LXR linux/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
  10struct 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#ifdef CONFIG_KMEMCHECK
  34#define ___GFP_NOTRACK          0x200000u
  35#else
  36#define ___GFP_NOTRACK          0
  37#endif
  38#define ___GFP_NO_KSWAPD        0x400000u
  39#define ___GFP_OTHER_NODE       0x800000u
  40#define ___GFP_WRITE            0x1000000u
  41
  42/*
  43 * GFP bitmasks..
  44 *
  45 * Zone modifiers (see linux/mmzone.h - low three bits)
  46 *
  47 * Do not put any conditional on these. If necessary modify the definitions
  48 * without the underscores and use them consistently. The definitions here may
  49 * be used in bit comparisons.
  50 */
  51#define __GFP_DMA       ((__force gfp_t)___GFP_DMA)
  52#define __GFP_HIGHMEM   ((__force gfp_t)___GFP_HIGHMEM)
  53#define __GFP_DMA32     ((__force gfp_t)___GFP_DMA32)
  54#define __GFP_MOVABLE   ((__force gfp_t)___GFP_MOVABLE)  /* Page is movable */
  55#define GFP_ZONEMASK    (__GFP_DMA|__GFP_HIGHMEM|__GFP_DMA32|__GFP_MOVABLE)
  56/*
  57 * Action modifiers - doesn't change the zoning
  58 *
  59 * __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
  60 * _might_ fail.  This depends upon the particular VM implementation.
  61 *
  62 * __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
  63 * cannot handle allocation failures.  This modifier is deprecated and no new
  64 * users should be added.
  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_NOTRACK   ((__force gfp_t)___GFP_NOTRACK)  /* Don't track with kmemcheck */
  92
  93#define __GFP_NO_KSWAPD ((__force gfp_t)___GFP_NO_KSWAPD)
  94#define __GFP_OTHER_NODE ((__force gfp_t)___GFP_OTHER_NODE) /* On behalf of other node */
  95#define __GFP_WRITE     ((__force gfp_t)___GFP_WRITE)   /* Allocator intends to dirty page */
  96
  97/*
  98 * This may seem redundant, but it's a way of annotating false positives vs.
  99 * allocations that simply cannot be supported (e.g. page tables).
 100 */
 101#define __GFP_NOTRACK_FALSE_POSITIVE (__GFP_NOTRACK)
 102
 103#define __GFP_BITS_SHIFT 25     /* Room for N __GFP_FOO bits */
 104#define __GFP_BITS_MASK ((__force gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
 105
 106/* This equals 0, but use constants in case they ever change */
 107#define GFP_NOWAIT      (GFP_ATOMIC & ~__GFP_HIGH)
 108/* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */
 109#define GFP_ATOMIC      (__GFP_HIGH)
 110#define GFP_NOIO        (__GFP_WAIT)
 111#define GFP_NOFS        (__GFP_WAIT | __GFP_IO)
 112#define GFP_KERNEL      (__GFP_WAIT | __GFP_IO | __GFP_FS)
 113#define GFP_TEMPORARY   (__GFP_WAIT | __GFP_IO | __GFP_FS | \
 114                         __GFP_RECLAIMABLE)
 115#define GFP_USER        (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL)
 116#define GFP_HIGHUSER    (__GFP_WAIT | __GFP_IO | __GFP_FS | __GFP_HARDWALL | \
 117                         __GFP_HIGHMEM)
 118#define GFP_HIGHUSER_MOVABLE    (__GFP_WAIT | __GFP_IO | __GFP_FS | \
 119                                 __GFP_HARDWALL | __GFP_HIGHMEM | \
 120                                 __GFP_MOVABLE)
 121#define GFP_IOFS        (__GFP_IO | __GFP_FS)
 122#define GFP_TRANSHUGE   (GFP_HIGHUSER_MOVABLE | __GFP_COMP | \
 123                         __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN | \
 124                         __GFP_NO_KSWAPD)
 125
 126#ifdef CONFIG_NUMA
 127#define GFP_THISNODE    (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY)
 128#else
 129#define GFP_THISNODE    ((__force gfp_t)0)
 130#endif
 131
 132/* This mask makes up all the page movable related flags */
 133#define GFP_MOVABLE_MASK (__GFP_RECLAIMABLE|__GFP_MOVABLE)
 134
 135/* Control page allocator reclaim behavior */
 136#define GFP_RECLAIM_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_FS|\
 137                        __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
 138                        __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
 139
 140/* Control slab gfp mask during early boot */
 141#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_WAIT|__GFP_IO|__GFP_FS))
 142
 143/* Control allocation constraints */
 144#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
 145
 146/* Do not use these with a slab allocator */
 147#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
 148
 149/* Flag - indicates that the buffer will be suitable for DMA.  Ignored on some
 150   platforms, used as appropriate on others */
 151
 152#define GFP_DMA         __GFP_DMA
 153
 154/* 4GB DMA on some platforms */
 155#define GFP_DMA32       __GFP_DMA32
 156
 157/* Convert GFP flags to their corresponding migrate type */
 158static inline int allocflags_to_migratetype(gfp_t gfp_flags)
 159{
 160        WARN_ON((gfp_flags & GFP_MOVABLE_MASK) == GFP_MOVABLE_MASK);
 161
 162        if (unlikely(page_group_by_mobility_disabled))
 163                return MIGRATE_UNMOVABLE;
 164
 165        /* Group based on mobility */
 166        return (((gfp_flags & __GFP_MOVABLE) != 0) << 1) |
 167                ((gfp_flags & __GFP_RECLAIMABLE) != 0);
 168}
 169
 170#ifdef CONFIG_HIGHMEM
 171#define OPT_ZONE_HIGHMEM ZONE_HIGHMEM
 172#else
 173#define OPT_ZONE_HIGHMEM ZONE_NORMAL
 174#endif
 175
 176#ifdef CONFIG_ZONE_DMA
 177#define OPT_ZONE_DMA ZONE_DMA
 178#else
 179#define OPT_ZONE_DMA ZONE_NORMAL
 180#endif
 181
 182#ifdef CONFIG_ZONE_DMA32
 183#define OPT_ZONE_DMA32 ZONE_DMA32
 184#else
 185#define OPT_ZONE_DMA32 ZONE_NORMAL
 186#endif
 187
 188/*
 189 * GFP_ZONE_TABLE is a word size bitstring that is used for looking up the
 190 * zone to use given the lowest 4 bits of gfp_t. Entries are ZONE_SHIFT long
 191 * and there are 16 of them to cover all possible combinations of
 192 * __GFP_DMA, __GFP_DMA32, __GFP_MOVABLE and __GFP_HIGHMEM.
 193 *
 194 * The zone fallback order is MOVABLE=>HIGHMEM=>NORMAL=>DMA32=>DMA.
 195 * But GFP_MOVABLE is not only a zone specifier but also an allocation
 196 * policy. Therefore __GFP_MOVABLE plus another zone selector is valid.
 197 * Only 1 bit of the lowest 3 bits (DMA,DMA32,HIGHMEM) can be set to "1".
 198 *
 199 *       bit       result
 200 *       =================
 201 *       0x0    => NORMAL
 202 *       0x1    => DMA or NORMAL
 203 *       0x2    => HIGHMEM or NORMAL
 204 *       0x3    => BAD (DMA+HIGHMEM)
 205 *       0x4    => DMA32 or DMA or NORMAL
 206 *       0x5    => BAD (DMA+DMA32)
 207 *       0x6    => BAD (HIGHMEM+DMA32)
 208 *       0x7    => BAD (HIGHMEM+DMA32+DMA)
 209 *       0x8    => NORMAL (MOVABLE+0)
 210 *       0x9    => DMA or NORMAL (MOVABLE+DMA)
 211 *       0xa    => MOVABLE (Movable is valid only if HIGHMEM is set too)
 212 *       0xb    => BAD (MOVABLE+HIGHMEM+DMA)
 213 *       0xc    => DMA32 (MOVABLE+HIGHMEM+DMA32)
 214 *       0xd    => BAD (MOVABLE+DMA32+DMA)
 215 *       0xe    => BAD (MOVABLE+DMA32+HIGHMEM)
 216 *       0xf    => BAD (MOVABLE+DMA32+HIGHMEM+DMA)
 217 *
 218 * ZONES_SHIFT must be <= 2 on 32 bit platforms.
 219 */
 220
 221#if 16 * ZONES_SHIFT > BITS_PER_LONG
 222#error ZONES_SHIFT too large to create GFP_ZONE_TABLE integer
 223#endif
 224
 225#define GFP_ZONE_TABLE ( \
 226        (ZONE_NORMAL << 0 * ZONES_SHIFT)                                      \
 227        | (OPT_ZONE_DMA << ___GFP_DMA * ZONES_SHIFT)                          \
 228        | (OPT_ZONE_HIGHMEM << ___GFP_HIGHMEM * ZONES_SHIFT)                  \
 229        | (OPT_ZONE_DMA32 << ___GFP_DMA32 * ZONES_SHIFT)                      \
 230        | (ZONE_NORMAL << ___GFP_MOVABLE * ZONES_SHIFT)                       \
 231        | (OPT_ZONE_DMA << (___GFP_MOVABLE | ___GFP_DMA) * ZONES_SHIFT)       \
 232        | (ZONE_MOVABLE << (___GFP_MOVABLE | ___GFP_HIGHMEM) * ZONES_SHIFT)   \
 233        | (OPT_ZONE_DMA32 << (___GFP_MOVABLE | ___GFP_DMA32) * ZONES_SHIFT)   \
 234)
 235
 236/*
 237 * GFP_ZONE_BAD is a bitmap for all combinations of __GFP_DMA, __GFP_DMA32
 238 * __GFP_HIGHMEM and __GFP_MOVABLE that are not permitted. One flag per
 239 * entry starting with bit 0. Bit is set if the combination is not
 240 * allowed.
 241 */
 242#define GFP_ZONE_BAD ( \
 243        1 << (___GFP_DMA | ___GFP_HIGHMEM)                                    \
 244        | 1 << (___GFP_DMA | ___GFP_DMA32)                                    \
 245        | 1 << (___GFP_DMA32 | ___GFP_HIGHMEM)                                \
 246        | 1 << (___GFP_DMA | ___GFP_DMA32 | ___GFP_HIGHMEM)                   \
 247        | 1 << (___GFP_MOVABLE | ___GFP_HIGHMEM | ___GFP_DMA)                 \
 248        | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA)                   \
 249        | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_HIGHMEM)               \
 250        | 1 << (___GFP_MOVABLE | ___GFP_DMA32 | ___GFP_DMA | ___GFP_HIGHMEM)  \
 251)
 252
 253static inline enum zone_type gfp_zone(gfp_t flags)
 254{
 255        enum zone_type z;
 256        int bit = (__force int) (flags & GFP_ZONEMASK);
 257
 258        z = (GFP_ZONE_TABLE >> (bit * ZONES_SHIFT)) &
 259                                         ((1 << ZONES_SHIFT) - 1);
 260        VM_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
 261        return z;
 262}
 263
 264/*
 265 * There is only one page-allocator function, and two main namespaces to
 266 * it. The alloc_page*() variants return 'struct page *' and as such
 267 * can allocate highmem pages, the *get*page*() variants return
 268 * virtual kernel addresses to the allocated page(s).
 269 */
 270
 271static inline int gfp_zonelist(gfp_t flags)
 272{
 273        if (NUMA_BUILD && unlikely(flags & __GFP_THISNODE))
 274                return 1;
 275
 276        return 0;
 277}
 278
 279/*
 280 * We get the zone list from the current node and the gfp_mask.
 281 * This zone list contains a maximum of MAXNODES*MAX_NR_ZONES zones.
 282 * There are two zonelists per node, one for all zones with memory and
 283 * one containing just zones from the node the zonelist belongs to.
 284 *
 285 * For the normal case of non-DISCONTIGMEM systems the NODE_DATA() gets
 286 * optimized to &contig_page_data at compile-time.
 287 */
 288static inline struct zonelist *node_zonelist(int nid, gfp_t flags)
 289{
 290        return NODE_DATA(nid)->node_zonelists + gfp_zonelist(flags);
 291}
 292
 293#ifndef HAVE_ARCH_FREE_PAGE
 294static inline void arch_free_page(struct page *page, int order) { }
 295#endif
 296#ifndef HAVE_ARCH_ALLOC_PAGE
 297static inline void arch_alloc_page(struct page *page, int order) { }
 298#endif
 299
 300struct page *
 301__alloc_pages_nodemask(gfp_t gfp_mask, unsigned int order,
 302                       struct zonelist *zonelist, nodemask_t *nodemask);
 303
 304static inline struct page *
 305__alloc_pages(gfp_t gfp_mask, unsigned int order,
 306                struct zonelist *zonelist)
 307{
 308        return __alloc_pages_nodemask(gfp_mask, order, zonelist, NULL);
 309}
 310
 311static inline struct page *alloc_pages_node(int nid, gfp_t gfp_mask,
 312                                                unsigned int order)
 313{
 314        /* Unknown node is current node */
 315        if (nid < 0)
 316                nid = numa_node_id();
 317
 318        return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
 319}
 320
 321static inline struct page *alloc_pages_exact_node(int nid, gfp_t gfp_mask,
 322                                                unsigned int order)
 323{
 324        VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES || !node_online(nid));
 325
 326        return __alloc_pages(gfp_mask, order, node_zonelist(nid, gfp_mask));
 327}
 328
 329#ifdef CONFIG_NUMA
 330extern struct page *alloc_pages_current(gfp_t gfp_mask, unsigned order);
 331
 332static inline struct page *
 333alloc_pages(gfp_t gfp_mask, unsigned int order)
 334{
 335        return alloc_pages_current(gfp_mask, order);
 336}
 337extern struct page *alloc_pages_vma(gfp_t gfp_mask, int order,
 338                        struct vm_area_struct *vma, unsigned long addr,
 339                        int node);
 340#else
 341#define alloc_pages(gfp_mask, order) \
 342                alloc_pages_node(numa_node_id(), gfp_mask, order)
 343#define alloc_pages_vma(gfp_mask, order, vma, addr, node)       \
 344        alloc_pages(gfp_mask, order)
 345#endif
 346#define alloc_page(gfp_mask) alloc_pages(gfp_mask, 0)
 347#define alloc_page_vma(gfp_mask, vma, addr)                     \
 348        alloc_pages_vma(gfp_mask, 0, vma, addr, numa_node_id())
 349#define alloc_page_vma_node(gfp_mask, vma, addr, node)          \
 350        alloc_pages_vma(gfp_mask, 0, vma, addr, node)
 351
 352extern unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order);
 353extern unsigned long get_zeroed_page(gfp_t gfp_mask);
 354
 355void *alloc_pages_exact(size_t size, gfp_t gfp_mask);
 356void free_pages_exact(void *virt, size_t size);
 357/* This is different from alloc_pages_exact_node !!! */
 358void *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
 366extern void __free_pages(struct page *page, unsigned int order);
 367extern void free_pages(unsigned long addr, unsigned int order);
 368extern void free_hot_cold_page(struct page *page, int cold);
 369extern void free_hot_cold_page_list(struct list_head *list, int cold);
 370
 371#define __free_page(page) __free_pages((page), 0)
 372#define free_page(addr) free_pages((addr), 0)
 373
 374void page_alloc_init(void);
 375void drain_zone_pages(struct zone *zone, struct per_cpu_pages *pcp);
 376void drain_all_pages(void);
 377void drain_local_pages(void *dummy);
 378
 379/*
 380 * gfp_allowed_mask is set to GFP_BOOT_MASK during early boot to restrict what
 381 * GFP flags are used before interrupts are enabled. Once interrupts are
 382 * enabled, it is set to __GFP_BITS_MASK while the system is running. During
 383 * hibernation, it is used by PM to avoid I/O during memory allocation while
 384 * devices are suspended.
 385 */
 386extern gfp_t gfp_allowed_mask;
 387
 388/* Returns true if the gfp_mask allows use of ALLOC_NO_WATERMARK */
 389bool gfp_pfmemalloc_allowed(gfp_t gfp_mask);
 390
 391extern void pm_restrict_gfp_mask(void);
 392extern void pm_restore_gfp_mask(void);
 393
 394#ifdef CONFIG_PM_SLEEP
 395extern bool pm_suspended_storage(void);
 396#else
 397static inline bool pm_suspended_storage(void)
 398{
 399        return false;
 400}
 401#endif /* CONFIG_PM_SLEEP */
 402
 403#ifdef CONFIG_CMA
 404
 405/* The below functions must be run on a range from a single zone. */
 406extern int alloc_contig_range(unsigned long start, unsigned long end,
 407                              unsigned migratetype);
 408extern void free_contig_range(unsigned long pfn, unsigned nr_pages);
 409
 410/* CMA stuff */
 411extern void init_cma_reserved_pageblock(struct page *page);
 412
 413#endif
 414
 415#endif /* __LINUX_GFP_H */
 416