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;
12 /* Plain integer GFP bitmasks. Do not use this directly. */
13 #define ___GFP_DMA 0x01u
14 #define ___GFP_HIGHMEM 0x02u
15 #define ___GFP_DMA32 0x04u
16 #define ___GFP_MOVABLE 0x08u
17 #define ___GFP_RECLAIMABLE 0x10u
18 #define ___GFP_HIGH 0x20u
19 #define ___GFP_IO 0x40u
20 #define ___GFP_FS 0x80u
21 #define ___GFP_COLD 0x100u
22 #define ___GFP_NOWARN 0x200u
23 #define ___GFP_REPEAT 0x400u
24 #define ___GFP_NOFAIL 0x800u
25 #define ___GFP_NORETRY 0x1000u
26 #define ___GFP_MEMALLOC 0x2000u
27 #define ___GFP_COMP 0x4000u
28 #define ___GFP_ZERO 0x8000u
29 #define ___GFP_NOMEMALLOC 0x10000u
30 #define ___GFP_HARDWALL 0x20000u
31 #define ___GFP_THISNODE 0x40000u
32 #define ___GFP_ATOMIC 0x80000u
33 #define ___GFP_NOACCOUNT 0x100000u
34 #define ___GFP_NOTRACK 0x200000u
35 #define ___GFP_DIRECT_RECLAIM 0x400000u
36 #define ___GFP_OTHER_NODE 0x800000u
37 #define ___GFP_WRITE 0x1000000u
38 #define ___GFP_KSWAPD_RECLAIM 0x2000000u
39 /* If the above are modified, __GFP_BITS_SHIFT may need updating */
44 * Zone modifiers (see linux/mmzone.h - low three bits)
46 * Do not put any conditional on these. If necessary modify the definitions
47 * without the underscores and use them consistently. The definitions here may
48 * be used in bit comparisons.
50 #define __GFP_DMA ((__force gfp_t)___GFP_DMA)
51 #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
52 #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
53 #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */
54 #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
56 * Action modifiers - doesn't change the zoning
58 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
59 * _might_ fail. This depends upon the particular VM implementation.
61 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
62 * cannot handle allocation failures. New users should be evaluated carefully
63 * (and the flag should be used only when there is no reasonable failure policy)
64 * but it is definitely preferable to use the flag rather than opencode endless
65 * loop around allocator.
67 * __GFP_NORETRY: The VM implementation must not retry indefinitely and will
68 * return NULL when direct reclaim and memory compaction have failed to allow
69 * the allocation to succeed. The OOM killer is not called with the current
72 * __GFP_MOVABLE: Flag that this page will be movable by the page migration
73 * mechanism or reclaimed
75 #define __GFP_ATOMIC ((__force gfp_t)___GFP_ATOMIC) /* Caller cannot wait or reschedule */
76 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */
77 #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */
78 #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */
79 #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */
80 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */
81 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */
82 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */
83 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */
84 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */
85 #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */
86 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */
87 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves.
88 * This takes precedence over the
89 * __GFP_MEMALLOC flag if both are
92 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */
93 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */
94 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */
95 #define __GFP_NOACCOUNT ((__force gfp_t)___GFP_NOACCOUNT) /* Don't account to kmemcg */
96 #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */
98 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */
99 #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE) /* Allocator intends to dirty page */
102 * A caller that is willing to wait may enter direct reclaim and will
103 * wake kswapd to reclaim pages in the background until the high
104 * watermark is met. A caller may wish to clear __GFP_DIRECT_RECLAIM to
105 * avoid unnecessary delays when a fallback option is available but
106 * still allow kswapd to reclaim in the background. The kswapd flag
107 * can be cleared when the reclaiming of pages would cause unnecessary
110 #define __GFP_WAIT ((__force gfp_t)(___GFP_DIRECT_RECLAIM|___GFP_KSWAPD_RECLAIM))
111 #define __GFP_DIRECT_RECLAIM ((__force gfp_t)___GFP_DIRECT_RECLAIM) /* Caller can reclaim */
112 #define __GFP_KSWAPD_RECLAIM ((__force gfp_t)___GFP_KSWAPD_RECLAIM) /* kswapd can wake */
115 * This may seem redundant, but it's a way of annotating false positives vs.
116 * allocations that simply cannot be supported (e.g. page tables).
118 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
120 #define __GFP_BITS_SHIFT 26 /* Room for N __GFP_FOO bits */
121 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
124 * GFP_ATOMIC callers can not sleep, need the allocation to succeed.
125 * A lower watermark is applied to allow access to "atomic reserves"
127 #define GFP_ATOMIC (__GFP_HIGH|__GFP_ATOMIC|__GFP_KSWAPD_RECLAIM)
128 #define GFP_NOWAIT (__GFP_KSWAPD_RECLAIM)
129 #define GFP_NOIO (__GFP_WAIT)
130 #define GFP_NOFS (__GFP_WAIT | __GFP_IO)
131 #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS)
132 #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \
134 #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
135 #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM)
136 #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE)
137 #define GFP_TRANSHUGE ((GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
138 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN) & \
139 ~__GFP_KSWAPD_RECLAIM)
141 /* This mask makes up all the page movable related flags */
142 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
143 #define GFP_MOVABLE_SHIFT 3
145 /* Control page allocator reclaim behavior */
146 #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
147 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
148 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
150 /* Control slab gfp mask during early boot */
151 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
153 /* Control allocation constraints */
154 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
156 /* Do not use these with a slab allocator */
157 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
159 /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some
160 platforms, used as appropriate on others */
162 #define GFP_DMA __GFP_DMA
164 /* 4GB DMA on some platforms */
165 #define GFP_DMA32 __GFP_DMA32
167 /* Convert GFP flags to their corresponding migrate type */
168 static inline int gfpflags_to_migratetype(const gfp_t gfp_flags)
170 VM_WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
171 BUILD_BUG_ON((1UL << GFP_MOVABLE_SHIFT) != ___GFP_MOVABLE);
172 BUILD_BUG_ON((___GFP_MOVABLE >> GFP_MOVABLE_SHIFT) != MIGRATE_MOVABLE);
174 if (unlikely(page_group_by_mobility_disabled))
175 return MIGRATE_UNMOVABLE;
177 /* Group based on mobility */
178 return (gfp_flags & GFP_MOVABLE_MASK) >> GFP_MOVABLE_SHIFT;
181 static inline bool gfpflags_allow_blocking(const gfp_t gfp_flags)
183 return gfp_flags & __GFP_DIRECT_RECLAIM;
186 #ifdef CONFIG_HIGHMEM
187 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
189 #define OPT_ZONE_HIGHMEM ZONE_NORMAL
192 #ifdef CONFIG_ZONE_DMA
193 #define OPT_ZONE_DMA ZONE_DMA
195 #define OPT_ZONE_DMA ZONE_NORMAL
198 #ifdef CONFIG_ZONE_DMA32
199 #define OPT_ZONE_DMA32 ZONE_DMA32
201 #define OPT_ZONE_DMA32 ZONE_NORMAL
205 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
206 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
207 * and there are 16 of them to cover all possible combinations of
208 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
210 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
211 * But GFP_MOVABLE is not only a zone specifier but also an allocation
212 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
213 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
218 * 0x1 => DMA or NORMAL
219 * 0x2 => HIGHMEM or NORMAL
220 * 0x3 => BAD (DMA+HIGHMEM)
221 * 0x4 => DMA32 or DMA or NORMAL
222 * 0x5 => BAD (DMA+DMA32)
223 * 0x6 => BAD (HIGHMEM+DMA32)
224 * 0x7 => BAD (HIGHMEM+DMA32+DMA)
225 * 0x8 => NORMAL (MOVABLE+0)
226 * 0x9 => DMA or NORMAL (MOVABLE+DMA)
227 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
228 * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
229 * 0xc => DMA32 (MOVABLE+DMA32)
230 * 0xd => BAD (MOVABLE+DMA32+DMA)
231 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
232 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
234 * ZONES_SHIFT must be <= 2 on 32 bit platforms.
237 #if 16 * ZONES_SHIFT > BITS_PER_LONG
238 #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
241 #define GFP_ZONE_TABLE ( \
242 (ZONE_NORMAL << 0 * ZONES_SHIFT) \
243 | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \
244 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \
245 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \
246 | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \
247 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \
248 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \
249 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \
253 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
254 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
255 * entry starting with bit 0. Bit is set if the combination is not
258 #define GFP_ZONE_BAD ( \
259 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
260 | 1 << (___GFP_DMA | ___GFP_DMA32) \
261 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
262 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
263 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
264 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
265 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
266 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
269 static inline enum zone_type gfp_zone(gfp_t flags)
272 int bit = (__force int) (flags & GFP_ZONEMASK);
274 z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
275 ((1 << ZONES_SHIFT) - 1);
276 VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
281 * There is only one page-allocator function, and two main namespaces to
282 * it. The alloc_page*() variants return 'struct page *' and as such
283 * can allocate highmem pages, the *get*page*() variants return
284 * virtual kernel addresses to the allocated page(s).
287 static inline int gfp_zonelist(gfp_t flags)
289 if (IS_ENABLED(CONFIG_NUMA) && unlikely(flags & __GFP_THISNODE))
296 * We get the zone list from the current node and the gfp_mask.
297 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
298 * There are two zonelists per node, one for all zones with memory and
299 * one containing just zones from the node the zonelist belongs to.
301 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
302 * optimized to &contig_page_data at compile-time.
304 static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
306 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
309 #ifndef HAVE_ARCH_FREE_PAGE
310 static inline void arch_free_page(struct page *page, int order) { }
312 #ifndef HAVE_ARCH_ALLOC_PAGE
313 static inline void arch_alloc_page(struct page *page, int order) { }
317 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
318 struct zonelist *zonelist, nodemask_t *nodemask);
320 static inline struct page *
321 __alloc_pages(gfp_t gfp_mask, unsigned int order,
322 struct zonelist *zonelist)
324 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
328 * Allocate pages, preferring the node given as nid. The node must be valid and
329 * online. For more general interface, see alloc_pages_node().
331 static inline struct page *
332 __alloc_pages_node(int nid, gfp_t gfp_mask, unsigned int order)
334 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
335 VM_WARN_ON(!node_online(nid));
337 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
341 * Allocate pages, preferring the node given as nid. When nid == NUMA_NO_NODE,
342 * prefer the current CPU's closest node. Otherwise node must be valid and
345 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
348 if (nid == NUMA_NO_NODE)
351 return __alloc_pages_node(nid, gfp_mask, order);
355 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
357 static inline struct page *
358 alloc_pages(gfp_t gfp_mask, unsigned int order)
360 return alloc_pages_current(gfp_mask, order);
362 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
363 struct vm_area_struct *vma, unsigned long addr,
364 int node, bool hugepage);
365 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
366 alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true)
368 #define alloc_pages(gfp_mask, order) \
369 alloc_pages_node(numa_node_id(), gfp_mask, order)
370 #define alloc_pages_vma(gfp_mask, order, vma, addr, node, false)\
371 alloc_pages(gfp_mask, order)
372 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
373 alloc_pages(gfp_mask, order)
375 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
376 #define alloc_page_vma(gfp_mask, vma, addr) \
377 alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id(), false)
378 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \
379 alloc_pages_vma(gfp_mask, 0, vma, addr, node, false)
381 extern struct page *alloc_kmem_pages(gfp_t gfp_mask, unsigned int order);
382 extern struct page *alloc_kmem_pages_node(int nid, gfp_t gfp_mask,
385 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
386 extern unsigned long get_zeroed_page(gfp_t gfp_mask);
388 void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
389 void free_pages_exact(void *virt, size_t size);
390 void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
392 #define __get_free_page(gfp_mask) \
393 __get_free_pages((gfp_mask), 0)
395 #define __get_dma_pages(gfp_mask, order) \
396 __get_free_pages((gfp_mask) | GFP_DMA, (order))
398 extern void __free_pages(struct page *page, unsigned int order);
399 extern void free_pages(unsigned long addr, unsigned int order);
400 extern void free_hot_cold_page(struct page *page, bool cold);
401 extern void free_hot_cold_page_list(struct list_head *list, bool cold);
403 struct page_frag_cache;
404 extern void *__alloc_page_frag(struct page_frag_cache *nc,
405 unsigned int fragsz, gfp_t gfp_mask);
406 extern void __free_page_frag(void *addr);
408 extern void __free_kmem_pages(struct page *page, unsigned int order);
409 extern void free_kmem_pages(unsigned long addr, unsigned int order);
411 #define __free_page(page) __free_pages((page), 0)
412 #define free_page(addr) free_pages((addr), 0)
414 void page_alloc_init(void);
415 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
416 void drain_all_pages(struct zone *zone);
417 void drain_local_pages(struct zone *zone);
419 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
420 void page_alloc_init_late(void);
422 static inline void page_alloc_init_late(void)
428 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
429 * GFP flags are used before interrupts are enabled. Once interrupts are
430 * enabled, it is set to __GFP_BITS_MASK while the system is running. During
431 * hibernation, it is used by PM to avoid I/O during memory allocation while
432 * devices are suspended.
434 extern gfp_t gfp_allowed_mask;
436 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
437 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask);
439 extern void pm_restrict_gfp_mask(void);
440 extern void pm_restore_gfp_mask(void);
442 #ifdef CONFIG_PM_SLEEP
443 extern bool pm_suspended_storage(void);
445 static inline bool pm_suspended_storage(void)
449 #endif /* CONFIG_PM_SLEEP */
453 /* The below functions must be run on a range from a single zone. */
454 extern int alloc_contig_range(unsigned long start, unsigned long end,
455 unsigned migratetype);
456 extern void free_contig_range(unsigned long pfn, unsigned nr_pages);
459 extern void init_cma_reserved_pageblock(struct page *page);
463 #endif /* __LINUX_GFP_H */