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