
| 294 | #define put_page_testzero(p) \ |
| 295 | ({ \ |
| 296 | BUG_ON(page_count(p) == 0); \ |
| 297 | atomic_add_negative(-1, &(p)->_count); \ |
| 298 | }) |
| 299 | |
| 300 | /* |
| 301 | * Grab a ref, return true if the page previously had a logical refcount of |
| 302 | * zero. ie: returns true if we just grabbed an already-deemed-to-be-free page |
| 303 | */ |
| 304 | #define get_page_testone(p) atomic_inc_and_test(&(p)->_count) |
| 305 | |
| 306 | #define set_page_count(p,v) atomic_set(&(p)->_count, v - 1) |
| 307 | #define __put_page(p) atomic_dec(&(p)->_count) |
| 308 | |
| 309 | extern void FASTCALL(__page_cache_release(struct page *)); |
| 310 | |
| 311 | #ifdef CONFIG_HUGETLB_PAGE |
| 312 | |
| 313 | static inline int page_count(struct page *p) |
| 314 | { |
| 315 | if (PageCompound(p)) |
| 316 | p = (struct page *)p->private; |
| 317 | return atomic_read(&(p)->_count) + 1; |
| 318 | } |
| 319 | |
| 320 | static inline void get_page(struct page *page) |
| 321 | { |
| 322 | if (unlikely(PageCompound(page))) |
| 323 | page = (struct page *)page->private; |
| 324 | atomic_inc(&page->_count); |
| 325 | } |
| 326 | |
| 327 | void put_page(struct page *page); |
| 328 | |
| 329 | #else /* CONFIG_HUGETLB_PAGE */ |
| 330 | |
| 331 | #define page_count(p) (atomic_read(&(p)->_count) + 1) |
| 332 | |
| 333 | static inline void get_page(struct page *page) |
| 334 | { |
| 335 | atomic_inc(&page->_count); |
| 336 | } |
| 337 | |
| 338 | static inline void put_page(struct page *page) |
| 339 | { |
| 340 | if (!PageReserved(page) && put_page_testzero(page)) |
| 341 | __page_cache_release(page); |
| 342 | } |
| 343 | |
| 344 | #endif /* CONFIG_HUGETLB_PAGE */ |
| 345 | |
| 346 | /* |
| 347 | * Multiple processes may "see" the same page. E.g. for untouched |
| 348 | * mappings of /dev/null, all processes see the same page full of |
| 349 | * zeroes, and text pages of executables and shared libraries have |
| 350 | * only one copy in memory, at most, normally. |
| 351 | * |
| 352 | * For the non-reserved pages, page_count(page) denotes a reference count. |
| 353 | * page_count() == 0 means the page is free. |
| 354 | * page_count() == 1 means the page is used for exactly one purpose |
| 355 | * (e.g. a private data page of one process). |
| 356 | * |
| 357 | * A page may be used for kmalloc() or anyone else who does a |
| 358 | * __get_free_page(). In this case the page_count() is at least 1, and |
| 359 | * all other fields are unused but should be 0 or NULL. The |
| 360 | * management of this page is the responsibility of the one who uses |
| 361 | * it. |
| 362 | * |
| 363 | * The other pages (we may call them "process pages") are completely |
| 364 | * managed by the Linux memory manager: I/O, buffers, swapping etc. |
| 365 | * The following discussion applies only to them. |
| 366 | * |
| 367 | * A page may belong to an inode's memory mapping. In this case, |
| 368 | * page->mapping is the pointer to the inode, and page->index is the |
| 369 | * file offset of the page, in units of PAGE_CACHE_SIZE. |
| 370 | * |
| 371 | * A page contains an opaque `private' member, which belongs to the |
| 372 | * page's address_space. Usually, this is the address of a circular |
| 373 | * list of the page's disk buffers. |
| 374 | * |
| 375 | * For pages belonging to inodes, the page_count() is the number of |
| 376 | * attaches, plus 1 if `private' contains something, plus one for |
| 377 | * the page cache itself. |
| 378 | * |
| 379 | * All pages belonging to an inode are in these doubly linked lists: |
| 380 | * mapping->clean_pages, mapping->dirty_pages and mapping->locked_pages; |
| 381 | * using the page->list list_head. These fields are also used for |
| 382 | * freelist managemet (when page_count()==0). |
| 383 | * |
| 384 | * There is also a per-mapping radix tree mapping index to the page |
| 385 | * in memory if present. The tree is rooted at mapping->root. |
| 386 | * |
| 387 | * All process pages can do I/O: |
| 388 | * - inode pages may need to be read from disk, |
| 389 | * - inode pages which have been modified and are MAP_SHARED may need |
| 390 | * to be written to disk, |
| 391 | * - private pages which have been modified may need to be swapped out |
| 392 | * to swap space and (later) to be read back into memory. |
| 393 | */ |
| 394 | |
| 395 | /* |
| 396 | * The zone field is never updated after free_area_init_core() |
| 397 | * sets it, so none of the operations on it need to be atomic. |
| 398 | * We'll have up to (MAX_NUMNODES * MAX_NR_ZONES) zones total, |
| 399 | * so we use (MAX_NODES_SHIFT + MAX_ZONES_SHIFT) here to get enough bits. |
| 400 | */ |
| 401 | #define NODEZONE_SHIFT (sizeof(page_flags_t)*8 - MAX_NODES_SHIFT - MAX_ZONES_SHIFT) |
| 402 | #define NODEZONE(node, zone) ((node << ZONES_SHIFT) | zone) |
| 403 | |
| 404 | static inline unsigned long page_zonenum(struct page *page) |
| 405 | { |
| 406 | return (page->flags >> NODEZONE_SHIFT) & (~(~0UL << ZONES_SHIFT)); |
| 407 | } |
| 408 | static inline unsigned long page_to_nid(struct page *page) |
| 409 | { |
| 410 | return (page->flags >> (NODEZONE_SHIFT + ZONES_SHIFT)); |
| 411 | } |
| 412 | |
| 413 | struct zone; |
| 414 | extern struct zone *zone_table[]; |
| 415 | |
| 416 | static inline struct zone *page_zone(struct page *page) |
| 417 | { |
| 418 | return zone_table[page->flags >> NODEZONE_SHIFT]; |
| 419 | } |
| 420 | |
| 421 | static inline void set_page_zone(struct page *page, unsigned long nodezone_num) |
| 422 | { |
| 423 | page->flags &= ~(~0UL << NODEZONE_SHIFT); |
| 424 | page->flags |= nodezone_num << NODEZONE_SHIFT; |
| 425 | } |
| 426 | |
| 427 | #ifndef CONFIG_DISCONTIGMEM |
| 428 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ |
| 429 | extern struct page *mem_map; |
| 430 | #endif |
| 431 | |
| 432 | static inline void *lowmem_page_address(struct page *page) |
| 433 | { |
| 434 | return __va(page_to_pfn(page) << PAGE_SHIFT); |
| 435 | } |
| 436 | |
| 437 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) |
| 438 | #define HASHED_PAGE_VIRTUAL |
| 439 | #endif |
| 440 | |
| 441 | #if defined(WANT_PAGE_VIRTUAL) |
| 442 | #define page_address(page) ((page)->virtual) |
| 443 | #define set_page_address(page, address) \ |
| 444 | do { \ |
| 445 | (page)->virtual = (address); \ |
| 446 | } while(0) |
| 447 | #define page_address_init() do { } while(0) |
| 448 | #endif |
| 449 | |
| 450 | #if defined(HASHED_PAGE_VIRTUAL) |
| 451 | void *page_address(struct page *page); |
| 452 | void set_page_address(struct page *page, void *virtual); |
| 453 | void page_address_init(void); |
| 454 | #endif |
| 455 | |
| 456 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) |
| 457 | #define page_address(page) lowmem_page_address(page) |
| 458 | #define set_page_address(page, address) do { } while(0) |
| 459 | #define page_address_init() do { } while(0) |
| 460 | #endif |
| 461 | |
| 462 | /* |
| 463 | * On an anonymous page mapped into a user virtual memory area, |
| 464 | * page->mapping points to its anon_vma, not to a struct address_space; |
| 465 | * with the PAGE_MAPPING_ANON bit set to distinguish it. |
| 466 | * |
| 467 | * Please note that, confusingly, "page_mapping" refers to the inode |
| 468 | * address_space which maps the page from disk; whereas "page_mapped" |
| 469 | * refers to user virtual address space into which the page is mapped. |
| 470 | */ |
| 471 | #define PAGE_MAPPING_ANON 1 |
| 472 | |
| 473 | extern struct address_space swapper_space; |
| 474 | static inline struct address_space *page_mapping(struct page *page) |
| 475 | { |
| 476 | struct address_space *mapping = page->mapping; |
| 477 | |
| 478 | if (unlikely(PageSwapCache(page))) |
| 479 | mapping = &swapper_space; |
| 480 | else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) |
| 481 | mapping = NULL; |
| 482 | return mapping; |
| 483 | } |
| 484 | |
| 485 | static inline int PageAnon(struct page *page) |
| 486 | { |
| 487 | return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; |
| 488 | } |
| 489 | |
| 490 | /* |
| 491 | * Return the pagecache index of the passed page. Regular pagecache pages |
| 492 | * use ->index whereas swapcache pages use ->private |
| 493 | */ |
| 494 | static inline pgoff_t page_index(struct page *page) |
| 495 | { |
| 496 | if (unlikely(PageSwapCache(page))) |
| 497 | return page->private; |
| 498 | return page->index; |
| 499 | } |
| 500 | |
| 501 | /* |
| 502 | * The atomic page->_mapcount, like _count, starts from -1: |
| 503 | * so that transitions both from it and to it can be tracked, |
| 504 | * using atomic_inc_and_test and atomic_add_negative(-1). |
| 505 | */ |
| 506 | static inline void reset_page_mapcount(struct page *page) |
| 507 | { |
| 508 | atomic_set(&(page)->_mapcount, -1); |
| 509 | } |
| 510 | |
| 511 | static inline int page_mapcount(struct page *page) |
| 512 | { |
| 513 | return atomic_read(&(page)->_mapcount) + 1; |
| 514 | } |
| 515 | |
| 516 | /* |
| 517 | * Return true if this page is mapped into pagetables. |
| 518 | */ |
| 519 | static inline int page_mapped(struct page *page) |
| 520 | { |
| 521 | return atomic_read(&(page)->_mapcount) >= 0; |
| 522 | } |
| 523 | |
| 524 | /* |
| 525 | * Error return values for the *_nopage functions |
| 526 | */ |
| 527 | #define NOPAGE_SIGBUS (NULL) |
| 528 | #define NOPAGE_OOM ((struct page *) (-1)) |
| 529 | |
| 530 | /* |
| 531 | * Different kinds of faults, as returned by handle_mm_fault(). |
| 532 | * Used to decide whether a process gets delivered SIGBUS or |
| 533 | * just gets major/minor fault counters bumped up. |
| 534 | */ |
| 535 | #define VM_FAULT_OOM (-1) |
| 536 | #define VM_FAULT_SIGBUS 0 |
| 537 | #define VM_FAULT_MINOR 1 |
| 538 | #define VM_FAULT_MAJOR 2 |
| 539 | |
| 540 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) |
| 541 | |
| 542 | extern void show_free_areas(void); |
| 543 | |
| 544 | #ifdef CONFIG_SHMEM |
| 545 | struct page *shmem_nopage(struct vm_area_struct *vma, |
| 546 | unsigned long address, int *type); |
| 547 | int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new); |
| 548 | struct mempolicy *shmem_get_policy(struct vm_area_struct *vma, |
| 549 | unsigned long addr); |
| 550 | int shmem_lock(struct file *file, int lock, struct user_struct *user); |
| 551 | #else |
| 552 | #define shmem_nopage filemap_nopage |
| 553 | #define shmem_lock(a, b, c) ({0;}) /* always in memory, no need to lock */ |
| 554 | #define shmem_set_policy(a, b) (0) |
| 555 | #define shmem_get_policy(a, b) (NULL) |
| 556 | #endif |
| 557 | struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags); |
| 558 | |
| 559 | int shmem_zero_setup(struct vm_area_struct *); |
| 560 | |
| 561 | static inline int can_do_mlock(void) |
| 562 | { |
| 563 | if (capable(CAP_IPC_LOCK)) |
| 564 | return 1; |
| 565 | if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0) |
| 566 | return 1; |
| 567 | return 0; |
| 568 | } |
| 569 | extern int user_shm_lock(size_t, struct user_struct *); |
| 570 | extern void user_shm_unlock(size_t, struct user_struct *); |
| 571 | |
| 572 | /* |
| 573 | * Parameter block passed down to zap_pte_range in exceptional cases. |
| 574 | */ |
| 575 | struct zap_details { |
| 576 | struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ |
| 577 | struct address_space *check_mapping; /* Check page->mapping if set */ |
| 578 | pgoff_t first_index; /* Lowest page->index to unmap */ |
| 579 | pgoff_t last_index; /* Highest page->index to unmap */ |
| 580 | spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 581 | unsigned long truncate_count; /* Compare vm_truncate_count */ |
| 582 | }; |
| 583 | |
Hugh Dickins | ee39b37 | 2005-04-19 13:29:15 -0700 | [diff] [blame] | 584 | 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] | 585 | unsigned long size, struct zap_details *); |
Hugh Dickins | ee39b37 | 2005-04-19 13:29:15 -0700 | [diff] [blame] | 586 | unsigned long unmap_vmas(struct mmu_gather **tlb, struct mm_struct *mm, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 587 | struct vm_area_struct *start_vma, unsigned long start_addr, |
| 588 | unsigned long end_addr, unsigned long *nr_accounted, |
| 589 | struct zap_details *); |
Hugh Dickins | 3bf5ee9 | 2005-04-19 13:29:16 -0700 | [diff] [blame] | 590 | void free_pgd_range(struct mmu_gather **tlb, unsigned long addr, |
| 591 | unsigned long end, unsigned long floor, unsigned long ceiling); |
| 592 | 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] | 593 | unsigned long floor, unsigned long ceiling); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 594 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, |
| 595 | struct vm_area_struct *vma); |
| 596 | int zeromap_page_range(struct vm_area_struct *vma, unsigned long from, |
| 597 | unsigned long size, pgprot_t prot); |
| 598 | void unmap_mapping_range(struct address_space *mapping, |
| 599 | loff_t const holebegin, loff_t const holelen, int even_cows); |
| 600 | |
| 601 | static inline void unmap_shared_mapping_range(struct address_space *mapping, |
| 602 | loff_t const holebegin, loff_t const holelen) |
| 603 | { |
| 604 | unmap_mapping_range(mapping, holebegin, holelen, 0); |
| 605 | } |
| 606 | |
| 607 | extern int vmtruncate(struct inode * inode, loff_t offset); |
| 608 | extern pud_t *FASTCALL(__pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)); |
| 609 | extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)); |
| 610 | extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); |
| 611 | extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); |
| 612 | extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); |
| 613 | extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot); |
| 614 | extern int handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access); |
| 615 | extern int make_pages_present(unsigned long addr, unsigned long end); |
| 616 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); |
| 617 | void install_arg_page(struct vm_area_struct *, struct page *, unsigned long); |
| 618 | |
| 619 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start, |
| 620 | int len, int write, int force, struct page **pages, struct vm_area_struct **vmas); |
| 621 | |
| 622 | int __set_page_dirty_buffers(struct page *page); |
| 623 | int __set_page_dirty_nobuffers(struct page *page); |
| 624 | int redirty_page_for_writepage(struct writeback_control *wbc, |
| 625 | struct page *page); |
| 626 | int FASTCALL(set_page_dirty(struct page *page)); |
| 627 | int set_page_dirty_lock(struct page *page); |
| 628 | int clear_page_dirty_for_io(struct page *page); |
| 629 | |
| 630 | extern unsigned long do_mremap(unsigned long addr, |
| 631 | unsigned long old_len, unsigned long new_len, |
| 632 | unsigned long flags, unsigned long new_addr); |
| 633 | |
| 634 | /* |
| 635 | * Prototype to add a shrinker callback for ageable caches. |
| 636 | * |
| 637 | * These functions are passed a count `nr_to_scan' and a gfpmask. They should |
| 638 | * scan `nr_to_scan' objects, attempting to free them. |
| 639 | * |
Domen Puncer | 845d343 | 2005-05-05 16:16:14 -0700 | [diff] [blame] | 640 | * 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] | 641 | * |
Domen Puncer | 845d343 | 2005-05-05 16:16:14 -0700 | [diff] [blame] | 642 | * 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] | 643 | * cache size, so a fastpath for that case is appropriate. |
| 644 | */ |
| 645 | typedef int (*shrinker_t)(int nr_to_scan, unsigned int gfp_mask); |
| 646 | |
| 647 | /* |
| 648 | * Add an aging callback. The int is the number of 'seeks' it takes |
| 649 | * to recreate one of the objects that these functions age. |
| 650 | */ |
| 651 | |
| 652 | #define DEFAULT_SEEKS 2 |
| 653 | struct shrinker; |
| 654 | extern struct shrinker *set_shrinker(int, shrinker_t); |
| 655 | extern void remove_shrinker(struct shrinker *shrinker); |
| 656 | |
| 657 | /* |
| 658 | * On a two-level or three-level page table, this ends up being trivial. Thus |
| 659 | * the inlining and the symmetry break with pte_alloc_map() that does all |
| 660 | * of this out-of-line. |
| 661 | */ |
| 662 | /* |
| 663 | * The following ifdef needed to get the 4level-fixup.h header to work. |
| 664 | * Remove it when 4level-fixup.h has been removed. |
| 665 | */ |
| 666 | #ifdef CONFIG_MMU |
| 667 | #ifndef __ARCH_HAS_4LEVEL_HACK |
| 668 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) |
| 669 | { |
| 670 | if (pgd_none(*pgd)) |
| 671 | return __pud_alloc(mm, pgd, address); |
| 672 | return pud_offset(pgd, address); |
| 673 | } |
| 674 | |
| 675 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) |
| 676 | { |
| 677 | if (pud_none(*pud)) |
| 678 | return __pmd_alloc(mm, pud, address); |
| 679 | return pmd_offset(pud, address); |
| 680 | } |
| 681 | #endif |
| 682 | #endif /* CONFIG_MMU */ |
| 683 | |
| 684 | extern void free_area_init(unsigned long * zones_size); |
| 685 | extern void free_area_init_node(int nid, pg_data_t *pgdat, |
| 686 | unsigned long * zones_size, unsigned long zone_start_pfn, |
| 687 | unsigned long *zholes_size); |
| 688 | extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long); |
| 689 | extern void mem_init(void); |
| 690 | extern void show_mem(void); |
| 691 | extern void si_meminfo(struct sysinfo * val); |
| 692 | extern void si_meminfo_node(struct sysinfo *val, int nid); |
| 693 | |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 694 | #ifdef CONFIG_NUMA |
| 695 | extern void setup_per_cpu_pageset(void); |
| 696 | #else |
| 697 | static inline void setup_per_cpu_pageset(void) {} |
| 698 | #endif |
| 699 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 700 | /* prio_tree.c */ |
| 701 | void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); |
| 702 | void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); |
| 703 | void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); |
| 704 | struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, |
| 705 | struct prio_tree_iter *iter); |
| 706 | |
| 707 | #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ |
| 708 | for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ |
| 709 | (vma = vma_prio_tree_next(vma, iter)); ) |
| 710 | |
| 711 | static inline void vma_nonlinear_insert(struct vm_area_struct *vma, |
| 712 | struct list_head *list) |
| 713 | { |
| 714 | vma->shared.vm_set.parent = NULL; |
| 715 | list_add_tail(&vma->shared.vm_set.list, list); |
| 716 | } |
| 717 | |
| 718 | /* mmap.c */ |
| 719 | extern int __vm_enough_memory(long pages, int cap_sys_admin); |
| 720 | extern void vma_adjust(struct vm_area_struct *vma, unsigned long start, |
| 721 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); |
| 722 | extern struct vm_area_struct *vma_merge(struct mm_struct *, |
| 723 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, |
| 724 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, |
| 725 | struct mempolicy *); |
| 726 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); |
| 727 | extern int split_vma(struct mm_struct *, |
| 728 | struct vm_area_struct *, unsigned long addr, int new_below); |
| 729 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); |
| 730 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, |
| 731 | struct rb_node **, struct rb_node *); |
| 732 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, |
| 733 | unsigned long addr, unsigned long len, pgoff_t pgoff); |
| 734 | extern void exit_mmap(struct mm_struct *); |
akpm@osdl.org | 119f657 | 2005-05-01 08:58:35 -0700 | [diff] [blame] | 735 | extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 736 | |
| 737 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); |
| 738 | |
| 739 | extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, |
| 740 | unsigned long len, unsigned long prot, |
| 741 | unsigned long flag, unsigned long pgoff); |
| 742 | |
| 743 | static inline unsigned long do_mmap(struct file *file, unsigned long addr, |
| 744 | unsigned long len, unsigned long prot, |
| 745 | unsigned long flag, unsigned long offset) |
| 746 | { |
| 747 | unsigned long ret = -EINVAL; |
| 748 | if ((offset + PAGE_ALIGN(len)) < offset) |
| 749 | goto out; |
| 750 | if (!(offset & ~PAGE_MASK)) |
| 751 | ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); |
| 752 | out: |
| 753 | return ret; |
| 754 | } |
| 755 | |
| 756 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); |
| 757 | |
| 758 | extern unsigned long do_brk(unsigned long, unsigned long); |
| 759 | |
| 760 | /* filemap.c */ |
| 761 | extern unsigned long page_unuse(struct page *); |
| 762 | extern void truncate_inode_pages(struct address_space *, loff_t); |
| 763 | |
| 764 | /* generic vm_area_ops exported for stackable file systems */ |
| 765 | extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *); |
| 766 | extern int filemap_populate(struct vm_area_struct *, unsigned long, |
| 767 | unsigned long, pgprot_t, unsigned long, int); |
| 768 | |
| 769 | /* mm/page-writeback.c */ |
| 770 | int write_one_page(struct page *page, int wait); |
| 771 | |
| 772 | /* readahead.c */ |
| 773 | #define VM_MAX_READAHEAD 128 /* kbytes */ |
| 774 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ |
| 775 | #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before |
| 776 | * turning readahead off */ |
| 777 | |
| 778 | int do_page_cache_readahead(struct address_space *mapping, struct file *filp, |
| 779 | unsigned long offset, unsigned long nr_to_read); |
| 780 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, |
| 781 | unsigned long offset, unsigned long nr_to_read); |
| 782 | unsigned long page_cache_readahead(struct address_space *mapping, |
| 783 | struct file_ra_state *ra, |
| 784 | struct file *filp, |
| 785 | unsigned long offset, |
| 786 | unsigned long size); |
| 787 | void handle_ra_miss(struct address_space *mapping, |
| 788 | struct file_ra_state *ra, pgoff_t offset); |
| 789 | unsigned long max_sane_readahead(unsigned long nr); |
| 790 | |
| 791 | /* Do stack extension */ |
| 792 | extern int expand_stack(struct vm_area_struct * vma, unsigned long address); |
| 793 | |
| 794 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ |
| 795 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); |
| 796 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, |
| 797 | struct vm_area_struct **pprev); |
| 798 | |
| 799 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, |
| 800 | NULL if none. Assume start_addr < end_addr. */ |
| 801 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) |
| 802 | { |
| 803 | struct vm_area_struct * vma = find_vma(mm,start_addr); |
| 804 | |
| 805 | if (vma && end_addr <= vma->vm_start) |
| 806 | vma = NULL; |
| 807 | return vma; |
| 808 | } |
| 809 | |
| 810 | static inline unsigned long vma_pages(struct vm_area_struct *vma) |
| 811 | { |
| 812 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
| 813 | } |
| 814 | |
| 815 | extern struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr); |
| 816 | |
| 817 | extern struct page * vmalloc_to_page(void *addr); |
| 818 | extern unsigned long vmalloc_to_pfn(void *addr); |
| 819 | extern struct page * follow_page(struct mm_struct *mm, unsigned long address, |
| 820 | int write); |
| 821 | extern int check_user_page_readable(struct mm_struct *mm, unsigned long address); |
| 822 | int remap_pfn_range(struct vm_area_struct *, unsigned long, |
| 823 | unsigned long, unsigned long, pgprot_t); |
| 824 | |
| 825 | #ifdef CONFIG_PROC_FS |
| 826 | void __vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); |
| 827 | #else |
| 828 | static inline void __vm_stat_account(struct mm_struct *mm, |
| 829 | unsigned long flags, struct file *file, long pages) |
| 830 | { |
| 831 | } |
| 832 | #endif /* CONFIG_PROC_FS */ |
| 833 | |
| 834 | static inline void vm_stat_account(struct vm_area_struct *vma) |
| 835 | { |
| 836 | __vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file, |
| 837 | vma_pages(vma)); |
| 838 | } |
| 839 | |
| 840 | static inline void vm_stat_unaccount(struct vm_area_struct *vma) |
| 841 | { |
| 842 | __vm_stat_account(vma->vm_mm, vma->vm_flags, vma->vm_file, |
| 843 | -vma_pages(vma)); |
| 844 | } |
| 845 | |
| 846 | /* update per process rss and vm hiwater data */ |
| 847 | extern void update_mem_hiwater(struct task_struct *tsk); |
| 848 | |
| 849 | #ifndef CONFIG_DEBUG_PAGEALLOC |
| 850 | static inline void |
| 851 | kernel_map_pages(struct page *page, int numpages, int enable) |
| 852 | { |
| 853 | } |
| 854 | #endif |
| 855 | |
| 856 | extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); |
| 857 | #ifdef __HAVE_ARCH_GATE_AREA |
| 858 | int in_gate_area_no_task(unsigned long addr); |
| 859 | int in_gate_area(struct task_struct *task, unsigned long addr); |
| 860 | #else |
| 861 | int in_gate_area_no_task(unsigned long addr); |
| 862 | #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) |
| 863 | #endif /* __HAVE_ARCH_GATE_AREA */ |
| 864 | |
Andrea Arcangeli | 79befd0 | 2005-04-16 15:24:05 -0700 | [diff] [blame] | 865 | /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */ |
| 866 | #define OOM_DISABLE -17 |
| 867 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 868 | #endif /* __KERNEL__ */ |
| 869 | #endif /* _LINUX_MM_H */ |