gfp.h 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414
  1. #ifndef __LINUX_GFP_H
  2. #define __LINUX_GFP_H
  3. #include <linux/mmdebug.h>
  4. #include <linux/mmzone.h>
  5. #include <linux/stddef.h>
  6. #include <linux/linkage.h>
  7. #include <linux/topology.h>
  8. struct vm_area_struct;
  9. /* Plain integer GFP bitmasks. Do not use this directly. */
  10. #define ___GFP_DMA 0x01u
  11. #define ___GFP_HIGHMEM 0x02u
  12. #define ___GFP_DMA32 0x04u
  13. #define ___GFP_MOVABLE 0x08u
  14. #define ___GFP_WAIT 0x10u
  15. #define ___GFP_HIGH 0x20u
  16. #define ___GFP_IO 0x40u
  17. #define ___GFP_FS 0x80u
  18. #define ___GFP_COLD 0x100u
  19. #define ___GFP_NOWARN 0x200u
  20. #define ___GFP_REPEAT 0x400u
  21. #define ___GFP_NOFAIL 0x800u
  22. #define ___GFP_NORETRY 0x1000u
  23. #define ___GFP_MEMALLOC 0x2000u
  24. #define ___GFP_COMP 0x4000u
  25. #define ___GFP_ZERO 0x8000u
  26. #define ___GFP_NOMEMALLOC 0x10000u
  27. #define ___GFP_HARDWALL 0x20000u
  28. #define ___GFP_THISNODE 0x40000u
  29. #define ___GFP_RECLAIMABLE 0x80000u
  30. #define ___GFP_NOTRACK 0x200000u
  31. #define ___GFP_NO_KSWAPD 0x400000u
  32. #define ___GFP_OTHER_NODE 0x800000u
  33. #define ___GFP_WRITE 0x1000000u
  34. /* If the above are modified, __GFP_BITS_SHIFT may need updating */
  35. /*
  36. * GFP bitmasks..
  37. *
  38. * Zone modifiers (see linux/mmzone.h - low three bits)
  39. *
  40. * Do not put any conditional on these. If necessary modify the definitions
  41. * without the underscores and use them consistently. The definitions here may
  42. * be used in bit comparisons.
  43. */
  44. #define __GFP_DMA ((__force gfp_t)___GFP_DMA)
  45. #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
  46. #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
  47. #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */
  48. #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
  49. /*
  50. * Action modifiers - doesn't change the zoning
  51. *
  52. * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
  53. * _might_ fail. This depends upon the particular VM implementation.
  54. *
  55. * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
  56. * cannot handle allocation failures. This modifier is deprecated and no new
  57. * users should be added.
  58. *
  59. * __GFP_NORETRY: The VM implementation must not retry indefinitely.
  60. *
  61. * __GFP_MOVABLE: Flag that this page will be movable by the page migration
  62. * mechanism or reclaimed
  63. */
  64. #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */
  65. #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */
  66. #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */
  67. #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */
  68. #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */
  69. #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */
  70. #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */
  71. #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */
  72. #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */
  73. #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */
  74. #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */
  75. #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */
  76. #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves.
  77. * This takes precedence over the
  78. * __GFP_MEMALLOC flag if both are
  79. * set
  80. */
  81. #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */
  82. #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */
  83. #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */
  84. #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */
  85. #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD)
  86. #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */
  87. #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE) /* Allocator intends to dirty page */
  88. /*
  89. * This may seem redundant, but it's a way of annotating false positives vs.
  90. * allocations that simply cannot be supported (e.g. page tables).
  91. */
  92. #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
  93. #define __GFP_BITS_SHIFT 25 /* Room for N __GFP_FOO bits */
  94. #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
  95. /* This equals 0, but use constants in case they ever change */
  96. #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH)
  97. /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
  98. #define GFP_ATOMIC (__GFP_HIGH)
  99. #define GFP_NOIO (__GFP_WAIT)
  100. #define GFP_NOFS (__GFP_WAIT | __GFP_IO)
  101. #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS)
  102. #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \
  103. __GFP_RECLAIMABLE)
  104. #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
  105. #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM)
  106. #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE)
  107. #define GFP_IOFS (__GFP_IO | __GFP_FS)
  108. #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
  109. __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \
  110. __GFP_NO_KSWAPD)
  111. /* This mask makes up all the page movable related flags */
  112. #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
  113. /* Control page allocator reclaim behavior */
  114. #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
  115. __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
  116. __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
  117. /* Control slab gfp mask during early boot */
  118. #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
  119. /* Control allocation constraints */
  120. #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
  121. /* Do not use these with a slab allocator */
  122. #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
  123. /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some
  124. platforms, used as appropriate on others */
  125. #define GFP_DMA __GFP_DMA
  126. /* 4GB DMA on some platforms */
  127. #define GFP_DMA32 __GFP_DMA32
  128. /* Convert GFP flags to their corresponding migrate type */
  129. static inline int gfpflags_to_migratetype(const gfp_t gfp_flags)
  130. {
  131. WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
  132. if (unlikely(page_group_by_mobility_disabled))
  133. return MIGRATE_UNMOVABLE;
  134. /* Group based on mobility */
  135. return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
  136. ((gfp_flags & __GFP_RECLAIMABLE) != 0);
  137. }
  138. #ifdef CONFIG_HIGHMEM
  139. #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
  140. #else
  141. #define OPT_ZONE_HIGHMEM ZONE_NORMAL
  142. #endif
  143. #ifdef CONFIG_ZONE_DMA
  144. #define OPT_ZONE_DMA ZONE_DMA
  145. #else
  146. #define OPT_ZONE_DMA ZONE_NORMAL
  147. #endif
  148. #ifdef CONFIG_ZONE_DMA32
  149. #define OPT_ZONE_DMA32 ZONE_DMA32
  150. #else
  151. #define OPT_ZONE_DMA32 ZONE_NORMAL
  152. #endif
  153. /*
  154. * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
  155. * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
  156. * and there are 16 of them to cover all possible combinations of
  157. * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
  158. *
  159. * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
  160. * But GFP_MOVABLE is not only a zone specifier but also an allocation
  161. * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
  162. * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
  163. *
  164. * bit result
  165. * =================
  166. * 0x0 => NORMAL
  167. * 0x1 => DMA or NORMAL
  168. * 0x2 => HIGHMEM or NORMAL
  169. * 0x3 => BAD (DMA+HIGHMEM)
  170. * 0x4 => DMA32 or DMA or NORMAL
  171. * 0x5 => BAD (DMA+DMA32)
  172. * 0x6 => BAD (HIGHMEM+DMA32)
  173. * 0x7 => BAD (HIGHMEM+DMA32+DMA)
  174. * 0x8 => NORMAL (MOVABLE+0)
  175. * 0x9 => DMA or NORMAL (MOVABLE+DMA)
  176. * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
  177. * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
  178. * 0xc => DMA32 (MOVABLE+DMA32)
  179. * 0xd => BAD (MOVABLE+DMA32+DMA)
  180. * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
  181. * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
  182. *
  183. * ZONES_SHIFT must be <= 2 on 32 bit platforms.
  184. */
  185. #if 16 * ZONES_SHIFT > BITS_PER_LONG
  186. #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
  187. #endif
  188. #define GFP_ZONE_TABLE ( \
  189. (ZONE_NORMAL << 0 * ZONES_SHIFT) \
  190. | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \
  191. | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \
  192. | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \
  193. | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \
  194. | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \
  195. | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \
  196. | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \
  197. )
  198. /*
  199. * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
  200. * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
  201. * entry starting with bit 0. Bit is set if the combination is not
  202. * allowed.
  203. */
  204. #define GFP_ZONE_BAD ( \
  205. 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
  206. | 1 << (___GFP_DMA | ___GFP_DMA32) \
  207. | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
  208. | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
  209. | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
  210. | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
  211. | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
  212. | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
  213. )
  214. static inline enum zone_type gfp_zone(gfp_t flags)
  215. {
  216. enum zone_type z;
  217. int bit = (__force int) (flags & GFP_ZONEMASK);
  218. z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
  219. ((1 << ZONES_SHIFT) - 1);
  220. VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
  221. return z;
  222. }
  223. /*
  224. * There is only one page-allocator function, and two main namespaces to
  225. * it. The alloc_page*() variants return 'struct page *' and as such
  226. * can allocate highmem pages, the *get*page*() variants return
  227. * virtual kernel addresses to the allocated page(s).
  228. */
  229. static inline int gfp_zonelist(gfp_t flags)
  230. {
  231. if (IS_ENABLED(CONFIG_NUMA) && unlikely(flags & __GFP_THISNODE))
  232. return 1;
  233. return 0;
  234. }
  235. /*
  236. * We get the zone list from the current node and the gfp_mask.
  237. * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
  238. * There are two zonelists per node, one for all zones with memory and
  239. * one containing just zones from the node the zonelist belongs to.
  240. *
  241. * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
  242. * optimized to &contig_page_data at compile-time.
  243. */
  244. static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
  245. {
  246. return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
  247. }
  248. #ifndef HAVE_ARCH_FREE_PAGE
  249. static inline void arch_free_page(struct page *page, int order) { }
  250. #endif
  251. #ifndef HAVE_ARCH_ALLOC_PAGE
  252. static inline void arch_alloc_page(struct page *page, int order) { }
  253. #endif
  254. struct page *
  255. __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
  256. struct zonelist *zonelist, nodemask_t *nodemask);
  257. static inline struct page *
  258. __alloc_pages(gfp_t gfp_mask, unsigned int order,
  259. struct zonelist *zonelist)
  260. {
  261. return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
  262. }
  263. static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
  264. unsigned int order)
  265. {
  266. /* Unknown node is current node */
  267. if (nid < 0)
  268. nid = numa_node_id();
  269. return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
  270. }
  271. static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask,
  272. unsigned int order)
  273. {
  274. VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES || !node_online(nid));
  275. return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
  276. }
  277. #ifdef CONFIG_NUMA
  278. extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
  279. static inline struct page *
  280. alloc_pages(gfp_t gfp_mask, unsigned int order)
  281. {
  282. return alloc_pages_current(gfp_mask, order);
  283. }
  284. extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
  285. struct vm_area_struct *vma, unsigned long addr,
  286. int node, bool hugepage);
  287. #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
  288. alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true)
  289. #else
  290. #define alloc_pages(gfp_mask, order) \
  291. alloc_pages_node(numa_node_id(), gfp_mask, order)
  292. #define alloc_pages_vma(gfp_mask, order, vma, addr, node, false)\
  293. alloc_pages(gfp_mask, order)
  294. #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
  295. alloc_pages(gfp_mask, order)
  296. #endif
  297. #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
  298. #define alloc_page_vma(gfp_mask, vma, addr) \
  299. alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id(), false)
  300. #define alloc_page_vma_node(gfp_mask, vma, addr, node) \
  301. alloc_pages_vma(gfp_mask, 0, vma, addr, node, false)
  302. extern struct page *alloc_kmem_pages(gfp_t gfp_mask, unsigned int order);
  303. extern struct page *alloc_kmem_pages_node(int nid, gfp_t gfp_mask,
  304. unsigned int order);
  305. extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
  306. extern unsigned long get_zeroed_page(gfp_t gfp_mask);
  307. void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
  308. void free_pages_exact(void *virt, size_t size);
  309. /* This is different from alloc_pages_exact_node !!! */
  310. void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
  311. #define __get_free_page(gfp_mask) \
  312. __get_free_pages((gfp_mask), 0)
  313. #define __get_dma_pages(gfp_mask, order) \
  314. __get_free_pages((gfp_mask) | GFP_DMA, (order))
  315. extern void __free_pages(struct page *page, unsigned int order);
  316. extern void free_pages(unsigned long addr, unsigned int order);
  317. extern void free_hot_cold_page(struct page *page, bool cold);
  318. extern void free_hot_cold_page_list(struct list_head *list, bool cold);
  319. extern void __free_kmem_pages(struct page *page, unsigned int order);
  320. extern void free_kmem_pages(unsigned long addr, unsigned int order);
  321. #define __free_page(page) __free_pages((page), 0)
  322. #define free_page(addr) free_pages((addr), 0)
  323. void page_alloc_init(void);
  324. void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
  325. void drain_all_pages(struct zone *zone);
  326. void drain_local_pages(struct zone *zone);
  327. /*
  328. * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
  329. * GFP flags are used before interrupts are enabled. Once interrupts are
  330. * enabled, it is set to __GFP_BITS_MASK while the system is running. During
  331. * hibernation, it is used by PM to avoid I/O during memory allocation while
  332. * devices are suspended.
  333. */
  334. extern gfp_t gfp_allowed_mask;
  335. /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
  336. bool gfp_pfmemalloc_allowed(gfp_t gfp_mask);
  337. extern void pm_restrict_gfp_mask(void);
  338. extern void pm_restore_gfp_mask(void);
  339. #ifdef CONFIG_PM_SLEEP
  340. extern bool pm_suspended_storage(void);
  341. #else
  342. static inline bool pm_suspended_storage(void)
  343. {
  344. return false;
  345. }
  346. #endif /* CONFIG_PM_SLEEP */
  347. #ifdef CONFIG_CMA
  348. /* The below functions must be run on a range from a single zone. */
  349. extern int alloc_contig_range(unsigned long start, unsigned long end,
  350. unsigned migratetype);
  351. extern void free_contig_range(unsigned long pfn, unsigned nr_pages);
  352. /* CMA stuff */
  353. extern void init_cma_reserved_pageblock(struct page *page);
  354. #endif
  355. #endif /* __LINUX_GFP_H */