~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/linux/mmzone.h

Version: ~ [ linux-5.16-rc3 ] ~ [ linux-5.15.5 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.82 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.162 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.218 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.256 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.291 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.293 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 migratetype {
 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 #  define is_migrate_cma_page(_page) (get_pageblock_migratetype(_page) == MIGRATE_CMA)
 72 #else
 73 #  define is_migrate_cma(migratetype) false
 74 #  define is_migrate_cma_page(_page) false
 75 #endif
 76 
 77 static inline bool is_migrate_movable(int mt)
 78 {
 79         return is_migrate_cma(mt) || mt == MIGRATE_MOVABLE;
 80 }
 81 
 82 #define for_each_migratetype_order(order, type) \
 83         for (order = 0; order < MAX_ORDER; order++) \
 84                 for (type = 0; type < MIGRATE_TYPES; type++)
 85 
 86 extern int page_group_by_mobility_disabled;
 87 
 88 #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
 89 #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
 90 
 91 #define get_pageblock_migratetype(page)                                 \
 92         get_pfnblock_flags_mask(page, page_to_pfn(page),                \
 93                         PB_migrate_end, MIGRATETYPE_MASK)
 94 
 95 struct free_area {
 96         struct list_head        free_list[MIGRATE_TYPES];
 97         unsigned long           nr_free;
 98 };
 99 
100 struct pglist_data;
101 
102 /*
103  * zone->lock and the zone lru_lock are two of the hottest locks in the kernel.
104  * So add a wild amount of padding here to ensure that they fall into separate
105  * cachelines.  There are very few zone structures in the machine, so space
106  * consumption is not a concern here.
107  */
108 #if defined(CONFIG_SMP)
109 struct zone_padding {
110         char x[0];
111 } ____cacheline_internodealigned_in_smp;
112 #define ZONE_PADDING(name)      struct zone_padding name;
113 #else
114 #define ZONE_PADDING(name)
115 #endif
116 
117 enum zone_stat_item {
118         /* First 128 byte cacheline (assuming 64 bit words) */
119         NR_FREE_PAGES,
120         NR_ZONE_LRU_BASE, /* Used only for compaction and reclaim retry */
121         NR_ZONE_INACTIVE_ANON = NR_ZONE_LRU_BASE,
122         NR_ZONE_ACTIVE_ANON,
123         NR_ZONE_INACTIVE_FILE,
124         NR_ZONE_ACTIVE_FILE,
125         NR_ZONE_UNEVICTABLE,
126         NR_ZONE_WRITE_PENDING,  /* Count of dirty, writeback and unstable pages */
127         NR_MLOCK,               /* mlock()ed pages found and moved off LRU */
128         NR_SLAB_RECLAIMABLE,
129         NR_SLAB_UNRECLAIMABLE,
130         NR_PAGETABLE,           /* used for pagetables */
131         NR_KERNEL_STACK_KB,     /* measured in KiB */
132         /* Second 128 byte cacheline */
133         NR_BOUNCE,
134 #if IS_ENABLED(CONFIG_ZSMALLOC)
135         NR_ZSPAGES,             /* allocated in zsmalloc */
136 #endif
137 #ifdef CONFIG_NUMA
138         NUMA_HIT,               /* allocated in intended node */
139         NUMA_MISS,              /* allocated in non intended node */
140         NUMA_FOREIGN,           /* was intended here, hit elsewhere */
141         NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
142         NUMA_LOCAL,             /* allocation from local node */
143         NUMA_OTHER,             /* allocation from other node */
144 #endif
145         NR_FREE_CMA_PAGES,
146         NR_VM_ZONE_STAT_ITEMS };
147 
148 enum node_stat_item {
149         NR_LRU_BASE,
150         NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
151         NR_ACTIVE_ANON,         /*  "     "     "   "       "         */
152         NR_INACTIVE_FILE,       /*  "     "     "   "       "         */
153         NR_ACTIVE_FILE,         /*  "     "     "   "       "         */
154         NR_UNEVICTABLE,         /*  "     "     "   "       "         */
155         NR_ISOLATED_ANON,       /* Temporary isolated pages from anon lru */
156         NR_ISOLATED_FILE,       /* Temporary isolated pages from file lru */
157         WORKINGSET_REFAULT,
158         WORKINGSET_ACTIVATE,
159         WORKINGSET_NODERECLAIM,
160         NR_ANON_MAPPED, /* Mapped anonymous pages */
161         NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
162                            only modified from process context */
163         NR_FILE_PAGES,
164         NR_FILE_DIRTY,
165         NR_WRITEBACK,
166         NR_WRITEBACK_TEMP,      /* Writeback using temporary buffers */
167         NR_SHMEM,               /* shmem pages (included tmpfs/GEM pages) */
168         NR_SHMEM_THPS,
169         NR_SHMEM_PMDMAPPED,
170         NR_ANON_THPS,
171         NR_UNSTABLE_NFS,        /* NFS unstable pages */
172         NR_VMSCAN_WRITE,
173         NR_VMSCAN_IMMEDIATE,    /* Prioritise for reclaim when writeback ends */
174         NR_DIRTIED,             /* page dirtyings since bootup */
175         NR_WRITTEN,             /* page writings since bootup */
176         NR_VM_NODE_STAT_ITEMS
177 };
178 
179 /*
180  * We do arithmetic on the LRU lists in various places in the code,
181  * so it is important to keep the active lists LRU_ACTIVE higher in
182  * the array than the corresponding inactive lists, and to keep
183  * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
184  *
185  * This has to be kept in sync with the statistics in zone_stat_item
186  * above and the descriptions in vmstat_text in mm/vmstat.c
187  */
188 #define LRU_BASE 0
189 #define LRU_ACTIVE 1
190 #define LRU_FILE 2
191 
192 enum lru_list {
193         LRU_INACTIVE_ANON = LRU_BASE,
194         LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
195         LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
196         LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
197         LRU_UNEVICTABLE,
198         NR_LRU_LISTS
199 };
200 
201 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
202 
203 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
204 
205 static inline int is_file_lru(enum lru_list lru)
206 {
207         return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
208 }
209 
210 static inline int is_active_lru(enum lru_list lru)
211 {
212         return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
213 }
214 
215 struct zone_reclaim_stat {
216         /*
217          * The pageout code in vmscan.c keeps track of how many of the
218          * mem/swap backed and file backed pages are referenced.
219          * The higher the rotated/scanned ratio, the more valuable
220          * that cache is.
221          *
222          * The anon LRU stats live in [0], file LRU stats in [1]
223          */
224         unsigned long           recent_rotated[2];
225         unsigned long           recent_scanned[2];
226 };
227 
228 struct lruvec {
229         struct list_head                lists[NR_LRU_LISTS];
230         struct zone_reclaim_stat        reclaim_stat;
231         /* Evictions & activations on the inactive file list */
232         atomic_long_t                   inactive_age;
233         /* Refaults at the time of last reclaim cycle */
234         unsigned long                   refaults;
235 #ifdef CONFIG_MEMCG
236         struct pglist_data *pgdat;
237 #endif
238 };
239 
240 /* Mask used at gathering information at once (see memcontrol.c) */
241 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
242 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
243 #define LRU_ALL      ((1 << NR_LRU_LISTS) - 1)
244 
245 /* Isolate unmapped file */
246 #define ISOLATE_UNMAPPED        ((__force isolate_mode_t)0x2)
247 /* Isolate for asynchronous migration */
248 #define ISOLATE_ASYNC_MIGRATE   ((__force isolate_mode_t)0x4)
249 /* Isolate unevictable pages */
250 #define ISOLATE_UNEVICTABLE     ((__force isolate_mode_t)0x8)
251 
252 /* LRU Isolation modes. */
253 typedef unsigned __bitwise isolate_mode_t;
254 
255 enum zone_watermarks {
256         WMARK_MIN,
257         WMARK_LOW,
258         WMARK_HIGH,
259         NR_WMARK
260 };
261 
262 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
263 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
264 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
265 
266 struct per_cpu_pages {
267         int count;              /* number of pages in the list */
268         int high;               /* high watermark, emptying needed */
269         int batch;              /* chunk size for buddy add/remove */
270 
271         /* Lists of pages, one per migrate type stored on the pcp-lists */
272         struct list_head lists[MIGRATE_PCPTYPES];
273 };
274 
275 struct per_cpu_pageset {
276         struct per_cpu_pages pcp;
277 #ifdef CONFIG_NUMA
278         s8 expire;
279 #endif
280 #ifdef CONFIG_SMP
281         s8 stat_threshold;
282         s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
283 #endif
284 };
285 
286 struct per_cpu_nodestat {
287         s8 stat_threshold;
288         s8 vm_node_stat_diff[NR_VM_NODE_STAT_ITEMS];
289 };
290 
291 #endif /* !__GENERATING_BOUNDS.H */
292 
293 enum zone_type {
294 #ifdef CONFIG_ZONE_DMA
295         /*
296          * ZONE_DMA is used when there are devices that are not able
297          * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
298          * carve out the portion of memory that is needed for these devices.
299          * The range is arch specific.
300          *
301          * Some examples
302          *
303          * Architecture         Limit
304          * ---------------------------
305          * parisc, ia64, sparc  <4G
306          * s390                 <2G
307          * arm                  Various
308          * alpha                Unlimited or 0-16MB.
309          *
310          * i386, x86_64 and multiple other arches
311          *                      <16M.
312          */
313         ZONE_DMA,
314 #endif
315 #ifdef CONFIG_ZONE_DMA32
316         /*
317          * x86_64 needs two ZONE_DMAs because it supports devices that are
318          * only able to do DMA to the lower 16M but also 32 bit devices that
319          * can only do DMA areas below 4G.
320          */
321         ZONE_DMA32,
322 #endif
323         /*
324          * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
325          * performed on pages in ZONE_NORMAL if the DMA devices support
326          * transfers to all addressable memory.
327          */
328         ZONE_NORMAL,
329 #ifdef CONFIG_HIGHMEM
330         /*
331          * A memory area that is only addressable by the kernel through
332          * mapping portions into its own address space. This is for example
333          * used by i386 to allow the kernel to address the memory beyond
334          * 900MB. The kernel will set up special mappings (page
335          * table entries on i386) for each page that the kernel needs to
336          * access.
337          */
338         ZONE_HIGHMEM,
339 #endif
340         ZONE_MOVABLE,
341 #ifdef CONFIG_ZONE_DEVICE
342         ZONE_DEVICE,
343 #endif
344         __MAX_NR_ZONES
345 
346 };
347 
348 #ifndef __GENERATING_BOUNDS_H
349 
350 struct zone {
351         /* Read-mostly fields */
352 
353         /* zone watermarks, access with *_wmark_pages(zone) macros */
354         unsigned long watermark[NR_WMARK];
355 
356         unsigned long nr_reserved_highatomic;
357 
358         /*
359          * We don't know if the memory that we're going to allocate will be
360          * freeable or/and it will be released eventually, so to avoid totally
361          * wasting several GB of ram we must reserve some of the lower zone
362          * memory (otherwise we risk to run OOM on the lower zones despite
363          * there being tons of freeable ram on the higher zones).  This array is
364          * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
365          * changes.
366          */
367         long lowmem_reserve[MAX_NR_ZONES];
368 
369 #ifdef CONFIG_NUMA
370         int node;
371 #endif
372         struct pglist_data      *zone_pgdat;
373         struct per_cpu_pageset __percpu *pageset;
374 
375 #ifndef CONFIG_SPARSEMEM
376         /*
377          * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
378          * In SPARSEMEM, this map is stored in struct mem_section
379          */
380         unsigned long           *pageblock_flags;
381 #endif /* CONFIG_SPARSEMEM */
382 
383         /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
384         unsigned long           zone_start_pfn;
385 
386         /*
387          * spanned_pages is the total pages spanned by the zone, including
388          * holes, which is calculated as:
389          *      spanned_pages = zone_end_pfn - zone_start_pfn;
390          *
391          * present_pages is physical pages existing within the zone, which
392          * is calculated as:
393          *      present_pages = spanned_pages - absent_pages(pages in holes);
394          *
395          * managed_pages is present pages managed by the buddy system, which
396          * is calculated as (reserved_pages includes pages allocated by the
397          * bootmem allocator):
398          *      managed_pages = present_pages - reserved_pages;
399          *
400          * So present_pages may be used by memory hotplug or memory power
401          * management logic to figure out unmanaged pages by checking
402          * (present_pages - managed_pages). And managed_pages should be used
403          * by page allocator and vm scanner to calculate all kinds of watermarks
404          * and thresholds.
405          *
406          * Locking rules:
407          *
408          * zone_start_pfn and spanned_pages are protected by span_seqlock.
409          * It is a seqlock because it has to be read outside of zone->lock,
410          * and it is done in the main allocator path.  But, it is written
411          * quite infrequently.
412          *
413          * The span_seq lock is declared along with zone->lock because it is
414          * frequently read in proximity to zone->lock.  It's good to
415          * give them a chance of being in the same cacheline.
416          *
417          * Write access to present_pages at runtime should be protected by
418          * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
419          * present_pages should get_online_mems() to get a stable value.
420          *
421          * Read access to managed_pages should be safe because it's unsigned
422          * long. Write access to zone->managed_pages and totalram_pages are
423          * protected by managed_page_count_lock at runtime. Idealy only
424          * adjust_managed_page_count() should be used instead of directly
425          * touching zone->managed_pages and totalram_pages.
426          */
427         unsigned long           managed_pages;
428         unsigned long           spanned_pages;
429         unsigned long           present_pages;
430 
431         const char              *name;
432 
433 #ifdef CONFIG_MEMORY_ISOLATION
434         /*
435          * Number of isolated pageblock. It is used to solve incorrect
436          * freepage counting problem due to racy retrieving migratetype
437          * of pageblock. Protected by zone->lock.
438          */
439         unsigned long           nr_isolate_pageblock;
440 #endif
441 
442 #ifdef CONFIG_MEMORY_HOTPLUG
443         /* see spanned/present_pages for more description */
444         seqlock_t               span_seqlock;
445 #endif
446 
447         int initialized;
448 
449         /* Write-intensive fields used from the page allocator */
450         ZONE_PADDING(_pad1_)
451 
452         /* free areas of different sizes */
453         struct free_area        free_area[MAX_ORDER];
454 
455         /* zone flags, see below */
456         unsigned long           flags;
457 
458         /* Primarily protects free_area */
459         spinlock_t              lock;
460 
461         /* Write-intensive fields used by compaction and vmstats. */
462         ZONE_PADDING(_pad2_)
463 
464         /*
465          * When free pages are below this point, additional steps are taken
466          * when reading the number of free pages to avoid per-cpu counter
467          * drift allowing watermarks to be breached
468          */
469         unsigned long percpu_drift_mark;
470 
471 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
472         /* pfn where compaction free scanner should start */
473         unsigned long           compact_cached_free_pfn;
474         /* pfn where async and sync compaction migration scanner should start */
475         unsigned long           compact_cached_migrate_pfn[2];
476 #endif
477 
478 #ifdef CONFIG_COMPACTION
479         /*
480          * On compaction failure, 1<<compact_defer_shift compactions
481          * are skipped before trying again. The number attempted since
482          * last failure is tracked with compact_considered.
483          */
484         unsigned int            compact_considered;
485         unsigned int            compact_defer_shift;
486         int                     compact_order_failed;
487 #endif
488 
489 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
490         /* Set to true when the PG_migrate_skip bits should be cleared */
491         bool                    compact_blockskip_flush;
492 #endif
493 
494         bool                    contiguous;
495 
496         ZONE_PADDING(_pad3_)
497         /* Zone statistics */
498         atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
499 } ____cacheline_internodealigned_in_smp;
500 
501 enum pgdat_flags {
502         PGDAT_CONGESTED,                /* pgdat has many dirty pages backed by
503                                          * a congested BDI
504                                          */
505         PGDAT_DIRTY,                    /* reclaim scanning has recently found
506                                          * many dirty file pages at the tail
507                                          * of the LRU.
508                                          */
509         PGDAT_WRITEBACK,                /* reclaim scanning has recently found
510                                          * many pages under writeback
511                                          */
512         PGDAT_RECLAIM_LOCKED,           /* prevents concurrent reclaim */
513 };
514 
515 static inline unsigned long zone_end_pfn(const struct zone *zone)
516 {
517         return zone->zone_start_pfn + zone->spanned_pages;
518 }
519 
520 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
521 {
522         return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
523 }
524 
525 static inline bool zone_is_initialized(struct zone *zone)
526 {
527         return zone->initialized;
528 }
529 
530 static inline bool zone_is_empty(struct zone *zone)
531 {
532         return zone->spanned_pages == 0;
533 }
534 
535 /*
536  * The "priority" of VM scanning is how much of the queues we will scan in one
537  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
538  * queues ("queue_length >> 12") during an aging round.
539  */
540 #define DEF_PRIORITY 12
541 
542 /* Maximum number of zones on a zonelist */
543 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
544 
545 enum {
546         ZONELIST_FALLBACK,      /* zonelist with fallback */
547 #ifdef CONFIG_NUMA
548         /*
549          * The NUMA zonelists are doubled because we need zonelists that
550          * restrict the allocations to a single node for __GFP_THISNODE.
551          */
552         ZONELIST_NOFALLBACK,    /* zonelist without fallback (__GFP_THISNODE) */
553 #endif
554         MAX_ZONELISTS
555 };
556 
557 /*
558  * This struct contains information about a zone in a zonelist. It is stored
559  * here to avoid dereferences into large structures and lookups of tables
560  */
561 struct zoneref {
562         struct zone *zone;      /* Pointer to actual zone */
563         int zone_idx;           /* zone_idx(zoneref->zone) */
564 };
565 
566 /*
567  * One allocation request operates on a zonelist. A zonelist
568  * is a list of zones, the first one is the 'goal' of the
569  * allocation, the other zones are fallback zones, in decreasing
570  * priority.
571  *
572  * To speed the reading of the zonelist, the zonerefs contain the zone index
573  * of the entry being read. Helper functions to access information given
574  * a struct zoneref are
575  *
576  * zonelist_zone()      - Return the struct zone * for an entry in _zonerefs
577  * zonelist_zone_idx()  - Return the index of the zone for an entry
578  * zonelist_node_idx()  - Return the index of the node for an entry
579  */
580 struct zonelist {
581         struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
582 };
583 
584 #ifndef CONFIG_DISCONTIGMEM
585 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
586 extern struct page *mem_map;
587 #endif
588 
589 /*
590  * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
591  * (mostly NUMA machines?) to denote a higher-level memory zone than the
592  * zone denotes.
593  *
594  * On NUMA machines, each NUMA node would have a pg_data_t to describe
595  * it's memory layout.
596  *
597  * Memory statistics and page replacement data structures are maintained on a
598  * per-zone basis.
599  */
600 struct bootmem_data;
601 typedef struct pglist_data {
602         struct zone node_zones[MAX_NR_ZONES];
603         struct zonelist node_zonelists[MAX_ZONELISTS];
604         int nr_zones;
605 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
606         struct page *node_mem_map;
607 #ifdef CONFIG_PAGE_EXTENSION
608         struct page_ext *node_page_ext;
609 #endif
610 #endif
611 #ifndef CONFIG_NO_BOOTMEM
612         struct bootmem_data *bdata;
613 #endif
614 #ifdef CONFIG_MEMORY_HOTPLUG
615         /*
616          * Must be held any time you expect node_start_pfn, node_present_pages
617          * or node_spanned_pages stay constant.  Holding this will also
618          * guarantee that any pfn_valid() stays that way.
619          *
620          * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
621          * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
622          *
623          * Nests above zone->lock and zone->span_seqlock
624          */
625         spinlock_t node_size_lock;
626 #endif
627         unsigned long node_start_pfn;
628         unsigned long node_present_pages; /* total number of physical pages */
629         unsigned long node_spanned_pages; /* total size of physical page
630                                              range, including holes */
631         int node_id;
632         wait_queue_head_t kswapd_wait;
633         wait_queue_head_t pfmemalloc_wait;
634         struct task_struct *kswapd;     /* Protected by
635                                            mem_hotplug_begin/end() */
636         int kswapd_order;
637         enum zone_type kswapd_classzone_idx;
638 
639         int kswapd_failures;            /* Number of 'reclaimed == 0' runs */
640 
641 #ifdef CONFIG_COMPACTION
642         int kcompactd_max_order;
643         enum zone_type kcompactd_classzone_idx;
644         wait_queue_head_t kcompactd_wait;
645         struct task_struct *kcompactd;
646 #endif
647 #ifdef CONFIG_NUMA_BALANCING
648         /* Lock serializing the migrate rate limiting window */
649         spinlock_t numabalancing_migrate_lock;
650 
651         /* Rate limiting time interval */
652         unsigned long numabalancing_migrate_next_window;
653 
654         /* Number of pages migrated during the rate limiting time interval */
655         unsigned long numabalancing_migrate_nr_pages;
656 #endif
657         /*
658          * This is a per-node reserve of pages that are not available
659          * to userspace allocations.
660          */
661         unsigned long           totalreserve_pages;
662 
663 #ifdef CONFIG_NUMA
664         /*
665          * zone reclaim becomes active if more unmapped pages exist.
666          */
667         unsigned long           min_unmapped_pages;
668         unsigned long           min_slab_pages;
669 #endif /* CONFIG_NUMA */
670 
671         /* Write-intensive fields used by page reclaim */
672         ZONE_PADDING(_pad1_)
673         spinlock_t              lru_lock;
674 
675 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
676         /*
677          * If memory initialisation on large machines is deferred then this
678          * is the first PFN that needs to be initialised.
679          */
680         unsigned long first_deferred_pfn;
681         unsigned long static_init_size;
682 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
683 
684 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
685         spinlock_t split_queue_lock;
686         struct list_head split_queue;
687         unsigned long split_queue_len;
688 #endif
689 
690         /* Fields commonly accessed by the page reclaim scanner */
691         struct lruvec           lruvec;
692 
693         /*
694          * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
695          * this node's LRU.  Maintained by the pageout code.
696          */
697         unsigned int inactive_ratio;
698 
699         unsigned long           flags;
700 
701         ZONE_PADDING(_pad2_)
702 
703         /* Per-node vmstats */
704         struct per_cpu_nodestat __percpu *per_cpu_nodestats;
705         atomic_long_t           vm_stat[NR_VM_NODE_STAT_ITEMS];
706 } pg_data_t;
707 
708 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
709 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
710 #ifdef CONFIG_FLAT_NODE_MEM_MAP
711 #define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
712 #else
713 #define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
714 #endif
715 #define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
716 
717 #define node_start_pfn(nid)     (NODE_DATA(nid)->node_start_pfn)
718 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
719 static inline spinlock_t *zone_lru_lock(struct zone *zone)
720 {
721         return &zone->zone_pgdat->lru_lock;
722 }
723 
724 static inline struct lruvec *node_lruvec(struct pglist_data *pgdat)
725 {
726         return &pgdat->lruvec;
727 }
728 
729 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
730 {
731         return pgdat->node_start_pfn + pgdat->node_spanned_pages;
732 }
733 
734 static inline bool pgdat_is_empty(pg_data_t *pgdat)
735 {
736         return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
737 }
738 
739 static inline int zone_id(const struct zone *zone)
740 {
741         struct pglist_data *pgdat = zone->zone_pgdat;
742 
743         return zone - pgdat->node_zones;
744 }
745 
746 #ifdef CONFIG_ZONE_DEVICE
747 static inline bool is_dev_zone(const struct zone *zone)
748 {
749         return zone_id(zone) == ZONE_DEVICE;
750 }
751 #else
752 static inline bool is_dev_zone(const struct zone *zone)
753 {
754         return false;
755 }
756 #endif
757 
758 #include <linux/memory_hotplug.h>
759 
760 extern struct mutex zonelists_mutex;
761 void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
762 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
763 bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
764                          int classzone_idx, unsigned int alloc_flags,
765                          long free_pages);
766 bool zone_watermark_ok(struct zone *z, unsigned int order,
767                 unsigned long mark, int classzone_idx,
768                 unsigned int alloc_flags);
769 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
770                 unsigned long mark, int classzone_idx);
771 enum memmap_context {
772         MEMMAP_EARLY,
773         MEMMAP_HOTPLUG,
774 };
775 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
776                                      unsigned long size);
777 
778 extern void lruvec_init(struct lruvec *lruvec);
779 
780 static inline struct pglist_data *lruvec_pgdat(struct lruvec *lruvec)
781 {
782 #ifdef CONFIG_MEMCG
783         return lruvec->pgdat;
784 #else
785         return container_of(lruvec, struct pglist_data, lruvec);
786 #endif
787 }
788 
789 extern unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru, int zone_idx);
790 
791 #ifdef CONFIG_HAVE_MEMORY_PRESENT
792 void memory_present(int nid, unsigned long start, unsigned long end);
793 #else
794 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
795 #endif
796 
797 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
798 int local_memory_node(int node_id);
799 #else
800 static inline int local_memory_node(int node_id) { return node_id; };
801 #endif
802 
803 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
804 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
805 #endif
806 
807 /*
808  * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
809  */
810 #define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
811 
812 /*
813  * Returns true if a zone has pages managed by the buddy allocator.
814  * All the reclaim decisions have to use this function rather than
815  * populated_zone(). If the whole zone is reserved then we can easily
816  * end up with populated_zone() && !managed_zone().
817  */
818 static inline bool managed_zone(struct zone *zone)
819 {
820         return zone->managed_pages;
821 }
822 
823 /* Returns true if a zone has memory */
824 static inline bool populated_zone(struct zone *zone)
825 {
826         return zone->present_pages;
827 }
828 
829 extern int movable_zone;
830 
831 #ifdef CONFIG_HIGHMEM
832 static inline int zone_movable_is_highmem(void)
833 {
834 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
835         return movable_zone == ZONE_HIGHMEM;
836 #else
837         return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
838 #endif
839 }
840 #endif
841 
842 static inline int is_highmem_idx(enum zone_type idx)
843 {
844 #ifdef CONFIG_HIGHMEM
845         return (idx == ZONE_HIGHMEM ||
846                 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
847 #else
848         return 0;
849 #endif
850 }
851 
852 /**
853  * is_highmem - helper function to quickly check if a struct zone is a 
854  *              highmem zone or not.  This is an attempt to keep references
855  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
856  * @zone - pointer to struct zone variable
857  */
858 static inline int is_highmem(struct zone *zone)
859 {
860 #ifdef CONFIG_HIGHMEM
861         return is_highmem_idx(zone_idx(zone));
862 #else
863         return 0;
864 #endif
865 }
866 
867 /* These two functions are used to setup the per zone pages min values */
868 struct ctl_table;
869 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
870                                         void __user *, size_t *, loff_t *);
871 int watermark_scale_factor_sysctl_handler(struct ctl_table *, int,
872                                         void __user *, size_t *, loff_t *);
873 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
874 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
875                                         void __user *, size_t *, loff_t *);
876 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
877                                         void __user *, size_t *, loff_t *);
878 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
879                         void __user *, size_t *, loff_t *);
880 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
881                         void __user *, size_t *, loff_t *);
882 
883 extern int numa_zonelist_order_handler(struct ctl_table *, int,
884                         void __user *, size_t *, loff_t *);
885 extern char numa_zonelist_order[];
886 #define NUMA_ZONELIST_ORDER_LEN 16      /* string buffer size */
887 
888 #ifndef CONFIG_NEED_MULTIPLE_NODES
889 
890 extern struct pglist_data contig_page_data;
891 #define NODE_DATA(nid)          (&contig_page_data)
892 #define NODE_MEM_MAP(nid)       mem_map
893 
894 #else /* CONFIG_NEED_MULTIPLE_NODES */
895 
896 #include <asm/mmzone.h>
897 
898 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
899 
900 extern struct pglist_data *first_online_pgdat(void);
901 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
902 extern struct zone *next_zone(struct zone *zone);
903 
904 /**
905  * for_each_online_pgdat - helper macro to iterate over all online nodes
906  * @pgdat - pointer to a pg_data_t variable
907  */
908 #define for_each_online_pgdat(pgdat)                    \
909         for (pgdat = first_online_pgdat();              \
910              pgdat;                                     \
911              pgdat = next_online_pgdat(pgdat))
912 /**
913  * for_each_zone - helper macro to iterate over all memory zones
914  * @zone - pointer to struct zone variable
915  *
916  * The user only needs to declare the zone variable, for_each_zone
917  * fills it in.
918  */
919 #define for_each_zone(zone)                             \
920         for (zone = (first_online_pgdat())->node_zones; \
921              zone;                                      \
922              zone = next_zone(zone))
923 
924 #define for_each_populated_zone(zone)                   \
925         for (zone = (first_online_pgdat())->node_zones; \
926              zone;                                      \
927              zone = next_zone(zone))                    \
928                 if (!populated_zone(zone))              \
929                         ; /* do nothing */              \
930                 else
931 
932 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
933 {
934         return zoneref->zone;
935 }
936 
937 static inline int zonelist_zone_idx(struct zoneref *zoneref)
938 {
939         return zoneref->zone_idx;
940 }
941 
942 static inline int zonelist_node_idx(struct zoneref *zoneref)
943 {
944 #ifdef CONFIG_NUMA
945         /* zone_to_nid not available in this context */
946         return zoneref->zone->node;
947 #else
948         return 0;
949 #endif /* CONFIG_NUMA */
950 }
951 
952 struct zoneref *__next_zones_zonelist(struct zoneref *z,
953                                         enum zone_type highest_zoneidx,
954                                         nodemask_t *nodes);
955 
956 /**
957  * 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
958  * @z - The cursor used as a starting point for the search
959  * @highest_zoneidx - The zone index of the highest zone to return
960  * @nodes - An optional nodemask to filter the zonelist with
961  *
962  * This function returns the next zone at or below a given zone index that is
963  * within the allowed nodemask using a cursor as the starting point for the
964  * search. The zoneref returned is a cursor that represents the current zone
965  * being examined. It should be advanced by one before calling
966  * next_zones_zonelist again.
967  */
968 static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,
969                                         enum zone_type highest_zoneidx,
970                                         nodemask_t *nodes)
971 {
972         if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))
973                 return z;
974         return __next_zones_zonelist(z, highest_zoneidx, nodes);
975 }
976 
977 /**
978  * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
979  * @zonelist - The zonelist to search for a suitable zone
980  * @highest_zoneidx - The zone index of the highest zone to return
981  * @nodes - An optional nodemask to filter the zonelist with
982  * @return - Zoneref pointer for the first suitable zone found (see below)
983  *
984  * This function returns the first zone at or below a given zone index that is
985  * within the allowed nodemask. The zoneref returned is a cursor that can be
986  * used to iterate the zonelist with next_zones_zonelist by advancing it by
987  * one before calling.
988  *
989  * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is
990  * never NULL). This may happen either genuinely, or due to concurrent nodemask
991  * update due to cpuset modification.
992  */
993 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
994                                         enum zone_type highest_zoneidx,
995                                         nodemask_t *nodes)
996 {
997         return next_zones_zonelist(zonelist->_zonerefs,
998                                                         highest_zoneidx, nodes);
999 }
1000 
1001 /**
1002  * 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
1003  * @zone - The current zone in the iterator
1004  * @z - The current pointer within zonelist->zones being iterated
1005  * @zlist - The zonelist being iterated
1006  * @highidx - The zone index of the highest zone to return
1007  * @nodemask - Nodemask allowed by the allocator
1008  *
1009  * This iterator iterates though all zones at or below a given zone index and
1010  * within a given nodemask
1011  */
1012 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1013         for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z);       \
1014                 zone;                                                   \
1015                 z = next_zones_zonelist(++z, highidx, nodemask),        \
1016                         zone = zonelist_zone(z))
1017 
1018 #define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1019         for (zone = z->zone;    \
1020                 zone;                                                   \
1021                 z = next_zones_zonelist(++z, highidx, nodemask),        \
1022                         zone = zonelist_zone(z))
1023 
1024 
1025 /**
1026  * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1027  * @zone - The current zone in the iterator
1028  * @z - The current pointer within zonelist->zones being iterated
1029  * @zlist - The zonelist being iterated
1030  * @highidx - The zone index of the highest zone to return
1031  *
1032  * This iterator iterates though all zones at or below a given zone index.
1033  */
1034 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
1035         for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1036 
1037 #ifdef CONFIG_SPARSEMEM
1038 #include <asm/sparsemem.h>
1039 #endif
1040 
1041 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1042         !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1043 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1044 {
1045         return 0;
1046 }
1047 #endif
1048 
1049 #ifdef CONFIG_FLATMEM
1050 #define pfn_to_nid(pfn)         (0)
1051 #endif
1052 
1053 #ifdef CONFIG_SPARSEMEM
1054 
1055 /*
1056  * SECTION_SHIFT                #bits space required to store a section #
1057  *
1058  * PA_SECTION_SHIFT             physical address to/from section number
1059  * PFN_SECTION_SHIFT            pfn to/from section number
1060  */
1061 #define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
1062 #define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
1063 
1064 #define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
1065 
1066 #define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
1067 #define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
1068 
1069 #define SECTION_BLOCKFLAGS_BITS \
1070         ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1071 
1072 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1073 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1074 #endif
1075 
1076 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1077 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1078 
1079 #define SECTION_ALIGN_UP(pfn)   (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1080 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1081 
1082 struct page;
1083 struct page_ext;
1084 struct mem_section {
1085         /*
1086          * This is, logically, a pointer to an array of struct
1087          * pages.  However, it is stored with some other magic.
1088          * (see sparse.c::sparse_init_one_section())
1089          *
1090          * Additionally during early boot we encode node id of
1091          * the location of the section here to guide allocation.
1092          * (see sparse.c::memory_present())
1093          *
1094          * Making it a UL at least makes someone do a cast
1095          * before using it wrong.
1096          */
1097         unsigned long section_mem_map;
1098 
1099         /* See declaration of similar field in struct zone */
1100         unsigned long *pageblock_flags;
1101 #ifdef CONFIG_PAGE_EXTENSION
1102         /*
1103          * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1104          * section. (see page_ext.h about this.)
1105          */
1106         struct page_ext *page_ext;
1107         unsigned long pad;
1108 #endif
1109         /*
1110          * WARNING: mem_section must be a power-of-2 in size for the
1111          * calculation and use of SECTION_ROOT_MASK to make sense.
1112          */
1113 };
1114 
1115 #ifdef CONFIG_SPARSEMEM_EXTREME
1116 #define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
1117 #else
1118 #define SECTIONS_PER_ROOT       1
1119 #endif
1120 
1121 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1122 #define NR_SECTION_ROOTS        DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1123 #define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
1124 
1125 #ifdef CONFIG_SPARSEMEM_EXTREME
1126 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1127 #else
1128 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1129 #endif
1130 
1131 static inline struct mem_section *__nr_to_section(unsigned long nr)
1132 {
1133         if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1134                 return NULL;
1135         return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1136 }
1137 extern int __section_nr(struct mem_section* ms);
1138 extern unsigned long usemap_size(void);
1139 
1140 /*
1141  * We use the lower bits of the mem_map pointer to store
1142  * a little bit of information.  There should be at least
1143  * 3 bits here due to 32-bit alignment.
1144  */
1145 #define SECTION_MARKED_PRESENT  (1UL<<0)
1146 #define SECTION_HAS_MEM_MAP     (1UL<<1)
1147 #define SECTION_MAP_LAST_BIT    (1UL<<2)
1148 #define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
1149 #define SECTION_NID_SHIFT       2
1150 
1151 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1152 {
1153         unsigned long map = section->section_mem_map;
1154         map &= SECTION_MAP_MASK;
1155         return (struct page *)map;
1156 }
1157 
1158 static inline int present_section(struct mem_section *section)
1159 {
1160         return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1161 }
1162 
1163 static inline int present_section_nr(unsigned long nr)
1164 {
1165         return present_section(__nr_to_section(nr));
1166 }
1167 
1168 static inline int valid_section(struct mem_section *section)
1169 {
1170         return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1171 }
1172 
1173 static inline int valid_section_nr(unsigned long nr)
1174 {
1175         return valid_section(__nr_to_section(nr));
1176 }
1177 
1178 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1179 {
1180         return __nr_to_section(pfn_to_section_nr(pfn));
1181 }
1182 
1183 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1184 static inline int pfn_valid(unsigned long pfn)
1185 {
1186         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1187                 return 0;
1188         return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1189 }
1190 #endif
1191 
1192 static inline int pfn_present(unsigned long pfn)
1193 {
1194         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1195                 return 0;
1196         return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1197 }
1198 
1199 /*
1200  * These are _only_ used during initialisation, therefore they
1201  * can use __initdata ...  They could have names to indicate
1202  * this restriction.
1203  */
1204 #ifdef CONFIG_NUMA
1205 #define pfn_to_nid(pfn)                                                 \
1206 ({                                                                      \
1207         unsigned long __pfn_to_nid_pfn = (pfn);                         \
1208         page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
1209 })
1210 #else
1211 #define pfn_to_nid(pfn)         (0)
1212 #endif
1213 
1214 #define early_pfn_valid(pfn)    pfn_valid(pfn)
1215 void sparse_init(void);
1216 #else
1217 #define sparse_init()   do {} while (0)
1218 #define sparse_index_init(_sec, _nid)  do {} while (0)
1219 #endif /* CONFIG_SPARSEMEM */
1220 
1221 /*
1222  * During memory init memblocks map pfns to nids. The search is expensive and
1223  * this caches recent lookups. The implementation of __early_pfn_to_nid
1224  * may treat start/end as pfns or sections.
1225  */
1226 struct mminit_pfnnid_cache {
1227         unsigned long last_start;
1228         unsigned long last_end;
1229         int last_nid;
1230 };
1231 
1232 #ifndef early_pfn_valid
1233 #define early_pfn_valid(pfn)    (1)
1234 #endif
1235 
1236 void memory_present(int nid, unsigned long start, unsigned long end);
1237 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1238 
1239 /*
1240  * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1241  * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1242  * pfn_valid_within() should be used in this case; we optimise this away
1243  * when we have no holes within a MAX_ORDER_NR_PAGES block.
1244  */
1245 #ifdef CONFIG_HOLES_IN_ZONE
1246 #define pfn_valid_within(pfn) pfn_valid(pfn)
1247 #else
1248 #define pfn_valid_within(pfn) (1)
1249 #endif
1250 
1251 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1252 /*
1253  * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1254  * associated with it or not. In FLATMEM, it is expected that holes always
1255  * have valid memmap as long as there is valid PFNs either side of the hole.
1256  * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1257  * entire section.
1258  *
1259  * However, an ARM, and maybe other embedded architectures in the future
1260  * free memmap backing holes to save memory on the assumption the memmap is
1261  * never used. The page_zone linkages are then broken even though pfn_valid()
1262  * returns true. A walker of the full memmap must then do this additional
1263  * check to ensure the memmap they are looking at is sane by making sure
1264  * the zone and PFN linkages are still valid. This is expensive, but walkers
1265  * of the full memmap are extremely rare.
1266  */
1267 bool memmap_valid_within(unsigned long pfn,
1268                                         struct page *page, struct zone *zone);
1269 #else
1270 static inline bool memmap_valid_within(unsigned long pfn,
1271                                         struct page *page, struct zone *zone)
1272 {
1273         return true;
1274 }
1275 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1276 
1277 #endif /* !__GENERATING_BOUNDS.H */
1278 #endif /* !__ASSEMBLY__ */
1279 #endif /* _LINUX_MMZONE_H */
1280 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp