1 #ifndef __LINUX_GFP_H 2 #define __LINUX_GFP_H 3 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> 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_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 */ 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. 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. 65 * 66 * __GFP_NORETRY: The VM implementation must not retry indefinitely. 67 * 68 * __GFP_MOVABLE: Flag that this page will be movable by the page migration 69 * mechanism or reclaimed 70 */ 71 #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */ 72 #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */ 73 #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */ 74 #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */ 75 #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */ 76 #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */ 77 #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */ 78 #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */ 79 #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */ 80 #define __GFP_MEMALLOC ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */ 81 #define __GFP_COMP ((__force gfp_t)___GFP_COMP) /* Add compound page metadata */ 82 #define __GFP_ZERO ((__force gfp_t)___GFP_ZERO) /* Return zeroed page on success */ 83 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves. 84 * This takes precedence over the 85 * __GFP_MEMALLOC flag if both are 86 * set 87 */ 88 #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */ 89 #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */ 90 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */ 91 #define __GFP_NOACCOUNT ((__force gfp_t)___GFP_NOACCOUNT) /* Don't account to kmemcg */ 92 #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */ 93 94 #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD) 95 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */ 96 #define __GFP_WRITE ((__force gfp_t)___GFP_WRITE) /* Allocator intends to dirty page */ 97 98 /* 99 * This may seem redundant, but it's a way of annotating false positives vs. 100 * allocations that simply cannot be supported (e.g. page tables). 101 */ 102 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK) 103 104 #define __GFP_BITS_SHIFT 25 /* Room for N __GFP_FOO bits */ 105 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1)) 106 107 /* This equals 0, but use constants in case they ever change */ 108 #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH) 109 /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */ 110 #define GFP_ATOMIC (__GFP_HIGH) 111 #define GFP_NOIO (__GFP_WAIT) 112 #define GFP_NOFS (__GFP_WAIT | __GFP_IO) 113 #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS) 114 #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \ 115 __GFP_RECLAIMABLE) 116 #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL) 117 #define GFP_HIGHUSER (GFP_USER | __GFP_HIGHMEM) 118 #define GFP_HIGHUSER_MOVABLE (GFP_HIGHUSER | __GFP_MOVABLE) 119 #define GFP_IOFS (__GFP_IO | __GFP_FS) 120 #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \ 121 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \ 122 __GFP_NO_KSWAPD) 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_MEMALLOC|__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 */ 150 static inline int gfpflags_to_migratetype(const 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+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 ) 244 245 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 VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1); 253 return z; 254 } 255 256 /* 257 * There is only one page-allocator function, and two main namespaces to 258 * it. The alloc_page*() variants return 'struct page *' and as such 259 * can allocate highmem pages, the *get*page*() variants return 260 * virtual kernel addresses to the allocated page(s). 261 */ 262 263 static inline int gfp_zonelist(gfp_t flags) 264 { 265 if (IS_ENABLED(CONFIG_NUMA) && unlikely(flags & __GFP_THISNODE)) 266 return 1; 267 268 return 0; 269 } 270 271 /* 272 * We get the zone list from the current node and the gfp_mask. 273 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones. 274 * There are two zonelists per node, one for all zones with memory and 275 * one containing just zones from the node the zonelist belongs to. 276 * 277 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets 278 * optimized to &contig_page_data at compile-time. 279 */ 280 static inline struct zonelist *node_zonelist(int nid, gfp_t flags) 281 { 282 return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags); 283 } 284 285 #ifndef HAVE_ARCH_FREE_PAGE 286 static inline void arch_free_page(struct page *page, int order) { } 287 #endif 288 #ifndef HAVE_ARCH_ALLOC_PAGE 289 static inline void arch_alloc_page(struct page *page, int order) { } 290 #endif 291 292 struct page * 293 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, 294 struct zonelist *zonelist, nodemask_t *nodemask); 295 296 static inline struct page * 297 __alloc_pages(gfp_t gfp_mask, unsigned int order, 298 struct zonelist *zonelist) 299 { 300 return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL); 301 } 302 303 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask, 304 unsigned int order) 305 { 306 /* Unknown node is current node */ 307 if (nid < 0) 308 nid = numa_node_id(); 309 310 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); 311 } 312 313 static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask, 314 unsigned int order) 315 { 316 VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES || !node_online(nid)); 317 318 return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); 319 } 320 321 #ifdef CONFIG_NUMA 322 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order); 323 324 static inline struct page * 325 alloc_pages(gfp_t gfp_mask, unsigned int order) 326 { 327 return alloc_pages_current(gfp_mask, order); 328 } 329 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order, 330 struct vm_area_struct *vma, unsigned long addr, 331 int node, bool hugepage); 332 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \ 333 alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true) 334 #else 335 #define alloc_pages(gfp_mask, order) \ 336 alloc_pages_node(numa_node_id(), gfp_mask, order) 337 #define alloc_pages_vma(gfp_mask, order, vma, addr, node, false)\ 338 alloc_pages(gfp_mask, order) 339 #define alloc_hugepage_vma(gfp_mask, vma, addr, order) \ 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(), false) 345 #define alloc_page_vma_node(gfp_mask, vma, addr, node) \ 346 alloc_pages_vma(gfp_mask, 0, vma, addr, node, false) 347 348 extern struct page *alloc_kmem_pages(gfp_t gfp_mask, unsigned int order); 349 extern struct page *alloc_kmem_pages_node(int nid, gfp_t gfp_mask, 350 unsigned int order); 351 352 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order); 353 extern unsigned long get_zeroed_page(gfp_t gfp_mask); 354 355 void *alloc_pages_exact(size_t size, gfp_t gfp_mask); 356 void free_pages_exact(void *virt, size_t size); 357 /* This is different from alloc_pages_exact_node !!! */ 358 void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask); 359 360 #define __get_free_page(gfp_mask) \ 361 __get_free_pages((gfp_mask), 0) 362 363 #define __get_dma_pages(gfp_mask, order) \ 364 __get_free_pages((gfp_mask) | GFP_DMA, (order)) 365 366 extern void __free_pages(struct page *page, unsigned int order); 367 extern void free_pages(unsigned long addr, unsigned int order); 368 extern void free_hot_cold_page(struct page *page, bool cold); 369 extern void free_hot_cold_page_list(struct list_head *list, bool cold); 370 371 struct page_frag_cache; 372 extern void *__alloc_page_frag(struct page_frag_cache *nc, 373 unsigned int fragsz, gfp_t gfp_mask); 374 extern void __free_page_frag(void *addr); 375 376 extern void __free_kmem_pages(struct page *page, unsigned int order); 377 extern void free_kmem_pages(unsigned long addr, unsigned int order); 378 379 #define __free_page(page) __free_pages((page), 0) 380 #define free_page(addr) free_pages((addr), 0) 381 382 void page_alloc_init(void); 383 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp); 384 void drain_all_pages(struct zone *zone); 385 void drain_local_pages(struct zone *zone); 386 387 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT 388 void page_alloc_init_late(void); 389 #else 390 static inline void page_alloc_init_late(void) 391 { 392 } 393 #endif 394 395 /* 396 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what 397 * GFP flags are used before interrupts are enabled. Once interrupts are 398 * enabled, it is set to __GFP_BITS_MASK while the system is running. During 399 * hibernation, it is used by PM to avoid I/O during memory allocation while 400 * devices are suspended. 401 */ 402 extern gfp_t gfp_allowed_mask; 403 404 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */ 405 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask); 406 407 extern void pm_restrict_gfp_mask(void); 408 extern void pm_restore_gfp_mask(void); 409 410 #ifdef CONFIG_PM_SLEEP 411 extern bool pm_suspended_storage(void); 412 #else 413 static inline bool pm_suspended_storage(void) 414 { 415 return false; 416 } 417 #endif /* CONFIG_PM_SLEEP */ 418 419 #ifdef CONFIG_CMA 420 421 /* The below functions must be run on a range from a single zone. */ 422 extern int alloc_contig_range(unsigned long start, unsigned long end, 423 unsigned migratetype); 424 extern void free_contig_range(unsigned long pfn, unsigned nr_pages); 425 426 /* CMA stuff */ 427 extern void init_cma_reserved_pageblock(struct page *page); 428 429 #endif 430 431 #endif /* __LINUX_GFP_H */ 432