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_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_RECLAIMABLE 0x80000u 33 #define ___GFP_NOTRACK 0x200000u 34 #define ___GFP_OTHER_NODE 0x800000u 35 #define ___GFP_WRITE 0x1000000u 36 37 /* 38 * GFP bitmasks.. 39 * 40 * Zone modifiers (see linux/mmzone.h - low three bits) 41 * 42 * Do not put any conditional on these. If necessary modify the definitions 43 * without the underscores and use them consistently. The definitions here may 44 * be used in bit comparisons. 45 */ 46 #define __GFP_DMA ((__force gfp_t)___GFP_DMA) 47 #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM) 48 #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32) 49 #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */ 50 #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE) 51 /* 52 * Action modifiers - doesn't change the zoning 53 * 54 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt 55 * _might_ fail. This depends upon the particular VM implementation. 56 * 57 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller 58 * cannot handle allocation failures. This modifier is deprecated and no new 59 * users should be added. 60 * 61 * __GFP_NORETRY: The VM implementation must not retry indefinitely. 62 * 63 * __GFP_MOVABLE: Flag that this page will be movable by the page migration 64 * mechanism or reclaimed 65 */ 66 #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */ 67 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */ 68 #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */ 69 #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */ 70 #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */ 71 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */ 72 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */ 73 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */ 74 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */ 75 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */ 76 #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */ 77 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */ 78 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves. 79 * This takes precedence over the 80 * __GFP_MEMALLOC flag if both are 81 * set 82 */ 83 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */ 84 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */ 85 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */ 86 #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */ 87 88 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */ 89 #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE) /* Allocator intends to dirty page */ 90 91 /* 92 * This may seem redundant, but it's a way of annotating false positives vs. 93 * allocations that simply cannot be supported (e.g. page tables). 94 */ 95 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK) 96 97 #define __GFP_BITS_SHIFT 25 /* Room for N __GFP_FOO bits */ 98 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1)) 99 100 /* This equals 0, but use constants in case they ever change */ 101 #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH) 102 /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */ 103 #define GFP_ATOMIC (__GFP_HIGH) 104 #define GFP_NOIO (__GFP_WAIT) 105 #define GFP_NOFS (__GFP_WAIT | __GFP_IO) 106 #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS) 107 #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \ 108 __GFP_RECLAIMABLE) 109 #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL) 110 #define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \ 111 __GFP_HIGHMEM) 112 #define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \ 113 __GFP_HARDWALL | __GFP_HIGHMEM | \ 114 __GFP_MOVABLE) 115 #define GFP_IOFS (__GFP_IO | __GFP_FS) 116 #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \ 117 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN) 118 119 #ifdef CONFIG_NUMA 120 #define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY) 121 #else 122 #define GFP_THISNODE ((__force gfp_t)0) 123 #endif 124 125 /* This mask makes up all the page movable related flags */ 126 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE) 127 128 /* Control page allocator reclaim behavior */ 129 #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\ 130 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\ 131 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC) 132 133 /* Control slab gfp mask during early boot */ 134 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS)) 135 136 /* Control allocation constraints */ 137 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) 138 139 /* Do not use these with a slab allocator */ 140 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) 141 142 /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some 143 platforms, used as appropriate on others */ 144 145 #define GFP_DMA __GFP_DMA 146 147 /* 4GB DMA on some platforms */ 148 #define GFP_DMA32 __GFP_DMA32 149 150 /* Convert GFP flags to their corresponding migrate type */ 151 static inline int allocflags_to_migratetype(gfp_t gfp_flags) 152 { 153 WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK); 154 155 if (unlikely(page_group_by_mobility_disabled)) 156 return MIGRATE_UNMOVABLE; 157 158 /* Group based on mobility */ 159 return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) | 160 ((gfp_flags & __GFP_RECLAIMABLE) != 0); 161 } 162 163 #ifdef CONFIG_HIGHMEM 164 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM 165 #else 166 #define OPT_ZONE_HIGHMEM ZONE_NORMAL 167 #endif 168 169 #ifdef CONFIG_ZONE_DMA 170 #define OPT_ZONE_DMA ZONE_DMA 171 #else 172 #define OPT_ZONE_DMA ZONE_NORMAL 173 #endif 174 175 #ifdef CONFIG_ZONE_DMA32 176 #define OPT_ZONE_DMA32 ZONE_DMA32 177 #else 178 #define OPT_ZONE_DMA32 ZONE_NORMAL 179 #endif 180 181 /* 182 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the 183 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long 184 * and there are 16 of them to cover all possible combinations of 185 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM. 186 * 187 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA. 188 * But GFP_MOVABLE is not only a zone specifier but also an allocation 189 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid. 190 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1". 191 * 192 * bit result 193 * ================= 194 * 0x0 => NORMAL 195 * 0x1 => DMA or NORMAL 196 * 0x2 => HIGHMEM or NORMAL 197 * 0x3 => BAD (DMA+HIGHMEM) 198 * 0x4 => DMA32 or DMA or NORMAL 199 * 0x5 => BAD (DMA+DMA32) 200 * 0x6 => BAD (HIGHMEM+DMA32) 201 * 0x7 => BAD (HIGHMEM+DMA32+DMA) 202 * 0x8 => NORMAL (MOVABLE+0) 203 * 0x9 => DMA or NORMAL (MOVABLE+DMA) 204 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too) 205 * 0xb => BAD (MOVABLE+HIGHMEM+DMA) 206 * 0xc => DMA32 (MOVABLE+HIGHMEM+DMA32) 207 * 0xd => BAD (MOVABLE+DMA32+DMA) 208 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM) 209 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA) 210 * 211 * ZONES_SHIFT must be <= 2 on 32 bit platforms. 212 */ 213 214 #if 16 * ZONES_SHIFT > BITS_PER_LONG 215 #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer 216 #endif 217 218 #define GFP_ZONE_TABLE ( \ 219 (ZONE_NORMAL << 0 * ZONES_SHIFT) \ 220 | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \ 221 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \ 222 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \ 223 | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \ 224 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \ 225 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \ 226 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \ 227 ) 228 229 /* 230 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32 231 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per 232 * entry starting with bit 0. Bit is set if the combination is not 233 * allowed. 234 */ 235 #define GFP_ZONE_BAD ( \ 236 1 << (___GFP_DMA | ___GFP_HIGHMEM) \ 237 | 1 << (___GFP_DMA | ___GFP_DMA32) \ 238 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \ 239 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \ 240 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \ 241 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \ 242 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \ 243 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \ 244 ) 245 246 static inline enum zone_type gfp_zone(gfp_t flags) 247 { 248 enum zone_type z; 249 int bit = (__force int) (flags & GFP_ZONEMASK); 250 251 z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) & 252 ((1 << ZONES_SHIFT) - 1); 253 VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1); 254 return z; 255 } 256 257 /* 258 * There is only one page-allocator function, and two main namespaces to 259 * it. The alloc_page*() variants return 'struct page *' and as such 260 * can allocate highmem pages, the *get*page*() variants return 261 * virtual kernel addresses to the allocated page(s). 262 */ 263 264 static inline int gfp_zonelist(gfp_t flags) 265 { 266 if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE)) 267 return 1; 268 269 return 0; 270 } 271 272 /* 273 * We get the zone list from the current node and the gfp_mask. 274 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones. 275 * There are two zonelists per node, one for all zones with memory and 276 * one containing just zones from the node the zonelist belongs to. 277 * 278 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets 279 * optimized to &contig_page_data at compile-time. 280 */ 281 static inline struct zonelist *node_zonelist(int nid, gfp_t flags) 282 { 283 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags); 284 } 285 286 #ifndef HAVE_ARCH_FREE_PAGE 287 static inline void arch_free_page(struct page *page, int order) { } 288 #endif 289 #ifndef HAVE_ARCH_ALLOC_PAGE 290 static inline void arch_alloc_page(struct page *page, int order) { } 291 #endif 292 293 struct page * 294 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, 295 struct zonelist *zonelist, nodemask_t *nodemask); 296 297 static inline struct page * 298 __alloc_pages(gfp_t gfp_mask, unsigned int order, 299 struct zonelist *zonelist) 300 { 301 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL); 302 } 303 304 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask, 305 unsigned int order) 306 { 307 /* Unknown node is current node */ 308 if (nid < 0) 309 nid = numa_node_id(); 310 311 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); 312 } 313 314 static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask, 315 unsigned int order) 316 { 317 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES || !node_online(nid)); 318 319 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); 320 } 321 322 #ifdef CONFIG_NUMA 323 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order); 324 325 static inline struct page * 326 alloc_pages(gfp_t gfp_mask, unsigned int order) 327 { 328 return alloc_pages_current(gfp_mask, order); 329 } 330 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order, 331 struct vm_area_struct *vma, unsigned long addr, 332 int node); 333 #else 334 #define alloc_pages(gfp_mask, order) \ 335 alloc_pages_node(numa_node_id(), gfp_mask, order) 336 #define alloc_pages_vma(gfp_mask, order, vma, addr, node) \ 337 alloc_pages(gfp_mask, order) 338 #endif 339 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0) 340 #define alloc_page_vma(gfp_mask, vma, addr) \ 341 alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id()) 342 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \ 343 alloc_pages_vma(gfp_mask, 0, vma, addr, node) 344 345 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order); 346 extern unsigned long get_zeroed_page(gfp_t gfp_mask); 347 348 void *alloc_pages_exact(size_t size, gfp_t gfp_mask); 349 void free_pages_exact(void *virt, size_t size); 350 /* This is different from alloc_pages_exact_node !!! */ 351 void *alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask); 352 353 #define __get_free_page(gfp_mask) \ 354 __get_free_pages((gfp_mask), 0) 355 356 #define __get_dma_pages(gfp_mask, order) \ 357 __get_free_pages((gfp_mask) | GFP_DMA, (order)) 358 359 extern void __free_pages(struct page *page, unsigned int order); 360 extern void free_pages(unsigned long addr, unsigned int order); 361 extern void free_hot_cold_page(struct page *page, int cold); 362 extern void free_hot_cold_page_list(struct list_head *list, int cold); 363 364 #define __free_page(page) __free_pages((page), 0) 365 #define free_page(addr) free_pages((addr), 0) 366 367 void page_alloc_init(void); 368 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp); 369 void drain_all_pages(void); 370 void drain_local_pages(void *dummy); 371 372 /* 373 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what 374 * GFP flags are used before interrupts are enabled. Once interrupts are 375 * enabled, it is set to __GFP_BITS_MASK while the system is running. During 376 * hibernation, it is used by PM to avoid I/O during memory allocation while 377 * devices are suspended. 378 */ 379 extern gfp_t gfp_allowed_mask; 380 381 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */ 382 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask); 383 384 extern void pm_restrict_gfp_mask(void); 385 extern void pm_restore_gfp_mask(void); 386 387 #ifdef CONFIG_PM_SLEEP 388 extern bool pm_suspended_storage(void); 389 #else 390 static inline bool pm_suspended_storage(void) 391 { 392 return false; 393 } 394 #endif /* CONFIG_PM_SLEEP */ 395 396 #ifdef CONFIG_CMA 397 398 /* The below functions must be run on a range from a single zone. */ 399 extern int alloc_contig_range(unsigned long start, unsigned long end, 400 unsigned migratetype); 401 extern void free_contig_range(unsigned long pfn, unsigned nr_pages); 402 403 /* CMA stuff */ 404 extern void init_cma_reserved_pageblock(struct page *page); 405 406 #endif 407 408 #endif /* __LINUX_GFP_H */ 409