include/linux/mmzone.h

   1 #ifndef _LINUX_MMZONE_H
   2 #define _LINUX_MMZONE_H
   3
   4 #ifdef __KERNEL__
   5 #ifndef __ASSEMBLY__
   6
   7 #include <linux/spinlock.h>
   8 #include <linux/list.h>
   9 #include <linux/wait.h>
  10 #include <linux/cache.h>
  11 #include <linux/threads.h>
  12 #include <linux/numa.h>
  13 #include <linux/init.h>
  14 #include <linux/seqlock.h>
  15 #include <linux/nodemask.h>
  16 #include <asm/atomic.h>
  17 #include <asm/page.h>
  18
  19 /* Free memory management - zoned buddy allocator.  */
  20 #ifndef CONFIG_FORCE_MAX_ZONEORDER
  21 #define MAX_ORDER 11
  22 #else
  23 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
  24 #endif
  25 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
  26
  27 struct free_area {
  28         struct list_head        free_list;
  29         unsigned long           nr_free;
  30 };
  31
  32 struct pglist_data;
  33
  34 /*
  35  * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
  36  * So add a wild amount of padding here to ensure that they fall into separate
  37  * cachelines.  There are very few zone structures in the machine, so space
  38  * consumption is not a concern here.
  39  */
  40 #if defined(CONFIG_SMP)
  41 struct zone_padding {
  42         char x[0];
  43 } ____cacheline_internodealigned_in_smp;
  44 #define ZONE_PADDING(name)      struct zone_padding name;
  45 #else
  46 #define ZONE_PADDING(name)
  47 #endif
  48
  49 enum zone_stat_item {
  50         NR_ANON_PAGES,  /* Mapped anonymous pages */
  51         NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
  52                            only modified from process context */
  53         NR_FILE_PAGES,
  54         NR_SLAB,        /* Pages used by slab allocator */
  55         NR_PAGETABLE,   /* used for pagetables */
  56         NR_FILE_DIRTY,
  57         NR_WRITEBACK,
  58         NR_UNSTABLE_NFS,        /* NFS unstable pages */
  59         NR_BOUNCE,
  60 #ifdef CONFIG_NUMA
  61         NUMA_HIT,               /* allocated in intended node */
  62         NUMA_MISS,              /* allocated in non intended node */
  63         NUMA_FOREIGN,           /* was intended here, hit elsewhere */
  64         NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
  65         NUMA_LOCAL,             /* allocation from local node */
  66         NUMA_OTHER,             /* allocation from other node */
  67 #endif
  68         NR_VM_ZONE_STAT_ITEMS };
  69
  70 struct per_cpu_pages {
  71         int count;              /* number of pages in the list */
  72         int high;               /* high watermark, emptying needed */
  73         int batch;              /* chunk size for buddy add/remove */
  74         struct list_head list;  /* the list of pages */
  75 };
  76
  77 struct per_cpu_pageset {
  78         struct per_cpu_pages pcp[2];    /* 0: hot.  1: cold */
  79 #ifdef CONFIG_SMP
  80         s8 stat_threshold;
  81         s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
  82 #endif
  83 } ____cacheline_aligned_in_smp;
  84
  85 #ifdef CONFIG_NUMA
  86 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
  87 #else
  88 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
  89 #endif
  90
  91 enum zone_type {
  92         /*
  93          * ZONE_DMA is used when there are devices that are not able
  94          * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
  95          * carve out the portion of memory that is needed for these devices.
  96          * The range is arch specific.
  97          *
  98          * Some examples
  99          *
 100          * Architecture         Limit
 101          * ---------------------------
 102          * parisc, ia64, sparc  <4G
 103          * s390                 <2G
 104          * arm26                <48M
 105          * arm                  Various
 106          * alpha                Unlimited or 0-16MB.
 107          *
 108          * i386, x86_64 and multiple other arches
 109          *                      <16M.
 110          */
 111         ZONE_DMA,
 112 #ifdef CONFIG_ZONE_DMA32
 113         /*
 114          * x86_64 needs two ZONE_DMAs because it supports devices that are
 115          * only able to do DMA to the lower 16M but also 32 bit devices that
 116          * can only do DMA areas below 4G.
 117          */
 118         ZONE_DMA32,
 119 #endif
 120         /*
 121          * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
 122          * performed on pages in ZONE_NORMAL if the DMA devices support
 123          * transfers to all addressable memory.
 124          */
 125         ZONE_NORMAL,
 126 #ifdef CONFIG_HIGHMEM
 127         /*
 128          * A memory area that is only addressable by the kernel through
 129          * mapping portions into its own address space. This is for example
 130          * used by i386 to allow the kernel to address the memory beyond
 131          * 900MB. The kernel will set up special mappings (page
 132          * table entries on i386) for each page that the kernel needs to
 133          * access.
 134          */
 135         ZONE_HIGHMEM,
 136 #endif
 137         MAX_NR_ZONES
 138 };
 139
 140 /*
 141  * When a memory allocation must conform to specific limitations (such
 142  * as being suitable for DMA) the caller will pass in hints to the
 143  * allocator in the gfp_mask, in the zone modifier bits.  These bits
 144  * are used to select a priority ordered list of memory zones which
 145  * match the requested limits. See gfp_zone() in include/linux/gfp.h
 146  */
 147
 148 #if !defined(CONFIG_ZONE_DMA32) && !defined(CONFIG_HIGHMEM)
 149 #define ZONES_SHIFT 1
 150 #else
 151 #define ZONES_SHIFT 2
 152 #endif
 153
 154 struct zone {
 155         /* Fields commonly accessed by the page allocator */
 156         unsigned long           free_pages;
 157         unsigned long           pages_min, pages_low, pages_high;
 158         /*
 159          * We don't know if the memory that we're going to allocate will be freeable
 160          * or/and it will be released eventually, so to avoid totally wasting several
 161          * GB of ram we must reserve some of the lower zone memory (otherwise we risk
 162          * to run OOM on the lower zones despite there's tons of freeable ram
 163          * on the higher zones). This array is recalculated at runtime if the
 164          * sysctl_lowmem_reserve_ratio sysctl changes.
 165          */
 166         unsigned long           lowmem_reserve[MAX_NR_ZONES];
 167
 168 #ifdef CONFIG_NUMA
 169         /*
 170          * zone reclaim becomes active if more unmapped pages exist.
 171          */
 172         unsigned long           min_unmapped_pages;
 173         struct per_cpu_pageset  *pageset[NR_CPUS];
 174 #else
 175         struct per_cpu_pageset  pageset[NR_CPUS];
 176 #endif
 177         /*
 178          * free areas of different sizes
 179          */
 180         spinlock_t              lock;
 181 #ifdef CONFIG_MEMORY_HOTPLUG
 182         /* see spanned/present_pages for more description */
 183         seqlock_t               span_seqlock;
 184 #endif
 185         struct free_area        free_area[MAX_ORDER];
 186
 187
 188         ZONE_PADDING(_pad1_)
 189
 190         /* Fields commonly accessed by the page reclaim scanner */
 191         spinlock_t              lru_lock;
 192         struct list_head        active_list;
 193         struct list_head        inactive_list;
 194         unsigned long           nr_scan_active;
 195         unsigned long           nr_scan_inactive;
 196         unsigned long           nr_active;
 197         unsigned long           nr_inactive;
 198         unsigned long           pages_scanned;     /* since last reclaim */
 199         int                     all_unreclaimable; /* All pages pinned */
 200
 201         /* A count of how many reclaimers are scanning this zone */
 202         atomic_t                reclaim_in_progress;
 203
 204         /* Zone statistics */
 205         atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
 206
 207         /*
 208          * prev_priority holds the scanning priority for this zone.  It is
 209          * defined as the scanning priority at which we achieved our reclaim
 210          * target at the previous try_to_free_pages() or balance_pgdat()
 211          * invokation.
 212          *
 213          * We use prev_priority as a measure of how much stress page reclaim is
 214          * under - it drives the swappiness decision: whether to unmap mapped
 215          * pages.
 216          *
 217          * temp_priority is used to remember the scanning priority at which
 218          * this zone was successfully refilled to free_pages == pages_high.
 219          *
 220          * Access to both these fields is quite racy even on uniprocessor.  But
 221          * it is expected to average out OK.
 222          */
 223         int temp_priority;
 224         int prev_priority;
 225
 226
 227         ZONE_PADDING(_pad2_)
 228         /* Rarely used or read-mostly fields */
 229
 230         /*
 231          * wait_table           -- the array holding the hash table
 232          * wait_table_hash_nr_entries   -- the size of the hash table array
 233          * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
 234          *
 235          * The purpose of all these is to keep track of the people
 236          * waiting for a page to become available and make them
 237          * runnable again when possible. The trouble is that this
 238          * consumes a lot of space, especially when so few things
 239          * wait on pages at a given time. So instead of using
 240          * per-page waitqueues, we use a waitqueue hash table.
 241          *
 242          * The bucket discipline is to sleep on the same queue when
 243          * colliding and wake all in that wait queue when removing.
 244          * When something wakes, it must check to be sure its page is
 245          * truly available, a la thundering herd. The cost of a
 246          * collision is great, but given the expected load of the
 247          * table, they should be so rare as to be outweighed by the
 248          * benefits from the saved space.
 249          *
 250          * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
 251          * primary users of these fields, and in mm/page_alloc.c
 252          * free_area_init_core() performs the initialization of them.
 253          */
 254         wait_queue_head_t       * wait_table;
 255         unsigned long           wait_table_hash_nr_entries;
 256         unsigned long           wait_table_bits;
 257
 258         /*
 259          * Discontig memory support fields.
 260          */
 261         struct pglist_data      *zone_pgdat;
 262         /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
 263         unsigned long           zone_start_pfn;
 264
 265         /*
 266          * zone_start_pfn, spanned_pages and present_pages are all
 267          * protected by span_seqlock.  It is a seqlock because it has
 268          * to be read outside of zone->lock, and it is done in the main
 269          * allocator path.  But, it is written quite infrequently.
 270          *
 271          * The lock is declared along with zone->lock because it is
 272          * frequently read in proximity to zone->lock.  It's good to
 273          * give them a chance of being in the same cacheline.
 274          */
 275         unsigned long           spanned_pages;  /* total size, including holes */
 276         unsigned long           present_pages;  /* amount of memory (excluding holes) */
 277
 278         /*
 279          * rarely used fields:
 280          */
 281         char                    *name;
 282 } ____cacheline_internodealigned_in_smp;
 283
 284 /*
 285  * The "priority" of VM scanning is how much of the queues we will scan in one
 286  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
 287  * queues ("queue_length >> 12") during an aging round.
 288  */
 289 #define DEF_PRIORITY 12
 290
 291 /*
 292  * One allocation request operates on a zonelist. A zonelist
 293  * is a list of zones, the first one is the 'goal' of the
 294  * allocation, the other zones are fallback zones, in decreasing
 295  * priority.
 296  *
 297  * Right now a zonelist takes up less than a cacheline. We never
 298  * modify it apart from boot-up, and only a few indices are used,
 299  * so despite the zonelist table being relatively big, the cache
 300  * footprint of this construct is very small.
 301  */
 302 struct zonelist {
 303         struct zone *zones[MAX_NUMNODES * MAX_NR_ZONES + 1]; // NULL delimited
 304 };
 305
 306
 307 /*
 308  * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
 309  * (mostly NUMA machines?) to denote a higher-level memory zone than the
 310  * zone denotes.
 311  *
 312  * On NUMA machines, each NUMA node would have a pg_data_t to describe
 313  * it's memory layout.
 314  *
 315  * Memory statistics and page replacement data structures are maintained on a
 316  * per-zone basis.
 317  */
 318 struct bootmem_data;
 319 typedef struct pglist_data {
 320         struct zone node_zones[MAX_NR_ZONES];
 321         struct zonelist node_zonelists[MAX_NR_ZONES];
 322         int nr_zones;
 323 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 324         struct page *node_mem_map;
 325 #endif
 326         struct bootmem_data *bdata;
 327 #ifdef CONFIG_MEMORY_HOTPLUG
 328         /*
 329          * Must be held any time you expect node_start_pfn, node_present_pages
 330          * or node_spanned_pages stay constant.  Holding this will also
 331          * guarantee that any pfn_valid() stays that way.
 332          *
 333          * Nests above zone->lock and zone->size_seqlock.
 334          */
 335         spinlock_t node_size_lock;
 336 #endif
 337         unsigned long node_start_pfn;
 338         unsigned long node_present_pages; /* total number of physical pages */
 339         unsigned long node_spanned_pages; /* total size of physical page
 340                                              range, including holes */
 341         int node_id;
 342         wait_queue_head_t kswapd_wait;
 343         struct task_struct *kswapd;
 344         int kswapd_max_order;
 345 } pg_data_t;
 346
 347 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
 348 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
 349 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 350 #define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
 351 #else
 352 #define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
 353 #endif
 354 #define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
 355
 356 #include <linux/memory_hotplug.h>
 357
 358 void __get_zone_counts(unsigned long *active, unsigned long *inactive,
 359                         unsigned long *free, struct pglist_data *pgdat);
 360 void get_zone_counts(unsigned long *active, unsigned long *inactive,
 361                         unsigned long *free);
 362 void build_all_zonelists(void);
 363 void wakeup_kswapd(struct zone *zone, int order);
 364 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
 365                 int classzone_idx, int alloc_flags);
 366
 367 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
 368                                      unsigned long size);
 369
 370 #ifdef CONFIG_HAVE_MEMORY_PRESENT
 371 void memory_present(int nid, unsigned long start, unsigned long end);
 372 #else
 373 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
 374 #endif
 375
 376 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
 377 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 378 #endif
 379
 380 /*
 381  * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
 382  */
 383 #define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
 384
 385 static inline int populated_zone(struct zone *zone)
 386 {
 387         return (!!zone->present_pages);
 388 }
 389
 390 static inline int is_highmem_idx(enum zone_type idx)
 391 {
 392 #ifdef CONFIG_HIGHMEM
 393         return (idx == ZONE_HIGHMEM);
 394 #else
 395         return 0;
 396 #endif
 397 }
 398
 399 static inline int is_normal_idx(enum zone_type idx)
 400 {
 401         return (idx == ZONE_NORMAL);
 402 }
 403
 404 /**
 405  * is_highmem - helper function to quickly check if a struct zone is a
 406  *              highmem zone or not.  This is an attempt to keep references
 407  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
 408  * @zone - pointer to struct zone variable
 409  */
 410 static inline int is_highmem(struct zone *zone)
 411 {
 412 #ifdef CONFIG_HIGHMEM
 413         return zone == zone->zone_pgdat->node_zones + ZONE_HIGHMEM;
 414 #else
 415         return 0;
 416 #endif
 417 }
 418
 419 static inline int is_normal(struct zone *zone)
 420 {
 421         return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
 422 }
 423
 424 static inline int is_dma32(struct zone *zone)
 425 {
 426 #ifdef CONFIG_ZONE_DMA32
 427         return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
 428 #else
 429         return 0;
 430 #endif
 431 }
 432
 433 static inline int is_dma(struct zone *zone)
 434 {
 435         return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
 436 }
 437
 438 /* These two functions are used to setup the per zone pages min values */
 439 struct ctl_table;
 440 struct file;
 441 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
 442                                         void __user *, size_t *, loff_t *);
 443 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
 444 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
 445                                         void __user *, size_t *, loff_t *);
 446 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
 447                                         void __user *, size_t *, loff_t *);
 448 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
 449                         struct file *, void __user *, size_t *, loff_t *);
 450
 451 #include <linux/topology.h>
 452 /* Returns the number of the current Node. */
 453 #ifndef numa_node_id
 454 #define numa_node_id()          (cpu_to_node(raw_smp_processor_id()))
 455 #endif
 456
 457 #ifndef CONFIG_NEED_MULTIPLE_NODES
 458
 459 extern struct pglist_data contig_page_data;
 460 #define NODE_DATA(nid)          (&contig_page_data)
 461 #define NODE_MEM_MAP(nid)       mem_map
 462 #define MAX_NODES_SHIFT         1
 463
 464 #else /* CONFIG_NEED_MULTIPLE_NODES */
 465
 466 #include <asm/mmzone.h>
 467
 468 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
 469
 470 extern struct pglist_data *first_online_pgdat(void);
 471 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
 472 extern struct zone *next_zone(struct zone *zone);
 473
 474 /**
 475  * for_each_pgdat - helper macro to iterate over all nodes
 476  * @pgdat - pointer to a pg_data_t variable
 477  */
 478 #define for_each_online_pgdat(pgdat)                    \
 479         for (pgdat = first_online_pgdat();              \
 480              pgdat;                                     \
 481              pgdat = next_online_pgdat(pgdat))
 482 /**
 483  * for_each_zone - helper macro to iterate over all memory zones
 484  * @zone - pointer to struct zone variable
 485  *
 486  * The user only needs to declare the zone variable, for_each_zone
 487  * fills it in.
 488  */
 489 #define for_each_zone(zone)                             \
 490         for (zone = (first_online_pgdat())->node_zones; \
 491              zone;                                      \
 492              zone = next_zone(zone))
 493
 494 #ifdef CONFIG_SPARSEMEM
 495 #include <asm/sparsemem.h>
 496 #endif
 497
 498 #if BITS_PER_LONG == 32
 499 /*
 500  * with 32 bit page->flags field, we reserve 9 bits for node/zone info.
 501  * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes.
 502  */
 503 #define FLAGS_RESERVED          9
 504
 505 #elif BITS_PER_LONG == 64
 506 /*
 507  * with 64 bit flags field, there's plenty of room.
 508  */
 509 #define FLAGS_RESERVED          32
 510
 511 #else
 512
 513 #error BITS_PER_LONG not defined
 514
 515 #endif
 516
 517 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
 518 #define early_pfn_to_nid(nid)  (0UL)
 519 #endif
 520
 521 #ifdef CONFIG_FLATMEM
 522 #define pfn_to_nid(pfn)         (0)
 523 #endif
 524
 525 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
 526 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
 527
 528 #ifdef CONFIG_SPARSEMEM
 529
 530 /*
 531  * SECTION_SHIFT                #bits space required to store a section #
 532  *
 533  * PA_SECTION_SHIFT             physical address to/from section number
 534  * PFN_SECTION_SHIFT            pfn to/from section number
 535  */
 536 #define SECTIONS_SHIFT          (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
 537
 538 #define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
 539 #define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
 540
 541 #define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
 542
 543 #define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
 544 #define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
 545
 546 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
 547 #error Allocator MAX_ORDER exceeds SECTION_SIZE
 548 #endif
 549
 550 struct page;
 551 struct mem_section {
 552         /*
 553          * This is, logically, a pointer to an array of struct
 554          * pages.  However, it is stored with some other magic.
 555          * (see sparse.c::sparse_init_one_section())
 556          *
 557          * Additionally during early boot we encode node id of
 558          * the location of the section here to guide allocation.
 559          * (see sparse.c::memory_present())
 560          *
 561          * Making it a UL at least makes someone do a cast
 562          * before using it wrong.
 563          */
 564         unsigned long section_mem_map;
 565 };
 566
 567 #ifdef CONFIG_SPARSEMEM_EXTREME
 568 #define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
 569 #else
 570 #define SECTIONS_PER_ROOT       1
 571 #endif
 572
 573 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
 574 #define NR_SECTION_ROOTS        (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
 575 #define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
 576
 577 #ifdef CONFIG_SPARSEMEM_EXTREME
 578 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
 579 #else
 580 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
 581 #endif
 582
 583 static inline struct mem_section *__nr_to_section(unsigned long nr)
 584 {
 585         if (!mem_section[SECTION_NR_TO_ROOT(nr)])
 586                 return NULL;
 587         return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
 588 }
 589 extern int __section_nr(struct mem_section* ms);
 590
 591 /*
 592  * We use the lower bits of the mem_map pointer to store
 593  * a little bit of information.  There should be at least
 594  * 3 bits here due to 32-bit alignment.
 595  */
 596 #define SECTION_MARKED_PRESENT  (1UL<<0)
 597 #define SECTION_HAS_MEM_MAP     (1UL<<1)
 598 #define SECTION_MAP_LAST_BIT    (1UL<<2)
 599 #define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
 600 #define SECTION_NID_SHIFT       2
 601
 602 static inline struct page *__section_mem_map_addr(struct mem_section *section)
 603 {
 604         unsigned long map = section->section_mem_map;
 605         map &= SECTION_MAP_MASK;
 606         return (struct page *)map;
 607 }
 608
 609 static inline int valid_section(struct mem_section *section)
 610 {
 611         return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
 612 }
 613
 614 static inline int section_has_mem_map(struct mem_section *section)
 615 {
 616         return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
 617 }
 618
 619 static inline int valid_section_nr(unsigned long nr)
 620 {
 621         return valid_section(__nr_to_section(nr));
 622 }
 623
 624 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
 625 {
 626         return __nr_to_section(pfn_to_section_nr(pfn));
 627 }
 628
 629 static inline int pfn_valid(unsigned long pfn)
 630 {
 631         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
 632                 return 0;
 633         return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
 634 }
 635
 636 /*
 637  * These are _only_ used during initialisation, therefore they
 638  * can use __initdata ...  They could have names to indicate
 639  * this restriction.
 640  */
 641 #ifdef CONFIG_NUMA
 642 #define pfn_to_nid(pfn)                                                 \
 643 ({                                                                      \
 644         unsigned long __pfn_to_nid_pfn = (pfn);                         \
 645         page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
 646 })
 647 #else
 648 #define pfn_to_nid(pfn)         (0)
 649 #endif
 650
 651 #define early_pfn_valid(pfn)    pfn_valid(pfn)
 652 void sparse_init(void);
 653 #else
 654 #define sparse_init()   do {} while (0)
 655 #define sparse_index_init(_sec, _nid)  do {} while (0)
 656 #endif /* CONFIG_SPARSEMEM */
 657
 658 #ifndef early_pfn_valid
 659 #define early_pfn_valid(pfn)    (1)
 660 #endif
 661
 662 void memory_present(int nid, unsigned long start, unsigned long end);
 663 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 664
 665 #endif /* !__ASSEMBLY__ */
 666 #endif /* __KERNEL__ */
 667 #endif /* _LINUX_MMZONE_H */