Version:  2.0.40 2.2.26 2.4.37 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17

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_RECLAIMABLE,
 41         MIGRATE_MOVABLE,
 42         MIGRATE_PCPTYPES,       /* the number of types on the pcp lists */
 43         MIGRATE_RESERVE = 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 #ifdef CONFIG_CMA
 67 #  define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
 68 #else
 69 #  define is_migrate_cma(migratetype) false
 70 #endif
 71 
 72 #define for_each_migratetype_order(order, type) \
 73         for (order = 0; order < MAX_ORDER; order++) \
 74                 for (type = 0; type < MIGRATE_TYPES; type++)
 75 
 76 extern int page_group_by_mobility_disabled;
 77 
 78 #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
 79 #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
 80 
 81 #define get_pageblock_migratetype(page)                                 \
 82         get_pfnblock_flags_mask(page, page_to_pfn(page),                \
 83                         PB_migrate_end, MIGRATETYPE_MASK)
 84 
 85 static inline int get_pfnblock_migratetype(struct page *page, unsigned long pfn)
 86 {
 87         BUILD_BUG_ON(PB_migrate_end - PB_migrate != 2);
 88         return get_pfnblock_flags_mask(page, pfn, PB_migrate_end,
 89                                         MIGRATETYPE_MASK);
 90 }
 91 
 92 struct free_area {
 93         struct list_head        free_list[MIGRATE_TYPES];
 94         unsigned long           nr_free;
 95 };
 96 
 97 struct pglist_data;
 98 
 99 /*
100  * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
101  * So add a wild amount of padding here to ensure that they fall into separate
102  * cachelines.  There are very few zone structures in the machine, so space
103  * consumption is not a concern here.
104  */
105 #if defined(CONFIG_SMP)
106 struct zone_padding {
107         char x[0];
108 } ____cacheline_internodealigned_in_smp;
109 #define ZONE_PADDING(name)      struct zone_padding name;
110 #else
111 #define ZONE_PADDING(name)
112 #endif
113 
114 enum zone_stat_item {
115         /* First 128 byte cacheline (assuming 64 bit words) */
116         NR_FREE_PAGES,
117         NR_ALLOC_BATCH,
118         NR_LRU_BASE,
119         NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
120         NR_ACTIVE_ANON,         /*  "     "     "   "       "         */
121         NR_INACTIVE_FILE,       /*  "     "     "   "       "         */
122         NR_ACTIVE_FILE,         /*  "     "     "   "       "         */
123         NR_UNEVICTABLE,         /*  "     "     "   "       "         */
124         NR_MLOCK,               /* mlock()ed pages found and moved off LRU */
125         NR_ANON_PAGES,  /* Mapped anonymous pages */
126         NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
127                            only modified from process context */
128         NR_FILE_PAGES,
129         NR_FILE_DIRTY,
130         NR_WRITEBACK,
131         NR_SLAB_RECLAIMABLE,
132         NR_SLAB_UNRECLAIMABLE,
133         NR_PAGETABLE,           /* used for pagetables */
134         NR_KERNEL_STACK,
135         /* Second 128 byte cacheline */
136         NR_UNSTABLE_NFS,        /* NFS unstable pages */
137         NR_BOUNCE,
138         NR_VMSCAN_WRITE,
139         NR_VMSCAN_IMMEDIATE,    /* Prioritise for reclaim when writeback ends */
140         NR_WRITEBACK_TEMP,      /* Writeback using temporary buffers */
141         NR_ISOLATED_ANON,       /* Temporary isolated pages from anon lru */
142         NR_ISOLATED_FILE,       /* Temporary isolated pages from file lru */
143         NR_SHMEM,               /* shmem pages (included tmpfs/GEM pages) */
144         NR_DIRTIED,             /* page dirtyings since bootup */
145         NR_WRITTEN,             /* page writings since bootup */
146         NR_PAGES_SCANNED,       /* pages scanned since last reclaim */
147 #ifdef CONFIG_NUMA
148         NUMA_HIT,               /* allocated in intended node */
149         NUMA_MISS,              /* allocated in non intended node */
150         NUMA_FOREIGN,           /* was intended here, hit elsewhere */
151         NUMA_INTERLEAVE_HIT,    /* interleaver preferred this zone */
152         NUMA_LOCAL,             /* allocation from local node */
153         NUMA_OTHER,             /* allocation from other node */
154 #endif
155         WORKINGSET_REFAULT,
156         WORKINGSET_ACTIVATE,
157         WORKINGSET_NODERECLAIM,
158         NR_ANON_TRANSPARENT_HUGEPAGES,
159         NR_FREE_CMA_PAGES,
160         NR_VM_ZONE_STAT_ITEMS };
161 
162 /*
163  * We do arithmetic on the LRU lists in various places in the code,
164  * so it is important to keep the active lists LRU_ACTIVE higher in
165  * the array than the corresponding inactive lists, and to keep
166  * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
167  *
168  * This has to be kept in sync with the statistics in zone_stat_item
169  * above and the descriptions in vmstat_text in mm/vmstat.c
170  */
171 #define LRU_BASE 0
172 #define LRU_ACTIVE 1
173 #define LRU_FILE 2
174 
175 enum lru_list {
176         LRU_INACTIVE_ANON = LRU_BASE,
177         LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
178         LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
179         LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
180         LRU_UNEVICTABLE,
181         NR_LRU_LISTS
182 };
183 
184 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
185 
186 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
187 
188 static inline int is_file_lru(enum lru_list lru)
189 {
190         return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
191 }
192 
193 static inline int is_active_lru(enum lru_list lru)
194 {
195         return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
196 }
197 
198 static inline int is_unevictable_lru(enum lru_list lru)
199 {
200         return (lru == LRU_UNEVICTABLE);
201 }
202 
203 struct zone_reclaim_stat {
204         /*
205          * The pageout code in vmscan.c keeps track of how many of the
206          * mem/swap backed and file backed pages are referenced.
207          * The higher the rotated/scanned ratio, the more valuable
208          * that cache is.
209          *
210          * The anon LRU stats live in [0], file LRU stats in [1]
211          */
212         unsigned long           recent_rotated[2];
213         unsigned long           recent_scanned[2];
214 };
215 
216 struct lruvec {
217         struct list_head lists[NR_LRU_LISTS];
218         struct zone_reclaim_stat reclaim_stat;
219 #ifdef CONFIG_MEMCG
220         struct zone *zone;
221 #endif
222 };
223 
224 /* Mask used at gathering information at once (see memcontrol.c) */
225 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
226 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
227 #define LRU_ALL      ((1 << NR_LRU_LISTS) - 1)
228 
229 /* Isolate clean file */
230 #define ISOLATE_CLEAN           ((__force isolate_mode_t)0x1)
231 /* Isolate unmapped file */
232 #define ISOLATE_UNMAPPED        ((__force isolate_mode_t)0x2)
233 /* Isolate for asynchronous migration */
234 #define ISOLATE_ASYNC_MIGRATE   ((__force isolate_mode_t)0x4)
235 /* Isolate unevictable pages */
236 #define ISOLATE_UNEVICTABLE     ((__force isolate_mode_t)0x8)
237 
238 /* LRU Isolation modes. */
239 typedef unsigned __bitwise__ isolate_mode_t;
240 
241 enum zone_watermarks {
242         WMARK_MIN,
243         WMARK_LOW,
244         WMARK_HIGH,
245         NR_WMARK
246 };
247 
248 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
249 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
250 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
251 
252 struct per_cpu_pages {
253         int count;              /* number of pages in the list */
254         int high;               /* high watermark, emptying needed */
255         int batch;              /* chunk size for buddy add/remove */
256 
257         /* Lists of pages, one per migrate type stored on the pcp-lists */
258         struct list_head lists[MIGRATE_PCPTYPES];
259 };
260 
261 struct per_cpu_pageset {
262         struct per_cpu_pages pcp;
263 #ifdef CONFIG_NUMA
264         s8 expire;
265 #endif
266 #ifdef CONFIG_SMP
267         s8 stat_threshold;
268         s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
269 #endif
270 };
271 
272 #endif /* !__GENERATING_BOUNDS.H */
273 
274 enum zone_type {
275 #ifdef CONFIG_ZONE_DMA
276         /*
277          * ZONE_DMA is used when there are devices that are not able
278          * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
279          * carve out the portion of memory that is needed for these devices.
280          * The range is arch specific.
281          *
282          * Some examples
283          *
284          * Architecture         Limit
285          * ---------------------------
286          * parisc, ia64, sparc  <4G
287          * s390                 <2G
288          * arm                  Various
289          * alpha                Unlimited or 0-16MB.
290          *
291          * i386, x86_64 and multiple other arches
292          *                      <16M.
293          */
294         ZONE_DMA,
295 #endif
296 #ifdef CONFIG_ZONE_DMA32
297         /*
298          * x86_64 needs two ZONE_DMAs because it supports devices that are
299          * only able to do DMA to the lower 16M but also 32 bit devices that
300          * can only do DMA areas below 4G.
301          */
302         ZONE_DMA32,
303 #endif
304         /*
305          * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
306          * performed on pages in ZONE_NORMAL if the DMA devices support
307          * transfers to all addressable memory.
308          */
309         ZONE_NORMAL,
310 #ifdef CONFIG_HIGHMEM
311         /*
312          * A memory area that is only addressable by the kernel through
313          * mapping portions into its own address space. This is for example
314          * used by i386 to allow the kernel to address the memory beyond
315          * 900MB. The kernel will set up special mappings (page
316          * table entries on i386) for each page that the kernel needs to
317          * access.
318          */
319         ZONE_HIGHMEM,
320 #endif
321         ZONE_MOVABLE,
322         __MAX_NR_ZONES
323 };
324 
325 #ifndef __GENERATING_BOUNDS_H
326 
327 struct zone {
328         /* Read-mostly fields */
329 
330         /* zone watermarks, access with *_wmark_pages(zone) macros */
331         unsigned long watermark[NR_WMARK];
332 
333         /*
334          * We don't know if the memory that we're going to allocate will be freeable
335          * or/and it will be released eventually, so to avoid totally wasting several
336          * GB of ram we must reserve some of the lower zone memory (otherwise we risk
337          * to run OOM on the lower zones despite there's tons of freeable ram
338          * on the higher zones). This array is recalculated at runtime if the
339          * sysctl_lowmem_reserve_ratio sysctl 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 should not be
358          * considered dirtyable memory.
359          */
360         unsigned long           dirty_balance_reserve;
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         /*
429          * Number of MIGRATE_RESEVE page block. To maintain for just
430          * optimization. Protected by zone->lock.
431          */
432         int                     nr_migrate_reserve_block;
433 
434 #ifdef CONFIG_MEMORY_HOTPLUG
435         /* see spanned/present_pages for more description */
436         seqlock_t               span_seqlock;
437 #endif
438 
439         /*
440          * wait_table           -- the array holding the hash table
441          * wait_table_hash_nr_entries   -- the size of the hash table array
442          * wait_table_bits      -- wait_table_size == (1 << wait_table_bits)
443          *
444          * The purpose of all these is to keep track of the people
445          * waiting for a page to become available and make them
446          * runnable again when possible. The trouble is that this
447          * consumes a lot of space, especially when so few things
448          * wait on pages at a given time. So instead of using
449          * per-page waitqueues, we use a waitqueue hash table.
450          *
451          * The bucket discipline is to sleep on the same queue when
452          * colliding and wake all in that wait queue when removing.
453          * When something wakes, it must check to be sure its page is
454          * truly available, a la thundering herd. The cost of a
455          * collision is great, but given the expected load of the
456          * table, they should be so rare as to be outweighed by the
457          * benefits from the saved space.
458          *
459          * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
460          * primary users of these fields, and in mm/page_alloc.c
461          * free_area_init_core() performs the initialization of them.
462          */
463         wait_queue_head_t       *wait_table;
464         unsigned long           wait_table_hash_nr_entries;
465         unsigned long           wait_table_bits;
466 
467         ZONE_PADDING(_pad1_)
468 
469         /* Write-intensive fields used from the page allocator */
470         spinlock_t              lock;
471 
472         /* free areas of different sizes */
473         struct free_area        free_area[MAX_ORDER];
474 
475         /* zone flags, see below */
476         unsigned long           flags;
477 
478         ZONE_PADDING(_pad2_)
479 
480         /* Write-intensive fields used by page reclaim */
481 
482         /* Fields commonly accessed by the page reclaim scanner */
483         spinlock_t              lru_lock;
484         struct lruvec           lruvec;
485 
486         /* Evictions & activations on the inactive file list */
487         atomic_long_t           inactive_age;
488 
489         /*
490          * When free pages are below this point, additional steps are taken
491          * when reading the number of free pages to avoid per-cpu counter
492          * drift allowing watermarks to be breached
493          */
494         unsigned long percpu_drift_mark;
495 
496 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
497         /* pfn where compaction free scanner should start */
498         unsigned long           compact_cached_free_pfn;
499         /* pfn where async and sync compaction migration scanner should start */
500         unsigned long           compact_cached_migrate_pfn[2];
501 #endif
502 
503 #ifdef CONFIG_COMPACTION
504         /*
505          * On compaction failure, 1<<compact_defer_shift compactions
506          * are skipped before trying again. The number attempted since
507          * last failure is tracked with compact_considered.
508          */
509         unsigned int            compact_considered;
510         unsigned int            compact_defer_shift;
511         int                     compact_order_failed;
512 #endif
513 
514 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
515         /* Set to true when the PG_migrate_skip bits should be cleared */
516         bool                    compact_blockskip_flush;
517 #endif
518 
519         ZONE_PADDING(_pad3_)
520         /* Zone statistics */
521         atomic_long_t           vm_stat[NR_VM_ZONE_STAT_ITEMS];
522 } ____cacheline_internodealigned_in_smp;
523 
524 typedef enum {
525         ZONE_RECLAIM_LOCKED,            /* prevents concurrent reclaim */
526         ZONE_OOM_LOCKED,                /* zone is in OOM killer zonelist */
527         ZONE_CONGESTED,                 /* zone has many dirty pages backed by
528                                          * a congested BDI
529                                          */
530         ZONE_TAIL_LRU_DIRTY,            /* reclaim scanning has recently found
531                                          * many dirty file pages at the tail
532                                          * of the LRU.
533                                          */
534         ZONE_WRITEBACK,                 /* reclaim scanning has recently found
535                                          * many pages under writeback
536                                          */
537         ZONE_FAIR_DEPLETED,             /* fair zone policy batch depleted */
538 } zone_flags_t;
539 
540 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
541 {
542         set_bit(flag, &zone->flags);
543 }
544 
545 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
546 {
547         return test_and_set_bit(flag, &zone->flags);
548 }
549 
550 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
551 {
552         clear_bit(flag, &zone->flags);
553 }
554 
555 static inline int zone_is_reclaim_congested(const struct zone *zone)
556 {
557         return test_bit(ZONE_CONGESTED, &zone->flags);
558 }
559 
560 static inline int zone_is_reclaim_dirty(const struct zone *zone)
561 {
562         return test_bit(ZONE_TAIL_LRU_DIRTY, &zone->flags);
563 }
564 
565 static inline int zone_is_reclaim_writeback(const struct zone *zone)
566 {
567         return test_bit(ZONE_WRITEBACK, &zone->flags);
568 }
569 
570 static inline int zone_is_reclaim_locked(const struct zone *zone)
571 {
572         return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
573 }
574 
575 static inline int zone_is_fair_depleted(const struct zone *zone)
576 {
577         return test_bit(ZONE_FAIR_DEPLETED, &zone->flags);
578 }
579 
580 static inline int zone_is_oom_locked(const struct zone *zone)
581 {
582         return test_bit(ZONE_OOM_LOCKED, &zone->flags);
583 }
584 
585 static inline unsigned long zone_end_pfn(const struct zone *zone)
586 {
587         return zone->zone_start_pfn + zone->spanned_pages;
588 }
589 
590 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
591 {
592         return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
593 }
594 
595 static inline bool zone_is_initialized(struct zone *zone)
596 {
597         return !!zone->wait_table;
598 }
599 
600 static inline bool zone_is_empty(struct zone *zone)
601 {
602         return zone->spanned_pages == 0;
603 }
604 
605 /*
606  * The "priority" of VM scanning is how much of the queues we will scan in one
607  * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
608  * queues ("queue_length >> 12") during an aging round.
609  */
610 #define DEF_PRIORITY 12
611 
612 /* Maximum number of zones on a zonelist */
613 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
614 
615 #ifdef CONFIG_NUMA
616 
617 /*
618  * The NUMA zonelists are doubled because we need zonelists that restrict the
619  * allocations to a single node for __GFP_THISNODE.
620  *
621  * [0]  : Zonelist with fallback
622  * [1]  : No fallback (__GFP_THISNODE)
623  */
624 #define MAX_ZONELISTS 2
625 
626 
627 /*
628  * We cache key information from each zonelist for smaller cache
629  * footprint when scanning for free pages in get_page_from_freelist().
630  *
631  * 1) The BITMAP fullzones tracks which zones in a zonelist have come
632  *    up short of free memory since the last time (last_fullzone_zap)
633  *    we zero'd fullzones.
634  * 2) The array z_to_n[] maps each zone in the zonelist to its node
635  *    id, so that we can efficiently evaluate whether that node is
636  *    set in the current tasks mems_allowed.
637  *
638  * Both fullzones and z_to_n[] are one-to-one with the zonelist,
639  * indexed by a zones offset in the zonelist zones[] array.
640  *
641  * The get_page_from_freelist() routine does two scans.  During the
642  * first scan, we skip zones whose corresponding bit in 'fullzones'
643  * is set or whose corresponding node in current->mems_allowed (which
644  * comes from cpusets) is not set.  During the second scan, we bypass
645  * this zonelist_cache, to ensure we look methodically at each zone.
646  *
647  * Once per second, we zero out (zap) fullzones, forcing us to
648  * reconsider nodes that might have regained more free memory.
649  * The field last_full_zap is the time we last zapped fullzones.
650  *
651  * This mechanism reduces the amount of time we waste repeatedly
652  * reexaming zones for free memory when they just came up low on
653  * memory momentarilly ago.
654  *
655  * The zonelist_cache struct members logically belong in struct
656  * zonelist.  However, the mempolicy zonelists constructed for
657  * MPOL_BIND are intentionally variable length (and usually much
658  * shorter).  A general purpose mechanism for handling structs with
659  * multiple variable length members is more mechanism than we want
660  * here.  We resort to some special case hackery instead.
661  *
662  * The MPOL_BIND zonelists don't need this zonelist_cache (in good
663  * part because they are shorter), so we put the fixed length stuff
664  * at the front of the zonelist struct, ending in a variable length
665  * zones[], as is needed by MPOL_BIND.
666  *
667  * Then we put the optional zonelist cache on the end of the zonelist
668  * struct.  This optional stuff is found by a 'zlcache_ptr' pointer in
669  * the fixed length portion at the front of the struct.  This pointer
670  * both enables us to find the zonelist cache, and in the case of
671  * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
672  * to know that the zonelist cache is not there.
673  *
674  * The end result is that struct zonelists come in two flavors:
675  *  1) The full, fixed length version, shown below, and
676  *  2) The custom zonelists for MPOL_BIND.
677  * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
678  *
679  * Even though there may be multiple CPU cores on a node modifying
680  * fullzones or last_full_zap in the same zonelist_cache at the same
681  * time, we don't lock it.  This is just hint data - if it is wrong now
682  * and then, the allocator will still function, perhaps a bit slower.
683  */
684 
685 
686 struct zonelist_cache {
687         unsigned short z_to_n[MAX_ZONES_PER_ZONELIST];          /* zone->nid */
688         DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST);      /* zone full? */
689         unsigned long last_full_zap;            /* when last zap'd (jiffies) */
690 };
691 #else
692 #define MAX_ZONELISTS 1
693 struct zonelist_cache;
694 #endif
695 
696 /*
697  * This struct contains information about a zone in a zonelist. It is stored
698  * here to avoid dereferences into large structures and lookups of tables
699  */
700 struct zoneref {
701         struct zone *zone;      /* Pointer to actual zone */
702         int zone_idx;           /* zone_idx(zoneref->zone) */
703 };
704 
705 /*
706  * One allocation request operates on a zonelist. A zonelist
707  * is a list of zones, the first one is the 'goal' of the
708  * allocation, the other zones are fallback zones, in decreasing
709  * priority.
710  *
711  * If zlcache_ptr is not NULL, then it is just the address of zlcache,
712  * as explained above.  If zlcache_ptr is NULL, there is no zlcache.
713  * *
714  * To speed the reading of the zonelist, the zonerefs contain the zone index
715  * of the entry being read. Helper functions to access information given
716  * a struct zoneref are
717  *
718  * zonelist_zone()      - Return the struct zone * for an entry in _zonerefs
719  * zonelist_zone_idx()  - Return the index of the zone for an entry
720  * zonelist_node_idx()  - Return the index of the node for an entry
721  */
722 struct zonelist {
723         struct zonelist_cache *zlcache_ptr;                  // NULL or &zlcache
724         struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
725 #ifdef CONFIG_NUMA
726         struct zonelist_cache zlcache;                       // optional ...
727 #endif
728 };
729 
730 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
731 struct node_active_region {
732         unsigned long start_pfn;
733         unsigned long end_pfn;
734         int nid;
735 };
736 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
737 
738 #ifndef CONFIG_DISCONTIGMEM
739 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
740 extern struct page *mem_map;
741 #endif
742 
743 /*
744  * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
745  * (mostly NUMA machines?) to denote a higher-level memory zone than the
746  * zone denotes.
747  *
748  * On NUMA machines, each NUMA node would have a pg_data_t to describe
749  * it's memory layout.
750  *
751  * Memory statistics and page replacement data structures are maintained on a
752  * per-zone basis.
753  */
754 struct bootmem_data;
755 typedef struct pglist_data {
756         struct zone node_zones[MAX_NR_ZONES];
757         struct zonelist node_zonelists[MAX_ZONELISTS];
758         int nr_zones;
759 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
760         struct page *node_mem_map;
761 #ifdef CONFIG_MEMCG
762         struct page_cgroup *node_page_cgroup;
763 #endif
764 #endif
765 #ifndef CONFIG_NO_BOOTMEM
766         struct bootmem_data *bdata;
767 #endif
768 #ifdef CONFIG_MEMORY_HOTPLUG
769         /*
770          * Must be held any time you expect node_start_pfn, node_present_pages
771          * or node_spanned_pages stay constant.  Holding this will also
772          * guarantee that any pfn_valid() stays that way.
773          *
774          * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
775          * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
776          *
777          * Nests above zone->lock and zone->span_seqlock
778          */
779         spinlock_t node_size_lock;
780 #endif
781         unsigned long node_start_pfn;
782         unsigned long node_present_pages; /* total number of physical pages */
783         unsigned long node_spanned_pages; /* total size of physical page
784                                              range, including holes */
785         int node_id;
786         wait_queue_head_t kswapd_wait;
787         wait_queue_head_t pfmemalloc_wait;
788         struct task_struct *kswapd;     /* Protected by
789                                            mem_hotplug_begin/end() */
790         int kswapd_max_order;
791         enum zone_type classzone_idx;
792 #ifdef CONFIG_NUMA_BALANCING
793         /* Lock serializing the migrate rate limiting window */
794         spinlock_t numabalancing_migrate_lock;
795 
796         /* Rate limiting time interval */
797         unsigned long numabalancing_migrate_next_window;
798 
799         /* Number of pages migrated during the rate limiting time interval */
800         unsigned long numabalancing_migrate_nr_pages;
801 #endif
802 } pg_data_t;
803 
804 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
805 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
806 #ifdef CONFIG_FLAT_NODE_MEM_MAP
807 #define pgdat_page_nr(pgdat, pagenr)    ((pgdat)->node_mem_map + (pagenr))
808 #else
809 #define pgdat_page_nr(pgdat, pagenr)    pfn_to_page((pgdat)->node_start_pfn + (pagenr))
810 #endif
811 #define nid_page_nr(nid, pagenr)        pgdat_page_nr(NODE_DATA(nid),(pagenr))
812 
813 #define node_start_pfn(nid)     (NODE_DATA(nid)->node_start_pfn)
814 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
815 
816 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
817 {
818         return pgdat->node_start_pfn + pgdat->node_spanned_pages;
819 }
820 
821 static inline bool pgdat_is_empty(pg_data_t *pgdat)
822 {
823         return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
824 }
825 
826 #include <linux/memory_hotplug.h>
827 
828 extern struct mutex zonelists_mutex;
829 void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
830 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
831 bool zone_watermark_ok(struct zone *z, unsigned int order,
832                 unsigned long mark, int classzone_idx, int alloc_flags);
833 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
834                 unsigned long mark, int classzone_idx, int alloc_flags);
835 enum memmap_context {
836         MEMMAP_EARLY,
837         MEMMAP_HOTPLUG,
838 };
839 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
840                                      unsigned long size,
841                                      enum memmap_context context);
842 
843 extern void lruvec_init(struct lruvec *lruvec);
844 
845 static inline struct zone *lruvec_zone(struct lruvec *lruvec)
846 {
847 #ifdef CONFIG_MEMCG
848         return lruvec->zone;
849 #else
850         return container_of(lruvec, struct zone, lruvec);
851 #endif
852 }
853 
854 #ifdef CONFIG_HAVE_MEMORY_PRESENT
855 void memory_present(int nid, unsigned long start, unsigned long end);
856 #else
857 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
858 #endif
859 
860 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
861 int local_memory_node(int node_id);
862 #else
863 static inline int local_memory_node(int node_id) { return node_id; };
864 #endif
865 
866 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
867 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
868 #endif
869 
870 /*
871  * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
872  */
873 #define zone_idx(zone)          ((zone) - (zone)->zone_pgdat->node_zones)
874 
875 static inline int populated_zone(struct zone *zone)
876 {
877         return (!!zone->present_pages);
878 }
879 
880 extern int movable_zone;
881 
882 static inline int zone_movable_is_highmem(void)
883 {
884 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
885         return movable_zone == ZONE_HIGHMEM;
886 #elif defined(CONFIG_HIGHMEM)
887         return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
888 #else
889         return 0;
890 #endif
891 }
892 
893 static inline int is_highmem_idx(enum zone_type idx)
894 {
895 #ifdef CONFIG_HIGHMEM
896         return (idx == ZONE_HIGHMEM ||
897                 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
898 #else
899         return 0;
900 #endif
901 }
902 
903 /**
904  * is_highmem - helper function to quickly check if a struct zone is a 
905  *              highmem zone or not.  This is an attempt to keep references
906  *              to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
907  * @zone - pointer to struct zone variable
908  */
909 static inline int is_highmem(struct zone *zone)
910 {
911 #ifdef CONFIG_HIGHMEM
912         int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
913         return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
914                (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
915                 zone_movable_is_highmem());
916 #else
917         return 0;
918 #endif
919 }
920 
921 /* These two functions are used to setup the per zone pages min values */
922 struct ctl_table;
923 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
924                                         void __user *, size_t *, loff_t *);
925 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
926 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
927                                         void __user *, size_t *, loff_t *);
928 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
929                                         void __user *, size_t *, loff_t *);
930 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
931                         void __user *, size_t *, loff_t *);
932 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
933                         void __user *, size_t *, loff_t *);
934 
935 extern int numa_zonelist_order_handler(struct ctl_table *, int,
936                         void __user *, size_t *, loff_t *);
937 extern char numa_zonelist_order[];
938 #define NUMA_ZONELIST_ORDER_LEN 16      /* string buffer size */
939 
940 #ifndef CONFIG_NEED_MULTIPLE_NODES
941 
942 extern struct pglist_data contig_page_data;
943 #define NODE_DATA(nid)          (&contig_page_data)
944 #define NODE_MEM_MAP(nid)       mem_map
945 
946 #else /* CONFIG_NEED_MULTIPLE_NODES */
947 
948 #include <asm/mmzone.h>
949 
950 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
951 
952 extern struct pglist_data *first_online_pgdat(void);
953 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
954 extern struct zone *next_zone(struct zone *zone);
955 
956 /**
957  * for_each_online_pgdat - helper macro to iterate over all online nodes
958  * @pgdat - pointer to a pg_data_t variable
959  */
960 #define for_each_online_pgdat(pgdat)                    \
961         for (pgdat = first_online_pgdat();              \
962              pgdat;                                     \
963              pgdat = next_online_pgdat(pgdat))
964 /**
965  * for_each_zone - helper macro to iterate over all memory zones
966  * @zone - pointer to struct zone variable
967  *
968  * The user only needs to declare the zone variable, for_each_zone
969  * fills it in.
970  */
971 #define for_each_zone(zone)                             \
972         for (zone = (first_online_pgdat())->node_zones; \
973              zone;                                      \
974              zone = next_zone(zone))
975 
976 #define for_each_populated_zone(zone)                   \
977         for (zone = (first_online_pgdat())->node_zones; \
978              zone;                                      \
979              zone = next_zone(zone))                    \
980                 if (!populated_zone(zone))              \
981                         ; /* do nothing */              \
982                 else
983 
984 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
985 {
986         return zoneref->zone;
987 }
988 
989 static inline int zonelist_zone_idx(struct zoneref *zoneref)
990 {
991         return zoneref->zone_idx;
992 }
993 
994 static inline int zonelist_node_idx(struct zoneref *zoneref)
995 {
996 #ifdef CONFIG_NUMA
997         /* zone_to_nid not available in this context */
998         return zoneref->zone->node;
999 #else
1000         return 0;
1001 #endif /* CONFIG_NUMA */
1002 }
1003 
1004 /**
1005  * 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
1006  * @z - The cursor used as a starting point for the search
1007  * @highest_zoneidx - The zone index of the highest zone to return
1008  * @nodes - An optional nodemask to filter the zonelist with
1009  * @zone - The first suitable zone found is returned via this parameter
1010  *
1011  * This function returns the next zone at or below a given zone index that is
1012  * within the allowed nodemask using a cursor as the starting point for the
1013  * search. The zoneref returned is a cursor that represents the current zone
1014  * being examined. It should be advanced by one before calling
1015  * next_zones_zonelist again.
1016  */
1017 struct zoneref *next_zones_zonelist(struct zoneref *z,
1018                                         enum zone_type highest_zoneidx,
1019                                         nodemask_t *nodes,
1020                                         struct zone **zone);
1021 
1022 /**
1023  * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
1024  * @zonelist - The zonelist to search for a suitable zone
1025  * @highest_zoneidx - The zone index of the highest zone to return
1026  * @nodes - An optional nodemask to filter the zonelist with
1027  * @zone - The first suitable zone found is returned via this parameter
1028  *
1029  * This function returns the first zone at or below a given zone index that is
1030  * within the allowed nodemask. The zoneref returned is a cursor that can be
1031  * used to iterate the zonelist with next_zones_zonelist by advancing it by
1032  * one before calling.
1033  */
1034 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
1035                                         enum zone_type highest_zoneidx,
1036                                         nodemask_t *nodes,
1037                                         struct zone **zone)
1038 {
1039         return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
1040                                                                 zone);
1041 }
1042 
1043 /**
1044  * 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
1045  * @zone - The current zone in the iterator
1046  * @z - The current pointer within zonelist->zones being iterated
1047  * @zlist - The zonelist being iterated
1048  * @highidx - The zone index of the highest zone to return
1049  * @nodemask - Nodemask allowed by the allocator
1050  *
1051  * This iterator iterates though all zones at or below a given zone index and
1052  * within a given nodemask
1053  */
1054 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1055         for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
1056                 zone;                                                   \
1057                 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
1058 
1059 /**
1060  * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1061  * @zone - The current zone in the iterator
1062  * @z - The current pointer within zonelist->zones being iterated
1063  * @zlist - The zonelist being iterated
1064  * @highidx - The zone index of the highest zone to return
1065  *
1066  * This iterator iterates though all zones at or below a given zone index.
1067  */
1068 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
1069         for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1070 
1071 #ifdef CONFIG_SPARSEMEM
1072 #include <asm/sparsemem.h>
1073 #endif
1074 
1075 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1076         !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1077 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1078 {
1079         return 0;
1080 }
1081 #endif
1082 
1083 #ifdef CONFIG_FLATMEM
1084 #define pfn_to_nid(pfn)         (0)
1085 #endif
1086 
1087 #ifdef CONFIG_SPARSEMEM
1088 
1089 /*
1090  * SECTION_SHIFT                #bits space required to store a section #
1091  *
1092  * PA_SECTION_SHIFT             physical address to/from section number
1093  * PFN_SECTION_SHIFT            pfn to/from section number
1094  */
1095 #define PA_SECTION_SHIFT        (SECTION_SIZE_BITS)
1096 #define PFN_SECTION_SHIFT       (SECTION_SIZE_BITS - PAGE_SHIFT)
1097 
1098 #define NR_MEM_SECTIONS         (1UL << SECTIONS_SHIFT)
1099 
1100 #define PAGES_PER_SECTION       (1UL << PFN_SECTION_SHIFT)
1101 #define PAGE_SECTION_MASK       (~(PAGES_PER_SECTION-1))
1102 
1103 #define SECTION_BLOCKFLAGS_BITS \
1104         ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1105 
1106 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1107 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1108 #endif
1109 
1110 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1111 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1112 
1113 #define SECTION_ALIGN_UP(pfn)   (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1114 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1115 
1116 struct page;
1117 struct page_cgroup;
1118 struct mem_section {
1119         /*
1120          * This is, logically, a pointer to an array of struct
1121          * pages.  However, it is stored with some other magic.
1122          * (see sparse.c::sparse_init_one_section())
1123          *
1124          * Additionally during early boot we encode node id of
1125          * the location of the section here to guide allocation.
1126          * (see sparse.c::memory_present())
1127          *
1128          * Making it a UL at least makes someone do a cast
1129          * before using it wrong.
1130          */
1131         unsigned long section_mem_map;
1132 
1133         /* See declaration of similar field in struct zone */
1134         unsigned long *pageblock_flags;
1135 #ifdef CONFIG_MEMCG
1136         /*
1137          * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1138          * section. (see memcontrol.h/page_cgroup.h about this.)
1139          */
1140         struct page_cgroup *page_cgroup;
1141         unsigned long pad;
1142 #endif
1143         /*
1144          * WARNING: mem_section must be a power-of-2 in size for the
1145          * calculation and use of SECTION_ROOT_MASK to make sense.
1146          */
1147 };
1148 
1149 #ifdef CONFIG_SPARSEMEM_EXTREME
1150 #define SECTIONS_PER_ROOT       (PAGE_SIZE / sizeof (struct mem_section))
1151 #else
1152 #define SECTIONS_PER_ROOT       1
1153 #endif
1154 
1155 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1156 #define NR_SECTION_ROOTS        DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1157 #define SECTION_ROOT_MASK       (SECTIONS_PER_ROOT - 1)
1158 
1159 #ifdef CONFIG_SPARSEMEM_EXTREME
1160 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1161 #else
1162 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1163 #endif
1164 
1165 static inline struct mem_section *__nr_to_section(unsigned long nr)
1166 {
1167         if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1168                 return NULL;
1169         return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1170 }
1171 extern int __section_nr(struct mem_section* ms);
1172 extern unsigned long usemap_size(void);
1173 
1174 /*
1175  * We use the lower bits of the mem_map pointer to store
1176  * a little bit of information.  There should be at least
1177  * 3 bits here due to 32-bit alignment.
1178  */
1179 #define SECTION_MARKED_PRESENT  (1UL<<0)
1180 #define SECTION_HAS_MEM_MAP     (1UL<<1)
1181 #define SECTION_MAP_LAST_BIT    (1UL<<2)
1182 #define SECTION_MAP_MASK        (~(SECTION_MAP_LAST_BIT-1))
1183 #define SECTION_NID_SHIFT       2
1184 
1185 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1186 {
1187         unsigned long map = section->section_mem_map;
1188         map &= SECTION_MAP_MASK;
1189         return (struct page *)map;
1190 }
1191 
1192 static inline int present_section(struct mem_section *section)
1193 {
1194         return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1195 }
1196 
1197 static inline int present_section_nr(unsigned long nr)
1198 {
1199         return present_section(__nr_to_section(nr));
1200 }
1201 
1202 static inline int valid_section(struct mem_section *section)
1203 {
1204         return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1205 }
1206 
1207 static inline int valid_section_nr(unsigned long nr)
1208 {
1209         return valid_section(__nr_to_section(nr));
1210 }
1211 
1212 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1213 {
1214         return __nr_to_section(pfn_to_section_nr(pfn));
1215 }
1216 
1217 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1218 static inline int pfn_valid(unsigned long pfn)
1219 {
1220         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1221                 return 0;
1222         return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1223 }
1224 #endif
1225 
1226 static inline int pfn_present(unsigned long pfn)
1227 {
1228         if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1229                 return 0;
1230         return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1231 }
1232 
1233 /*
1234  * These are _only_ used during initialisation, therefore they
1235  * can use __initdata ...  They could have names to indicate
1236  * this restriction.
1237  */
1238 #ifdef CONFIG_NUMA
1239 #define pfn_to_nid(pfn)                                                 \
1240 ({                                                                      \
1241         unsigned long __pfn_to_nid_pfn = (pfn);                         \
1242         page_to_nid(pfn_to_page(__pfn_to_nid_pfn));                     \
1243 })
1244 #else
1245 #define pfn_to_nid(pfn)         (0)
1246 #endif
1247 
1248 #define early_pfn_valid(pfn)    pfn_valid(pfn)
1249 void sparse_init(void);
1250 #else
1251 #define sparse_init()   do {} while (0)
1252 #define sparse_index_init(_sec, _nid)  do {} while (0)
1253 #endif /* CONFIG_SPARSEMEM */
1254 
1255 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1256 bool early_pfn_in_nid(unsigned long pfn, int nid);
1257 #else
1258 #define early_pfn_in_nid(pfn, nid)      (1)
1259 #endif
1260 
1261 #ifndef early_pfn_valid
1262 #define early_pfn_valid(pfn)    (1)
1263 #endif
1264 
1265 void memory_present(int nid, unsigned long start, unsigned long end);
1266 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1267 
1268 /*
1269  * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1270  * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1271  * pfn_valid_within() should be used in this case; we optimise this away
1272  * when we have no holes within a MAX_ORDER_NR_PAGES block.
1273  */
1274 #ifdef CONFIG_HOLES_IN_ZONE
1275 #define pfn_valid_within(pfn) pfn_valid(pfn)
1276 #else
1277 #define pfn_valid_within(pfn) (1)
1278 #endif
1279 
1280 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1281 /*
1282  * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1283  * associated with it or not. In FLATMEM, it is expected that holes always
1284  * have valid memmap as long as there is valid PFNs either side of the hole.
1285  * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1286  * entire section.
1287  *
1288  * However, an ARM, and maybe other embedded architectures in the future
1289  * free memmap backing holes to save memory on the assumption the memmap is
1290  * never used. The page_zone linkages are then broken even though pfn_valid()
1291  * returns true. A walker of the full memmap must then do this additional
1292  * check to ensure the memmap they are looking at is sane by making sure
1293  * the zone and PFN linkages are still valid. This is expensive, but walkers
1294  * of the full memmap are extremely rare.
1295  */
1296 int memmap_valid_within(unsigned long pfn,
1297                                         struct page *page, struct zone *zone);
1298 #else
1299 static inline int memmap_valid_within(unsigned long pfn,
1300                                         struct page *page, struct zone *zone)
1301 {
1302         return 1;
1303 }
1304 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1305 
1306 #endif /* !__GENERATING_BOUNDS.H */
1307 #endif /* !__ASSEMBLY__ */
1308 #endif /* _LINUX_MMZONE_H */
1309 

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