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#else
  67#  define is_migrate_cma(migratetype) false
  68#endif
  69
  70#define for_each_migratetype_order(order, type) \
  71        for (order = 0; order < MAX_ORDER; order++) \
  72                for (type = 0; type < MIGRATE_TYPES; type++)
  73
  74extern int page_group_by_mobility_disabled;
  75
  76static inline int get_pageblock_migratetype(struct page *page)
  77{
  78        return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
  79}
  80
  81struct free_area {
  82        struct list_head        free_list[MIGRATE_TYPES];
  83        unsigned long           nr_free;
  84};
  85
  86struct pglist_data;
  87
  88/*
  89 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
  90 * So add a wild amount of padding here to ensure that they fall into separate
  91 * cachelines.  There are very few zone structures in the machine, so space
  92 * consumption is not a concern here.
  93 */
  94#if defined(CONFIG_SMP)
  95struct zone_padding {
  96        char x[0];
  97} ____cacheline_internodealigned_in_smp;
  98#define ZONE_PADDING(name)      struct zone_padding name;
  99#else
 100#define ZONE_PADDING(name)
 101#endif
 102
 103enum zone_stat_item {
 104        /* First 128 byte cacheline (assuming 64 bit words) */
 105        NR_FREE_PAGES,
 106        NR_LRU_BASE,
 107        NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
 108        NR_ACTIVE_ANON,         /*  "     "     "   "       "         */
 109        NR_INACTIVE_FILE,       /*  "     "     "   "       "         */
 110        NR_ACTIVE_FILE,         /*  "     "     "   "       "         */
 111        NR_UNEVICTABLE,         /*  "     "     "   "       "         */
 112        NR_MLOCK,               /* mlock()ed pages found and moved off LRU */
 113        NR_ANON_PAGES,  /* Mapped anonymous pages */
 114        NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
 115                           only modified from process context */
 116        NR_FILE_PAGES,
 117        NR_FILE_DIRTY,
 118        NR_WRITEBACK,
 119        NR_SLAB_RECLAIMABLE,
 120        NR_SLAB_UNRECLAIMABLE,
 121        NR_PAGETABLE,           /* used for pagetables */
 122        NR_KERNEL_STACK,
 123        /* Second 128 byte cacheline */
 124        NR_UNSTABLE_NFS,        /* NFS unstable pages */
 125        NR_BOUNCE,
 126        NR_VMSCAN_WRITE,
 127        NR_VMSCAN_IMMEDIATE,    /* Prioritise for reclaim when writeback ends */
 128        NR_WRITEBACK_TEMP,      /* Writeback using temporary buffers */
 129        NR_ISOLATED_ANON,       /* Temporary isolated pages from anon lru */
 130        NR_ISOLATED_FILE,       /* Temporary isolated pages from file lru */
 131        NR_SHMEM,               /* shmem pages (included tmpfs/GEM pages) */
 132        NR_DIRTIED,             /* page dirtyings since bootup */
 133        NR_WRITTEN,             /* page writings since bootup */
 134#ifdef CONFIG_NUMA
 135        NUMA_HIT,               /* allocated in intended node */
 136        NUMA_MISS,              /* allocated in non intended node */
 137        NUMA_FOREIGN,           /* was intended here, hit elsewhere */
 138        NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
 139        NUMA_LOCAL,             /* allocation from local node */
 140        NUMA_OTHER,             /* allocation from other node */
 141#endif
 142        NR_ANON_TRANSPARENT_HUGEPAGES,
 143        NR_FREE_CMA_PAGES,
 144        NR_VM_ZONE_STAT_ITEMS };
 145
 146/*
 147 * We do arithmetic on the LRU lists in various places in the code,
 148 * so it is important to keep the active lists LRU_ACTIVE higher in
 149 * the array than the corresponding inactive lists, and to keep
 150 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
 151 *
 152 * This has to be kept in sync with the statistics in zone_stat_item
 153 * above and the descriptions in vmstat_text in mm/vmstat.c
 154 */
 155#define LRU_BASE 0
 156#define LRU_ACTIVE 1
 157#define LRU_FILE 2
 158
 159enum lru_list {
 160        LRU_INACTIVE_ANON = LRU_BASE,
 161        LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
 162        LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
 163        LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
 164        LRU_UNEVICTABLE,
 165        NR_LRU_LISTS
 166};
 167
 168#define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
 169
 170#define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
 171
 172static inline int is_file_lru(enum lru_list lru)
 173{
 174        return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
 175}
 176
 177static inline int is_active_lru(enum lru_list lru)
 178{
 179        return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
 180}
 181
 182static inline int is_unevictable_lru(enum lru_list lru)
 183{
 184        return (lru == LRU_UNEVICTABLE);
 185}
 186
 187struct zone_reclaim_stat {
 188        /*
 189         * The pageout code in vmscan.c keeps track of how many of the
 190         * mem/swap backed and file backed pages are referenced.
 191         * The higher the rotated/scanned ratio, the more valuable
 192         * that cache is.
 193         *
 194         * The anon LRU stats live in [0], file LRU stats in [1]
 195         */
 196        unsigned long           recent_rotated[2];
 197        unsigned long           recent_scanned[2];
 198};
 199
 200struct lruvec {
 201        struct list_head lists[NR_LRU_LISTS];
 202        struct zone_reclaim_stat reclaim_stat;
 203#ifdef CONFIG_MEMCG
 204        struct zone *zone;
 205#endif
 206};
 207
 208/* Mask used at gathering information at once (see memcontrol.c) */
 209#define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
 210#define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
 211#define LRU_ALL      ((1 << NR_LRU_LISTS) - 1)
 212
 213/* Isolate clean file */
 214#define ISOLATE_CLEAN           ((__force isolate_mode_t)0x1)
 215/* Isolate unmapped file */
 216#define ISOLATE_UNMAPPED        ((__force isolate_mode_t)0x2)
 217/* Isolate for asynchronous migration */
 218#define ISOLATE_ASYNC_MIGRATE   ((__force isolate_mode_t)0x4)
 219/* Isolate unevictable pages */
 220#define ISOLATE_UNEVICTABLE     ((__force isolate_mode_t)0x8)
 221
 222/* LRU Isolation modes. */
 223typedef unsigned __bitwise__ isolate_mode_t;
 224
 225enum zone_watermarks {
 226        WMARK_MIN,
 227        WMARK_LOW,
 228        WMARK_HIGH,
 229        NR_WMARK
 230};
 231
 232#define min_wmark_pages(z) (z->watermark[WMARK_MIN])
 233#define low_wmark_pages(z) (z->watermark[WMARK_LOW])
 234#define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
 235
 236struct per_cpu_pages {
 237        int count;              /* number of pages in the list */
 238        int high;               /* high watermark, emptying needed */
 239        int batch;              /* chunk size for buddy add/remove */
 240
 241        /* Lists of pages, one per migrate type stored on the pcp-lists */
 242        struct list_head lists[MIGRATE_PCPTYPES];
 243};
 244
 245struct per_cpu_pageset {
 246        struct per_cpu_pages pcp;
 247#ifdef CONFIG_NUMA
 248        s8 expire;
 249#endif
 250#ifdef CONFIG_SMP
 251        s8 stat_threshold;
 252        s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
 253#endif
 254};
 255
 256#endif /* !__GENERATING_BOUNDS.H */
 257
 258enum zone_type {
 259#ifdef CONFIG_ZONE_DMA
 260        /*
 261         * ZONE_DMA is used when there are devices that are not able
 262         * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
 263         * carve out the portion of memory that is needed for these devices.
 264         * The range is arch specific.
 265         *
 266         * Some examples
 267         *
 268         * Architecture         Limit
 269         * ---------------------------
 270         * parisc, ia64, sparc  <4G
 271         * s390                 <2G
 272         * arm                  Various
 273         * alpha                Unlimited or 0-16MB.
 274         *
 275         * i386, x86_64 and multiple other arches
 276         *                      <16M.
 277         */
 278        ZONE_DMA,
 279#endif
 280#ifdef CONFIG_ZONE_DMA32
 281        /*
 282         * x86_64 needs two ZONE_DMAs because it supports devices that are
 283         * only able to do DMA to the lower 16M but also 32 bit devices that
 284         * can only do DMA areas below 4G.
 285         */
 286        ZONE_DMA32,
 287#endif
 288        /*
 289         * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
 290         * performed on pages in ZONE_NORMAL if the DMA devices support
 291         * transfers to all addressable memory.
 292         */
 293        ZONE_NORMAL,
 294#ifdef CONFIG_HIGHMEM
 295        /*
 296         * A memory area that is only addressable by the kernel through
 297         * mapping portions into its own address space. This is for example
 298         * used by i386 to allow the kernel to address the memory beyond
 299         * 900MB. The kernel will set up special mappings (page
 300         * table entries on i386) for each page that the kernel needs to
 301         * access.
 302         */
 303        ZONE_HIGHMEM,
 304#endif
 305        ZONE_MOVABLE,
 306        __MAX_NR_ZONES
 307};
 308
 309#ifndef __GENERATING_BOUNDS_H
 310
 311/*
 312 * When a memory allocation must conform to specific limitations (such
 313 * as being suitable for DMA) the caller will pass in hints to the
 314 * allocator in the gfp_mask, in the zone modifier bits.  These bits
 315 * are used to select a priority ordered list of memory zones which
 316 * match the requested limits. See gfp_zone() in include/linux/gfp.h
 317 */
 318
 319#if MAX_NR_ZONES < 2
 320#define ZONES_SHIFT 0
 321#elif MAX_NR_ZONES <= 2
 322#define ZONES_SHIFT 1
 323#elif MAX_NR_ZONES <= 4
 324#define ZONES_SHIFT 2
 325#else
 326#error ZONES_SHIFT -- too many zones configured adjust calculation
 327#endif
 328
 329struct zone {
 330        /* Fields commonly accessed by the page allocator */
 331
 332        /* zone watermarks, access with *_wmark_pages(zone) macros */
 333        unsigned long watermark[NR_WMARK];
 334
 335        /*
 336         * When free pages are below this point, additional steps are taken
 337         * when reading the number of free pages to avoid per-cpu counter
 338         * drift allowing watermarks to be breached
 339         */
 340        unsigned long percpu_drift_mark;
 341
 342        /*
 343         * We don't know if the memory that we're going to allocate will be freeable
 344         * or/and it will be released eventually, so to avoid totally wasting several
 345         * GB of ram we must reserve some of the lower zone memory (otherwise we risk
 346         * to run OOM on the lower zones despite there's tons of freeable ram
 347         * on the higher zones). This array is recalculated at runtime if the
 348         * sysctl_lowmem_reserve_ratio sysctl changes.
 349         */
 350        unsigned long           lowmem_reserve[MAX_NR_ZONES];
 351
 352        /*
 353         * This is a per-zone reserve of pages that should not be
 354         * considered dirtyable memory.
 355         */
 356        unsigned long           dirty_balance_reserve;
 357
 358#ifdef CONFIG_NUMA
 359        int node;
 360        /*
 361         * zone reclaim becomes active if more unmapped pages exist.
 362         */
 363        unsigned long           min_unmapped_pages;
 364        unsigned long           min_slab_pages;
 365#endif
 366        struct per_cpu_pageset __percpu *pageset;
 367        /*
 368         * free areas of different sizes
 369         */
 370        spinlock_t              lock;
 371        int                     all_unreclaimable; /* All pages pinned */
 372#if defined CONFIG_COMPACTION || defined CONFIG_CMA
 373        /* Set to true when the PG_migrate_skip bits should be cleared */
 374        bool                    compact_blockskip_flush;
 375
 376        /* pfns where compaction scanners should start */
 377        unsigned long           compact_cached_free_pfn;
 378        unsigned long           compact_cached_migrate_pfn;
 379#endif
 380#ifdef CONFIG_MEMORY_HOTPLUG
 381        /* see spanned/present_pages for more description */
 382        seqlock_t               span_seqlock;
 383#endif
 384        struct free_area        free_area[MAX_ORDER];
 385
 386#ifndef CONFIG_SPARSEMEM
 387        /*
 388         * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
 389         * In SPARSEMEM, this map is stored in struct mem_section
 390         */
 391        unsigned long           *pageblock_flags;
 392#endif /* CONFIG_SPARSEMEM */
 393
 394#ifdef CONFIG_COMPACTION
 395        /*
 396         * On compaction failure, 1<<compact_defer_shift compactions
 397         * are skipped before trying again. The number attempted since
 398         * last failure is tracked with compact_considered.
 399         */
 400        unsigned int            compact_considered;
 401        unsigned int            compact_defer_shift;
 402        int                     compact_order_failed;
 403#endif
 404
 405        ZONE_PADDING(_pad1_)
 406
 407        /* Fields commonly accessed by the page reclaim scanner */
 408        spinlock_t              lru_lock;
 409        struct lruvec           lruvec;
 410
 411        unsigned long           pages_scanned;     /* since last reclaim */
 412        unsigned long           flags;             /* zone flags, see below */
 413
 414        /* Zone statistics */
 415        atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
 416
 417        /*
 418         * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
 419         * this zone's LRU.  Maintained by the pageout code.
 420         */
 421        unsigned int inactive_ratio;
 422
 423
 424        ZONE_PADDING(_pad2_)
 425        /* Rarely used or read-mostly fields */
 426
 427        /*
 428         * wait_table           -- the array holding the hash table
 429         * wait_table_hash_nr_entries   -- the size of the hash table array
 430         * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
 431         *
 432         * The purpose of all these is to keep track of the people
 433         * waiting for a page to become available and make them
 434         * runnable again when possible. The trouble is that this
 435         * consumes a lot of space, especially when so few things
 436         * wait on pages at a given time. So instead of using
 437         * per-page waitqueues, we use a waitqueue hash table.
 438         *
 439         * The bucket discipline is to sleep on the same queue when
 440         * colliding and wake all in that wait queue when removing.
 441         * When something wakes, it must check to be sure its page is
 442         * truly available, a la thundering herd. The cost of a
 443         * collision is great, but given the expected load of the
 444         * table, they should be so rare as to be outweighed by the
 445         * benefits from the saved space.
 446         *
 447         * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
 448         * primary users of these fields, and in mm/page_alloc.c
 449         * free_area_init_core() performs the initialization of them.
 450         */
 451        wait_queue_head_t       * wait_table;
 452        unsigned long           wait_table_hash_nr_entries;
 453        unsigned long           wait_table_bits;
 454
 455        /*
 456         * Discontig memory support fields.
 457         */
 458        struct pglist_data      *zone_pgdat;
 459        /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
 460        unsigned long           zone_start_pfn;
 461
 462        /*
 463         * spanned_pages is the total pages spanned by the zone, including
 464         * holes, which is calculated as:
 465         *      spanned_pages = zone_end_pfn - zone_start_pfn;
 466         *
 467         * present_pages is physical pages existing within the zone, which
 468         * is calculated as:
 469         *      present_pages = spanned_pages - absent_pages(pags in holes);
 470         *
 471         * managed_pages is present pages managed by the buddy system, which
 472         * is calculated as (reserved_pages includes pages allocated by the
 473         * bootmem allocator):
 474         *      managed_pages = present_pages - reserved_pages;
 475         *
 476         * So present_pages may be used by memory hotplug or memory power
 477         * management logic to figure out unmanaged pages by checking
 478         * (present_pages - managed_pages). And managed_pages should be used
 479         * by page allocator and vm scanner to calculate all kinds of watermarks
 480         * and thresholds.
 481         *
 482         * Locking rules:
 483         *
 484         * zone_start_pfn and spanned_pages are protected by span_seqlock.
 485         * It is a seqlock because it has to be read outside of zone->lock,
 486         * and it is done in the main allocator path.  But, it is written
 487         * quite infrequently.
 488         *
 489         * The span_seq lock is declared along with zone->lock because it is
 490         * frequently read in proximity to zone->lock.  It's good to
 491         * give them a chance of being in the same cacheline.
 492         *
 493         * Write access to present_pages and managed_pages at runtime should
 494         * be protected by lock_memory_hotplug()/unlock_memory_hotplug().
 495         * Any reader who can't tolerant drift of present_pages and
 496         * managed_pages should hold memory hotplug lock to get a stable value.
 497         */
 498        unsigned long           spanned_pages;
 499        unsigned long           present_pages;
 500        unsigned long           managed_pages;
 501
 502        /*
 503         * rarely used fields:
 504         */
 505        const char              *name;
 506} ____cacheline_internodealigned_in_smp;
 507
 508typedef enum {
 509        ZONE_RECLAIM_LOCKED,            /* prevents concurrent reclaim */
 510        ZONE_OOM_LOCKED,                /* zone is in OOM killer zonelist */
 511        ZONE_CONGESTED,                 /* zone has many dirty pages backed by
 512                                         * a congested BDI
 513                                         */
 514} zone_flags_t;
 515
 516static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
 517{
 518        set_bit(flag, &zone->flags);
 519}
 520
 521static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
 522{
 523        return test_and_set_bit(flag, &zone->flags);
 524}
 525
 526static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
 527{
 528        clear_bit(flag, &zone->flags);
 529}
 530
 531static inline int zone_is_reclaim_congested(const struct zone *zone)
 532{
 533        return test_bit(ZONE_CONGESTED, &zone->flags);
 534}
 535
 536static inline int zone_is_reclaim_locked(const struct zone *zone)
 537{
 538        return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
 539}
 540
 541static inline int zone_is_oom_locked(const struct zone *zone)
 542{
 543        return test_bit(ZONE_OOM_LOCKED, &zone->flags);
 544}
 545
 546/*
 547 * The "priority" of VM scanning is how much of the queues we will scan in one
 548 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
 549 * queues ("queue_length >> 12") during an aging round.
 550 */
 551#define DEF_PRIORITY 12
 552
 553/* Maximum number of zones on a zonelist */
 554#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
 555
 556#ifdef CONFIG_NUMA
 557
 558/*
 559 * The NUMA zonelists are doubled because we need zonelists that restrict the
 560 * allocations to a single node for GFP_THISNODE.
 561 *
 562 * [0]  : Zonelist with fallback
 563 * [1]  : No fallback (GFP_THISNODE)
 564 */
 565#define MAX_ZONELISTS 2
 566
 567
 568/*
 569 * We cache key information from each zonelist for smaller cache
 570 * footprint when scanning for free pages in get_page_from_freelist().
 571 *
 572 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
 573 *    up short of free memory since the last time (last_fullzone_zap)
 574 *    we zero'd fullzones.
 575 * 2) The array z_to_n[] maps each zone in the zonelist to its node
 576 *    id, so that we can efficiently evaluate whether that node is
 577 *    set in the current tasks mems_allowed.
 578 *
 579 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
 580 * indexed by a zones offset in the zonelist zones[] array.
 581 *
 582 * The get_page_from_freelist() routine does two scans.  During the
 583 * first scan, we skip zones whose corresponding bit in 'fullzones'
 584 * is set or whose corresponding node in current->mems_allowed (which
 585 * comes from cpusets) is not set.  During the second scan, we bypass
 586 * this zonelist_cache, to ensure we look methodically at each zone.
 587 *
 588 * Once per second, we zero out (zap) fullzones, forcing us to
 589 * reconsider nodes that might have regained more free memory.
 590 * The field last_full_zap is the time we last zapped fullzones.
 591 *
 592 * This mechanism reduces the amount of time we waste repeatedly
 593 * reexaming zones for free memory when they just came up low on
 594 * memory momentarilly ago.
 595 *
 596 * The zonelist_cache struct members logically belong in struct
 597 * zonelist.  However, the mempolicy zonelists constructed for
 598 * MPOL_BIND are intentionally variable length (and usually much
 599 * shorter).  A general purpose mechanism for handling structs with
 600 * multiple variable length members is more mechanism than we want
 601 * here.  We resort to some special case hackery instead.
 602 *
 603 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
 604 * part because they are shorter), so we put the fixed length stuff
 605 * at the front of the zonelist struct, ending in a variable length
 606 * zones[], as is needed by MPOL_BIND.
 607 *
 608 * Then we put the optional zonelist cache on the end of the zonelist
 609 * struct.  This optional stuff is found by a 'zlcache_ptr' pointer in
 610 * the fixed length portion at the front of the struct.  This pointer
 611 * both enables us to find the zonelist cache, and in the case of
 612 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
 613 * to know that the zonelist cache is not there.
 614 *
 615 * The end result is that struct zonelists come in two flavors:
 616 *  1) The full, fixed length version, shown below, and
 617 *  2) The custom zonelists for MPOL_BIND.
 618 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
 619 *
 620 * Even though there may be multiple CPU cores on a node modifying
 621 * fullzones or last_full_zap in the same zonelist_cache at the same
 622 * time, we don't lock it.  This is just hint data - if it is wrong now
 623 * and then, the allocator will still function, perhaps a bit slower.
 624 */
 625
 626
 627struct zonelist_cache {
 628        unsigned short z_to_n[MAX_ZONES_PER_ZONELIST];          /* zone->nid */
 629        DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST);      /* zone full? */
 630        unsigned long last_full_zap;            /* when last zap'd (jiffies) */
 631};
 632#else
 633#define MAX_ZONELISTS 1
 634struct zonelist_cache;
 635#endif
 636
 637/*
 638 * This struct contains information about a zone in a zonelist. It is stored
 639 * here to avoid dereferences into large structures and lookups of tables
 640 */
 641struct zoneref {
 642        struct zone *zone;      /* Pointer to actual zone */
 643        int zone_idx;           /* zone_idx(zoneref->zone) */
 644};
 645
 646/*
 647 * One allocation request operates on a zonelist. A zonelist
 648 * is a list of zones, the first one is the 'goal' of the
 649 * allocation, the other zones are fallback zones, in decreasing
 650 * priority.
 651 *
 652 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
 653 * as explained above.  If zlcache_ptr is NULL, there is no zlcache.
 654 * *
 655 * To speed the reading of the zonelist, the zonerefs contain the zone index
 656 * of the entry being read. Helper functions to access information given
 657 * a struct zoneref are
 658 *
 659 * zonelist_zone()      - Return the struct zone * for an entry in _zonerefs
 660 * zonelist_zone_idx()  - Return the index of the zone for an entry
 661 * zonelist_node_idx()  - Return the index of the node for an entry
 662 */
 663struct zonelist {
 664        struct zonelist_cache *zlcache_ptr;                  // NULL or &zlcache
 665        struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
 666#ifdef CONFIG_NUMA
 667        struct zonelist_cache zlcache;                       // optional ...
 668#endif
 669};
 670
 671#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
 672struct node_active_region {
 673        unsigned long start_pfn;
 674        unsigned long end_pfn;
 675        int nid;
 676};
 677#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
 678
 679#ifndef CONFIG_DISCONTIGMEM
 680/* The array of struct pages - for discontigmem use pgdat->lmem_map */
 681extern struct page *mem_map;
 682#endif
 683
 684/*
 685 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
 686 * (mostly NUMA machines?) to denote a higher-level memory zone than the
 687 * zone denotes.
 688 *
 689 * On NUMA machines, each NUMA node would have a pg_data_t to describe
 690 * it's memory layout.
 691 *
 692 * Memory statistics and page replacement data structures are maintained on a
 693 * per-zone basis.
 694 */
 695struct bootmem_data;
 696typedef struct pglist_data {
 697        struct zone node_zones[MAX_NR_ZONES];
 698        struct zonelist node_zonelists[MAX_ZONELISTS];
 699        int nr_zones;
 700#ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
 701        struct page *node_mem_map;
 702#ifdef CONFIG_MEMCG
 703        struct page_cgroup *node_page_cgroup;
 704#endif
 705#endif
 706#ifndef CONFIG_NO_BOOTMEM
 707        struct bootmem_data *bdata;
 708#endif
 709#ifdef CONFIG_MEMORY_HOTPLUG
 710        /*
 711         * Must be held any time you expect node_start_pfn, node_present_pages
 712         * or node_spanned_pages stay constant.  Holding this will also
 713         * guarantee that any pfn_valid() stays that way.
 714         *
 715         * Nests above zone->lock and zone->size_seqlock.
 716         */
 717        spinlock_t node_size_lock;
 718#endif
 719        unsigned long node_start_pfn;
 720        unsigned long node_present_pages; /* total number of physical pages */
 721        unsigned long node_spanned_pages; /* total size of physical page
 722                                             range, including holes */
 723        int node_id;
 724        nodemask_t reclaim_nodes;       /* Nodes allowed to reclaim from */
 725        wait_queue_head_t kswapd_wait;
 726        wait_queue_head_t pfmemalloc_wait;
 727        struct task_struct *kswapd;     /* Protected by lock_memory_hotplug() */
 728        int kswapd_max_order;
 729        enum zone_type classzone_idx;
 730#ifdef CONFIG_NUMA_BALANCING
 731        /*
 732         * Lock serializing the per destination node AutoNUMA memory
 733         * migration rate limiting data.
 734         */
 735        spinlock_t numabalancing_migrate_lock;
 736
 737        /* Rate limiting time interval */
 738        unsigned long numabalancing_migrate_next_window;
 739
 740        /* Number of pages migrated during the rate limiting time interval */
 741        unsigned long numabalancing_migrate_nr_pages;
 742#endif
 743} pg_data_t;
 744
 745#define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
 746#define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
 747#ifdef CONFIG_FLAT_NODE_MEM_MAP
 748#define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
 749#else
 750#define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
 751#endif
 752#define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
 753
 754#define node_start_pfn(nid)     (NODE_DATA(nid)->node_start_pfn)
 755
 756#define node_end_pfn(nid) ({\
 757        pg_data_t *__pgdat = NODE_DATA(nid);\
 758        __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
 759})
 760
 761#include <linux/memory_hotplug.h>
 762
 763extern struct mutex zonelists_mutex;
 764void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
 765void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
 766bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
 767                int classzone_idx, int alloc_flags);
 768bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
 769                int classzone_idx, int alloc_flags);
 770enum memmap_context {
 771        MEMMAP_EARLY,
 772        MEMMAP_HOTPLUG,
 773};
 774extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
 775                                     unsigned long size,
 776                                     enum memmap_context context);
 777
 778extern void lruvec_init(struct lruvec *lruvec);
 779
 780static inline struct zone *lruvec_zone(struct lruvec *lruvec)
 781{
 782#ifdef CONFIG_MEMCG
 783        return lruvec->zone;
 784#else
 785        return container_of(lruvec, struct zone, lruvec);
 786#endif
 787}
 788
 789#ifdef CONFIG_HAVE_MEMORY_PRESENT
 790void memory_present(int nid, unsigned long start, unsigned long end);
 791#else
 792static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
 793#endif
 794
 795#ifdef CONFIG_HAVE_MEMORYLESS_NODES
 796int local_memory_node(int node_id);
 797#else
 798static inline int local_memory_node(int node_id) { return node_id; };
 799#endif
 800
 801#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
 802unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 803#endif
 804
 805/*
 806 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
 807 */
 808#define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
 809
 810static inline int populated_zone(struct zone *zone)
 811{
 812        return (!!zone->present_pages);
 813}
 814
 815extern int movable_zone;
 816
 817static inline int zone_movable_is_highmem(void)
 818{
 819#if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
 820        return movable_zone == ZONE_HIGHMEM;
 821#else
 822        return 0;
 823#endif
 824}
 825
 826static inline int is_highmem_idx(enum zone_type idx)
 827{
 828#ifdef CONFIG_HIGHMEM
 829        return (idx == ZONE_HIGHMEM ||
 830                (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
 831#else
 832        return 0;
 833#endif
 834}
 835
 836static inline int is_normal_idx(enum zone_type idx)
 837{
 838        return (idx == ZONE_NORMAL);
 839}
 840
 841/**
 842 * is_highmem - helper function to quickly check if a struct zone is a 
 843 *              highmem zone or not.  This is an attempt to keep references
 844 *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
 845 * @zone - pointer to struct zone variable
 846 */
 847static inline int is_highmem(struct zone *zone)
 848{
 849#ifdef CONFIG_HIGHMEM
 850        int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
 851        return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
 852               (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
 853                zone_movable_is_highmem());
 854#else
 855        return 0;
 856#endif
 857}
 858
 859static inline int is_normal(struct zone *zone)
 860{
 861        return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
 862}
 863
 864static inline int is_dma32(struct zone *zone)
 865{
 866#ifdef CONFIG_ZONE_DMA32
 867        return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
 868#else
 869        return 0;
 870#endif
 871}
 872
 873static inline int is_dma(struct zone *zone)
 874{
 875#ifdef CONFIG_ZONE_DMA
 876        return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
 877#else
 878        return 0;
 879#endif
 880}
 881
 882/* These two functions are used to setup the per zone pages min values */
 883struct ctl_table;
 884int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
 885                                        void __user *, size_t *, loff_t *);
 886extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
 887int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
 888                                        void __user *, size_t *, loff_t *);
 889int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
 890                                        void __user *, size_t *, loff_t *);
 891int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
 892                        void __user *, size_t *, loff_t *);
 893int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
 894                        void __user *, size_t *, loff_t *);
 895
 896extern int numa_zonelist_order_handler(struct ctl_table *, int,
 897                        void __user *, size_t *, loff_t *);
 898extern char numa_zonelist_order[];
 899#define NUMA_ZONELIST_ORDER_LEN 16      /* string buffer size */
 900
 901#ifndef CONFIG_NEED_MULTIPLE_NODES
 902
 903extern struct pglist_data contig_page_data;
 904#define NODE_DATA(nid)          (&contig_page_data)
 905#define NODE_MEM_MAP(nid)       mem_map
 906
 907#else /* CONFIG_NEED_MULTIPLE_NODES */
 908
 909#include <asm/mmzone.h>
 910
 911#endif /* !CONFIG_NEED_MULTIPLE_NODES */
 912
 913extern struct pglist_data *first_online_pgdat(void);
 914extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
 915extern struct zone *next_zone(struct zone *zone);
 916
 917/**
 918 * for_each_online_pgdat - helper macro to iterate over all online nodes
 919 * @pgdat - pointer to a pg_data_t variable
 920 */
 921#define for_each_online_pgdat(pgdat)                    \
 922        for (pgdat = first_online_pgdat();              \
 923             pgdat;                                     \
 924             pgdat = next_online_pgdat(pgdat))
 925/**
 926 * for_each_zone - helper macro to iterate over all memory zones
 927 * @zone - pointer to struct zone variable
 928 *
 929 * The user only needs to declare the zone variable, for_each_zone
 930 * fills it in.
 931 */
 932#define for_each_zone(zone)                             \
 933        for (zone = (first_online_pgdat())->node_zones; \
 934             zone;                                      \
 935             zone = next_zone(zone))
 936
 937#define for_each_populated_zone(zone)                   \
 938        for (zone = (first_online_pgdat())->node_zones; \
 939             zone;                                      \
 940             zone = next_zone(zone))                    \
 941                if (!populated_zone(zone))              \
 942                        ; /* do nothing */              \
 943                else
 944
 945static inline struct zone *zonelist_zone(struct zoneref *zoneref)
 946{
 947        return zoneref->zone;
 948}
 949
 950static inline int zonelist_zone_idx(struct zoneref *zoneref)
 951{
 952        return zoneref->zone_idx;
 953}
 954
 955static inline int zonelist_node_idx(struct zoneref *zoneref)
 956{
 957#ifdef CONFIG_NUMA
 958        /* zone_to_nid not available in this context */
 959        return zoneref->zone->node;
 960#else
 961        return 0;
 962#endif /* CONFIG_NUMA */
 963}
 964
 965/**
 966 * 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
 967 * @z - The cursor used as a starting point for the search
 968 * @highest_zoneidx - The zone index of the highest zone to return
 969 * @nodes - An optional nodemask to filter the zonelist with
 970 * @zone - The first suitable zone found is returned via this parameter
 971 *
 972 * This function returns the next zone at or below a given zone index that is
 973 * within the allowed nodemask using a cursor as the starting point for the
 974 * search. The zoneref returned is a cursor that represents the current zone
 975 * being examined. It should be advanced by one before calling
 976 * next_zones_zonelist again.
 977 */
 978struct zoneref *next_zones_zonelist(struct zoneref *z,
 979                                        enum zone_type highest_zoneidx,
 980                                        nodemask_t *nodes,
 981                                        struct zone **zone);
 982
 983/**
 984 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
 985 * @zonelist - The zonelist to search for a suitable zone
 986 * @highest_zoneidx - The zone index of the highest zone to return
 987 * @nodes - An optional nodemask to filter the zonelist with
 988 * @zone - The first suitable zone found is returned via this parameter
 989 *
 990 * This function returns the first zone at or below a given zone index that is
 991 * within the allowed nodemask. The zoneref returned is a cursor that can be
 992 * used to iterate the zonelist with next_zones_zonelist by advancing it by
 993 * one before calling.
 994 */
 995static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
 996                                        enum zone_type highest_zoneidx,
 997                                        nodemask_t *nodes,
 998                                        struct zone **zone)
 999{
1000        return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
1001                                                                zone);
1002}
1003
1004/**
1005 * 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
1006 * @zone - The current zone in the iterator
1007 * @z - The current pointer within zonelist->zones being iterated
1008 * @zlist - The zonelist being iterated
1009 * @highidx - The zone index of the highest zone to return
1010 * @nodemask - Nodemask allowed by the allocator
1011 *
1012 * This iterator iterates though all zones at or below a given zone index and
1013 * within a given nodemask
1014 */
1015#define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1016        for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
1017                zone;                                                   \
1018                z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
1019
1020/**
1021 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1022 * @zone - The current zone in the iterator
1023 * @z - The current pointer within zonelist->zones being iterated
1024 * @zlist - The zonelist being iterated
1025 * @highidx - The zone index of the highest zone to return
1026 *
1027 * This iterator iterates though all zones at or below a given zone index.
1028 */
1029#define for_each_zone_zonelist(zone, z, zlist, highidx) \
1030        for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1031
1032#ifdef CONFIG_SPARSEMEM
1033#include <asm/sparsemem.h>
1034#endif
1035
1036#if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1037        !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1038static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1039{
1040        return 0;
1041}
1042#endif
1043
1044#ifdef CONFIG_FLATMEM
1045#define pfn_to_nid(pfn)         (0)
1046#endif
1047
1048#ifdef CONFIG_SPARSEMEM
1049
1050/*
1051 * SECTION_SHIFT                #bits space required to store a section #
1052 *
1053 * PA_SECTION_SHIFT             physical address to/from section number
1054 * PFN_SECTION_SHIFT            pfn to/from section number
1055 */
1056#define SECTIONS_SHIFT          (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
1057
1058#define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
1059#define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
1060
1061#define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
1062
1063#define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
1064#define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
1065
1066#define SECTION_BLOCKFLAGS_BITS \
1067        ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1068
1069#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1070#error Allocator MAX_ORDER exceeds SECTION_SIZE
1071#endif
1072
1073#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1074#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1075
1076#define SECTION_ALIGN_UP(pfn)   (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1077#define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1078
1079struct page;
1080struct page_cgroup;
1081struct mem_section {
1082        /*
1083         * This is, logically, a pointer to an array of struct
1084         * pages.  However, it is stored with some other magic.
1085         * (see sparse.c::sparse_init_one_section())
1086         *
1087         * Additionally during early boot we encode node id of
1088         * the location of the section here to guide allocation.
1089         * (see sparse.c::memory_present())
1090         *
1091         * Making it a UL at least makes someone do a cast
1092         * before using it wrong.
1093         */
1094        unsigned long section_mem_map;
1095
1096        /* See declaration of similar field in struct zone */
1097        unsigned long *pageblock_flags;
1098#ifdef CONFIG_MEMCG
1099        /*
1100         * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1101         * section. (see memcontrol.h/page_cgroup.h about this.)
1102         */
1103        struct page_cgroup *page_cgroup;
1104        unsigned long pad;
1105#endif
1106};
1107
1108#ifdef CONFIG_SPARSEMEM_EXTREME
1109#define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
1110#else
1111#define SECTIONS_PER_ROOT       1
1112#endif
1113
1114#define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1115#define NR_SECTION_ROOTS        DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1116#define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
1117
1118#ifdef CONFIG_SPARSEMEM_EXTREME
1119extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1120#else
1121extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1122#endif
1123
1124static inline struct mem_section *__nr_to_section(unsigned long nr)
1125{
1126        if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1127                return NULL;
1128        return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1129}
1130extern int __section_nr(struct mem_section* ms);
1131extern unsigned long usemap_size(void);
1132
1133/*
1134 * We use the lower bits of the mem_map pointer to store
1135 * a little bit of information.  There should be at least
1136 * 3 bits here due to 32-bit alignment.
1137 */
1138#define SECTION_MARKED_PRESENT  (1UL<<0)
1139#define SECTION_HAS_MEM_MAP     (1UL<<1)
1140#define SECTION_MAP_LAST_BIT    (1UL<<2)
1141#define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
1142#define SECTION_NID_SHIFT       2
1143
1144static inline struct page *__section_mem_map_addr(struct mem_section *section)
1145{
1146        unsigned long map = section->section_mem_map;
1147        map &= SECTION_MAP_MASK;
1148        return (struct page *)map;
1149}
1150
1151static inline int present_section(struct mem_section *section)
1152{
1153        return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1154}
1155
1156static inline int present_section_nr(unsigned long nr)
1157{
1158        return present_section(__nr_to_section(nr));
1159}
1160
1161static inline int valid_section(struct mem_section *section)
1162{
1163        return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1164}
1165
1166static inline int valid_section_nr(unsigned long nr)
1167{
1168        return valid_section(__nr_to_section(nr));
1169}
1170
1171static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1172{
1173        return __nr_to_section(pfn_to_section_nr(pfn));
1174}
1175
1176#ifndef CONFIG_HAVE_ARCH_PFN_VALID
1177static inline int pfn_valid(unsigned long pfn)
1178{
1179        if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1180                return 0;
1181        return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1182}
1183#endif
1184
1185static inline int pfn_present(unsigned long pfn)
1186{
1187        if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1188                return 0;
1189        return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1190}
1191
1192/*
1193 * These are _only_ used during initialisation, therefore they
1194 * can use __initdata ...  They could have names to indicate
1195 * this restriction.
1196 */
1197#ifdef CONFIG_NUMA
1198#define pfn_to_nid(pfn)                                                 \
1199({                                                                      \
1200        unsigned long __pfn_to_nid_pfn = (pfn);                         \
1201        page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
1202})
1203#else
1204#define pfn_to_nid(pfn)         (0)
1205#endif
1206
1207#define early_pfn_valid(pfn)    pfn_valid(pfn)
1208void sparse_init(void);
1209#else
1210#define sparse_init()   do {} while (0)
1211#define sparse_index_init(_sec, _nid)  do {} while (0)
1212#endif /* CONFIG_SPARSEMEM */
1213
1214#ifdef CONFIG_NODES_SPAN_OTHER_NODES
1215bool early_pfn_in_nid(unsigned long pfn, int nid);
1216#else
1217#define early_pfn_in_nid(pfn, nid)      (1)
1218#endif
1219
1220#ifndef early_pfn_valid
1221#define early_pfn_valid(pfn)    (1)
1222#endif
1223
1224void memory_present(int nid, unsigned long start, unsigned long end);
1225unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1226
1227/*
1228 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1229 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1230 * pfn_valid_within() should be used in this case; we optimise this away
1231 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1232 */
1233#ifdef CONFIG_HOLES_IN_ZONE
1234#define pfn_valid_within(pfn) pfn_valid(pfn)
1235#else
1236#define pfn_valid_within(pfn) (1)
1237#endif
1238
1239#ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1240/*
1241 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1242 * associated with it or not. In FLATMEM, it is expected that holes always
1243 * have valid memmap as long as there is valid PFNs either side of the hole.
1244 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1245 * entire section.
1246 *
1247 * However, an ARM, and maybe other embedded architectures in the future
1248 * free memmap backing holes to save memory on the assumption the memmap is
1249 * never used. The page_zone linkages are then broken even though pfn_valid()
1250 * returns true. A walker of the full memmap must then do this additional
1251 * check to ensure the memmap they are looking at is sane by making sure
1252 * the zone and PFN linkages are still valid. This is expensive, but walkers
1253 * of the full memmap are extremely rare.
1254 */
1255int memmap_valid_within(unsigned long pfn,
1256                                        struct page *page, struct zone *zone);
1257#else
1258static inline int memmap_valid_within(unsigned long pfn,
1259                                        struct page *page, struct zone *zone)
1260{
1261        return 1;
1262}
1263#endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1264
1265#endif /* !__GENERATING_BOUNDS.H */
1266#endif /* !__ASSEMBLY__ */
1267#endif /* _LINUX_MMZONE_H */
1268
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.