mm.h 35 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061
  1. #ifndef _LINUX_MM_H
  2. #define _LINUX_MM_H
  3. #include <linux/sched.h>
  4. #include <linux/errno.h>
  5. #include <linux/capability.h>
  6. #ifdef __KERNEL__
  7. #include <linux/config.h>
  8. #include <linux/gfp.h>
  9. #include <linux/list.h>
  10. #include <linux/mmzone.h>
  11. #include <linux/rbtree.h>
  12. #include <linux/prio_tree.h>
  13. #include <linux/fs.h>
  14. #include <linux/mutex.h>
  15. struct mempolicy;
  16. struct anon_vma;
  17. #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
  18. extern unsigned long max_mapnr;
  19. #endif
  20. extern unsigned long num_physpages;
  21. extern void * high_memory;
  22. extern unsigned long vmalloc_earlyreserve;
  23. extern int page_cluster;
  24. #ifdef CONFIG_SYSCTL
  25. extern int sysctl_legacy_va_layout;
  26. #else
  27. #define sysctl_legacy_va_layout 0
  28. #endif
  29. #include <asm/page.h>
  30. #include <asm/pgtable.h>
  31. #include <asm/processor.h>
  32. #include <asm/atomic.h>
  33. #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  34. /*
  35. * Linux kernel virtual memory manager primitives.
  36. * The idea being to have a "virtual" mm in the same way
  37. * we have a virtual fs - giving a cleaner interface to the
  38. * mm details, and allowing different kinds of memory mappings
  39. * (from shared memory to executable loading to arbitrary
  40. * mmap() functions).
  41. */
  42. /*
  43. * This struct defines a memory VMM memory area. There is one of these
  44. * per VM-area/task. A VM area is any part of the process virtual memory
  45. * space that has a special rule for the page-fault handlers (ie a shared
  46. * library, the executable area etc).
  47. */
  48. struct vm_area_struct {
  49. struct mm_struct * vm_mm; /* The address space we belong to. */
  50. unsigned long vm_start; /* Our start address within vm_mm. */
  51. unsigned long vm_end; /* The first byte after our end address
  52. within vm_mm. */
  53. /* linked list of VM areas per task, sorted by address */
  54. struct vm_area_struct *vm_next;
  55. pgprot_t vm_page_prot; /* Access permissions of this VMA. */
  56. unsigned long vm_flags; /* Flags, listed below. */
  57. struct rb_node vm_rb;
  58. /*
  59. * For areas with an address space and backing store,
  60. * linkage into the address_space->i_mmap prio tree, or
  61. * linkage to the list of like vmas hanging off its node, or
  62. * linkage of vma in the address_space->i_mmap_nonlinear list.
  63. */
  64. union {
  65. struct {
  66. struct list_head list;
  67. void *parent; /* aligns with prio_tree_node parent */
  68. struct vm_area_struct *head;
  69. } vm_set;
  70. struct raw_prio_tree_node prio_tree_node;
  71. } shared;
  72. /*
  73. * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
  74. * list, after a COW of one of the file pages. A MAP_SHARED vma
  75. * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
  76. * or brk vma (with NULL file) can only be in an anon_vma list.
  77. */
  78. struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
  79. struct anon_vma *anon_vma; /* Serialized by page_table_lock */
  80. /* Function pointers to deal with this struct. */
  81. struct vm_operations_struct * vm_ops;
  82. /* Information about our backing store: */
  83. unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
  84. units, *not* PAGE_CACHE_SIZE */
  85. struct file * vm_file; /* File we map to (can be NULL). */
  86. void * vm_private_data; /* was vm_pte (shared mem) */
  87. unsigned long vm_truncate_count;/* truncate_count or restart_addr */
  88. #ifndef CONFIG_MMU
  89. atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
  90. #endif
  91. #ifdef CONFIG_NUMA
  92. struct mempolicy *vm_policy; /* NUMA policy for the VMA */
  93. #endif
  94. };
  95. /*
  96. * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
  97. * disabled, then there's a single shared list of VMAs maintained by the
  98. * system, and mm's subscribe to these individually
  99. */
  100. struct vm_list_struct {
  101. struct vm_list_struct *next;
  102. struct vm_area_struct *vma;
  103. };
  104. #ifndef CONFIG_MMU
  105. extern struct rb_root nommu_vma_tree;
  106. extern struct rw_semaphore nommu_vma_sem;
  107. extern unsigned int kobjsize(const void *objp);
  108. #endif
  109. /*
  110. * vm_flags..
  111. */
  112. #define VM_READ 0x00000001 /* currently active flags */
  113. #define VM_WRITE 0x00000002
  114. #define VM_EXEC 0x00000004
  115. #define VM_SHARED 0x00000008
  116. /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  117. #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
  118. #define VM_MAYWRITE 0x00000020
  119. #define VM_MAYEXEC 0x00000040
  120. #define VM_MAYSHARE 0x00000080
  121. #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
  122. #define VM_GROWSUP 0x00000200
  123. #define VM_SHM 0x00000000 /* Means nothing: delete it later */
  124. #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
  125. #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
  126. #define VM_EXECUTABLE 0x00001000
  127. #define VM_LOCKED 0x00002000
  128. #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
  129. /* Used by sys_madvise() */
  130. #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
  131. #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
  132. #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
  133. #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
  134. #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
  135. #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
  136. #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
  137. #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
  138. #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
  139. #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
  140. #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
  141. #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
  142. #endif
  143. #ifdef CONFIG_STACK_GROWSUP
  144. #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  145. #else
  146. #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  147. #endif
  148. #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
  149. #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
  150. #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
  151. #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
  152. #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
  153. /*
  154. * mapping from the currently active vm_flags protection bits (the
  155. * low four bits) to a page protection mask..
  156. */
  157. extern pgprot_t protection_map[16];
  158. /*
  159. * These are the virtual MM functions - opening of an area, closing and
  160. * unmapping it (needed to keep files on disk up-to-date etc), pointer
  161. * to the functions called when a no-page or a wp-page exception occurs.
  162. */
  163. struct vm_operations_struct {
  164. void (*open)(struct vm_area_struct * area);
  165. void (*close)(struct vm_area_struct * area);
  166. struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
  167. int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
  168. #ifdef CONFIG_NUMA
  169. int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
  170. struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
  171. unsigned long addr);
  172. #endif
  173. };
  174. struct mmu_gather;
  175. struct inode;
  176. /*
  177. * Each physical page in the system has a struct page associated with
  178. * it to keep track of whatever it is we are using the page for at the
  179. * moment. Note that we have no way to track which tasks are using
  180. * a page.
  181. */
  182. struct page {
  183. unsigned long flags; /* Atomic flags, some possibly
  184. * updated asynchronously */
  185. atomic_t _count; /* Usage count, see below. */
  186. atomic_t _mapcount; /* Count of ptes mapped in mms,
  187. * to show when page is mapped
  188. * & limit reverse map searches.
  189. */
  190. union {
  191. struct {
  192. unsigned long private; /* Mapping-private opaque data:
  193. * usually used for buffer_heads
  194. * if PagePrivate set; used for
  195. * swp_entry_t if PageSwapCache.
  196. * When page is free, this
  197. * indicates order in the buddy
  198. * system.
  199. */
  200. struct address_space *mapping; /* If low bit clear, points to
  201. * inode address_space, or NULL.
  202. * If page mapped as anonymous
  203. * memory, low bit is set, and
  204. * it points to anon_vma object:
  205. * see PAGE_MAPPING_ANON below.
  206. */
  207. };
  208. #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
  209. spinlock_t ptl;
  210. #endif
  211. };
  212. pgoff_t index; /* Our offset within mapping. */
  213. struct list_head lru; /* Pageout list, eg. active_list
  214. * protected by zone->lru_lock !
  215. */
  216. /*
  217. * On machines where all RAM is mapped into kernel address space,
  218. * we can simply calculate the virtual address. On machines with
  219. * highmem some memory is mapped into kernel virtual memory
  220. * dynamically, so we need a place to store that address.
  221. * Note that this field could be 16 bits on x86 ... ;)
  222. *
  223. * Architectures with slow multiplication can define
  224. * WANT_PAGE_VIRTUAL in asm/page.h
  225. */
  226. #if defined(WANT_PAGE_VIRTUAL)
  227. void *virtual; /* Kernel virtual address (NULL if
  228. not kmapped, ie. highmem) */
  229. #endif /* WANT_PAGE_VIRTUAL */
  230. };
  231. #define page_private(page) ((page)->private)
  232. #define set_page_private(page, v) ((page)->private = (v))
  233. /*
  234. * FIXME: take this include out, include page-flags.h in
  235. * files which need it (119 of them)
  236. */
  237. #include <linux/page-flags.h>
  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.
  249. */
  250. /*
  251. * Drop a ref, return true if the logical refcount fell to zero (the page has
  252. * no users)
  253. */
  254. static inline int put_page_testzero(struct page *page)
  255. {
  256. BUG_ON(atomic_read(&page->_count) == 0);
  257. return atomic_dec_and_test(&page->_count);
  258. }
  259. /*
  260. * Try to grab a ref unless the page has a refcount of zero, return false if
  261. * that is the case.
  262. */
  263. static inline int get_page_unless_zero(struct page *page)
  264. {
  265. return atomic_inc_not_zero(&page->_count);
  266. }
  267. #define set_page_count(p,v) atomic_set(&(p)->_count, (v))
  268. #define __put_page(p) atomic_dec(&(p)->_count)
  269. extern void FASTCALL(__page_cache_release(struct page *));
  270. static inline int page_count(struct page *page)
  271. {
  272. if (PageCompound(page))
  273. page = (struct page *)page_private(page);
  274. return atomic_read(&page->_count);
  275. }
  276. static inline void get_page(struct page *page)
  277. {
  278. if (unlikely(PageCompound(page)))
  279. page = (struct page *)page_private(page);
  280. atomic_inc(&page->_count);
  281. }
  282. void put_page(struct page *page);
  283. #ifdef CONFIG_MMU
  284. void split_page(struct page *page, unsigned int order);
  285. #else
  286. static inline void split_page(struct page *page, unsigned int order) {}
  287. #endif
  288. /*
  289. * Multiple processes may "see" the same page. E.g. for untouched
  290. * mappings of /dev/null, all processes see the same page full of
  291. * zeroes, and text pages of executables and shared libraries have
  292. * only one copy in memory, at most, normally.
  293. *
  294. * For the non-reserved pages, page_count(page) denotes a reference count.
  295. * page_count() == 0 means the page is free. page->lru is then used for
  296. * freelist management in the buddy allocator.
  297. * page_count() == 1 means the page is used for exactly one purpose
  298. * (e.g. a private data page of one process).
  299. *
  300. * A page may be used for kmalloc() or anyone else who does a
  301. * __get_free_page(). In this case the page_count() is at least 1, and
  302. * all other fields are unused but should be 0 or NULL. The
  303. * management of this page is the responsibility of the one who uses
  304. * it.
  305. *
  306. * The other pages (we may call them "process pages") are completely
  307. * managed by the Linux memory manager: I/O, buffers, swapping etc.
  308. * The following discussion applies only to them.
  309. *
  310. * A page may belong to an inode's memory mapping. In this case,
  311. * page->mapping is the pointer to the inode, and page->index is the
  312. * file offset of the page, in units of PAGE_CACHE_SIZE.
  313. *
  314. * A page contains an opaque `private' member, which belongs to the
  315. * page's address_space. Usually, this is the address of a circular
  316. * list of the page's disk buffers.
  317. *
  318. * For pages belonging to inodes, the page_count() is the number of
  319. * attaches, plus 1 if `private' contains something, plus one for
  320. * the page cache itself.
  321. *
  322. * Instead of keeping dirty/clean pages in per address-space lists, we instead
  323. * now tag pages as dirty/under writeback in the radix tree.
  324. *
  325. * There is also a per-mapping radix tree mapping index to the page
  326. * in memory if present. The tree is rooted at mapping->root.
  327. *
  328. * All process pages can do I/O:
  329. * - inode pages may need to be read from disk,
  330. * - inode pages which have been modified and are MAP_SHARED may need
  331. * to be written to disk,
  332. * - private pages which have been modified may need to be swapped out
  333. * to swap space and (later) to be read back into memory.
  334. */
  335. /*
  336. * The zone field is never updated after free_area_init_core()
  337. * sets it, so none of the operations on it need to be atomic.
  338. */
  339. /*
  340. * page->flags layout:
  341. *
  342. * There are three possibilities for how page->flags get
  343. * laid out. The first is for the normal case, without
  344. * sparsemem. The second is for sparsemem when there is
  345. * plenty of space for node and section. The last is when
  346. * we have run out of space and have to fall back to an
  347. * alternate (slower) way of determining the node.
  348. *
  349. * No sparsemem: | NODE | ZONE | ... | FLAGS |
  350. * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
  351. * no space for node: | SECTION | ZONE | ... | FLAGS |
  352. */
  353. #ifdef CONFIG_SPARSEMEM
  354. #define SECTIONS_WIDTH SECTIONS_SHIFT
  355. #else
  356. #define SECTIONS_WIDTH 0
  357. #endif
  358. #define ZONES_WIDTH ZONES_SHIFT
  359. #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
  360. #define NODES_WIDTH NODES_SHIFT
  361. #else
  362. #define NODES_WIDTH 0
  363. #endif
  364. /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
  365. #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
  366. #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
  367. #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
  368. /*
  369. * We are going to use the flags for the page to node mapping if its in
  370. * there. This includes the case where there is no node, so it is implicit.
  371. */
  372. #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
  373. #ifndef PFN_SECTION_SHIFT
  374. #define PFN_SECTION_SHIFT 0
  375. #endif
  376. /*
  377. * Define the bit shifts to access each section. For non-existant
  378. * sections we define the shift as 0; that plus a 0 mask ensures
  379. * the compiler will optimise away reference to them.
  380. */
  381. #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
  382. #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
  383. #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
  384. /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
  385. #if FLAGS_HAS_NODE
  386. #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
  387. #else
  388. #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
  389. #endif
  390. #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
  391. #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
  392. #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
  393. #endif
  394. #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
  395. #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
  396. #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
  397. #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
  398. static inline unsigned long page_zonenum(struct page *page)
  399. {
  400. return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
  401. }
  402. struct zone;
  403. extern struct zone *zone_table[];
  404. static inline struct zone *page_zone(struct page *page)
  405. {
  406. return zone_table[(page->flags >> ZONETABLE_PGSHIFT) &
  407. ZONETABLE_MASK];
  408. }
  409. static inline unsigned long page_to_nid(struct page *page)
  410. {
  411. if (FLAGS_HAS_NODE)
  412. return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
  413. else
  414. return page_zone(page)->zone_pgdat->node_id;
  415. }
  416. static inline unsigned long page_to_section(struct page *page)
  417. {
  418. return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
  419. }
  420. static inline void set_page_zone(struct page *page, unsigned long zone)
  421. {
  422. page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
  423. page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
  424. }
  425. static inline void set_page_node(struct page *page, unsigned long node)
  426. {
  427. page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
  428. page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
  429. }
  430. static inline void set_page_section(struct page *page, unsigned long section)
  431. {
  432. page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
  433. page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
  434. }
  435. static inline void set_page_links(struct page *page, unsigned long zone,
  436. unsigned long node, unsigned long pfn)
  437. {
  438. set_page_zone(page, zone);
  439. set_page_node(page, node);
  440. set_page_section(page, pfn_to_section_nr(pfn));
  441. }
  442. #ifndef CONFIG_DISCONTIGMEM
  443. /* The array of struct pages - for discontigmem use pgdat->lmem_map */
  444. extern struct page *mem_map;
  445. #endif
  446. static __always_inline void *lowmem_page_address(struct page *page)
  447. {
  448. return __va(page_to_pfn(page) << PAGE_SHIFT);
  449. }
  450. #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
  451. #define HASHED_PAGE_VIRTUAL
  452. #endif
  453. #if defined(WANT_PAGE_VIRTUAL)
  454. #define page_address(page) ((page)->virtual)
  455. #define set_page_address(page, address) \
  456. do { \
  457. (page)->virtual = (address); \
  458. } while(0)
  459. #define page_address_init() do { } while(0)
  460. #endif
  461. #if defined(HASHED_PAGE_VIRTUAL)
  462. void *page_address(struct page *page);
  463. void set_page_address(struct page *page, void *virtual);
  464. void page_address_init(void);
  465. #endif
  466. #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
  467. #define page_address(page) lowmem_page_address(page)
  468. #define set_page_address(page, address) do { } while(0)
  469. #define page_address_init() do { } while(0)
  470. #endif
  471. /*
  472. * On an anonymous page mapped into a user virtual memory area,
  473. * page->mapping points to its anon_vma, not to a struct address_space;
  474. * with the PAGE_MAPPING_ANON bit set to distinguish it.
  475. *
  476. * Please note that, confusingly, "page_mapping" refers to the inode
  477. * address_space which maps the page from disk; whereas "page_mapped"
  478. * refers to user virtual address space into which the page is mapped.
  479. */
  480. #define PAGE_MAPPING_ANON 1
  481. extern struct address_space swapper_space;
  482. static inline struct address_space *page_mapping(struct page *page)
  483. {
  484. struct address_space *mapping = page->mapping;
  485. if (unlikely(PageSwapCache(page)))
  486. mapping = &swapper_space;
  487. else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
  488. mapping = NULL;
  489. return mapping;
  490. }
  491. static inline int PageAnon(struct page *page)
  492. {
  493. return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
  494. }
  495. /*
  496. * Return the pagecache index of the passed page. Regular pagecache pages
  497. * use ->index whereas swapcache pages use ->private
  498. */
  499. static inline pgoff_t page_index(struct page *page)
  500. {
  501. if (unlikely(PageSwapCache(page)))
  502. return page_private(page);
  503. return page->index;
  504. }
  505. /*
  506. * The atomic page->_mapcount, like _count, starts from -1:
  507. * so that transitions both from it and to it can be tracked,
  508. * using atomic_inc_and_test and atomic_add_negative(-1).
  509. */
  510. static inline void reset_page_mapcount(struct page *page)
  511. {
  512. atomic_set(&(page)->_mapcount, -1);
  513. }
  514. static inline int page_mapcount(struct page *page)
  515. {
  516. return atomic_read(&(page)->_mapcount) + 1;
  517. }
  518. /*
  519. * Return true if this page is mapped into pagetables.
  520. */
  521. static inline int page_mapped(struct page *page)
  522. {
  523. return atomic_read(&(page)->_mapcount) >= 0;
  524. }
  525. /*
  526. * Error return values for the *_nopage functions
  527. */
  528. #define NOPAGE_SIGBUS (NULL)
  529. #define NOPAGE_OOM ((struct page *) (-1))
  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 0x00
  536. #define VM_FAULT_SIGBUS 0x01
  537. #define VM_FAULT_MINOR 0x02
  538. #define VM_FAULT_MAJOR 0x03
  539. /*
  540. * Special case for get_user_pages.
  541. * Must be in a distinct bit from the above VM_FAULT_ flags.
  542. */
  543. #define VM_FAULT_WRITE 0x10
  544. #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
  545. extern void show_free_areas(void);
  546. #ifdef CONFIG_SHMEM
  547. struct page *shmem_nopage(struct vm_area_struct *vma,
  548. unsigned long address, int *type);
  549. int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
  550. struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
  551. unsigned long addr);
  552. int shmem_lock(struct file *file, int lock, struct user_struct *user);
  553. #else
  554. #define shmem_nopage filemap_nopage
  555. static inline int shmem_lock(struct file *file, int lock,
  556. struct user_struct *user)
  557. {
  558. return 0;
  559. }
  560. static inline int shmem_set_policy(struct vm_area_struct *vma,
  561. struct mempolicy *new)
  562. {
  563. return 0;
  564. }
  565. static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
  566. unsigned long addr)
  567. {
  568. return NULL;
  569. }
  570. #endif
  571. struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
  572. extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
  573. int shmem_zero_setup(struct vm_area_struct *);
  574. #ifndef CONFIG_MMU
  575. extern unsigned long shmem_get_unmapped_area(struct file *file,
  576. unsigned long addr,
  577. unsigned long len,
  578. unsigned long pgoff,
  579. unsigned long flags);
  580. #endif
  581. static inline int can_do_mlock(void)
  582. {
  583. if (capable(CAP_IPC_LOCK))
  584. return 1;
  585. if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
  586. return 1;
  587. return 0;
  588. }
  589. extern int user_shm_lock(size_t, struct user_struct *);
  590. extern void user_shm_unlock(size_t, struct user_struct *);
  591. /*
  592. * Parameter block passed down to zap_pte_range in exceptional cases.
  593. */
  594. struct zap_details {
  595. struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
  596. struct address_space *check_mapping; /* Check page->mapping if set */
  597. pgoff_t first_index; /* Lowest page->index to unmap */
  598. pgoff_t last_index; /* Highest page->index to unmap */
  599. spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
  600. unsigned long truncate_count; /* Compare vm_truncate_count */
  601. };
  602. struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
  603. unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
  604. unsigned long size, struct zap_details *);
  605. unsigned long unmap_vmas(struct mmu_gather **tlb,
  606. struct vm_area_struct *start_vma, unsigned long start_addr,
  607. unsigned long end_addr, unsigned long *nr_accounted,
  608. struct zap_details *);
  609. void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
  610. unsigned long end, unsigned long floor, unsigned long ceiling);
  611. void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
  612. unsigned long floor, unsigned long ceiling);
  613. int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
  614. struct vm_area_struct *vma);
  615. int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
  616. unsigned long size, pgprot_t prot);
  617. void unmap_mapping_range(struct address_space *mapping,
  618. loff_t const holebegin, loff_t const holelen, int even_cows);
  619. static inline void unmap_shared_mapping_range(struct address_space *mapping,
  620. loff_t const holebegin, loff_t const holelen)
  621. {
  622. unmap_mapping_range(mapping, holebegin, holelen, 0);
  623. }
  624. extern int vmtruncate(struct inode * inode, loff_t offset);
  625. extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
  626. extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
  627. extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
  628. #ifdef CONFIG_MMU
  629. extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
  630. unsigned long address, int write_access);
  631. static inline int handle_mm_fault(struct mm_struct *mm,
  632. struct vm_area_struct *vma, unsigned long address,
  633. int write_access)
  634. {
  635. return __handle_mm_fault(mm, vma, address, write_access) &
  636. (~VM_FAULT_WRITE);
  637. }
  638. #else
  639. static inline int handle_mm_fault(struct mm_struct *mm,
  640. struct vm_area_struct *vma, unsigned long address,
  641. int write_access)
  642. {
  643. /* should never happen if there's no MMU */
  644. BUG();
  645. return VM_FAULT_SIGBUS;
  646. }
  647. #endif
  648. extern int make_pages_present(unsigned long addr, unsigned long end);
  649. extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
  650. void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
  651. int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
  652. int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
  653. void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
  654. int __set_page_dirty_buffers(struct page *page);
  655. int __set_page_dirty_nobuffers(struct page *page);
  656. int redirty_page_for_writepage(struct writeback_control *wbc,
  657. struct page *page);
  658. int FASTCALL(set_page_dirty(struct page *page));
  659. int set_page_dirty_lock(struct page *page);
  660. int clear_page_dirty_for_io(struct page *page);
  661. extern unsigned long do_mremap(unsigned long addr,
  662. unsigned long old_len, unsigned long new_len,
  663. unsigned long flags, unsigned long new_addr);
  664. /*
  665. * Prototype to add a shrinker callback for ageable caches.
  666. *
  667. * These functions are passed a count `nr_to_scan' and a gfpmask. They should
  668. * scan `nr_to_scan' objects, attempting to free them.
  669. *
  670. * The callback must return the number of objects which remain in the cache.
  671. *
  672. * The callback will be passed nr_to_scan == 0 when the VM is querying the
  673. * cache size, so a fastpath for that case is appropriate.
  674. */
  675. typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
  676. /*
  677. * Add an aging callback. The int is the number of 'seeks' it takes
  678. * to recreate one of the objects that these functions age.
  679. */
  680. #define DEFAULT_SEEKS 2
  681. struct shrinker;
  682. extern struct shrinker *set_shrinker(int, shrinker_t);
  683. extern void remove_shrinker(struct shrinker *shrinker);
  684. extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
  685. int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
  686. int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
  687. int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
  688. int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
  689. /*
  690. * The following ifdef needed to get the 4level-fixup.h header to work.
  691. * Remove it when 4level-fixup.h has been removed.
  692. */
  693. #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
  694. static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
  695. {
  696. return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
  697. NULL: pud_offset(pgd, address);
  698. }
  699. static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
  700. {
  701. return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
  702. NULL: pmd_offset(pud, address);
  703. }
  704. #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
  705. #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
  706. /*
  707. * We tuck a spinlock to guard each pagetable page into its struct page,
  708. * at page->private, with BUILD_BUG_ON to make sure that this will not
  709. * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
  710. * When freeing, reset page->mapping so free_pages_check won't complain.
  711. */
  712. #define __pte_lockptr(page) &((page)->ptl)
  713. #define pte_lock_init(_page) do { \
  714. spin_lock_init(__pte_lockptr(_page)); \
  715. } while (0)
  716. #define pte_lock_deinit(page) ((page)->mapping = NULL)
  717. #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
  718. #else
  719. /*
  720. * We use mm->page_table_lock to guard all pagetable pages of the mm.
  721. */
  722. #define pte_lock_init(page) do {} while (0)
  723. #define pte_lock_deinit(page) do {} while (0)
  724. #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
  725. #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
  726. #define pte_offset_map_lock(mm, pmd, address, ptlp) \
  727. ({ \
  728. spinlock_t *__ptl = pte_lockptr(mm, pmd); \
  729. pte_t *__pte = pte_offset_map(pmd, address); \
  730. *(ptlp) = __ptl; \
  731. spin_lock(__ptl); \
  732. __pte; \
  733. })
  734. #define pte_unmap_unlock(pte, ptl) do { \
  735. spin_unlock(ptl); \
  736. pte_unmap(pte); \
  737. } while (0)
  738. #define pte_alloc_map(mm, pmd, address) \
  739. ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
  740. NULL: pte_offset_map(pmd, address))
  741. #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
  742. ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
  743. NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
  744. #define pte_alloc_kernel(pmd, address) \
  745. ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
  746. NULL: pte_offset_kernel(pmd, address))
  747. extern void free_area_init(unsigned long * zones_size);
  748. extern void free_area_init_node(int nid, pg_data_t *pgdat,
  749. unsigned long * zones_size, unsigned long zone_start_pfn,
  750. unsigned long *zholes_size);
  751. extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
  752. extern void setup_per_zone_pages_min(void);
  753. extern void mem_init(void);
  754. extern void show_mem(void);
  755. extern void si_meminfo(struct sysinfo * val);
  756. extern void si_meminfo_node(struct sysinfo *val, int nid);
  757. #ifdef CONFIG_NUMA
  758. extern void setup_per_cpu_pageset(void);
  759. #else
  760. static inline void setup_per_cpu_pageset(void) {}
  761. #endif
  762. /* prio_tree.c */
  763. void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
  764. void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
  765. void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
  766. struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
  767. struct prio_tree_iter *iter);
  768. #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
  769. for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
  770. (vma = vma_prio_tree_next(vma, iter)); )
  771. static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
  772. struct list_head *list)
  773. {
  774. vma->shared.vm_set.parent = NULL;
  775. list_add_tail(&vma->shared.vm_set.list, list);
  776. }
  777. /* mmap.c */
  778. extern int __vm_enough_memory(long pages, int cap_sys_admin);
  779. extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
  780. unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
  781. extern struct vm_area_struct *vma_merge(struct mm_struct *,
  782. struct vm_area_struct *prev, unsigned long addr, unsigned long end,
  783. unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
  784. struct mempolicy *);
  785. extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
  786. extern int split_vma(struct mm_struct *,
  787. struct vm_area_struct *, unsigned long addr, int new_below);
  788. extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
  789. extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
  790. struct rb_node **, struct rb_node *);
  791. extern void unlink_file_vma(struct vm_area_struct *);
  792. extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
  793. unsigned long addr, unsigned long len, pgoff_t pgoff);
  794. extern void exit_mmap(struct mm_struct *);
  795. extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
  796. extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
  797. extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
  798. unsigned long len, unsigned long prot,
  799. unsigned long flag, unsigned long pgoff);
  800. static inline unsigned long do_mmap(struct file *file, unsigned long addr,
  801. unsigned long len, unsigned long prot,
  802. unsigned long flag, unsigned long offset)
  803. {
  804. unsigned long ret = -EINVAL;
  805. if ((offset + PAGE_ALIGN(len)) < offset)
  806. goto out;
  807. if (!(offset & ~PAGE_MASK))
  808. ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
  809. out:
  810. return ret;
  811. }
  812. extern int do_munmap(struct mm_struct *, unsigned long, size_t);
  813. extern unsigned long do_brk(unsigned long, unsigned long);
  814. /* filemap.c */
  815. extern unsigned long page_unuse(struct page *);
  816. extern void truncate_inode_pages(struct address_space *, loff_t);
  817. extern void truncate_inode_pages_range(struct address_space *,
  818. loff_t lstart, loff_t lend);
  819. /* generic vm_area_ops exported for stackable file systems */
  820. extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
  821. extern int filemap_populate(struct vm_area_struct *, unsigned long,
  822. unsigned long, pgprot_t, unsigned long, int);
  823. /* mm/page-writeback.c */
  824. int write_one_page(struct page *page, int wait);
  825. /* readahead.c */
  826. #define VM_MAX_READAHEAD 128 /* kbytes */
  827. #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
  828. #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
  829. * turning readahead off */
  830. int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
  831. pgoff_t offset, unsigned long nr_to_read);
  832. int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
  833. pgoff_t offset, unsigned long nr_to_read);
  834. unsigned long page_cache_readahead(struct address_space *mapping,
  835. struct file_ra_state *ra,
  836. struct file *filp,
  837. pgoff_t offset,
  838. unsigned long size);
  839. void handle_ra_miss(struct address_space *mapping,
  840. struct file_ra_state *ra, pgoff_t offset);
  841. unsigned long max_sane_readahead(unsigned long nr);
  842. /* Do stack extension */
  843. extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
  844. #ifdef CONFIG_IA64
  845. extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
  846. #endif
  847. /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
  848. extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
  849. extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
  850. struct vm_area_struct **pprev);
  851. /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
  852. NULL if none. Assume start_addr < end_addr. */
  853. static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
  854. {
  855. struct vm_area_struct * vma = find_vma(mm,start_addr);
  856. if (vma && end_addr <= vma->vm_start)
  857. vma = NULL;
  858. return vma;
  859. }
  860. static inline unsigned long vma_pages(struct vm_area_struct *vma)
  861. {
  862. return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
  863. }
  864. struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
  865. struct page *vmalloc_to_page(void *addr);
  866. unsigned long vmalloc_to_pfn(void *addr);
  867. int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
  868. unsigned long pfn, unsigned long size, pgprot_t);
  869. int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
  870. struct page *follow_page(struct vm_area_struct *, unsigned long address,
  871. unsigned int foll_flags);
  872. #define FOLL_WRITE 0x01 /* check pte is writable */
  873. #define FOLL_TOUCH 0x02 /* mark page accessed */
  874. #define FOLL_GET 0x04 /* do get_page on page */
  875. #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
  876. #ifdef CONFIG_PROC_FS
  877. void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
  878. #else
  879. static inline void vm_stat_account(struct mm_struct *mm,
  880. unsigned long flags, struct file *file, long pages)
  881. {
  882. }
  883. #endif /* CONFIG_PROC_FS */
  884. #ifndef CONFIG_DEBUG_PAGEALLOC
  885. static inline void
  886. kernel_map_pages(struct page *page, int numpages, int enable)
  887. {
  888. if (!PageHighMem(page) && !enable)
  889. mutex_debug_check_no_locks_freed(page_address(page),
  890. numpages * PAGE_SIZE);
  891. }
  892. #endif
  893. extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
  894. #ifdef __HAVE_ARCH_GATE_AREA
  895. int in_gate_area_no_task(unsigned long addr);
  896. int in_gate_area(struct task_struct *task, unsigned long addr);
  897. #else
  898. int in_gate_area_no_task(unsigned long addr);
  899. #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
  900. #endif /* __HAVE_ARCH_GATE_AREA */
  901. /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
  902. #define OOM_DISABLE -17
  903. int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
  904. void __user *, size_t *, loff_t *);
  905. int shrink_slab(unsigned long scanned, gfp_t gfp_mask,
  906. unsigned long lru_pages);
  907. void drop_pagecache(void);
  908. void drop_slab(void);
  909. #ifndef CONFIG_MMU
  910. #define randomize_va_space 0
  911. #else
  912. extern int randomize_va_space;
  913. #endif
  914. #endif /* __KERNEL__ */
  915. #endif /* _LINUX_MM_H */