mm.h 36 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082
  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/gfp.h>
  8. #include <linux/list.h>
  9. #include <linux/mmzone.h>
  10. #include <linux/rbtree.h>
  11. #include <linux/prio_tree.h>
  12. #include <linux/fs.h>
  13. #include <linux/mutex.h>
  14. struct mempolicy;
  15. struct anon_vma;
  16. #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
  17. extern unsigned long max_mapnr;
  18. #endif
  19. extern unsigned long num_physpages;
  20. extern void * high_memory;
  21. extern unsigned long vmalloc_earlyreserve;
  22. extern int page_cluster;
  23. #ifdef CONFIG_SYSCTL
  24. extern int sysctl_legacy_va_layout;
  25. #else
  26. #define sysctl_legacy_va_layout 0
  27. #endif
  28. #include <asm/page.h>
  29. #include <asm/pgtable.h>
  30. #include <asm/processor.h>
  31. #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  32. /*
  33. * Linux kernel virtual memory manager primitives.
  34. * The idea being to have a "virtual" mm in the same way
  35. * we have a virtual fs - giving a cleaner interface to the
  36. * mm details, and allowing different kinds of memory mappings
  37. * (from shared memory to executable loading to arbitrary
  38. * mmap() functions).
  39. */
  40. /*
  41. * This struct defines a memory VMM memory area. There is one of these
  42. * per VM-area/task. A VM area is any part of the process virtual memory
  43. * space that has a special rule for the page-fault handlers (ie a shared
  44. * library, the executable area etc).
  45. */
  46. struct vm_area_struct {
  47. struct mm_struct * vm_mm; /* The address space we belong to. */
  48. unsigned long vm_start; /* Our start address within vm_mm. */
  49. unsigned long vm_end; /* The first byte after our end address
  50. within vm_mm. */
  51. /* linked list of VM areas per task, sorted by address */
  52. struct vm_area_struct *vm_next;
  53. pgprot_t vm_page_prot; /* Access permissions of this VMA. */
  54. unsigned long vm_flags; /* Flags, listed below. */
  55. struct rb_node vm_rb;
  56. /*
  57. * For areas with an address space and backing store,
  58. * linkage into the address_space->i_mmap prio tree, or
  59. * linkage to the list of like vmas hanging off its node, or
  60. * linkage of vma in the address_space->i_mmap_nonlinear list.
  61. */
  62. union {
  63. struct {
  64. struct list_head list;
  65. void *parent; /* aligns with prio_tree_node parent */
  66. struct vm_area_struct *head;
  67. } vm_set;
  68. struct raw_prio_tree_node prio_tree_node;
  69. } shared;
  70. /*
  71. * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
  72. * list, after a COW of one of the file pages. A MAP_SHARED vma
  73. * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
  74. * or brk vma (with NULL file) can only be in an anon_vma list.
  75. */
  76. struct list_head anon_vma_node; /* Serialized by anon_vma->lock */
  77. struct anon_vma *anon_vma; /* Serialized by page_table_lock */
  78. /* Function pointers to deal with this struct. */
  79. struct vm_operations_struct * vm_ops;
  80. /* Information about our backing store: */
  81. unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
  82. units, *not* PAGE_CACHE_SIZE */
  83. struct file * vm_file; /* File we map to (can be NULL). */
  84. void * vm_private_data; /* was vm_pte (shared mem) */
  85. unsigned long vm_truncate_count;/* truncate_count or restart_addr */
  86. #ifndef CONFIG_MMU
  87. atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
  88. #endif
  89. #ifdef CONFIG_NUMA
  90. struct mempolicy *vm_policy; /* NUMA policy for the VMA */
  91. #endif
  92. };
  93. /*
  94. * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
  95. * disabled, then there's a single shared list of VMAs maintained by the
  96. * system, and mm's subscribe to these individually
  97. */
  98. struct vm_list_struct {
  99. struct vm_list_struct *next;
  100. struct vm_area_struct *vma;
  101. };
  102. #ifndef CONFIG_MMU
  103. extern struct rb_root nommu_vma_tree;
  104. extern struct rw_semaphore nommu_vma_sem;
  105. extern unsigned int kobjsize(const void *objp);
  106. #endif
  107. /*
  108. * vm_flags..
  109. */
  110. #define VM_READ 0x00000001 /* currently active flags */
  111. #define VM_WRITE 0x00000002
  112. #define VM_EXEC 0x00000004
  113. #define VM_SHARED 0x00000008
  114. /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  115. #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
  116. #define VM_MAYWRITE 0x00000020
  117. #define VM_MAYEXEC 0x00000040
  118. #define VM_MAYSHARE 0x00000080
  119. #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
  120. #define VM_GROWSUP 0x00000200
  121. #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
  122. #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
  123. #define VM_EXECUTABLE 0x00001000
  124. #define VM_LOCKED 0x00002000
  125. #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
  126. /* Used by sys_madvise() */
  127. #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
  128. #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
  129. #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
  130. #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
  131. #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
  132. #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
  133. #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
  134. #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
  135. #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
  136. #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
  137. #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
  138. #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
  139. #endif
  140. #ifdef CONFIG_STACK_GROWSUP
  141. #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  142. #else
  143. #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  144. #endif
  145. #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
  146. #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
  147. #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
  148. #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
  149. #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
  150. /*
  151. * mapping from the currently active vm_flags protection bits (the
  152. * low four bits) to a page protection mask..
  153. */
  154. extern pgprot_t protection_map[16];
  155. /*
  156. * These are the virtual MM functions - opening of an area, closing and
  157. * unmapping it (needed to keep files on disk up-to-date etc), pointer
  158. * to the functions called when a no-page or a wp-page exception occurs.
  159. */
  160. struct vm_operations_struct {
  161. void (*open)(struct vm_area_struct * area);
  162. void (*close)(struct vm_area_struct * area);
  163. struct page * (*nopage)(struct vm_area_struct * area, unsigned long address, int *type);
  164. int (*populate)(struct vm_area_struct * area, unsigned long address, unsigned long len, pgprot_t prot, unsigned long pgoff, int nonblock);
  165. /* notification that a previously read-only page is about to become
  166. * writable, if an error is returned it will cause a SIGBUS */
  167. int (*page_mkwrite)(struct vm_area_struct *vma, struct page *page);
  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. int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
  173. const nodemask_t *to, unsigned long flags);
  174. #endif
  175. };
  176. struct mmu_gather;
  177. struct inode;
  178. /*
  179. * Each physical page in the system has a struct page associated with
  180. * it to keep track of whatever it is we are using the page for at the
  181. * moment. Note that we have no way to track which tasks are using
  182. * a page.
  183. */
  184. struct page {
  185. unsigned long flags; /* Atomic flags, some possibly
  186. * updated asynchronously */
  187. atomic_t _count; /* Usage count, see below. */
  188. atomic_t _mapcount; /* Count of ptes mapped in mms,
  189. * to show when page is mapped
  190. * & limit reverse map searches.
  191. */
  192. union {
  193. struct {
  194. unsigned long private; /* Mapping-private opaque data:
  195. * usually used for buffer_heads
  196. * if PagePrivate set; used for
  197. * swp_entry_t if PageSwapCache;
  198. * indicates order in the buddy
  199. * system if PG_buddy is set.
  200. */
  201. struct address_space *mapping; /* If low bit clear, points to
  202. * inode address_space, or NULL.
  203. * If page mapped as anonymous
  204. * memory, low bit is set, and
  205. * it points to anon_vma object:
  206. * see PAGE_MAPPING_ANON below.
  207. */
  208. };
  209. #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
  210. spinlock_t ptl;
  211. #endif
  212. };
  213. pgoff_t index; /* Our offset within mapping. */
  214. struct list_head lru; /* Pageout list, eg. active_list
  215. * protected by zone->lru_lock !
  216. */
  217. /*
  218. * On machines where all RAM is mapped into kernel address space,
  219. * we can simply calculate the virtual address. On machines with
  220. * highmem some memory is mapped into kernel virtual memory
  221. * dynamically, so we need a place to store that address.
  222. * Note that this field could be 16 bits on x86 ... ;)
  223. *
  224. * Architectures with slow multiplication can define
  225. * WANT_PAGE_VIRTUAL in asm/page.h
  226. */
  227. #if defined(WANT_PAGE_VIRTUAL)
  228. void *virtual; /* Kernel virtual address (NULL if
  229. not kmapped, ie. highmem) */
  230. #endif /* WANT_PAGE_VIRTUAL */
  231. };
  232. #define page_private(page) ((page)->private)
  233. #define set_page_private(page, v) ((page)->private = (v))
  234. /*
  235. * FIXME: take this include out, include page-flags.h in
  236. * files which need it (119 of them)
  237. */
  238. #include <linux/page-flags.h>
  239. /*
  240. * Methods to modify the page usage count.
  241. *
  242. * What counts for a page usage:
  243. * - cache mapping (page->mapping)
  244. * - private data (page->private)
  245. * - page mapped in a task's page tables, each mapping
  246. * is counted separately
  247. *
  248. * Also, many kernel routines increase the page count before a critical
  249. * routine so they can be sure the page doesn't go away from under them.
  250. */
  251. /*
  252. * Drop a ref, return true if the logical refcount fell to zero (the page has
  253. * no users)
  254. */
  255. static inline int put_page_testzero(struct page *page)
  256. {
  257. BUG_ON(atomic_read(&page->_count) == 0);
  258. return atomic_dec_and_test(&page->_count);
  259. }
  260. /*
  261. * Try to grab a ref unless the page has a refcount of zero, return false if
  262. * that is the case.
  263. */
  264. static inline int get_page_unless_zero(struct page *page)
  265. {
  266. return atomic_inc_not_zero(&page->_count);
  267. }
  268. extern void FASTCALL(__page_cache_release(struct page *));
  269. static inline int page_count(struct page *page)
  270. {
  271. if (unlikely(PageCompound(page)))
  272. page = (struct page *)page_private(page);
  273. return atomic_read(&page->_count);
  274. }
  275. static inline void get_page(struct page *page)
  276. {
  277. if (unlikely(PageCompound(page)))
  278. page = (struct page *)page_private(page);
  279. atomic_inc(&page->_count);
  280. }
  281. /*
  282. * Setup the page count before being freed into the page allocator for
  283. * the first time (boot or memory hotplug)
  284. */
  285. static inline void init_page_count(struct page *page)
  286. {
  287. atomic_set(&page->_count, 1);
  288. }
  289. void put_page(struct page *page);
  290. void split_page(struct page *page, unsigned int order);
  291. /*
  292. * Multiple processes may "see" the same page. E.g. for untouched
  293. * mappings of /dev/null, all processes see the same page full of
  294. * zeroes, and text pages of executables and shared libraries have
  295. * only one copy in memory, at most, normally.
  296. *
  297. * For the non-reserved pages, page_count(page) denotes a reference count.
  298. * page_count() == 0 means the page is free. page->lru is then used for
  299. * freelist management in the buddy allocator.
  300. * page_count() == 1 means the page is used for exactly one purpose
  301. * (e.g. a private data page of one process).
  302. *
  303. * A page may be used for kmalloc() or anyone else who does a
  304. * __get_free_page(). In this case the page_count() is at least 1, and
  305. * all other fields are unused but should be 0 or NULL. The
  306. * management of this page is the responsibility of the one who uses
  307. * it.
  308. *
  309. * The other pages (we may call them "process pages") are completely
  310. * managed by the Linux memory manager: I/O, buffers, swapping etc.
  311. * The following discussion applies only to them.
  312. *
  313. * A page may belong to an inode's memory mapping. In this case,
  314. * page->mapping is the pointer to the inode, and page->index is the
  315. * file offset of the page, in units of PAGE_CACHE_SIZE.
  316. *
  317. * A page contains an opaque `private' member, which belongs to the
  318. * page's address_space. Usually, this is the address of a circular
  319. * list of the page's disk buffers.
  320. *
  321. * For pages belonging to inodes, the page_count() is the number of
  322. * attaches, plus 1 if `private' contains something, plus one for
  323. * the page cache itself.
  324. *
  325. * Instead of keeping dirty/clean pages in per address-space lists, we instead
  326. * now tag pages as dirty/under writeback in the radix tree.
  327. *
  328. * There is also a per-mapping radix tree mapping index to the page
  329. * in memory if present. The tree is rooted at mapping->root.
  330. *
  331. * All process pages can do I/O:
  332. * - inode pages may need to be read from disk,
  333. * - inode pages which have been modified and are MAP_SHARED may need
  334. * to be written to disk,
  335. * - private pages which have been modified may need to be swapped out
  336. * to swap space and (later) to be read back into memory.
  337. */
  338. /*
  339. * The zone field is never updated after free_area_init_core()
  340. * sets it, so none of the operations on it need to be atomic.
  341. */
  342. /*
  343. * page->flags layout:
  344. *
  345. * There are three possibilities for how page->flags get
  346. * laid out. The first is for the normal case, without
  347. * sparsemem. The second is for sparsemem when there is
  348. * plenty of space for node and section. The last is when
  349. * we have run out of space and have to fall back to an
  350. * alternate (slower) way of determining the node.
  351. *
  352. * No sparsemem: | NODE | ZONE | ... | FLAGS |
  353. * with space for node: | SECTION | NODE | ZONE | ... | FLAGS |
  354. * no space for node: | SECTION | ZONE | ... | FLAGS |
  355. */
  356. #ifdef CONFIG_SPARSEMEM
  357. #define SECTIONS_WIDTH SECTIONS_SHIFT
  358. #else
  359. #define SECTIONS_WIDTH 0
  360. #endif
  361. #define ZONES_WIDTH ZONES_SHIFT
  362. #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED
  363. #define NODES_WIDTH NODES_SHIFT
  364. #else
  365. #define NODES_WIDTH 0
  366. #endif
  367. /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
  368. #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
  369. #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
  370. #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
  371. /*
  372. * We are going to use the flags for the page to node mapping if its in
  373. * there. This includes the case where there is no node, so it is implicit.
  374. */
  375. #define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0)
  376. #ifndef PFN_SECTION_SHIFT
  377. #define PFN_SECTION_SHIFT 0
  378. #endif
  379. /*
  380. * Define the bit shifts to access each section. For non-existant
  381. * sections we define the shift as 0; that plus a 0 mask ensures
  382. * the compiler will optimise away reference to them.
  383. */
  384. #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
  385. #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
  386. #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
  387. /* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */
  388. #if FLAGS_HAS_NODE
  389. #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT)
  390. #else
  391. #define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
  392. #endif
  393. #define ZONETABLE_PGSHIFT ZONES_PGSHIFT
  394. #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
  395. #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
  396. #endif
  397. #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
  398. #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
  399. #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
  400. #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1)
  401. static inline unsigned long page_zonenum(struct page *page)
  402. {
  403. return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
  404. }
  405. struct zone;
  406. extern struct zone *zone_table[];
  407. static inline int page_zone_id(struct page *page)
  408. {
  409. return (page->flags >> ZONETABLE_PGSHIFT) & ZONETABLE_MASK;
  410. }
  411. static inline struct zone *page_zone(struct page *page)
  412. {
  413. return zone_table[page_zone_id(page)];
  414. }
  415. static inline unsigned long page_to_nid(struct page *page)
  416. {
  417. if (FLAGS_HAS_NODE)
  418. return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
  419. else
  420. return page_zone(page)->zone_pgdat->node_id;
  421. }
  422. static inline unsigned long page_to_section(struct page *page)
  423. {
  424. return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
  425. }
  426. static inline void set_page_zone(struct page *page, unsigned long zone)
  427. {
  428. page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
  429. page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
  430. }
  431. static inline void set_page_node(struct page *page, unsigned long node)
  432. {
  433. page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
  434. page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
  435. }
  436. static inline void set_page_section(struct page *page, unsigned long section)
  437. {
  438. page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
  439. page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
  440. }
  441. static inline void set_page_links(struct page *page, unsigned long zone,
  442. unsigned long node, unsigned long pfn)
  443. {
  444. set_page_zone(page, zone);
  445. set_page_node(page, node);
  446. set_page_section(page, pfn_to_section_nr(pfn));
  447. }
  448. /*
  449. * Some inline functions in vmstat.h depend on page_zone()
  450. */
  451. #include <linux/vmstat.h>
  452. #ifndef CONFIG_DISCONTIGMEM
  453. /* The array of struct pages - for discontigmem use pgdat->lmem_map */
  454. extern struct page *mem_map;
  455. #endif
  456. static __always_inline void *lowmem_page_address(struct page *page)
  457. {
  458. return __va(page_to_pfn(page) << PAGE_SHIFT);
  459. }
  460. #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
  461. #define HASHED_PAGE_VIRTUAL
  462. #endif
  463. #if defined(WANT_PAGE_VIRTUAL)
  464. #define page_address(page) ((page)->virtual)
  465. #define set_page_address(page, address) \
  466. do { \
  467. (page)->virtual = (address); \
  468. } while(0)
  469. #define page_address_init() do { } while(0)
  470. #endif
  471. #if defined(HASHED_PAGE_VIRTUAL)
  472. void *page_address(struct page *page);
  473. void set_page_address(struct page *page, void *virtual);
  474. void page_address_init(void);
  475. #endif
  476. #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
  477. #define page_address(page) lowmem_page_address(page)
  478. #define set_page_address(page, address) do { } while(0)
  479. #define page_address_init() do { } while(0)
  480. #endif
  481. /*
  482. * On an anonymous page mapped into a user virtual memory area,
  483. * page->mapping points to its anon_vma, not to a struct address_space;
  484. * with the PAGE_MAPPING_ANON bit set to distinguish it.
  485. *
  486. * Please note that, confusingly, "page_mapping" refers to the inode
  487. * address_space which maps the page from disk; whereas "page_mapped"
  488. * refers to user virtual address space into which the page is mapped.
  489. */
  490. #define PAGE_MAPPING_ANON 1
  491. extern struct address_space swapper_space;
  492. static inline struct address_space *page_mapping(struct page *page)
  493. {
  494. struct address_space *mapping = page->mapping;
  495. if (unlikely(PageSwapCache(page)))
  496. mapping = &swapper_space;
  497. else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
  498. mapping = NULL;
  499. return mapping;
  500. }
  501. static inline int PageAnon(struct page *page)
  502. {
  503. return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
  504. }
  505. /*
  506. * Return the pagecache index of the passed page. Regular pagecache pages
  507. * use ->index whereas swapcache pages use ->private
  508. */
  509. static inline pgoff_t page_index(struct page *page)
  510. {
  511. if (unlikely(PageSwapCache(page)))
  512. return page_private(page);
  513. return page->index;
  514. }
  515. /*
  516. * The atomic page->_mapcount, like _count, starts from -1:
  517. * so that transitions both from it and to it can be tracked,
  518. * using atomic_inc_and_test and atomic_add_negative(-1).
  519. */
  520. static inline void reset_page_mapcount(struct page *page)
  521. {
  522. atomic_set(&(page)->_mapcount, -1);
  523. }
  524. static inline int page_mapcount(struct page *page)
  525. {
  526. return atomic_read(&(page)->_mapcount) + 1;
  527. }
  528. /*
  529. * Return true if this page is mapped into pagetables.
  530. */
  531. static inline int page_mapped(struct page *page)
  532. {
  533. return atomic_read(&(page)->_mapcount) >= 0;
  534. }
  535. /*
  536. * Error return values for the *_nopage functions
  537. */
  538. #define NOPAGE_SIGBUS (NULL)
  539. #define NOPAGE_OOM ((struct page *) (-1))
  540. /*
  541. * Different kinds of faults, as returned by handle_mm_fault().
  542. * Used to decide whether a process gets delivered SIGBUS or
  543. * just gets major/minor fault counters bumped up.
  544. */
  545. #define VM_FAULT_OOM 0x00
  546. #define VM_FAULT_SIGBUS 0x01
  547. #define VM_FAULT_MINOR 0x02
  548. #define VM_FAULT_MAJOR 0x03
  549. /*
  550. * Special case for get_user_pages.
  551. * Must be in a distinct bit from the above VM_FAULT_ flags.
  552. */
  553. #define VM_FAULT_WRITE 0x10
  554. #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
  555. extern void show_free_areas(void);
  556. #ifdef CONFIG_SHMEM
  557. struct page *shmem_nopage(struct vm_area_struct *vma,
  558. unsigned long address, int *type);
  559. int shmem_set_policy(struct vm_area_struct *vma, struct mempolicy *new);
  560. struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
  561. unsigned long addr);
  562. int shmem_lock(struct file *file, int lock, struct user_struct *user);
  563. #else
  564. #define shmem_nopage filemap_nopage
  565. static inline int shmem_lock(struct file *file, int lock,
  566. struct user_struct *user)
  567. {
  568. return 0;
  569. }
  570. static inline int shmem_set_policy(struct vm_area_struct *vma,
  571. struct mempolicy *new)
  572. {
  573. return 0;
  574. }
  575. static inline struct mempolicy *shmem_get_policy(struct vm_area_struct *vma,
  576. unsigned long addr)
  577. {
  578. return NULL;
  579. }
  580. #endif
  581. struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
  582. extern int shmem_mmap(struct file *file, struct vm_area_struct *vma);
  583. int shmem_zero_setup(struct vm_area_struct *);
  584. #ifndef CONFIG_MMU
  585. extern unsigned long shmem_get_unmapped_area(struct file *file,
  586. unsigned long addr,
  587. unsigned long len,
  588. unsigned long pgoff,
  589. unsigned long flags);
  590. #endif
  591. static inline int can_do_mlock(void)
  592. {
  593. if (capable(CAP_IPC_LOCK))
  594. return 1;
  595. if (current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur != 0)
  596. return 1;
  597. return 0;
  598. }
  599. extern int user_shm_lock(size_t, struct user_struct *);
  600. extern void user_shm_unlock(size_t, struct user_struct *);
  601. /*
  602. * Parameter block passed down to zap_pte_range in exceptional cases.
  603. */
  604. struct zap_details {
  605. struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
  606. struct address_space *check_mapping; /* Check page->mapping if set */
  607. pgoff_t first_index; /* Lowest page->index to unmap */
  608. pgoff_t last_index; /* Highest page->index to unmap */
  609. spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */
  610. unsigned long truncate_count; /* Compare vm_truncate_count */
  611. };
  612. struct page *vm_normal_page(struct vm_area_struct *, unsigned long, pte_t);
  613. unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
  614. unsigned long size, struct zap_details *);
  615. unsigned long unmap_vmas(struct mmu_gather **tlb,
  616. struct vm_area_struct *start_vma, unsigned long start_addr,
  617. unsigned long end_addr, unsigned long *nr_accounted,
  618. struct zap_details *);
  619. void free_pgd_range(struct mmu_gather **tlb, unsigned long addr,
  620. unsigned long end, unsigned long floor, unsigned long ceiling);
  621. void free_pgtables(struct mmu_gather **tlb, struct vm_area_struct *start_vma,
  622. unsigned long floor, unsigned long ceiling);
  623. int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
  624. struct vm_area_struct *vma);
  625. int zeromap_page_range(struct vm_area_struct *vma, unsigned long from,
  626. unsigned long size, pgprot_t prot);
  627. void unmap_mapping_range(struct address_space *mapping,
  628. loff_t const holebegin, loff_t const holelen, int even_cows);
  629. static inline void unmap_shared_mapping_range(struct address_space *mapping,
  630. loff_t const holebegin, loff_t const holelen)
  631. {
  632. unmap_mapping_range(mapping, holebegin, holelen, 0);
  633. }
  634. extern int vmtruncate(struct inode * inode, loff_t offset);
  635. extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
  636. extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot);
  637. extern int install_file_pte(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, unsigned long pgoff, pgprot_t prot);
  638. #ifdef CONFIG_MMU
  639. extern int __handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma,
  640. unsigned long address, int write_access);
  641. static inline int handle_mm_fault(struct mm_struct *mm,
  642. struct vm_area_struct *vma, unsigned long address,
  643. int write_access)
  644. {
  645. return __handle_mm_fault(mm, vma, address, write_access) &
  646. (~VM_FAULT_WRITE);
  647. }
  648. #else
  649. static inline int handle_mm_fault(struct mm_struct *mm,
  650. struct vm_area_struct *vma, unsigned long address,
  651. int write_access)
  652. {
  653. /* should never happen if there's no MMU */
  654. BUG();
  655. return VM_FAULT_SIGBUS;
  656. }
  657. #endif
  658. extern int make_pages_present(unsigned long addr, unsigned long end);
  659. extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
  660. void install_arg_page(struct vm_area_struct *, struct page *, unsigned long);
  661. int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
  662. int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
  663. void print_bad_pte(struct vm_area_struct *, pte_t, unsigned long);
  664. int __set_page_dirty_buffers(struct page *page);
  665. int __set_page_dirty_nobuffers(struct page *page);
  666. int redirty_page_for_writepage(struct writeback_control *wbc,
  667. struct page *page);
  668. int FASTCALL(set_page_dirty(struct page *page));
  669. int set_page_dirty_lock(struct page *page);
  670. int clear_page_dirty_for_io(struct page *page);
  671. extern unsigned long do_mremap(unsigned long addr,
  672. unsigned long old_len, unsigned long new_len,
  673. unsigned long flags, unsigned long new_addr);
  674. /*
  675. * Prototype to add a shrinker callback for ageable caches.
  676. *
  677. * These functions are passed a count `nr_to_scan' and a gfpmask. They should
  678. * scan `nr_to_scan' objects, attempting to free them.
  679. *
  680. * The callback must return the number of objects which remain in the cache.
  681. *
  682. * The callback will be passed nr_to_scan == 0 when the VM is querying the
  683. * cache size, so a fastpath for that case is appropriate.
  684. */
  685. typedef int (*shrinker_t)(int nr_to_scan, gfp_t gfp_mask);
  686. /*
  687. * Add an aging callback. The int is the number of 'seeks' it takes
  688. * to recreate one of the objects that these functions age.
  689. */
  690. #define DEFAULT_SEEKS 2
  691. struct shrinker;
  692. extern struct shrinker *set_shrinker(int, shrinker_t);
  693. extern void remove_shrinker(struct shrinker *shrinker);
  694. extern pte_t *FASTCALL(get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl));
  695. int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
  696. int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
  697. int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
  698. int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
  699. /*
  700. * The following ifdef needed to get the 4level-fixup.h header to work.
  701. * Remove it when 4level-fixup.h has been removed.
  702. */
  703. #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
  704. static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
  705. {
  706. return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
  707. NULL: pud_offset(pgd, address);
  708. }
  709. static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
  710. {
  711. return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
  712. NULL: pmd_offset(pud, address);
  713. }
  714. #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
  715. #if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
  716. /*
  717. * We tuck a spinlock to guard each pagetable page into its struct page,
  718. * at page->private, with BUILD_BUG_ON to make sure that this will not
  719. * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
  720. * When freeing, reset page->mapping so free_pages_check won't complain.
  721. */
  722. #define __pte_lockptr(page) &((page)->ptl)
  723. #define pte_lock_init(_page) do { \
  724. spin_lock_init(__pte_lockptr(_page)); \
  725. } while (0)
  726. #define pte_lock_deinit(page) ((page)->mapping = NULL)
  727. #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
  728. #else
  729. /*
  730. * We use mm->page_table_lock to guard all pagetable pages of the mm.
  731. */
  732. #define pte_lock_init(page) do {} while (0)
  733. #define pte_lock_deinit(page) do {} while (0)
  734. #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
  735. #endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
  736. #define pte_offset_map_lock(mm, pmd, address, ptlp) \
  737. ({ \
  738. spinlock_t *__ptl = pte_lockptr(mm, pmd); \
  739. pte_t *__pte = pte_offset_map(pmd, address); \
  740. *(ptlp) = __ptl; \
  741. spin_lock(__ptl); \
  742. __pte; \
  743. })
  744. #define pte_unmap_unlock(pte, ptl) do { \
  745. spin_unlock(ptl); \
  746. pte_unmap(pte); \
  747. } while (0)
  748. #define pte_alloc_map(mm, pmd, address) \
  749. ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
  750. NULL: pte_offset_map(pmd, address))
  751. #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
  752. ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
  753. NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
  754. #define pte_alloc_kernel(pmd, address) \
  755. ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
  756. NULL: pte_offset_kernel(pmd, address))
  757. extern void free_area_init(unsigned long * zones_size);
  758. extern void free_area_init_node(int nid, pg_data_t *pgdat,
  759. unsigned long * zones_size, unsigned long zone_start_pfn,
  760. unsigned long *zholes_size);
  761. extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long);
  762. extern void setup_per_zone_pages_min(void);
  763. extern void mem_init(void);
  764. extern void show_mem(void);
  765. extern void si_meminfo(struct sysinfo * val);
  766. extern void si_meminfo_node(struct sysinfo *val, int nid);
  767. #ifdef CONFIG_NUMA
  768. extern void setup_per_cpu_pageset(void);
  769. #else
  770. static inline void setup_per_cpu_pageset(void) {}
  771. #endif
  772. /* prio_tree.c */
  773. void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
  774. void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
  775. void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
  776. struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
  777. struct prio_tree_iter *iter);
  778. #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
  779. for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
  780. (vma = vma_prio_tree_next(vma, iter)); )
  781. static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
  782. struct list_head *list)
  783. {
  784. vma->shared.vm_set.parent = NULL;
  785. list_add_tail(&vma->shared.vm_set.list, list);
  786. }
  787. /* mmap.c */
  788. extern int __vm_enough_memory(long pages, int cap_sys_admin);
  789. extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
  790. unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
  791. extern struct vm_area_struct *vma_merge(struct mm_struct *,
  792. struct vm_area_struct *prev, unsigned long addr, unsigned long end,
  793. unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
  794. struct mempolicy *);
  795. extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
  796. extern int split_vma(struct mm_struct *,
  797. struct vm_area_struct *, unsigned long addr, int new_below);
  798. extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
  799. extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
  800. struct rb_node **, struct rb_node *);
  801. extern void unlink_file_vma(struct vm_area_struct *);
  802. extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
  803. unsigned long addr, unsigned long len, pgoff_t pgoff);
  804. extern void exit_mmap(struct mm_struct *);
  805. extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
  806. extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
  807. extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
  808. unsigned long len, unsigned long prot,
  809. unsigned long flag, unsigned long pgoff);
  810. static inline unsigned long do_mmap(struct file *file, unsigned long addr,
  811. unsigned long len, unsigned long prot,
  812. unsigned long flag, unsigned long offset)
  813. {
  814. unsigned long ret = -EINVAL;
  815. if ((offset + PAGE_ALIGN(len)) < offset)
  816. goto out;
  817. if (!(offset & ~PAGE_MASK))
  818. ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
  819. out:
  820. return ret;
  821. }
  822. extern int do_munmap(struct mm_struct *, unsigned long, size_t);
  823. extern unsigned long do_brk(unsigned long, unsigned long);
  824. /* filemap.c */
  825. extern unsigned long page_unuse(struct page *);
  826. extern void truncate_inode_pages(struct address_space *, loff_t);
  827. extern void truncate_inode_pages_range(struct address_space *,
  828. loff_t lstart, loff_t lend);
  829. /* generic vm_area_ops exported for stackable file systems */
  830. extern struct page *filemap_nopage(struct vm_area_struct *, unsigned long, int *);
  831. extern int filemap_populate(struct vm_area_struct *, unsigned long,
  832. unsigned long, pgprot_t, unsigned long, int);
  833. /* mm/page-writeback.c */
  834. int write_one_page(struct page *page, int wait);
  835. /* readahead.c */
  836. #define VM_MAX_READAHEAD 128 /* kbytes */
  837. #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
  838. #define VM_MAX_CACHE_HIT 256 /* max pages in a row in cache before
  839. * turning readahead off */
  840. int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
  841. pgoff_t offset, unsigned long nr_to_read);
  842. int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
  843. pgoff_t offset, unsigned long nr_to_read);
  844. unsigned long page_cache_readahead(struct address_space *mapping,
  845. struct file_ra_state *ra,
  846. struct file *filp,
  847. pgoff_t offset,
  848. unsigned long size);
  849. void handle_ra_miss(struct address_space *mapping,
  850. struct file_ra_state *ra, pgoff_t offset);
  851. unsigned long max_sane_readahead(unsigned long nr);
  852. /* Do stack extension */
  853. extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
  854. #ifdef CONFIG_IA64
  855. extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
  856. #endif
  857. /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
  858. extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
  859. extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
  860. struct vm_area_struct **pprev);
  861. /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
  862. NULL if none. Assume start_addr < end_addr. */
  863. static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
  864. {
  865. struct vm_area_struct * vma = find_vma(mm,start_addr);
  866. if (vma && end_addr <= vma->vm_start)
  867. vma = NULL;
  868. return vma;
  869. }
  870. static inline unsigned long vma_pages(struct vm_area_struct *vma)
  871. {
  872. return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
  873. }
  874. struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
  875. struct page *vmalloc_to_page(void *addr);
  876. unsigned long vmalloc_to_pfn(void *addr);
  877. int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
  878. unsigned long pfn, unsigned long size, pgprot_t);
  879. int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
  880. struct page *follow_page(struct vm_area_struct *, unsigned long address,
  881. unsigned int foll_flags);
  882. #define FOLL_WRITE 0x01 /* check pte is writable */
  883. #define FOLL_TOUCH 0x02 /* mark page accessed */
  884. #define FOLL_GET 0x04 /* do get_page on page */
  885. #define FOLL_ANON 0x08 /* give ZERO_PAGE if no pgtable */
  886. #ifdef CONFIG_PROC_FS
  887. void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
  888. #else
  889. static inline void vm_stat_account(struct mm_struct *mm,
  890. unsigned long flags, struct file *file, long pages)
  891. {
  892. }
  893. #endif /* CONFIG_PROC_FS */
  894. static inline void
  895. debug_check_no_locks_freed(const void *from, unsigned long len)
  896. {
  897. mutex_debug_check_no_locks_freed(from, len);
  898. rt_mutex_debug_check_no_locks_freed(from, len);
  899. }
  900. #ifndef CONFIG_DEBUG_PAGEALLOC
  901. static inline void
  902. kernel_map_pages(struct page *page, int numpages, int enable)
  903. {
  904. if (!PageHighMem(page) && !enable)
  905. debug_check_no_locks_freed(page_address(page),
  906. numpages * PAGE_SIZE);
  907. }
  908. #endif
  909. extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
  910. #ifdef __HAVE_ARCH_GATE_AREA
  911. int in_gate_area_no_task(unsigned long addr);
  912. int in_gate_area(struct task_struct *task, unsigned long addr);
  913. #else
  914. int in_gate_area_no_task(unsigned long addr);
  915. #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
  916. #endif /* __HAVE_ARCH_GATE_AREA */
  917. /* /proc/<pid>/oom_adj set to -17 protects from the oom-killer */
  918. #define OOM_DISABLE -17
  919. int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
  920. void __user *, size_t *, loff_t *);
  921. unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
  922. unsigned long lru_pages);
  923. void drop_pagecache(void);
  924. void drop_slab(void);
  925. #ifndef CONFIG_MMU
  926. #define randomize_va_space 0
  927. #else
  928. extern int randomize_va_space;
  929. #endif
  930. const char *arch_vma_name(struct vm_area_struct *vma);
  931. #endif /* __KERNEL__ */
  932. #endif /* _LINUX_MM_H */