| 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 | |
| 39 | /* |
| 40 | * GFP bitmasks.. |
| 41 | * |
| 42 | * Zone modifiers (see linux/mmzone.h - low three bits) |
| 43 | * |
| 44 | * Do not put any conditional on these. If necessary modify the definitions |
| 45 | * without the underscores and use them consistently. The definitions here may |
| 46 | * be used in bit comparisons. |
| 47 | */ |
| 48 | #define __GFP_DMA ((__force gfp_t)___GFP_DMA) |
| 49 | #define __GFP_HIGHMEM ((__force gfp_t)___GFP_HIGHMEM) |
| 50 | #define __GFP_DMA32 ((__force gfp_t)___GFP_DMA32) |
| 51 | #define __GFP_MOVABLE ((__force gfp_t)___GFP_MOVABLE) /* Page is movable */ |
| 52 | #define GFP_ZONEMASK (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE) |
| 53 | /* |
| 54 | * Action modifiers - doesn't change the zoning |
| 55 | * |
| 56 | * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt |
| 57 | * _might_ fail. This depends upon the particular VM implementation. |
| 58 | * |
| 59 | * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller |
| 60 | * cannot handle allocation failures. This modifier is deprecated and no new |
| 61 | * users should be added. |
| 62 | * |
| 63 | * __GFP_NORETRY: The VM implementation must not retry indefinitely. |
| 64 | * |
| 65 | * __GFP_MOVABLE: Flag that this page will be movable by the page migration |
| 66 | * mechanism or reclaimed |
| 67 | */ |
| 68 | #define __GFP_WAIT ((__force gfp_t)___GFP_WAIT) /* Can wait and reschedule? */ |
| 69 | #define __GFP_HIGH ((__force gfp_t)___GFP_HIGH) /* Should access emergency pools? */ |
| 70 | #define __GFP_IO ((__force gfp_t)___GFP_IO) /* Can start physical IO? */ |
| 71 | #define __GFP_FS ((__force gfp_t)___GFP_FS) /* Can call down to low-level FS? */ |
| 72 | #define __GFP_COLD ((__force gfp_t)___GFP_COLD) /* Cache-cold page required */ |
| 73 | #define __GFP_NOWARN ((__force gfp_t)___GFP_NOWARN) /* Suppress page allocation failure warning */ |
| 74 | #define __GFP_REPEAT ((__force gfp_t)___GFP_REPEAT) /* See above */ |
| 75 | #define __GFP_NOFAIL ((__force gfp_t)___GFP_NOFAIL) /* See above */ |
| 76 | #define __GFP_NORETRY ((__force gfp_t)___GFP_NORETRY) /* See above */ |
| 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 | #define __GFP_HARDWALL ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */ |
| 81 | #define __GFP_THISNODE ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */ |
| 82 | #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */ |
| 83 | #define __GFP_NOTRACK ((__force gfp_t)___GFP_NOTRACK) /* Don't track with kmemcheck */ |
| 84 | |
| 85 | #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD) |
| 86 | |
| 87 | /* |
| 88 | * This may seem redundant, but it's a way of annotating false positives vs. |
| 89 | * allocations that simply cannot be supported (e.g. page tables). |
| 90 | */ |
| 91 | #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK) |
| 92 | |
| 93 | #define __GFP_BITS_SHIFT 23 /* Room for 23 __GFP_FOO bits */ |
| 94 | #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1)) |
| 95 | |
| 96 | /* This equals 0, but use constants in case they ever change */ |
| 97 | #define GFP_NOWAIT (GFP_ATOMIC & ~__GFP_HIGH) |
| 98 | /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */ |
| 99 | #define GFP_ATOMIC (__GFP_HIGH) |
| 100 | #define GFP_NOIO (__GFP_WAIT) |
| 101 | #define GFP_NOFS (__GFP_WAIT | __GFP_IO) |
| 102 | #define GFP_KERNEL (__GFP_WAIT | __GFP_IO | __GFP_FS) |
| 103 | #define GFP_TEMPORARY (__GFP_WAIT | __GFP_IO | __GFP_FS | \ |
| 104 | __GFP_RECLAIMABLE) |
| 105 | #define GFP_USER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL) |
| 106 | #define GFP_HIGHUSER (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \ |
| 107 | __GFP_HIGHMEM) |
| 108 | #define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \ |
| 109 | __GFP_HARDWALL | __GFP_HIGHMEM | \ |
| 110 | __GFP_MOVABLE) |
| 111 | #define GFP_IOFS (__GFP_IO | __GFP_FS) |
| 112 | #define GFP_TRANSHUGE (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \ |
| 113 | __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \ |
| 114 | __GFP_NO_KSWAPD) |
| 115 | |
| 116 | #ifdef CONFIG_NUMA |
| 117 | #define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY) |
| 118 | #else |
| 119 | #define GFP_THISNODE ((__force gfp_t)0) |
| 120 | #endif |
| 121 | |
| 122 | /* This mask makes up all the page movable related flags */ |
| 123 | #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE) |
| 124 | |
| 125 | /* Control page allocator reclaim behavior */ |
| 126 | #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\ |
| 127 | __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\ |
| 128 | __GFP_NORETRY|__GFP_NOMEMALLOC) |
| 129 | |
| 130 | /* Control slab gfp mask during early boot */ |
| 131 | #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS)) |
| 132 | |
| 133 | /* Control allocation constraints */ |
| 134 | #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) |
| 135 | |
| 136 | /* Do not use these with a slab allocator */ |
| 137 | #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) |
| 138 | |
| 139 | /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some |
| 140 | platforms, used as appropriate on others */ |
| 141 | |
| 142 | #define GFP_DMA __GFP_DMA |
| 143 | |
| 144 | /* 4GB DMA on some platforms */ |
| 145 | #define GFP_DMA32 __GFP_DMA32 |
| 146 | |
| 147 | /* Convert GFP flags to their corresponding migrate type */ |
| 148 | static inline int allocflags_to_migratetype(gfp_t gfp_flags) |
| 149 | { |
| 150 | WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK); |
| 151 | |
| 152 | if (unlikely(page_group_by_mobility_disabled)) |
| 153 | return MIGRATE_UNMOVABLE; |
| 154 | |
| 155 | /* Group based on mobility */ |
| 156 | return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) | |
| 157 | ((gfp_flags & __GFP_RECLAIMABLE) != 0); |
| 158 | } |
| 159 | |
| 160 | #ifdef CONFIG_HIGHMEM |
| 161 | #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM |
| 162 | #else |
| 163 | #define OPT_ZONE_HIGHMEM ZONE_NORMAL |
| 164 | #endif |
| 165 | |
| 166 | #ifdef CONFIG_ZONE_DMA |
| 167 | #define OPT_ZONE_DMA ZONE_DMA |
| 168 | #else |
| 169 | #define OPT_ZONE_DMA ZONE_NORMAL |
| 170 | #endif |
| 171 | |
| 172 | #ifdef CONFIG_ZONE_DMA32 |
| 173 | #define OPT_ZONE_DMA32 ZONE_DMA32 |
| 174 | #else |
| 175 | #define OPT_ZONE_DMA32 ZONE_NORMAL |
| 176 | #endif |
| 177 | |
| 178 | /* |
| 179 | * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the |
| 180 | * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long |
| 181 | * and there are 16 of them to cover all possible combinations of |
| 182 | * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM. |
| 183 | * |
| 184 | * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA. |
| 185 | * But GFP_MOVABLE is not only a zone specifier but also an allocation |
| 186 | * policy. Therefore __GFP_MOVABLE plus another zone selector is valid. |
| 187 | * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1". |
| 188 | * |
| 189 | * bit result |
| 190 | * ================= |
| 191 | * 0x0 => NORMAL |
| 192 | * 0x1 => DMA or NORMAL |
| 193 | * 0x2 => HIGHMEM or NORMAL |
| 194 | * 0x3 => BAD (DMA+HIGHMEM) |
| 195 | * 0x4 => DMA32 or DMA or NORMAL |
| 196 | * 0x5 => BAD (DMA+DMA32) |
| 197 | * 0x6 => BAD (HIGHMEM+DMA32) |
| 198 | * 0x7 => BAD (HIGHMEM+DMA32+DMA) |
| 199 | * 0x8 => NORMAL (MOVABLE+0) |
| 200 | * 0x9 => DMA or NORMAL (MOVABLE+DMA) |
| 201 | * 0xa => MOVABLE (Movable is valid only if HIGHMEM is set too) |
| 202 | * 0xb => BAD (MOVABLE+HIGHMEM+DMA) |
| 203 | * 0xc => DMA32 (MOVABLE+HIGHMEM+DMA32) |
| 204 | * 0xd => BAD (MOVABLE+DMA32+DMA) |
| 205 | * 0xe => BAD (MOVABLE+DMA32+HIGHMEM) |
| 206 | * 0xf => BAD (MOVABLE+DMA32+HIGHMEM+DMA) |
| 207 | * |
| 208 | * ZONES_SHIFT must be <= 2 on 32 bit platforms. |
| 209 | */ |
| 210 | |
| 211 | #if 16 * ZONES_SHIFT > BITS_PER_LONG |
| 212 | #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer |
| 213 | #endif |
| 214 | |
| 215 | #define GFP_ZONE_TABLE ( \ |
| 216 | (ZONE_NORMAL << 0 * ZONES_SHIFT) \ |
| 217 | | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT) \ |
| 218 | | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT) \ |
| 219 | | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT) \ |
| 220 | | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT) \ |
| 221 | | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT) \ |
| 222 | | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT) \ |
| 223 | | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT) \ |
| 224 | ) |
| 225 | |
| 226 | /* |
| 227 | * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32 |
| 228 | * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per |
| 229 | * entry starting with bit 0. Bit is set if the combination is not |
| 230 | * allowed. |
| 231 | */ |
| 232 | #define GFP_ZONE_BAD ( \ |
| 233 | 1 << (___GFP_DMA | ___GFP_HIGHMEM) \ |
| 234 | | 1 << (___GFP_DMA | ___GFP_DMA32) \ |
| 235 | | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM) \ |
| 236 | | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM) \ |
| 237 | | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA) \ |
| 238 | | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA) \ |
| 239 | | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM) \ |
| 240 | | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM) \ |
| 241 | ) |
| 242 | |
| 243 | static inline enum zone_type gfp_zone(gfp_t flags) |
| 244 | { |
| 245 | enum zone_type z; |
| 246 | int bit = (__force int) (flags & GFP_ZONEMASK); |
| 247 | |
| 248 | z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) & |
| 249 | ((1 << ZONES_SHIFT) - 1); |
| 250 | |
| 251 | if (__builtin_constant_p(bit)) |
| 252 | BUILD_BUG_ON((GFP_ZONE_BAD >> bit) & 1); |
| 253 | else { |
| 254 | #ifdef CONFIG_DEBUG_VM |
| 255 | BUG_ON((GFP_ZONE_BAD >> bit) & 1); |
| 256 | #endif |
| 257 | } |
| 258 | return z; |
| 259 | } |
| 260 | |
| 261 | /* |
| 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). |
| 266 | */ |
| 267 | |
| 268 | static inline int gfp_zonelist(gfp_t flags) |
| 269 | { |
| 270 | if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE)) |
| 271 | return 1; |
| 272 | |
| 273 | return 0; |
| 274 | } |
| 275 | |
| 276 | /* |
| 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. |
| 281 | * |
| 282 | * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets |
| 283 | * optimized to &contig_page_data at compile-time. |
| 284 | */ |
| 285 | static inline struct zonelist *node_zonelist(int nid, gfp_t flags) |
| 286 | { |
| 287 | return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags); |
| 288 | } |
| 289 | |
| 290 | #ifndef HAVE_ARCH_FREE_PAGE |
| 291 | static inline void arch_free_page(struct page *page, int order) { } |
| 292 | #endif |
| 293 | #ifndef HAVE_ARCH_ALLOC_PAGE |
| 294 | static inline void arch_alloc_page(struct page *page, int order) { } |
| 295 | #endif |
| 296 | |
| 297 | struct page * |
| 298 | __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order, |
| 299 | struct zonelist *zonelist, nodemask_t *nodemask); |
| 300 | |
| 301 | static inline struct page * |
| 302 | __alloc_pages(gfp_t gfp_mask, unsigned int order, |
| 303 | struct zonelist *zonelist) |
| 304 | { |
| 305 | return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL); |
| 306 | } |
| 307 | |
| 308 | static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask, |
| 309 | unsigned int order) |
| 310 | { |
| 311 | /* Unknown node is current node */ |
| 312 | if (nid < 0) |
| 313 | nid = numa_node_id(); |
| 314 | |
| 315 | return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); |
| 316 | } |
| 317 | |
| 318 | static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask, |
| 319 | unsigned int order) |
| 320 | { |
| 321 | VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES); |
| 322 | |
| 323 | return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask)); |
| 324 | } |
| 325 | |
| 326 | #ifdef CONFIG_NUMA |
| 327 | extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order); |
| 328 | |
| 329 | static inline struct page * |
| 330 | alloc_pages(gfp_t gfp_mask, unsigned int order) |
| 331 | { |
| 332 | return alloc_pages_current(gfp_mask, order); |
| 333 | } |
| 334 | extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order, |
| 335 | struct vm_area_struct *vma, unsigned long addr, |
| 336 | int node); |
| 337 | #else |
| 338 | #define alloc_pages(gfp_mask, order) \ |
| 339 | alloc_pages_node(numa_node_id(), gfp_mask, order) |
| 340 | #define alloc_pages_vma(gfp_mask, order, vma, addr, node) \ |
| 341 | alloc_pages(gfp_mask, order) |
| 342 | #endif |
| 343 | #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0) |
| 344 | #define alloc_page_vma(gfp_mask, vma, addr) \ |
| 345 | alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id()) |
| 346 | #define alloc_page_vma_node(gfp_mask, vma, addr, node) \ |
| 347 | alloc_pages_vma(gfp_mask, 0, vma, addr, node) |
| 348 | |
| 349 | extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order); |
| 350 | extern unsigned long get_zeroed_page(gfp_t gfp_mask); |
| 351 | |
| 352 | void *alloc_pages_exact(size_t size, gfp_t gfp_mask); |
| 353 | void free_pages_exact(void *virt, size_t size); |
| 354 | |
| 355 | #define __get_free_page(gfp_mask) \ |
| 356 | __get_free_pages((gfp_mask), 0) |
| 357 | |
| 358 | #define __get_dma_pages(gfp_mask, order) \ |
| 359 | __get_free_pages((gfp_mask) | GFP_DMA, (order)) |
| 360 | |
| 361 | extern void __free_pages(struct page *page, unsigned int order); |
| 362 | extern void free_pages(unsigned long addr, unsigned int order); |
| 363 | extern void free_hot_cold_page(struct page *page, int cold); |
| 364 | |
| 365 | #define __free_page(page) __free_pages((page), 0) |
| 366 | #define free_page(addr) free_pages((addr), 0) |
| 367 | |
| 368 | void page_alloc_init(void); |
| 369 | void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp); |
| 370 | void drain_all_pages(void); |
| 371 | void drain_local_pages(void *dummy); |
| 372 | |
| 373 | extern gfp_t gfp_allowed_mask; |
| 374 | |
| 375 | extern void pm_restrict_gfp_mask(void); |
| 376 | extern void pm_restore_gfp_mask(void); |
| 377 | |
| 378 | #endif /* __LINUX_GFP_H */ |