1 #ifndef __LINUX_GFP_H
2 #define __LINUX_GFP_H
3
4 #include <linux/mmzone.h>
5 #include <linux/stddef.h>
6 #include <linux/linkage.h>
7 #include <linux/topology.h>
8 #include <linux/mmdebug.h>
9
10 struct vm_area_struct;
11
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_WAIT 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_COMP 0x4000u
27 #define ___GFP_ZERO 0x8000u
28 #define ___GFP_NOMEMALLOC 0x10000u
29 #define ___GFP_HARDWALL 0x20000u
30 #define ___GFP_THISNODE 0x40000u
31 #define ___GFP_RECLAIMABLE 0x80000u
32 #ifdef CONFIG_KMEMCHECK
33 #define ___GFP_NOTRACK 0x200000u
34 #else
35 #define ___GFP_NOTRACK 0
36 #endif
37 #define ___GFP_NO_KSWAPD 0x400000u
38 #define ___GFP_OTHER_NODE 0x800000u
39
40 /*
41 * GFP bitmasks..
42 *
43 * Zone modifiers (see linux/mmzone.h - low three bits)
44 *
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.
48 */
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)
54 /*
55 * Action modifiers - doesn't change the zoning
56 *
57 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
58 * _might_ fail. This depends upon the particular VM implementation.
59 *
60 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
61 * cannot handle allocation failures. This modifier is deprecated and no new
62 * users should be added.
63 *
64 * __GFP_NORETRY: The VM implementation must not retry indefinitely.
65 *
66 * __GFP_MOVABLE: Flag that this page will be movable by the page migration
67 * mechanism or reclaimed
68 */
69 #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */
70 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */
71 #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */
72 #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */
73 #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */
74 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */
75 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */
76 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */
77 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */
78 #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */
79 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */
80 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves */
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
86 #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD)
87 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */
88
89 /*
90 * This may seem redundant, but it's a way of annotating false positives vs.
91 * allocations that simply cannot be supported (e.g. page tables).
92 */
93 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
94
95 #define __GFP_BITS_SHIFT 23 /* Room for 23 __GFP_FOO bits */
96 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
97
98 /* This equals 0, but use constants in case they ever change */
99 #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH)
100 /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
101 #define GFP_ATOMIC (__GFP_HIGH)
102 #define GFP_NOIO (__GFP_WAIT)
103 #define GFP_NOFS (__GFP_WAIT | __GFP_IO)
104 #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS)
105 #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \
106 __GFP_RECLAIMABLE)
107 #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
108 #define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \
109 __GFP_HIGHMEM)
110 #define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \
111 __GFP_HARDWALL | __GFP_HIGHMEM | \
112 __GFP_MOVABLE)
113 #define GFP_IOFS (__GFP_IO | __GFP_FS)
114 #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
115 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \
116 __GFP_NO_KSWAPD)
117
118 #ifdef CONFIG_NUMA
119 #define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY)
120 #else
121 #define GFP_THISNODE ((__force gfp_t)0)
122 #endif
123
124 /* This mask makes up all the page movable related flags */
125 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
126
127 /* Control page allocator reclaim behavior */
128 #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
129 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
130 __GFP_NORETRY|__GFP_NOMEMALLOC)
131
132 /* Control slab gfp mask during early boot */
133 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
134
135 /* Control allocation constraints */
136 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
137
138 /* Do not use these with a slab allocator */
139 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
140
141 /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some
142 platforms, used as appropriate on others */
143
144 #define GFP_DMA __GFP_DMA
145
146 /* 4GB DMA on some platforms */
147 #define GFP_DMA32 __GFP_DMA32
148
149 /* Convert GFP flags to their corresponding migrate type */
allocflags_to_migratetype(gfp_t gfp_flags)150 static inline int allocflags_to_migratetype(gfp_t gfp_flags)
151 {
152 WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
153
154 if (unlikely(page_group_by_mobility_disabled))
155 return MIGRATE_UNMOVABLE;
156
157 /* Group based on mobility */
158 return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
159 ((gfp_flags & __GFP_RECLAIMABLE) != 0);
160 }
161
162 #ifdef CONFIG_HIGHMEM
163 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
164 #else
165 #define OPT_ZONE_HIGHMEM ZONE_NORMAL
166 #endif
167
168 #ifdef CONFIG_ZONE_DMA
169 #define OPT_ZONE_DMA ZONE_DMA
170 #else
171 #define OPT_ZONE_DMA ZONE_NORMAL
172 #endif
173
174 #ifdef CONFIG_ZONE_DMA32
175 #define OPT_ZONE_DMA32 ZONE_DMA32
176 #else
177 #define OPT_ZONE_DMA32 ZONE_NORMAL
178 #endif
179
180 /*
181 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
182 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
183 * and there are 16 of them to cover all possible combinations of
184 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
185 *
186 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
187 * But GFP_MOVABLE is not only a zone specifier but also an allocation
188 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
189 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
190 *
191 * bit result
192 * =================
193 * 0x0 => NORMAL
194 * 0x1 => DMA or NORMAL
195 * 0x2 => HIGHMEM or NORMAL
196 * 0x3 => BAD (DMA+HIGHMEM)
197 * 0x4 => DMA32 or DMA or NORMAL
198 * 0x5 => BAD (DMA+DMA32)
199 * 0x6 => BAD (HIGHMEM+DMA32)
200 * 0x7 => BAD (HIGHMEM+DMA32+DMA)
201 * 0x8 => NORMAL (MOVABLE+0)
202 * 0x9 => DMA or NORMAL (MOVABLE+DMA)
203 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too)
204 * 0xb => BAD (MOVABLE+HIGHMEM+DMA)
205 * 0xc => DMA32 (MOVABLE+HIGHMEM+DMA32)
206 * 0xd => BAD (MOVABLE+DMA32+DMA)
207 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM)
208 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
209 *
210 * ZONES_SHIFT must be <= 2 on 32 bit platforms.
211 */
212
213 #if 16 * ZONES_SHIFT > BITS_PER_LONG
214 #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
215 #endif
216
217 #define GFP_ZONE_TABLE ( \
218 (ZONE_NORMAL << 0 * ZONES_SHIFT) \
219 | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \
220 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \
221 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \
222 | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \
223 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \
224 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \
225 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \
226 )
227
228 /*
229 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
230 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
231 * entry starting with bit 0. Bit is set if the combination is not
232 * allowed.
233 */
234 #define GFP_ZONE_BAD ( \
235 1 << (___GFP_DMA | ___GFP_HIGHMEM) \
236 | 1 << (___GFP_DMA | ___GFP_DMA32) \
237 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \
238 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \
239 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \
240 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \
241 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \
242 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \
243 )
244
gfp_zone(gfp_t flags)245 static inline enum zone_type gfp_zone(gfp_t flags)
246 {
247 enum zone_type z;
248 int bit = (__force int) (flags & GFP_ZONEMASK);
249
250 z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
251 ((1 << ZONES_SHIFT) - 1);
252
253 if (__builtin_constant_p(bit))
254 BUILD_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
255 else {
256 #ifdef CONFIG_DEBUG_VM
257 BUG_ON((GFP_ZONE_BAD >> bit) & 1);
258 #endif
259 }
260 return z;
261 }
262
263 /*
264 * There is only one page-allocator function, and two main namespaces to
265 * it. The alloc_page*() variants return 'struct page *' and as such
266 * can allocate highmem pages, the *get*page*() variants return
267 * virtual kernel addresses to the allocated page(s).
268 */
269
gfp_zonelist(gfp_t flags)270 static inline int gfp_zonelist(gfp_t flags)
271 {
272 if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE))
273 return 1;
274
275 return 0;
276 }
277
278 /*
279 * We get the zone list from the current node and the gfp_mask.
280 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
281 * There are two zonelists per node, one for all zones with memory and
282 * one containing just zones from the node the zonelist belongs to.
283 *
284 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
285 * optimized to &contig_page_data at compile-time.
286 */
node_zonelist(int nid,gfp_t flags)287 static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
288 {
289 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
290 }
291
292 #ifndef HAVE_ARCH_FREE_PAGE
arch_free_page(struct page * page,int order)293 static inline void arch_free_page(struct page *page, int order) { }
294 #endif
295 #ifndef HAVE_ARCH_ALLOC_PAGE
arch_alloc_page(struct page * page,int order)296 static inline void arch_alloc_page(struct page *page, int order) { }
297 #endif
298
299 struct page *
300 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
301 struct zonelist *zonelist, nodemask_t *nodemask);
302
303 static inline struct page *
__alloc_pages(gfp_t gfp_mask,unsigned int order,struct zonelist * zonelist)304 __alloc_pages(gfp_t gfp_mask, unsigned int order,
305 struct zonelist *zonelist)
306 {
307 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
308 }
309
alloc_pages_node(int nid,gfp_t gfp_mask,unsigned int order)310 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
311 unsigned int order)
312 {
313 /* Unknown node is current node */
314 if (nid < 0)
315 nid = numa_node_id();
316
317 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
318 }
319
alloc_pages_exact_node(int nid,gfp_t gfp_mask,unsigned int order)320 static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask,
321 unsigned int order)
322 {
323 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
324
325 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
326 }
327
328 #ifdef CONFIG_NUMA
329 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
330
331 static inline struct page *
alloc_pages(gfp_t gfp_mask,unsigned int order)332 alloc_pages(gfp_t gfp_mask, unsigned int order)
333 {
334 return alloc_pages_current(gfp_mask, order);
335 }
336 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
337 struct vm_area_struct *vma, unsigned long addr,
338 int node);
339 #else
340 #define alloc_pages(gfp_mask, order) \
341 alloc_pages_node(numa_node_id(), gfp_mask, order)
342 #define alloc_pages_vma(gfp_mask, order, vma, addr, node) \
343 alloc_pages(gfp_mask, order)
344 #endif
345 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
346 #define alloc_page_vma(gfp_mask, vma, addr) \
347 alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id())
348 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \
349 alloc_pages_vma(gfp_mask, 0, vma, addr, node)
350
351 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
352 extern unsigned long get_zeroed_page(gfp_t gfp_mask);
353
354 void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
355 void free_pages_exact(void *virt, size_t size);
356 /* This is different from alloc_pages_exact_node !!! */
357 void *alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
358
359 #define __get_free_page(gfp_mask) \
360 __get_free_pages((gfp_mask), 0)
361
362 #define __get_dma_pages(gfp_mask, order) \
363 __get_free_pages((gfp_mask) | GFP_DMA, (order))
364
365 extern void __free_pages(struct page *page, unsigned int order);
366 extern void free_pages(unsigned long addr, unsigned int order);
367 extern void free_hot_cold_page(struct page *page, int cold);
368
369 #define __free_page(page) __free_pages((page), 0)
370 #define free_page(addr) free_pages((addr), 0)
371
372 void page_alloc_init(void);
373 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
374 void drain_all_pages(void);
375 void drain_local_pages(void *dummy);
376
377 extern gfp_t gfp_allowed_mask;
378
379 extern void pm_restrict_gfp_mask(void);
380 extern void pm_restore_gfp_mask(void);
381
382 #endif /* __LINUX_GFP_H */
383