gfp.h 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554
  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_RECLAIMABLE 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_ATOMIC 0x80000u
  30. #define ___GFP_NOACCOUNT 0x100000u
  31. #define ___GFP_NOTRACK 0x200000u
  32. #define ___GFP_DIRECT_RECLAIM 0x400000u
  33. #define ___GFP_OTHER_NODE 0x800000u
  34. #define ___GFP_WRITE 0x1000000u
  35. #define ___GFP_KSWAPD_RECLAIM 0x2000000u
  36. /* If the above are modified, __GFP_BITS_SHIFT may need updating */
  37. /*
  38. * Physical address zone modifiers (see linux/mmzone.h - low four 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_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* ZONE_MOVABLE allowed */
  49. #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
  50. /*
  51. * Page mobility and placement hints
  52. *
  53. * These flags provide hints about how mobile the page is. Pages with similar
  54. * mobility are placed within the same pageblocks to minimise problems due
  55. * to external fragmentation.
  56. *
  57. * __GFP_MOVABLE (also a zone modifier) indicates that the page can be
  58. * moved by page migration during memory compaction or can be reclaimed.
  59. *
  60. * __GFP_RECLAIMABLE is used for slab allocations that specify
  61. * SLAB_RECLAIM_ACCOUNT and whose pages can be freed via shrinkers.
  62. *
  63. * __GFP_WRITE indicates the caller intends to dirty the page. Where possible,
  64. * these pages will be spread between local zones to avoid all the dirty
  65. * pages being in one zone (fair zone allocation policy).
  66. *
  67. * __GFP_HARDWALL enforces the cpuset memory allocation policy.
  68. *
  69. * __GFP_THISNODE forces the allocation to be satisified from the requested
  70. * node with no fallbacks or placement policy enforcements.
  71. */
  72. #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE)
  73. #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE)
  74. #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL)
  75. #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)
  76. /*
  77. * Watermark modifiers -- controls access to emergency reserves
  78. *
  79. * __GFP_HIGH indicates that the caller is high-priority and that granting
  80. * the request is necessary before the system can make forward progress.
  81. * For example, creating an IO context to clean pages.
  82. *
  83. * __GFP_ATOMIC indicates that the caller cannot reclaim or sleep and is
  84. * high priority. Users are typically interrupt handlers. This may be
  85. * used in conjunction with __GFP_HIGH
  86. *
  87. * __GFP_MEMALLOC allows access to all memory. This should only be used when
  88. * the caller guarantees the allocation will allow more memory to be freed
  89. * very shortly e.g. process exiting or swapping. Users either should
  90. * be the MM or co-ordinating closely with the VM (e.g. swap over NFS).
  91. *
  92. * __GFP_NOMEMALLOC is used to explicitly forbid access to emergency reserves.
  93. * This takes precedence over the __GFP_MEMALLOC flag if both are set.
  94. *
  95. * __GFP_NOACCOUNT ignores the accounting for kmemcg limit enforcement.
  96. */
  97. #define __GFP_ATOMIC ((__force gfp_t)___GFP_ATOMIC)
  98. #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH)
  99. #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)
  100. #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC)
  101. #define __GFP_NOACCOUNT ((__force gfp_t)___GFP_NOACCOUNT)
  102. /*
  103. * Reclaim modifiers
  104. *
  105. * __GFP_IO can start physical IO.
  106. *
  107. * __GFP_FS can call down to the low-level FS. Clearing the flag avoids the
  108. * allocator recursing into the filesystem which might already be holding
  109. * locks.
  110. *
  111. * __GFP_DIRECT_RECLAIM indicates that the caller may enter direct reclaim.
  112. * This flag can be cleared to avoid unnecessary delays when a fallback
  113. * option is available.
  114. *
  115. * __GFP_KSWAPD_RECLAIM indicates that the caller wants to wake kswapd when
  116. * the low watermark is reached and have it reclaim pages until the high
  117. * watermark is reached. A caller may wish to clear this flag when fallback
  118. * options are available and the reclaim is likely to disrupt the system. The
  119. * canonical example is THP allocation where a fallback is cheap but
  120. * reclaim/compaction may cause indirect stalls.
  121. *
  122. * __GFP_RECLAIM is shorthand to allow/forbid both direct and kswapd reclaim.
  123. *
  124. * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
  125. * _might_ fail. This depends upon the particular VM implementation.
  126. *
  127. * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
  128. * cannot handle allocation failures. New users should be evaluated carefully
  129. * (and the flag should be used only when there is no reasonable failure
  130. * policy) but it is definitely preferable to use the flag rather than
  131. * opencode endless loop around allocator.
  132. *
  133. * __GFP_NORETRY: The VM implementation must not retry indefinitely and will
  134. * return NULL when direct reclaim and memory compaction have failed to allow
  135. * the allocation to succeed. The OOM killer is not called with the current
  136. * implementation.
  137. */
  138. #define __GFP_IO ((__force gfp_t)___GFP_IO)
  139. #define __GFP_FS ((__force gfp_t)___GFP_FS)
  140. #define __GFP_DIRECT_RECLAIM ((__force gfp_t)___GFP_DIRECT_RECLAIM) /* Caller can reclaim */
  141. #define __GFP_KSWAPD_RECLAIM ((__force gfp_t)___GFP_KSWAPD_RECLAIM) /* kswapd can wake */
  142. #define __GFP_RECLAIM ((__force gfp_t)(___GFP_DIRECT_RECLAIM|___GFP_KSWAPD_RECLAIM))
  143. #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT)
  144. #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL)
  145. #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY)
  146. /*
  147. * Action modifiers
  148. *
  149. * __GFP_COLD indicates that the caller does not expect to be used in the near
  150. * future. Where possible, a cache-cold page will be returned.
  151. *
  152. * __GFP_NOWARN suppresses allocation failure reports.
  153. *
  154. * __GFP_COMP address compound page metadata.
  155. *
  156. * __GFP_ZERO returns a zeroed page on success.
  157. *
  158. * __GFP_NOTRACK avoids tracking with kmemcheck.
  159. *
  160. * __GFP_NOTRACK_FALSE_POSITIVE is an alias of __GFP_NOTRACK. It's a means of
  161. * distinguishing in the source between false positives and allocations that
  162. * cannot be supported (e.g. page tables).
  163. *
  164. * __GFP_OTHER_NODE is for allocations that are on a remote node but that
  165. * should not be accounted for as a remote allocation in vmstat. A
  166. * typical user would be khugepaged collapsing a huge page on a remote
  167. * node.
  168. */
  169. #define __GFP_COLD ((__force gfp_t)___GFP_COLD)
  170. #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN)
  171. #define __GFP_COMP ((__force gfp_t)___GFP_COMP)
  172. #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO)
  173. #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK)
  174. #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
  175. #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE)
  176. /* Room for N __GFP_FOO bits */
  177. #define __GFP_BITS_SHIFT 26
  178. #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
  179. /*
  180. * Useful GFP flag combinations that are commonly used. It is recommended
  181. * that subsystems start with one of these combinations and then set/clear
  182. * __GFP_FOO flags as necessary.
  183. *
  184. * GFP_ATOMIC users can not sleep and need the allocation to succeed. A lower
  185. * watermark is applied to allow access to "atomic reserves"
  186. *
  187. * GFP_KERNEL is typical for kernel-internal allocations. The caller requires
  188. * ZONE_NORMAL or a lower zone for direct access but can direct reclaim.
  189. *
  190. * GFP_NOWAIT is for kernel allocations that should not stall for direct
  191. * reclaim, start physical IO or use any filesystem callback.
  192. *
  193. * GFP_NOIO will use direct reclaim to discard clean pages or slab pages
  194. * that do not require the starting of any physical IO.
  195. *
  196. * GFP_NOFS will use direct reclaim but will not use any filesystem interfaces.
  197. *
  198. * GFP_USER is for userspace allocations that also need to be directly
  199. * accessibly by the kernel or hardware. It is typically used by hardware
  200. * for buffers that are mapped to userspace (e.g. graphics) that hardware
  201. * still must DMA to. cpuset limits are enforced for these allocations.
  202. *
  203. * GFP_DMA exists for historical reasons and should be avoided where possible.
  204. * The flags indicates that the caller requires that the lowest zone be
  205. * used (ZONE_DMA or 16M on x86-64). Ideally, this would be removed but
  206. * it would require careful auditing as some users really require it and
  207. * others use the flag to avoid lowmem reserves in ZONE_DMA and treat the
  208. * lowest zone as a type of emergency reserve.
  209. *
  210. * GFP_DMA32 is similar to GFP_DMA except that the caller requires a 32-bit
  211. * address.
  212. *
  213. * GFP_HIGHUSER is for userspace allocations that may be mapped to userspace,
  214. * do not need to be directly accessible by the kernel but that cannot
  215. * move once in use. An example may be a hardware allocation that maps
  216. * data directly into userspace but has no addressing limitations.
  217. *
  218. * GFP_HIGHUSER_MOVABLE is for userspace allocations that the kernel does not
  219. * need direct access to but can use kmap() when access is required. They
  220. * are expected to be movable via page reclaim or page migration. Typically,
  221. * pages on the LRU would also be allocated with GFP_HIGHUSER_MOVABLE.
  222. *
  223. * GFP_TRANSHUGE is used for THP allocations. They are compound allocations
  224. * that will fail quickly if memory is not available and will not wake
  225. * kswapd on failure.
  226. */
  227. #define GFP_ATOMIC (__GFP_HIGH|__GFP_ATOMIC|__GFP_KSWAPD_RECLAIM)
  228. #define GFP_KERNEL (__GFP_RECLAIM | __GFP_IO | __GFP_FS)
  229. #define GFP_NOWAIT (__GFP_KSWAPD_RECLAIM)
  230. #define GFP_NOIO (__GFP_RECLAIM)
  231. #define GFP_NOFS (__GFP_RECLAIM | __GFP_IO)
  232. #define GFP_TEMPORARY (__GFP_RECLAIM | __GFP_IO | __GFP_FS | \
  233. __GFP_RECLAIMABLE)
  234. #define GFP_USER (__GFP_RECLAIM | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
  235. #define GFP_DMA __GFP_DMA
  236. #define GFP_DMA32 __GFP_DMA32
  237. #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM)
  238. #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE)
  239. #define GFP_TRANSHUGE ((GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
  240. __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN) & \
  241. ~__GFP_KSWAPD_RECLAIM)
  242. /* Convert GFP flags to their corresponding migrate type */
  243. #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
  244. #define GFP_MOVABLE_SHIFT 3
  245. static inline int gfpflags_to_migratetype(const gfp_t gfp_flags)
  246. {
  247. VM_WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
  248. BUILD_BUG_ON((1UL << GFP_MOVABLE_SHIFT) != ___GFP_MOVABLE);
  249. BUILD_BUG_ON((___GFP_MOVABLE >> GFP_MOVABLE_SHIFT) != MIGRATE_MOVABLE);
  250. if (unlikely(page_group_by_mobility_disabled))
  251. return MIGRATE_UNMOVABLE;
  252. /* Group based on mobility */
  253. return (gfp_flags & GFP_MOVABLE_MASK) >> GFP_MOVABLE_SHIFT;
  254. }
  255. #undef GFP_MOVABLE_MASK
  256. #undef GFP_MOVABLE_SHIFT
  257. static inline bool gfpflags_allow_blocking(const gfp_t gfp_flags)
  258. {
  259. return (bool __force)(gfp_flags & __GFP_DIRECT_RECLAIM);
  260. }
  261. #ifdef CONFIG_HIGHMEM
  262. #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
  263. #else
  264. #define OPT_ZONE_HIGHMEM ZONE_NORMAL
  265. #endif
  266. #ifdef CONFIG_ZONE_DMA
  267. #define OPT_ZONE_DMA ZONE_DMA
  268. #else
  269. #define OPT_ZONE_DMA ZONE_NORMAL
  270. #endif
  271. #ifdef CONFIG_ZONE_DMA32
  272. #define OPT_ZONE_DMA32 ZONE_DMA32
  273. #else
  274. #define OPT_ZONE_DMA32 ZONE_NORMAL
  275. #endif
  276. /*
  277. * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
  278. * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
  279. * and there are 16 of them to cover all possible combinations of
  280. * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
  281. *
  282. * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
  283. * But GFP_MOVABLE is not only a zone specifier but also an allocation
  284. * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
  285. * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
  286. *
  287. * bit result
  288. * =================
  289. * 0x0 => NORMAL
  290. * 0x1 => DMA or NORMAL
  291. * 0x2 => HIGHMEM or NORMAL
  292. * 0x3 => BAD (DMA+HIGHMEM)
  293. * 0x4 => DMA32 or DMA or NORMAL
  294. * 0x5 => BAD (DMA+DMA32)
  295. * 0x6 => BAD (HIGHMEM+DMA32)
  296. * 0x7 => BAD (HIGHMEM+DMA32+DMA)
  297. * 0x8 => NORMAL (MOVABLE+0)
  298. * 0x9 => DMA or NORMAL (MOVABLE+DMA)
  299. * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
  300. * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
  301. * 0xc => DMA32 (MOVABLE+DMA32)
  302. * 0xd => BAD (MOVABLE+DMA32+DMA)
  303. * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
  304. * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
  305. *
  306. * ZONES_SHIFT must be <= 2 on 32 bit platforms.
  307. */
  308. #if 16 * ZONES_SHIFT > BITS_PER_LONG
  309. #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
  310. #endif
  311. #define GFP_ZONE_TABLE ( \
  312. (ZONE_NORMAL << 0 * ZONES_SHIFT) \
  313. | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \
  314. | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \
  315. | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \
  316. | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \
  317. | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \
  318. | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \
  319. | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \
  320. )
  321. /*
  322. * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
  323. * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
  324. * entry starting with bit 0. Bit is set if the combination is not
  325. * allowed.
  326. */
  327. #define GFP_ZONE_BAD ( \
  328. 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
  329. | 1 << (___GFP_DMA | ___GFP_DMA32) \
  330. | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
  331. | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
  332. | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
  333. | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
  334. | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
  335. | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
  336. )
  337. static inline enum zone_type gfp_zone(gfp_t flags)
  338. {
  339. enum zone_type z;
  340. int bit = (__force int) (flags & GFP_ZONEMASK);
  341. z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
  342. ((1 << ZONES_SHIFT) - 1);
  343. VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
  344. return z;
  345. }
  346. /*
  347. * There is only one page-allocator function, and two main namespaces to
  348. * it. The alloc_page*() variants return 'struct page *' and as such
  349. * can allocate highmem pages, the *get*page*() variants return
  350. * virtual kernel addresses to the allocated page(s).
  351. */
  352. static inline int gfp_zonelist(gfp_t flags)
  353. {
  354. if (IS_ENABLED(CONFIG_NUMA) && unlikely(flags & __GFP_THISNODE))
  355. return 1;
  356. return 0;
  357. }
  358. /*
  359. * We get the zone list from the current node and the gfp_mask.
  360. * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
  361. * There are two zonelists per node, one for all zones with memory and
  362. * one containing just zones from the node the zonelist belongs to.
  363. *
  364. * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
  365. * optimized to &contig_page_data at compile-time.
  366. */
  367. static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
  368. {
  369. return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
  370. }
  371. #ifndef HAVE_ARCH_FREE_PAGE
  372. static inline void arch_free_page(struct page *page, int order) { }
  373. #endif
  374. #ifndef HAVE_ARCH_ALLOC_PAGE
  375. static inline void arch_alloc_page(struct page *page, int order) { }
  376. #endif
  377. struct page *
  378. __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
  379. struct zonelist *zonelist, nodemask_t *nodemask);
  380. static inline struct page *
  381. __alloc_pages(gfp_t gfp_mask, unsigned int order,
  382. struct zonelist *zonelist)
  383. {
  384. return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
  385. }
  386. /*
  387. * Allocate pages, preferring the node given as nid. The node must be valid and
  388. * online. For more general interface, see alloc_pages_node().
  389. */
  390. static inline struct page *
  391. __alloc_pages_node(int nid, gfp_t gfp_mask, unsigned int order)
  392. {
  393. VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
  394. VM_WARN_ON(!node_online(nid));
  395. return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
  396. }
  397. /*
  398. * Allocate pages, preferring the node given as nid. When nid == NUMA_NO_NODE,
  399. * prefer the current CPU's closest node. Otherwise node must be valid and
  400. * online.
  401. */
  402. static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
  403. unsigned int order)
  404. {
  405. if (nid == NUMA_NO_NODE)
  406. nid = numa_mem_id();
  407. return __alloc_pages_node(nid, gfp_mask, order);
  408. }
  409. #ifdef CONFIG_NUMA
  410. extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
  411. static inline struct page *
  412. alloc_pages(gfp_t gfp_mask, unsigned int order)
  413. {
  414. return alloc_pages_current(gfp_mask, order);
  415. }
  416. extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
  417. struct vm_area_struct *vma, unsigned long addr,
  418. int node, bool hugepage);
  419. #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
  420. alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true)
  421. #else
  422. #define alloc_pages(gfp_mask, order) \
  423. alloc_pages_node(numa_node_id(), gfp_mask, order)
  424. #define alloc_pages_vma(gfp_mask, order, vma, addr, node, false)\
  425. alloc_pages(gfp_mask, order)
  426. #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
  427. alloc_pages(gfp_mask, order)
  428. #endif
  429. #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
  430. #define alloc_page_vma(gfp_mask, vma, addr) \
  431. alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id(), false)
  432. #define alloc_page_vma_node(gfp_mask, vma, addr, node) \
  433. alloc_pages_vma(gfp_mask, 0, vma, addr, node, false)
  434. extern struct page *alloc_kmem_pages(gfp_t gfp_mask, unsigned int order);
  435. extern struct page *alloc_kmem_pages_node(int nid, gfp_t gfp_mask,
  436. unsigned int order);
  437. extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
  438. extern unsigned long get_zeroed_page(gfp_t gfp_mask);
  439. void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
  440. void free_pages_exact(void *virt, size_t size);
  441. void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
  442. #define __get_free_page(gfp_mask) \
  443. __get_free_pages((gfp_mask), 0)
  444. #define __get_dma_pages(gfp_mask, order) \
  445. __get_free_pages((gfp_mask) | GFP_DMA, (order))
  446. extern void __free_pages(struct page *page, unsigned int order);
  447. extern void free_pages(unsigned long addr, unsigned int order);
  448. extern void free_hot_cold_page(struct page *page, bool cold);
  449. extern void free_hot_cold_page_list(struct list_head *list, bool cold);
  450. struct page_frag_cache;
  451. extern void *__alloc_page_frag(struct page_frag_cache *nc,
  452. unsigned int fragsz, gfp_t gfp_mask);
  453. extern void __free_page_frag(void *addr);
  454. extern void __free_kmem_pages(struct page *page, unsigned int order);
  455. extern void free_kmem_pages(unsigned long addr, unsigned int order);
  456. #define __free_page(page) __free_pages((page), 0)
  457. #define free_page(addr) free_pages((addr), 0)
  458. void page_alloc_init(void);
  459. void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
  460. void drain_all_pages(struct zone *zone);
  461. void drain_local_pages(struct zone *zone);
  462. #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
  463. void page_alloc_init_late(void);
  464. #else
  465. static inline void page_alloc_init_late(void)
  466. {
  467. }
  468. #endif
  469. /*
  470. * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
  471. * GFP flags are used before interrupts are enabled. Once interrupts are
  472. * enabled, it is set to __GFP_BITS_MASK while the system is running. During
  473. * hibernation, it is used by PM to avoid I/O during memory allocation while
  474. * devices are suspended.
  475. */
  476. extern gfp_t gfp_allowed_mask;
  477. /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
  478. bool gfp_pfmemalloc_allowed(gfp_t gfp_mask);
  479. extern void pm_restrict_gfp_mask(void);
  480. extern void pm_restore_gfp_mask(void);
  481. #ifdef CONFIG_PM_SLEEP
  482. extern bool pm_suspended_storage(void);
  483. #else
  484. static inline bool pm_suspended_storage(void)
  485. {
  486. return false;
  487. }
  488. #endif /* CONFIG_PM_SLEEP */
  489. #ifdef CONFIG_CMA
  490. /* The below functions must be run on a range from a single zone. */
  491. extern int alloc_contig_range(unsigned long start, unsigned long end,
  492. unsigned migratetype);
  493. extern void free_contig_range(unsigned long pfn, unsigned nr_pages);
  494. /* CMA stuff */
  495. extern void init_cma_reserved_pageblock(struct page *page);
  496. #endif
  497. #endif /* __LINUX_GFP_H */