mm.h 65 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048
  1. #ifndef _LINUX_MM_H
  2. #define _LINUX_MM_H
  3. #include <linux/errno.h>
  4. #ifdef __KERNEL__
  5. #include <linux/gfp.h>
  6. #include <linux/bug.h>
  7. #include <linux/list.h>
  8. #include <linux/mmzone.h>
  9. #include <linux/rbtree.h>
  10. #include <linux/atomic.h>
  11. #include <linux/debug_locks.h>
  12. #include <linux/mm_types.h>
  13. #include <linux/range.h>
  14. #include <linux/pfn.h>
  15. #include <linux/bit_spinlock.h>
  16. #include <linux/shrinker.h>
  17. struct mempolicy;
  18. struct anon_vma;
  19. struct anon_vma_chain;
  20. struct file_ra_state;
  21. struct user_struct;
  22. struct writeback_control;
  23. #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
  24. extern unsigned long max_mapnr;
  25. static inline void set_max_mapnr(unsigned long limit)
  26. {
  27. max_mapnr = limit;
  28. }
  29. #else
  30. static inline void set_max_mapnr(unsigned long limit) { }
  31. #endif
  32. extern unsigned long totalram_pages;
  33. extern void * high_memory;
  34. extern int page_cluster;
  35. #ifdef CONFIG_SYSCTL
  36. extern int sysctl_legacy_va_layout;
  37. #else
  38. #define sysctl_legacy_va_layout 0
  39. #endif
  40. #include <asm/page.h>
  41. #include <asm/pgtable.h>
  42. #include <asm/processor.h>
  43. #ifndef __pa_symbol
  44. #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
  45. #endif
  46. extern unsigned long sysctl_user_reserve_kbytes;
  47. extern unsigned long sysctl_admin_reserve_kbytes;
  48. #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  49. /* to align the pointer to the (next) page boundary */
  50. #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  51. /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
  52. #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
  53. /*
  54. * Linux kernel virtual memory manager primitives.
  55. * The idea being to have a "virtual" mm in the same way
  56. * we have a virtual fs - giving a cleaner interface to the
  57. * mm details, and allowing different kinds of memory mappings
  58. * (from shared memory to executable loading to arbitrary
  59. * mmap() functions).
  60. */
  61. extern struct kmem_cache *vm_area_cachep;
  62. #ifndef CONFIG_MMU
  63. extern struct rb_root nommu_region_tree;
  64. extern struct rw_semaphore nommu_region_sem;
  65. extern unsigned int kobjsize(const void *objp);
  66. #endif
  67. /*
  68. * vm_flags in vm_area_struct, see mm_types.h.
  69. */
  70. #define VM_NONE 0x00000000
  71. #define VM_READ 0x00000001 /* currently active flags */
  72. #define VM_WRITE 0x00000002
  73. #define VM_EXEC 0x00000004
  74. #define VM_SHARED 0x00000008
  75. /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  76. #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
  77. #define VM_MAYWRITE 0x00000020
  78. #define VM_MAYEXEC 0x00000040
  79. #define VM_MAYSHARE 0x00000080
  80. #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
  81. #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
  82. #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
  83. #define VM_LOCKED 0x00002000
  84. #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
  85. /* Used by sys_madvise() */
  86. #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
  87. #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
  88. #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
  89. #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
  90. #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
  91. #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
  92. #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
  93. #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
  94. #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
  95. #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
  96. #ifdef CONFIG_MEM_SOFT_DIRTY
  97. # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
  98. #else
  99. # define VM_SOFTDIRTY 0
  100. #endif
  101. #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
  102. #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
  103. #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
  104. #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
  105. #if defined(CONFIG_X86)
  106. # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
  107. #elif defined(CONFIG_PPC)
  108. # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
  109. #elif defined(CONFIG_PARISC)
  110. # define VM_GROWSUP VM_ARCH_1
  111. #elif defined(CONFIG_METAG)
  112. # define VM_GROWSUP VM_ARCH_1
  113. #elif defined(CONFIG_IA64)
  114. # define VM_GROWSUP VM_ARCH_1
  115. #elif !defined(CONFIG_MMU)
  116. # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
  117. #endif
  118. #ifndef VM_GROWSUP
  119. # define VM_GROWSUP VM_NONE
  120. #endif
  121. /* Bits set in the VMA until the stack is in its final location */
  122. #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
  123. #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
  124. #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
  125. #endif
  126. #ifdef CONFIG_STACK_GROWSUP
  127. #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  128. #else
  129. #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  130. #endif
  131. /*
  132. * Special vmas that are non-mergable, non-mlock()able.
  133. * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
  134. */
  135. #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
  136. /*
  137. * mapping from the currently active vm_flags protection bits (the
  138. * low four bits) to a page protection mask..
  139. */
  140. extern pgprot_t protection_map[16];
  141. #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
  142. #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
  143. #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
  144. #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
  145. #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
  146. #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
  147. #define FAULT_FLAG_TRIED 0x40 /* second try */
  148. #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
  149. /*
  150. * vm_fault is filled by the the pagefault handler and passed to the vma's
  151. * ->fault function. The vma's ->fault is responsible for returning a bitmask
  152. * of VM_FAULT_xxx flags that give details about how the fault was handled.
  153. *
  154. * pgoff should be used in favour of virtual_address, if possible. If pgoff
  155. * is used, one may implement ->remap_pages to get nonlinear mapping support.
  156. */
  157. struct vm_fault {
  158. unsigned int flags; /* FAULT_FLAG_xxx flags */
  159. pgoff_t pgoff; /* Logical page offset based on vma */
  160. void __user *virtual_address; /* Faulting virtual address */
  161. struct page *page; /* ->fault handlers should return a
  162. * page here, unless VM_FAULT_NOPAGE
  163. * is set (which is also implied by
  164. * VM_FAULT_ERROR).
  165. */
  166. };
  167. /*
  168. * These are the virtual MM functions - opening of an area, closing and
  169. * unmapping it (needed to keep files on disk up-to-date etc), pointer
  170. * to the functions called when a no-page or a wp-page exception occurs.
  171. */
  172. struct vm_operations_struct {
  173. void (*open)(struct vm_area_struct * area);
  174. void (*close)(struct vm_area_struct * area);
  175. int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
  176. /* notification that a previously read-only page is about to become
  177. * writable, if an error is returned it will cause a SIGBUS */
  178. int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
  179. /* called by access_process_vm when get_user_pages() fails, typically
  180. * for use by special VMAs that can switch between memory and hardware
  181. */
  182. int (*access)(struct vm_area_struct *vma, unsigned long addr,
  183. void *buf, int len, int write);
  184. #ifdef CONFIG_NUMA
  185. /*
  186. * set_policy() op must add a reference to any non-NULL @new mempolicy
  187. * to hold the policy upon return. Caller should pass NULL @new to
  188. * remove a policy and fall back to surrounding context--i.e. do not
  189. * install a MPOL_DEFAULT policy, nor the task or system default
  190. * mempolicy.
  191. */
  192. int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
  193. /*
  194. * get_policy() op must add reference [mpol_get()] to any policy at
  195. * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
  196. * in mm/mempolicy.c will do this automatically.
  197. * get_policy() must NOT add a ref if the policy at (vma,addr) is not
  198. * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
  199. * If no [shared/vma] mempolicy exists at the addr, get_policy() op
  200. * must return NULL--i.e., do not "fallback" to task or system default
  201. * policy.
  202. */
  203. struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
  204. unsigned long addr);
  205. int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
  206. const nodemask_t *to, unsigned long flags);
  207. #endif
  208. /* called by sys_remap_file_pages() to populate non-linear mapping */
  209. int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
  210. unsigned long size, pgoff_t pgoff);
  211. };
  212. struct mmu_gather;
  213. struct inode;
  214. #define page_private(page) ((page)->private)
  215. #define set_page_private(page, v) ((page)->private = (v))
  216. /* It's valid only if the page is free path or free_list */
  217. static inline void set_freepage_migratetype(struct page *page, int migratetype)
  218. {
  219. page->index = migratetype;
  220. }
  221. /* It's valid only if the page is free path or free_list */
  222. static inline int get_freepage_migratetype(struct page *page)
  223. {
  224. return page->index;
  225. }
  226. /*
  227. * FIXME: take this include out, include page-flags.h in
  228. * files which need it (119 of them)
  229. */
  230. #include <linux/page-flags.h>
  231. #include <linux/huge_mm.h>
  232. /*
  233. * Methods to modify the page usage count.
  234. *
  235. * What counts for a page usage:
  236. * - cache mapping (page->mapping)
  237. * - private data (page->private)
  238. * - page mapped in a task's page tables, each mapping
  239. * is counted separately
  240. *
  241. * Also, many kernel routines increase the page count before a critical
  242. * routine so they can be sure the page doesn't go away from under them.
  243. */
  244. /*
  245. * Drop a ref, return true if the refcount fell to zero (the page has no users)
  246. */
  247. static inline int put_page_testzero(struct page *page)
  248. {
  249. VM_BUG_ON(atomic_read(&page->_count) == 0);
  250. return atomic_dec_and_test(&page->_count);
  251. }
  252. /*
  253. * Try to grab a ref unless the page has a refcount of zero, return false if
  254. * that is the case.
  255. * This can be called when MMU is off so it must not access
  256. * any of the virtual mappings.
  257. */
  258. static inline int get_page_unless_zero(struct page *page)
  259. {
  260. return atomic_inc_not_zero(&page->_count);
  261. }
  262. /*
  263. * Try to drop a ref unless the page has a refcount of one, return false if
  264. * that is the case.
  265. * This is to make sure that the refcount won't become zero after this drop.
  266. * This can be called when MMU is off so it must not access
  267. * any of the virtual mappings.
  268. */
  269. static inline int put_page_unless_one(struct page *page)
  270. {
  271. return atomic_add_unless(&page->_count, -1, 1);
  272. }
  273. extern int page_is_ram(unsigned long pfn);
  274. /* Support for virtually mapped pages */
  275. struct page *vmalloc_to_page(const void *addr);
  276. unsigned long vmalloc_to_pfn(const void *addr);
  277. /*
  278. * Determine if an address is within the vmalloc range
  279. *
  280. * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
  281. * is no special casing required.
  282. */
  283. static inline int is_vmalloc_addr(const void *x)
  284. {
  285. #ifdef CONFIG_MMU
  286. unsigned long addr = (unsigned long)x;
  287. return addr >= VMALLOC_START && addr < VMALLOC_END;
  288. #else
  289. return 0;
  290. #endif
  291. }
  292. #ifdef CONFIG_MMU
  293. extern int is_vmalloc_or_module_addr(const void *x);
  294. #else
  295. static inline int is_vmalloc_or_module_addr(const void *x)
  296. {
  297. return 0;
  298. }
  299. #endif
  300. static inline void compound_lock(struct page *page)
  301. {
  302. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  303. VM_BUG_ON(PageSlab(page));
  304. bit_spin_lock(PG_compound_lock, &page->flags);
  305. #endif
  306. }
  307. static inline void compound_unlock(struct page *page)
  308. {
  309. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  310. VM_BUG_ON(PageSlab(page));
  311. bit_spin_unlock(PG_compound_lock, &page->flags);
  312. #endif
  313. }
  314. static inline unsigned long compound_lock_irqsave(struct page *page)
  315. {
  316. unsigned long uninitialized_var(flags);
  317. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  318. local_irq_save(flags);
  319. compound_lock(page);
  320. #endif
  321. return flags;
  322. }
  323. static inline void compound_unlock_irqrestore(struct page *page,
  324. unsigned long flags)
  325. {
  326. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  327. compound_unlock(page);
  328. local_irq_restore(flags);
  329. #endif
  330. }
  331. static inline struct page *compound_head(struct page *page)
  332. {
  333. if (unlikely(PageTail(page)))
  334. return page->first_page;
  335. return page;
  336. }
  337. /*
  338. * The atomic page->_mapcount, starts from -1: so that transitions
  339. * both from it and to it can be tracked, using atomic_inc_and_test
  340. * and atomic_add_negative(-1).
  341. */
  342. static inline void page_mapcount_reset(struct page *page)
  343. {
  344. atomic_set(&(page)->_mapcount, -1);
  345. }
  346. static inline int page_mapcount(struct page *page)
  347. {
  348. return atomic_read(&(page)->_mapcount) + 1;
  349. }
  350. static inline int page_count(struct page *page)
  351. {
  352. return atomic_read(&compound_head(page)->_count);
  353. }
  354. #ifdef CONFIG_HUGETLB_PAGE
  355. extern int PageHeadHuge(struct page *page_head);
  356. #else /* CONFIG_HUGETLB_PAGE */
  357. static inline int PageHeadHuge(struct page *page_head)
  358. {
  359. return 0;
  360. }
  361. #endif /* CONFIG_HUGETLB_PAGE */
  362. static inline bool __compound_tail_refcounted(struct page *page)
  363. {
  364. return !PageSlab(page) && !PageHeadHuge(page);
  365. }
  366. /*
  367. * This takes a head page as parameter and tells if the
  368. * tail page reference counting can be skipped.
  369. *
  370. * For this to be safe, PageSlab and PageHeadHuge must remain true on
  371. * any given page where they return true here, until all tail pins
  372. * have been released.
  373. */
  374. static inline bool compound_tail_refcounted(struct page *page)
  375. {
  376. VM_BUG_ON(!PageHead(page));
  377. return __compound_tail_refcounted(page);
  378. }
  379. static inline void get_huge_page_tail(struct page *page)
  380. {
  381. /*
  382. * __split_huge_page_refcount() cannot run
  383. * from under us.
  384. * In turn no need of compound_trans_head here.
  385. */
  386. VM_BUG_ON(page_mapcount(page) < 0);
  387. VM_BUG_ON(atomic_read(&page->_count) != 0);
  388. if (compound_tail_refcounted(compound_head(page)))
  389. atomic_inc(&page->_mapcount);
  390. }
  391. extern bool __get_page_tail(struct page *page);
  392. static inline void get_page(struct page *page)
  393. {
  394. if (unlikely(PageTail(page)))
  395. if (likely(__get_page_tail(page)))
  396. return;
  397. /*
  398. * Getting a normal page or the head of a compound page
  399. * requires to already have an elevated page->_count.
  400. */
  401. VM_BUG_ON(atomic_read(&page->_count) <= 0);
  402. atomic_inc(&page->_count);
  403. }
  404. static inline struct page *virt_to_head_page(const void *x)
  405. {
  406. struct page *page = virt_to_page(x);
  407. return compound_head(page);
  408. }
  409. /*
  410. * Setup the page count before being freed into the page allocator for
  411. * the first time (boot or memory hotplug)
  412. */
  413. static inline void init_page_count(struct page *page)
  414. {
  415. atomic_set(&page->_count, 1);
  416. }
  417. /*
  418. * PageBuddy() indicate that the page is free and in the buddy system
  419. * (see mm/page_alloc.c).
  420. *
  421. * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
  422. * -2 so that an underflow of the page_mapcount() won't be mistaken
  423. * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
  424. * efficiently by most CPU architectures.
  425. */
  426. #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
  427. static inline int PageBuddy(struct page *page)
  428. {
  429. return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
  430. }
  431. static inline void __SetPageBuddy(struct page *page)
  432. {
  433. VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
  434. atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
  435. }
  436. static inline void __ClearPageBuddy(struct page *page)
  437. {
  438. VM_BUG_ON(!PageBuddy(page));
  439. atomic_set(&page->_mapcount, -1);
  440. }
  441. void put_page(struct page *page);
  442. void put_pages_list(struct list_head *pages);
  443. void split_page(struct page *page, unsigned int order);
  444. int split_free_page(struct page *page);
  445. /*
  446. * Compound pages have a destructor function. Provide a
  447. * prototype for that function and accessor functions.
  448. * These are _only_ valid on the head of a PG_compound page.
  449. */
  450. typedef void compound_page_dtor(struct page *);
  451. static inline void set_compound_page_dtor(struct page *page,
  452. compound_page_dtor *dtor)
  453. {
  454. page[1].lru.next = (void *)dtor;
  455. }
  456. static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
  457. {
  458. return (compound_page_dtor *)page[1].lru.next;
  459. }
  460. static inline int compound_order(struct page *page)
  461. {
  462. if (!PageHead(page))
  463. return 0;
  464. return (unsigned long)page[1].lru.prev;
  465. }
  466. static inline void set_compound_order(struct page *page, unsigned long order)
  467. {
  468. page[1].lru.prev = (void *)order;
  469. }
  470. #ifdef CONFIG_MMU
  471. /*
  472. * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
  473. * servicing faults for write access. In the normal case, do always want
  474. * pte_mkwrite. But get_user_pages can cause write faults for mappings
  475. * that do not have writing enabled, when used by access_process_vm.
  476. */
  477. static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
  478. {
  479. if (likely(vma->vm_flags & VM_WRITE))
  480. pte = pte_mkwrite(pte);
  481. return pte;
  482. }
  483. #endif
  484. /*
  485. * Multiple processes may "see" the same page. E.g. for untouched
  486. * mappings of /dev/null, all processes see the same page full of
  487. * zeroes, and text pages of executables and shared libraries have
  488. * only one copy in memory, at most, normally.
  489. *
  490. * For the non-reserved pages, page_count(page) denotes a reference count.
  491. * page_count() == 0 means the page is free. page->lru is then used for
  492. * freelist management in the buddy allocator.
  493. * page_count() > 0 means the page has been allocated.
  494. *
  495. * Pages are allocated by the slab allocator in order to provide memory
  496. * to kmalloc and kmem_cache_alloc. In this case, the management of the
  497. * page, and the fields in 'struct page' are the responsibility of mm/slab.c
  498. * unless a particular usage is carefully commented. (the responsibility of
  499. * freeing the kmalloc memory is the caller's, of course).
  500. *
  501. * A page may be used by anyone else who does a __get_free_page().
  502. * In this case, page_count still tracks the references, and should only
  503. * be used through the normal accessor functions. The top bits of page->flags
  504. * and page->virtual store page management information, but all other fields
  505. * are unused and could be used privately, carefully. The management of this
  506. * page is the responsibility of the one who allocated it, and those who have
  507. * subsequently been given references to it.
  508. *
  509. * The other pages (we may call them "pagecache pages") are completely
  510. * managed by the Linux memory manager: I/O, buffers, swapping etc.
  511. * The following discussion applies only to them.
  512. *
  513. * A pagecache page contains an opaque `private' member, which belongs to the
  514. * page's address_space. Usually, this is the address of a circular list of
  515. * the page's disk buffers. PG_private must be set to tell the VM to call
  516. * into the filesystem to release these pages.
  517. *
  518. * A page may belong to an inode's memory mapping. In this case, page->mapping
  519. * is the pointer to the inode, and page->index is the file offset of the page,
  520. * in units of PAGE_CACHE_SIZE.
  521. *
  522. * If pagecache pages are not associated with an inode, they are said to be
  523. * anonymous pages. These may become associated with the swapcache, and in that
  524. * case PG_swapcache is set, and page->private is an offset into the swapcache.
  525. *
  526. * In either case (swapcache or inode backed), the pagecache itself holds one
  527. * reference to the page. Setting PG_private should also increment the
  528. * refcount. The each user mapping also has a reference to the page.
  529. *
  530. * The pagecache pages are stored in a per-mapping radix tree, which is
  531. * rooted at mapping->page_tree, and indexed by offset.
  532. * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
  533. * lists, we instead now tag pages as dirty/writeback in the radix tree.
  534. *
  535. * All pagecache pages may be subject to I/O:
  536. * - inode pages may need to be read from disk,
  537. * - inode pages which have been modified and are MAP_SHARED may need
  538. * to be written back to the inode on disk,
  539. * - anonymous pages (including MAP_PRIVATE file mappings) which have been
  540. * modified may need to be swapped out to swap space and (later) to be read
  541. * back into memory.
  542. */
  543. /*
  544. * The zone field is never updated after free_area_init_core()
  545. * sets it, so none of the operations on it need to be atomic.
  546. */
  547. /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
  548. #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
  549. #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
  550. #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
  551. #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
  552. /*
  553. * Define the bit shifts to access each section. For non-existent
  554. * sections we define the shift as 0; that plus a 0 mask ensures
  555. * the compiler will optimise away reference to them.
  556. */
  557. #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
  558. #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
  559. #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
  560. #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
  561. /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
  562. #ifdef NODE_NOT_IN_PAGE_FLAGS
  563. #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
  564. #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
  565. SECTIONS_PGOFF : ZONES_PGOFF)
  566. #else
  567. #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
  568. #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
  569. NODES_PGOFF : ZONES_PGOFF)
  570. #endif
  571. #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
  572. #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
  573. #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
  574. #endif
  575. #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
  576. #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
  577. #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
  578. #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_WIDTH) - 1)
  579. #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
  580. static inline enum zone_type page_zonenum(const struct page *page)
  581. {
  582. return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
  583. }
  584. #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
  585. #define SECTION_IN_PAGE_FLAGS
  586. #endif
  587. /*
  588. * The identification function is mainly used by the buddy allocator for
  589. * determining if two pages could be buddies. We are not really identifying
  590. * the zone since we could be using the section number id if we do not have
  591. * node id available in page flags.
  592. * We only guarantee that it will return the same value for two combinable
  593. * pages in a zone.
  594. */
  595. static inline int page_zone_id(struct page *page)
  596. {
  597. return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
  598. }
  599. static inline int zone_to_nid(struct zone *zone)
  600. {
  601. #ifdef CONFIG_NUMA
  602. return zone->node;
  603. #else
  604. return 0;
  605. #endif
  606. }
  607. #ifdef NODE_NOT_IN_PAGE_FLAGS
  608. extern int page_to_nid(const struct page *page);
  609. #else
  610. static inline int page_to_nid(const struct page *page)
  611. {
  612. return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
  613. }
  614. #endif
  615. #ifdef CONFIG_NUMA_BALANCING
  616. static inline int cpu_pid_to_cpupid(int cpu, int pid)
  617. {
  618. return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
  619. }
  620. static inline int cpupid_to_pid(int cpupid)
  621. {
  622. return cpupid & LAST__PID_MASK;
  623. }
  624. static inline int cpupid_to_cpu(int cpupid)
  625. {
  626. return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
  627. }
  628. static inline int cpupid_to_nid(int cpupid)
  629. {
  630. return cpu_to_node(cpupid_to_cpu(cpupid));
  631. }
  632. static inline bool cpupid_pid_unset(int cpupid)
  633. {
  634. return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
  635. }
  636. static inline bool cpupid_cpu_unset(int cpupid)
  637. {
  638. return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
  639. }
  640. static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
  641. {
  642. return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
  643. }
  644. #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
  645. #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
  646. static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
  647. {
  648. return xchg(&page->_last_cpupid, cpupid);
  649. }
  650. static inline int page_cpupid_last(struct page *page)
  651. {
  652. return page->_last_cpupid;
  653. }
  654. static inline void page_cpupid_reset_last(struct page *page)
  655. {
  656. page->_last_cpupid = -1;
  657. }
  658. #else
  659. static inline int page_cpupid_last(struct page *page)
  660. {
  661. return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
  662. }
  663. extern int page_cpupid_xchg_last(struct page *page, int cpupid);
  664. static inline void page_cpupid_reset_last(struct page *page)
  665. {
  666. int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
  667. page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
  668. page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
  669. }
  670. #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
  671. #else /* !CONFIG_NUMA_BALANCING */
  672. static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
  673. {
  674. return page_to_nid(page); /* XXX */
  675. }
  676. static inline int page_cpupid_last(struct page *page)
  677. {
  678. return page_to_nid(page); /* XXX */
  679. }
  680. static inline int cpupid_to_nid(int cpupid)
  681. {
  682. return -1;
  683. }
  684. static inline int cpupid_to_pid(int cpupid)
  685. {
  686. return -1;
  687. }
  688. static inline int cpupid_to_cpu(int cpupid)
  689. {
  690. return -1;
  691. }
  692. static inline int cpu_pid_to_cpupid(int nid, int pid)
  693. {
  694. return -1;
  695. }
  696. static inline bool cpupid_pid_unset(int cpupid)
  697. {
  698. return 1;
  699. }
  700. static inline void page_cpupid_reset_last(struct page *page)
  701. {
  702. }
  703. static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
  704. {
  705. return false;
  706. }
  707. #endif /* CONFIG_NUMA_BALANCING */
  708. static inline struct zone *page_zone(const struct page *page)
  709. {
  710. return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
  711. }
  712. #ifdef SECTION_IN_PAGE_FLAGS
  713. static inline void set_page_section(struct page *page, unsigned long section)
  714. {
  715. page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
  716. page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
  717. }
  718. static inline unsigned long page_to_section(const struct page *page)
  719. {
  720. return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
  721. }
  722. #endif
  723. static inline void set_page_zone(struct page *page, enum zone_type zone)
  724. {
  725. page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
  726. page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
  727. }
  728. static inline void set_page_node(struct page *page, unsigned long node)
  729. {
  730. page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
  731. page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
  732. }
  733. static inline void set_page_links(struct page *page, enum zone_type zone,
  734. unsigned long node, unsigned long pfn)
  735. {
  736. set_page_zone(page, zone);
  737. set_page_node(page, node);
  738. #ifdef SECTION_IN_PAGE_FLAGS
  739. set_page_section(page, pfn_to_section_nr(pfn));
  740. #endif
  741. }
  742. /*
  743. * Some inline functions in vmstat.h depend on page_zone()
  744. */
  745. #include <linux/vmstat.h>
  746. static __always_inline void *lowmem_page_address(const struct page *page)
  747. {
  748. return __va(PFN_PHYS(page_to_pfn(page)));
  749. }
  750. #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
  751. #define HASHED_PAGE_VIRTUAL
  752. #endif
  753. #if defined(WANT_PAGE_VIRTUAL)
  754. static inline void *page_address(const struct page *page)
  755. {
  756. return page->virtual;
  757. }
  758. static inline void set_page_address(struct page *page, void *address)
  759. {
  760. page->virtual = address;
  761. }
  762. #define page_address_init() do { } while(0)
  763. #endif
  764. #if defined(HASHED_PAGE_VIRTUAL)
  765. void *page_address(const struct page *page);
  766. void set_page_address(struct page *page, void *virtual);
  767. void page_address_init(void);
  768. #endif
  769. #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
  770. #define page_address(page) lowmem_page_address(page)
  771. #define set_page_address(page, address) do { } while(0)
  772. #define page_address_init() do { } while(0)
  773. #endif
  774. /*
  775. * On an anonymous page mapped into a user virtual memory area,
  776. * page->mapping points to its anon_vma, not to a struct address_space;
  777. * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
  778. *
  779. * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
  780. * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
  781. * and then page->mapping points, not to an anon_vma, but to a private
  782. * structure which KSM associates with that merged page. See ksm.h.
  783. *
  784. * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
  785. *
  786. * Please note that, confusingly, "page_mapping" refers to the inode
  787. * address_space which maps the page from disk; whereas "page_mapped"
  788. * refers to user virtual address space into which the page is mapped.
  789. */
  790. #define PAGE_MAPPING_ANON 1
  791. #define PAGE_MAPPING_KSM 2
  792. #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
  793. extern struct address_space *page_mapping(struct page *page);
  794. /* Neutral page->mapping pointer to address_space or anon_vma or other */
  795. static inline void *page_rmapping(struct page *page)
  796. {
  797. return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
  798. }
  799. extern struct address_space *__page_file_mapping(struct page *);
  800. static inline
  801. struct address_space *page_file_mapping(struct page *page)
  802. {
  803. if (unlikely(PageSwapCache(page)))
  804. return __page_file_mapping(page);
  805. return page->mapping;
  806. }
  807. static inline int PageAnon(struct page *page)
  808. {
  809. return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
  810. }
  811. /*
  812. * Return the pagecache index of the passed page. Regular pagecache pages
  813. * use ->index whereas swapcache pages use ->private
  814. */
  815. static inline pgoff_t page_index(struct page *page)
  816. {
  817. if (unlikely(PageSwapCache(page)))
  818. return page_private(page);
  819. return page->index;
  820. }
  821. extern pgoff_t __page_file_index(struct page *page);
  822. /*
  823. * Return the file index of the page. Regular pagecache pages use ->index
  824. * whereas swapcache pages use swp_offset(->private)
  825. */
  826. static inline pgoff_t page_file_index(struct page *page)
  827. {
  828. if (unlikely(PageSwapCache(page)))
  829. return __page_file_index(page);
  830. return page->index;
  831. }
  832. /*
  833. * Return true if this page is mapped into pagetables.
  834. */
  835. static inline int page_mapped(struct page *page)
  836. {
  837. return atomic_read(&(page)->_mapcount) >= 0;
  838. }
  839. /*
  840. * Different kinds of faults, as returned by handle_mm_fault().
  841. * Used to decide whether a process gets delivered SIGBUS or
  842. * just gets major/minor fault counters bumped up.
  843. */
  844. #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
  845. #define VM_FAULT_OOM 0x0001
  846. #define VM_FAULT_SIGBUS 0x0002
  847. #define VM_FAULT_MAJOR 0x0004
  848. #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
  849. #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
  850. #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
  851. #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
  852. #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
  853. #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
  854. #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
  855. #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
  856. #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
  857. VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
  858. /* Encode hstate index for a hwpoisoned large page */
  859. #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
  860. #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
  861. /*
  862. * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
  863. */
  864. extern void pagefault_out_of_memory(void);
  865. #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
  866. /*
  867. * Flags passed to show_mem() and show_free_areas() to suppress output in
  868. * various contexts.
  869. */
  870. #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
  871. #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
  872. extern void show_free_areas(unsigned int flags);
  873. extern bool skip_free_areas_node(unsigned int flags, int nid);
  874. int shmem_zero_setup(struct vm_area_struct *);
  875. extern int can_do_mlock(void);
  876. extern int user_shm_lock(size_t, struct user_struct *);
  877. extern void user_shm_unlock(size_t, struct user_struct *);
  878. /*
  879. * Parameter block passed down to zap_pte_range in exceptional cases.
  880. */
  881. struct zap_details {
  882. struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
  883. struct address_space *check_mapping; /* Check page->mapping if set */
  884. pgoff_t first_index; /* Lowest page->index to unmap */
  885. pgoff_t last_index; /* Highest page->index to unmap */
  886. };
  887. struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
  888. pte_t pte);
  889. int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
  890. unsigned long size);
  891. void zap_page_range(struct vm_area_struct *vma, unsigned long address,
  892. unsigned long size, struct zap_details *);
  893. void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  894. unsigned long start, unsigned long end);
  895. /**
  896. * mm_walk - callbacks for walk_page_range
  897. * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
  898. * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
  899. * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
  900. * this handler is required to be able to handle
  901. * pmd_trans_huge() pmds. They may simply choose to
  902. * split_huge_page() instead of handling it explicitly.
  903. * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
  904. * @pte_hole: if set, called for each hole at all levels
  905. * @hugetlb_entry: if set, called for each hugetlb entry
  906. * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
  907. * is used.
  908. *
  909. * (see walk_page_range for more details)
  910. */
  911. struct mm_walk {
  912. int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
  913. unsigned long next, struct mm_walk *walk);
  914. int (*pud_entry)(pud_t *pud, unsigned long addr,
  915. unsigned long next, struct mm_walk *walk);
  916. int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
  917. unsigned long next, struct mm_walk *walk);
  918. int (*pte_entry)(pte_t *pte, unsigned long addr,
  919. unsigned long next, struct mm_walk *walk);
  920. int (*pte_hole)(unsigned long addr, unsigned long next,
  921. struct mm_walk *walk);
  922. int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
  923. unsigned long addr, unsigned long next,
  924. struct mm_walk *walk);
  925. struct mm_struct *mm;
  926. void *private;
  927. };
  928. int walk_page_range(unsigned long addr, unsigned long end,
  929. struct mm_walk *walk);
  930. void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
  931. unsigned long end, unsigned long floor, unsigned long ceiling);
  932. int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
  933. struct vm_area_struct *vma);
  934. void unmap_mapping_range(struct address_space *mapping,
  935. loff_t const holebegin, loff_t const holelen, int even_cows);
  936. int follow_pfn(struct vm_area_struct *vma, unsigned long address,
  937. unsigned long *pfn);
  938. int follow_phys(struct vm_area_struct *vma, unsigned long address,
  939. unsigned int flags, unsigned long *prot, resource_size_t *phys);
  940. int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
  941. void *buf, int len, int write);
  942. static inline void unmap_shared_mapping_range(struct address_space *mapping,
  943. loff_t const holebegin, loff_t const holelen)
  944. {
  945. unmap_mapping_range(mapping, holebegin, holelen, 0);
  946. }
  947. extern void truncate_pagecache(struct inode *inode, loff_t new);
  948. extern void truncate_setsize(struct inode *inode, loff_t newsize);
  949. void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
  950. int truncate_inode_page(struct address_space *mapping, struct page *page);
  951. int generic_error_remove_page(struct address_space *mapping, struct page *page);
  952. int invalidate_inode_page(struct page *page);
  953. #ifdef CONFIG_MMU
  954. extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
  955. unsigned long address, unsigned int flags);
  956. extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
  957. unsigned long address, unsigned int fault_flags);
  958. #else
  959. static inline int handle_mm_fault(struct mm_struct *mm,
  960. struct vm_area_struct *vma, unsigned long address,
  961. unsigned int flags)
  962. {
  963. /* should never happen if there's no MMU */
  964. BUG();
  965. return VM_FAULT_SIGBUS;
  966. }
  967. static inline int fixup_user_fault(struct task_struct *tsk,
  968. struct mm_struct *mm, unsigned long address,
  969. unsigned int fault_flags)
  970. {
  971. /* should never happen if there's no MMU */
  972. BUG();
  973. return -EFAULT;
  974. }
  975. #endif
  976. extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
  977. extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
  978. void *buf, int len, int write);
  979. long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
  980. unsigned long start, unsigned long nr_pages,
  981. unsigned int foll_flags, struct page **pages,
  982. struct vm_area_struct **vmas, int *nonblocking);
  983. long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
  984. unsigned long start, unsigned long nr_pages,
  985. int write, int force, struct page **pages,
  986. struct vm_area_struct **vmas);
  987. int get_user_pages_fast(unsigned long start, int nr_pages, int write,
  988. struct page **pages);
  989. struct kvec;
  990. int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
  991. struct page **pages);
  992. int get_kernel_page(unsigned long start, int write, struct page **pages);
  993. struct page *get_dump_page(unsigned long addr);
  994. extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
  995. extern void do_invalidatepage(struct page *page, unsigned int offset,
  996. unsigned int length);
  997. int __set_page_dirty_nobuffers(struct page *page);
  998. int __set_page_dirty_no_writeback(struct page *page);
  999. int redirty_page_for_writepage(struct writeback_control *wbc,
  1000. struct page *page);
  1001. void account_page_dirtied(struct page *page, struct address_space *mapping);
  1002. void account_page_writeback(struct page *page);
  1003. int set_page_dirty(struct page *page);
  1004. int set_page_dirty_lock(struct page *page);
  1005. int clear_page_dirty_for_io(struct page *page);
  1006. /* Is the vma a continuation of the stack vma above it? */
  1007. static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
  1008. {
  1009. return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
  1010. }
  1011. static inline int stack_guard_page_start(struct vm_area_struct *vma,
  1012. unsigned long addr)
  1013. {
  1014. return (vma->vm_flags & VM_GROWSDOWN) &&
  1015. (vma->vm_start == addr) &&
  1016. !vma_growsdown(vma->vm_prev, addr);
  1017. }
  1018. /* Is the vma a continuation of the stack vma below it? */
  1019. static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
  1020. {
  1021. return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
  1022. }
  1023. static inline int stack_guard_page_end(struct vm_area_struct *vma,
  1024. unsigned long addr)
  1025. {
  1026. return (vma->vm_flags & VM_GROWSUP) &&
  1027. (vma->vm_end == addr) &&
  1028. !vma_growsup(vma->vm_next, addr);
  1029. }
  1030. extern pid_t
  1031. vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
  1032. extern unsigned long move_page_tables(struct vm_area_struct *vma,
  1033. unsigned long old_addr, struct vm_area_struct *new_vma,
  1034. unsigned long new_addr, unsigned long len,
  1035. bool need_rmap_locks);
  1036. extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
  1037. unsigned long end, pgprot_t newprot,
  1038. int dirty_accountable, int prot_numa);
  1039. extern int mprotect_fixup(struct vm_area_struct *vma,
  1040. struct vm_area_struct **pprev, unsigned long start,
  1041. unsigned long end, unsigned long newflags);
  1042. /*
  1043. * doesn't attempt to fault and will return short.
  1044. */
  1045. int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
  1046. struct page **pages);
  1047. /*
  1048. * per-process(per-mm_struct) statistics.
  1049. */
  1050. static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
  1051. {
  1052. long val = atomic_long_read(&mm->rss_stat.count[member]);
  1053. #ifdef SPLIT_RSS_COUNTING
  1054. /*
  1055. * counter is updated in asynchronous manner and may go to minus.
  1056. * But it's never be expected number for users.
  1057. */
  1058. if (val < 0)
  1059. val = 0;
  1060. #endif
  1061. return (unsigned long)val;
  1062. }
  1063. static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
  1064. {
  1065. atomic_long_add(value, &mm->rss_stat.count[member]);
  1066. }
  1067. static inline void inc_mm_counter(struct mm_struct *mm, int member)
  1068. {
  1069. atomic_long_inc(&mm->rss_stat.count[member]);
  1070. }
  1071. static inline void dec_mm_counter(struct mm_struct *mm, int member)
  1072. {
  1073. atomic_long_dec(&mm->rss_stat.count[member]);
  1074. }
  1075. static inline unsigned long get_mm_rss(struct mm_struct *mm)
  1076. {
  1077. return get_mm_counter(mm, MM_FILEPAGES) +
  1078. get_mm_counter(mm, MM_ANONPAGES);
  1079. }
  1080. static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
  1081. {
  1082. return max(mm->hiwater_rss, get_mm_rss(mm));
  1083. }
  1084. static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
  1085. {
  1086. return max(mm->hiwater_vm, mm->total_vm);
  1087. }
  1088. static inline void update_hiwater_rss(struct mm_struct *mm)
  1089. {
  1090. unsigned long _rss = get_mm_rss(mm);
  1091. if ((mm)->hiwater_rss < _rss)
  1092. (mm)->hiwater_rss = _rss;
  1093. }
  1094. static inline void update_hiwater_vm(struct mm_struct *mm)
  1095. {
  1096. if (mm->hiwater_vm < mm->total_vm)
  1097. mm->hiwater_vm = mm->total_vm;
  1098. }
  1099. static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
  1100. struct mm_struct *mm)
  1101. {
  1102. unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
  1103. if (*maxrss < hiwater_rss)
  1104. *maxrss = hiwater_rss;
  1105. }
  1106. #if defined(SPLIT_RSS_COUNTING)
  1107. void sync_mm_rss(struct mm_struct *mm);
  1108. #else
  1109. static inline void sync_mm_rss(struct mm_struct *mm)
  1110. {
  1111. }
  1112. #endif
  1113. int vma_wants_writenotify(struct vm_area_struct *vma);
  1114. extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
  1115. spinlock_t **ptl);
  1116. static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
  1117. spinlock_t **ptl)
  1118. {
  1119. pte_t *ptep;
  1120. __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
  1121. return ptep;
  1122. }
  1123. #ifdef __PAGETABLE_PUD_FOLDED
  1124. static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
  1125. unsigned long address)
  1126. {
  1127. return 0;
  1128. }
  1129. #else
  1130. int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
  1131. #endif
  1132. #ifdef __PAGETABLE_PMD_FOLDED
  1133. static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
  1134. unsigned long address)
  1135. {
  1136. return 0;
  1137. }
  1138. #else
  1139. int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
  1140. #endif
  1141. int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
  1142. pmd_t *pmd, unsigned long address);
  1143. int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
  1144. /*
  1145. * The following ifdef needed to get the 4level-fixup.h header to work.
  1146. * Remove it when 4level-fixup.h has been removed.
  1147. */
  1148. #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
  1149. static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
  1150. {
  1151. return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
  1152. NULL: pud_offset(pgd, address);
  1153. }
  1154. static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
  1155. {
  1156. return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
  1157. NULL: pmd_offset(pud, address);
  1158. }
  1159. #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
  1160. #if USE_SPLIT_PTE_PTLOCKS
  1161. #if ALLOC_SPLIT_PTLOCKS
  1162. extern bool ptlock_alloc(struct page *page);
  1163. extern void ptlock_free(struct page *page);
  1164. static inline spinlock_t *ptlock_ptr(struct page *page)
  1165. {
  1166. return page->ptl;
  1167. }
  1168. #else /* ALLOC_SPLIT_PTLOCKS */
  1169. static inline bool ptlock_alloc(struct page *page)
  1170. {
  1171. return true;
  1172. }
  1173. static inline void ptlock_free(struct page *page)
  1174. {
  1175. }
  1176. static inline spinlock_t *ptlock_ptr(struct page *page)
  1177. {
  1178. return &page->ptl;
  1179. }
  1180. #endif /* ALLOC_SPLIT_PTLOCKS */
  1181. static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1182. {
  1183. return ptlock_ptr(pmd_page(*pmd));
  1184. }
  1185. static inline bool ptlock_init(struct page *page)
  1186. {
  1187. /*
  1188. * prep_new_page() initialize page->private (and therefore page->ptl)
  1189. * with 0. Make sure nobody took it in use in between.
  1190. *
  1191. * It can happen if arch try to use slab for page table allocation:
  1192. * slab code uses page->slab_cache and page->first_page (for tail
  1193. * pages), which share storage with page->ptl.
  1194. */
  1195. VM_BUG_ON(*(unsigned long *)&page->ptl);
  1196. if (!ptlock_alloc(page))
  1197. return false;
  1198. spin_lock_init(ptlock_ptr(page));
  1199. return true;
  1200. }
  1201. /* Reset page->mapping so free_pages_check won't complain. */
  1202. static inline void pte_lock_deinit(struct page *page)
  1203. {
  1204. page->mapping = NULL;
  1205. ptlock_free(page);
  1206. }
  1207. #else /* !USE_SPLIT_PTE_PTLOCKS */
  1208. /*
  1209. * We use mm->page_table_lock to guard all pagetable pages of the mm.
  1210. */
  1211. static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1212. {
  1213. return &mm->page_table_lock;
  1214. }
  1215. static inline bool ptlock_init(struct page *page) { return true; }
  1216. static inline void pte_lock_deinit(struct page *page) {}
  1217. #endif /* USE_SPLIT_PTE_PTLOCKS */
  1218. static inline bool pgtable_page_ctor(struct page *page)
  1219. {
  1220. inc_zone_page_state(page, NR_PAGETABLE);
  1221. return ptlock_init(page);
  1222. }
  1223. static inline void pgtable_page_dtor(struct page *page)
  1224. {
  1225. pte_lock_deinit(page);
  1226. dec_zone_page_state(page, NR_PAGETABLE);
  1227. }
  1228. #define pte_offset_map_lock(mm, pmd, address, ptlp) \
  1229. ({ \
  1230. spinlock_t *__ptl = pte_lockptr(mm, pmd); \
  1231. pte_t *__pte = pte_offset_map(pmd, address); \
  1232. *(ptlp) = __ptl; \
  1233. spin_lock(__ptl); \
  1234. __pte; \
  1235. })
  1236. #define pte_unmap_unlock(pte, ptl) do { \
  1237. spin_unlock(ptl); \
  1238. pte_unmap(pte); \
  1239. } while (0)
  1240. #define pte_alloc_map(mm, vma, pmd, address) \
  1241. ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
  1242. pmd, address))? \
  1243. NULL: pte_offset_map(pmd, address))
  1244. #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
  1245. ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
  1246. pmd, address))? \
  1247. NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
  1248. #define pte_alloc_kernel(pmd, address) \
  1249. ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
  1250. NULL: pte_offset_kernel(pmd, address))
  1251. #if USE_SPLIT_PMD_PTLOCKS
  1252. static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1253. {
  1254. return ptlock_ptr(virt_to_page(pmd));
  1255. }
  1256. static inline bool pgtable_pmd_page_ctor(struct page *page)
  1257. {
  1258. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  1259. page->pmd_huge_pte = NULL;
  1260. #endif
  1261. return ptlock_init(page);
  1262. }
  1263. static inline void pgtable_pmd_page_dtor(struct page *page)
  1264. {
  1265. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  1266. VM_BUG_ON(page->pmd_huge_pte);
  1267. #endif
  1268. ptlock_free(page);
  1269. }
  1270. #define pmd_huge_pte(mm, pmd) (virt_to_page(pmd)->pmd_huge_pte)
  1271. #else
  1272. static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1273. {
  1274. return &mm->page_table_lock;
  1275. }
  1276. static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
  1277. static inline void pgtable_pmd_page_dtor(struct page *page) {}
  1278. #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
  1279. #endif
  1280. static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
  1281. {
  1282. spinlock_t *ptl = pmd_lockptr(mm, pmd);
  1283. spin_lock(ptl);
  1284. return ptl;
  1285. }
  1286. extern void free_area_init(unsigned long * zones_size);
  1287. extern void free_area_init_node(int nid, unsigned long * zones_size,
  1288. unsigned long zone_start_pfn, unsigned long *zholes_size);
  1289. extern void free_initmem(void);
  1290. /*
  1291. * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
  1292. * into the buddy system. The freed pages will be poisoned with pattern
  1293. * "poison" if it's within range [0, UCHAR_MAX].
  1294. * Return pages freed into the buddy system.
  1295. */
  1296. extern unsigned long free_reserved_area(void *start, void *end,
  1297. int poison, char *s);
  1298. #ifdef CONFIG_HIGHMEM
  1299. /*
  1300. * Free a highmem page into the buddy system, adjusting totalhigh_pages
  1301. * and totalram_pages.
  1302. */
  1303. extern void free_highmem_page(struct page *page);
  1304. #endif
  1305. extern void adjust_managed_page_count(struct page *page, long count);
  1306. extern void mem_init_print_info(const char *str);
  1307. /* Free the reserved page into the buddy system, so it gets managed. */
  1308. static inline void __free_reserved_page(struct page *page)
  1309. {
  1310. ClearPageReserved(page);
  1311. init_page_count(page);
  1312. __free_page(page);
  1313. }
  1314. static inline void free_reserved_page(struct page *page)
  1315. {
  1316. __free_reserved_page(page);
  1317. adjust_managed_page_count(page, 1);
  1318. }
  1319. static inline void mark_page_reserved(struct page *page)
  1320. {
  1321. SetPageReserved(page);
  1322. adjust_managed_page_count(page, -1);
  1323. }
  1324. /*
  1325. * Default method to free all the __init memory into the buddy system.
  1326. * The freed pages will be poisoned with pattern "poison" if it's within
  1327. * range [0, UCHAR_MAX].
  1328. * Return pages freed into the buddy system.
  1329. */
  1330. static inline unsigned long free_initmem_default(int poison)
  1331. {
  1332. extern char __init_begin[], __init_end[];
  1333. return free_reserved_area(&__init_begin, &__init_end,
  1334. poison, "unused kernel");
  1335. }
  1336. static inline unsigned long get_num_physpages(void)
  1337. {
  1338. int nid;
  1339. unsigned long phys_pages = 0;
  1340. for_each_online_node(nid)
  1341. phys_pages += node_present_pages(nid);
  1342. return phys_pages;
  1343. }
  1344. #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
  1345. /*
  1346. * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
  1347. * zones, allocate the backing mem_map and account for memory holes in a more
  1348. * architecture independent manner. This is a substitute for creating the
  1349. * zone_sizes[] and zholes_size[] arrays and passing them to
  1350. * free_area_init_node()
  1351. *
  1352. * An architecture is expected to register range of page frames backed by
  1353. * physical memory with memblock_add[_node]() before calling
  1354. * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
  1355. * usage, an architecture is expected to do something like
  1356. *
  1357. * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
  1358. * max_highmem_pfn};
  1359. * for_each_valid_physical_page_range()
  1360. * memblock_add_node(base, size, nid)
  1361. * free_area_init_nodes(max_zone_pfns);
  1362. *
  1363. * free_bootmem_with_active_regions() calls free_bootmem_node() for each
  1364. * registered physical page range. Similarly
  1365. * sparse_memory_present_with_active_regions() calls memory_present() for
  1366. * each range when SPARSEMEM is enabled.
  1367. *
  1368. * See mm/page_alloc.c for more information on each function exposed by
  1369. * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
  1370. */
  1371. extern void free_area_init_nodes(unsigned long *max_zone_pfn);
  1372. unsigned long node_map_pfn_alignment(void);
  1373. unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
  1374. unsigned long end_pfn);
  1375. extern unsigned long absent_pages_in_range(unsigned long start_pfn,
  1376. unsigned long end_pfn);
  1377. extern void get_pfn_range_for_nid(unsigned int nid,
  1378. unsigned long *start_pfn, unsigned long *end_pfn);
  1379. extern unsigned long find_min_pfn_with_active_regions(void);
  1380. extern void free_bootmem_with_active_regions(int nid,
  1381. unsigned long max_low_pfn);
  1382. extern void sparse_memory_present_with_active_regions(int nid);
  1383. #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
  1384. #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
  1385. !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
  1386. static inline int __early_pfn_to_nid(unsigned long pfn)
  1387. {
  1388. return 0;
  1389. }
  1390. #else
  1391. /* please see mm/page_alloc.c */
  1392. extern int __meminit early_pfn_to_nid(unsigned long pfn);
  1393. #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
  1394. /* there is a per-arch backend function. */
  1395. extern int __meminit __early_pfn_to_nid(unsigned long pfn);
  1396. #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
  1397. #endif
  1398. extern void set_dma_reserve(unsigned long new_dma_reserve);
  1399. extern void memmap_init_zone(unsigned long, int, unsigned long,
  1400. unsigned long, enum memmap_context);
  1401. extern void setup_per_zone_wmarks(void);
  1402. extern int __meminit init_per_zone_wmark_min(void);
  1403. extern void mem_init(void);
  1404. extern void __init mmap_init(void);
  1405. extern void show_mem(unsigned int flags);
  1406. extern void si_meminfo(struct sysinfo * val);
  1407. extern void si_meminfo_node(struct sysinfo *val, int nid);
  1408. extern __printf(3, 4)
  1409. void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
  1410. extern void setup_per_cpu_pageset(void);
  1411. extern void zone_pcp_update(struct zone *zone);
  1412. extern void zone_pcp_reset(struct zone *zone);
  1413. /* page_alloc.c */
  1414. extern int min_free_kbytes;
  1415. /* nommu.c */
  1416. extern atomic_long_t mmap_pages_allocated;
  1417. extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
  1418. /* interval_tree.c */
  1419. void vma_interval_tree_insert(struct vm_area_struct *node,
  1420. struct rb_root *root);
  1421. void vma_interval_tree_insert_after(struct vm_area_struct *node,
  1422. struct vm_area_struct *prev,
  1423. struct rb_root *root);
  1424. void vma_interval_tree_remove(struct vm_area_struct *node,
  1425. struct rb_root *root);
  1426. struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
  1427. unsigned long start, unsigned long last);
  1428. struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
  1429. unsigned long start, unsigned long last);
  1430. #define vma_interval_tree_foreach(vma, root, start, last) \
  1431. for (vma = vma_interval_tree_iter_first(root, start, last); \
  1432. vma; vma = vma_interval_tree_iter_next(vma, start, last))
  1433. static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
  1434. struct list_head *list)
  1435. {
  1436. list_add_tail(&vma->shared.nonlinear, list);
  1437. }
  1438. void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
  1439. struct rb_root *root);
  1440. void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
  1441. struct rb_root *root);
  1442. struct anon_vma_chain *anon_vma_interval_tree_iter_first(
  1443. struct rb_root *root, unsigned long start, unsigned long last);
  1444. struct anon_vma_chain *anon_vma_interval_tree_iter_next(
  1445. struct anon_vma_chain *node, unsigned long start, unsigned long last);
  1446. #ifdef CONFIG_DEBUG_VM_RB
  1447. void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
  1448. #endif
  1449. #define anon_vma_interval_tree_foreach(avc, root, start, last) \
  1450. for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
  1451. avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
  1452. /* mmap.c */
  1453. extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
  1454. extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
  1455. unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
  1456. extern struct vm_area_struct *vma_merge(struct mm_struct *,
  1457. struct vm_area_struct *prev, unsigned long addr, unsigned long end,
  1458. unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
  1459. struct mempolicy *);
  1460. extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
  1461. extern int split_vma(struct mm_struct *,
  1462. struct vm_area_struct *, unsigned long addr, int new_below);
  1463. extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
  1464. extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
  1465. struct rb_node **, struct rb_node *);
  1466. extern void unlink_file_vma(struct vm_area_struct *);
  1467. extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
  1468. unsigned long addr, unsigned long len, pgoff_t pgoff,
  1469. bool *need_rmap_locks);
  1470. extern void exit_mmap(struct mm_struct *);
  1471. extern int mm_take_all_locks(struct mm_struct *mm);
  1472. extern void mm_drop_all_locks(struct mm_struct *mm);
  1473. extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
  1474. extern struct file *get_mm_exe_file(struct mm_struct *mm);
  1475. extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
  1476. extern int install_special_mapping(struct mm_struct *mm,
  1477. unsigned long addr, unsigned long len,
  1478. unsigned long flags, struct page **pages);
  1479. extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
  1480. extern unsigned long mmap_region(struct file *file, unsigned long addr,
  1481. unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
  1482. extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
  1483. unsigned long len, unsigned long prot, unsigned long flags,
  1484. unsigned long pgoff, unsigned long *populate);
  1485. extern int do_munmap(struct mm_struct *, unsigned long, size_t);
  1486. #ifdef CONFIG_MMU
  1487. extern int __mm_populate(unsigned long addr, unsigned long len,
  1488. int ignore_errors);
  1489. static inline void mm_populate(unsigned long addr, unsigned long len)
  1490. {
  1491. /* Ignore errors */
  1492. (void) __mm_populate(addr, len, 1);
  1493. }
  1494. #else
  1495. static inline void mm_populate(unsigned long addr, unsigned long len) {}
  1496. #endif
  1497. /* These take the mm semaphore themselves */
  1498. extern unsigned long vm_brk(unsigned long, unsigned long);
  1499. extern int vm_munmap(unsigned long, size_t);
  1500. extern unsigned long vm_mmap(struct file *, unsigned long,
  1501. unsigned long, unsigned long,
  1502. unsigned long, unsigned long);
  1503. struct vm_unmapped_area_info {
  1504. #define VM_UNMAPPED_AREA_TOPDOWN 1
  1505. unsigned long flags;
  1506. unsigned long length;
  1507. unsigned long low_limit;
  1508. unsigned long high_limit;
  1509. unsigned long align_mask;
  1510. unsigned long align_offset;
  1511. };
  1512. extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
  1513. extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
  1514. /*
  1515. * Search for an unmapped address range.
  1516. *
  1517. * We are looking for a range that:
  1518. * - does not intersect with any VMA;
  1519. * - is contained within the [low_limit, high_limit) interval;
  1520. * - is at least the desired size.
  1521. * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
  1522. */
  1523. static inline unsigned long
  1524. vm_unmapped_area(struct vm_unmapped_area_info *info)
  1525. {
  1526. if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
  1527. return unmapped_area(info);
  1528. else
  1529. return unmapped_area_topdown(info);
  1530. }
  1531. /* truncate.c */
  1532. extern void truncate_inode_pages(struct address_space *, loff_t);
  1533. extern void truncate_inode_pages_range(struct address_space *,
  1534. loff_t lstart, loff_t lend);
  1535. /* generic vm_area_ops exported for stackable file systems */
  1536. extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
  1537. extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
  1538. /* mm/page-writeback.c */
  1539. int write_one_page(struct page *page, int wait);
  1540. void task_dirty_inc(struct task_struct *tsk);
  1541. /* readahead.c */
  1542. #define VM_MAX_READAHEAD 128 /* kbytes */
  1543. #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
  1544. int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
  1545. pgoff_t offset, unsigned long nr_to_read);
  1546. void page_cache_sync_readahead(struct address_space *mapping,
  1547. struct file_ra_state *ra,
  1548. struct file *filp,
  1549. pgoff_t offset,
  1550. unsigned long size);
  1551. void page_cache_async_readahead(struct address_space *mapping,
  1552. struct file_ra_state *ra,
  1553. struct file *filp,
  1554. struct page *pg,
  1555. pgoff_t offset,
  1556. unsigned long size);
  1557. unsigned long max_sane_readahead(unsigned long nr);
  1558. unsigned long ra_submit(struct file_ra_state *ra,
  1559. struct address_space *mapping,
  1560. struct file *filp);
  1561. /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
  1562. extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
  1563. /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
  1564. extern int expand_downwards(struct vm_area_struct *vma,
  1565. unsigned long address);
  1566. #if VM_GROWSUP
  1567. extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
  1568. #else
  1569. #define expand_upwards(vma, address) do { } while (0)
  1570. #endif
  1571. /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
  1572. extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
  1573. extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
  1574. struct vm_area_struct **pprev);
  1575. /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
  1576. NULL if none. Assume start_addr < end_addr. */
  1577. static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
  1578. {
  1579. struct vm_area_struct * vma = find_vma(mm,start_addr);
  1580. if (vma && end_addr <= vma->vm_start)
  1581. vma = NULL;
  1582. return vma;
  1583. }
  1584. static inline unsigned long vma_pages(struct vm_area_struct *vma)
  1585. {
  1586. return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
  1587. }
  1588. /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
  1589. static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
  1590. unsigned long vm_start, unsigned long vm_end)
  1591. {
  1592. struct vm_area_struct *vma = find_vma(mm, vm_start);
  1593. if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
  1594. vma = NULL;
  1595. return vma;
  1596. }
  1597. #ifdef CONFIG_MMU
  1598. pgprot_t vm_get_page_prot(unsigned long vm_flags);
  1599. #else
  1600. static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
  1601. {
  1602. return __pgprot(0);
  1603. }
  1604. #endif
  1605. #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
  1606. unsigned long change_prot_numa(struct vm_area_struct *vma,
  1607. unsigned long start, unsigned long end);
  1608. #endif
  1609. struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
  1610. int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
  1611. unsigned long pfn, unsigned long size, pgprot_t);
  1612. int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
  1613. int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
  1614. unsigned long pfn);
  1615. int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
  1616. unsigned long pfn);
  1617. int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
  1618. struct page *follow_page_mask(struct vm_area_struct *vma,
  1619. unsigned long address, unsigned int foll_flags,
  1620. unsigned int *page_mask);
  1621. static inline struct page *follow_page(struct vm_area_struct *vma,
  1622. unsigned long address, unsigned int foll_flags)
  1623. {
  1624. unsigned int unused_page_mask;
  1625. return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
  1626. }
  1627. #define FOLL_WRITE 0x01 /* check pte is writable */
  1628. #define FOLL_TOUCH 0x02 /* mark page accessed */
  1629. #define FOLL_GET 0x04 /* do get_page on page */
  1630. #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
  1631. #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
  1632. #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
  1633. * and return without waiting upon it */
  1634. #define FOLL_MLOCK 0x40 /* mark page as mlocked */
  1635. #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
  1636. #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
  1637. #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
  1638. #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
  1639. typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
  1640. void *data);
  1641. extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
  1642. unsigned long size, pte_fn_t fn, void *data);
  1643. #ifdef CONFIG_PROC_FS
  1644. void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
  1645. #else
  1646. static inline void vm_stat_account(struct mm_struct *mm,
  1647. unsigned long flags, struct file *file, long pages)
  1648. {
  1649. mm->total_vm += pages;
  1650. }
  1651. #endif /* CONFIG_PROC_FS */
  1652. #ifdef CONFIG_DEBUG_PAGEALLOC
  1653. extern void kernel_map_pages(struct page *page, int numpages, int enable);
  1654. #ifdef CONFIG_HIBERNATION
  1655. extern bool kernel_page_present(struct page *page);
  1656. #endif /* CONFIG_HIBERNATION */
  1657. #else
  1658. static inline void
  1659. kernel_map_pages(struct page *page, int numpages, int enable) {}
  1660. #ifdef CONFIG_HIBERNATION
  1661. static inline bool kernel_page_present(struct page *page) { return true; }
  1662. #endif /* CONFIG_HIBERNATION */
  1663. #endif
  1664. extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
  1665. #ifdef __HAVE_ARCH_GATE_AREA
  1666. int in_gate_area_no_mm(unsigned long addr);
  1667. int in_gate_area(struct mm_struct *mm, unsigned long addr);
  1668. #else
  1669. int in_gate_area_no_mm(unsigned long addr);
  1670. #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
  1671. #endif /* __HAVE_ARCH_GATE_AREA */
  1672. #ifdef CONFIG_SYSCTL
  1673. extern int sysctl_drop_caches;
  1674. int drop_caches_sysctl_handler(struct ctl_table *, int,
  1675. void __user *, size_t *, loff_t *);
  1676. #endif
  1677. unsigned long shrink_slab(struct shrink_control *shrink,
  1678. unsigned long nr_pages_scanned,
  1679. unsigned long lru_pages);
  1680. #ifndef CONFIG_MMU
  1681. #define randomize_va_space 0
  1682. #else
  1683. extern int randomize_va_space;
  1684. #endif
  1685. const char * arch_vma_name(struct vm_area_struct *vma);
  1686. void print_vma_addr(char *prefix, unsigned long rip);
  1687. void sparse_mem_maps_populate_node(struct page **map_map,
  1688. unsigned long pnum_begin,
  1689. unsigned long pnum_end,
  1690. unsigned long map_count,
  1691. int nodeid);
  1692. struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
  1693. pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
  1694. pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
  1695. pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
  1696. pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
  1697. void *vmemmap_alloc_block(unsigned long size, int node);
  1698. void *vmemmap_alloc_block_buf(unsigned long size, int node);
  1699. void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
  1700. int vmemmap_populate_basepages(unsigned long start, unsigned long end,
  1701. int node);
  1702. int vmemmap_populate(unsigned long start, unsigned long end, int node);
  1703. void vmemmap_populate_print_last(void);
  1704. #ifdef CONFIG_MEMORY_HOTPLUG
  1705. void vmemmap_free(unsigned long start, unsigned long end);
  1706. #endif
  1707. void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
  1708. unsigned long size);
  1709. enum mf_flags {
  1710. MF_COUNT_INCREASED = 1 << 0,
  1711. MF_ACTION_REQUIRED = 1 << 1,
  1712. MF_MUST_KILL = 1 << 2,
  1713. MF_SOFT_OFFLINE = 1 << 3,
  1714. };
  1715. extern int memory_failure(unsigned long pfn, int trapno, int flags);
  1716. extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
  1717. extern int unpoison_memory(unsigned long pfn);
  1718. extern int sysctl_memory_failure_early_kill;
  1719. extern int sysctl_memory_failure_recovery;
  1720. extern void shake_page(struct page *p, int access);
  1721. extern atomic_long_t num_poisoned_pages;
  1722. extern int soft_offline_page(struct page *page, int flags);
  1723. extern void dump_page(struct page *page);
  1724. #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
  1725. extern void clear_huge_page(struct page *page,
  1726. unsigned long addr,
  1727. unsigned int pages_per_huge_page);
  1728. extern void copy_user_huge_page(struct page *dst, struct page *src,
  1729. unsigned long addr, struct vm_area_struct *vma,
  1730. unsigned int pages_per_huge_page);
  1731. #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
  1732. #ifdef CONFIG_DEBUG_PAGEALLOC
  1733. extern unsigned int _debug_guardpage_minorder;
  1734. static inline unsigned int debug_guardpage_minorder(void)
  1735. {
  1736. return _debug_guardpage_minorder;
  1737. }
  1738. static inline bool page_is_guard(struct page *page)
  1739. {
  1740. return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
  1741. }
  1742. #else
  1743. static inline unsigned int debug_guardpage_minorder(void) { return 0; }
  1744. static inline bool page_is_guard(struct page *page) { return false; }
  1745. #endif /* CONFIG_DEBUG_PAGEALLOC */
  1746. #if MAX_NUMNODES > 1
  1747. void __init setup_nr_node_ids(void);
  1748. #else
  1749. static inline void setup_nr_node_ids(void) {}
  1750. #endif
  1751. #endif /* __KERNEL__ */
  1752. #endif /* _LINUX_MM_H */