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_RECLAIMABLE,
  55         NR_SLAB_UNRECLAIMABLE,
  56         NR_PAGETABLE,   /* used for pagetables */
  57         NR_FILE_DIRTY,
  58         NR_WRITEBACK,
  59         NR_UNSTABLE_NFS,        /* NFS unstable pages */
  60         NR_BOUNCE,
  61         NR_VMSCAN_WRITE,
  62 #ifdef CONFIG_NUMA
  63         NUMA_HIT,               /* allocated in intended node */
  64         NUMA_MISS,              /* allocated in non intended node */
  65         NUMA_FOREIGN,           /* was intended here, hit elsewhere */
  66         NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
  67         NUMA_LOCAL,             /* allocation from local node */
  68         NUMA_OTHER,             /* allocation from other node */
  69 #endif
  70         NR_VM_ZONE_STAT_ITEMS };
  71
  72 struct per_cpu_pages {
  73         int count;              /* number of pages in the list */
  74         int high;               /* high watermark, emptying needed */
  75         int batch;              /* chunk size for buddy add/remove */
  76         struct list_head list;  /* the list of pages */
  77 };
  78
  79 struct per_cpu_pageset {
  80         struct per_cpu_pages pcp[2];    /* 0: hot.  1: cold */
  81 #ifdef CONFIG_SMP
  82         s8 stat_threshold;
  83         s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
  84 #endif
  85 } ____cacheline_aligned_in_smp;
  86
  87 #ifdef CONFIG_NUMA
  88 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
  89 #else
  90 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
  91 #endif
  92
  93 enum zone_type {
  94         /*
  95          * ZONE_DMA is used when there are devices that are not able
  96          * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
  97          * carve out the portion of memory that is needed for these devices.
  98          * The range is arch specific.
  99          *
 100          * Some examples
 101          *
 102          * Architecture         Limit
 103          * ---------------------------
 104          * parisc, ia64, sparc  <4G
 105          * s390                 <2G
 106          * arm26                <48M
 107          * arm                  Various
 108          * alpha                Unlimited or 0-16MB.
 109          *
 110          * i386, x86_64 and multiple other arches
 111          *                      <16M.
 112          */
 113         ZONE_DMA,
 114 #ifdef CONFIG_ZONE_DMA32
 115         /*
 116          * x86_64 needs two ZONE_DMAs because it supports devices that are
 117          * only able to do DMA to the lower 16M but also 32 bit devices that
 118          * can only do DMA areas below 4G.
 119          */
 120         ZONE_DMA32,
 121 #endif
 122         /*
 123          * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
 124          * performed on pages in ZONE_NORMAL if the DMA devices support
 125          * transfers to all addressable memory.
 126          */
 127         ZONE_NORMAL,
 128 #ifdef CONFIG_HIGHMEM
 129         /*
 130          * A memory area that is only addressable by the kernel through
 131          * mapping portions into its own address space. This is for example
 132          * used by i386 to allow the kernel to address the memory beyond
 133          * 900MB. The kernel will set up special mappings (page
 134          * table entries on i386) for each page that the kernel needs to
 135          * access.
 136          */
 137         ZONE_HIGHMEM,
 138 #endif
 139         MAX_NR_ZONES
 140 };
 141
 142 /*
 143  * When a memory allocation must conform to specific limitations (such
 144  * as being suitable for DMA) the caller will pass in hints to the
 145  * allocator in the gfp_mask, in the zone modifier bits.  These bits
 146  * are used to select a priority ordered list of memory zones which
 147  * match the requested limits. See gfp_zone() in include/linux/gfp.h
 148  */
 149
 150 #if !defined(CONFIG_ZONE_DMA32) && !defined(CONFIG_HIGHMEM)
 151 #define ZONES_SHIFT 1
 152 #else
 153 #define ZONES_SHIFT 2
 154 #endif
 155
 156 struct zone {
 157         /* Fields commonly accessed by the page allocator */
 158         unsigned long           free_pages;
 159         unsigned long           pages_min, pages_low, pages_high;
 160         /*
 161          * We don't know if the memory that we're going to allocate will be freeable
 162          * or/and it will be released eventually, so to avoid totally wasting several
 163          * GB of ram we must reserve some of the lower zone memory (otherwise we risk
 164          * to run OOM on the lower zones despite there's tons of freeable ram
 165          * on the higher zones). This array is recalculated at runtime if the
 166          * sysctl_lowmem_reserve_ratio sysctl changes.
 167          */
 168         unsigned long           lowmem_reserve[MAX_NR_ZONES];
 169
 170 #ifdef CONFIG_NUMA
 171         int node;
 172         /*
 173          * zone reclaim becomes active if more unmapped pages exist.
 174          */
 175         unsigned long           min_unmapped_pages;
 176         unsigned long           min_slab_pages;
 177         struct per_cpu_pageset  *pageset[NR_CPUS];
 178 #else
 179         struct per_cpu_pageset  pageset[NR_CPUS];
 180 #endif
 181         /*
 182          * free areas of different sizes
 183          */
 184         spinlock_t              lock;
 185 #ifdef CONFIG_MEMORY_HOTPLUG
 186         /* see spanned/present_pages for more description */
 187         seqlock_t               span_seqlock;
 188 #endif
 189         struct free_area        free_area[MAX_ORDER];
 190
 191
 192         ZONE_PADDING(_pad1_)
 193
 194         /* Fields commonly accessed by the page reclaim scanner */
 195         spinlock_t              lru_lock;
 196         struct list_head        active_list;
 197         struct list_head        inactive_list;
 198         unsigned long           nr_scan_active;
 199         unsigned long           nr_scan_inactive;
 200         unsigned long           nr_active;
 201         unsigned long           nr_inactive;
 202         unsigned long           pages_scanned;     /* since last reclaim */
 203         int                     all_unreclaimable; /* All pages pinned */
 204
 205         /* A count of how many reclaimers are scanning this zone */
 206         atomic_t                reclaim_in_progress;
 207
 208         /* Zone statistics */
 209         atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
 210
 211         /*
 212          * prev_priority holds the scanning priority for this zone.  It is
 213          * defined as the scanning priority at which we achieved our reclaim
 214          * target at the previous try_to_free_pages() or balance_pgdat()
 215          * invokation.
 216          *
 217          * We use prev_priority as a measure of how much stress page reclaim is
 218          * under - it drives the swappiness decision: whether to unmap mapped
 219          * pages.
 220          *
 221          * temp_priority is used to remember the scanning priority at which
 222          * this zone was successfully refilled to free_pages == pages_high.
 223          *
 224          * Access to both these fields is quite racy even on uniprocessor.  But
 225          * it is expected to average out OK.
 226          */
 227         int temp_priority;
 228         int prev_priority;
 229
 230
 231         ZONE_PADDING(_pad2_)
 232         /* Rarely used or read-mostly fields */
 233
 234         /*
 235          * wait_table           -- the array holding the hash table
 236          * wait_table_hash_nr_entries   -- the size of the hash table array
 237          * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
 238          *
 239          * The purpose of all these is to keep track of the people
 240          * waiting for a page to become available and make them
 241          * runnable again when possible. The trouble is that this
 242          * consumes a lot of space, especially when so few things
 243          * wait on pages at a given time. So instead of using
 244          * per-page waitqueues, we use a waitqueue hash table.
 245          *
 246          * The bucket discipline is to sleep on the same queue when
 247          * colliding and wake all in that wait queue when removing.
 248          * When something wakes, it must check to be sure its page is
 249          * truly available, a la thundering herd. The cost of a
 250          * collision is great, but given the expected load of the
 251          * table, they should be so rare as to be outweighed by the
 252          * benefits from the saved space.
 253          *
 254          * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
 255          * primary users of these fields, and in mm/page_alloc.c
 256          * free_area_init_core() performs the initialization of them.
 257          */
 258         wait_queue_head_t       * wait_table;
 259         unsigned long           wait_table_hash_nr_entries;
 260         unsigned long           wait_table_bits;
 261
 262         /*
 263          * Discontig memory support fields.
 264          */
 265         struct pglist_data      *zone_pgdat;
 266         /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
 267         unsigned long           zone_start_pfn;
 268
 269         /*
 270          * zone_start_pfn, spanned_pages and present_pages are all
 271          * protected by span_seqlock.  It is a seqlock because it has
 272          * to be read outside of zone->lock, and it is done in the main
 273          * allocator path.  But, it is written quite infrequently.
 274          *
 275          * The lock is declared along with zone->lock because it is
 276          * frequently read in proximity to zone->lock.  It's good to
 277          * give them a chance of being in the same cacheline.
 278          */
 279         unsigned long           spanned_pages;  /* total size, including holes */
 280         unsigned long           present_pages;  /* amount of memory (excluding holes) */
 281
 282         /*
 283          * rarely used fields:
 284          */
 285         char                    *name;
 286 } ____cacheline_internodealigned_in_smp;
 287
 288 /*
 289  * The "priority" of VM scanning is how much of the queues we will scan in one
 290  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
 291  * queues ("queue_length >> 12") during an aging round.
 292  */
 293 #define DEF_PRIORITY 12
 294
 295 /*
 296  * One allocation request operates on a zonelist. A zonelist
 297  * is a list of zones, the first one is the 'goal' of the
 298  * allocation, the other zones are fallback zones, in decreasing
 299  * priority.
 300  *
 301  * Right now a zonelist takes up less than a cacheline. We never
 302  * modify it apart from boot-up, and only a few indices are used,
 303  * so despite the zonelist table being relatively big, the cache
 304  * footprint of this construct is very small.
 305  */
 306 struct zonelist {
 307         struct zone *zones[MAX_NUMNODES * MAX_NR_ZONES + 1]; // NULL delimited
 308 };
 309
 310 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
 311 struct node_active_region {
 312         unsigned long start_pfn;
 313         unsigned long end_pfn;
 314         int nid;
 315 };
 316 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
 317
 318 #ifndef CONFIG_DISCONTIGMEM
 319 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
 320 extern struct page *mem_map;
 321 #endif
 322
 323 /*
 324  * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
 325  * (mostly NUMA machines?) to denote a higher-level memory zone than the
 326  * zone denotes.
 327  *
 328  * On NUMA machines, each NUMA node would have a pg_data_t to describe
 329  * it's memory layout.
 330  *
 331  * Memory statistics and page replacement data structures are maintained on a
 332  * per-zone basis.
 333  */
 334 struct bootmem_data;
 335 typedef struct pglist_data {
 336         struct zone node_zones[MAX_NR_ZONES];
 337         struct zonelist node_zonelists[MAX_NR_ZONES];
 338         int nr_zones;
 339 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 340         struct page *node_mem_map;
 341 #endif
 342         struct bootmem_data *bdata;
 343 #ifdef CONFIG_MEMORY_HOTPLUG
 344         /*
 345          * Must be held any time you expect node_start_pfn, node_present_pages
 346          * or node_spanned_pages stay constant.  Holding this will also
 347          * guarantee that any pfn_valid() stays that way.
 348          *
 349          * Nests above zone->lock and zone->size_seqlock.
 350          */
 351         spinlock_t node_size_lock;
 352 #endif
 353         unsigned long node_start_pfn;
 354         unsigned long node_present_pages; /* total number of physical pages */
 355         unsigned long node_spanned_pages; /* total size of physical page
 356                                              range, including holes */
 357         int node_id;
 358         wait_queue_head_t kswapd_wait;
 359         struct task_struct *kswapd;
 360         int kswapd_max_order;
 361 } pg_data_t;
 362
 363 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
 364 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
 365 #ifdef CONFIG_FLAT_NODE_MEM_MAP
 366 #define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
 367 #else
 368 #define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
 369 #endif
 370 #define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
 371
 372 #include <linux/memory_hotplug.h>
 373
 374 void __get_zone_counts(unsigned long *active, unsigned long *inactive,
 375                         unsigned long *free, struct pglist_data *pgdat);
 376 void get_zone_counts(unsigned long *active, unsigned long *inactive,
 377                         unsigned long *free);
 378 void build_all_zonelists(void);
 379 void wakeup_kswapd(struct zone *zone, int order);
 380 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
 381                 int classzone_idx, int alloc_flags);
 382
 383 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
 384                                      unsigned long size);
 385
 386 #ifdef CONFIG_HAVE_MEMORY_PRESENT
 387 void memory_present(int nid, unsigned long start, unsigned long end);
 388 #else
 389 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
 390 #endif
 391
 392 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
 393 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 394 #endif
 395
 396 /*
 397  * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
 398  */
 399 #define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
 400
 401 static inline int populated_zone(struct zone *zone)
 402 {
 403         return (!!zone->present_pages);
 404 }
 405
 406 static inline int is_highmem_idx(enum zone_type idx)
 407 {
 408 #ifdef CONFIG_HIGHMEM
 409         return (idx == ZONE_HIGHMEM);
 410 #else
 411         return 0;
 412 #endif
 413 }
 414
 415 static inline int is_normal_idx(enum zone_type idx)
 416 {
 417         return (idx == ZONE_NORMAL);
 418 }
 419
 420 /**
 421  * is_highmem - helper function to quickly check if a struct zone is a
 422  *              highmem zone or not.  This is an attempt to keep references
 423  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
 424  * @zone - pointer to struct zone variable
 425  */
 426 static inline int is_highmem(struct zone *zone)
 427 {
 428 #ifdef CONFIG_HIGHMEM
 429         return zone == zone->zone_pgdat->node_zones + ZONE_HIGHMEM;
 430 #else
 431         return 0;
 432 #endif
 433 }
 434
 435 static inline int is_normal(struct zone *zone)
 436 {
 437         return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
 438 }
 439
 440 static inline int is_dma32(struct zone *zone)
 441 {
 442 #ifdef CONFIG_ZONE_DMA32
 443         return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
 444 #else
 445         return 0;
 446 #endif
 447 }
 448
 449 static inline int is_dma(struct zone *zone)
 450 {
 451         return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
 452 }
 453
 454 /* These two functions are used to setup the per zone pages min values */
 455 struct ctl_table;
 456 struct file;
 457 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
 458                                         void __user *, size_t *, loff_t *);
 459 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
 460 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
 461                                         void __user *, size_t *, loff_t *);
 462 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
 463                                         void __user *, size_t *, loff_t *);
 464 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
 465                         struct file *, void __user *, size_t *, loff_t *);
 466 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
 467                         struct file *, void __user *, size_t *, loff_t *);
 468
 469 #include <linux/topology.h>
 470 /* Returns the number of the current Node. */
 471 #ifndef numa_node_id
 472 #define numa_node_id()          (cpu_to_node(raw_smp_processor_id()))
 473 #endif
 474
 475 #ifndef CONFIG_NEED_MULTIPLE_NODES
 476
 477 extern struct pglist_data contig_page_data;
 478 #define NODE_DATA(nid)          (&contig_page_data)
 479 #define NODE_MEM_MAP(nid)       mem_map
 480 #define MAX_NODES_SHIFT         1
 481
 482 #else /* CONFIG_NEED_MULTIPLE_NODES */
 483
 484 #include <asm/mmzone.h>
 485
 486 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
 487
 488 extern struct pglist_data *first_online_pgdat(void);
 489 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
 490 extern struct zone *next_zone(struct zone *zone);
 491
 492 /**
 493  * for_each_pgdat - helper macro to iterate over all nodes
 494  * @pgdat - pointer to a pg_data_t variable
 495  */
 496 #define for_each_online_pgdat(pgdat)                    \
 497         for (pgdat = first_online_pgdat();              \
 498              pgdat;                                     \
 499              pgdat = next_online_pgdat(pgdat))
 500 /**
 501  * for_each_zone - helper macro to iterate over all memory zones
 502  * @zone - pointer to struct zone variable
 503  *
 504  * The user only needs to declare the zone variable, for_each_zone
 505  * fills it in.
 506  */
 507 #define for_each_zone(zone)                             \
 508         for (zone = (first_online_pgdat())->node_zones; \
 509              zone;                                      \
 510              zone = next_zone(zone))
 511
 512 #ifdef CONFIG_SPARSEMEM
 513 #include <asm/sparsemem.h>
 514 #endif
 515
 516 #if BITS_PER_LONG == 32
 517 /*
 518  * with 32 bit page->flags field, we reserve 9 bits for node/zone info.
 519  * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes.
 520  */
 521 #define FLAGS_RESERVED          9
 522
 523 #elif BITS_PER_LONG == 64
 524 /*
 525  * with 64 bit flags field, there's plenty of room.
 526  */
 527 #define FLAGS_RESERVED          32
 528
 529 #else
 530
 531 #error BITS_PER_LONG not defined
 532
 533 #endif
 534
 535 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
 536         !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
 537 #define early_pfn_to_nid(nid)  (0UL)
 538 #endif
 539
 540 #ifdef CONFIG_FLATMEM
 541 #define pfn_to_nid(pfn)         (0)
 542 #endif
 543
 544 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
 545 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
 546
 547 #ifdef CONFIG_SPARSEMEM
 548
 549 /*
 550  * SECTION_SHIFT                #bits space required to store a section #
 551  *
 552  * PA_SECTION_SHIFT             physical address to/from section number
 553  * PFN_SECTION_SHIFT            pfn to/from section number
 554  */
 555 #define SECTIONS_SHIFT          (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
 556
 557 #define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
 558 #define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
 559
 560 #define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
 561
 562 #define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
 563 #define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
 564
 565 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
 566 #error Allocator MAX_ORDER exceeds SECTION_SIZE
 567 #endif
 568
 569 struct page;
 570 struct mem_section {
 571         /*
 572          * This is, logically, a pointer to an array of struct
 573          * pages.  However, it is stored with some other magic.
 574          * (see sparse.c::sparse_init_one_section())
 575          *
 576          * Additionally during early boot we encode node id of
 577          * the location of the section here to guide allocation.
 578          * (see sparse.c::memory_present())
 579          *
 580          * Making it a UL at least makes someone do a cast
 581          * before using it wrong.
 582          */
 583         unsigned long section_mem_map;
 584 };
 585
 586 #ifdef CONFIG_SPARSEMEM_EXTREME
 587 #define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
 588 #else
 589 #define SECTIONS_PER_ROOT       1
 590 #endif
 591
 592 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
 593 #define NR_SECTION_ROOTS        (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
 594 #define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
 595
 596 #ifdef CONFIG_SPARSEMEM_EXTREME
 597 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
 598 #else
 599 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
 600 #endif
 601
 602 static inline struct mem_section *__nr_to_section(unsigned long nr)
 603 {
 604         if (!mem_section[SECTION_NR_TO_ROOT(nr)])
 605                 return NULL;
 606         return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
 607 }
 608 extern int __section_nr(struct mem_section* ms);
 609
 610 /*
 611  * We use the lower bits of the mem_map pointer to store
 612  * a little bit of information.  There should be at least
 613  * 3 bits here due to 32-bit alignment.
 614  */
 615 #define SECTION_MARKED_PRESENT  (1UL<<0)
 616 #define SECTION_HAS_MEM_MAP     (1UL<<1)
 617 #define SECTION_MAP_LAST_BIT    (1UL<<2)
 618 #define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
 619 #define SECTION_NID_SHIFT       2
 620
 621 static inline struct page *__section_mem_map_addr(struct mem_section *section)
 622 {
 623         unsigned long map = section->section_mem_map;
 624         map &= SECTION_MAP_MASK;
 625         return (struct page *)map;
 626 }
 627
 628 static inline int valid_section(struct mem_section *section)
 629 {
 630         return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
 631 }
 632
 633 static inline int section_has_mem_map(struct mem_section *section)
 634 {
 635         return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
 636 }
 637
 638 static inline int valid_section_nr(unsigned long nr)
 639 {
 640         return valid_section(__nr_to_section(nr));
 641 }
 642
 643 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
 644 {
 645         return __nr_to_section(pfn_to_section_nr(pfn));
 646 }
 647
 648 static inline int pfn_valid(unsigned long pfn)
 649 {
 650         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
 651                 return 0;
 652         return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
 653 }
 654
 655 /*
 656  * These are _only_ used during initialisation, therefore they
 657  * can use __initdata ...  They could have names to indicate
 658  * this restriction.
 659  */
 660 #ifdef CONFIG_NUMA
 661 #define pfn_to_nid(pfn)                                                 \
 662 ({                                                                      \
 663         unsigned long __pfn_to_nid_pfn = (pfn);                         \
 664         page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
 665 })
 666 #else
 667 #define pfn_to_nid(pfn)         (0)
 668 #endif
 669
 670 #define early_pfn_valid(pfn)    pfn_valid(pfn)
 671 void sparse_init(void);
 672 #else
 673 #define sparse_init()   do {} while (0)
 674 #define sparse_index_init(_sec, _nid)  do {} while (0)
 675 #endif /* CONFIG_SPARSEMEM */
 676
 677 #ifndef early_pfn_valid
 678 #define early_pfn_valid(pfn)    (1)
 679 #endif
 680
 681 void memory_present(int nid, unsigned long start, unsigned long end);
 682 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
 683
 684 #endif /* !__ASSEMBLY__ */
 685 #endif /* __KERNEL__ */
 686 #endif /* _LINUX_MMZONE_H */