Version:  2.0.40 2.2.26 2.4.37 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7

Linux/include/linux/mmzone.h

  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 <linux/page-flags-layout.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 
 38 enum {
 39         MIGRATE_UNMOVABLE,
 40         MIGRATE_MOVABLE,
 41         MIGRATE_RECLAIMABLE,
 42         MIGRATE_PCPTYPES,       /* the number of types on the pcp lists */
 43         MIGRATE_HIGHATOMIC = 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 #ifdef CONFIG_MEMORY_ISOLATION
 61         MIGRATE_ISOLATE,        /* can't allocate from here */
 62 #endif
 63         MIGRATE_TYPES
 64 };
 65 
 66 /* In mm/page_alloc.c; keep in sync also with show_migration_types() there */
 67 extern char * const migratetype_names[MIGRATE_TYPES];
 68 
 69 #ifdef CONFIG_CMA
 70 #  define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
 71 #else
 72 #  define is_migrate_cma(migratetype) false
 73 #endif
 74 
 75 #define for_each_migratetype_order(order, type) \
 76         for (order = 0; order < MAX_ORDER; order++) \
 77                 for (type = 0; type < MIGRATE_TYPES; type++)
 78 
 79 extern int page_group_by_mobility_disabled;
 80 
 81 #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
 82 #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
 83 
 84 #define get_pageblock_migratetype(page)                                 \
 85         get_pfnblock_flags_mask(page, page_to_pfn(page),                \
 86                         PB_migrate_end, MIGRATETYPE_MASK)
 87 
 88 struct free_area {
 89         struct list_head        free_list[MIGRATE_TYPES];
 90         unsigned long           nr_free;
 91 };
 92 
 93 struct pglist_data;
 94 
 95 /*
 96  * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
 97  * So add a wild amount of padding here to ensure that they fall into separate
 98  * cachelines.  There are very few zone structures in the machine, so space
 99  * consumption is not a concern here.
100  */
101 #if defined(CONFIG_SMP)
102 struct zone_padding {
103         char x[0];
104 } ____cacheline_internodealigned_in_smp;
105 #define ZONE_PADDING(name)      struct zone_padding name;
106 #else
107 #define ZONE_PADDING(name)
108 #endif
109 
110 enum zone_stat_item {
111         /* First 128 byte cacheline (assuming 64 bit words) */
112         NR_FREE_PAGES,
113         NR_ALLOC_BATCH,
114         NR_LRU_BASE,
115         NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
116         NR_ACTIVE_ANON,         /*  "     "     "   "       "         */
117         NR_INACTIVE_FILE,       /*  "     "     "   "       "         */
118         NR_ACTIVE_FILE,         /*  "     "     "   "       "         */
119         NR_UNEVICTABLE,         /*  "     "     "   "       "         */
120         NR_MLOCK,               /* mlock()ed pages found and moved off LRU */
121         NR_ANON_PAGES,  /* Mapped anonymous pages */
122         NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
123                            only modified from process context */
124         NR_FILE_PAGES,
125         NR_FILE_DIRTY,
126         NR_WRITEBACK,
127         NR_SLAB_RECLAIMABLE,
128         NR_SLAB_UNRECLAIMABLE,
129         NR_PAGETABLE,           /* used for pagetables */
130         NR_KERNEL_STACK,
131         /* Second 128 byte cacheline */
132         NR_UNSTABLE_NFS,        /* NFS unstable pages */
133         NR_BOUNCE,
134         NR_VMSCAN_WRITE,
135         NR_VMSCAN_IMMEDIATE,    /* Prioritise for reclaim when writeback ends */
136         NR_WRITEBACK_TEMP,      /* Writeback using temporary buffers */
137         NR_ISOLATED_ANON,       /* Temporary isolated pages from anon lru */
138         NR_ISOLATED_FILE,       /* Temporary isolated pages from file lru */
139         NR_SHMEM,               /* shmem pages (included tmpfs/GEM pages) */
140         NR_DIRTIED,             /* page dirtyings since bootup */
141         NR_WRITTEN,             /* page writings since bootup */
142         NR_PAGES_SCANNED,       /* pages scanned since last reclaim */
143 #ifdef CONFIG_NUMA
144         NUMA_HIT,               /* allocated in intended node */
145         NUMA_MISS,              /* allocated in non intended node */
146         NUMA_FOREIGN,           /* was intended here, hit elsewhere */
147         NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
148         NUMA_LOCAL,             /* allocation from local node */
149         NUMA_OTHER,             /* allocation from other node */
150 #endif
151         WORKINGSET_REFAULT,
152         WORKINGSET_ACTIVATE,
153         WORKINGSET_NODERECLAIM,
154         NR_ANON_TRANSPARENT_HUGEPAGES,
155         NR_FREE_CMA_PAGES,
156         NR_VM_ZONE_STAT_ITEMS };
157 
158 /*
159  * We do arithmetic on the LRU lists in various places in the code,
160  * so it is important to keep the active lists LRU_ACTIVE higher in
161  * the array than the corresponding inactive lists, and to keep
162  * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
163  *
164  * This has to be kept in sync with the statistics in zone_stat_item
165  * above and the descriptions in vmstat_text in mm/vmstat.c
166  */
167 #define LRU_BASE 0
168 #define LRU_ACTIVE 1
169 #define LRU_FILE 2
170 
171 enum lru_list {
172         LRU_INACTIVE_ANON = LRU_BASE,
173         LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
174         LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
175         LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
176         LRU_UNEVICTABLE,
177         NR_LRU_LISTS
178 };
179 
180 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
181 
182 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
183 
184 static inline int is_file_lru(enum lru_list lru)
185 {
186         return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
187 }
188 
189 static inline int is_active_lru(enum lru_list lru)
190 {
191         return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
192 }
193 
194 struct zone_reclaim_stat {
195         /*
196          * The pageout code in vmscan.c keeps track of how many of the
197          * mem/swap backed and file backed pages are referenced.
198          * The higher the rotated/scanned ratio, the more valuable
199          * that cache is.
200          *
201          * The anon LRU stats live in [0], file LRU stats in [1]
202          */
203         unsigned long           recent_rotated[2];
204         unsigned long           recent_scanned[2];
205 };
206 
207 struct lruvec {
208         struct list_head                lists[NR_LRU_LISTS];
209         struct zone_reclaim_stat        reclaim_stat;
210         /* Evictions & activations on the inactive file list */
211         atomic_long_t                   inactive_age;
212 #ifdef CONFIG_MEMCG
213         struct zone                     *zone;
214 #endif
215 };
216 
217 /* Mask used at gathering information at once (see memcontrol.c) */
218 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
219 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
220 #define LRU_ALL      ((1 << NR_LRU_LISTS) - 1)
221 
222 /* Isolate clean file */
223 #define ISOLATE_CLEAN           ((__force isolate_mode_t)0x1)
224 /* Isolate unmapped file */
225 #define ISOLATE_UNMAPPED        ((__force isolate_mode_t)0x2)
226 /* Isolate for asynchronous migration */
227 #define ISOLATE_ASYNC_MIGRATE   ((__force isolate_mode_t)0x4)
228 /* Isolate unevictable pages */
229 #define ISOLATE_UNEVICTABLE     ((__force isolate_mode_t)0x8)
230 
231 /* LRU Isolation modes. */
232 typedef unsigned __bitwise__ isolate_mode_t;
233 
234 enum zone_watermarks {
235         WMARK_MIN,
236         WMARK_LOW,
237         WMARK_HIGH,
238         NR_WMARK
239 };
240 
241 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
242 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
243 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
244 
245 struct per_cpu_pages {
246         int count;              /* number of pages in the list */
247         int high;               /* high watermark, emptying needed */
248         int batch;              /* chunk size for buddy add/remove */
249 
250         /* Lists of pages, one per migrate type stored on the pcp-lists */
251         struct list_head lists[MIGRATE_PCPTYPES];
252 };
253 
254 struct per_cpu_pageset {
255         struct per_cpu_pages pcp;
256 #ifdef CONFIG_NUMA
257         s8 expire;
258 #endif
259 #ifdef CONFIG_SMP
260         s8 stat_threshold;
261         s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
262 #endif
263 };
264 
265 #endif /* !__GENERATING_BOUNDS.H */
266 
267 enum zone_type {
268 #ifdef CONFIG_ZONE_DMA
269         /*
270          * ZONE_DMA is used when there are devices that are not able
271          * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
272          * carve out the portion of memory that is needed for these devices.
273          * The range is arch specific.
274          *
275          * Some examples
276          *
277          * Architecture         Limit
278          * ---------------------------
279          * parisc, ia64, sparc  <4G
280          * s390                 <2G
281          * arm                  Various
282          * alpha                Unlimited or 0-16MB.
283          *
284          * i386, x86_64 and multiple other arches
285          *                      <16M.
286          */
287         ZONE_DMA,
288 #endif
289 #ifdef CONFIG_ZONE_DMA32
290         /*
291          * x86_64 needs two ZONE_DMAs because it supports devices that are
292          * only able to do DMA to the lower 16M but also 32 bit devices that
293          * can only do DMA areas below 4G.
294          */
295         ZONE_DMA32,
296 #endif
297         /*
298          * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
299          * performed on pages in ZONE_NORMAL if the DMA devices support
300          * transfers to all addressable memory.
301          */
302         ZONE_NORMAL,
303 #ifdef CONFIG_HIGHMEM
304         /*
305          * A memory area that is only addressable by the kernel through
306          * mapping portions into its own address space. This is for example
307          * used by i386 to allow the kernel to address the memory beyond
308          * 900MB. The kernel will set up special mappings (page
309          * table entries on i386) for each page that the kernel needs to
310          * access.
311          */
312         ZONE_HIGHMEM,
313 #endif
314         ZONE_MOVABLE,
315 #ifdef CONFIG_ZONE_DEVICE
316         ZONE_DEVICE,
317 #endif
318         __MAX_NR_ZONES
319 
320 };
321 
322 #ifndef __GENERATING_BOUNDS_H
323 
324 struct zone {
325         /* Read-mostly fields */
326 
327         /* zone watermarks, access with *_wmark_pages(zone) macros */
328         unsigned long watermark[NR_WMARK];
329 
330         unsigned long nr_reserved_highatomic;
331 
332         /*
333          * We don't know if the memory that we're going to allocate will be
334          * freeable or/and it will be released eventually, so to avoid totally
335          * wasting several GB of ram we must reserve some of the lower zone
336          * memory (otherwise we risk to run OOM on the lower zones despite
337          * there being tons of freeable ram on the higher zones).  This array is
338          * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
339          * changes.
340          */
341         long lowmem_reserve[MAX_NR_ZONES];
342 
343 #ifdef CONFIG_NUMA
344         int node;
345 #endif
346 
347         /*
348          * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
349          * this zone's LRU.  Maintained by the pageout code.
350          */
351         unsigned int inactive_ratio;
352 
353         struct pglist_data      *zone_pgdat;
354         struct per_cpu_pageset __percpu *pageset;
355 
356         /*
357          * This is a per-zone reserve of pages that are not available
358          * to userspace allocations.
359          */
360         unsigned long           totalreserve_pages;
361 
362 #ifndef CONFIG_SPARSEMEM
363         /*
364          * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
365          * In SPARSEMEM, this map is stored in struct mem_section
366          */
367         unsigned long           *pageblock_flags;
368 #endif /* CONFIG_SPARSEMEM */
369 
370 #ifdef CONFIG_NUMA
371         /*
372          * zone reclaim becomes active if more unmapped pages exist.
373          */
374         unsigned long           min_unmapped_pages;
375         unsigned long           min_slab_pages;
376 #endif /* CONFIG_NUMA */
377 
378         /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
379         unsigned long           zone_start_pfn;
380 
381         /*
382          * spanned_pages is the total pages spanned by the zone, including
383          * holes, which is calculated as:
384          *      spanned_pages = zone_end_pfn - zone_start_pfn;
385          *
386          * present_pages is physical pages existing within the zone, which
387          * is calculated as:
388          *      present_pages = spanned_pages - absent_pages(pages in holes);
389          *
390          * managed_pages is present pages managed by the buddy system, which
391          * is calculated as (reserved_pages includes pages allocated by the
392          * bootmem allocator):
393          *      managed_pages = present_pages - reserved_pages;
394          *
395          * So present_pages may be used by memory hotplug or memory power
396          * management logic to figure out unmanaged pages by checking
397          * (present_pages - managed_pages). And managed_pages should be used
398          * by page allocator and vm scanner to calculate all kinds of watermarks
399          * and thresholds.
400          *
401          * Locking rules:
402          *
403          * zone_start_pfn and spanned_pages are protected by span_seqlock.
404          * It is a seqlock because it has to be read outside of zone->lock,
405          * and it is done in the main allocator path.  But, it is written
406          * quite infrequently.
407          *
408          * The span_seq lock is declared along with zone->lock because it is
409          * frequently read in proximity to zone->lock.  It's good to
410          * give them a chance of being in the same cacheline.
411          *
412          * Write access to present_pages at runtime should be protected by
413          * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
414          * present_pages should get_online_mems() to get a stable value.
415          *
416          * Read access to managed_pages should be safe because it's unsigned
417          * long. Write access to zone->managed_pages and totalram_pages are
418          * protected by managed_page_count_lock at runtime. Idealy only
419          * adjust_managed_page_count() should be used instead of directly
420          * touching zone->managed_pages and totalram_pages.
421          */
422         unsigned long           managed_pages;
423         unsigned long           spanned_pages;
424         unsigned long           present_pages;
425 
426         const char              *name;
427 
428 #ifdef CONFIG_MEMORY_ISOLATION
429         /*
430          * Number of isolated pageblock. It is used to solve incorrect
431          * freepage counting problem due to racy retrieving migratetype
432          * of pageblock. Protected by zone->lock.
433          */
434         unsigned long           nr_isolate_pageblock;
435 #endif
436 
437 #ifdef CONFIG_MEMORY_HOTPLUG
438         /* see spanned/present_pages for more description */
439         seqlock_t               span_seqlock;
440 #endif
441 
442         /*
443          * wait_table           -- the array holding the hash table
444          * wait_table_hash_nr_entries   -- the size of the hash table array
445          * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
446          *
447          * The purpose of all these is to keep track of the people
448          * waiting for a page to become available and make them
449          * runnable again when possible. The trouble is that this
450          * consumes a lot of space, especially when so few things
451          * wait on pages at a given time. So instead of using
452          * per-page waitqueues, we use a waitqueue hash table.
453          *
454          * The bucket discipline is to sleep on the same queue when
455          * colliding and wake all in that wait queue when removing.
456          * When something wakes, it must check to be sure its page is
457          * truly available, a la thundering herd. The cost of a
458          * collision is great, but given the expected load of the
459          * table, they should be so rare as to be outweighed by the
460          * benefits from the saved space.
461          *
462          * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
463          * primary users of these fields, and in mm/page_alloc.c
464          * free_area_init_core() performs the initialization of them.
465          */
466         wait_queue_head_t       *wait_table;
467         unsigned long           wait_table_hash_nr_entries;
468         unsigned long           wait_table_bits;
469 
470         ZONE_PADDING(_pad1_)
471         /* free areas of different sizes */
472         struct free_area        free_area[MAX_ORDER];
473 
474         /* zone flags, see below */
475         unsigned long           flags;
476 
477         /* Write-intensive fields used from the page allocator */
478         spinlock_t              lock;
479 
480         ZONE_PADDING(_pad2_)
481 
482         /* Write-intensive fields used by page reclaim */
483 
484         /* Fields commonly accessed by the page reclaim scanner */
485         spinlock_t              lru_lock;
486         struct lruvec           lruvec;
487 
488         /*
489          * When free pages are below this point, additional steps are taken
490          * when reading the number of free pages to avoid per-cpu counter
491          * drift allowing watermarks to be breached
492          */
493         unsigned long percpu_drift_mark;
494 
495 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
496         /* pfn where compaction free scanner should start */
497         unsigned long           compact_cached_free_pfn;
498         /* pfn where async and sync compaction migration scanner should start */
499         unsigned long           compact_cached_migrate_pfn[2];
500 #endif
501 
502 #ifdef CONFIG_COMPACTION
503         /*
504          * On compaction failure, 1<<compact_defer_shift compactions
505          * are skipped before trying again. The number attempted since
506          * last failure is tracked with compact_considered.
507          */
508         unsigned int            compact_considered;
509         unsigned int            compact_defer_shift;
510         int                     compact_order_failed;
511 #endif
512 
513 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
514         /* Set to true when the PG_migrate_skip bits should be cleared */
515         bool                    compact_blockskip_flush;
516 #endif
517 
518         bool                    contiguous;
519 
520         ZONE_PADDING(_pad3_)
521         /* Zone statistics */
522         atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
523 } ____cacheline_internodealigned_in_smp;
524 
525 enum zone_flags {
526         ZONE_RECLAIM_LOCKED,            /* prevents concurrent reclaim */
527         ZONE_OOM_LOCKED,                /* zone is in OOM killer zonelist */
528         ZONE_CONGESTED,                 /* zone has many dirty pages backed by
529                                          * a congested BDI
530                                          */
531         ZONE_DIRTY,                     /* reclaim scanning has recently found
532                                          * many dirty file pages at the tail
533                                          * of the LRU.
534                                          */
535         ZONE_WRITEBACK,                 /* reclaim scanning has recently found
536                                          * many pages under writeback
537                                          */
538         ZONE_FAIR_DEPLETED,             /* fair zone policy batch depleted */
539 };
540 
541 static inline unsigned long zone_end_pfn(const struct zone *zone)
542 {
543         return zone->zone_start_pfn + zone->spanned_pages;
544 }
545 
546 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
547 {
548         return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
549 }
550 
551 static inline bool zone_is_initialized(struct zone *zone)
552 {
553         return !!zone->wait_table;
554 }
555 
556 static inline bool zone_is_empty(struct zone *zone)
557 {
558         return zone->spanned_pages == 0;
559 }
560 
561 /*
562  * The "priority" of VM scanning is how much of the queues we will scan in one
563  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
564  * queues ("queue_length >> 12") during an aging round.
565  */
566 #define DEF_PRIORITY 12
567 
568 /* Maximum number of zones on a zonelist */
569 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
570 
571 enum {
572         ZONELIST_FALLBACK,      /* zonelist with fallback */
573 #ifdef CONFIG_NUMA
574         /*
575          * The NUMA zonelists are doubled because we need zonelists that
576          * restrict the allocations to a single node for __GFP_THISNODE.
577          */
578         ZONELIST_NOFALLBACK,    /* zonelist without fallback (__GFP_THISNODE) */
579 #endif
580         MAX_ZONELISTS
581 };
582 
583 /*
584  * This struct contains information about a zone in a zonelist. It is stored
585  * here to avoid dereferences into large structures and lookups of tables
586  */
587 struct zoneref {
588         struct zone *zone;      /* Pointer to actual zone */
589         int zone_idx;           /* zone_idx(zoneref->zone) */
590 };
591 
592 /*
593  * One allocation request operates on a zonelist. A zonelist
594  * is a list of zones, the first one is the 'goal' of the
595  * allocation, the other zones are fallback zones, in decreasing
596  * priority.
597  *
598  * To speed the reading of the zonelist, the zonerefs contain the zone index
599  * of the entry being read. Helper functions to access information given
600  * a struct zoneref are
601  *
602  * zonelist_zone()      - Return the struct zone * for an entry in _zonerefs
603  * zonelist_zone_idx()  - Return the index of the zone for an entry
604  * zonelist_node_idx()  - Return the index of the node for an entry
605  */
606 struct zonelist {
607         struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
608 };
609 
610 #ifndef CONFIG_DISCONTIGMEM
611 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
612 extern struct page *mem_map;
613 #endif
614 
615 /*
616  * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
617  * (mostly NUMA machines?) to denote a higher-level memory zone than the
618  * zone denotes.
619  *
620  * On NUMA machines, each NUMA node would have a pg_data_t to describe
621  * it's memory layout.
622  *
623  * Memory statistics and page replacement data structures are maintained on a
624  * per-zone basis.
625  */
626 struct bootmem_data;
627 typedef struct pglist_data {
628         struct zone node_zones[MAX_NR_ZONES];
629         struct zonelist node_zonelists[MAX_ZONELISTS];
630         int nr_zones;
631 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
632         struct page *node_mem_map;
633 #ifdef CONFIG_PAGE_EXTENSION
634         struct page_ext *node_page_ext;
635 #endif
636 #endif
637 #ifndef CONFIG_NO_BOOTMEM
638         struct bootmem_data *bdata;
639 #endif
640 #ifdef CONFIG_MEMORY_HOTPLUG
641         /*
642          * Must be held any time you expect node_start_pfn, node_present_pages
643          * or node_spanned_pages stay constant.  Holding this will also
644          * guarantee that any pfn_valid() stays that way.
645          *
646          * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
647          * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
648          *
649          * Nests above zone->lock and zone->span_seqlock
650          */
651         spinlock_t node_size_lock;
652 #endif
653         unsigned long node_start_pfn;
654         unsigned long node_present_pages; /* total number of physical pages */
655         unsigned long node_spanned_pages; /* total size of physical page
656                                              range, including holes */
657         int node_id;
658         wait_queue_head_t kswapd_wait;
659         wait_queue_head_t pfmemalloc_wait;
660         struct task_struct *kswapd;     /* Protected by
661                                            mem_hotplug_begin/end() */
662         int kswapd_max_order;
663         enum zone_type classzone_idx;
664 #ifdef CONFIG_COMPACTION
665         int kcompactd_max_order;
666         enum zone_type kcompactd_classzone_idx;
667         wait_queue_head_t kcompactd_wait;
668         struct task_struct *kcompactd;
669 #endif
670 #ifdef CONFIG_NUMA_BALANCING
671         /* Lock serializing the migrate rate limiting window */
672         spinlock_t numabalancing_migrate_lock;
673 
674         /* Rate limiting time interval */
675         unsigned long numabalancing_migrate_next_window;
676 
677         /* Number of pages migrated during the rate limiting time interval */
678         unsigned long numabalancing_migrate_nr_pages;
679 #endif
680 
681 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
682         /*
683          * If memory initialisation on large machines is deferred then this
684          * is the first PFN that needs to be initialised.
685          */
686         unsigned long first_deferred_pfn;
687 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
688 
689 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
690         spinlock_t split_queue_lock;
691         struct list_head split_queue;
692         unsigned long split_queue_len;
693 #endif
694 } pg_data_t;
695 
696 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
697 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
698 #ifdef CONFIG_FLAT_NODE_MEM_MAP
699 #define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
700 #else
701 #define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
702 #endif
703 #define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
704 
705 #define node_start_pfn(nid)     (NODE_DATA(nid)->node_start_pfn)
706 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
707 
708 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
709 {
710         return pgdat->node_start_pfn + pgdat->node_spanned_pages;
711 }
712 
713 static inline bool pgdat_is_empty(pg_data_t *pgdat)
714 {
715         return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
716 }
717 
718 static inline int zone_id(const struct zone *zone)
719 {
720         struct pglist_data *pgdat = zone->zone_pgdat;
721 
722         return zone - pgdat->node_zones;
723 }
724 
725 #ifdef CONFIG_ZONE_DEVICE
726 static inline bool is_dev_zone(const struct zone *zone)
727 {
728         return zone_id(zone) == ZONE_DEVICE;
729 }
730 #else
731 static inline bool is_dev_zone(const struct zone *zone)
732 {
733         return false;
734 }
735 #endif
736 
737 #include <linux/memory_hotplug.h>
738 
739 extern struct mutex zonelists_mutex;
740 void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
741 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
742 bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
743                          int classzone_idx, unsigned int alloc_flags,
744                          long free_pages);
745 bool zone_watermark_ok(struct zone *z, unsigned int order,
746                 unsigned long mark, int classzone_idx,
747                 unsigned int alloc_flags);
748 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
749                 unsigned long mark, int classzone_idx);
750 enum memmap_context {
751         MEMMAP_EARLY,
752         MEMMAP_HOTPLUG,
753 };
754 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
755                                      unsigned long size);
756 
757 extern void lruvec_init(struct lruvec *lruvec);
758 
759 static inline struct zone *lruvec_zone(struct lruvec *lruvec)
760 {
761 #ifdef CONFIG_MEMCG
762         return lruvec->zone;
763 #else
764         return container_of(lruvec, struct zone, lruvec);
765 #endif
766 }
767 
768 extern unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru);
769 
770 #ifdef CONFIG_HAVE_MEMORY_PRESENT
771 void memory_present(int nid, unsigned long start, unsigned long end);
772 #else
773 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
774 #endif
775 
776 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
777 int local_memory_node(int node_id);
778 #else
779 static inline int local_memory_node(int node_id) { return node_id; };
780 #endif
781 
782 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
783 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
784 #endif
785 
786 /*
787  * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
788  */
789 #define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
790 
791 static inline int populated_zone(struct zone *zone)
792 {
793         return (!!zone->present_pages);
794 }
795 
796 extern int movable_zone;
797 
798 #ifdef CONFIG_HIGHMEM
799 static inline int zone_movable_is_highmem(void)
800 {
801 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
802         return movable_zone == ZONE_HIGHMEM;
803 #else
804         return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
805 #endif
806 }
807 #endif
808 
809 static inline int is_highmem_idx(enum zone_type idx)
810 {
811 #ifdef CONFIG_HIGHMEM
812         return (idx == ZONE_HIGHMEM ||
813                 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
814 #else
815         return 0;
816 #endif
817 }
818 
819 /**
820  * is_highmem - helper function to quickly check if a struct zone is a 
821  *              highmem zone or not.  This is an attempt to keep references
822  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
823  * @zone - pointer to struct zone variable
824  */
825 static inline int is_highmem(struct zone *zone)
826 {
827 #ifdef CONFIG_HIGHMEM
828         return is_highmem_idx(zone_idx(zone));
829 #else
830         return 0;
831 #endif
832 }
833 
834 /* These two functions are used to setup the per zone pages min values */
835 struct ctl_table;
836 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
837                                         void __user *, size_t *, loff_t *);
838 int watermark_scale_factor_sysctl_handler(struct ctl_table *, int,
839                                         void __user *, size_t *, loff_t *);
840 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
841 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
842                                         void __user *, size_t *, loff_t *);
843 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
844                                         void __user *, size_t *, loff_t *);
845 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
846                         void __user *, size_t *, loff_t *);
847 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
848                         void __user *, size_t *, loff_t *);
849 
850 extern int numa_zonelist_order_handler(struct ctl_table *, int,
851                         void __user *, size_t *, loff_t *);
852 extern char numa_zonelist_order[];
853 #define NUMA_ZONELIST_ORDER_LEN 16      /* string buffer size */
854 
855 #ifndef CONFIG_NEED_MULTIPLE_NODES
856 
857 extern struct pglist_data contig_page_data;
858 #define NODE_DATA(nid)          (&contig_page_data)
859 #define NODE_MEM_MAP(nid)       mem_map
860 
861 #else /* CONFIG_NEED_MULTIPLE_NODES */
862 
863 #include <asm/mmzone.h>
864 
865 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
866 
867 extern struct pglist_data *first_online_pgdat(void);
868 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
869 extern struct zone *next_zone(struct zone *zone);
870 
871 /**
872  * for_each_online_pgdat - helper macro to iterate over all online nodes
873  * @pgdat - pointer to a pg_data_t variable
874  */
875 #define for_each_online_pgdat(pgdat)                    \
876         for (pgdat = first_online_pgdat();              \
877              pgdat;                                     \
878              pgdat = next_online_pgdat(pgdat))
879 /**
880  * for_each_zone - helper macro to iterate over all memory zones
881  * @zone - pointer to struct zone variable
882  *
883  * The user only needs to declare the zone variable, for_each_zone
884  * fills it in.
885  */
886 #define for_each_zone(zone)                             \
887         for (zone = (first_online_pgdat())->node_zones; \
888              zone;                                      \
889              zone = next_zone(zone))
890 
891 #define for_each_populated_zone(zone)                   \
892         for (zone = (first_online_pgdat())->node_zones; \
893              zone;                                      \
894              zone = next_zone(zone))                    \
895                 if (!populated_zone(zone))              \
896                         ; /* do nothing */              \
897                 else
898 
899 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
900 {
901         return zoneref->zone;
902 }
903 
904 static inline int zonelist_zone_idx(struct zoneref *zoneref)
905 {
906         return zoneref->zone_idx;
907 }
908 
909 static inline int zonelist_node_idx(struct zoneref *zoneref)
910 {
911 #ifdef CONFIG_NUMA
912         /* zone_to_nid not available in this context */
913         return zoneref->zone->node;
914 #else
915         return 0;
916 #endif /* CONFIG_NUMA */
917 }
918 
919 struct zoneref *__next_zones_zonelist(struct zoneref *z,
920                                         enum zone_type highest_zoneidx,
921                                         nodemask_t *nodes);
922 
923 /**
924  * 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
925  * @z - The cursor used as a starting point for the search
926  * @highest_zoneidx - The zone index of the highest zone to return
927  * @nodes - An optional nodemask to filter the zonelist with
928  *
929  * This function returns the next zone at or below a given zone index that is
930  * within the allowed nodemask using a cursor as the starting point for the
931  * search. The zoneref returned is a cursor that represents the current zone
932  * being examined. It should be advanced by one before calling
933  * next_zones_zonelist again.
934  */
935 static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,
936                                         enum zone_type highest_zoneidx,
937                                         nodemask_t *nodes)
938 {
939         if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))
940                 return z;
941         return __next_zones_zonelist(z, highest_zoneidx, nodes);
942 }
943 
944 /**
945  * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
946  * @zonelist - The zonelist to search for a suitable zone
947  * @highest_zoneidx - The zone index of the highest zone to return
948  * @nodes - An optional nodemask to filter the zonelist with
949  * @zone - The first suitable zone found is returned via this parameter
950  *
951  * This function returns the first zone at or below a given zone index that is
952  * within the allowed nodemask. The zoneref returned is a cursor that can be
953  * used to iterate the zonelist with next_zones_zonelist by advancing it by
954  * one before calling.
955  */
956 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
957                                         enum zone_type highest_zoneidx,
958                                         nodemask_t *nodes)
959 {
960         return next_zones_zonelist(zonelist->_zonerefs,
961                                                         highest_zoneidx, nodes);
962 }
963 
964 /**
965  * 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
966  * @zone - The current zone in the iterator
967  * @z - The current pointer within zonelist->zones being iterated
968  * @zlist - The zonelist being iterated
969  * @highidx - The zone index of the highest zone to return
970  * @nodemask - Nodemask allowed by the allocator
971  *
972  * This iterator iterates though all zones at or below a given zone index and
973  * within a given nodemask
974  */
975 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
976         for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z);       \
977                 zone;                                                   \
978                 z = next_zones_zonelist(++z, highidx, nodemask),        \
979                         zone = zonelist_zone(z))
980 
981 #define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
982         for (zone = z->zone;    \
983                 zone;                                                   \
984                 z = next_zones_zonelist(++z, highidx, nodemask),        \
985                         zone = zonelist_zone(z))
986 
987 
988 /**
989  * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
990  * @zone - The current zone in the iterator
991  * @z - The current pointer within zonelist->zones being iterated
992  * @zlist - The zonelist being iterated
993  * @highidx - The zone index of the highest zone to return
994  *
995  * This iterator iterates though all zones at or below a given zone index.
996  */
997 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
998         for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
999 
1000 #ifdef CONFIG_SPARSEMEM
1001 #include <asm/sparsemem.h>
1002 #endif
1003 
1004 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1005         !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1006 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1007 {
1008         return 0;
1009 }
1010 #endif
1011 
1012 #ifdef CONFIG_FLATMEM
1013 #define pfn_to_nid(pfn)         (0)
1014 #endif
1015 
1016 #ifdef CONFIG_SPARSEMEM
1017 
1018 /*
1019  * SECTION_SHIFT                #bits space required to store a section #
1020  *
1021  * PA_SECTION_SHIFT             physical address to/from section number
1022  * PFN_SECTION_SHIFT            pfn to/from section number
1023  */
1024 #define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
1025 #define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
1026 
1027 #define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
1028 
1029 #define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
1030 #define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
1031 
1032 #define SECTION_BLOCKFLAGS_BITS \
1033         ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1034 
1035 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1036 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1037 #endif
1038 
1039 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1040 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1041 
1042 #define SECTION_ALIGN_UP(pfn)   (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1043 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1044 
1045 struct page;
1046 struct page_ext;
1047 struct mem_section {
1048         /*
1049          * This is, logically, a pointer to an array of struct
1050          * pages.  However, it is stored with some other magic.
1051          * (see sparse.c::sparse_init_one_section())
1052          *
1053          * Additionally during early boot we encode node id of
1054          * the location of the section here to guide allocation.
1055          * (see sparse.c::memory_present())
1056          *
1057          * Making it a UL at least makes someone do a cast
1058          * before using it wrong.
1059          */
1060         unsigned long section_mem_map;
1061 
1062         /* See declaration of similar field in struct zone */
1063         unsigned long *pageblock_flags;
1064 #ifdef CONFIG_PAGE_EXTENSION
1065         /*
1066          * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1067          * section. (see page_ext.h about this.)
1068          */
1069         struct page_ext *page_ext;
1070         unsigned long pad;
1071 #endif
1072         /*
1073          * WARNING: mem_section must be a power-of-2 in size for the
1074          * calculation and use of SECTION_ROOT_MASK to make sense.
1075          */
1076 };
1077 
1078 #ifdef CONFIG_SPARSEMEM_EXTREME
1079 #define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
1080 #else
1081 #define SECTIONS_PER_ROOT       1
1082 #endif
1083 
1084 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1085 #define NR_SECTION_ROOTS        DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1086 #define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
1087 
1088 #ifdef CONFIG_SPARSEMEM_EXTREME
1089 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1090 #else
1091 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1092 #endif
1093 
1094 static inline struct mem_section *__nr_to_section(unsigned long nr)
1095 {
1096         if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1097                 return NULL;
1098         return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1099 }
1100 extern int __section_nr(struct mem_section* ms);
1101 extern unsigned long usemap_size(void);
1102 
1103 /*
1104  * We use the lower bits of the mem_map pointer to store
1105  * a little bit of information.  There should be at least
1106  * 3 bits here due to 32-bit alignment.
1107  */
1108 #define SECTION_MARKED_PRESENT  (1UL<<0)
1109 #define SECTION_HAS_MEM_MAP     (1UL<<1)
1110 #define SECTION_MAP_LAST_BIT    (1UL<<2)
1111 #define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
1112 #define SECTION_NID_SHIFT       2
1113 
1114 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1115 {
1116         unsigned long map = section->section_mem_map;
1117         map &= SECTION_MAP_MASK;
1118         return (struct page *)map;
1119 }
1120 
1121 static inline int present_section(struct mem_section *section)
1122 {
1123         return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1124 }
1125 
1126 static inline int present_section_nr(unsigned long nr)
1127 {
1128         return present_section(__nr_to_section(nr));
1129 }
1130 
1131 static inline int valid_section(struct mem_section *section)
1132 {
1133         return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1134 }
1135 
1136 static inline int valid_section_nr(unsigned long nr)
1137 {
1138         return valid_section(__nr_to_section(nr));
1139 }
1140 
1141 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1142 {
1143         return __nr_to_section(pfn_to_section_nr(pfn));
1144 }
1145 
1146 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1147 static inline int pfn_valid(unsigned long pfn)
1148 {
1149         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1150                 return 0;
1151         return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1152 }
1153 #endif
1154 
1155 static inline int pfn_present(unsigned long pfn)
1156 {
1157         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1158                 return 0;
1159         return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1160 }
1161 
1162 /*
1163  * These are _only_ used during initialisation, therefore they
1164  * can use __initdata ...  They could have names to indicate
1165  * this restriction.
1166  */
1167 #ifdef CONFIG_NUMA
1168 #define pfn_to_nid(pfn)                                                 \
1169 ({                                                                      \
1170         unsigned long __pfn_to_nid_pfn = (pfn);                         \
1171         page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
1172 })
1173 #else
1174 #define pfn_to_nid(pfn)         (0)
1175 #endif
1176 
1177 #define early_pfn_valid(pfn)    pfn_valid(pfn)
1178 void sparse_init(void);
1179 #else
1180 #define sparse_init()   do {} while (0)
1181 #define sparse_index_init(_sec, _nid)  do {} while (0)
1182 #endif /* CONFIG_SPARSEMEM */
1183 
1184 /*
1185  * During memory init memblocks map pfns to nids. The search is expensive and
1186  * this caches recent lookups. The implementation of __early_pfn_to_nid
1187  * may treat start/end as pfns or sections.
1188  */
1189 struct mminit_pfnnid_cache {
1190         unsigned long last_start;
1191         unsigned long last_end;
1192         int last_nid;
1193 };
1194 
1195 #ifndef early_pfn_valid
1196 #define early_pfn_valid(pfn)    (1)
1197 #endif
1198 
1199 void memory_present(int nid, unsigned long start, unsigned long end);
1200 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1201 
1202 /*
1203  * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1204  * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1205  * pfn_valid_within() should be used in this case; we optimise this away
1206  * when we have no holes within a MAX_ORDER_NR_PAGES block.
1207  */
1208 #ifdef CONFIG_HOLES_IN_ZONE
1209 #define pfn_valid_within(pfn) pfn_valid(pfn)
1210 #else
1211 #define pfn_valid_within(pfn) (1)
1212 #endif
1213 
1214 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1215 /*
1216  * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1217  * associated with it or not. In FLATMEM, it is expected that holes always
1218  * have valid memmap as long as there is valid PFNs either side of the hole.
1219  * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1220  * entire section.
1221  *
1222  * However, an ARM, and maybe other embedded architectures in the future
1223  * free memmap backing holes to save memory on the assumption the memmap is
1224  * never used. The page_zone linkages are then broken even though pfn_valid()
1225  * returns true. A walker of the full memmap must then do this additional
1226  * check to ensure the memmap they are looking at is sane by making sure
1227  * the zone and PFN linkages are still valid. This is expensive, but walkers
1228  * of the full memmap are extremely rare.
1229  */
1230 bool memmap_valid_within(unsigned long pfn,
1231                                         struct page *page, struct zone *zone);
1232 #else
1233 static inline bool memmap_valid_within(unsigned long pfn,
1234                                         struct page *page, struct zone *zone)
1235 {
1236         return true;
1237 }
1238 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1239 
1240 #endif /* !__GENERATING_BOUNDS.H */
1241 #endif /* !__ASSEMBLY__ */
1242 #endif /* _LINUX_MMZONE_H */
1243 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us