include/linux/gfp.h

   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 and will
  67  * return NULL when direct reclaim and memory compaction have failed to allow
  68  * the allocation to succeed.  The OOM killer is not called with the current
  69  * implementation.
  70  *
  71  * __GFP_MOVABLE: Flag that this page will be movable by the page migration
  72  * mechanism or reclaimed
  73  */
  74 #define __GFP_WAIT      ((__force gfp_t)___GFP_WAIT)    /* Can wait and reschedule? */
  75 #define __GFP_HIGH      ((__force gfp_t)___GFP_HIGH)    /* Should access emergency pools? */
  76 #define __GFP_IO        ((__force gfp_t)___GFP_IO)      /* Can start physical IO? */
  77 #define __GFP_FS        ((__force gfp_t)___GFP_FS)      /* Can call down to low-level FS? */
  78 #define __GFP_COLD      ((__force gfp_t)___GFP_COLD)    /* Cache-cold page required */
  79 #define __GFP_NOWARN    ((__force gfp_t)___GFP_NOWARN)  /* Suppress page allocation failure warning */
  80 #define __GFP_REPEAT    ((__force gfp_t)___GFP_REPEAT)  /* See above */
  81 #define __GFP_NOFAIL    ((__force gfp_t)___GFP_NOFAIL)  /* See above */
  82 #define __GFP_NORETRY   ((__force gfp_t)___GFP_NORETRY) /* See above */
  83 #define __GFP_MEMALLOC  ((__force gfp_t)___GFP_MEMALLOC)/* Allow access to emergency reserves */
  84 #define __GFP_COMP      ((__force gfp_t)___GFP_COMP)    /* Add compound page metadata */
  85 #define __GFP_ZERO      ((__force gfp_t)___GFP_ZERO)    /* Return zeroed page on success */
  86 #define __GFP_NOMEMALLOC ((__force gfp_t)___GFP_NOMEMALLOC) /* Don't use emergency reserves.
  87                                                          * This takes precedence over the
  88                                                          * __GFP_MEMALLOC flag if both are
  89                                                          * set
  90                                                          */
  91 #define __GFP_HARDWALL   ((__force gfp_t)___GFP_HARDWALL) /* Enforce hardwall cpuset memory allocs */
  92 #define __GFP_THISNODE  ((__force gfp_t)___GFP_THISNODE)/* No fallback, no policies */
  93 #define __GFP_RECLAIMABLE ((__force gfp_t)___GFP_RECLAIMABLE) /* Page is reclaimable */
  94 #define __GFP_NOACCOUNT ((__force gfp_t)___GFP_NOACCOUNT) /* Don't account to kmemcg */
  95 #define __GFP_NOTRACK   ((__force gfp_t)___GFP_NOTRACK)  /* Don't track with kmemcheck */
  96
  97 #define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD)
  98 #define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */
  99 #define __GFP_WRITE     ((__force gfp_t)___GFP_WRITE)   /* Allocator intends to dirty page */
 100
 101 /*
 102  * This may seem redundant, but it's a way of annotating false positives vs.
 103  * allocations that simply cannot be supported (e.g. page tables).
 104  */
 105 #define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
 106
 107 #define __GFP_BITS_SHIFT 25     /* Room for N __GFP_FOO bits */
 108 #define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 109
 110 /* This equals 0, but use constants in case they ever change */
 111 #define GFP_NOWAIT      (GFP_ATOMIC & ~__GFP_HIGH)
 112 /* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
 113 #define GFP_ATOMIC      (__GFP_HIGH)
 114 #define GFP_NOIO        (__GFP_WAIT)
 115 #define GFP_NOFS        (__GFP_WAIT | __GFP_IO)
 116 #define GFP_KERNEL      (__GFP_WAIT | __GFP_IO | __GFP_FS)
 117 #define GFP_TEMPORARY   (__GFP_WAIT | __GFP_IO | __GFP_FS | \
 118                          __GFP_RECLAIMABLE)
 119 #define GFP_USER        (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
 120 #define GFP_HIGHUSER    (GFP_USER | __GFP_HIGHMEM)
 121 #define GFP_HIGHUSER_MOVABLE    (GFP_HIGHUSER | __GFP_MOVABLE)
 122 #define GFP_IOFS        (__GFP_IO | __GFP_FS)
 123 #define GFP_TRANSHUGE   (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
 124                          __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \
 125                          __GFP_NO_KSWAPD)
 126
 127 /* This mask makes up all the page movable related flags */
 128 #define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
 129
 130 /* Control page allocator reclaim behavior */
 131 #define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
 132                         __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
 133                         __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
 134
 135 /* Control slab gfp mask during early boot */
 136 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
 137
 138 /* Control allocation constraints */
 139 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
 140
 141 /* Do not use these with a slab allocator */
 142 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
 143
 144 /* Flag - indicates that the buffer will be suitable for DMA.  Ignored on some
 145    platforms, used as appropriate on others */
 146
 147 #define GFP_DMA         __GFP_DMA
 148
 149 /* 4GB DMA on some platforms */
 150 #define GFP_DMA32       __GFP_DMA32
 151
 152 /* Convert GFP flags to their corresponding migrate type */
 153 static inline int gfpflags_to_migratetype(const gfp_t gfp_flags)
 154 {
 155         WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
 156
 157         if (unlikely(page_group_by_mobility_disabled))
 158                 return MIGRATE_UNMOVABLE;
 159
 160         /* Group based on mobility */
 161         return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
 162                 ((gfp_flags & __GFP_RECLAIMABLE) != 0);
 163 }
 164
 165 #ifdef CONFIG_HIGHMEM
 166 #define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
 167 #else
 168 #define OPT_ZONE_HIGHMEM ZONE_NORMAL
 169 #endif
 170
 171 #ifdef CONFIG_ZONE_DMA
 172 #define OPT_ZONE_DMA ZONE_DMA
 173 #else
 174 #define OPT_ZONE_DMA ZONE_NORMAL
 175 #endif
 176
 177 #ifdef CONFIG_ZONE_DMA32
 178 #define OPT_ZONE_DMA32 ZONE_DMA32
 179 #else
 180 #define OPT_ZONE_DMA32 ZONE_NORMAL
 181 #endif
 182
 183 /*
 184  * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
 185  * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
 186  * and there are 16 of them to cover all possible combinations of
 187  * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
 188  *
 189  * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
 190  * But GFP_MOVABLE is not only a zone specifier but also an allocation
 191  * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
 192  * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
 193  *
 194  *       bit       result
 195  *       =================
 196  *       0x0    => NORMAL
 197  *       0x1    => DMA or NORMAL
 198  *       0x2    => HIGHMEM or NORMAL
 199  *       0x3    => BAD (DMA+HIGHMEM)
 200  *       0x4    => DMA32 or DMA or NORMAL
 201  *       0x5    => BAD (DMA+DMA32)
 202  *       0x6    => BAD (HIGHMEM+DMA32)
 203  *       0x7    => BAD (HIGHMEM+DMA32+DMA)
 204  *       0x8    => NORMAL (MOVABLE+0)
 205  *       0x9    => DMA or NORMAL (MOVABLE+DMA)
 206  *       0xa    => MOVABLE (Movable is valid only if HIGHMEM is set too)
 207  *       0xb    => BAD (MOVABLE+HIGHMEM+DMA)
 208  *       0xc    => DMA32 (MOVABLE+DMA32)
 209  *       0xd    => BAD (MOVABLE+DMA32+DMA)
 210  *       0xe    => BAD (MOVABLE+DMA32+HIGHMEM)
 211  *       0xf    => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
 212  *
 213  * ZONES_SHIFT must be <= 2 on 32 bit platforms.
 214  */
 215
 216 #if 16 * ZONES_SHIFT > BITS_PER_LONG
 217 #error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
 218 #endif
 219
 220 #define GFP_ZONE_TABLE ( \
 221         (ZONE_NORMAL << 0 * ZONES_SHIFT)                                      \
 222         | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT)                          \
 223         | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT)                  \
 224         | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT)                      \
 225         | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT)                       \
 226         | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT)       \
 227         | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT)   \
 228         | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT)   \
 229 )
 230
 231 /*
 232  * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
 233  * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
 234  * entry starting with bit 0. Bit is set if the combination is not
 235  * allowed.
 236  */
 237 #define GFP_ZONE_BAD ( \
 238         1 << (___GFP_DMA | ___GFP_HIGHMEM)                                    \
 239         | 1 << (___GFP_DMA | ___GFP_DMA32)                                    \
 240         | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM)                                \
 241         | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM)                   \
 242         | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA)                 \
 243         | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA)                   \
 244         | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM)               \
 245         | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM)  \
 246 )
 247
 248 static inline enum zone_type gfp_zone(gfp_t flags)
 249 {
 250         enum zone_type z;
 251         int bit = (__force int) (flags & GFP_ZONEMASK);
 252
 253         z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
 254                                          ((1 << ZONES_SHIFT) - 1);
 255         VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
 256         return z;
 257 }
 258
 259 /*
 260  * There is only one page-allocator function, and two main namespaces to
 261  * it. The alloc_page*() variants return 'struct page *' and as such
 262  * can allocate highmem pages, the *get*page*() variants return
 263  * virtual kernel addresses to the allocated page(s).
 264  */
 265
 266 static inline int gfp_zonelist(gfp_t flags)
 267 {
 268         if (IS_ENABLED(CONFIG_NUMA) && unlikely(flags & __GFP_THISNODE))
 269                 return 1;
 270
 271         return 0;
 272 }
 273
 274 /*
 275  * We get the zone list from the current node and the gfp_mask.
 276  * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
 277  * There are two zonelists per node, one for all zones with memory and
 278  * one containing just zones from the node the zonelist belongs to.
 279  *
 280  * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
 281  * optimized to &contig_page_data at compile-time.
 282  */
 283 static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
 284 {
 285         return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
 286 }
 287
 288 #ifndef HAVE_ARCH_FREE_PAGE
 289 static inline void arch_free_page(struct page *page, int order) { }
 290 #endif
 291 #ifndef HAVE_ARCH_ALLOC_PAGE
 292 static inline void arch_alloc_page(struct page *page, int order) { }
 293 #endif
 294
 295 struct page *
 296 __alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
 297                        struct zonelist *zonelist, nodemask_t *nodemask);
 298
 299 static inline struct page *
 300 __alloc_pages(gfp_t gfp_mask, unsigned int order,
 301                 struct zonelist *zonelist)
 302 {
 303         return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
 304 }
 305
 306 static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
 307                                                 unsigned int order)
 308 {
 309         /* Unknown node is current node */
 310         if (nid < 0)
 311                 nid = numa_node_id();
 312
 313         return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
 314 }
 315
 316 static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask,
 317                                                 unsigned int order)
 318 {
 319         VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES || !node_online(nid));
 320
 321         return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
 322 }
 323
 324 #ifdef CONFIG_NUMA
 325 extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
 326
 327 static inline struct page *
 328 alloc_pages(gfp_t gfp_mask, unsigned int order)
 329 {
 330         return alloc_pages_current(gfp_mask, order);
 331 }
 332 extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
 333                         struct vm_area_struct *vma, unsigned long addr,
 334                         int node, bool hugepage);
 335 #define alloc_hugepage_vma(gfp_mask, vma, addr, order)  \
 336         alloc_pages_vma(gfp_mask, order, vma, addr, numa_node_id(), true)
 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, false)\
 341         alloc_pages(gfp_mask, order)
 342 #define alloc_hugepage_vma(gfp_mask, vma, addr, order)  \
 343         alloc_pages(gfp_mask, order)
 344 #endif
 345 #define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
 346 #define alloc_page_vma(gfp_mask, vma, addr)                     \
 347         alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id(), false)
 348 #define alloc_page_vma_node(gfp_mask, vma, addr, node)          \
 349         alloc_pages_vma(gfp_mask, 0, vma, addr, node, false)
 350
 351 extern struct page *alloc_kmem_pages(gfp_t gfp_mask, unsigned int order);
 352 extern struct page *alloc_kmem_pages_node(int nid, gfp_t gfp_mask,
 353                                           unsigned int order);
 354
 355 extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
 356 extern unsigned long get_zeroed_page(gfp_t gfp_mask);
 357
 358 void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
 359 void free_pages_exact(void *virt, size_t size);
 360 /* This is different from alloc_pages_exact_node !!! */
 361 void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask);
 362
 363 #define __get_free_page(gfp_mask) \
 364                 __get_free_pages((gfp_mask), 0)
 365
 366 #define __get_dma_pages(gfp_mask, order) \
 367                 __get_free_pages((gfp_mask) | GFP_DMA, (order))
 368
 369 extern void __free_pages(struct page *page, unsigned int order);
 370 extern void free_pages(unsigned long addr, unsigned int order);
 371 extern void free_hot_cold_page(struct page *page, bool cold);
 372 extern void free_hot_cold_page_list(struct list_head *list, bool cold);
 373
 374 struct page_frag_cache;
 375 extern void *__alloc_page_frag(struct page_frag_cache *nc,
 376                                unsigned int fragsz, gfp_t gfp_mask);
 377 extern void __free_page_frag(void *addr);
 378
 379 extern void __free_kmem_pages(struct page *page, unsigned int order);
 380 extern void free_kmem_pages(unsigned long addr, unsigned int order);
 381
 382 #define __free_page(page) __free_pages((page), 0)
 383 #define free_page(addr) free_pages((addr), 0)
 384
 385 void page_alloc_init(void);
 386 void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
 387 void drain_all_pages(struct zone *zone);
 388 void drain_local_pages(struct zone *zone);
 389
 390 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
 391 void page_alloc_init_late(void);
 392 #else
 393 static inline void page_alloc_init_late(void)
 394 {
 395 }
 396 #endif
 397
 398 /*
 399  * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
 400  * GFP flags are used before interrupts are enabled. Once interrupts are
 401  * enabled, it is set to __GFP_BITS_MASK while the system is running. During
 402  * hibernation, it is used by PM to avoid I/O during memory allocation while
 403  * devices are suspended.
 404  */
 405 extern gfp_t gfp_allowed_mask;
 406
 407 /* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
 408 bool gfp_pfmemalloc_allowed(gfp_t gfp_mask);
 409
 410 extern void pm_restrict_gfp_mask(void);
 411 extern void pm_restore_gfp_mask(void);
 412
 413 #ifdef CONFIG_PM_SLEEP
 414 extern bool pm_suspended_storage(void);
 415 #else
 416 static inline bool pm_suspended_storage(void)
 417 {
 418         return false;
 419 }
 420 #endif /* CONFIG_PM_SLEEP */
 421
 422 #ifdef CONFIG_CMA
 423
 424 /* The below functions must be run on a range from a single zone. */
 425 extern int alloc_contig_range(unsigned long start, unsigned long end,
 426                               unsigned migratetype);
 427 extern void free_contig_range(unsigned long pfn, unsigned nr_pages);
 428
 429 /* CMA stuff */
 430 extern void init_cma_reserved_pageblock(struct page *page);
 431
 432 #endif
 433
 434 #endif /* __LINUX_GFP_H */