mm.h 83 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660
  1. /* SPDX-License-Identifier: GPL-2.0 */
  2. #ifndef _LINUX_MM_H
  3. #define _LINUX_MM_H
  4. #include <linux/errno.h>
  5. #ifdef __KERNEL__
  6. #include <linux/mmdebug.h>
  7. #include <linux/gfp.h>
  8. #include <linux/bug.h>
  9. #include <linux/list.h>
  10. #include <linux/mmzone.h>
  11. #include <linux/rbtree.h>
  12. #include <linux/atomic.h>
  13. #include <linux/debug_locks.h>
  14. #include <linux/mm_types.h>
  15. #include <linux/range.h>
  16. #include <linux/pfn.h>
  17. #include <linux/percpu-refcount.h>
  18. #include <linux/bit_spinlock.h>
  19. #include <linux/shrinker.h>
  20. #include <linux/resource.h>
  21. #include <linux/page_ext.h>
  22. #include <linux/err.h>
  23. #include <linux/page_ref.h>
  24. #include <linux/memremap.h>
  25. struct mempolicy;
  26. struct anon_vma;
  27. struct anon_vma_chain;
  28. struct file_ra_state;
  29. struct user_struct;
  30. struct writeback_control;
  31. struct bdi_writeback;
  32. void init_mm_internals(void);
  33. #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
  34. extern unsigned long max_mapnr;
  35. static inline void set_max_mapnr(unsigned long limit)
  36. {
  37. max_mapnr = limit;
  38. }
  39. #else
  40. static inline void set_max_mapnr(unsigned long limit) { }
  41. #endif
  42. extern unsigned long totalram_pages;
  43. extern void * high_memory;
  44. extern int page_cluster;
  45. #ifdef CONFIG_SYSCTL
  46. extern int sysctl_legacy_va_layout;
  47. #else
  48. #define sysctl_legacy_va_layout 0
  49. #endif
  50. #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
  51. extern const int mmap_rnd_bits_min;
  52. extern const int mmap_rnd_bits_max;
  53. extern int mmap_rnd_bits __read_mostly;
  54. #endif
  55. #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
  56. extern const int mmap_rnd_compat_bits_min;
  57. extern const int mmap_rnd_compat_bits_max;
  58. extern int mmap_rnd_compat_bits __read_mostly;
  59. #endif
  60. #include <asm/page.h>
  61. #include <asm/pgtable.h>
  62. #include <asm/processor.h>
  63. #ifndef __pa_symbol
  64. #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
  65. #endif
  66. #ifndef page_to_virt
  67. #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
  68. #endif
  69. #ifndef lm_alias
  70. #define lm_alias(x) __va(__pa_symbol(x))
  71. #endif
  72. /*
  73. * To prevent common memory management code establishing
  74. * a zero page mapping on a read fault.
  75. * This macro should be defined within <asm/pgtable.h>.
  76. * s390 does this to prevent multiplexing of hardware bits
  77. * related to the physical page in case of virtualization.
  78. */
  79. #ifndef mm_forbids_zeropage
  80. #define mm_forbids_zeropage(X) (0)
  81. #endif
  82. /*
  83. * On some architectures it is expensive to call memset() for small sizes.
  84. * Those architectures should provide their own implementation of "struct page"
  85. * zeroing by defining this macro in <asm/pgtable.h>.
  86. */
  87. #ifndef mm_zero_struct_page
  88. #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
  89. #endif
  90. /*
  91. * Default maximum number of active map areas, this limits the number of vmas
  92. * per mm struct. Users can overwrite this number by sysctl but there is a
  93. * problem.
  94. *
  95. * When a program's coredump is generated as ELF format, a section is created
  96. * per a vma. In ELF, the number of sections is represented in unsigned short.
  97. * This means the number of sections should be smaller than 65535 at coredump.
  98. * Because the kernel adds some informative sections to a image of program at
  99. * generating coredump, we need some margin. The number of extra sections is
  100. * 1-3 now and depends on arch. We use "5" as safe margin, here.
  101. *
  102. * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
  103. * not a hard limit any more. Although some userspace tools can be surprised by
  104. * that.
  105. */
  106. #define MAPCOUNT_ELF_CORE_MARGIN (5)
  107. #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
  108. extern int sysctl_max_map_count;
  109. extern unsigned long sysctl_user_reserve_kbytes;
  110. extern unsigned long sysctl_admin_reserve_kbytes;
  111. extern int sysctl_overcommit_memory;
  112. extern int sysctl_overcommit_ratio;
  113. extern unsigned long sysctl_overcommit_kbytes;
  114. extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
  115. size_t *, loff_t *);
  116. extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
  117. size_t *, loff_t *);
  118. #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  119. /* to align the pointer to the (next) page boundary */
  120. #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  121. /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
  122. #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
  123. /*
  124. * Linux kernel virtual memory manager primitives.
  125. * The idea being to have a "virtual" mm in the same way
  126. * we have a virtual fs - giving a cleaner interface to the
  127. * mm details, and allowing different kinds of memory mappings
  128. * (from shared memory to executable loading to arbitrary
  129. * mmap() functions).
  130. */
  131. extern struct kmem_cache *vm_area_cachep;
  132. #ifndef CONFIG_MMU
  133. extern struct rb_root nommu_region_tree;
  134. extern struct rw_semaphore nommu_region_sem;
  135. extern unsigned int kobjsize(const void *objp);
  136. #endif
  137. /*
  138. * vm_flags in vm_area_struct, see mm_types.h.
  139. * When changing, update also include/trace/events/mmflags.h
  140. */
  141. #define VM_NONE 0x00000000
  142. #define VM_READ 0x00000001 /* currently active flags */
  143. #define VM_WRITE 0x00000002
  144. #define VM_EXEC 0x00000004
  145. #define VM_SHARED 0x00000008
  146. /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  147. #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
  148. #define VM_MAYWRITE 0x00000020
  149. #define VM_MAYEXEC 0x00000040
  150. #define VM_MAYSHARE 0x00000080
  151. #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
  152. #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
  153. #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
  154. #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
  155. #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
  156. #define VM_LOCKED 0x00002000
  157. #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
  158. /* Used by sys_madvise() */
  159. #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
  160. #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
  161. #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
  162. #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
  163. #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
  164. #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
  165. #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
  166. #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
  167. #define VM_SYNC 0x00800000 /* Synchronous page faults */
  168. #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
  169. #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
  170. #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
  171. #ifdef CONFIG_MEM_SOFT_DIRTY
  172. # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
  173. #else
  174. # define VM_SOFTDIRTY 0
  175. #endif
  176. #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
  177. #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
  178. #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
  179. #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
  180. #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
  181. #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
  182. #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
  183. #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
  184. #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
  185. #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
  186. #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
  187. #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
  188. #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
  189. #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
  190. #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
  191. #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
  192. #if defined(CONFIG_X86)
  193. # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
  194. #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
  195. # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
  196. # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
  197. # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
  198. # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
  199. # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
  200. #endif
  201. #elif defined(CONFIG_PPC)
  202. # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
  203. #elif defined(CONFIG_PARISC)
  204. # define VM_GROWSUP VM_ARCH_1
  205. #elif defined(CONFIG_METAG)
  206. # define VM_GROWSUP VM_ARCH_1
  207. #elif defined(CONFIG_IA64)
  208. # define VM_GROWSUP VM_ARCH_1
  209. #elif !defined(CONFIG_MMU)
  210. # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
  211. #endif
  212. #if defined(CONFIG_X86_INTEL_MPX)
  213. /* MPX specific bounds table or bounds directory */
  214. # define VM_MPX VM_HIGH_ARCH_4
  215. #else
  216. # define VM_MPX VM_NONE
  217. #endif
  218. #ifndef VM_GROWSUP
  219. # define VM_GROWSUP VM_NONE
  220. #endif
  221. /* Bits set in the VMA until the stack is in its final location */
  222. #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
  223. #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
  224. #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
  225. #endif
  226. #ifdef CONFIG_STACK_GROWSUP
  227. #define VM_STACK VM_GROWSUP
  228. #else
  229. #define VM_STACK VM_GROWSDOWN
  230. #endif
  231. #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
  232. /*
  233. * Special vmas that are non-mergable, non-mlock()able.
  234. * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
  235. */
  236. #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
  237. /* This mask defines which mm->def_flags a process can inherit its parent */
  238. #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
  239. /* This mask is used to clear all the VMA flags used by mlock */
  240. #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
  241. /*
  242. * mapping from the currently active vm_flags protection bits (the
  243. * low four bits) to a page protection mask..
  244. */
  245. extern pgprot_t protection_map[16];
  246. #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
  247. #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
  248. #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
  249. #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
  250. #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
  251. #define FAULT_FLAG_TRIED 0x20 /* Second try */
  252. #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
  253. #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
  254. #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
  255. #define FAULT_FLAG_TRACE \
  256. { FAULT_FLAG_WRITE, "WRITE" }, \
  257. { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
  258. { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
  259. { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
  260. { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
  261. { FAULT_FLAG_TRIED, "TRIED" }, \
  262. { FAULT_FLAG_USER, "USER" }, \
  263. { FAULT_FLAG_REMOTE, "REMOTE" }, \
  264. { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
  265. /*
  266. * vm_fault is filled by the the pagefault handler and passed to the vma's
  267. * ->fault function. The vma's ->fault is responsible for returning a bitmask
  268. * of VM_FAULT_xxx flags that give details about how the fault was handled.
  269. *
  270. * MM layer fills up gfp_mask for page allocations but fault handler might
  271. * alter it if its implementation requires a different allocation context.
  272. *
  273. * pgoff should be used in favour of virtual_address, if possible.
  274. */
  275. struct vm_fault {
  276. struct vm_area_struct *vma; /* Target VMA */
  277. unsigned int flags; /* FAULT_FLAG_xxx flags */
  278. gfp_t gfp_mask; /* gfp mask to be used for allocations */
  279. pgoff_t pgoff; /* Logical page offset based on vma */
  280. unsigned long address; /* Faulting virtual address */
  281. pmd_t *pmd; /* Pointer to pmd entry matching
  282. * the 'address' */
  283. pud_t *pud; /* Pointer to pud entry matching
  284. * the 'address'
  285. */
  286. pte_t orig_pte; /* Value of PTE at the time of fault */
  287. struct page *cow_page; /* Page handler may use for COW fault */
  288. struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
  289. struct page *page; /* ->fault handlers should return a
  290. * page here, unless VM_FAULT_NOPAGE
  291. * is set (which is also implied by
  292. * VM_FAULT_ERROR).
  293. */
  294. /* These three entries are valid only while holding ptl lock */
  295. pte_t *pte; /* Pointer to pte entry matching
  296. * the 'address'. NULL if the page
  297. * table hasn't been allocated.
  298. */
  299. spinlock_t *ptl; /* Page table lock.
  300. * Protects pte page table if 'pte'
  301. * is not NULL, otherwise pmd.
  302. */
  303. pgtable_t prealloc_pte; /* Pre-allocated pte page table.
  304. * vm_ops->map_pages() calls
  305. * alloc_set_pte() from atomic context.
  306. * do_fault_around() pre-allocates
  307. * page table to avoid allocation from
  308. * atomic context.
  309. */
  310. };
  311. /* page entry size for vm->huge_fault() */
  312. enum page_entry_size {
  313. PE_SIZE_PTE = 0,
  314. PE_SIZE_PMD,
  315. PE_SIZE_PUD,
  316. };
  317. /*
  318. * These are the virtual MM functions - opening of an area, closing and
  319. * unmapping it (needed to keep files on disk up-to-date etc), pointer
  320. * to the functions called when a no-page or a wp-page exception occurs.
  321. */
  322. struct vm_operations_struct {
  323. void (*open)(struct vm_area_struct * area);
  324. void (*close)(struct vm_area_struct * area);
  325. int (*split)(struct vm_area_struct * area, unsigned long addr);
  326. int (*mremap)(struct vm_area_struct * area);
  327. int (*fault)(struct vm_fault *vmf);
  328. int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
  329. void (*map_pages)(struct vm_fault *vmf,
  330. pgoff_t start_pgoff, pgoff_t end_pgoff);
  331. /* notification that a previously read-only page is about to become
  332. * writable, if an error is returned it will cause a SIGBUS */
  333. int (*page_mkwrite)(struct vm_fault *vmf);
  334. /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
  335. int (*pfn_mkwrite)(struct vm_fault *vmf);
  336. /* called by access_process_vm when get_user_pages() fails, typically
  337. * for use by special VMAs that can switch between memory and hardware
  338. */
  339. int (*access)(struct vm_area_struct *vma, unsigned long addr,
  340. void *buf, int len, int write);
  341. /* Called by the /proc/PID/maps code to ask the vma whether it
  342. * has a special name. Returning non-NULL will also cause this
  343. * vma to be dumped unconditionally. */
  344. const char *(*name)(struct vm_area_struct *vma);
  345. #ifdef CONFIG_NUMA
  346. /*
  347. * set_policy() op must add a reference to any non-NULL @new mempolicy
  348. * to hold the policy upon return. Caller should pass NULL @new to
  349. * remove a policy and fall back to surrounding context--i.e. do not
  350. * install a MPOL_DEFAULT policy, nor the task or system default
  351. * mempolicy.
  352. */
  353. int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
  354. /*
  355. * get_policy() op must add reference [mpol_get()] to any policy at
  356. * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
  357. * in mm/mempolicy.c will do this automatically.
  358. * get_policy() must NOT add a ref if the policy at (vma,addr) is not
  359. * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
  360. * If no [shared/vma] mempolicy exists at the addr, get_policy() op
  361. * must return NULL--i.e., do not "fallback" to task or system default
  362. * policy.
  363. */
  364. struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
  365. unsigned long addr);
  366. #endif
  367. /*
  368. * Called by vm_normal_page() for special PTEs to find the
  369. * page for @addr. This is useful if the default behavior
  370. * (using pte_page()) would not find the correct page.
  371. */
  372. struct page *(*find_special_page)(struct vm_area_struct *vma,
  373. unsigned long addr);
  374. };
  375. struct mmu_gather;
  376. struct inode;
  377. #define page_private(page) ((page)->private)
  378. #define set_page_private(page, v) ((page)->private = (v))
  379. #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
  380. static inline int pmd_devmap(pmd_t pmd)
  381. {
  382. return 0;
  383. }
  384. static inline int pud_devmap(pud_t pud)
  385. {
  386. return 0;
  387. }
  388. static inline int pgd_devmap(pgd_t pgd)
  389. {
  390. return 0;
  391. }
  392. #endif
  393. /*
  394. * FIXME: take this include out, include page-flags.h in
  395. * files which need it (119 of them)
  396. */
  397. #include <linux/page-flags.h>
  398. #include <linux/huge_mm.h>
  399. /*
  400. * Methods to modify the page usage count.
  401. *
  402. * What counts for a page usage:
  403. * - cache mapping (page->mapping)
  404. * - private data (page->private)
  405. * - page mapped in a task's page tables, each mapping
  406. * is counted separately
  407. *
  408. * Also, many kernel routines increase the page count before a critical
  409. * routine so they can be sure the page doesn't go away from under them.
  410. */
  411. /*
  412. * Drop a ref, return true if the refcount fell to zero (the page has no users)
  413. */
  414. static inline int put_page_testzero(struct page *page)
  415. {
  416. VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
  417. return page_ref_dec_and_test(page);
  418. }
  419. /*
  420. * Try to grab a ref unless the page has a refcount of zero, return false if
  421. * that is the case.
  422. * This can be called when MMU is off so it must not access
  423. * any of the virtual mappings.
  424. */
  425. static inline int get_page_unless_zero(struct page *page)
  426. {
  427. return page_ref_add_unless(page, 1, 0);
  428. }
  429. extern int page_is_ram(unsigned long pfn);
  430. enum {
  431. REGION_INTERSECTS,
  432. REGION_DISJOINT,
  433. REGION_MIXED,
  434. };
  435. int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
  436. unsigned long desc);
  437. /* Support for virtually mapped pages */
  438. struct page *vmalloc_to_page(const void *addr);
  439. unsigned long vmalloc_to_pfn(const void *addr);
  440. /*
  441. * Determine if an address is within the vmalloc range
  442. *
  443. * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
  444. * is no special casing required.
  445. */
  446. static inline bool is_vmalloc_addr(const void *x)
  447. {
  448. #ifdef CONFIG_MMU
  449. unsigned long addr = (unsigned long)x;
  450. return addr >= VMALLOC_START && addr < VMALLOC_END;
  451. #else
  452. return false;
  453. #endif
  454. }
  455. #ifdef CONFIG_MMU
  456. extern int is_vmalloc_or_module_addr(const void *x);
  457. #else
  458. static inline int is_vmalloc_or_module_addr(const void *x)
  459. {
  460. return 0;
  461. }
  462. #endif
  463. extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
  464. static inline void *kvmalloc(size_t size, gfp_t flags)
  465. {
  466. return kvmalloc_node(size, flags, NUMA_NO_NODE);
  467. }
  468. static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
  469. {
  470. return kvmalloc_node(size, flags | __GFP_ZERO, node);
  471. }
  472. static inline void *kvzalloc(size_t size, gfp_t flags)
  473. {
  474. return kvmalloc(size, flags | __GFP_ZERO);
  475. }
  476. static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
  477. {
  478. if (size != 0 && n > SIZE_MAX / size)
  479. return NULL;
  480. return kvmalloc(n * size, flags);
  481. }
  482. extern void kvfree(const void *addr);
  483. static inline atomic_t *compound_mapcount_ptr(struct page *page)
  484. {
  485. return &page[1].compound_mapcount;
  486. }
  487. static inline int compound_mapcount(struct page *page)
  488. {
  489. VM_BUG_ON_PAGE(!PageCompound(page), page);
  490. page = compound_head(page);
  491. return atomic_read(compound_mapcount_ptr(page)) + 1;
  492. }
  493. /*
  494. * The atomic page->_mapcount, starts from -1: so that transitions
  495. * both from it and to it can be tracked, using atomic_inc_and_test
  496. * and atomic_add_negative(-1).
  497. */
  498. static inline void page_mapcount_reset(struct page *page)
  499. {
  500. atomic_set(&(page)->_mapcount, -1);
  501. }
  502. int __page_mapcount(struct page *page);
  503. static inline int page_mapcount(struct page *page)
  504. {
  505. VM_BUG_ON_PAGE(PageSlab(page), page);
  506. if (unlikely(PageCompound(page)))
  507. return __page_mapcount(page);
  508. return atomic_read(&page->_mapcount) + 1;
  509. }
  510. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  511. int total_mapcount(struct page *page);
  512. int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
  513. #else
  514. static inline int total_mapcount(struct page *page)
  515. {
  516. return page_mapcount(page);
  517. }
  518. static inline int page_trans_huge_mapcount(struct page *page,
  519. int *total_mapcount)
  520. {
  521. int mapcount = page_mapcount(page);
  522. if (total_mapcount)
  523. *total_mapcount = mapcount;
  524. return mapcount;
  525. }
  526. #endif
  527. static inline struct page *virt_to_head_page(const void *x)
  528. {
  529. struct page *page = virt_to_page(x);
  530. return compound_head(page);
  531. }
  532. void __put_page(struct page *page);
  533. void put_pages_list(struct list_head *pages);
  534. void split_page(struct page *page, unsigned int order);
  535. /*
  536. * Compound pages have a destructor function. Provide a
  537. * prototype for that function and accessor functions.
  538. * These are _only_ valid on the head of a compound page.
  539. */
  540. typedef void compound_page_dtor(struct page *);
  541. /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
  542. enum compound_dtor_id {
  543. NULL_COMPOUND_DTOR,
  544. COMPOUND_PAGE_DTOR,
  545. #ifdef CONFIG_HUGETLB_PAGE
  546. HUGETLB_PAGE_DTOR,
  547. #endif
  548. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  549. TRANSHUGE_PAGE_DTOR,
  550. #endif
  551. NR_COMPOUND_DTORS,
  552. };
  553. extern compound_page_dtor * const compound_page_dtors[];
  554. static inline void set_compound_page_dtor(struct page *page,
  555. enum compound_dtor_id compound_dtor)
  556. {
  557. VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
  558. page[1].compound_dtor = compound_dtor;
  559. }
  560. static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
  561. {
  562. VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
  563. return compound_page_dtors[page[1].compound_dtor];
  564. }
  565. static inline unsigned int compound_order(struct page *page)
  566. {
  567. if (!PageHead(page))
  568. return 0;
  569. return page[1].compound_order;
  570. }
  571. static inline void set_compound_order(struct page *page, unsigned int order)
  572. {
  573. page[1].compound_order = order;
  574. }
  575. void free_compound_page(struct page *page);
  576. #ifdef CONFIG_MMU
  577. /*
  578. * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
  579. * servicing faults for write access. In the normal case, do always want
  580. * pte_mkwrite. But get_user_pages can cause write faults for mappings
  581. * that do not have writing enabled, when used by access_process_vm.
  582. */
  583. static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
  584. {
  585. if (likely(vma->vm_flags & VM_WRITE))
  586. pte = pte_mkwrite(pte);
  587. return pte;
  588. }
  589. int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
  590. struct page *page);
  591. int finish_fault(struct vm_fault *vmf);
  592. int finish_mkwrite_fault(struct vm_fault *vmf);
  593. #endif
  594. /*
  595. * Multiple processes may "see" the same page. E.g. for untouched
  596. * mappings of /dev/null, all processes see the same page full of
  597. * zeroes, and text pages of executables and shared libraries have
  598. * only one copy in memory, at most, normally.
  599. *
  600. * For the non-reserved pages, page_count(page) denotes a reference count.
  601. * page_count() == 0 means the page is free. page->lru is then used for
  602. * freelist management in the buddy allocator.
  603. * page_count() > 0 means the page has been allocated.
  604. *
  605. * Pages are allocated by the slab allocator in order to provide memory
  606. * to kmalloc and kmem_cache_alloc. In this case, the management of the
  607. * page, and the fields in 'struct page' are the responsibility of mm/slab.c
  608. * unless a particular usage is carefully commented. (the responsibility of
  609. * freeing the kmalloc memory is the caller's, of course).
  610. *
  611. * A page may be used by anyone else who does a __get_free_page().
  612. * In this case, page_count still tracks the references, and should only
  613. * be used through the normal accessor functions. The top bits of page->flags
  614. * and page->virtual store page management information, but all other fields
  615. * are unused and could be used privately, carefully. The management of this
  616. * page is the responsibility of the one who allocated it, and those who have
  617. * subsequently been given references to it.
  618. *
  619. * The other pages (we may call them "pagecache pages") are completely
  620. * managed by the Linux memory manager: I/O, buffers, swapping etc.
  621. * The following discussion applies only to them.
  622. *
  623. * A pagecache page contains an opaque `private' member, which belongs to the
  624. * page's address_space. Usually, this is the address of a circular list of
  625. * the page's disk buffers. PG_private must be set to tell the VM to call
  626. * into the filesystem to release these pages.
  627. *
  628. * A page may belong to an inode's memory mapping. In this case, page->mapping
  629. * is the pointer to the inode, and page->index is the file offset of the page,
  630. * in units of PAGE_SIZE.
  631. *
  632. * If pagecache pages are not associated with an inode, they are said to be
  633. * anonymous pages. These may become associated with the swapcache, and in that
  634. * case PG_swapcache is set, and page->private is an offset into the swapcache.
  635. *
  636. * In either case (swapcache or inode backed), the pagecache itself holds one
  637. * reference to the page. Setting PG_private should also increment the
  638. * refcount. The each user mapping also has a reference to the page.
  639. *
  640. * The pagecache pages are stored in a per-mapping radix tree, which is
  641. * rooted at mapping->page_tree, and indexed by offset.
  642. * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
  643. * lists, we instead now tag pages as dirty/writeback in the radix tree.
  644. *
  645. * All pagecache pages may be subject to I/O:
  646. * - inode pages may need to be read from disk,
  647. * - inode pages which have been modified and are MAP_SHARED may need
  648. * to be written back to the inode on disk,
  649. * - anonymous pages (including MAP_PRIVATE file mappings) which have been
  650. * modified may need to be swapped out to swap space and (later) to be read
  651. * back into memory.
  652. */
  653. /*
  654. * The zone field is never updated after free_area_init_core()
  655. * sets it, so none of the operations on it need to be atomic.
  656. */
  657. /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
  658. #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
  659. #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
  660. #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
  661. #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
  662. /*
  663. * Define the bit shifts to access each section. For non-existent
  664. * sections we define the shift as 0; that plus a 0 mask ensures
  665. * the compiler will optimise away reference to them.
  666. */
  667. #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
  668. #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
  669. #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
  670. #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
  671. /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
  672. #ifdef NODE_NOT_IN_PAGE_FLAGS
  673. #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
  674. #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
  675. SECTIONS_PGOFF : ZONES_PGOFF)
  676. #else
  677. #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
  678. #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
  679. NODES_PGOFF : ZONES_PGOFF)
  680. #endif
  681. #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
  682. #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
  683. #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
  684. #endif
  685. #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
  686. #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
  687. #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
  688. #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
  689. #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
  690. static inline enum zone_type page_zonenum(const struct page *page)
  691. {
  692. return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
  693. }
  694. #ifdef CONFIG_ZONE_DEVICE
  695. static inline bool is_zone_device_page(const struct page *page)
  696. {
  697. return page_zonenum(page) == ZONE_DEVICE;
  698. }
  699. #else
  700. static inline bool is_zone_device_page(const struct page *page)
  701. {
  702. return false;
  703. }
  704. #endif
  705. #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
  706. void put_zone_device_private_or_public_page(struct page *page);
  707. DECLARE_STATIC_KEY_FALSE(device_private_key);
  708. #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
  709. static inline bool is_device_private_page(const struct page *page);
  710. static inline bool is_device_public_page(const struct page *page);
  711. #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
  712. static inline void put_zone_device_private_or_public_page(struct page *page)
  713. {
  714. }
  715. #define IS_HMM_ENABLED 0
  716. static inline bool is_device_private_page(const struct page *page)
  717. {
  718. return false;
  719. }
  720. static inline bool is_device_public_page(const struct page *page)
  721. {
  722. return false;
  723. }
  724. #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
  725. static inline void get_page(struct page *page)
  726. {
  727. page = compound_head(page);
  728. /*
  729. * Getting a normal page or the head of a compound page
  730. * requires to already have an elevated page->_refcount.
  731. */
  732. VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
  733. page_ref_inc(page);
  734. }
  735. static inline void put_page(struct page *page)
  736. {
  737. page = compound_head(page);
  738. /*
  739. * For private device pages we need to catch refcount transition from
  740. * 2 to 1, when refcount reach one it means the private device page is
  741. * free and we need to inform the device driver through callback. See
  742. * include/linux/memremap.h and HMM for details.
  743. */
  744. if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
  745. unlikely(is_device_public_page(page)))) {
  746. put_zone_device_private_or_public_page(page);
  747. return;
  748. }
  749. if (put_page_testzero(page))
  750. __put_page(page);
  751. }
  752. #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
  753. #define SECTION_IN_PAGE_FLAGS
  754. #endif
  755. /*
  756. * The identification function is mainly used by the buddy allocator for
  757. * determining if two pages could be buddies. We are not really identifying
  758. * the zone since we could be using the section number id if we do not have
  759. * node id available in page flags.
  760. * We only guarantee that it will return the same value for two combinable
  761. * pages in a zone.
  762. */
  763. static inline int page_zone_id(struct page *page)
  764. {
  765. return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
  766. }
  767. static inline int zone_to_nid(struct zone *zone)
  768. {
  769. #ifdef CONFIG_NUMA
  770. return zone->node;
  771. #else
  772. return 0;
  773. #endif
  774. }
  775. #ifdef NODE_NOT_IN_PAGE_FLAGS
  776. extern int page_to_nid(const struct page *page);
  777. #else
  778. static inline int page_to_nid(const struct page *page)
  779. {
  780. return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
  781. }
  782. #endif
  783. #ifdef CONFIG_NUMA_BALANCING
  784. static inline int cpu_pid_to_cpupid(int cpu, int pid)
  785. {
  786. return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
  787. }
  788. static inline int cpupid_to_pid(int cpupid)
  789. {
  790. return cpupid & LAST__PID_MASK;
  791. }
  792. static inline int cpupid_to_cpu(int cpupid)
  793. {
  794. return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
  795. }
  796. static inline int cpupid_to_nid(int cpupid)
  797. {
  798. return cpu_to_node(cpupid_to_cpu(cpupid));
  799. }
  800. static inline bool cpupid_pid_unset(int cpupid)
  801. {
  802. return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
  803. }
  804. static inline bool cpupid_cpu_unset(int cpupid)
  805. {
  806. return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
  807. }
  808. static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
  809. {
  810. return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
  811. }
  812. #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
  813. #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
  814. static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
  815. {
  816. return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
  817. }
  818. static inline int page_cpupid_last(struct page *page)
  819. {
  820. return page->_last_cpupid;
  821. }
  822. static inline void page_cpupid_reset_last(struct page *page)
  823. {
  824. page->_last_cpupid = -1 & LAST_CPUPID_MASK;
  825. }
  826. #else
  827. static inline int page_cpupid_last(struct page *page)
  828. {
  829. return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
  830. }
  831. extern int page_cpupid_xchg_last(struct page *page, int cpupid);
  832. static inline void page_cpupid_reset_last(struct page *page)
  833. {
  834. page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
  835. }
  836. #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
  837. #else /* !CONFIG_NUMA_BALANCING */
  838. static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
  839. {
  840. return page_to_nid(page); /* XXX */
  841. }
  842. static inline int page_cpupid_last(struct page *page)
  843. {
  844. return page_to_nid(page); /* XXX */
  845. }
  846. static inline int cpupid_to_nid(int cpupid)
  847. {
  848. return -1;
  849. }
  850. static inline int cpupid_to_pid(int cpupid)
  851. {
  852. return -1;
  853. }
  854. static inline int cpupid_to_cpu(int cpupid)
  855. {
  856. return -1;
  857. }
  858. static inline int cpu_pid_to_cpupid(int nid, int pid)
  859. {
  860. return -1;
  861. }
  862. static inline bool cpupid_pid_unset(int cpupid)
  863. {
  864. return 1;
  865. }
  866. static inline void page_cpupid_reset_last(struct page *page)
  867. {
  868. }
  869. static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
  870. {
  871. return false;
  872. }
  873. #endif /* CONFIG_NUMA_BALANCING */
  874. static inline struct zone *page_zone(const struct page *page)
  875. {
  876. return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
  877. }
  878. static inline pg_data_t *page_pgdat(const struct page *page)
  879. {
  880. return NODE_DATA(page_to_nid(page));
  881. }
  882. #ifdef SECTION_IN_PAGE_FLAGS
  883. static inline void set_page_section(struct page *page, unsigned long section)
  884. {
  885. page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
  886. page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
  887. }
  888. static inline unsigned long page_to_section(const struct page *page)
  889. {
  890. return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
  891. }
  892. #endif
  893. static inline void set_page_zone(struct page *page, enum zone_type zone)
  894. {
  895. page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
  896. page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
  897. }
  898. static inline void set_page_node(struct page *page, unsigned long node)
  899. {
  900. page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
  901. page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
  902. }
  903. static inline void set_page_links(struct page *page, enum zone_type zone,
  904. unsigned long node, unsigned long pfn)
  905. {
  906. set_page_zone(page, zone);
  907. set_page_node(page, node);
  908. #ifdef SECTION_IN_PAGE_FLAGS
  909. set_page_section(page, pfn_to_section_nr(pfn));
  910. #endif
  911. }
  912. #ifdef CONFIG_MEMCG
  913. static inline struct mem_cgroup *page_memcg(struct page *page)
  914. {
  915. return page->mem_cgroup;
  916. }
  917. static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
  918. {
  919. WARN_ON_ONCE(!rcu_read_lock_held());
  920. return READ_ONCE(page->mem_cgroup);
  921. }
  922. #else
  923. static inline struct mem_cgroup *page_memcg(struct page *page)
  924. {
  925. return NULL;
  926. }
  927. static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
  928. {
  929. WARN_ON_ONCE(!rcu_read_lock_held());
  930. return NULL;
  931. }
  932. #endif
  933. /*
  934. * Some inline functions in vmstat.h depend on page_zone()
  935. */
  936. #include <linux/vmstat.h>
  937. static __always_inline void *lowmem_page_address(const struct page *page)
  938. {
  939. return page_to_virt(page);
  940. }
  941. #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
  942. #define HASHED_PAGE_VIRTUAL
  943. #endif
  944. #if defined(WANT_PAGE_VIRTUAL)
  945. static inline void *page_address(const struct page *page)
  946. {
  947. return page->virtual;
  948. }
  949. static inline void set_page_address(struct page *page, void *address)
  950. {
  951. page->virtual = address;
  952. }
  953. #define page_address_init() do { } while(0)
  954. #endif
  955. #if defined(HASHED_PAGE_VIRTUAL)
  956. void *page_address(const struct page *page);
  957. void set_page_address(struct page *page, void *virtual);
  958. void page_address_init(void);
  959. #endif
  960. #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
  961. #define page_address(page) lowmem_page_address(page)
  962. #define set_page_address(page, address) do { } while(0)
  963. #define page_address_init() do { } while(0)
  964. #endif
  965. extern void *page_rmapping(struct page *page);
  966. extern struct anon_vma *page_anon_vma(struct page *page);
  967. extern struct address_space *page_mapping(struct page *page);
  968. extern struct address_space *__page_file_mapping(struct page *);
  969. static inline
  970. struct address_space *page_file_mapping(struct page *page)
  971. {
  972. if (unlikely(PageSwapCache(page)))
  973. return __page_file_mapping(page);
  974. return page->mapping;
  975. }
  976. extern pgoff_t __page_file_index(struct page *page);
  977. /*
  978. * Return the pagecache index of the passed page. Regular pagecache pages
  979. * use ->index whereas swapcache pages use swp_offset(->private)
  980. */
  981. static inline pgoff_t page_index(struct page *page)
  982. {
  983. if (unlikely(PageSwapCache(page)))
  984. return __page_file_index(page);
  985. return page->index;
  986. }
  987. bool page_mapped(struct page *page);
  988. struct address_space *page_mapping(struct page *page);
  989. /*
  990. * Return true only if the page has been allocated with
  991. * ALLOC_NO_WATERMARKS and the low watermark was not
  992. * met implying that the system is under some pressure.
  993. */
  994. static inline bool page_is_pfmemalloc(struct page *page)
  995. {
  996. /*
  997. * Page index cannot be this large so this must be
  998. * a pfmemalloc page.
  999. */
  1000. return page->index == -1UL;
  1001. }
  1002. /*
  1003. * Only to be called by the page allocator on a freshly allocated
  1004. * page.
  1005. */
  1006. static inline void set_page_pfmemalloc(struct page *page)
  1007. {
  1008. page->index = -1UL;
  1009. }
  1010. static inline void clear_page_pfmemalloc(struct page *page)
  1011. {
  1012. page->index = 0;
  1013. }
  1014. /*
  1015. * Different kinds of faults, as returned by handle_mm_fault().
  1016. * Used to decide whether a process gets delivered SIGBUS or
  1017. * just gets major/minor fault counters bumped up.
  1018. */
  1019. #define VM_FAULT_OOM 0x0001
  1020. #define VM_FAULT_SIGBUS 0x0002
  1021. #define VM_FAULT_MAJOR 0x0004
  1022. #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
  1023. #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
  1024. #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
  1025. #define VM_FAULT_SIGSEGV 0x0040
  1026. #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
  1027. #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
  1028. #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
  1029. #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
  1030. #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
  1031. #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
  1032. * and needs fsync() to complete (for
  1033. * synchronous page faults in DAX) */
  1034. #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
  1035. VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
  1036. VM_FAULT_FALLBACK)
  1037. #define VM_FAULT_RESULT_TRACE \
  1038. { VM_FAULT_OOM, "OOM" }, \
  1039. { VM_FAULT_SIGBUS, "SIGBUS" }, \
  1040. { VM_FAULT_MAJOR, "MAJOR" }, \
  1041. { VM_FAULT_WRITE, "WRITE" }, \
  1042. { VM_FAULT_HWPOISON, "HWPOISON" }, \
  1043. { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
  1044. { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
  1045. { VM_FAULT_NOPAGE, "NOPAGE" }, \
  1046. { VM_FAULT_LOCKED, "LOCKED" }, \
  1047. { VM_FAULT_RETRY, "RETRY" }, \
  1048. { VM_FAULT_FALLBACK, "FALLBACK" }, \
  1049. { VM_FAULT_DONE_COW, "DONE_COW" }, \
  1050. { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
  1051. /* Encode hstate index for a hwpoisoned large page */
  1052. #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
  1053. #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
  1054. /*
  1055. * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
  1056. */
  1057. extern void pagefault_out_of_memory(void);
  1058. #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
  1059. /*
  1060. * Flags passed to show_mem() and show_free_areas() to suppress output in
  1061. * various contexts.
  1062. */
  1063. #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
  1064. extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
  1065. extern bool can_do_mlock(void);
  1066. extern int user_shm_lock(size_t, struct user_struct *);
  1067. extern void user_shm_unlock(size_t, struct user_struct *);
  1068. /*
  1069. * Parameter block passed down to zap_pte_range in exceptional cases.
  1070. */
  1071. struct zap_details {
  1072. struct address_space *check_mapping; /* Check page->mapping if set */
  1073. pgoff_t first_index; /* Lowest page->index to unmap */
  1074. pgoff_t last_index; /* Highest page->index to unmap */
  1075. };
  1076. struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
  1077. pte_t pte, bool with_public_device);
  1078. #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
  1079. struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
  1080. pmd_t pmd);
  1081. int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
  1082. unsigned long size);
  1083. void zap_page_range(struct vm_area_struct *vma, unsigned long address,
  1084. unsigned long size);
  1085. void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  1086. unsigned long start, unsigned long end);
  1087. /**
  1088. * mm_walk - callbacks for walk_page_range
  1089. * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
  1090. * this handler should only handle pud_trans_huge() puds.
  1091. * the pmd_entry or pte_entry callbacks will be used for
  1092. * regular PUDs.
  1093. * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
  1094. * this handler is required to be able to handle
  1095. * pmd_trans_huge() pmds. They may simply choose to
  1096. * split_huge_page() instead of handling it explicitly.
  1097. * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
  1098. * @pte_hole: if set, called for each hole at all levels
  1099. * @hugetlb_entry: if set, called for each hugetlb entry
  1100. * @test_walk: caller specific callback function to determine whether
  1101. * we walk over the current vma or not. Returning 0
  1102. * value means "do page table walk over the current vma,"
  1103. * and a negative one means "abort current page table walk
  1104. * right now." 1 means "skip the current vma."
  1105. * @mm: mm_struct representing the target process of page table walk
  1106. * @vma: vma currently walked (NULL if walking outside vmas)
  1107. * @private: private data for callbacks' usage
  1108. *
  1109. * (see the comment on walk_page_range() for more details)
  1110. */
  1111. struct mm_walk {
  1112. int (*pud_entry)(pud_t *pud, unsigned long addr,
  1113. unsigned long next, struct mm_walk *walk);
  1114. int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
  1115. unsigned long next, struct mm_walk *walk);
  1116. int (*pte_entry)(pte_t *pte, unsigned long addr,
  1117. unsigned long next, struct mm_walk *walk);
  1118. int (*pte_hole)(unsigned long addr, unsigned long next,
  1119. struct mm_walk *walk);
  1120. int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
  1121. unsigned long addr, unsigned long next,
  1122. struct mm_walk *walk);
  1123. int (*test_walk)(unsigned long addr, unsigned long next,
  1124. struct mm_walk *walk);
  1125. struct mm_struct *mm;
  1126. struct vm_area_struct *vma;
  1127. void *private;
  1128. };
  1129. int walk_page_range(unsigned long addr, unsigned long end,
  1130. struct mm_walk *walk);
  1131. int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
  1132. void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
  1133. unsigned long end, unsigned long floor, unsigned long ceiling);
  1134. int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
  1135. struct vm_area_struct *vma);
  1136. void unmap_mapping_range(struct address_space *mapping,
  1137. loff_t const holebegin, loff_t const holelen, int even_cows);
  1138. int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
  1139. unsigned long *start, unsigned long *end,
  1140. pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
  1141. int follow_pfn(struct vm_area_struct *vma, unsigned long address,
  1142. unsigned long *pfn);
  1143. int follow_phys(struct vm_area_struct *vma, unsigned long address,
  1144. unsigned int flags, unsigned long *prot, resource_size_t *phys);
  1145. int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
  1146. void *buf, int len, int write);
  1147. static inline void unmap_shared_mapping_range(struct address_space *mapping,
  1148. loff_t const holebegin, loff_t const holelen)
  1149. {
  1150. unmap_mapping_range(mapping, holebegin, holelen, 0);
  1151. }
  1152. extern void truncate_pagecache(struct inode *inode, loff_t new);
  1153. extern void truncate_setsize(struct inode *inode, loff_t newsize);
  1154. void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
  1155. void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
  1156. int truncate_inode_page(struct address_space *mapping, struct page *page);
  1157. int generic_error_remove_page(struct address_space *mapping, struct page *page);
  1158. int invalidate_inode_page(struct page *page);
  1159. #ifdef CONFIG_MMU
  1160. extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
  1161. unsigned int flags);
  1162. extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
  1163. unsigned long address, unsigned int fault_flags,
  1164. bool *unlocked);
  1165. #else
  1166. static inline int handle_mm_fault(struct vm_area_struct *vma,
  1167. unsigned long address, unsigned int flags)
  1168. {
  1169. /* should never happen if there's no MMU */
  1170. BUG();
  1171. return VM_FAULT_SIGBUS;
  1172. }
  1173. static inline int fixup_user_fault(struct task_struct *tsk,
  1174. struct mm_struct *mm, unsigned long address,
  1175. unsigned int fault_flags, bool *unlocked)
  1176. {
  1177. /* should never happen if there's no MMU */
  1178. BUG();
  1179. return -EFAULT;
  1180. }
  1181. #endif
  1182. extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
  1183. unsigned int gup_flags);
  1184. extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
  1185. void *buf, int len, unsigned int gup_flags);
  1186. extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
  1187. unsigned long addr, void *buf, int len, unsigned int gup_flags);
  1188. long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
  1189. unsigned long start, unsigned long nr_pages,
  1190. unsigned int gup_flags, struct page **pages,
  1191. struct vm_area_struct **vmas, int *locked);
  1192. long get_user_pages(unsigned long start, unsigned long nr_pages,
  1193. unsigned int gup_flags, struct page **pages,
  1194. struct vm_area_struct **vmas);
  1195. long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
  1196. unsigned int gup_flags, struct page **pages, int *locked);
  1197. long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
  1198. struct page **pages, unsigned int gup_flags);
  1199. int get_user_pages_fast(unsigned long start, int nr_pages, int write,
  1200. struct page **pages);
  1201. /* Container for pinned pfns / pages */
  1202. struct frame_vector {
  1203. unsigned int nr_allocated; /* Number of frames we have space for */
  1204. unsigned int nr_frames; /* Number of frames stored in ptrs array */
  1205. bool got_ref; /* Did we pin pages by getting page ref? */
  1206. bool is_pfns; /* Does array contain pages or pfns? */
  1207. void *ptrs[0]; /* Array of pinned pfns / pages. Use
  1208. * pfns_vector_pages() or pfns_vector_pfns()
  1209. * for access */
  1210. };
  1211. struct frame_vector *frame_vector_create(unsigned int nr_frames);
  1212. void frame_vector_destroy(struct frame_vector *vec);
  1213. int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
  1214. unsigned int gup_flags, struct frame_vector *vec);
  1215. void put_vaddr_frames(struct frame_vector *vec);
  1216. int frame_vector_to_pages(struct frame_vector *vec);
  1217. void frame_vector_to_pfns(struct frame_vector *vec);
  1218. static inline unsigned int frame_vector_count(struct frame_vector *vec)
  1219. {
  1220. return vec->nr_frames;
  1221. }
  1222. static inline struct page **frame_vector_pages(struct frame_vector *vec)
  1223. {
  1224. if (vec->is_pfns) {
  1225. int err = frame_vector_to_pages(vec);
  1226. if (err)
  1227. return ERR_PTR(err);
  1228. }
  1229. return (struct page **)(vec->ptrs);
  1230. }
  1231. static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
  1232. {
  1233. if (!vec->is_pfns)
  1234. frame_vector_to_pfns(vec);
  1235. return (unsigned long *)(vec->ptrs);
  1236. }
  1237. struct kvec;
  1238. int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
  1239. struct page **pages);
  1240. int get_kernel_page(unsigned long start, int write, struct page **pages);
  1241. struct page *get_dump_page(unsigned long addr);
  1242. extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
  1243. extern void do_invalidatepage(struct page *page, unsigned int offset,
  1244. unsigned int length);
  1245. int __set_page_dirty_nobuffers(struct page *page);
  1246. int __set_page_dirty_no_writeback(struct page *page);
  1247. int redirty_page_for_writepage(struct writeback_control *wbc,
  1248. struct page *page);
  1249. void account_page_dirtied(struct page *page, struct address_space *mapping);
  1250. void account_page_cleaned(struct page *page, struct address_space *mapping,
  1251. struct bdi_writeback *wb);
  1252. int set_page_dirty(struct page *page);
  1253. int set_page_dirty_lock(struct page *page);
  1254. void __cancel_dirty_page(struct page *page);
  1255. static inline void cancel_dirty_page(struct page *page)
  1256. {
  1257. /* Avoid atomic ops, locking, etc. when not actually needed. */
  1258. if (PageDirty(page))
  1259. __cancel_dirty_page(page);
  1260. }
  1261. int clear_page_dirty_for_io(struct page *page);
  1262. int get_cmdline(struct task_struct *task, char *buffer, int buflen);
  1263. static inline bool vma_is_anonymous(struct vm_area_struct *vma)
  1264. {
  1265. return !vma->vm_ops;
  1266. }
  1267. #ifdef CONFIG_SHMEM
  1268. /*
  1269. * The vma_is_shmem is not inline because it is used only by slow
  1270. * paths in userfault.
  1271. */
  1272. bool vma_is_shmem(struct vm_area_struct *vma);
  1273. #else
  1274. static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
  1275. #endif
  1276. int vma_is_stack_for_current(struct vm_area_struct *vma);
  1277. extern unsigned long move_page_tables(struct vm_area_struct *vma,
  1278. unsigned long old_addr, struct vm_area_struct *new_vma,
  1279. unsigned long new_addr, unsigned long len,
  1280. bool need_rmap_locks);
  1281. extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
  1282. unsigned long end, pgprot_t newprot,
  1283. int dirty_accountable, int prot_numa);
  1284. extern int mprotect_fixup(struct vm_area_struct *vma,
  1285. struct vm_area_struct **pprev, unsigned long start,
  1286. unsigned long end, unsigned long newflags);
  1287. /*
  1288. * doesn't attempt to fault and will return short.
  1289. */
  1290. int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
  1291. struct page **pages);
  1292. /*
  1293. * per-process(per-mm_struct) statistics.
  1294. */
  1295. static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
  1296. {
  1297. long val = atomic_long_read(&mm->rss_stat.count[member]);
  1298. #ifdef SPLIT_RSS_COUNTING
  1299. /*
  1300. * counter is updated in asynchronous manner and may go to minus.
  1301. * But it's never be expected number for users.
  1302. */
  1303. if (val < 0)
  1304. val = 0;
  1305. #endif
  1306. return (unsigned long)val;
  1307. }
  1308. static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
  1309. {
  1310. atomic_long_add(value, &mm->rss_stat.count[member]);
  1311. }
  1312. static inline void inc_mm_counter(struct mm_struct *mm, int member)
  1313. {
  1314. atomic_long_inc(&mm->rss_stat.count[member]);
  1315. }
  1316. static inline void dec_mm_counter(struct mm_struct *mm, int member)
  1317. {
  1318. atomic_long_dec(&mm->rss_stat.count[member]);
  1319. }
  1320. /* Optimized variant when page is already known not to be PageAnon */
  1321. static inline int mm_counter_file(struct page *page)
  1322. {
  1323. if (PageSwapBacked(page))
  1324. return MM_SHMEMPAGES;
  1325. return MM_FILEPAGES;
  1326. }
  1327. static inline int mm_counter(struct page *page)
  1328. {
  1329. if (PageAnon(page))
  1330. return MM_ANONPAGES;
  1331. return mm_counter_file(page);
  1332. }
  1333. static inline unsigned long get_mm_rss(struct mm_struct *mm)
  1334. {
  1335. return get_mm_counter(mm, MM_FILEPAGES) +
  1336. get_mm_counter(mm, MM_ANONPAGES) +
  1337. get_mm_counter(mm, MM_SHMEMPAGES);
  1338. }
  1339. static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
  1340. {
  1341. return max(mm->hiwater_rss, get_mm_rss(mm));
  1342. }
  1343. static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
  1344. {
  1345. return max(mm->hiwater_vm, mm->total_vm);
  1346. }
  1347. static inline void update_hiwater_rss(struct mm_struct *mm)
  1348. {
  1349. unsigned long _rss = get_mm_rss(mm);
  1350. if ((mm)->hiwater_rss < _rss)
  1351. (mm)->hiwater_rss = _rss;
  1352. }
  1353. static inline void update_hiwater_vm(struct mm_struct *mm)
  1354. {
  1355. if (mm->hiwater_vm < mm->total_vm)
  1356. mm->hiwater_vm = mm->total_vm;
  1357. }
  1358. static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
  1359. {
  1360. mm->hiwater_rss = get_mm_rss(mm);
  1361. }
  1362. static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
  1363. struct mm_struct *mm)
  1364. {
  1365. unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
  1366. if (*maxrss < hiwater_rss)
  1367. *maxrss = hiwater_rss;
  1368. }
  1369. #if defined(SPLIT_RSS_COUNTING)
  1370. void sync_mm_rss(struct mm_struct *mm);
  1371. #else
  1372. static inline void sync_mm_rss(struct mm_struct *mm)
  1373. {
  1374. }
  1375. #endif
  1376. #ifndef __HAVE_ARCH_PTE_DEVMAP
  1377. static inline int pte_devmap(pte_t pte)
  1378. {
  1379. return 0;
  1380. }
  1381. #endif
  1382. int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
  1383. extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
  1384. spinlock_t **ptl);
  1385. static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
  1386. spinlock_t **ptl)
  1387. {
  1388. pte_t *ptep;
  1389. __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
  1390. return ptep;
  1391. }
  1392. #ifdef __PAGETABLE_P4D_FOLDED
  1393. static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
  1394. unsigned long address)
  1395. {
  1396. return 0;
  1397. }
  1398. #else
  1399. int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
  1400. #endif
  1401. #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
  1402. static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
  1403. unsigned long address)
  1404. {
  1405. return 0;
  1406. }
  1407. static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
  1408. static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
  1409. #else
  1410. int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
  1411. static inline void mm_inc_nr_puds(struct mm_struct *mm)
  1412. {
  1413. atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
  1414. }
  1415. static inline void mm_dec_nr_puds(struct mm_struct *mm)
  1416. {
  1417. atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
  1418. }
  1419. #endif
  1420. #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
  1421. static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
  1422. unsigned long address)
  1423. {
  1424. return 0;
  1425. }
  1426. static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
  1427. static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
  1428. #else
  1429. int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
  1430. static inline void mm_inc_nr_pmds(struct mm_struct *mm)
  1431. {
  1432. atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
  1433. }
  1434. static inline void mm_dec_nr_pmds(struct mm_struct *mm)
  1435. {
  1436. atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
  1437. }
  1438. #endif
  1439. #ifdef CONFIG_MMU
  1440. static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
  1441. {
  1442. atomic_long_set(&mm->pgtables_bytes, 0);
  1443. }
  1444. static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
  1445. {
  1446. return atomic_long_read(&mm->pgtables_bytes);
  1447. }
  1448. static inline void mm_inc_nr_ptes(struct mm_struct *mm)
  1449. {
  1450. atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
  1451. }
  1452. static inline void mm_dec_nr_ptes(struct mm_struct *mm)
  1453. {
  1454. atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
  1455. }
  1456. #else
  1457. static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
  1458. static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
  1459. {
  1460. return 0;
  1461. }
  1462. static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
  1463. static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
  1464. #endif
  1465. int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
  1466. int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
  1467. /*
  1468. * The following ifdef needed to get the 4level-fixup.h header to work.
  1469. * Remove it when 4level-fixup.h has been removed.
  1470. */
  1471. #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
  1472. #ifndef __ARCH_HAS_5LEVEL_HACK
  1473. static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
  1474. unsigned long address)
  1475. {
  1476. return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
  1477. NULL : p4d_offset(pgd, address);
  1478. }
  1479. static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
  1480. unsigned long address)
  1481. {
  1482. return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
  1483. NULL : pud_offset(p4d, address);
  1484. }
  1485. #endif /* !__ARCH_HAS_5LEVEL_HACK */
  1486. static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
  1487. {
  1488. return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
  1489. NULL: pmd_offset(pud, address);
  1490. }
  1491. #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
  1492. #if USE_SPLIT_PTE_PTLOCKS
  1493. #if ALLOC_SPLIT_PTLOCKS
  1494. void __init ptlock_cache_init(void);
  1495. extern bool ptlock_alloc(struct page *page);
  1496. extern void ptlock_free(struct page *page);
  1497. static inline spinlock_t *ptlock_ptr(struct page *page)
  1498. {
  1499. return page->ptl;
  1500. }
  1501. #else /* ALLOC_SPLIT_PTLOCKS */
  1502. static inline void ptlock_cache_init(void)
  1503. {
  1504. }
  1505. static inline bool ptlock_alloc(struct page *page)
  1506. {
  1507. return true;
  1508. }
  1509. static inline void ptlock_free(struct page *page)
  1510. {
  1511. }
  1512. static inline spinlock_t *ptlock_ptr(struct page *page)
  1513. {
  1514. return &page->ptl;
  1515. }
  1516. #endif /* ALLOC_SPLIT_PTLOCKS */
  1517. static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1518. {
  1519. return ptlock_ptr(pmd_page(*pmd));
  1520. }
  1521. static inline bool ptlock_init(struct page *page)
  1522. {
  1523. /*
  1524. * prep_new_page() initialize page->private (and therefore page->ptl)
  1525. * with 0. Make sure nobody took it in use in between.
  1526. *
  1527. * It can happen if arch try to use slab for page table allocation:
  1528. * slab code uses page->slab_cache, which share storage with page->ptl.
  1529. */
  1530. VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
  1531. if (!ptlock_alloc(page))
  1532. return false;
  1533. spin_lock_init(ptlock_ptr(page));
  1534. return true;
  1535. }
  1536. /* Reset page->mapping so free_pages_check won't complain. */
  1537. static inline void pte_lock_deinit(struct page *page)
  1538. {
  1539. page->mapping = NULL;
  1540. ptlock_free(page);
  1541. }
  1542. #else /* !USE_SPLIT_PTE_PTLOCKS */
  1543. /*
  1544. * We use mm->page_table_lock to guard all pagetable pages of the mm.
  1545. */
  1546. static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1547. {
  1548. return &mm->page_table_lock;
  1549. }
  1550. static inline void ptlock_cache_init(void) {}
  1551. static inline bool ptlock_init(struct page *page) { return true; }
  1552. static inline void pte_lock_deinit(struct page *page) {}
  1553. #endif /* USE_SPLIT_PTE_PTLOCKS */
  1554. static inline void pgtable_init(void)
  1555. {
  1556. ptlock_cache_init();
  1557. pgtable_cache_init();
  1558. }
  1559. static inline bool pgtable_page_ctor(struct page *page)
  1560. {
  1561. if (!ptlock_init(page))
  1562. return false;
  1563. inc_zone_page_state(page, NR_PAGETABLE);
  1564. return true;
  1565. }
  1566. static inline void pgtable_page_dtor(struct page *page)
  1567. {
  1568. pte_lock_deinit(page);
  1569. dec_zone_page_state(page, NR_PAGETABLE);
  1570. }
  1571. #define pte_offset_map_lock(mm, pmd, address, ptlp) \
  1572. ({ \
  1573. spinlock_t *__ptl = pte_lockptr(mm, pmd); \
  1574. pte_t *__pte = pte_offset_map(pmd, address); \
  1575. *(ptlp) = __ptl; \
  1576. spin_lock(__ptl); \
  1577. __pte; \
  1578. })
  1579. #define pte_unmap_unlock(pte, ptl) do { \
  1580. spin_unlock(ptl); \
  1581. pte_unmap(pte); \
  1582. } while (0)
  1583. #define pte_alloc(mm, pmd, address) \
  1584. (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
  1585. #define pte_alloc_map(mm, pmd, address) \
  1586. (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
  1587. #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
  1588. (pte_alloc(mm, pmd, address) ? \
  1589. NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
  1590. #define pte_alloc_kernel(pmd, address) \
  1591. ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
  1592. NULL: pte_offset_kernel(pmd, address))
  1593. #if USE_SPLIT_PMD_PTLOCKS
  1594. static struct page *pmd_to_page(pmd_t *pmd)
  1595. {
  1596. unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
  1597. return virt_to_page((void *)((unsigned long) pmd & mask));
  1598. }
  1599. static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1600. {
  1601. return ptlock_ptr(pmd_to_page(pmd));
  1602. }
  1603. static inline bool pgtable_pmd_page_ctor(struct page *page)
  1604. {
  1605. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  1606. page->pmd_huge_pte = NULL;
  1607. #endif
  1608. return ptlock_init(page);
  1609. }
  1610. static inline void pgtable_pmd_page_dtor(struct page *page)
  1611. {
  1612. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  1613. VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
  1614. #endif
  1615. ptlock_free(page);
  1616. }
  1617. #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
  1618. #else
  1619. static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
  1620. {
  1621. return &mm->page_table_lock;
  1622. }
  1623. static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
  1624. static inline void pgtable_pmd_page_dtor(struct page *page) {}
  1625. #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
  1626. #endif
  1627. static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
  1628. {
  1629. spinlock_t *ptl = pmd_lockptr(mm, pmd);
  1630. spin_lock(ptl);
  1631. return ptl;
  1632. }
  1633. /*
  1634. * No scalability reason to split PUD locks yet, but follow the same pattern
  1635. * as the PMD locks to make it easier if we decide to. The VM should not be
  1636. * considered ready to switch to split PUD locks yet; there may be places
  1637. * which need to be converted from page_table_lock.
  1638. */
  1639. static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
  1640. {
  1641. return &mm->page_table_lock;
  1642. }
  1643. static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
  1644. {
  1645. spinlock_t *ptl = pud_lockptr(mm, pud);
  1646. spin_lock(ptl);
  1647. return ptl;
  1648. }
  1649. extern void __init pagecache_init(void);
  1650. extern void free_area_init(unsigned long * zones_size);
  1651. extern void free_area_init_node(int nid, unsigned long * zones_size,
  1652. unsigned long zone_start_pfn, unsigned long *zholes_size);
  1653. extern void free_initmem(void);
  1654. /*
  1655. * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
  1656. * into the buddy system. The freed pages will be poisoned with pattern
  1657. * "poison" if it's within range [0, UCHAR_MAX].
  1658. * Return pages freed into the buddy system.
  1659. */
  1660. extern unsigned long free_reserved_area(void *start, void *end,
  1661. int poison, char *s);
  1662. #ifdef CONFIG_HIGHMEM
  1663. /*
  1664. * Free a highmem page into the buddy system, adjusting totalhigh_pages
  1665. * and totalram_pages.
  1666. */
  1667. extern void free_highmem_page(struct page *page);
  1668. #endif
  1669. extern void adjust_managed_page_count(struct page *page, long count);
  1670. extern void mem_init_print_info(const char *str);
  1671. extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
  1672. /* Free the reserved page into the buddy system, so it gets managed. */
  1673. static inline void __free_reserved_page(struct page *page)
  1674. {
  1675. ClearPageReserved(page);
  1676. init_page_count(page);
  1677. __free_page(page);
  1678. }
  1679. static inline void free_reserved_page(struct page *page)
  1680. {
  1681. __free_reserved_page(page);
  1682. adjust_managed_page_count(page, 1);
  1683. }
  1684. static inline void mark_page_reserved(struct page *page)
  1685. {
  1686. SetPageReserved(page);
  1687. adjust_managed_page_count(page, -1);
  1688. }
  1689. /*
  1690. * Default method to free all the __init memory into the buddy system.
  1691. * The freed pages will be poisoned with pattern "poison" if it's within
  1692. * range [0, UCHAR_MAX].
  1693. * Return pages freed into the buddy system.
  1694. */
  1695. static inline unsigned long free_initmem_default(int poison)
  1696. {
  1697. extern char __init_begin[], __init_end[];
  1698. return free_reserved_area(&__init_begin, &__init_end,
  1699. poison, "unused kernel");
  1700. }
  1701. static inline unsigned long get_num_physpages(void)
  1702. {
  1703. int nid;
  1704. unsigned long phys_pages = 0;
  1705. for_each_online_node(nid)
  1706. phys_pages += node_present_pages(nid);
  1707. return phys_pages;
  1708. }
  1709. #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
  1710. /*
  1711. * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
  1712. * zones, allocate the backing mem_map and account for memory holes in a more
  1713. * architecture independent manner. This is a substitute for creating the
  1714. * zone_sizes[] and zholes_size[] arrays and passing them to
  1715. * free_area_init_node()
  1716. *
  1717. * An architecture is expected to register range of page frames backed by
  1718. * physical memory with memblock_add[_node]() before calling
  1719. * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
  1720. * usage, an architecture is expected to do something like
  1721. *
  1722. * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
  1723. * max_highmem_pfn};
  1724. * for_each_valid_physical_page_range()
  1725. * memblock_add_node(base, size, nid)
  1726. * free_area_init_nodes(max_zone_pfns);
  1727. *
  1728. * free_bootmem_with_active_regions() calls free_bootmem_node() for each
  1729. * registered physical page range. Similarly
  1730. * sparse_memory_present_with_active_regions() calls memory_present() for
  1731. * each range when SPARSEMEM is enabled.
  1732. *
  1733. * See mm/page_alloc.c for more information on each function exposed by
  1734. * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
  1735. */
  1736. extern void free_area_init_nodes(unsigned long *max_zone_pfn);
  1737. unsigned long node_map_pfn_alignment(void);
  1738. unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
  1739. unsigned long end_pfn);
  1740. extern unsigned long absent_pages_in_range(unsigned long start_pfn,
  1741. unsigned long end_pfn);
  1742. extern void get_pfn_range_for_nid(unsigned int nid,
  1743. unsigned long *start_pfn, unsigned long *end_pfn);
  1744. extern unsigned long find_min_pfn_with_active_regions(void);
  1745. extern void free_bootmem_with_active_regions(int nid,
  1746. unsigned long max_low_pfn);
  1747. extern void sparse_memory_present_with_active_regions(int nid);
  1748. #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
  1749. #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
  1750. !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
  1751. static inline int __early_pfn_to_nid(unsigned long pfn,
  1752. struct mminit_pfnnid_cache *state)
  1753. {
  1754. return 0;
  1755. }
  1756. #else
  1757. /* please see mm/page_alloc.c */
  1758. extern int __meminit early_pfn_to_nid(unsigned long pfn);
  1759. /* there is a per-arch backend function. */
  1760. extern int __meminit __early_pfn_to_nid(unsigned long pfn,
  1761. struct mminit_pfnnid_cache *state);
  1762. #endif
  1763. #ifdef CONFIG_HAVE_MEMBLOCK
  1764. void zero_resv_unavail(void);
  1765. #else
  1766. static inline void zero_resv_unavail(void) {}
  1767. #endif
  1768. extern void set_dma_reserve(unsigned long new_dma_reserve);
  1769. extern void memmap_init_zone(unsigned long, int, unsigned long,
  1770. unsigned long, enum memmap_context);
  1771. extern void setup_per_zone_wmarks(void);
  1772. extern int __meminit init_per_zone_wmark_min(void);
  1773. extern void mem_init(void);
  1774. extern void __init mmap_init(void);
  1775. extern void show_mem(unsigned int flags, nodemask_t *nodemask);
  1776. extern long si_mem_available(void);
  1777. extern void si_meminfo(struct sysinfo * val);
  1778. extern void si_meminfo_node(struct sysinfo *val, int nid);
  1779. #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
  1780. extern unsigned long arch_reserved_kernel_pages(void);
  1781. #endif
  1782. extern __printf(3, 4)
  1783. void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
  1784. extern void setup_per_cpu_pageset(void);
  1785. extern void zone_pcp_update(struct zone *zone);
  1786. extern void zone_pcp_reset(struct zone *zone);
  1787. /* page_alloc.c */
  1788. extern int min_free_kbytes;
  1789. extern int watermark_scale_factor;
  1790. /* nommu.c */
  1791. extern atomic_long_t mmap_pages_allocated;
  1792. extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
  1793. /* interval_tree.c */
  1794. void vma_interval_tree_insert(struct vm_area_struct *node,
  1795. struct rb_root_cached *root);
  1796. void vma_interval_tree_insert_after(struct vm_area_struct *node,
  1797. struct vm_area_struct *prev,
  1798. struct rb_root_cached *root);
  1799. void vma_interval_tree_remove(struct vm_area_struct *node,
  1800. struct rb_root_cached *root);
  1801. struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
  1802. unsigned long start, unsigned long last);
  1803. struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
  1804. unsigned long start, unsigned long last);
  1805. #define vma_interval_tree_foreach(vma, root, start, last) \
  1806. for (vma = vma_interval_tree_iter_first(root, start, last); \
  1807. vma; vma = vma_interval_tree_iter_next(vma, start, last))
  1808. void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
  1809. struct rb_root_cached *root);
  1810. void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
  1811. struct rb_root_cached *root);
  1812. struct anon_vma_chain *
  1813. anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
  1814. unsigned long start, unsigned long last);
  1815. struct anon_vma_chain *anon_vma_interval_tree_iter_next(
  1816. struct anon_vma_chain *node, unsigned long start, unsigned long last);
  1817. #ifdef CONFIG_DEBUG_VM_RB
  1818. void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
  1819. #endif
  1820. #define anon_vma_interval_tree_foreach(avc, root, start, last) \
  1821. for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
  1822. avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
  1823. /* mmap.c */
  1824. extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
  1825. extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
  1826. unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
  1827. struct vm_area_struct *expand);
  1828. static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
  1829. unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
  1830. {
  1831. return __vma_adjust(vma, start, end, pgoff, insert, NULL);
  1832. }
  1833. extern struct vm_area_struct *vma_merge(struct mm_struct *,
  1834. struct vm_area_struct *prev, unsigned long addr, unsigned long end,
  1835. unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
  1836. struct mempolicy *, struct vm_userfaultfd_ctx);
  1837. extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
  1838. extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
  1839. unsigned long addr, int new_below);
  1840. extern int split_vma(struct mm_struct *, struct vm_area_struct *,
  1841. unsigned long addr, int new_below);
  1842. extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
  1843. extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
  1844. struct rb_node **, struct rb_node *);
  1845. extern void unlink_file_vma(struct vm_area_struct *);
  1846. extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
  1847. unsigned long addr, unsigned long len, pgoff_t pgoff,
  1848. bool *need_rmap_locks);
  1849. extern void exit_mmap(struct mm_struct *);
  1850. static inline int check_data_rlimit(unsigned long rlim,
  1851. unsigned long new,
  1852. unsigned long start,
  1853. unsigned long end_data,
  1854. unsigned long start_data)
  1855. {
  1856. if (rlim < RLIM_INFINITY) {
  1857. if (((new - start) + (end_data - start_data)) > rlim)
  1858. return -ENOSPC;
  1859. }
  1860. return 0;
  1861. }
  1862. extern int mm_take_all_locks(struct mm_struct *mm);
  1863. extern void mm_drop_all_locks(struct mm_struct *mm);
  1864. extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
  1865. extern struct file *get_mm_exe_file(struct mm_struct *mm);
  1866. extern struct file *get_task_exe_file(struct task_struct *task);
  1867. extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
  1868. extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
  1869. extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
  1870. const struct vm_special_mapping *sm);
  1871. extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
  1872. unsigned long addr, unsigned long len,
  1873. unsigned long flags,
  1874. const struct vm_special_mapping *spec);
  1875. /* This is an obsolete alternative to _install_special_mapping. */
  1876. extern int install_special_mapping(struct mm_struct *mm,
  1877. unsigned long addr, unsigned long len,
  1878. unsigned long flags, struct page **pages);
  1879. extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
  1880. extern unsigned long mmap_region(struct file *file, unsigned long addr,
  1881. unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
  1882. struct list_head *uf);
  1883. extern unsigned long do_mmap(struct file *file, unsigned long addr,
  1884. unsigned long len, unsigned long prot, unsigned long flags,
  1885. vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
  1886. struct list_head *uf);
  1887. extern int do_munmap(struct mm_struct *, unsigned long, size_t,
  1888. struct list_head *uf);
  1889. static inline unsigned long
  1890. do_mmap_pgoff(struct file *file, unsigned long addr,
  1891. unsigned long len, unsigned long prot, unsigned long flags,
  1892. unsigned long pgoff, unsigned long *populate,
  1893. struct list_head *uf)
  1894. {
  1895. return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
  1896. }
  1897. #ifdef CONFIG_MMU
  1898. extern int __mm_populate(unsigned long addr, unsigned long len,
  1899. int ignore_errors);
  1900. static inline void mm_populate(unsigned long addr, unsigned long len)
  1901. {
  1902. /* Ignore errors */
  1903. (void) __mm_populate(addr, len, 1);
  1904. }
  1905. #else
  1906. static inline void mm_populate(unsigned long addr, unsigned long len) {}
  1907. #endif
  1908. /* These take the mm semaphore themselves */
  1909. extern int __must_check vm_brk(unsigned long, unsigned long);
  1910. extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
  1911. extern int vm_munmap(unsigned long, size_t);
  1912. extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
  1913. unsigned long, unsigned long,
  1914. unsigned long, unsigned long);
  1915. struct vm_unmapped_area_info {
  1916. #define VM_UNMAPPED_AREA_TOPDOWN 1
  1917. unsigned long flags;
  1918. unsigned long length;
  1919. unsigned long low_limit;
  1920. unsigned long high_limit;
  1921. unsigned long align_mask;
  1922. unsigned long align_offset;
  1923. };
  1924. extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
  1925. extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
  1926. /*
  1927. * Search for an unmapped address range.
  1928. *
  1929. * We are looking for a range that:
  1930. * - does not intersect with any VMA;
  1931. * - is contained within the [low_limit, high_limit) interval;
  1932. * - is at least the desired size.
  1933. * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
  1934. */
  1935. static inline unsigned long
  1936. vm_unmapped_area(struct vm_unmapped_area_info *info)
  1937. {
  1938. if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
  1939. return unmapped_area_topdown(info);
  1940. else
  1941. return unmapped_area(info);
  1942. }
  1943. /* truncate.c */
  1944. extern void truncate_inode_pages(struct address_space *, loff_t);
  1945. extern void truncate_inode_pages_range(struct address_space *,
  1946. loff_t lstart, loff_t lend);
  1947. extern void truncate_inode_pages_final(struct address_space *);
  1948. /* generic vm_area_ops exported for stackable file systems */
  1949. extern int filemap_fault(struct vm_fault *vmf);
  1950. extern void filemap_map_pages(struct vm_fault *vmf,
  1951. pgoff_t start_pgoff, pgoff_t end_pgoff);
  1952. extern int filemap_page_mkwrite(struct vm_fault *vmf);
  1953. /* mm/page-writeback.c */
  1954. int __must_check write_one_page(struct page *page);
  1955. void task_dirty_inc(struct task_struct *tsk);
  1956. /* readahead.c */
  1957. #define VM_MAX_READAHEAD 128 /* kbytes */
  1958. #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
  1959. int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
  1960. pgoff_t offset, unsigned long nr_to_read);
  1961. void page_cache_sync_readahead(struct address_space *mapping,
  1962. struct file_ra_state *ra,
  1963. struct file *filp,
  1964. pgoff_t offset,
  1965. unsigned long size);
  1966. void page_cache_async_readahead(struct address_space *mapping,
  1967. struct file_ra_state *ra,
  1968. struct file *filp,
  1969. struct page *pg,
  1970. pgoff_t offset,
  1971. unsigned long size);
  1972. extern unsigned long stack_guard_gap;
  1973. /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
  1974. extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
  1975. /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
  1976. extern int expand_downwards(struct vm_area_struct *vma,
  1977. unsigned long address);
  1978. #if VM_GROWSUP
  1979. extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
  1980. #else
  1981. #define expand_upwards(vma, address) (0)
  1982. #endif
  1983. /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
  1984. extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
  1985. extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
  1986. struct vm_area_struct **pprev);
  1987. /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
  1988. NULL if none. Assume start_addr < end_addr. */
  1989. static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
  1990. {
  1991. struct vm_area_struct * vma = find_vma(mm,start_addr);
  1992. if (vma && end_addr <= vma->vm_start)
  1993. vma = NULL;
  1994. return vma;
  1995. }
  1996. static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
  1997. {
  1998. unsigned long vm_start = vma->vm_start;
  1999. if (vma->vm_flags & VM_GROWSDOWN) {
  2000. vm_start -= stack_guard_gap;
  2001. if (vm_start > vma->vm_start)
  2002. vm_start = 0;
  2003. }
  2004. return vm_start;
  2005. }
  2006. static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
  2007. {
  2008. unsigned long vm_end = vma->vm_end;
  2009. if (vma->vm_flags & VM_GROWSUP) {
  2010. vm_end += stack_guard_gap;
  2011. if (vm_end < vma->vm_end)
  2012. vm_end = -PAGE_SIZE;
  2013. }
  2014. return vm_end;
  2015. }
  2016. static inline unsigned long vma_pages(struct vm_area_struct *vma)
  2017. {
  2018. return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
  2019. }
  2020. /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
  2021. static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
  2022. unsigned long vm_start, unsigned long vm_end)
  2023. {
  2024. struct vm_area_struct *vma = find_vma(mm, vm_start);
  2025. if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
  2026. vma = NULL;
  2027. return vma;
  2028. }
  2029. #ifdef CONFIG_MMU
  2030. pgprot_t vm_get_page_prot(unsigned long vm_flags);
  2031. void vma_set_page_prot(struct vm_area_struct *vma);
  2032. #else
  2033. static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
  2034. {
  2035. return __pgprot(0);
  2036. }
  2037. static inline void vma_set_page_prot(struct vm_area_struct *vma)
  2038. {
  2039. vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
  2040. }
  2041. #endif
  2042. #ifdef CONFIG_NUMA_BALANCING
  2043. unsigned long change_prot_numa(struct vm_area_struct *vma,
  2044. unsigned long start, unsigned long end);
  2045. #endif
  2046. struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
  2047. int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
  2048. unsigned long pfn, unsigned long size, pgprot_t);
  2049. int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
  2050. int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
  2051. unsigned long pfn);
  2052. int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
  2053. unsigned long pfn, pgprot_t pgprot);
  2054. int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
  2055. pfn_t pfn);
  2056. int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
  2057. pfn_t pfn);
  2058. int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
  2059. struct page *follow_page_mask(struct vm_area_struct *vma,
  2060. unsigned long address, unsigned int foll_flags,
  2061. unsigned int *page_mask);
  2062. static inline struct page *follow_page(struct vm_area_struct *vma,
  2063. unsigned long address, unsigned int foll_flags)
  2064. {
  2065. unsigned int unused_page_mask;
  2066. return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
  2067. }
  2068. #define FOLL_WRITE 0x01 /* check pte is writable */
  2069. #define FOLL_TOUCH 0x02 /* mark page accessed */
  2070. #define FOLL_GET 0x04 /* do get_page on page */
  2071. #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
  2072. #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
  2073. #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
  2074. * and return without waiting upon it */
  2075. #define FOLL_POPULATE 0x40 /* fault in page */
  2076. #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
  2077. #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
  2078. #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
  2079. #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
  2080. #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
  2081. #define FOLL_MLOCK 0x1000 /* lock present pages */
  2082. #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
  2083. #define FOLL_COW 0x4000 /* internal GUP flag */
  2084. static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
  2085. {
  2086. if (vm_fault & VM_FAULT_OOM)
  2087. return -ENOMEM;
  2088. if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
  2089. return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
  2090. if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
  2091. return -EFAULT;
  2092. return 0;
  2093. }
  2094. typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
  2095. void *data);
  2096. extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
  2097. unsigned long size, pte_fn_t fn, void *data);
  2098. #ifdef CONFIG_PAGE_POISONING
  2099. extern bool page_poisoning_enabled(void);
  2100. extern void kernel_poison_pages(struct page *page, int numpages, int enable);
  2101. extern bool page_is_poisoned(struct page *page);
  2102. #else
  2103. static inline bool page_poisoning_enabled(void) { return false; }
  2104. static inline void kernel_poison_pages(struct page *page, int numpages,
  2105. int enable) { }
  2106. static inline bool page_is_poisoned(struct page *page) { return false; }
  2107. #endif
  2108. #ifdef CONFIG_DEBUG_PAGEALLOC
  2109. extern bool _debug_pagealloc_enabled;
  2110. extern void __kernel_map_pages(struct page *page, int numpages, int enable);
  2111. static inline bool debug_pagealloc_enabled(void)
  2112. {
  2113. return _debug_pagealloc_enabled;
  2114. }
  2115. static inline void
  2116. kernel_map_pages(struct page *page, int numpages, int enable)
  2117. {
  2118. if (!debug_pagealloc_enabled())
  2119. return;
  2120. __kernel_map_pages(page, numpages, enable);
  2121. }
  2122. #ifdef CONFIG_HIBERNATION
  2123. extern bool kernel_page_present(struct page *page);
  2124. #endif /* CONFIG_HIBERNATION */
  2125. #else /* CONFIG_DEBUG_PAGEALLOC */
  2126. static inline void
  2127. kernel_map_pages(struct page *page, int numpages, int enable) {}
  2128. #ifdef CONFIG_HIBERNATION
  2129. static inline bool kernel_page_present(struct page *page) { return true; }
  2130. #endif /* CONFIG_HIBERNATION */
  2131. static inline bool debug_pagealloc_enabled(void)
  2132. {
  2133. return false;
  2134. }
  2135. #endif /* CONFIG_DEBUG_PAGEALLOC */
  2136. #ifdef __HAVE_ARCH_GATE_AREA
  2137. extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
  2138. extern int in_gate_area_no_mm(unsigned long addr);
  2139. extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
  2140. #else
  2141. static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
  2142. {
  2143. return NULL;
  2144. }
  2145. static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
  2146. static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
  2147. {
  2148. return 0;
  2149. }
  2150. #endif /* __HAVE_ARCH_GATE_AREA */
  2151. extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
  2152. #ifdef CONFIG_SYSCTL
  2153. extern int sysctl_drop_caches;
  2154. int drop_caches_sysctl_handler(struct ctl_table *, int,
  2155. void __user *, size_t *, loff_t *);
  2156. #endif
  2157. void drop_slab(void);
  2158. void drop_slab_node(int nid);
  2159. #ifndef CONFIG_MMU
  2160. #define randomize_va_space 0
  2161. #else
  2162. extern int randomize_va_space;
  2163. #endif
  2164. const char * arch_vma_name(struct vm_area_struct *vma);
  2165. void print_vma_addr(char *prefix, unsigned long rip);
  2166. void sparse_mem_maps_populate_node(struct page **map_map,
  2167. unsigned long pnum_begin,
  2168. unsigned long pnum_end,
  2169. unsigned long map_count,
  2170. int nodeid);
  2171. struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
  2172. pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
  2173. p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
  2174. pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
  2175. pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
  2176. pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
  2177. void *vmemmap_alloc_block(unsigned long size, int node);
  2178. struct vmem_altmap;
  2179. void *__vmemmap_alloc_block_buf(unsigned long size, int node,
  2180. struct vmem_altmap *altmap);
  2181. static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
  2182. {
  2183. return __vmemmap_alloc_block_buf(size, node, NULL);
  2184. }
  2185. void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
  2186. int vmemmap_populate_basepages(unsigned long start, unsigned long end,
  2187. int node);
  2188. int vmemmap_populate(unsigned long start, unsigned long end, int node);
  2189. void vmemmap_populate_print_last(void);
  2190. #ifdef CONFIG_MEMORY_HOTPLUG
  2191. void vmemmap_free(unsigned long start, unsigned long end);
  2192. #endif
  2193. void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
  2194. unsigned long nr_pages);
  2195. enum mf_flags {
  2196. MF_COUNT_INCREASED = 1 << 0,
  2197. MF_ACTION_REQUIRED = 1 << 1,
  2198. MF_MUST_KILL = 1 << 2,
  2199. MF_SOFT_OFFLINE = 1 << 3,
  2200. };
  2201. extern int memory_failure(unsigned long pfn, int trapno, int flags);
  2202. extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
  2203. extern int unpoison_memory(unsigned long pfn);
  2204. extern int get_hwpoison_page(struct page *page);
  2205. #define put_hwpoison_page(page) put_page(page)
  2206. extern int sysctl_memory_failure_early_kill;
  2207. extern int sysctl_memory_failure_recovery;
  2208. extern void shake_page(struct page *p, int access);
  2209. extern atomic_long_t num_poisoned_pages;
  2210. extern int soft_offline_page(struct page *page, int flags);
  2211. /*
  2212. * Error handlers for various types of pages.
  2213. */
  2214. enum mf_result {
  2215. MF_IGNORED, /* Error: cannot be handled */
  2216. MF_FAILED, /* Error: handling failed */
  2217. MF_DELAYED, /* Will be handled later */
  2218. MF_RECOVERED, /* Successfully recovered */
  2219. };
  2220. enum mf_action_page_type {
  2221. MF_MSG_KERNEL,
  2222. MF_MSG_KERNEL_HIGH_ORDER,
  2223. MF_MSG_SLAB,
  2224. MF_MSG_DIFFERENT_COMPOUND,
  2225. MF_MSG_POISONED_HUGE,
  2226. MF_MSG_HUGE,
  2227. MF_MSG_FREE_HUGE,
  2228. MF_MSG_UNMAP_FAILED,
  2229. MF_MSG_DIRTY_SWAPCACHE,
  2230. MF_MSG_CLEAN_SWAPCACHE,
  2231. MF_MSG_DIRTY_MLOCKED_LRU,
  2232. MF_MSG_CLEAN_MLOCKED_LRU,
  2233. MF_MSG_DIRTY_UNEVICTABLE_LRU,
  2234. MF_MSG_CLEAN_UNEVICTABLE_LRU,
  2235. MF_MSG_DIRTY_LRU,
  2236. MF_MSG_CLEAN_LRU,
  2237. MF_MSG_TRUNCATED_LRU,
  2238. MF_MSG_BUDDY,
  2239. MF_MSG_BUDDY_2ND,
  2240. MF_MSG_UNKNOWN,
  2241. };
  2242. #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
  2243. extern void clear_huge_page(struct page *page,
  2244. unsigned long addr_hint,
  2245. unsigned int pages_per_huge_page);
  2246. extern void copy_user_huge_page(struct page *dst, struct page *src,
  2247. unsigned long addr, struct vm_area_struct *vma,
  2248. unsigned int pages_per_huge_page);
  2249. extern long copy_huge_page_from_user(struct page *dst_page,
  2250. const void __user *usr_src,
  2251. unsigned int pages_per_huge_page,
  2252. bool allow_pagefault);
  2253. #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
  2254. extern struct page_ext_operations debug_guardpage_ops;
  2255. #ifdef CONFIG_DEBUG_PAGEALLOC
  2256. extern unsigned int _debug_guardpage_minorder;
  2257. extern bool _debug_guardpage_enabled;
  2258. static inline unsigned int debug_guardpage_minorder(void)
  2259. {
  2260. return _debug_guardpage_minorder;
  2261. }
  2262. static inline bool debug_guardpage_enabled(void)
  2263. {
  2264. return _debug_guardpage_enabled;
  2265. }
  2266. static inline bool page_is_guard(struct page *page)
  2267. {
  2268. struct page_ext *page_ext;
  2269. if (!debug_guardpage_enabled())
  2270. return false;
  2271. page_ext = lookup_page_ext(page);
  2272. if (unlikely(!page_ext))
  2273. return false;
  2274. return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
  2275. }
  2276. #else
  2277. static inline unsigned int debug_guardpage_minorder(void) { return 0; }
  2278. static inline bool debug_guardpage_enabled(void) { return false; }
  2279. static inline bool page_is_guard(struct page *page) { return false; }
  2280. #endif /* CONFIG_DEBUG_PAGEALLOC */
  2281. #if MAX_NUMNODES > 1
  2282. void __init setup_nr_node_ids(void);
  2283. #else
  2284. static inline void setup_nr_node_ids(void) {}
  2285. #endif
  2286. #endif /* __KERNEL__ */
  2287. #endif /* _LINUX_MM_H */