internal.h 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508
  1. /* internal.h: mm/ internal definitions
  2. *
  3. * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
  4. * Written by David Howells (dhowells@redhat.com)
  5. *
  6. * This program is free software; you can redistribute it and/or
  7. * modify it under the terms of the GNU General Public License
  8. * as published by the Free Software Foundation; either version
  9. * 2 of the License, or (at your option) any later version.
  10. */
  11. #ifndef __MM_INTERNAL_H
  12. #define __MM_INTERNAL_H
  13. #include <linux/fs.h>
  14. #include <linux/mm.h>
  15. #include <linux/pagemap.h>
  16. #include <linux/tracepoint-defs.h>
  17. /*
  18. * The set of flags that only affect watermark checking and reclaim
  19. * behaviour. This is used by the MM to obey the caller constraints
  20. * about IO, FS and watermark checking while ignoring placement
  21. * hints such as HIGHMEM usage.
  22. */
  23. #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
  24. __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
  25. __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
  26. __GFP_ATOMIC)
  27. /* The GFP flags allowed during early boot */
  28. #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
  29. /* Control allocation cpuset and node placement constraints */
  30. #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
  31. /* Do not use these with a slab allocator */
  32. #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
  33. void page_writeback_init(void);
  34. int do_swap_page(struct vm_fault *vmf);
  35. void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  36. unsigned long floor, unsigned long ceiling);
  37. static inline bool can_madv_dontneed_vma(struct vm_area_struct *vma)
  38. {
  39. return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
  40. }
  41. void unmap_page_range(struct mmu_gather *tlb,
  42. struct vm_area_struct *vma,
  43. unsigned long addr, unsigned long end,
  44. struct zap_details *details);
  45. extern int __do_page_cache_readahead(struct address_space *mapping,
  46. struct file *filp, pgoff_t offset, unsigned long nr_to_read,
  47. unsigned long lookahead_size);
  48. /*
  49. * Submit IO for the read-ahead request in file_ra_state.
  50. */
  51. static inline unsigned long ra_submit(struct file_ra_state *ra,
  52. struct address_space *mapping, struct file *filp)
  53. {
  54. return __do_page_cache_readahead(mapping, filp,
  55. ra->start, ra->size, ra->async_size);
  56. }
  57. /*
  58. * Turn a non-refcounted page (->_refcount == 0) into refcounted with
  59. * a count of one.
  60. */
  61. static inline void set_page_refcounted(struct page *page)
  62. {
  63. VM_BUG_ON_PAGE(PageTail(page), page);
  64. VM_BUG_ON_PAGE(page_ref_count(page), page);
  65. set_page_count(page, 1);
  66. }
  67. extern unsigned long highest_memmap_pfn;
  68. /*
  69. * in mm/vmscan.c:
  70. */
  71. extern int isolate_lru_page(struct page *page);
  72. extern void putback_lru_page(struct page *page);
  73. extern bool pgdat_reclaimable(struct pglist_data *pgdat);
  74. /*
  75. * in mm/rmap.c:
  76. */
  77. extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  78. /*
  79. * in mm/page_alloc.c
  80. */
  81. /*
  82. * Structure for holding the mostly immutable allocation parameters passed
  83. * between functions involved in allocations, including the alloc_pages*
  84. * family of functions.
  85. *
  86. * nodemask, migratetype and high_zoneidx are initialized only once in
  87. * __alloc_pages_nodemask() and then never change.
  88. *
  89. * zonelist, preferred_zone and classzone_idx are set first in
  90. * __alloc_pages_nodemask() for the fast path, and might be later changed
  91. * in __alloc_pages_slowpath(). All other functions pass the whole strucure
  92. * by a const pointer.
  93. */
  94. struct alloc_context {
  95. struct zonelist *zonelist;
  96. nodemask_t *nodemask;
  97. struct zoneref *preferred_zoneref;
  98. int migratetype;
  99. enum zone_type high_zoneidx;
  100. bool spread_dirty_pages;
  101. };
  102. #define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
  103. /*
  104. * Locate the struct page for both the matching buddy in our
  105. * pair (buddy1) and the combined O(n+1) page they form (page).
  106. *
  107. * 1) Any buddy B1 will have an order O twin B2 which satisfies
  108. * the following equation:
  109. * B2 = B1 ^ (1 << O)
  110. * For example, if the starting buddy (buddy2) is #8 its order
  111. * 1 buddy is #10:
  112. * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
  113. *
  114. * 2) Any buddy B will have an order O+1 parent P which
  115. * satisfies the following equation:
  116. * P = B & ~(1 << O)
  117. *
  118. * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
  119. */
  120. static inline unsigned long
  121. __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
  122. {
  123. return page_pfn ^ (1 << order);
  124. }
  125. extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
  126. unsigned long end_pfn, struct zone *zone);
  127. static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
  128. unsigned long end_pfn, struct zone *zone)
  129. {
  130. if (zone->contiguous)
  131. return pfn_to_page(start_pfn);
  132. return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
  133. }
  134. extern int __isolate_free_page(struct page *page, unsigned int order);
  135. extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
  136. unsigned int order);
  137. extern void prep_compound_page(struct page *page, unsigned int order);
  138. extern void post_alloc_hook(struct page *page, unsigned int order,
  139. gfp_t gfp_flags);
  140. extern int user_min_free_kbytes;
  141. #if defined CONFIG_COMPACTION || defined CONFIG_CMA
  142. /*
  143. * in mm/compaction.c
  144. */
  145. /*
  146. * compact_control is used to track pages being migrated and the free pages
  147. * they are being migrated to during memory compaction. The free_pfn starts
  148. * at the end of a zone and migrate_pfn begins at the start. Movable pages
  149. * are moved to the end of a zone during a compaction run and the run
  150. * completes when free_pfn <= migrate_pfn
  151. */
  152. struct compact_control {
  153. struct list_head freepages; /* List of free pages to migrate to */
  154. struct list_head migratepages; /* List of pages being migrated */
  155. unsigned long nr_freepages; /* Number of isolated free pages */
  156. unsigned long nr_migratepages; /* Number of pages to migrate */
  157. unsigned long total_migrate_scanned;
  158. unsigned long total_free_scanned;
  159. unsigned long free_pfn; /* isolate_freepages search base */
  160. unsigned long migrate_pfn; /* isolate_migratepages search base */
  161. unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
  162. enum migrate_mode mode; /* Async or sync migration mode */
  163. bool ignore_skip_hint; /* Scan blocks even if marked skip */
  164. bool ignore_block_suitable; /* Scan blocks considered unsuitable */
  165. bool direct_compaction; /* False from kcompactd or /proc/... */
  166. bool whole_zone; /* Whole zone should/has been scanned */
  167. int order; /* order a direct compactor needs */
  168. const gfp_t gfp_mask; /* gfp mask of a direct compactor */
  169. const unsigned int alloc_flags; /* alloc flags of a direct compactor */
  170. const int classzone_idx; /* zone index of a direct compactor */
  171. struct zone *zone;
  172. bool contended; /* Signal lock or sched contention */
  173. };
  174. unsigned long
  175. isolate_freepages_range(struct compact_control *cc,
  176. unsigned long start_pfn, unsigned long end_pfn);
  177. unsigned long
  178. isolate_migratepages_range(struct compact_control *cc,
  179. unsigned long low_pfn, unsigned long end_pfn);
  180. int find_suitable_fallback(struct free_area *area, unsigned int order,
  181. int migratetype, bool only_stealable, bool *can_steal);
  182. #endif
  183. /*
  184. * This function returns the order of a free page in the buddy system. In
  185. * general, page_zone(page)->lock must be held by the caller to prevent the
  186. * page from being allocated in parallel and returning garbage as the order.
  187. * If a caller does not hold page_zone(page)->lock, it must guarantee that the
  188. * page cannot be allocated or merged in parallel. Alternatively, it must
  189. * handle invalid values gracefully, and use page_order_unsafe() below.
  190. */
  191. static inline unsigned int page_order(struct page *page)
  192. {
  193. /* PageBuddy() must be checked by the caller */
  194. return page_private(page);
  195. }
  196. /*
  197. * Like page_order(), but for callers who cannot afford to hold the zone lock.
  198. * PageBuddy() should be checked first by the caller to minimize race window,
  199. * and invalid values must be handled gracefully.
  200. *
  201. * READ_ONCE is used so that if the caller assigns the result into a local
  202. * variable and e.g. tests it for valid range before using, the compiler cannot
  203. * decide to remove the variable and inline the page_private(page) multiple
  204. * times, potentially observing different values in the tests and the actual
  205. * use of the result.
  206. */
  207. #define page_order_unsafe(page) READ_ONCE(page_private(page))
  208. static inline bool is_cow_mapping(vm_flags_t flags)
  209. {
  210. return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
  211. }
  212. /*
  213. * These three helpers classifies VMAs for virtual memory accounting.
  214. */
  215. /*
  216. * Executable code area - executable, not writable, not stack
  217. */
  218. static inline bool is_exec_mapping(vm_flags_t flags)
  219. {
  220. return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
  221. }
  222. /*
  223. * Stack area - atomatically grows in one direction
  224. *
  225. * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
  226. * do_mmap() forbids all other combinations.
  227. */
  228. static inline bool is_stack_mapping(vm_flags_t flags)
  229. {
  230. return (flags & VM_STACK) == VM_STACK;
  231. }
  232. /*
  233. * Data area - private, writable, not stack
  234. */
  235. static inline bool is_data_mapping(vm_flags_t flags)
  236. {
  237. return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
  238. }
  239. /* mm/util.c */
  240. void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
  241. struct vm_area_struct *prev, struct rb_node *rb_parent);
  242. #ifdef CONFIG_MMU
  243. extern long populate_vma_page_range(struct vm_area_struct *vma,
  244. unsigned long start, unsigned long end, int *nonblocking);
  245. extern void munlock_vma_pages_range(struct vm_area_struct *vma,
  246. unsigned long start, unsigned long end);
  247. static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
  248. {
  249. munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
  250. }
  251. /*
  252. * must be called with vma's mmap_sem held for read or write, and page locked.
  253. */
  254. extern void mlock_vma_page(struct page *page);
  255. extern unsigned int munlock_vma_page(struct page *page);
  256. /*
  257. * Clear the page's PageMlocked(). This can be useful in a situation where
  258. * we want to unconditionally remove a page from the pagecache -- e.g.,
  259. * on truncation or freeing.
  260. *
  261. * It is legal to call this function for any page, mlocked or not.
  262. * If called for a page that is still mapped by mlocked vmas, all we do
  263. * is revert to lazy LRU behaviour -- semantics are not broken.
  264. */
  265. extern void clear_page_mlock(struct page *page);
  266. /*
  267. * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
  268. * (because that does not go through the full procedure of migration ptes):
  269. * to migrate the Mlocked page flag; update statistics.
  270. */
  271. static inline void mlock_migrate_page(struct page *newpage, struct page *page)
  272. {
  273. if (TestClearPageMlocked(page)) {
  274. int nr_pages = hpage_nr_pages(page);
  275. /* Holding pmd lock, no change in irq context: __mod is safe */
  276. __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
  277. SetPageMlocked(newpage);
  278. __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
  279. }
  280. }
  281. extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
  282. /*
  283. * At what user virtual address is page expected in @vma?
  284. */
  285. static inline unsigned long
  286. __vma_address(struct page *page, struct vm_area_struct *vma)
  287. {
  288. pgoff_t pgoff = page_to_pgoff(page);
  289. return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
  290. }
  291. static inline unsigned long
  292. vma_address(struct page *page, struct vm_area_struct *vma)
  293. {
  294. unsigned long start, end;
  295. start = __vma_address(page, vma);
  296. end = start + PAGE_SIZE * (hpage_nr_pages(page) - 1);
  297. /* page should be within @vma mapping range */
  298. VM_BUG_ON_VMA(end < vma->vm_start || start >= vma->vm_end, vma);
  299. return max(start, vma->vm_start);
  300. }
  301. #else /* !CONFIG_MMU */
  302. static inline void clear_page_mlock(struct page *page) { }
  303. static inline void mlock_vma_page(struct page *page) { }
  304. static inline void mlock_migrate_page(struct page *new, struct page *old) { }
  305. #endif /* !CONFIG_MMU */
  306. /*
  307. * Return the mem_map entry representing the 'offset' subpage within
  308. * the maximally aligned gigantic page 'base'. Handle any discontiguity
  309. * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
  310. */
  311. static inline struct page *mem_map_offset(struct page *base, int offset)
  312. {
  313. if (unlikely(offset >= MAX_ORDER_NR_PAGES))
  314. return nth_page(base, offset);
  315. return base + offset;
  316. }
  317. /*
  318. * Iterator over all subpages within the maximally aligned gigantic
  319. * page 'base'. Handle any discontiguity in the mem_map.
  320. */
  321. static inline struct page *mem_map_next(struct page *iter,
  322. struct page *base, int offset)
  323. {
  324. if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
  325. unsigned long pfn = page_to_pfn(base) + offset;
  326. if (!pfn_valid(pfn))
  327. return NULL;
  328. return pfn_to_page(pfn);
  329. }
  330. return iter + 1;
  331. }
  332. /*
  333. * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
  334. * so all functions starting at paging_init should be marked __init
  335. * in those cases. SPARSEMEM, however, allows for memory hotplug,
  336. * and alloc_bootmem_node is not used.
  337. */
  338. #ifdef CONFIG_SPARSEMEM
  339. #define __paginginit __meminit
  340. #else
  341. #define __paginginit __init
  342. #endif
  343. /* Memory initialisation debug and verification */
  344. enum mminit_level {
  345. MMINIT_WARNING,
  346. MMINIT_VERIFY,
  347. MMINIT_TRACE
  348. };
  349. #ifdef CONFIG_DEBUG_MEMORY_INIT
  350. extern int mminit_loglevel;
  351. #define mminit_dprintk(level, prefix, fmt, arg...) \
  352. do { \
  353. if (level < mminit_loglevel) { \
  354. if (level <= MMINIT_WARNING) \
  355. pr_warn("mminit::" prefix " " fmt, ##arg); \
  356. else \
  357. printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
  358. } \
  359. } while (0)
  360. extern void mminit_verify_pageflags_layout(void);
  361. extern void mminit_verify_zonelist(void);
  362. #else
  363. static inline void mminit_dprintk(enum mminit_level level,
  364. const char *prefix, const char *fmt, ...)
  365. {
  366. }
  367. static inline void mminit_verify_pageflags_layout(void)
  368. {
  369. }
  370. static inline void mminit_verify_zonelist(void)
  371. {
  372. }
  373. #endif /* CONFIG_DEBUG_MEMORY_INIT */
  374. /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
  375. #if defined(CONFIG_SPARSEMEM)
  376. extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
  377. unsigned long *end_pfn);
  378. #else
  379. static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
  380. unsigned long *end_pfn)
  381. {
  382. }
  383. #endif /* CONFIG_SPARSEMEM */
  384. #define NODE_RECLAIM_NOSCAN -2
  385. #define NODE_RECLAIM_FULL -1
  386. #define NODE_RECLAIM_SOME 0
  387. #define NODE_RECLAIM_SUCCESS 1
  388. extern int hwpoison_filter(struct page *p);
  389. extern u32 hwpoison_filter_dev_major;
  390. extern u32 hwpoison_filter_dev_minor;
  391. extern u64 hwpoison_filter_flags_mask;
  392. extern u64 hwpoison_filter_flags_value;
  393. extern u64 hwpoison_filter_memcg;
  394. extern u32 hwpoison_filter_enable;
  395. extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
  396. unsigned long, unsigned long,
  397. unsigned long, unsigned long);
  398. extern void set_pageblock_order(void);
  399. unsigned long reclaim_clean_pages_from_list(struct zone *zone,
  400. struct list_head *page_list);
  401. /* The ALLOC_WMARK bits are used as an index to zone->watermark */
  402. #define ALLOC_WMARK_MIN WMARK_MIN
  403. #define ALLOC_WMARK_LOW WMARK_LOW
  404. #define ALLOC_WMARK_HIGH WMARK_HIGH
  405. #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
  406. /* Mask to get the watermark bits */
  407. #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
  408. #define ALLOC_HARDER 0x10 /* try to alloc harder */
  409. #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
  410. #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
  411. #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
  412. enum ttu_flags;
  413. struct tlbflush_unmap_batch;
  414. /*
  415. * only for MM internal work items which do not depend on
  416. * any allocations or locks which might depend on allocations
  417. */
  418. extern struct workqueue_struct *mm_percpu_wq;
  419. #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
  420. void try_to_unmap_flush(void);
  421. void try_to_unmap_flush_dirty(void);
  422. #else
  423. static inline void try_to_unmap_flush(void)
  424. {
  425. }
  426. static inline void try_to_unmap_flush_dirty(void)
  427. {
  428. }
  429. #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
  430. extern const struct trace_print_flags pageflag_names[];
  431. extern const struct trace_print_flags vmaflag_names[];
  432. extern const struct trace_print_flags gfpflag_names[];
  433. #endif /* __MM_INTERNAL_H */