internal.h 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372
  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/mm.h>
  14. void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  15. unsigned long floor, unsigned long ceiling);
  16. static inline void set_page_count(struct page *page, int v)
  17. {
  18. atomic_set(&page->_count, v);
  19. }
  20. /*
  21. * Turn a non-refcounted page (->_count == 0) into refcounted with
  22. * a count of one.
  23. */
  24. static inline void set_page_refcounted(struct page *page)
  25. {
  26. VM_BUG_ON(PageTail(page));
  27. VM_BUG_ON(atomic_read(&page->_count));
  28. set_page_count(page, 1);
  29. }
  30. static inline void __put_page(struct page *page)
  31. {
  32. atomic_dec(&page->_count);
  33. }
  34. static inline void __get_page_tail_foll(struct page *page,
  35. bool get_page_head)
  36. {
  37. /*
  38. * If we're getting a tail page, the elevated page->_count is
  39. * required only in the head page and we will elevate the head
  40. * page->_count and tail page->_mapcount.
  41. *
  42. * We elevate page_tail->_mapcount for tail pages to force
  43. * page_tail->_count to be zero at all times to avoid getting
  44. * false positives from get_page_unless_zero() with
  45. * speculative page access (like in
  46. * page_cache_get_speculative()) on tail pages.
  47. */
  48. VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
  49. VM_BUG_ON(atomic_read(&page->_count) != 0);
  50. VM_BUG_ON(page_mapcount(page) < 0);
  51. if (get_page_head)
  52. atomic_inc(&page->first_page->_count);
  53. atomic_inc(&page->_mapcount);
  54. }
  55. /*
  56. * This is meant to be called as the FOLL_GET operation of
  57. * follow_page() and it must be called while holding the proper PT
  58. * lock while the pte (or pmd_trans_huge) is still mapping the page.
  59. */
  60. static inline void get_page_foll(struct page *page)
  61. {
  62. if (unlikely(PageTail(page)))
  63. /*
  64. * This is safe only because
  65. * __split_huge_page_refcount() can't run under
  66. * get_page_foll() because we hold the proper PT lock.
  67. */
  68. __get_page_tail_foll(page, true);
  69. else {
  70. /*
  71. * Getting a normal page or the head of a compound page
  72. * requires to already have an elevated page->_count.
  73. */
  74. VM_BUG_ON(atomic_read(&page->_count) <= 0);
  75. atomic_inc(&page->_count);
  76. }
  77. }
  78. extern unsigned long highest_memmap_pfn;
  79. /*
  80. * in mm/vmscan.c:
  81. */
  82. extern int isolate_lru_page(struct page *page);
  83. extern void putback_lru_page(struct page *page);
  84. extern unsigned long zone_reclaimable_pages(struct zone *zone);
  85. extern bool zone_reclaimable(struct zone *zone);
  86. /*
  87. * in mm/page_alloc.c
  88. */
  89. extern void __free_pages_bootmem(struct page *page, unsigned int order);
  90. extern void prep_compound_page(struct page *page, unsigned long order);
  91. #ifdef CONFIG_MEMORY_FAILURE
  92. extern bool is_free_buddy_page(struct page *page);
  93. #endif
  94. #if defined CONFIG_COMPACTION || defined CONFIG_CMA
  95. /*
  96. * in mm/compaction.c
  97. */
  98. /*
  99. * compact_control is used to track pages being migrated and the free pages
  100. * they are being migrated to during memory compaction. The free_pfn starts
  101. * at the end of a zone and migrate_pfn begins at the start. Movable pages
  102. * are moved to the end of a zone during a compaction run and the run
  103. * completes when free_pfn <= migrate_pfn
  104. */
  105. struct compact_control {
  106. struct list_head freepages; /* List of free pages to migrate to */
  107. struct list_head migratepages; /* List of pages being migrated */
  108. unsigned long nr_freepages; /* Number of isolated free pages */
  109. unsigned long nr_migratepages; /* Number of pages to migrate */
  110. unsigned long free_pfn; /* isolate_freepages search base */
  111. unsigned long migrate_pfn; /* isolate_migratepages search base */
  112. bool sync; /* Synchronous migration */
  113. bool ignore_skip_hint; /* Scan blocks even if marked skip */
  114. bool finished_update_free; /* True when the zone cached pfns are
  115. * no longer being updated
  116. */
  117. bool finished_update_migrate;
  118. int order; /* order a direct compactor needs */
  119. int migratetype; /* MOVABLE, RECLAIMABLE etc */
  120. struct zone *zone;
  121. bool contended; /* True if a lock was contended */
  122. };
  123. unsigned long
  124. isolate_freepages_range(struct compact_control *cc,
  125. unsigned long start_pfn, unsigned long end_pfn);
  126. unsigned long
  127. isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
  128. unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
  129. #endif
  130. /*
  131. * function for dealing with page's order in buddy system.
  132. * zone->lock is already acquired when we use these.
  133. * So, we don't need atomic page->flags operations here.
  134. */
  135. static inline unsigned long page_order(struct page *page)
  136. {
  137. /* PageBuddy() must be checked by the caller */
  138. return page_private(page);
  139. }
  140. /* mm/util.c */
  141. void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
  142. struct vm_area_struct *prev, struct rb_node *rb_parent);
  143. #ifdef CONFIG_MMU
  144. extern long mlock_vma_pages_range(struct vm_area_struct *vma,
  145. unsigned long start, unsigned long end);
  146. extern void munlock_vma_pages_range(struct vm_area_struct *vma,
  147. unsigned long start, unsigned long end);
  148. static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
  149. {
  150. munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
  151. }
  152. /*
  153. * Called only in fault path, to determine if a new page is being
  154. * mapped into a LOCKED vma. If it is, mark page as mlocked.
  155. */
  156. static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
  157. struct page *page)
  158. {
  159. VM_BUG_ON(PageLRU(page));
  160. if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
  161. return 0;
  162. if (!TestSetPageMlocked(page)) {
  163. inc_zone_page_state(page, NR_MLOCK);
  164. count_vm_event(UNEVICTABLE_PGMLOCKED);
  165. }
  166. return 1;
  167. }
  168. /*
  169. * must be called with vma's mmap_sem held for read or write, and page locked.
  170. */
  171. extern void mlock_vma_page(struct page *page);
  172. extern void munlock_vma_page(struct page *page);
  173. /*
  174. * Clear the page's PageMlocked(). This can be useful in a situation where
  175. * we want to unconditionally remove a page from the pagecache -- e.g.,
  176. * on truncation or freeing.
  177. *
  178. * It is legal to call this function for any page, mlocked or not.
  179. * If called for a page that is still mapped by mlocked vmas, all we do
  180. * is revert to lazy LRU behaviour -- semantics are not broken.
  181. */
  182. extern void __clear_page_mlock(struct page *page);
  183. static inline void clear_page_mlock(struct page *page)
  184. {
  185. if (unlikely(TestClearPageMlocked(page)))
  186. __clear_page_mlock(page);
  187. }
  188. /*
  189. * mlock_migrate_page - called only from migrate_page_copy() to
  190. * migrate the Mlocked page flag; update statistics.
  191. */
  192. static inline void mlock_migrate_page(struct page *newpage, struct page *page)
  193. {
  194. if (TestClearPageMlocked(page)) {
  195. unsigned long flags;
  196. local_irq_save(flags);
  197. __dec_zone_page_state(page, NR_MLOCK);
  198. SetPageMlocked(newpage);
  199. __inc_zone_page_state(newpage, NR_MLOCK);
  200. local_irq_restore(flags);
  201. }
  202. }
  203. #ifdef CONFIG_TRANSPARENT_HUGEPAGE
  204. extern unsigned long vma_address(struct page *page,
  205. struct vm_area_struct *vma);
  206. #endif
  207. #else /* !CONFIG_MMU */
  208. static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
  209. {
  210. return 0;
  211. }
  212. static inline void clear_page_mlock(struct page *page) { }
  213. static inline void mlock_vma_page(struct page *page) { }
  214. static inline void mlock_migrate_page(struct page *new, struct page *old) { }
  215. #endif /* !CONFIG_MMU */
  216. /*
  217. * Return the mem_map entry representing the 'offset' subpage within
  218. * the maximally aligned gigantic page 'base'. Handle any discontiguity
  219. * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
  220. */
  221. static inline struct page *mem_map_offset(struct page *base, int offset)
  222. {
  223. if (unlikely(offset >= MAX_ORDER_NR_PAGES))
  224. return pfn_to_page(page_to_pfn(base) + offset);
  225. return base + offset;
  226. }
  227. /*
  228. * Iterator over all subpages within the maximally aligned gigantic
  229. * page 'base'. Handle any discontiguity in the mem_map.
  230. */
  231. static inline struct page *mem_map_next(struct page *iter,
  232. struct page *base, int offset)
  233. {
  234. if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
  235. unsigned long pfn = page_to_pfn(base) + offset;
  236. if (!pfn_valid(pfn))
  237. return NULL;
  238. return pfn_to_page(pfn);
  239. }
  240. return iter + 1;
  241. }
  242. /*
  243. * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
  244. * so all functions starting at paging_init should be marked __init
  245. * in those cases. SPARSEMEM, however, allows for memory hotplug,
  246. * and alloc_bootmem_node is not used.
  247. */
  248. #ifdef CONFIG_SPARSEMEM
  249. #define __paginginit __meminit
  250. #else
  251. #define __paginginit __init
  252. #endif
  253. /* Memory initialisation debug and verification */
  254. enum mminit_level {
  255. MMINIT_WARNING,
  256. MMINIT_VERIFY,
  257. MMINIT_TRACE
  258. };
  259. #ifdef CONFIG_DEBUG_MEMORY_INIT
  260. extern int mminit_loglevel;
  261. #define mminit_dprintk(level, prefix, fmt, arg...) \
  262. do { \
  263. if (level < mminit_loglevel) { \
  264. printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
  265. printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
  266. } \
  267. } while (0)
  268. extern void mminit_verify_pageflags_layout(void);
  269. extern void mminit_verify_page_links(struct page *page,
  270. enum zone_type zone, unsigned long nid, unsigned long pfn);
  271. extern void mminit_verify_zonelist(void);
  272. #else
  273. static inline void mminit_dprintk(enum mminit_level level,
  274. const char *prefix, const char *fmt, ...)
  275. {
  276. }
  277. static inline void mminit_verify_pageflags_layout(void)
  278. {
  279. }
  280. static inline void mminit_verify_page_links(struct page *page,
  281. enum zone_type zone, unsigned long nid, unsigned long pfn)
  282. {
  283. }
  284. static inline void mminit_verify_zonelist(void)
  285. {
  286. }
  287. #endif /* CONFIG_DEBUG_MEMORY_INIT */
  288. /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
  289. #if defined(CONFIG_SPARSEMEM)
  290. extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
  291. unsigned long *end_pfn);
  292. #else
  293. static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
  294. unsigned long *end_pfn)
  295. {
  296. }
  297. #endif /* CONFIG_SPARSEMEM */
  298. #define ZONE_RECLAIM_NOSCAN -2
  299. #define ZONE_RECLAIM_FULL -1
  300. #define ZONE_RECLAIM_SOME 0
  301. #define ZONE_RECLAIM_SUCCESS 1
  302. #endif
  303. extern int hwpoison_filter(struct page *p);
  304. extern u32 hwpoison_filter_dev_major;
  305. extern u32 hwpoison_filter_dev_minor;
  306. extern u64 hwpoison_filter_flags_mask;
  307. extern u64 hwpoison_filter_flags_value;
  308. extern u64 hwpoison_filter_memcg;
  309. extern u32 hwpoison_filter_enable;
  310. /* The ALLOC_WMARK bits are used as an index to zone->watermark */
  311. #define ALLOC_WMARK_MIN WMARK_MIN
  312. #define ALLOC_WMARK_LOW WMARK_LOW
  313. #define ALLOC_WMARK_HIGH WMARK_HIGH
  314. #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
  315. /* Mask to get the watermark bits */
  316. #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
  317. #define ALLOC_HARDER 0x10 /* try to alloc harder */
  318. #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
  319. #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
  320. #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
  321. unsigned long reclaim_clean_pages_from_list(struct zone *zone,
  322. struct list_head *page_list);
  323. extern void set_pageblock_order(void);