4 #include <linux/mmdebug.h>
5 #include <linux/mmzone.h>
6 #include <linux/stddef.h>
7 #include <linux/linkage.h>
8 #include <linux/topology.h>
10 struct vm_area_struct;
13 * In case of changes, please don't forget to update
14 * include/trace/events/mmflags.h and tools/perf/builtin-kmem.c
17 /* Plain integer GFP bitmasks. Do not use this directly. */
18 #define ___GFP_DMA 0x01u
19 #define ___GFP_HIGHMEM 0x02u
20 #define ___GFP_DMA32 0x04u
21 #define ___GFP_MOVABLE 0x08u
22 #define ___GFP_RECLAIMABLE 0x10u
23 #define ___GFP_HIGH 0x20u
24 #define ___GFP_IO 0x40u
25 #define ___GFP_FS 0x80u
26 #define ___GFP_COLD 0x100u
27 #define ___GFP_NOWARN 0x200u
28 #define ___GFP_REPEAT 0x400u
29 #define ___GFP_NOFAIL 0x800u
30 #define ___GFP_NORETRY 0x1000u
31 #define ___GFP_MEMALLOC 0x2000u
32 #define ___GFP_COMP 0x4000u
33 #define ___GFP_ZERO 0x8000u
34 #define ___GFP_NOMEMALLOC 0x10000u
35 #define ___GFP_HARDWALL 0x20000u
36 #define ___GFP_THISNODE 0x40000u
37 #define ___GFP_ATOMIC 0x80000u
38 #define ___GFP_ACCOUNT 0x100000u
39 #define ___GFP_NOTRACK 0x200000u
40 #define ___GFP_DIRECT_RECLAIM 0x400000u
41 #define ___GFP_OTHER_NODE 0x800000u
42 #define ___GFP_WRITE 0x1000000u
43 #define ___GFP_KSWAPD_RECLAIM 0x2000000u
44 /* If the above are modified, __GFP_BITS_SHIFT may need updating */
47 * Physical address zone modifiers (see linux/mmzone.h - low four bits)
49 * Do not put any conditional on these. If necessary modify the definitions
50 * without the underscores and use them consistently. The definitions here may
51 * be used in bit comparisons.
53 #define __GFP_DMA ((__force gfp_t)___GFP_DMA)
54 #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
55 #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
56 #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* ZONE_MOVABLE allowed */
57 #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
60 * Page mobility and placement hints
62 * These flags provide hints about how mobile the page is. Pages with similar
63 * mobility are placed within the same pageblocks to minimise problems due
64 * to external fragmentation.
66 * __GFP_MOVABLE (also a zone modifier) indicates that the page can be
67 * moved by page migration during memory compaction or can be reclaimed.
69 * __GFP_RECLAIMABLE is used for slab allocations that specify
70 * SLAB_RECLAIM_ACCOUNT and whose pages can be freed via shrinkers.
72 * __GFP_WRITE indicates the caller intends to dirty the page. Where possible,
73 * these pages will be spread between local zones to avoid all the dirty
74 * pages being in one zone (fair zone allocation policy).
76 * __GFP_HARDWALL enforces the cpuset memory allocation policy.
78 * __GFP_THISNODE forces the allocation to be satisified from the requested
79 * node with no fallbacks or placement policy enforcements.
81 * __GFP_ACCOUNT causes the allocation to be accounted to kmemcg.
83 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE)
84 #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE)
85 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL)
86 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)
87 #define __GFP_ACCOUNT ((__force gfp_t)___GFP_ACCOUNT)
90 * Watermark modifiers -- controls access to emergency reserves
92 * __GFP_HIGH indicates that the caller is high-priority and that granting
93 * the request is necessary before the system can make forward progress.
94 * For example, creating an IO context to clean pages.
96 * __GFP_ATOMIC indicates that the caller cannot reclaim or sleep and is
97 * high priority. Users are typically interrupt handlers. This may be
98 * used in conjunction with __GFP_HIGH
100 * __GFP_MEMALLOC allows access to all memory. This should only be used when
101 * the caller guarantees the allocation will allow more memory to be freed
102 * very shortly e.g. process exiting or swapping. Users either should
103 * be the MM or co-ordinating closely with the VM (e.g. swap over NFS).
105 * __GFP_NOMEMALLOC is used to explicitly forbid access to emergency reserves.
106 * This takes precedence over the __GFP_MEMALLOC flag if both are set.
108 #define __GFP_ATOMIC ((__force gfp_t)___GFP_ATOMIC)
109 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH)
110 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)
111 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC)
116 * __GFP_IO can start physical IO.
118 * __GFP_FS can call down to the low-level FS. Clearing the flag avoids the
119 * allocator recursing into the filesystem which might already be holding
122 * __GFP_DIRECT_RECLAIM indicates that the caller may enter direct reclaim.
123 * This flag can be cleared to avoid unnecessary delays when a fallback
124 * option is available.
126 * __GFP_KSWAPD_RECLAIM indicates that the caller wants to wake kswapd when
127 * the low watermark is reached and have it reclaim pages until the high
128 * watermark is reached. A caller may wish to clear this flag when fallback
129 * options are available and the reclaim is likely to disrupt the system. The
130 * canonical example is THP allocation where a fallback is cheap but
131 * reclaim/compaction may cause indirect stalls.
133 * __GFP_RECLAIM is shorthand to allow/forbid both direct and kswapd reclaim.
135 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
136 * _might_ fail. This depends upon the particular VM implementation.
138 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
139 * cannot handle allocation failures. New users should be evaluated carefully
140 * (and the flag should be used only when there is no reasonable failure
141 * policy) but it is definitely preferable to use the flag rather than
142 * opencode endless loop around allocator.
144 * __GFP_NORETRY: The VM implementation must not retry indefinitely and will
145 * return NULL when direct reclaim and memory compaction have failed to allow
146 * the allocation to succeed. The OOM killer is not called with the current
149 #define __GFP_IO ((__force gfp_t)___GFP_IO)
150 #define __GFP_FS ((__force gfp_t)___GFP_FS)
151 #define __GFP_DIRECT_RECLAIM ((__force gfp_t)___GFP_DIRECT_RECLAIM) /* Caller can reclaim */
152 #define __GFP_KSWAPD_RECLAIM ((__force gfp_t)___GFP_KSWAPD_RECLAIM) /* kswapd can wake */
153 #define __GFP_RECLAIM ((__force gfp_t)(___GFP_DIRECT_RECLAIM|___GFP_KSWAPD_RECLAIM))
154 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT)
155 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL)
156 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY)
161 * __GFP_COLD indicates that the caller does not expect to be used in the near
162 * future. Where possible, a cache-cold page will be returned.
164 * __GFP_NOWARN suppresses allocation failure reports.
166 * __GFP_COMP address compound page metadata.
168 * __GFP_ZERO returns a zeroed page on success.
170 * __GFP_NOTRACK avoids tracking with kmemcheck.
172 * __GFP_NOTRACK_FALSE_POSITIVE is an alias of __GFP_NOTRACK. It's a means of
173 * distinguishing in the source between false positives and allocations that
174 * cannot be supported (e.g. page tables).
176 * __GFP_OTHER_NODE is for allocations that are on a remote node but that
177 * should not be accounted for as a remote allocation in vmstat. A
178 * typical user would be khugepaged collapsing a huge page on a remote
181 #define __GFP_COLD ((__force gfp_t)___GFP_COLD)
182 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN)
183 #define __GFP_COMP ((__force gfp_t)___GFP_COMP)
184 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO)
185 #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK)
186 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
187 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE)
189 /* Room for N __GFP_FOO bits */
190 #define __GFP_BITS_SHIFT 26
191 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
194 * Useful GFP flag combinations that are commonly used. It is recommended
195 * that subsystems start with one of these combinations and then set/clear
196 * __GFP_FOO flags as necessary.
198 * GFP_ATOMIC users can not sleep and need the allocation to succeed. A lower
199 * watermark is applied to allow access to "atomic reserves"
201 * GFP_KERNEL is typical for kernel-internal allocations. The caller requires
202 * ZONE_NORMAL or a lower zone for direct access but can direct reclaim.
204 * GFP_KERNEL_ACCOUNT is the same as GFP_KERNEL, except the allocation is
205 * accounted to kmemcg.
207 * GFP_NOWAIT is for kernel allocations that should not stall for direct
208 * reclaim, start physical IO or use any filesystem callback.
210 * GFP_NOIO will use direct reclaim to discard clean pages or slab pages
211 * that do not require the starting of any physical IO.
213 * GFP_NOFS will use direct reclaim but will not use any filesystem interfaces.
215 * GFP_USER is for userspace allocations that also need to be directly
216 * accessibly by the kernel or hardware. It is typically used by hardware
217 * for buffers that are mapped to userspace (e.g. graphics) that hardware
218 * still must DMA to. cpuset limits are enforced for these allocations.
220 * GFP_DMA exists for historical reasons and should be avoided where possible.
221 * The flags indicates that the caller requires that the lowest zone be
222 * used (ZONE_DMA or 16M on x86-64). Ideally, this would be removed but
223 * it would require careful auditing as some users really require it and
224 * others use the flag to avoid lowmem reserves in ZONE_DMA and treat the
225 * lowest zone as a type of emergency reserve.
227 * GFP_DMA32 is similar to GFP_DMA except that the caller requires a 32-bit
230 * GFP_HIGHUSER is for userspace allocations that may be mapped to userspace,
231 * do not need to be directly accessible by the kernel but that cannot
232 * move once in use. An example may be a hardware allocation that maps
233 * data directly into userspace but has no addressing limitations.
235 * GFP_HIGHUSER_MOVABLE is for userspace allocations that the kernel does not
236 * need direct access to but can use kmap() when access is required. They
237 * are expected to be movable via page reclaim or page migration. Typically,
238 * pages on the LRU would also be allocated with GFP_HIGHUSER_MOVABLE.
240 * GFP_TRANSHUGE and GFP_TRANSHUGE_LIGHT are used for THP allocations. They are
241 * compound allocations that will generally fail quickly if memory is not
242 * available and will not wake kswapd/kcompactd on failure. The _LIGHT
243 * version does not attempt reclaim/compaction at all and is by default used
244 * in page fault path, while the non-light is used by khugepaged.
246 #define GFP_ATOMIC (__GFP_HIGH|__GFP_ATOMIC|__GFP_KSWAPD_RECLAIM)
247 #define GFP_KERNEL (__GFP_RECLAIM | __GFP_IO | __GFP_FS)
248 #define GFP_KERNEL_ACCOUNT (GFP_KERNEL | __GFP_ACCOUNT)
249 #define GFP_NOWAIT (__GFP_KSWAPD_RECLAIM)
250 #define GFP_NOIO (__GFP_RECLAIM)
251 #define GFP_NOFS (__GFP_RECLAIM | __GFP_IO)
252 #define GFP_TEMPORARY (__GFP_RECLAIM | __GFP_IO | __GFP_FS | \
254 #define GFP_USER (__GFP_RECLAIM | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
255 #define GFP_DMA __GFP_DMA
256 #define GFP_DMA32 __GFP_DMA32
257 #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM)
258 #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE)
259 #define GFP_TRANSHUGE_LIGHT ((GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
260 __GFP_NOMEMALLOC | __GFP_NOWARN) & ~__GFP_RECLAIM)
261 #define GFP_TRANSHUGE (GFP_TRANSHUGE_LIGHT | __GFP_DIRECT_RECLAIM)
263 /* Convert GFP flags to their corresponding migrate type */
264 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
265 #define GFP_MOVABLE_SHIFT 3
267 static inline int gfpflags_to_migratetype(const gfp_t gfp_flags)
269 VM_WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
270 BUILD_BUG_ON((1UL << GFP_MOVABLE_SHIFT) != ___GFP_MOVABLE);
271 BUILD_BUG_ON((___GFP_MOVABLE >> GFP_MOVABLE_SHIFT) != MIGRATE_MOVABLE);
273 if (unlikely(page_group_by_mobility_disabled))
274 return MIGRATE_UNMOVABLE;
276 /* Group based on mobility */
277 return (gfp_flags & GFP_MOVABLE_MASK) >> GFP_MOVABLE_SHIFT;
279 #undef GFP_MOVABLE_MASK
280 #undef GFP_MOVABLE_SHIFT
282 static inline bool gfpflags_allow_blocking(const gfp_t gfp_flags)
284 return !!(gfp_flags & __GFP_DIRECT_RECLAIM);
288 * gfpflags_normal_context - is gfp_flags a normal sleepable context?
289 * @gfp_flags: gfp_flags to test
291 * Test whether @gfp_flags indicates that the allocation is from the
292 * %current context and allowed to sleep.
294 * An allocation being allowed to block doesn't mean it owns the %current
295 * context. When direct reclaim path tries to allocate memory, the
296 * allocation context is nested inside whatever %current was doing at the
297 * time of the original allocation. The nested allocation may be allowed
298 * to block but modifying anything %current owns can corrupt the outer
299 * context's expectations.
301 * %true result from this function indicates that the allocation context
302 * can sleep and use anything that's associated with %current.
304 static inline bool gfpflags_normal_context(const gfp_t gfp_flags)
306 return (gfp_flags & (__GFP_DIRECT_RECLAIM | __GFP_MEMALLOC)) ==
307 __GFP_DIRECT_RECLAIM;
310 #ifdef CONFIG_HIGHMEM
311 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
313 #define OPT_ZONE_HIGHMEM ZONE_NORMAL
316 #ifdef CONFIG_ZONE_DMA
317 #define OPT_ZONE_DMA ZONE_DMA
319 #define OPT_ZONE_DMA ZONE_NORMAL
322 #ifdef CONFIG_ZONE_DMA32
323 #define OPT_ZONE_DMA32 ZONE_DMA32
325 #define OPT_ZONE_DMA32 ZONE_NORMAL
329 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
330 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
331 * and there are 16 of them to cover all possible combinations of
332 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
334 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
335 * But GFP_MOVABLE is not only a zone specifier but also an allocation
336 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
337 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
342 * 0x1 => DMA or NORMAL
343 * 0x2 => HIGHMEM or NORMAL
344 * 0x3 => BAD (DMA+HIGHMEM)
345 * 0x4 => DMA32 or DMA or NORMAL
346 * 0x5 => BAD (DMA+DMA32)
347 * 0x6 => BAD (HIGHMEM+DMA32)
348 * 0x7 => BAD (HIGHMEM+DMA32+DMA)
349 * 0x8 => NORMAL (MOVABLE+0)
350 * 0x9 => DMA or NORMAL (MOVABLE+DMA)
351 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
352 * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
353 * 0xc => DMA32 (MOVABLE+DMA32)
354 * 0xd => BAD (MOVABLE+DMA32+DMA)
355 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
356 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
358 * GFP_ZONES_SHIFT must be <= 2 on 32 bit platforms.
361 #if defined(CONFIG_ZONE_DEVICE) && (MAX_NR_ZONES-1) <= 4
362 /* ZONE_DEVICE is not a valid GFP zone specifier */
363 #define GFP_ZONES_SHIFT 2
365 #define GFP_ZONES_SHIFT ZONES_SHIFT
368 #if 16 * GFP_ZONES_SHIFT > BITS_PER_LONG
369 #error GFP_ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
372 #define GFP_ZONE_TABLE ( \
373 (ZONE_NORMAL << 0 * GFP_ZONES_SHIFT) \
374 | (OPT_ZONE_DMA << ___GFP_DMA * GFP_ZONES_SHIFT) \
375 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * GFP_ZONES_SHIFT) \
376 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * GFP_ZONES_SHIFT) \
377 | (ZONE_NORMAL << ___GFP_MOVABLE * GFP_ZONES_SHIFT) \
378 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * GFP_ZONES_SHIFT) \
379 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * GFP_ZONES_SHIFT)\
380 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * GFP_ZONES_SHIFT)\
384 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
385 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
386 * entry starting with bit 0. Bit is set if the combination is not
389 #define GFP_ZONE_BAD ( \
390 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
391 | 1 << (___GFP_DMA | ___GFP_DMA32) \
392 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
393 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
394 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
395 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
396 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
397 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
400 static inline enum zone_type gfp_zone(gfp_t flags)
403 int bit = (__force int) (flags & GFP_ZONEMASK);
405 z = (GFP_ZONE_TABLE >> (bit * GFP_ZONES_SHIFT)) &
406 ((1 << GFP_ZONES_SHIFT) - 1);
407 VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
412 * There is only one page-allocator function, and two main namespaces to
413 * it. The alloc_page*() variants return 'struct page *' and as such
414 * can allocate highmem pages, the *get*page*() variants return
415 * virtual kernel addresses to the allocated page(s).
418 static inline int gfp_zonelist(gfp_t flags)
421 if (unlikely(flags & __GFP_THISNODE))
422 return ZONELIST_NOFALLBACK;
424 return ZONELIST_FALLBACK;
428 * We get the zone list from the current node and the gfp_mask.
429 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
430 * There are two zonelists per node, one for all zones with memory and
431 * one containing just zones from the node the zonelist belongs to.
433 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
434 * optimized to &contig_page_data at compile-time.
436 static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
438 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
441 #ifndef HAVE_ARCH_FREE_PAGE
442 static inline void arch_free_page(struct page *page, int order) { }
444 #ifndef HAVE_ARCH_ALLOC_PAGE
445 static inline void arch_alloc_page(struct page *page, int order) { }
449 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
450 struct zonelist *zonelist, nodemask_t *nodemask);
452 static inline struct page *
453 __alloc_pages(gfp_t gfp_mask, unsigned int order,
454 struct zonelist *zonelist)
456 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
460 * Allocate pages, preferring the node given as nid. The node must be valid and
461 * online. For more general interface, see alloc_pages_node().
463 static inline struct page *
464 __alloc_pages_node(int nid, gfp_t gfp_mask, unsigned int order)
466 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
467 VM_WARN_ON(!node_online(nid));
469 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
473 * Allocate pages, preferring the node given as nid. When nid == NUMA_NO_NODE,
474 * prefer the current CPU's closest node. Otherwise node must be valid and
477 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
480 if (nid == NUMA_NO_NODE)
483 return __alloc_pages_node(nid, gfp_mask, order);
487 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
489 static inline struct page *
490 alloc_pages(gfp_t gfp_mask, unsigned int order)
492 return alloc_pages_current(gfp_mask, order);
494 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
495 struct vm_area_struct *vma, unsigned long addr,
496 int node, bool hugepage);
497 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
498 alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true)
500 #define alloc_pages(gfp_mask, order) \
501 alloc_pages_node(numa_node_id(), gfp_mask, order)
502 #define alloc_pages_vma(gfp_mask, order, vma, addr, node, false)\
503 alloc_pages(gfp_mask, order)
504 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
505 alloc_pages(gfp_mask, order)
507 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
508 #define alloc_page_vma(gfp_mask, vma, addr) \
509 alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id(), false)
510 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \
511 alloc_pages_vma(gfp_mask, 0, vma, addr, node, false)
513 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
514 extern unsigned long get_zeroed_page(gfp_t gfp_mask);
516 void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
517 void free_pages_exact(void *virt, size_t size);
518 void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
520 #define __get_free_page(gfp_mask) \
521 __get_free_pages((gfp_mask), 0)
523 #define __get_dma_pages(gfp_mask, order) \
524 __get_free_pages((gfp_mask) | GFP_DMA, (order))
526 extern void __free_pages(struct page *page, unsigned int order);
527 extern void free_pages(unsigned long addr, unsigned int order);
528 extern void free_hot_cold_page(struct page *page, bool cold);
529 extern void free_hot_cold_page_list(struct list_head *list, bool cold);
531 struct page_frag_cache;
532 extern void *__alloc_page_frag(struct page_frag_cache *nc,
533 unsigned int fragsz, gfp_t gfp_mask);
534 extern void __free_page_frag(void *addr);
536 #define __free_page(page) __free_pages((page), 0)
537 #define free_page(addr) free_pages((addr), 0)
539 void page_alloc_init(void);
540 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
541 void drain_all_pages(struct zone *zone);
542 void drain_local_pages(struct zone *zone);
544 void page_alloc_init_late(void);
547 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
548 * GFP flags are used before interrupts are enabled. Once interrupts are
549 * enabled, it is set to __GFP_BITS_MASK while the system is running. During
550 * hibernation, it is used by PM to avoid I/O during memory allocation while
551 * devices are suspended.
553 extern gfp_t gfp_allowed_mask;
555 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
556 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask);
558 extern void pm_restrict_gfp_mask(void);
559 extern void pm_restore_gfp_mask(void);
561 #ifdef CONFIG_PM_SLEEP
562 extern bool pm_suspended_storage(void);
564 static inline bool pm_suspended_storage(void)
568 #endif /* CONFIG_PM_SLEEP */
570 #if (defined(CONFIG_MEMORY_ISOLATION) && defined(CONFIG_COMPACTION)) || defined(CONFIG_CMA)
571 /* The below functions must be run on a range from a single zone. */
572 extern int alloc_contig_range(unsigned long start, unsigned long end,
573 unsigned migratetype);
574 extern void free_contig_range(unsigned long pfn, unsigned nr_pages);
579 extern void init_cma_reserved_pageblock(struct page *page);
582 #endif /* __LINUX_GFP_H */