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 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
  81 #endif
  82 } ____cacheline_aligned_in_smp;
  83
  84 #ifdef CONFIG_NUMA
  85 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
  86 #else
  87 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
  88 #endif
  89
  90 #define ZONE_DMA                0
  91 #define ZONE_DMA32              1
  92 #define ZONE_NORMAL             2
  93 #define ZONE_HIGHMEM            3
  94
  95 #define MAX_NR_ZONES            4       /* Sync this with ZONES_SHIFT */
  96 #define ZONES_SHIFT             2       /* ceil(log2(MAX_NR_ZONES)) */
  97
  98
  99 /*
 100  * When a memory allocation must conform to specific limitations (such
 101  * as being suitable for DMA) the caller will pass in hints to the
 102  * allocator in the gfp_mask, in the zone modifier bits.  These bits
 103  * are used to select a priority ordered list of memory zones which
 104  * match the requested limits.  GFP_ZONEMASK defines which bits within
 105  * the gfp_mask should be considered as zone modifiers.  Each valid
 106  * combination of the zone modifier bits has a corresponding list
 107  * of zones (in node_zonelists).  Thus for two zone modifiers there
 108  * will be a maximum of 4 (2 ** 2) zonelists, for 3 modifiers there will
 109  * be 8 (2 ** 3) zonelists.  GFP_ZONETYPES defines the number of possible
 110  * combinations of zone modifiers in "zone modifier space".
 111  *
 112  * As an optimisation any zone modifier bits which are only valid when
 113  * no other zone modifier bits are set (loners) should be placed in
 114  * the highest order bits of this field.  This allows us to reduce the
 115  * extent of the zonelists thus saving space.  For example in the case
 116  * of three zone modifier bits, we could require up to eight zonelists.
 117  * If the left most zone modifier is a "loner" then the highest valid
 118  * zonelist would be four allowing us to allocate only five zonelists.
 119  * Use the first form for GFP_ZONETYPES when the left most bit is not
 120  * a "loner", otherwise use the second.
 121  *
 122  * NOTE! Make sure this matches the zones in <linux/gfp.h>
 123  */
 124 #define GFP_ZONEMASK    0x07
 125 /* #define GFP_ZONETYPES       (GFP_ZONEMASK + 1) */           /* Non-loner */
 126 #define GFP_ZONETYPES  ((GFP_ZONEMASK + 1) / 2 + 1)            /* Loner */
 127
 128 /*
 129  * On machines where it is needed (eg PCs) we divide physical memory
 130  * into multiple physical zones. On a 32bit PC we have 4 zones:
 131  *
 132  * ZONE_DMA       < 16 MB       ISA DMA capable memory
 133  * ZONE_DMA32        0 MB       Empty
 134  * ZONE_NORMAL  16-896 MB       direct mapped by the kernel
 135  * ZONE_HIGHMEM  > 896 MB       only page cache and user processes
 136  */
 137
 138 struct zone {
 139         /* Fields commonly accessed by the page allocator */
 140         unsigned long           free_pages;
 141         unsigned long           pages_min, pages_low, pages_high;
 142         /*
 143          * We don't know if the memory that we're going to allocate will be freeable
 144          * or/and it will be released eventually, so to avoid totally wasting several
 145          * GB of ram we must reserve some of the lower zone memory (otherwise we risk
 146          * to run OOM on the lower zones despite there's tons of freeable ram
 147          * on the higher zones). This array is recalculated at runtime if the
 148          * sysctl_lowmem_reserve_ratio sysctl changes.
 149          */
 150         unsigned long           lowmem_reserve[MAX_NR_ZONES];
 151
 152 #ifdef CONFIG_NUMA
 153         /*
 154          * zone reclaim becomes active if more unmapped pages exist.
 155          */
 156         unsigned long           min_unmapped_ratio;
 157         struct per_cpu_pageset  *pageset[NR_CPUS];
 158 #else
 159         struct per_cpu_pageset  pageset[NR_CPUS];
 160 #endif
 161         /*
 162          * free areas of different sizes
 163          */
 164         spinlock_t              lock;
 165 #ifdef CONFIG_MEMORY_HOTPLUG
 166         /* see spanned/present_pages for more description */
 167         seqlock_t               span_seqlock;
 168 #endif
 169         struct free_area        free_area[MAX_ORDER];
 170
 171
 172         ZONE_PADDING(_pad1_)
 173
 174         /* Fields commonly accessed by the page reclaim scanner */
 175         spinlock_t              lru_lock;
 176         struct list_head        active_list;
 177         struct list_head        inactive_list;
 178         unsigned long           nr_scan_active;
 179         unsigned long           nr_scan_inactive;
 180         unsigned long           nr_active;
 181         unsigned long           nr_inactive;
 182         unsigned long           pages_scanned;     /* since last reclaim */
 183         int                     all_unreclaimable; /* All pages pinned */
 184
 185         /* A count of how many reclaimers are scanning this zone */
 186         atomic_t                reclaim_in_progress;
 187
 188         /* Zone statistics */
 189         atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
 190
 191         /*
 192          * prev_priority holds the scanning priority for this zone.  It is
 193          * defined as the scanning priority at which we achieved our reclaim
 194          * target at the previous try_to_free_pages() or balance_pgdat()
 195          * invokation.
 196          *
 197          * We use prev_priority as a measure of how much stress page reclaim is
 198          * under - it drives the swappiness decision: whether to unmap mapped
 199          * pages.
 200          *
 201          * temp_priority is used to remember the scanning priority at which
 202          * this zone was successfully refilled to free_pages == pages_high.
 203          *
 204          * Access to both these fields is quite racy even on uniprocessor.  But
 205          * it is expected to average out OK.
 206          */
 207         int temp_priority;
 208         int prev_priority;
 209
 210
 211         ZONE_PADDING(_pad2_)
 212         /* Rarely used or read-mostly fields */
 213
 214         /*
 215          * wait_table           -- the array holding the hash table
 216          * wait_table_hash_nr_entries   -- the size of the hash table array
 217          * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
 218          *
 219          * The purpose of all these is to keep track of the people
 220          * waiting for a page to become available and make them
 221          * runnable again when possible. The trouble is that this
 222          * consumes a lot of space, especially when so few things
 223          * wait on pages at a given time. So instead of using
 224          * per-page waitqueues, we use a waitqueue hash table.
 225          *
 226          * The bucket discipline is to sleep on the same queue when
 227          * colliding and wake all in that wait queue when removing.
 228          * When something wakes, it must check to be sure its page is
 229          * truly available, a la thundering herd. The cost of a
 230          * collision is great, but given the expected load of the
 231          * table, they should be so rare as to be outweighed by the
 232          * benefits from the saved space.
 233          *
 234          * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
 235          * primary users of these fields, and in mm/page_alloc.c
 236          * free_area_init_core() performs the initialization of them.
 237          */
 238         wait_queue_head_t       * wait_table;
 239         unsigned long           wait_table_hash_nr_entries;
 240         unsigned long           wait_table_bits;
 241
 242         /*
 243          * Discontig memory support fields.
 244          */
 245         struct pglist_data      *zone_pgdat;
 246         /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
 247         unsigned long           zone_start_pfn;
 248
 249         /*
 250          * zone_start_pfn, spanned_pages and present_pages are all
 251          * protected by span_seqlock.  It is a seqlock because it has
 252          * to be read outside of zone->lock, and it is done in the main
 253          * allocator path.  But, it is written quite infrequently.
 254          *
 255          * The lock is declared along with zone->lock because it is
 256          * frequently read in proximity to zone->lock.  It's good to
 257          * give them a chance of being in the same cacheline.
 258          */
 259         unsigned long           spanned_pages;  /* total size, including holes */
 260         unsigned long           present_pages;  /* amount of memory (excluding holes) */
 261
 262         /*
 263          * rarely used fields:
 264          */
 265         char                    *name;
 266 } ____cacheline_internodealigned_in_smp;
 267
 268
 269 /*
 270  * The "priority" of VM scanning is how much of the queues we will scan in one
 271  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
 272  * queues ("queue_length >> 12") during an aging round.
 273  */
 274 #define DEF_PRIORITY 12
 275
 276 /*
 277  * One allocation request operates on a zonelist. A zonelist
 278  * is a list of zones, the first one is the 'goal' of the
 279  * allocation, the other zones are fallback zones, in decreasing
 280  * priority.
 281  *
 282  * Right now a zonelist takes up less than a cacheline. We never
 283  * modify it apart from boot-up, and only a few indices are used,
 284  * so despite the zonelist table being relatively big, the cache
 285  * footprint of this construct is very small.
 286  */
 287 struct zonelist {
 288         struct zone *zones[MAX_NUMNODES * MAX_NR_ZONES + 1]; // NULL delimited
 289 };
 290
 291
 292 /*
 293  * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
 294  * (mostly NUMA machines?) to denote a higher-level memory zone than the
 295  * zone denotes.
 296  *
 297  * On NUMA machines, each NUMA node would have a pg_data_t to describe
 298  * it's memory layout.
 299  *
 300  * Memory statistics and page replacement data structures are maintained on a
 301  * per-zone basis.
 302  */
 303 struct bootmem_data;
 304 typedef struct pglist_data {
 305         struct zone node_zones[MAX_NR_ZONES];
 306         struct zonelist node_zonelists[GFP_ZONETYPES];
 307         int nr_zones;
 308 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 309         struct page *node_mem_map;
 310 #endif
 311         struct bootmem_data *bdata;
 312 #ifdef CONFIG_MEMORY_HOTPLUG
 313         /*
 314          * Must be held any time you expect node_start_pfn, node_present_pages
 315          * or node_spanned_pages stay constant.  Holding this will also
 316          * guarantee that any pfn_valid() stays that way.
 317          *
 318          * Nests above zone->lock and zone->size_seqlock.
 319          */
 320         spinlock_t node_size_lock;
 321 #endif
 322         unsigned long node_start_pfn;
 323         unsigned long node_present_pages; /* total number of physical pages */
 324         unsigned long node_spanned_pages; /* total size of physical page
 325                                              range, including holes */
 326         int node_id;
 327         wait_queue_head_t kswapd_wait;
 328         struct task_struct *kswapd;
 329         int kswapd_max_order;
 330 } pg_data_t;
 331
 332 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
 333 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
 334 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 335 #define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
 336 #else
 337 #define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
 338 #endif
 339 #define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
 340
 341 #include <linux/memory_hotplug.h>
 342
 343 void __get_zone_counts(unsigned long *active, unsigned long *inactive,
 344                         unsigned long *free, struct pglist_data *pgdat);
 345 void get_zone_counts(unsigned long *active, unsigned long *inactive,
 346                         unsigned long *free);
 347 void build_all_zonelists(void);
 348 void wakeup_kswapd(struct zone *zone, int order);
 349 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
 350                 int classzone_idx, int alloc_flags);
 351
 352 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
 353                                      unsigned long size);
 354
 355 #ifdef CONFIG_HAVE_MEMORY_PRESENT
 356 void memory_present(int nid, unsigned long start, unsigned long end);
 357 #else
 358 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
 359 #endif
 360
 361 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
 362 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 363 #endif
 364
 365 /*
 366  * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
 367  */
 368 #define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
 369
 370 static inline int populated_zone(struct zone *zone)
 371 {
 372         return (!!zone->present_pages);
 373 }
 374
 375 static inline int is_highmem_idx(int idx)
 376 {
 377         return (idx == ZONE_HIGHMEM);
 378 }
 379
 380 static inline int is_normal_idx(int idx)
 381 {
 382         return (idx == ZONE_NORMAL);
 383 }
 384
 385 /**
 386  * is_highmem - helper function to quickly check if a struct zone is a
 387  *              highmem zone or not.  This is an attempt to keep references
 388  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
 389  * @zone - pointer to struct zone variable
 390  */
 391 static inline int is_highmem(struct zone *zone)
 392 {
 393         return zone == zone->zone_pgdat->node_zones + ZONE_HIGHMEM;
 394 }
 395
 396 static inline int is_normal(struct zone *zone)
 397 {
 398         return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
 399 }
 400
 401 static inline int is_dma32(struct zone *zone)
 402 {
 403         return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
 404 }
 405
 406 static inline int is_dma(struct zone *zone)
 407 {
 408         return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
 409 }
 410
 411 /* These two functions are used to setup the per zone pages min values */
 412 struct ctl_table;
 413 struct file;
 414 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
 415                                         void __user *, size_t *, loff_t *);
 416 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
 417 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
 418                                         void __user *, size_t *, loff_t *);
 419 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
 420                                         void __user *, size_t *, loff_t *);
 421 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
 422                         struct file *, void __user *, size_t *, loff_t *);
 423
 424 #include <linux/topology.h>
 425 /* Returns the number of the current Node. */
 426 #ifndef numa_node_id
 427 #define numa_node_id()          (cpu_to_node(raw_smp_processor_id()))
 428 #endif
 429
 430 #ifndef CONFIG_NEED_MULTIPLE_NODES
 431
 432 extern struct pglist_data contig_page_data;
 433 #define NODE_DATA(nid)          (&contig_page_data)
 434 #define NODE_MEM_MAP(nid)       mem_map
 435 #define MAX_NODES_SHIFT         1
 436
 437 #else /* CONFIG_NEED_MULTIPLE_NODES */
 438
 439 #include <asm/mmzone.h>
 440
 441 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
 442
 443 extern struct pglist_data *first_online_pgdat(void);
 444 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
 445 extern struct zone *next_zone(struct zone *zone);
 446
 447 /**
 448  * for_each_pgdat - helper macro to iterate over all nodes
 449  * @pgdat - pointer to a pg_data_t variable
 450  */
 451 #define for_each_online_pgdat(pgdat)                    \
 452         for (pgdat = first_online_pgdat();              \
 453              pgdat;                                     \
 454              pgdat = next_online_pgdat(pgdat))
 455 /**
 456  * for_each_zone - helper macro to iterate over all memory zones
 457  * @zone - pointer to struct zone variable
 458  *
 459  * The user only needs to declare the zone variable, for_each_zone
 460  * fills it in.
 461  */
 462 #define for_each_zone(zone)                             \
 463         for (zone = (first_online_pgdat())->node_zones; \
 464              zone;                                      \
 465              zone = next_zone(zone))
 466
 467 #ifdef CONFIG_SPARSEMEM
 468 #include <asm/sparsemem.h>
 469 #endif
 470
 471 #if BITS_PER_LONG == 32
 472 /*
 473  * with 32 bit page->flags field, we reserve 9 bits for node/zone info.
 474  * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes.
 475  */
 476 #define FLAGS_RESERVED          9
 477
 478 #elif BITS_PER_LONG == 64
 479 /*
 480  * with 64 bit flags field, there's plenty of room.
 481  */
 482 #define FLAGS_RESERVED          32
 483
 484 #else
 485
 486 #error BITS_PER_LONG not defined
 487
 488 #endif
 489
 490 #ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
 491 #define early_pfn_to_nid(nid)  (0UL)
 492 #endif
 493
 494 #ifdef CONFIG_FLATMEM
 495 #define pfn_to_nid(pfn)         (0)
 496 #endif
 497
 498 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
 499 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
 500
 501 #ifdef CONFIG_SPARSEMEM
 502
 503 /*
 504  * SECTION_SHIFT                #bits space required to store a section #
 505  *
 506  * PA_SECTION_SHIFT             physical address to/from section number
 507  * PFN_SECTION_SHIFT            pfn to/from section number
 508  */
 509 #define SECTIONS_SHIFT          (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
 510
 511 #define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
 512 #define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
 513
 514 #define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
 515
 516 #define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
 517 #define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
 518
 519 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
 520 #error Allocator MAX_ORDER exceeds SECTION_SIZE
 521 #endif
 522
 523 struct page;
 524 struct mem_section {
 525         /*
 526          * This is, logically, a pointer to an array of struct
 527          * pages.  However, it is stored with some other magic.
 528          * (see sparse.c::sparse_init_one_section())
 529          *
 530          * Additionally during early boot we encode node id of
 531          * the location of the section here to guide allocation.
 532          * (see sparse.c::memory_present())
 533          *
 534          * Making it a UL at least makes someone do a cast
 535          * before using it wrong.
 536          */
 537         unsigned long section_mem_map;
 538 };
 539
 540 #ifdef CONFIG_SPARSEMEM_EXTREME
 541 #define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
 542 #else
 543 #define SECTIONS_PER_ROOT       1
 544 #endif
 545
 546 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
 547 #define NR_SECTION_ROOTS        (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
 548 #define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
 549
 550 #ifdef CONFIG_SPARSEMEM_EXTREME
 551 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
 552 #else
 553 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
 554 #endif
 555
 556 static inline struct mem_section *__nr_to_section(unsigned long nr)
 557 {
 558         if (!mem_section[SECTION_NR_TO_ROOT(nr)])
 559                 return NULL;
 560         return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
 561 }
 562 extern int __section_nr(struct mem_section* ms);
 563
 564 /*
 565  * We use the lower bits of the mem_map pointer to store
 566  * a little bit of information.  There should be at least
 567  * 3 bits here due to 32-bit alignment.
 568  */
 569 #define SECTION_MARKED_PRESENT  (1UL<<0)
 570 #define SECTION_HAS_MEM_MAP     (1UL<<1)
 571 #define SECTION_MAP_LAST_BIT    (1UL<<2)
 572 #define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
 573 #define SECTION_NID_SHIFT       2
 574
 575 static inline struct page *__section_mem_map_addr(struct mem_section *section)
 576 {
 577         unsigned long map = section->section_mem_map;
 578         map &= SECTION_MAP_MASK;
 579         return (struct page *)map;
 580 }
 581
 582 static inline int valid_section(struct mem_section *section)
 583 {
 584         return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
 585 }
 586
 587 static inline int section_has_mem_map(struct mem_section *section)
 588 {
 589         return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
 590 }
 591
 592 static inline int valid_section_nr(unsigned long nr)
 593 {
 594         return valid_section(__nr_to_section(nr));
 595 }
 596
 597 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
 598 {
 599         return __nr_to_section(pfn_to_section_nr(pfn));
 600 }
 601
 602 static inline int pfn_valid(unsigned long pfn)
 603 {
 604         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
 605                 return 0;
 606         return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
 607 }
 608
 609 /*
 610  * These are _only_ used during initialisation, therefore they
 611  * can use __initdata ...  They could have names to indicate
 612  * this restriction.
 613  */
 614 #ifdef CONFIG_NUMA
 615 #define pfn_to_nid(pfn)                                                 \
 616 ({                                                                      \
 617         unsigned long __pfn_to_nid_pfn = (pfn);                         \
 618         page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
 619 })
 620 #else
 621 #define pfn_to_nid(pfn)         (0)
 622 #endif
 623
 624 #define early_pfn_valid(pfn)    pfn_valid(pfn)
 625 void sparse_init(void);
 626 #else
 627 #define sparse_init()   do {} while (0)
 628 #define sparse_index_init(_sec, _nid)  do {} while (0)
 629 #endif /* CONFIG_SPARSEMEM */
 630
 631 #ifndef early_pfn_valid
 632 #define early_pfn_valid(pfn)    (1)
 633 #endif
 634
 635 void memory_present(int nid, unsigned long start, unsigned long end);
 636 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 637
 638 #endif /* !__ASSEMBLY__ */
 639 #endif /* __KERNEL__ */
 640 #endif /* _LINUX_MMZONE_H */