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_NO_KSWAPD 0x400000u 35 #define ___GFP_OTHER_NODE 0x800000u 36 #define ___GFP_WRITE 0x1000000u 37 38 /* 39 * GFP bitmasks.. 40 * 41 * Zone modifiers (see linux/mmzone.h - low three bits) 42 * 43 * Do not put any conditional on these. If necessary modify the definitions 44 * without the underscores and use them consistently. The definitions here may 45 * be used in bit comparisons. 46 */ 47 #define __GFP_DMA ((__force gfp_t)___GFP_DMA) 48 #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM) 49 #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32) 50 #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */ 51 #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE) 52 /* 53 * Action modifiers - doesn't change the zoning 54 * 55 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt 56 * _might_ fail. This depends upon the particular VM implementation. 57 * 58 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller 59 * cannot handle allocation failures. This modifier is deprecated and no new 60 * users should be added. 61 * 62 * __GFP_NORETRY: The VM implementation must not retry indefinitely. 63 * 64 * __GFP_MOVABLE: Flag that this page will be movable by the page migration 65 * mechanism or reclaimed 66 */ 67 #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */ 68 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */ 69 #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */ 70 #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */ 71 #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */ 72 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */ 73 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */ 74 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */ 75 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */ 76 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */ 77 #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */ 78 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */ 79 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves. 80 * This takes precedence over the 81 * __GFP_MEMALLOC flag if both are 82 * set 83 */ 84 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */ 85 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */ 86 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */ 87 #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */ 88 89 #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD) 90 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */ 91 #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE) /* Allocator intends to dirty page */ 92 93 /* 94 * This may seem redundant, but it's a way of annotating false positives vs. 95 * allocations that simply cannot be supported (e.g. page tables). 96 */ 97 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK) 98 99 #define __GFP_BITS_SHIFT 25 /* Room for N __GFP_FOO bits */ 100 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1)) 101 102 /* This equals 0, but use constants in case they ever change */ 103 #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH) 104 /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */ 105 #define GFP_ATOMIC (__GFP_HIGH) 106 #define GFP_NOIO (__GFP_WAIT) 107 #define GFP_NOFS (__GFP_WAIT | __GFP_IO) 108 #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS) 109 #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \ 110 __GFP_RECLAIMABLE) 111 #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL) 112 #define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \ 113 __GFP_HIGHMEM) 114 #define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \ 115 __GFP_HARDWALL | __GFP_HIGHMEM | \ 116 __GFP_MOVABLE) 117 #define GFP_IOFS (__GFP_IO | __GFP_FS) 118 #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \ 119 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \ 120 __GFP_NO_KSWAPD) 121 122 #ifdef CONFIG_NUMA 123 #define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY) 124 #else 125 #define GFP_THISNODE ((__force gfp_t)0) 126 #endif 127 128 /* This mask makes up all the page movable related flags */ 129 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE) 130 131 /* 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) 135 136 /* Control slab gfp mask during early boot */ 137 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS)) 138 139 /* Control allocation constraints */ 140 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) 141 142 /* Do not use these with a slab allocator */ 143 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) 144 145 /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some 146 platforms, used as appropriate on others */ 147 148 #define GFP_DMA __GFP_DMA 149 150 /* 4GB DMA on some platforms */ 151 #define GFP_DMA32 __GFP_DMA32 152 153 /* Convert GFP flags to their corresponding migrate type */ 154 static inline int allocflags_to_migratetype(gfp_t gfp_flags) 155 { 156 WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK); 157 158 if (unlikely(page_group_by_mobility_disabled)) 159 return MIGRATE_UNMOVABLE; 160 161 /* Group based on mobility */ 162 return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) | 163 ((gfp_flags & __GFP_RECLAIMABLE) != 0); 164 } 165 166 #ifdef CONFIG_HIGHMEM 167 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM 168 #else 169 #define OPT_ZONE_HIGHMEM ZONE_NORMAL 170 #endif 171 172 #ifdef CONFIG_ZONE_DMA 173 #define OPT_ZONE_DMA ZONE_DMA 174 #else 175 #define OPT_ZONE_DMA ZONE_NORMAL 176 #endif 177 178 #ifdef CONFIG_ZONE_DMA32 179 #define OPT_ZONE_DMA32 ZONE_DMA32 180 #else 181 #define OPT_ZONE_DMA32 ZONE_NORMAL 182 #endif 183 184 /* 185 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the 186 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long 187 * and there are 16 of them to cover all possible combinations of 188 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM. 189 * 190 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA. 191 * But GFP_MOVABLE is not only a zone specifier but also an allocation 192 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid. 193 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1". 194 * 195 * bit result 196 * ================= 197 * 0x0 => NORMAL 198 * 0x1 => DMA or NORMAL 199 * 0x2 => HIGHMEM or NORMAL 200 * 0x3 => BAD (DMA+HIGHMEM) 201 * 0x4 => DMA32 or DMA or NORMAL 202 * 0x5 => BAD (DMA+DMA32) 203 * 0x6 => BAD (HIGHMEM+DMA32) 204 * 0x7 => BAD (HIGHMEM+DMA32+DMA) 205 * 0x8 => NORMAL (MOVABLE+0) 206 * 0x9 => DMA or NORMAL (MOVABLE+DMA) 207 * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too) 208 * 0xb => BAD (MOVABLE+HIGHMEM+DMA) 209 * 0xc => DMA32 (MOVABLE+HIGHMEM+DMA32) 210 * 0xd => BAD (MOVABLE+DMA32+DMA) 211 * 0xe => BAD (MOVABLE+DMA32+HIGHMEM) 212 * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA) 213 * 214 * ZONES_SHIFT must be <= 2 on 32 bit platforms. 215 */ 216 217 #if 16 * ZONES_SHIFT > BITS_PER_LONG 218 #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer 219 #endif 220 221 #define GFP_ZONE_TABLE ( \ 222 (ZONE_NORMAL << 0 * ZONES_SHIFT) \ 223 | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \ 224 | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \ 225 | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \ 226 | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \ 227 | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \ 228 | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \ 229 | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \ 230 ) 231 232 /* 233 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32 234 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per 235 * entry starting with bit 0. Bit is set if the combination is not 236 * allowed. 237 */ 238 #define GFP_ZONE_BAD ( \ 239 1 << (___GFP_DMA | ___GFP_HIGHMEM) \ 240 | 1 << (___GFP_DMA | ___GFP_DMA32) \ 241 | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \ 242 | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \ 243 | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \ 244 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \ 245 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \ 246 | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \ 247 ) 248 249 static inline enum zone_type gfp_zone(gfp_t flags) 250 { 251 enum zone_type z; 252 int bit = (__force int) (flags & GFP_ZONEMASK); 253 254 z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) & 255 ((1 << ZONES_SHIFT) - 1); 256 VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1); 257 return z; 258 } 259 260 /* 261 * There is only one page-allocator function, and two main namespaces to 262 * it. The alloc_page*() variants return 'struct page *' and as such 263 * can allocate highmem pages, the *get*page*() variants return 264 * virtual kernel addresses to the allocated page(s). 265 */ 266 267 static inline int gfp_zonelist(gfp_t flags) 268 { 269 if (IS_ENABLED(CONFIG_NUMA) && unlikely(flags & __GFP_THISNODE)) 270 return 1; 271 272 return 0; 273 } 274 275 /* 276 * We get the zone list from the current node and the gfp_mask. 277 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones. 278 * There are two zonelists per node, one for all zones with memory and 279 * one containing just zones from the node the zonelist belongs to. 280 * 281 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets 282 * optimized to &contig_page_data at compile-time. 283 */ 284 static inline struct zonelist *node_zonelist(int nid, gfp_t flags) 285 { 286 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags); 287 } 288 289 #ifndef HAVE_ARCH_FREE_PAGE 290 static inline void arch_free_page(struct page *page, int order) { } 291 #endif 292 #ifndef HAVE_ARCH_ALLOC_PAGE 293 static inline void arch_alloc_page(struct page *page, int order) { } 294 #endif 295 296 struct page * 297 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, 298 struct zonelist *zonelist, nodemask_t *nodemask); 299 300 static inline struct page * 301 __alloc_pages(gfp_t gfp_mask, unsigned int order, 302 struct zonelist *zonelist) 303 { 304 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL); 305 } 306 307 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask, 308 unsigned int order) 309 { 310 /* Unknown node is current node */ 311 if (nid < 0) 312 nid = numa_node_id(); 313 314 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); 315 } 316 317 static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask, 318 unsigned int order) 319 { 320 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES || !node_online(nid)); 321 322 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); 323 } 324 325 #ifdef CONFIG_NUMA 326 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order); 327 328 static inline struct page * 329 alloc_pages(gfp_t gfp_mask, unsigned int order) 330 { 331 return alloc_pages_current(gfp_mask, order); 332 } 333 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order, 334 struct vm_area_struct *vma, unsigned long addr, 335 int node); 336 #else 337 #define alloc_pages(gfp_mask, order) \ 338 alloc_pages_node(numa_node_id(), gfp_mask, order) 339 #define alloc_pages_vma(gfp_mask, order, vma, addr, node) \ 340 alloc_pages(gfp_mask, order) 341 #endif 342 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0) 343 #define alloc_page_vma(gfp_mask, vma, addr) \ 344 alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id()) 345 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \ 346 alloc_pages_vma(gfp_mask, 0, vma, addr, node) 347 348 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order); 349 extern unsigned long get_zeroed_page(gfp_t gfp_mask); 350 351 void *alloc_pages_exact(size_t size, gfp_t gfp_mask); 352 void free_pages_exact(void *virt, size_t size); 353 /* This is different from alloc_pages_exact_node !!! */ 354 void *alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask); 355 356 #define __get_free_page(gfp_mask) \ 357 __get_free_pages((gfp_mask), 0) 358 359 #define __get_dma_pages(gfp_mask, order) \ 360 __get_free_pages((gfp_mask) | GFP_DMA, (order)) 361 362 extern void __free_pages(struct page *page, unsigned int order); 363 extern void free_pages(unsigned long addr, unsigned int order); 364 extern void free_hot_cold_page(struct page *page, int cold); 365 extern void free_hot_cold_page_list(struct list_head *list, int cold); 366 367 #define __free_page(page) __free_pages((page), 0) 368 #define free_page(addr) free_pages((addr), 0) 369 370 void page_alloc_init(void); 371 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp); 372 void drain_all_pages(void); 373 void drain_local_pages(void *dummy); 374 375 /* 376 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what 377 * GFP flags are used before interrupts are enabled. Once interrupts are 378 * enabled, it is set to __GFP_BITS_MASK while the system is running. During 379 * hibernation, it is used by PM to avoid I/O during memory allocation while 380 * devices are suspended. 381 */ 382 extern gfp_t gfp_allowed_mask; 383 384 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */ 385 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask); 386 387 extern void pm_restrict_gfp_mask(void); 388 extern void pm_restore_gfp_mask(void); 389 390 #ifdef CONFIG_PM_SLEEP 391 extern bool pm_suspended_storage(void); 392 #else 393 static inline bool pm_suspended_storage(void) 394 { 395 return false; 396 } 397 #endif /* CONFIG_PM_SLEEP */ 398 399 #ifdef CONFIG_CMA 400 401 /* The below functions must be run on a range from a single zone. */ 402 extern int alloc_contig_range(unsigned long start, unsigned long end, 403 unsigned migratetype); 404 extern void free_contig_range(unsigned long pfn, unsigned nr_pages); 405 406 /* CMA stuff */ 407 extern void init_cma_reserved_pageblock(struct page *page); 408 409 #endif 410 411 #endif /* __LINUX_GFP_H */ 412