include/linux/gfp.h

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