linux/include/linux/mmzone.h
<<
>>
Prefs
   1#ifndef _LINUX_MMZONE_H
   2#define _LINUX_MMZONE_H
   3
   4#ifndef __ASSEMBLY__
   5#ifndef __GENERATING_BOUNDS_H
   6
   7#include <linux/spinlock.h>
   8#include <linux/list.h>
   9#include <linux/wait.h>
  10#include <linux/bitops.h>
  11#include <linux/cache.h>
  12#include <linux/threads.h>
  13#include <linux/numa.h>
  14#include <linux/init.h>
  15#include <linux/seqlock.h>
  16#include <linux/nodemask.h>
  17#include <linux/pageblock-flags.h>
  18#include <generated/bounds.h>
  19#include <linux/atomic.h>
  20#include <asm/page.h>
  21
  22/* Free memory management - zoned buddy allocator.  */
  23#ifndef CONFIG_FORCE_MAX_ZONEORDER
  24#define MAX_ORDER 11
  25#else
  26#define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
  27#endif
  28#define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
  29
  30/*
  31 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
  32 * costly to service.  That is between allocation orders which should
  33 * coalesce naturally under reasonable reclaim pressure and those which
  34 * will not.
  35 */
  36#define PAGE_ALLOC_COSTLY_ORDER 3
  37
  38enum {
  39        MIGRATE_UNMOVABLE,
  40        MIGRATE_RECLAIMABLE,
  41        MIGRATE_MOVABLE,
  42        MIGRATE_PCPTYPES,       /* the number of types on the pcp lists */
  43        MIGRATE_RESERVE = MIGRATE_PCPTYPES,
  44#ifdef CONFIG_CMA
  45        /*
  46         * MIGRATE_CMA migration type is designed to mimic the way
  47         * ZONE_MOVABLE works.  Only movable pages can be allocated
  48         * from MIGRATE_CMA pageblocks and page allocator never
  49         * implicitly change migration type of MIGRATE_CMA pageblock.
  50         *
  51         * The way to use it is to change migratetype of a range of
  52         * pageblocks to MIGRATE_CMA which can be done by
  53         * __free_pageblock_cma() function.  What is important though
  54         * is that a range of pageblocks must be aligned to
  55         * MAX_ORDER_NR_PAGES should biggest page be bigger then
  56         * a single pageblock.
  57         */
  58        MIGRATE_CMA,
  59#endif
  60        MIGRATE_ISOLATE,        /* can't allocate from here */
  61        MIGRATE_TYPES
  62};
  63
  64#ifdef CONFIG_CMA
  65#  define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
  66#  define cma_wmark_pages(zone) zone->min_cma_pages
  67#else
  68#  define is_migrate_cma(migratetype) false
  69#  define cma_wmark_pages(zone) 0
  70#endif
  71
  72#define for_each_migratetype_order(order, type) \
  73        for (order = 0; order < MAX_ORDER; order++) \
  74                for (type = 0; type < MIGRATE_TYPES; type++)
  75
  76extern int page_group_by_mobility_disabled;
  77
  78static inline int get_pageblock_migratetype(struct page *page)
  79{
  80        return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
  81}
  82
  83struct free_area {
  84        struct list_head        free_list[MIGRATE_TYPES];
  85        unsigned long           nr_free;
  86};
  87
  88struct pglist_data;
  89
  90/*
  91 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
  92 * So add a wild amount of padding here to ensure that they fall into separate
  93 * cachelines.  There are very few zone structures in the machine, so space
  94 * consumption is not a concern here.
  95 */
  96#if defined(CONFIG_SMP)
  97struct zone_padding {
  98        char x[0];
  99} ____cacheline_internodealigned_in_smp;
 100#define ZONE_PADDING(name)      struct zone_padding name;
 101#else
 102#define ZONE_PADDING(name)
 103#endif
 104
 105enum zone_stat_item {
 106        /* First 128 byte cacheline (assuming 64 bit words) */
 107        NR_FREE_PAGES,
 108        NR_LRU_BASE,
 109        NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
 110        NR_ACTIVE_ANON,         /*  "     "     "   "       "         */
 111        NR_INACTIVE_FILE,       /*  "     "     "   "       "         */
 112        NR_ACTIVE_FILE,         /*  "     "     "   "       "         */
 113        NR_UNEVICTABLE,         /*  "     "     "   "       "         */
 114        NR_MLOCK,               /* mlock()ed pages found and moved off LRU */
 115        NR_ANON_PAGES,  /* Mapped anonymous pages */
 116        NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
 117                           only modified from process context */
 118        NR_FILE_PAGES,
 119        NR_FILE_DIRTY,
 120        NR_WRITEBACK,
 121        NR_SLAB_RECLAIMABLE,
 122        NR_SLAB_UNRECLAIMABLE,
 123        NR_PAGETABLE,           /* used for pagetables */
 124        NR_KERNEL_STACK,
 125        /* Second 128 byte cacheline */
 126        NR_UNSTABLE_NFS,        /* NFS unstable pages */
 127        NR_BOUNCE,
 128        NR_VMSCAN_WRITE,
 129        NR_VMSCAN_IMMEDIATE,    /* Prioritise for reclaim when writeback ends */
 130        NR_WRITEBACK_TEMP,      /* Writeback using temporary buffers */
 131        NR_ISOLATED_ANON,       /* Temporary isolated pages from anon lru */
 132        NR_ISOLATED_FILE,       /* Temporary isolated pages from file lru */
 133        NR_SHMEM,               /* shmem pages (included tmpfs/GEM pages) */
 134        NR_DIRTIED,             /* page dirtyings since bootup */
 135        NR_WRITTEN,             /* page writings since bootup */
 136#ifdef CONFIG_NUMA
 137        NUMA_HIT,               /* allocated in intended node */
 138        NUMA_MISS,              /* allocated in non intended node */
 139        NUMA_FOREIGN,           /* was intended here, hit elsewhere */
 140        NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
 141        NUMA_LOCAL,             /* allocation from local node */
 142        NUMA_OTHER,             /* allocation from other node */
 143#endif
 144        NR_ANON_TRANSPARENT_HUGEPAGES,
 145        NR_VM_ZONE_STAT_ITEMS };
 146
 147/*
 148 * We do arithmetic on the LRU lists in various places in the code,
 149 * so it is important to keep the active lists LRU_ACTIVE higher in
 150 * the array than the corresponding inactive lists, and to keep
 151 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
 152 *
 153 * This has to be kept in sync with the statistics in zone_stat_item
 154 * above and the descriptions in vmstat_text in mm/vmstat.c
 155 */
 156#define LRU_BASE 0
 157#define LRU_ACTIVE 1
 158#define LRU_FILE 2
 159
 160enum lru_list {
 161        LRU_INACTIVE_ANON = LRU_BASE,
 162        LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
 163        LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
 164        LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
 165        LRU_UNEVICTABLE,
 166        NR_LRU_LISTS
 167};
 168
 169#define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
 170
 171#define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
 172
 173static inline int is_file_lru(enum lru_list lru)
 174{
 175        return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
 176}
 177
 178static inline int is_active_lru(enum lru_list lru)
 179{
 180        return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
 181}
 182
 183static inline int is_unevictable_lru(enum lru_list lru)
 184{
 185        return (lru == LRU_UNEVICTABLE);
 186}
 187
 188struct zone_reclaim_stat {
 189        /*
 190         * The pageout code in vmscan.c keeps track of how many of the
 191         * mem/swap backed and file backed pages are referenced.
 192         * The higher the rotated/scanned ratio, the more valuable
 193         * that cache is.
 194         *
 195         * The anon LRU stats live in [0], file LRU stats in [1]
 196         */
 197        unsigned long           recent_rotated[2];
 198        unsigned long           recent_scanned[2];
 199};
 200
 201struct lruvec {
 202        struct list_head lists[NR_LRU_LISTS];
 203        struct zone_reclaim_stat reclaim_stat;
 204#ifdef CONFIG_MEMCG
 205        struct zone *zone;
 206#endif
 207};
 208
 209/* Mask used at gathering information at once (see memcontrol.c) */
 210#define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
 211#define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
 212#define LRU_ALL      ((1 << NR_LRU_LISTS) - 1)
 213
 214/* Isolate clean file */
 215#define ISOLATE_CLEAN           ((__force isolate_mode_t)0x1)
 216/* Isolate unmapped file */
 217#define ISOLATE_UNMAPPED        ((__force isolate_mode_t)0x2)
 218/* Isolate for asynchronous migration */
 219#define ISOLATE_ASYNC_MIGRATE   ((__force isolate_mode_t)0x4)
 220
 221/* LRU Isolation modes. */
 222typedef unsigned __bitwise__ isolate_mode_t;
 223
 224enum zone_watermarks {
 225        WMARK_MIN,
 226        WMARK_LOW,
 227        WMARK_HIGH,
 228        NR_WMARK
 229};
 230
 231#define min_wmark_pages(z) (z->watermark[WMARK_MIN])
 232#define low_wmark_pages(z) (z->watermark[WMARK_LOW])
 233#define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
 234
 235struct per_cpu_pages {
 236        int count;              /* number of pages in the list */
 237        int high;               /* high watermark, emptying needed */
 238        int batch;              /* chunk size for buddy add/remove */
 239
 240        /* Lists of pages, one per migrate type stored on the pcp-lists */
 241        struct list_head lists[MIGRATE_PCPTYPES];
 242};
 243
 244struct per_cpu_pageset {
 245        struct per_cpu_pages pcp;
 246#ifdef CONFIG_NUMA
 247        s8 expire;
 248#endif
 249#ifdef CONFIG_SMP
 250        s8 stat_threshold;
 251        s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
 252#endif
 253};
 254
 255#endif /* !__GENERATING_BOUNDS.H */
 256
 257enum zone_type {
 258#ifdef CONFIG_ZONE_DMA
 259        /*
 260         * ZONE_DMA is used when there are devices that are not able
 261         * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
 262         * carve out the portion of memory that is needed for these devices.
 263         * The range is arch specific.
 264         *
 265         * Some examples
 266         *
 267         * Architecture         Limit
 268         * ---------------------------
 269         * parisc, ia64, sparc  <4G
 270         * s390                 <2G
 271         * arm                  Various
 272         * alpha                Unlimited or 0-16MB.
 273         *
 274         * i386, x86_64 and multiple other arches
 275         *                      <16M.
 276         */
 277        ZONE_DMA,
 278#endif
 279#ifdef CONFIG_ZONE_DMA32
 280        /*
 281         * x86_64 needs two ZONE_DMAs because it supports devices that are
 282         * only able to do DMA to the lower 16M but also 32 bit devices that
 283         * can only do DMA areas below 4G.
 284         */
 285        ZONE_DMA32,
 286#endif
 287        /*
 288         * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
 289         * performed on pages in ZONE_NORMAL if the DMA devices support
 290         * transfers to all addressable memory.
 291         */
 292        ZONE_NORMAL,
 293#ifdef CONFIG_HIGHMEM
 294        /*
 295         * A memory area that is only addressable by the kernel through
 296         * mapping portions into its own address space. This is for example
 297         * used by i386 to allow the kernel to address the memory beyond
 298         * 900MB. The kernel will set up special mappings (page
 299         * table entries on i386) for each page that the kernel needs to
 300         * access.
 301         */
 302        ZONE_HIGHMEM,
 303#endif
 304        ZONE_MOVABLE,
 305        __MAX_NR_ZONES
 306};
 307
 308#ifndef __GENERATING_BOUNDS_H
 309
 310/*
 311 * When a memory allocation must conform to specific limitations (such
 312 * as being suitable for DMA) the caller will pass in hints to the
 313 * allocator in the gfp_mask, in the zone modifier bits.  These bits
 314 * are used to select a priority ordered list of memory zones which
 315 * match the requested limits. See gfp_zone() in include/linux/gfp.h
 316 */
 317
 318#if MAX_NR_ZONES < 2
 319#define ZONES_SHIFT 0
 320#elif MAX_NR_ZONES <= 2
 321#define ZONES_SHIFT 1
 322#elif MAX_NR_ZONES <= 4
 323#define ZONES_SHIFT 2
 324#else
 325#error ZONES_SHIFT -- too many zones configured adjust calculation
 326#endif
 327
 328struct zone {
 329        /* Fields commonly accessed by the page allocator */
 330
 331        /* zone watermarks, access with *_wmark_pages(zone) macros */
 332        unsigned long watermark[NR_WMARK];
 333
 334        /*
 335         * When free pages are below this point, additional steps are taken
 336         * when reading the number of free pages to avoid per-cpu counter
 337         * drift allowing watermarks to be breached
 338         */
 339        unsigned long percpu_drift_mark;
 340
 341        /*
 342         * We don't know if the memory that we're going to allocate will be freeable
 343         * or/and it will be released eventually, so to avoid totally wasting several
 344         * GB of ram we must reserve some of the lower zone memory (otherwise we risk
 345         * to run OOM on the lower zones despite there's tons of freeable ram
 346         * on the higher zones). This array is recalculated at runtime if the
 347         * sysctl_lowmem_reserve_ratio sysctl changes.
 348         */
 349        unsigned long           lowmem_reserve[MAX_NR_ZONES];
 350
 351        /*
 352         * This is a per-zone reserve of pages that should not be
 353         * considered dirtyable memory.
 354         */
 355        unsigned long           dirty_balance_reserve;
 356
 357#ifdef CONFIG_NUMA
 358        int node;
 359        /*
 360         * zone reclaim becomes active if more unmapped pages exist.
 361         */
 362        unsigned long           min_unmapped_pages;
 363        unsigned long           min_slab_pages;
 364#endif
 365        struct per_cpu_pageset __percpu *pageset;
 366        /*
 367         * free areas of different sizes
 368         */
 369        spinlock_t              lock;
 370        int                     all_unreclaimable; /* All pages pinned */
 371#if defined CONFIG_COMPACTION || defined CONFIG_CMA
 372        /* pfn where the last incremental compaction isolated free pages */
 373        unsigned long           compact_cached_free_pfn;
 374#endif
 375#ifdef CONFIG_MEMORY_HOTPLUG
 376        /* see spanned/present_pages for more description */
 377        seqlock_t               span_seqlock;
 378#endif
 379#ifdef CONFIG_CMA
 380        /*
 381         * CMA needs to increase watermark levels during the allocation
 382         * process to make sure that the system is not starved.
 383         */
 384        unsigned long           min_cma_pages;
 385#endif
 386        struct free_area        free_area[MAX_ORDER];
 387
 388#ifndef CONFIG_SPARSEMEM
 389        /*
 390         * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
 391         * In SPARSEMEM, this map is stored in struct mem_section
 392         */
 393        unsigned long           *pageblock_flags;
 394#endif /* CONFIG_SPARSEMEM */
 395
 396#ifdef CONFIG_COMPACTION
 397        /*
 398         * On compaction failure, 1<<compact_defer_shift compactions
 399         * are skipped before trying again. The number attempted since
 400         * last failure is tracked with compact_considered.
 401         */
 402        unsigned int            compact_considered;
 403        unsigned int            compact_defer_shift;
 404        int                     compact_order_failed;
 405#endif
 406
 407        ZONE_PADDING(_pad1_)
 408
 409        /* Fields commonly accessed by the page reclaim scanner */
 410        spinlock_t              lru_lock;
 411        struct lruvec           lruvec;
 412
 413        unsigned long           pages_scanned;     /* since last reclaim */
 414        unsigned long           flags;             /* zone flags, see below */
 415
 416        /* Zone statistics */
 417        atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
 418
 419        /*
 420         * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
 421         * this zone's LRU.  Maintained by the pageout code.
 422         */
 423        unsigned int inactive_ratio;
 424
 425
 426        ZONE_PADDING(_pad2_)
 427        /* Rarely used or read-mostly fields */
 428
 429        /*
 430         * wait_table           -- the array holding the hash table
 431         * wait_table_hash_nr_entries   -- the size of the hash table array
 432         * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
 433         *
 434         * The purpose of all these is to keep track of the people
 435         * waiting for a page to become available and make them
 436         * runnable again when possible. The trouble is that this
 437         * consumes a lot of space, especially when so few things
 438         * wait on pages at a given time. So instead of using
 439         * per-page waitqueues, we use a waitqueue hash table.
 440         *
 441         * The bucket discipline is to sleep on the same queue when
 442         * colliding and wake all in that wait queue when removing.
 443         * When something wakes, it must check to be sure its page is
 444         * truly available, a la thundering herd. The cost of a
 445         * collision is great, but given the expected load of the
 446         * table, they should be so rare as to be outweighed by the
 447         * benefits from the saved space.
 448         *
 449         * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
 450         * primary users of these fields, and in mm/page_alloc.c
 451         * free_area_init_core() performs the initialization of them.
 452         */
 453        wait_queue_head_t       * wait_table;
 454        unsigned long           wait_table_hash_nr_entries;
 455        unsigned long           wait_table_bits;
 456
 457        /*
 458         * Discontig memory support fields.
 459         */
 460        struct pglist_data      *zone_pgdat;
 461        /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
 462        unsigned long           zone_start_pfn;
 463
 464        /*
 465         * zone_start_pfn, spanned_pages and present_pages are all
 466         * protected by span_seqlock.  It is a seqlock because it has
 467         * to be read outside of zone->lock, and it is done in the main
 468         * allocator path.  But, it is written quite infrequently.
 469         *
 470         * The lock is declared along with zone->lock because it is
 471         * frequently read in proximity to zone->lock.  It's good to
 472         * give them a chance of being in the same cacheline.
 473         */
 474        unsigned long           spanned_pages;  /* total size, including holes */
 475        unsigned long           present_pages;  /* amount of memory (excluding holes) */
 476
 477        /*
 478         * rarely used fields:
 479         */
 480        const char              *name;
 481#ifdef CONFIG_MEMORY_ISOLATION
 482        /*
 483         * the number of MIGRATE_ISOLATE *pageblock*.
 484         * We need this for free page counting. Look at zone_watermark_ok_safe.
 485         * It's protected by zone->lock
 486         */
 487        int             nr_pageblock_isolate;
 488#endif
 489} ____cacheline_internodealigned_in_smp;
 490
 491typedef enum {
 492        ZONE_RECLAIM_LOCKED,            /* prevents concurrent reclaim */
 493        ZONE_OOM_LOCKED,                /* zone is in OOM killer zonelist */
 494        ZONE_CONGESTED,                 /* zone has many dirty pages backed by
 495                                         * a congested BDI
 496                                         */
 497} zone_flags_t;
 498
 499static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
 500{
 501        set_bit(flag, &zone->flags);
 502}
 503
 504static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
 505{
 506        return test_and_set_bit(flag, &zone->flags);
 507}
 508
 509static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
 510{
 511        clear_bit(flag, &zone->flags);
 512}
 513
 514static inline int zone_is_reclaim_congested(const struct zone *zone)
 515{
 516        return test_bit(ZONE_CONGESTED, &zone->flags);
 517}
 518
 519static inline int zone_is_reclaim_locked(const struct zone *zone)
 520{
 521        return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
 522}
 523
 524static inline int zone_is_oom_locked(const struct zone *zone)
 525{
 526        return test_bit(ZONE_OOM_LOCKED, &zone->flags);
 527}
 528
 529/*
 530 * The "priority" of VM scanning is how much of the queues we will scan in one
 531 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
 532 * queues ("queue_length >> 12") during an aging round.
 533 */
 534#define DEF_PRIORITY 12
 535
 536/* Maximum number of zones on a zonelist */
 537#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
 538
 539#ifdef CONFIG_NUMA
 540
 541/*
 542 * The NUMA zonelists are doubled because we need zonelists that restrict the
 543 * allocations to a single node for GFP_THISNODE.
 544 *
 545 * [0]  : Zonelist with fallback
 546 * [1]  : No fallback (GFP_THISNODE)
 547 */
 548#define MAX_ZONELISTS 2
 549
 550
 551/*
 552 * We cache key information from each zonelist for smaller cache
 553 * footprint when scanning for free pages in get_page_from_freelist().
 554 *
 555 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
 556 *    up short of free memory since the last time (last_fullzone_zap)
 557 *    we zero'd fullzones.
 558 * 2) The array z_to_n[] maps each zone in the zonelist to its node
 559 *    id, so that we can efficiently evaluate whether that node is
 560 *    set in the current tasks mems_allowed.
 561 *
 562 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
 563 * indexed by a zones offset in the zonelist zones[] array.
 564 *
 565 * The get_page_from_freelist() routine does two scans.  During the
 566 * first scan, we skip zones whose corresponding bit in 'fullzones'
 567 * is set or whose corresponding node in current->mems_allowed (which
 568 * comes from cpusets) is not set.  During the second scan, we bypass
 569 * this zonelist_cache, to ensure we look methodically at each zone.
 570 *
 571 * Once per second, we zero out (zap) fullzones, forcing us to
 572 * reconsider nodes that might have regained more free memory.
 573 * The field last_full_zap is the time we last zapped fullzones.
 574 *
 575 * This mechanism reduces the amount of time we waste repeatedly
 576 * reexaming zones for free memory when they just came up low on
 577 * memory momentarilly ago.
 578 *
 579 * The zonelist_cache struct members logically belong in struct
 580 * zonelist.  However, the mempolicy zonelists constructed for
 581 * MPOL_BIND are intentionally variable length (and usually much
 582 * shorter).  A general purpose mechanism for handling structs with
 583 * multiple variable length members is more mechanism than we want
 584 * here.  We resort to some special case hackery instead.
 585 *
 586 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
 587 * part because they are shorter), so we put the fixed length stuff
 588 * at the front of the zonelist struct, ending in a variable length
 589 * zones[], as is needed by MPOL_BIND.
 590 *
 591 * Then we put the optional zonelist cache on the end of the zonelist
 592 * struct.  This optional stuff is found by a 'zlcache_ptr' pointer in
 593 * the fixed length portion at the front of the struct.  This pointer
 594 * both enables us to find the zonelist cache, and in the case of
 595 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
 596 * to know that the zonelist cache is not there.
 597 *
 598 * The end result is that struct zonelists come in two flavors:
 599 *  1) The full, fixed length version, shown below, and
 600 *  2) The custom zonelists for MPOL_BIND.
 601 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
 602 *
 603 * Even though there may be multiple CPU cores on a node modifying
 604 * fullzones or last_full_zap in the same zonelist_cache at the same
 605 * time, we don't lock it.  This is just hint data - if it is wrong now
 606 * and then, the allocator will still function, perhaps a bit slower.
 607 */
 608
 609
 610struct zonelist_cache {
 611        unsigned short z_to_n[MAX_ZONES_PER_ZONELIST];          /* zone->nid */
 612        DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST);      /* zone full? */
 613        unsigned long last_full_zap;            /* when last zap'd (jiffies) */
 614};
 615#else
 616#define MAX_ZONELISTS 1
 617struct zonelist_cache;
 618#endif
 619
 620/*
 621 * This struct contains information about a zone in a zonelist. It is stored
 622 * here to avoid dereferences into large structures and lookups of tables
 623 */
 624struct zoneref {
 625        struct zone *zone;      /* Pointer to actual zone */
 626        int zone_idx;           /* zone_idx(zoneref->zone) */
 627};
 628
 629/*
 630 * One allocation request operates on a zonelist. A zonelist
 631 * is a list of zones, the first one is the 'goal' of the
 632 * allocation, the other zones are fallback zones, in decreasing
 633 * priority.
 634 *
 635 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
 636 * as explained above.  If zlcache_ptr is NULL, there is no zlcache.
 637 * *
 638 * To speed the reading of the zonelist, the zonerefs contain the zone index
 639 * of the entry being read. Helper functions to access information given
 640 * a struct zoneref are
 641 *
 642 * zonelist_zone()      - Return the struct zone * for an entry in _zonerefs
 643 * zonelist_zone_idx()  - Return the index of the zone for an entry
 644 * zonelist_node_idx()  - Return the index of the node for an entry
 645 */
 646struct zonelist {
 647        struct zonelist_cache *zlcache_ptr;                  // NULL or &zlcache
 648        struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
 649#ifdef CONFIG_NUMA
 650        struct zonelist_cache zlcache;                       // optional ...
 651#endif
 652};
 653
 654#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
 655struct node_active_region {
 656        unsigned long start_pfn;
 657        unsigned long end_pfn;
 658        int nid;
 659};
 660#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
 661
 662#ifndef CONFIG_DISCONTIGMEM
 663/* The array of struct pages - for discontigmem use pgdat->lmem_map */
 664extern struct page *mem_map;
 665#endif
 666
 667/*
 668 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
 669 * (mostly NUMA machines?) to denote a higher-level memory zone than the
 670 * zone denotes.
 671 *
 672 * On NUMA machines, each NUMA node would have a pg_data_t to describe
 673 * it's memory layout.
 674 *
 675 * Memory statistics and page replacement data structures are maintained on a
 676 * per-zone basis.
 677 */
 678struct bootmem_data;
 679typedef struct pglist_data {
 680        struct zone node_zones[MAX_NR_ZONES];
 681        struct zonelist node_zonelists[MAX_ZONELISTS];
 682        int nr_zones;
 683#ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
 684        struct page *node_mem_map;
 685#ifdef CONFIG_MEMCG
 686        struct page_cgroup *node_page_cgroup;
 687#endif
 688#endif
 689#ifndef CONFIG_NO_BOOTMEM
 690        struct bootmem_data *bdata;
 691#endif
 692#ifdef CONFIG_MEMORY_HOTPLUG
 693        /*
 694         * Must be held any time you expect node_start_pfn, node_present_pages
 695         * or node_spanned_pages stay constant.  Holding this will also
 696         * guarantee that any pfn_valid() stays that way.
 697         *
 698         * Nests above zone->lock and zone->size_seqlock.
 699         */
 700        spinlock_t node_size_lock;
 701#endif
 702        unsigned long node_start_pfn;
 703        unsigned long node_present_pages; /* total number of physical pages */
 704        unsigned long node_spanned_pages; /* total size of physical page
 705                                             range, including holes */
 706        int node_id;
 707        wait_queue_head_t kswapd_wait;
 708        wait_queue_head_t pfmemalloc_wait;
 709        struct task_struct *kswapd;     /* Protected by lock_memory_hotplug() */
 710        int kswapd_max_order;
 711        enum zone_type classzone_idx;
 712} pg_data_t;
 713
 714#define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
 715#define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
 716#ifdef CONFIG_FLAT_NODE_MEM_MAP
 717#define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
 718#else
 719#define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
 720#endif
 721#define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
 722
 723#define node_start_pfn(nid)     (NODE_DATA(nid)->node_start_pfn)
 724
 725#define node_end_pfn(nid) ({\
 726        pg_data_t *__pgdat = NODE_DATA(nid);\
 727        __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
 728})
 729
 730#include <linux/memory_hotplug.h>
 731
 732extern struct mutex zonelists_mutex;
 733void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
 734void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
 735bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
 736                int classzone_idx, int alloc_flags);
 737bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
 738                int classzone_idx, int alloc_flags);
 739enum memmap_context {
 740        MEMMAP_EARLY,
 741        MEMMAP_HOTPLUG,
 742};
 743extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
 744                                     unsigned long size,
 745                                     enum memmap_context context);
 746
 747extern void lruvec_init(struct lruvec *lruvec, struct zone *zone);
 748
 749static inline struct zone *lruvec_zone(struct lruvec *lruvec)
 750{
 751#ifdef CONFIG_MEMCG
 752        return lruvec->zone;
 753#else
 754        return container_of(lruvec, struct zone, lruvec);
 755#endif
 756}
 757
 758#ifdef CONFIG_HAVE_MEMORY_PRESENT
 759void memory_present(int nid, unsigned long start, unsigned long end);
 760#else
 761static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
 762#endif
 763
 764#ifdef CONFIG_HAVE_MEMORYLESS_NODES
 765int local_memory_node(int node_id);
 766#else
 767static inline int local_memory_node(int node_id) { return node_id; };
 768#endif
 769
 770#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
 771unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 772#endif
 773
 774/*
 775 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
 776 */
 777#define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
 778
 779static inline int populated_zone(struct zone *zone)
 780{
 781        return (!!zone->present_pages);
 782}
 783
 784extern int movable_zone;
 785
 786static inline int zone_movable_is_highmem(void)
 787{
 788#if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
 789        return movable_zone == ZONE_HIGHMEM;
 790#else
 791        return 0;
 792#endif
 793}
 794
 795static inline int is_highmem_idx(enum zone_type idx)
 796{
 797#ifdef CONFIG_HIGHMEM
 798        return (idx == ZONE_HIGHMEM ||
 799                (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
 800#else
 801        return 0;
 802#endif
 803}
 804
 805static inline int is_normal_idx(enum zone_type idx)
 806{
 807        return (idx == ZONE_NORMAL);
 808}
 809
 810/**
 811 * is_highmem - helper function to quickly check if a struct zone is a 
 812 *              highmem zone or not.  This is an attempt to keep references
 813 *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
 814 * @zone - pointer to struct zone variable
 815 */
 816static inline int is_highmem(struct zone *zone)
 817{
 818#ifdef CONFIG_HIGHMEM
 819        int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
 820        return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
 821               (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
 822                zone_movable_is_highmem());
 823#else
 824        return 0;
 825#endif
 826}
 827
 828static inline int is_normal(struct zone *zone)
 829{
 830        return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
 831}
 832
 833static inline int is_dma32(struct zone *zone)
 834{
 835#ifdef CONFIG_ZONE_DMA32
 836        return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
 837#else
 838        return 0;
 839#endif
 840}
 841
 842static inline int is_dma(struct zone *zone)
 843{
 844#ifdef CONFIG_ZONE_DMA
 845        return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
 846#else
 847        return 0;
 848#endif
 849}
 850
 851/* These two functions are used to setup the per zone pages min values */
 852struct ctl_table;
 853int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
 854                                        void __user *, size_t *, loff_t *);
 855extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
 856int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
 857                                        void __user *, size_t *, loff_t *);
 858int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
 859                                        void __user *, size_t *, loff_t *);
 860int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
 861                        void __user *, size_t *, loff_t *);
 862int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
 863                        void __user *, size_t *, loff_t *);
 864
 865extern int numa_zonelist_order_handler(struct ctl_table *, int,
 866                        void __user *, size_t *, loff_t *);
 867extern char numa_zonelist_order[];
 868#define NUMA_ZONELIST_ORDER_LEN 16      /* string buffer size */
 869
 870#ifndef CONFIG_NEED_MULTIPLE_NODES
 871
 872extern struct pglist_data contig_page_data;
 873#define NODE_DATA(nid)          (&contig_page_data)
 874#define NODE_MEM_MAP(nid)       mem_map
 875
 876#else /* CONFIG_NEED_MULTIPLE_NODES */
 877
 878#include <asm/mmzone.h>
 879
 880#endif /* !CONFIG_NEED_MULTIPLE_NODES */
 881
 882extern struct pglist_data *first_online_pgdat(void);
 883extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
 884extern struct zone *next_zone(struct zone *zone);
 885
 886/**
 887 * for_each_online_pgdat - helper macro to iterate over all online nodes
 888 * @pgdat - pointer to a pg_data_t variable
 889 */
 890#define for_each_online_pgdat(pgdat)                    \
 891        for (pgdat = first_online_pgdat();              \
 892             pgdat;                                     \
 893             pgdat = next_online_pgdat(pgdat))
 894/**
 895 * for_each_zone - helper macro to iterate over all memory zones
 896 * @zone - pointer to struct zone variable
 897 *
 898 * The user only needs to declare the zone variable, for_each_zone
 899 * fills it in.
 900 */
 901#define for_each_zone(zone)                             \
 902        for (zone = (first_online_pgdat())->node_zones; \
 903             zone;                                      \
 904             zone = next_zone(zone))
 905
 906#define for_each_populated_zone(zone)                   \
 907        for (zone = (first_online_pgdat())->node_zones; \
 908             zone;                                      \
 909             zone = next_zone(zone))                    \
 910                if (!populated_zone(zone))              \
 911                        ; /* do nothing */              \
 912                else
 913
 914static inline struct zone *zonelist_zone(struct zoneref *zoneref)
 915{
 916        return zoneref->zone;
 917}
 918
 919static inline int zonelist_zone_idx(struct zoneref *zoneref)
 920{
 921        return zoneref->zone_idx;
 922}
 923
 924static inline int zonelist_node_idx(struct zoneref *zoneref)
 925{
 926#ifdef CONFIG_NUMA
 927        /* zone_to_nid not available in this context */
 928        return zoneref->zone->node;
 929#else
 930        return 0;
 931#endif /* CONFIG_NUMA */
 932}
 933
 934/**
 935 * next_zones_zonelist - Returns the next zone at or below highest_zoneidx within the allowed nodemask using a cursor within a zonelist as a starting point
 936 * @z - The cursor used as a starting point for the search
 937 * @highest_zoneidx - The zone index of the highest zone to return
 938 * @nodes - An optional nodemask to filter the zonelist with
 939 * @zone - The first suitable zone found is returned via this parameter
 940 *
 941 * This function returns the next zone at or below a given zone index that is
 942 * within the allowed nodemask using a cursor as the starting point for the
 943 * search. The zoneref returned is a cursor that represents the current zone
 944 * being examined. It should be advanced by one before calling
 945 * next_zones_zonelist again.
 946 */
 947struct zoneref *next_zones_zonelist(struct zoneref *z,
 948                                        enum zone_type highest_zoneidx,
 949                                        nodemask_t *nodes,
 950                                        struct zone **zone);
 951
 952/**
 953 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
 954 * @zonelist - The zonelist to search for a suitable zone
 955 * @highest_zoneidx - The zone index of the highest zone to return
 956 * @nodes - An optional nodemask to filter the zonelist with
 957 * @zone - The first suitable zone found is returned via this parameter
 958 *
 959 * This function returns the first zone at or below a given zone index that is
 960 * within the allowed nodemask. The zoneref returned is a cursor that can be
 961 * used to iterate the zonelist with next_zones_zonelist by advancing it by
 962 * one before calling.
 963 */
 964static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
 965                                        enum zone_type highest_zoneidx,
 966                                        nodemask_t *nodes,
 967                                        struct zone **zone)
 968{
 969        return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
 970                                                                zone);
 971}
 972
 973/**
 974 * for_each_zone_zonelist_nodemask - helper macro to iterate over valid zones in a zonelist at or below a given zone index and within a nodemask
 975 * @zone - The current zone in the iterator
 976 * @z - The current pointer within zonelist->zones being iterated
 977 * @zlist - The zonelist being iterated
 978 * @highidx - The zone index of the highest zone to return
 979 * @nodemask - Nodemask allowed by the allocator
 980 *
 981 * This iterator iterates though all zones at or below a given zone index and
 982 * within a given nodemask
 983 */
 984#define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
 985        for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
 986                zone;                                                   \
 987                z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
 988
 989/**
 990 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
 991 * @zone - The current zone in the iterator
 992 * @z - The current pointer within zonelist->zones being iterated
 993 * @zlist - The zonelist being iterated
 994 * @highidx - The zone index of the highest zone to return
 995 *
 996 * This iterator iterates though all zones at or below a given zone index.
 997 */
 998#define for_each_zone_zonelist(zone, z, zlist, highidx) \
 999        for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1000
1001#ifdef CONFIG_SPARSEMEM
1002#include <asm/sparsemem.h>
1003#endif
1004
1005#if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1006        !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1007static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1008{
1009        return 0;
1010}
1011#endif
1012
1013#ifdef CONFIG_FLATMEM
1014#define pfn_to_nid(pfn)         (0)
1015#endif
1016
1017#ifdef CONFIG_SPARSEMEM
1018
1019/*
1020 * SECTION_SHIFT                #bits space required to store a section #
1021 *
1022 * PA_SECTION_SHIFT             physical address to/from section number
1023 * PFN_SECTION_SHIFT            pfn to/from section number
1024 */
1025#define SECTIONS_SHIFT          (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
1026
1027#define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
1028#define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
1029
1030#define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
1031
1032#define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
1033#define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
1034
1035#define SECTION_BLOCKFLAGS_BITS \
1036        ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1037
1038#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1039#error Allocator MAX_ORDER exceeds SECTION_SIZE
1040#endif
1041
1042#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1043#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1044
1045#define SECTION_ALIGN_UP(pfn)   (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1046#define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1047
1048struct page;
1049struct page_cgroup;
1050struct mem_section {
1051        /*
1052         * This is, logically, a pointer to an array of struct
1053         * pages.  However, it is stored with some other magic.
1054         * (see sparse.c::sparse_init_one_section())
1055         *
1056         * Additionally during early boot we encode node id of
1057         * the location of the section here to guide allocation.
1058         * (see sparse.c::memory_present())
1059         *
1060         * Making it a UL at least makes someone do a cast
1061         * before using it wrong.
1062         */
1063        unsigned long section_mem_map;
1064
1065        /* See declaration of similar field in struct zone */
1066        unsigned long *pageblock_flags;
1067#ifdef CONFIG_MEMCG
1068        /*
1069         * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1070         * section. (see memcontrol.h/page_cgroup.h about this.)
1071         */
1072        struct page_cgroup *page_cgroup;
1073        unsigned long pad;
1074#endif
1075};
1076
1077#ifdef CONFIG_SPARSEMEM_EXTREME
1078#define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
1079#else
1080#define SECTIONS_PER_ROOT       1
1081#endif
1082
1083#define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1084#define NR_SECTION_ROOTS        DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1085#define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
1086
1087#ifdef CONFIG_SPARSEMEM_EXTREME
1088extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1089#else
1090extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1091#endif
1092
1093static inline struct mem_section *__nr_to_section(unsigned long nr)
1094{
1095        if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1096                return NULL;
1097        return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1098}
1099extern int __section_nr(struct mem_section* ms);
1100extern unsigned long usemap_size(void);
1101
1102/*
1103 * We use the lower bits of the mem_map pointer to store
1104 * a little bit of information.  There should be at least
1105 * 3 bits here due to 32-bit alignment.
1106 */
1107#define SECTION_MARKED_PRESENT  (1UL<<0)
1108#define SECTION_HAS_MEM_MAP     (1UL<<1)
1109#define SECTION_MAP_LAST_BIT    (1UL<<2)
1110#define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
1111#define SECTION_NID_SHIFT       2
1112
1113static inline struct page *__section_mem_map_addr(struct mem_section *section)
1114{
1115        unsigned long map = section->section_mem_map;
1116        map &= SECTION_MAP_MASK;
1117        return (struct page *)map;
1118}
1119
1120static inline int present_section(struct mem_section *section)
1121{
1122        return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1123}
1124
1125static inline int present_section_nr(unsigned long nr)
1126{
1127        return present_section(__nr_to_section(nr));
1128}
1129
1130static inline int valid_section(struct mem_section *section)
1131{
1132        return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1133}
1134
1135static inline int valid_section_nr(unsigned long nr)
1136{
1137        return valid_section(__nr_to_section(nr));
1138}
1139
1140static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1141{
1142        return __nr_to_section(pfn_to_section_nr(pfn));
1143}
1144
1145#ifndef CONFIG_HAVE_ARCH_PFN_VALID
1146static inline int pfn_valid(unsigned long pfn)
1147{
1148        if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1149                return 0;
1150        return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1151}
1152#endif
1153
1154static inline int pfn_present(unsigned long pfn)
1155{
1156        if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1157                return 0;
1158        return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1159}
1160
1161/*
1162 * These are _only_ used during initialisation, therefore they
1163 * can use __initdata ...  They could have names to indicate
1164 * this restriction.
1165 */
1166#ifdef CONFIG_NUMA
1167#define pfn_to_nid(pfn)                                                 \
1168({                                                                      \
1169        unsigned long __pfn_to_nid_pfn = (pfn);                         \
1170        page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
1171})
1172#else
1173#define pfn_to_nid(pfn)         (0)
1174#endif
1175
1176#define early_pfn_valid(pfn)    pfn_valid(pfn)
1177void sparse_init(void);
1178#else
1179#define sparse_init()   do {} while (0)
1180#define sparse_index_init(_sec, _nid)  do {} while (0)
1181#endif /* CONFIG_SPARSEMEM */
1182
1183#ifdef CONFIG_NODES_SPAN_OTHER_NODES
1184bool early_pfn_in_nid(unsigned long pfn, int nid);
1185#else
1186#define early_pfn_in_nid(pfn, nid)      (1)
1187#endif
1188
1189#ifndef early_pfn_valid
1190#define early_pfn_valid(pfn)    (1)
1191#endif
1192
1193void memory_present(int nid, unsigned long start, unsigned long end);
1194unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1195
1196/*
1197 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1198 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1199 * pfn_valid_within() should be used in this case; we optimise this away
1200 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1201 */
1202#ifdef CONFIG_HOLES_IN_ZONE
1203#define pfn_valid_within(pfn) pfn_valid(pfn)
1204#else
1205#define pfn_valid_within(pfn) (1)
1206#endif
1207
1208#ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1209/*
1210 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1211 * associated with it or not. In FLATMEM, it is expected that holes always
1212 * have valid memmap as long as there is valid PFNs either side of the hole.
1213 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1214 * entire section.
1215 *
1216 * However, an ARM, and maybe other embedded architectures in the future
1217 * free memmap backing holes to save memory on the assumption the memmap is
1218 * never used. The page_zone linkages are then broken even though pfn_valid()
1219 * returns true. A walker of the full memmap must then do this additional
1220 * check to ensure the memmap they are looking at is sane by making sure
1221 * the zone and PFN linkages are still valid. This is expensive, but walkers
1222 * of the full memmap are extremely rare.
1223 */
1224int memmap_valid_within(unsigned long pfn,
1225                                        struct page *page, struct zone *zone);
1226#else
1227static inline int memmap_valid_within(unsigned long pfn,
1228                                        struct page *page, struct zone *zone)
1229{
1230        return 1;
1231}
1232#endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1233
1234#endif /* !__GENERATING_BOUNDS.H */
1235#endif /* !__ASSEMBLY__ */
1236#endif /* _LINUX_MMZONE_H */
1237
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.