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_WAIT 0x80000u
33 #define ___GFP_NOACCOUNT 0x100000u
34 #define ___GFP_NOTRACK 0x200000u
35 #define ___GFP_NO_KSWAPD 0x400000u
36 #define ___GFP_OTHER_NODE 0x800000u
37 #define ___GFP_WRITE 0x1000000u
38 /* If the above are modified, __GFP_BITS_SHIFT may need updating */
43 * Zone modifiers (see linux/mmzone.h - low three bits)
45 * Do not put any conditional on these. If necessary modify the definitions
46 * without the underscores and use them consistently. The definitions here may
47 * be used in bit comparisons.
49 #define __GFP_DMA ((__force gfp_t)___GFP_DMA)
50 #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM)
51 #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32)
52 #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */
53 #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
55 * Action modifiers - doesn't change the zoning
57 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
58 * _might_ fail. This depends upon the particular VM implementation.
60 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
61 * cannot handle allocation failures. New users should be evaluated carefully
62 * (and the flag should be used only when there is no reasonable failure policy)
63 * but it is definitely preferable to use the flag rather than opencode endless
64 * loop around allocator.
66 * __GFP_NORETRY: The VM implementation must not retry indefinitely and will
67 * return NULL when direct reclaim and memory compaction have failed to allow
68 * the allocation to succeed. The OOM killer is not called with the current
71 * __GFP_MOVABLE: Flag that this page will be movable by the page migration
72 * mechanism or reclaimed
74 #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */
75 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */
76 #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */
77 #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */
78 #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */
79 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */
80 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */
81 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */
82 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */
83 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */
84 #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */
85 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */
86 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves.
87 * This takes precedence over the
88 * __GFP_MEMALLOC flag if both are
91 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */
92 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */
93 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */
94 #define __GFP_NOACCOUNT ((__force gfp_t)___GFP_NOACCOUNT) /* Don't account to kmemcg */
95 #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */
97 #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD)
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 * This may seem redundant, but it's a way of annotating false positives vs.
103 * allocations that simply cannot be supported (e.g. page tables).
105 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
107 #define __GFP_BITS_SHIFT 25 /* Room for N __GFP_FOO bits */
108 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
110 /* This equals 0, but use constants in case they ever change */
111 #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH)
112 /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
113 #define GFP_ATOMIC (__GFP_HIGH)
114 #define GFP_NOIO (__GFP_WAIT)
115 #define GFP_NOFS (__GFP_WAIT | __GFP_IO)
116 #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS)
117 #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \
119 #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
120 #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM)
121 #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE)
122 #define GFP_IOFS (__GFP_IO | __GFP_FS)
123 #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
124 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \
127 /* This mask makes up all the page movable related flags */
128 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
129 #define GFP_MOVABLE_SHIFT 3
131 /* Control page allocator reclaim behavior */
132 #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
133 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
134 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
136 /* Control slab gfp mask during early boot */
137 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
139 /* Control allocation constraints */
140 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
142 /* Do not use these with a slab allocator */
143 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
145 /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some
146 platforms, used as appropriate on others */
148 #define GFP_DMA __GFP_DMA
150 /* 4GB DMA on some platforms */
151 #define GFP_DMA32 __GFP_DMA32
153 /* Convert GFP flags to their corresponding migrate type */
154 static inline int gfpflags_to_migratetype(const gfp_t gfp_flags
)
156 VM_WARN_ON((gfp_flags
& GFP_MOVABLE_MASK
) == GFP_MOVABLE_MASK
);
157 BUILD_BUG_ON((1UL << GFP_MOVABLE_SHIFT
) != ___GFP_MOVABLE
);
158 BUILD_BUG_ON((___GFP_MOVABLE
>> GFP_MOVABLE_SHIFT
) != MIGRATE_MOVABLE
);
160 if (unlikely(page_group_by_mobility_disabled
))
161 return MIGRATE_UNMOVABLE
;
163 /* Group based on mobility */
164 return (gfp_flags
& GFP_MOVABLE_MASK
) >> GFP_MOVABLE_SHIFT
;
167 #ifdef CONFIG_HIGHMEM
168 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
170 #define OPT_ZONE_HIGHMEM ZONE_NORMAL
173 #ifdef CONFIG_ZONE_DMA
174 #define OPT_ZONE_DMA ZONE_DMA
176 #define OPT_ZONE_DMA ZONE_NORMAL
179 #ifdef CONFIG_ZONE_DMA32
180 #define OPT_ZONE_DMA32 ZONE_DMA32
182 #define OPT_ZONE_DMA32 ZONE_NORMAL
186 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
187 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
188 * and there are 16 of them to cover all possible combinations of
189 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
191 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
192 * But GFP_MOVABLE is not only a zone specifier but also an allocation
193 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
194 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
199 * 0x1 => DMA or NORMAL
200 * 0x2 => HIGHMEM or NORMAL
201 * 0x3 => BAD (DMA+HIGHMEM)
202 * 0x4 => DMA32 or DMA or NORMAL
203 * 0x5 => BAD (DMA+DMA32)
204 * 0x6 => BAD (HIGHMEM+DMA32)
205 * 0x7 => BAD (HIGHMEM+DMA32+DMA)
206 * 0x8 => NORMAL (MOVABLE+0)
207 * 0x9 => DMA or NORMAL (MOVABLE+DMA)
208 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
209 * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
210 * 0xc => DMA32 (MOVABLE+DMA32)
211 * 0xd => BAD (MOVABLE+DMA32+DMA)
212 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
213 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
215 * ZONES_SHIFT must be <= 2 on 32 bit platforms.
218 #if 16 * ZONES_SHIFT > BITS_PER_LONG
219 #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
222 #define GFP_ZONE_TABLE ( \
223 (ZONE_NORMAL << 0 * ZONES_SHIFT) \
224 | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \
225 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \
226 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \
227 | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \
228 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \
229 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \
230 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \
234 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
235 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
236 * entry starting with bit 0. Bit is set if the combination is not
239 #define GFP_ZONE_BAD ( \
240 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
241 | 1 << (___GFP_DMA | ___GFP_DMA32) \
242 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
243 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
244 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
245 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
246 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
247 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
250 static inline enum zone_type
gfp_zone(gfp_t flags
)
253 int bit
= (__force
int) (flags
& GFP_ZONEMASK
);
255 z
= (GFP_ZONE_TABLE
>> (bit
* ZONES_SHIFT
)) &
256 ((1 << ZONES_SHIFT
) - 1);
257 VM_BUG_ON((GFP_ZONE_BAD
>> bit
) & 1);
262 * There is only one page-allocator function, and two main namespaces to
263 * it. The alloc_page*() variants return 'struct page *' and as such
264 * can allocate highmem pages, the *get*page*() variants return
265 * virtual kernel addresses to the allocated page(s).
268 static inline int gfp_zonelist(gfp_t flags
)
270 if (IS_ENABLED(CONFIG_NUMA
) && unlikely(flags
& __GFP_THISNODE
))
277 * We get the zone list from the current node and the gfp_mask.
278 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
279 * There are two zonelists per node, one for all zones with memory and
280 * one containing just zones from the node the zonelist belongs to.
282 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
283 * optimized to &contig_page_data at compile-time.
285 static inline struct zonelist
*node_zonelist(int nid
, gfp_t flags
)
287 return NODE_DATA(nid
)->node_zonelists
+ gfp_zonelist(flags
);
290 #ifndef HAVE_ARCH_FREE_PAGE
291 static inline void arch_free_page(struct page
*page
, int order
) { }
293 #ifndef HAVE_ARCH_ALLOC_PAGE
294 static inline void arch_alloc_page(struct page
*page
, int order
) { }
298 __alloc_pages_nodemask(gfp_t gfp_mask
, unsigned int order
,
299 struct zonelist
*zonelist
, nodemask_t
*nodemask
);
301 static inline struct page
*
302 __alloc_pages(gfp_t gfp_mask
, unsigned int order
,
303 struct zonelist
*zonelist
)
305 return __alloc_pages_nodemask(gfp_mask
, order
, zonelist
, NULL
);
309 * Allocate pages, preferring the node given as nid. The node must be valid and
310 * online. For more general interface, see alloc_pages_node().
312 static inline struct page
*
313 __alloc_pages_node(int nid
, gfp_t gfp_mask
, unsigned int order
)
315 VM_BUG_ON(nid
< 0 || nid
>= MAX_NUMNODES
);
316 VM_WARN_ON(!node_online(nid
));
318 return __alloc_pages(gfp_mask
, order
, node_zonelist(nid
, gfp_mask
));
322 * Allocate pages, preferring the node given as nid. When nid == NUMA_NO_NODE,
323 * prefer the current CPU's closest node. Otherwise node must be valid and
326 static inline struct page
*alloc_pages_node(int nid
, gfp_t gfp_mask
,
329 if (nid
== NUMA_NO_NODE
)
332 return __alloc_pages_node(nid
, gfp_mask
, order
);
336 extern struct page
*alloc_pages_current(gfp_t gfp_mask
, unsigned order
);
338 static inline struct page
*
339 alloc_pages(gfp_t gfp_mask
, unsigned int order
)
341 return alloc_pages_current(gfp_mask
, order
);
343 extern struct page
*alloc_pages_vma(gfp_t gfp_mask
, int order
,
344 struct vm_area_struct
*vma
, unsigned long addr
,
345 int node
, bool hugepage
);
346 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
347 alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true)
349 #define alloc_pages(gfp_mask, order) \
350 alloc_pages_node(numa_node_id(), gfp_mask, order)
351 #define alloc_pages_vma(gfp_mask, order, vma, addr, node, false)\
352 alloc_pages(gfp_mask, order)
353 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \
354 alloc_pages(gfp_mask, order)
356 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
357 #define alloc_page_vma(gfp_mask, vma, addr) \
358 alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id(), false)
359 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \
360 alloc_pages_vma(gfp_mask, 0, vma, addr, node, false)
362 extern struct page
*alloc_kmem_pages(gfp_t gfp_mask
, unsigned int order
);
363 extern struct page
*alloc_kmem_pages_node(int nid
, gfp_t gfp_mask
,
366 extern unsigned long __get_free_pages(gfp_t gfp_mask
, unsigned int order
);
367 extern unsigned long get_zeroed_page(gfp_t gfp_mask
);
369 void *alloc_pages_exact(size_t size
, gfp_t gfp_mask
);
370 void free_pages_exact(void *virt
, size_t size
);
371 void * __meminit
alloc_pages_exact_nid(int nid
, size_t size
, gfp_t gfp_mask
);
373 #define __get_free_page(gfp_mask) \
374 __get_free_pages((gfp_mask), 0)
376 #define __get_dma_pages(gfp_mask, order) \
377 __get_free_pages((gfp_mask) | GFP_DMA, (order))
379 extern void __free_pages(struct page
*page
, unsigned int order
);
380 extern void free_pages(unsigned long addr
, unsigned int order
);
381 extern void free_hot_cold_page(struct page
*page
, bool cold
);
382 extern void free_hot_cold_page_list(struct list_head
*list
, bool cold
);
384 struct page_frag_cache
;
385 extern void *__alloc_page_frag(struct page_frag_cache
*nc
,
386 unsigned int fragsz
, gfp_t gfp_mask
);
387 extern void __free_page_frag(void *addr
);
389 extern void __free_kmem_pages(struct page
*page
, unsigned int order
);
390 extern void free_kmem_pages(unsigned long addr
, unsigned int order
);
392 #define __free_page(page) __free_pages((page), 0)
393 #define free_page(addr) free_pages((addr), 0)
395 void page_alloc_init(void);
396 void drain_zone_pages(struct zone
*zone
, struct per_cpu_pages
*pcp
);
397 void drain_all_pages(struct zone
*zone
);
398 void drain_local_pages(struct zone
*zone
);
400 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
401 void page_alloc_init_late(void);
403 static inline void page_alloc_init_late(void)
409 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
410 * GFP flags are used before interrupts are enabled. Once interrupts are
411 * enabled, it is set to __GFP_BITS_MASK while the system is running. During
412 * hibernation, it is used by PM to avoid I/O during memory allocation while
413 * devices are suspended.
415 extern gfp_t gfp_allowed_mask
;
417 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
418 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask
);
420 extern void pm_restrict_gfp_mask(void);
421 extern void pm_restore_gfp_mask(void);
423 #ifdef CONFIG_PM_SLEEP
424 extern bool pm_suspended_storage(void);
426 static inline bool pm_suspended_storage(void)
430 #endif /* CONFIG_PM_SLEEP */
434 /* The below functions must be run on a range from a single zone. */
435 extern int alloc_contig_range(unsigned long start
, unsigned long end
,
436 unsigned migratetype
);
437 extern void free_contig_range(unsigned long pfn
, unsigned nr_pages
);
440 extern void init_cma_reserved_pageblock(struct page
*page
);
444 #endif /* __LINUX_GFP_H */