1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
5 #ifndef __GENERATING_BOUNDS_H
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>
22 /* Free memory management - zoned buddy allocator. */
23 #ifndef CONFIG_FORCE_MAX_ZONEORDER
26 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
28 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
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
36 #define PAGE_ALLOC_COSTLY_ORDER 3
44 * MIGRATE_CMA migration type is designed to mimic the way
45 * ZONE_MOVABLE works. Only movable pages can be allocated
46 * from MIGRATE_CMA pageblocks and page allocator never
47 * implicitly change migration type of MIGRATE_CMA pageblock.
49 * The way to use it is to change migratetype of a range of
50 * pageblocks to MIGRATE_CMA which can be done by
51 * __free_pageblock_cma() function. What is important though
52 * is that a range of pageblocks must be aligned to
53 * MAX_ORDER_NR_PAGES should biggest page be bigger then
58 #ifdef CONFIG_MEMORY_ISOLATION
59 MIGRATE_ISOLATE, /* can't allocate from here */
64 #define MIGRATE_PCPTYPES (MIGRATE_RECLAIMABLE+1)
67 # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
69 # define is_migrate_cma(migratetype) false
72 #define for_each_migratetype_order(order, type) \
73 for (order = 0; order < MAX_ORDER; order++) \
74 for (type = 0; type < MIGRATE_TYPES; type++)
76 extern int page_group_by_mobility_disabled;
78 #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
79 #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
81 #define get_pageblock_migratetype(page) \
82 get_pfnblock_flags_mask(page, page_to_pfn(page), \
83 PB_migrate_end, MIGRATETYPE_MASK)
85 static inline int get_pfnblock_migratetype(struct page *page, unsigned long pfn)
87 BUILD_BUG_ON(PB_migrate_end - PB_migrate != 2);
88 return get_pfnblock_flags_mask(page, pfn, PB_migrate_end,
93 struct list_head free_list[MIGRATE_TYPES];
94 unsigned long nr_free;
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.
105 #if defined(CONFIG_SMP)
106 struct zone_padding {
108 } ____cacheline_internodealigned_in_smp;
109 #define ZONE_PADDING(name) struct zone_padding name;
111 #define ZONE_PADDING(name)
114 enum zone_stat_item {
115 /* First 128 byte cacheline (assuming 64 bit words) */
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 */
132 NR_SLAB_UNRECLAIMABLE,
133 NR_PAGETABLE, /* used for pagetables */
135 /* Second 128 byte cacheline */
136 NR_UNSTABLE_NFS, /* NFS unstable pages */
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 */
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 */
157 WORKINGSET_NODERECLAIM,
158 NR_ANON_TRANSPARENT_HUGEPAGES,
160 NR_VM_ZONE_STAT_ITEMS };
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.
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
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,
184 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
186 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
188 static inline int is_file_lru(enum lru_list lru)
190 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
193 static inline int is_active_lru(enum lru_list lru)
195 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
198 static inline int is_unevictable_lru(enum lru_list lru)
200 return (lru == LRU_UNEVICTABLE);
203 struct zone_reclaim_stat {
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
210 * The anon LRU stats live in [0], file LRU stats in [1]
212 unsigned long recent_rotated[2];
213 unsigned long recent_scanned[2];
217 struct list_head lists[NR_LRU_LISTS];
218 struct zone_reclaim_stat reclaim_stat;
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)
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)
238 /* LRU Isolation modes. */
239 typedef unsigned __bitwise__ isolate_mode_t;
241 enum zone_watermarks {
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])
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 */
257 /* Lists of pages, one per migrate type stored on the pcp-lists */
258 struct list_head lists[MIGRATE_PCPTYPES];
261 struct per_cpu_pageset {
262 struct per_cpu_pages pcp;
268 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
272 #endif /* !__GENERATING_BOUNDS.H */
275 #ifdef CONFIG_ZONE_DMA
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.
285 * ---------------------------
286 * parisc, ia64, sparc <4G
289 * alpha Unlimited or 0-16MB.
291 * i386, x86_64 and multiple other arches
296 #ifdef CONFIG_ZONE_DMA32
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.
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.
310 #ifdef CONFIG_HIGHMEM
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
322 #ifdef CONFIG_ZONE_DEVICE
329 #ifndef __GENERATING_BOUNDS_H
332 /* Read-mostly fields */
334 /* zone watermarks, access with *_wmark_pages(zone) macros */
335 unsigned long watermark[NR_WMARK];
338 * We don't know if the memory that we're going to allocate will be freeable
339 * or/and it will be released eventually, so to avoid totally wasting several
340 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
341 * to run OOM on the lower zones despite there's tons of freeable ram
342 * on the higher zones). This array is recalculated at runtime if the
343 * sysctl_lowmem_reserve_ratio sysctl changes.
345 long lowmem_reserve[MAX_NR_ZONES];
352 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
353 * this zone's LRU. Maintained by the pageout code.
355 unsigned int inactive_ratio;
357 struct pglist_data *zone_pgdat;
358 struct per_cpu_pageset __percpu *pageset;
361 * This is a per-zone reserve of pages that should not be
362 * considered dirtyable memory.
364 unsigned long dirty_balance_reserve;
366 #ifndef CONFIG_SPARSEMEM
368 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
369 * In SPARSEMEM, this map is stored in struct mem_section
371 unsigned long *pageblock_flags;
372 #endif /* CONFIG_SPARSEMEM */
376 * zone reclaim becomes active if more unmapped pages exist.
378 unsigned long min_unmapped_pages;
379 unsigned long min_slab_pages;
380 #endif /* CONFIG_NUMA */
382 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
383 unsigned long zone_start_pfn;
386 * spanned_pages is the total pages spanned by the zone, including
387 * holes, which is calculated as:
388 * spanned_pages = zone_end_pfn - zone_start_pfn;
390 * present_pages is physical pages existing within the zone, which
392 * present_pages = spanned_pages - absent_pages(pages in holes);
394 * managed_pages is present pages managed by the buddy system, which
395 * is calculated as (reserved_pages includes pages allocated by the
396 * bootmem allocator):
397 * managed_pages = present_pages - reserved_pages;
399 * So present_pages may be used by memory hotplug or memory power
400 * management logic to figure out unmanaged pages by checking
401 * (present_pages - managed_pages). And managed_pages should be used
402 * by page allocator and vm scanner to calculate all kinds of watermarks
407 * zone_start_pfn and spanned_pages are protected by span_seqlock.
408 * It is a seqlock because it has to be read outside of zone->lock,
409 * and it is done in the main allocator path. But, it is written
410 * quite infrequently.
412 * The span_seq lock is declared along with zone->lock because it is
413 * frequently read in proximity to zone->lock. It's good to
414 * give them a chance of being in the same cacheline.
416 * Write access to present_pages at runtime should be protected by
417 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
418 * present_pages should get_online_mems() to get a stable value.
420 * Read access to managed_pages should be safe because it's unsigned
421 * long. Write access to zone->managed_pages and totalram_pages are
422 * protected by managed_page_count_lock at runtime. Idealy only
423 * adjust_managed_page_count() should be used instead of directly
424 * touching zone->managed_pages and totalram_pages.
426 unsigned long managed_pages;
427 unsigned long spanned_pages;
428 unsigned long present_pages;
432 #ifdef CONFIG_MEMORY_ISOLATION
434 * Number of isolated pageblock. It is used to solve incorrect
435 * freepage counting problem due to racy retrieving migratetype
436 * of pageblock. Protected by zone->lock.
438 unsigned long nr_isolate_pageblock;
441 #ifdef CONFIG_MEMORY_HOTPLUG
442 /* see spanned/present_pages for more description */
443 seqlock_t span_seqlock;
447 * wait_table -- the array holding the hash table
448 * wait_table_hash_nr_entries -- the size of the hash table array
449 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
451 * The purpose of all these is to keep track of the people
452 * waiting for a page to become available and make them
453 * runnable again when possible. The trouble is that this
454 * consumes a lot of space, especially when so few things
455 * wait on pages at a given time. So instead of using
456 * per-page waitqueues, we use a waitqueue hash table.
458 * The bucket discipline is to sleep on the same queue when
459 * colliding and wake all in that wait queue when removing.
460 * When something wakes, it must check to be sure its page is
461 * truly available, a la thundering herd. The cost of a
462 * collision is great, but given the expected load of the
463 * table, they should be so rare as to be outweighed by the
464 * benefits from the saved space.
466 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
467 * primary users of these fields, and in mm/page_alloc.c
468 * free_area_init_core() performs the initialization of them.
470 wait_queue_head_t *wait_table;
471 unsigned long wait_table_hash_nr_entries;
472 unsigned long wait_table_bits;
475 /* free areas of different sizes */
476 struct free_area free_area[MAX_ORDER];
478 /* zone flags, see below */
481 /* Write-intensive fields used from the page allocator */
486 /* Write-intensive fields used by page reclaim */
488 /* Fields commonly accessed by the page reclaim scanner */
490 struct lruvec lruvec;
492 /* Evictions & activations on the inactive file list */
493 atomic_long_t inactive_age;
496 * When free pages are below this point, additional steps are taken
497 * when reading the number of free pages to avoid per-cpu counter
498 * drift allowing watermarks to be breached
500 unsigned long percpu_drift_mark;
502 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
503 /* pfn where compaction free scanner should start */
504 unsigned long compact_cached_free_pfn;
505 /* pfn where async and sync compaction migration scanner should start */
506 unsigned long compact_cached_migrate_pfn[2];
509 #ifdef CONFIG_COMPACTION
511 * On compaction failure, 1<<compact_defer_shift compactions
512 * are skipped before trying again. The number attempted since
513 * last failure is tracked with compact_considered.
515 unsigned int compact_considered;
516 unsigned int compact_defer_shift;
517 int compact_order_failed;
520 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
521 /* Set to true when the PG_migrate_skip bits should be cleared */
522 bool compact_blockskip_flush;
526 /* Zone statistics */
527 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
528 } ____cacheline_internodealigned_in_smp;
531 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
532 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
533 ZONE_CONGESTED, /* zone has many dirty pages backed by
536 ZONE_DIRTY, /* reclaim scanning has recently found
537 * many dirty file pages at the tail
540 ZONE_WRITEBACK, /* reclaim scanning has recently found
541 * many pages under writeback
543 ZONE_FAIR_DEPLETED, /* fair zone policy batch depleted */
546 static inline unsigned long zone_end_pfn(const struct zone *zone)
548 return zone->zone_start_pfn + zone->spanned_pages;
551 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
553 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
556 static inline bool zone_is_initialized(struct zone *zone)
558 return !!zone->wait_table;
561 static inline bool zone_is_empty(struct zone *zone)
563 return zone->spanned_pages == 0;
567 * The "priority" of VM scanning is how much of the queues we will scan in one
568 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
569 * queues ("queue_length >> 12") during an aging round.
571 #define DEF_PRIORITY 12
573 /* Maximum number of zones on a zonelist */
574 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
579 * The NUMA zonelists are doubled because we need zonelists that restrict the
580 * allocations to a single node for __GFP_THISNODE.
582 * [0] : Zonelist with fallback
583 * [1] : No fallback (__GFP_THISNODE)
585 #define MAX_ZONELISTS 2
587 #define MAX_ZONELISTS 1
591 * This struct contains information about a zone in a zonelist. It is stored
592 * here to avoid dereferences into large structures and lookups of tables
595 struct zone *zone; /* Pointer to actual zone */
596 int zone_idx; /* zone_idx(zoneref->zone) */
600 * One allocation request operates on a zonelist. A zonelist
601 * is a list of zones, the first one is the 'goal' of the
602 * allocation, the other zones are fallback zones, in decreasing
605 * To speed the reading of the zonelist, the zonerefs contain the zone index
606 * of the entry being read. Helper functions to access information given
607 * a struct zoneref are
609 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
610 * zonelist_zone_idx() - Return the index of the zone for an entry
611 * zonelist_node_idx() - Return the index of the node for an entry
614 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
617 #ifndef CONFIG_DISCONTIGMEM
618 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
619 extern struct page *mem_map;
623 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
624 * (mostly NUMA machines?) to denote a higher-level memory zone than the
627 * On NUMA machines, each NUMA node would have a pg_data_t to describe
628 * it's memory layout.
630 * Memory statistics and page replacement data structures are maintained on a
634 typedef struct pglist_data {
635 struct zone node_zones[MAX_NR_ZONES];
636 struct zonelist node_zonelists[MAX_ZONELISTS];
638 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
639 struct page *node_mem_map;
640 #ifdef CONFIG_PAGE_EXTENSION
641 struct page_ext *node_page_ext;
644 #ifndef CONFIG_NO_BOOTMEM
645 struct bootmem_data *bdata;
647 #ifdef CONFIG_MEMORY_HOTPLUG
649 * Must be held any time you expect node_start_pfn, node_present_pages
650 * or node_spanned_pages stay constant. Holding this will also
651 * guarantee that any pfn_valid() stays that way.
653 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
654 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
656 * Nests above zone->lock and zone->span_seqlock
658 spinlock_t node_size_lock;
660 unsigned long node_start_pfn;
661 unsigned long node_present_pages; /* total number of physical pages */
662 unsigned long node_spanned_pages; /* total size of physical page
663 range, including holes */
665 wait_queue_head_t kswapd_wait;
666 wait_queue_head_t pfmemalloc_wait;
667 struct task_struct *kswapd; /* Protected by
668 mem_hotplug_begin/end() */
669 int kswapd_max_order;
670 enum zone_type classzone_idx;
671 #ifdef CONFIG_NUMA_BALANCING
672 /* Lock serializing the migrate rate limiting window */
673 spinlock_t numabalancing_migrate_lock;
675 /* Rate limiting time interval */
676 unsigned long numabalancing_migrate_next_window;
678 /* Number of pages migrated during the rate limiting time interval */
679 unsigned long numabalancing_migrate_nr_pages;
682 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
684 * If memory initialisation on large machines is deferred then this
685 * is the first PFN that needs to be initialised.
687 unsigned long first_deferred_pfn;
688 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
691 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
692 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
693 #ifdef CONFIG_FLAT_NODE_MEM_MAP
694 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
696 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
698 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
700 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
701 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
703 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
705 return pgdat->node_start_pfn + pgdat->node_spanned_pages;
708 static inline bool pgdat_is_empty(pg_data_t *pgdat)
710 return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
713 static inline int zone_id(const struct zone *zone)
715 struct pglist_data *pgdat = zone->zone_pgdat;
717 return zone - pgdat->node_zones;
720 #ifdef CONFIG_ZONE_DEVICE
721 static inline bool is_dev_zone(const struct zone *zone)
723 return zone_id(zone) == ZONE_DEVICE;
726 static inline bool is_dev_zone(const struct zone *zone)
732 #include <linux/memory_hotplug.h>
734 extern struct mutex zonelists_mutex;
735 void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
736 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
737 bool zone_watermark_ok(struct zone *z, unsigned int order,
738 unsigned long mark, int classzone_idx, int alloc_flags);
739 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
740 unsigned long mark, int classzone_idx);
741 enum memmap_context {
745 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
748 extern void lruvec_init(struct lruvec *lruvec);
750 static inline struct zone *lruvec_zone(struct lruvec *lruvec)
755 return container_of(lruvec, struct zone, lruvec);
759 #ifdef CONFIG_HAVE_MEMORY_PRESENT
760 void memory_present(int nid, unsigned long start, unsigned long end);
762 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
765 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
766 int local_memory_node(int node_id);
768 static inline int local_memory_node(int node_id) { return node_id; };
771 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
772 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
776 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
778 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
780 static inline int populated_zone(struct zone *zone)
782 return (!!zone->present_pages);
785 extern int movable_zone;
787 #ifdef CONFIG_HIGHMEM
788 static inline int zone_movable_is_highmem(void)
790 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
791 return movable_zone == ZONE_HIGHMEM;
793 return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
798 static inline int is_highmem_idx(enum zone_type idx)
800 #ifdef CONFIG_HIGHMEM
801 return (idx == ZONE_HIGHMEM ||
802 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
809 * is_highmem - helper function to quickly check if a struct zone is a
810 * highmem zone or not. This is an attempt to keep references
811 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
812 * @zone - pointer to struct zone variable
814 static inline int is_highmem(struct zone *zone)
816 #ifdef CONFIG_HIGHMEM
817 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
818 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
819 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
820 zone_movable_is_highmem());
826 /* These two functions are used to setup the per zone pages min values */
828 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
829 void __user *, size_t *, loff_t *);
830 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
831 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
832 void __user *, size_t *, loff_t *);
833 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
834 void __user *, size_t *, loff_t *);
835 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
836 void __user *, size_t *, loff_t *);
837 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
838 void __user *, size_t *, loff_t *);
840 extern int numa_zonelist_order_handler(struct ctl_table *, int,
841 void __user *, size_t *, loff_t *);
842 extern char numa_zonelist_order[];
843 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
845 #ifndef CONFIG_NEED_MULTIPLE_NODES
847 extern struct pglist_data contig_page_data;
848 #define NODE_DATA(nid) (&contig_page_data)
849 #define NODE_MEM_MAP(nid) mem_map
851 #else /* CONFIG_NEED_MULTIPLE_NODES */
853 #include <asm/mmzone.h>
855 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
857 extern struct pglist_data *first_online_pgdat(void);
858 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
859 extern struct zone *next_zone(struct zone *zone);
862 * for_each_online_pgdat - helper macro to iterate over all online nodes
863 * @pgdat - pointer to a pg_data_t variable
865 #define for_each_online_pgdat(pgdat) \
866 for (pgdat = first_online_pgdat(); \
868 pgdat = next_online_pgdat(pgdat))
870 * for_each_zone - helper macro to iterate over all memory zones
871 * @zone - pointer to struct zone variable
873 * The user only needs to declare the zone variable, for_each_zone
876 #define for_each_zone(zone) \
877 for (zone = (first_online_pgdat())->node_zones; \
879 zone = next_zone(zone))
881 #define for_each_populated_zone(zone) \
882 for (zone = (first_online_pgdat())->node_zones; \
884 zone = next_zone(zone)) \
885 if (!populated_zone(zone)) \
889 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
891 return zoneref->zone;
894 static inline int zonelist_zone_idx(struct zoneref *zoneref)
896 return zoneref->zone_idx;
899 static inline int zonelist_node_idx(struct zoneref *zoneref)
902 /* zone_to_nid not available in this context */
903 return zoneref->zone->node;
906 #endif /* CONFIG_NUMA */
910 * 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
911 * @z - The cursor used as a starting point for the search
912 * @highest_zoneidx - The zone index of the highest zone to return
913 * @nodes - An optional nodemask to filter the zonelist with
915 * This function returns the next zone at or below a given zone index that is
916 * within the allowed nodemask using a cursor as the starting point for the
917 * search. The zoneref returned is a cursor that represents the current zone
918 * being examined. It should be advanced by one before calling
919 * next_zones_zonelist again.
921 struct zoneref *next_zones_zonelist(struct zoneref *z,
922 enum zone_type highest_zoneidx,
926 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
927 * @zonelist - The zonelist to search for a suitable zone
928 * @highest_zoneidx - The zone index of the highest zone to return
929 * @nodes - An optional nodemask to filter the zonelist with
930 * @zone - The first suitable zone found is returned via this parameter
932 * This function returns the first zone at or below a given zone index that is
933 * within the allowed nodemask. The zoneref returned is a cursor that can be
934 * used to iterate the zonelist with next_zones_zonelist by advancing it by
935 * one before calling.
937 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
938 enum zone_type highest_zoneidx,
942 struct zoneref *z = next_zones_zonelist(zonelist->_zonerefs,
943 highest_zoneidx, nodes);
944 *zone = zonelist_zone(z);
949 * 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
950 * @zone - The current zone in the iterator
951 * @z - The current pointer within zonelist->zones being iterated
952 * @zlist - The zonelist being iterated
953 * @highidx - The zone index of the highest zone to return
954 * @nodemask - Nodemask allowed by the allocator
956 * This iterator iterates though all zones at or below a given zone index and
957 * within a given nodemask
959 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
960 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
962 z = next_zones_zonelist(++z, highidx, nodemask), \
963 zone = zonelist_zone(z)) \
966 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
967 * @zone - The current zone in the iterator
968 * @z - The current pointer within zonelist->zones being iterated
969 * @zlist - The zonelist being iterated
970 * @highidx - The zone index of the highest zone to return
972 * This iterator iterates though all zones at or below a given zone index.
974 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
975 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
977 #ifdef CONFIG_SPARSEMEM
978 #include <asm/sparsemem.h>
981 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
982 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
983 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
989 #ifdef CONFIG_FLATMEM
990 #define pfn_to_nid(pfn) (0)
993 #ifdef CONFIG_SPARSEMEM
996 * SECTION_SHIFT #bits space required to store a section #
998 * PA_SECTION_SHIFT physical address to/from section number
999 * PFN_SECTION_SHIFT pfn to/from section number
1001 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1002 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1004 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1006 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1007 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1009 #define SECTION_BLOCKFLAGS_BITS \
1010 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1012 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1013 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1016 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1017 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1019 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1020 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1024 struct mem_section {
1026 * This is, logically, a pointer to an array of struct
1027 * pages. However, it is stored with some other magic.
1028 * (see sparse.c::sparse_init_one_section())
1030 * Additionally during early boot we encode node id of
1031 * the location of the section here to guide allocation.
1032 * (see sparse.c::memory_present())
1034 * Making it a UL at least makes someone do a cast
1035 * before using it wrong.
1037 unsigned long section_mem_map;
1039 /* See declaration of similar field in struct zone */
1040 unsigned long *pageblock_flags;
1041 #ifdef CONFIG_PAGE_EXTENSION
1043 * If !SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1044 * section. (see page_ext.h about this.)
1046 struct page_ext *page_ext;
1050 * WARNING: mem_section must be a power-of-2 in size for the
1051 * calculation and use of SECTION_ROOT_MASK to make sense.
1055 #ifdef CONFIG_SPARSEMEM_EXTREME
1056 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1058 #define SECTIONS_PER_ROOT 1
1061 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1062 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1063 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1065 #ifdef CONFIG_SPARSEMEM_EXTREME
1066 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1068 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1071 static inline struct mem_section *__nr_to_section(unsigned long nr)
1073 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1075 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1077 extern int __section_nr(struct mem_section* ms);
1078 extern unsigned long usemap_size(void);
1081 * We use the lower bits of the mem_map pointer to store
1082 * a little bit of information. There should be at least
1083 * 3 bits here due to 32-bit alignment.
1085 #define SECTION_MARKED_PRESENT (1UL<<0)
1086 #define SECTION_HAS_MEM_MAP (1UL<<1)
1087 #define SECTION_MAP_LAST_BIT (1UL<<2)
1088 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1089 #define SECTION_NID_SHIFT 2
1091 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1093 unsigned long map = section->section_mem_map;
1094 map &= SECTION_MAP_MASK;
1095 return (struct page *)map;
1098 static inline int present_section(struct mem_section *section)
1100 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1103 static inline int present_section_nr(unsigned long nr)
1105 return present_section(__nr_to_section(nr));
1108 static inline int valid_section(struct mem_section *section)
1110 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1113 static inline int valid_section_nr(unsigned long nr)
1115 return valid_section(__nr_to_section(nr));
1118 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1120 return __nr_to_section(pfn_to_section_nr(pfn));
1123 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1124 static inline int pfn_valid(unsigned long pfn)
1126 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1128 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1132 static inline int pfn_present(unsigned long pfn)
1134 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1136 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1140 * These are _only_ used during initialisation, therefore they
1141 * can use __initdata ... They could have names to indicate
1145 #define pfn_to_nid(pfn) \
1147 unsigned long __pfn_to_nid_pfn = (pfn); \
1148 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1151 #define pfn_to_nid(pfn) (0)
1154 #define early_pfn_valid(pfn) pfn_valid(pfn)
1155 void sparse_init(void);
1157 #define sparse_init() do {} while (0)
1158 #define sparse_index_init(_sec, _nid) do {} while (0)
1159 #endif /* CONFIG_SPARSEMEM */
1162 * During memory init memblocks map pfns to nids. The search is expensive and
1163 * this caches recent lookups. The implementation of __early_pfn_to_nid
1164 * may treat start/end as pfns or sections.
1166 struct mminit_pfnnid_cache {
1167 unsigned long last_start;
1168 unsigned long last_end;
1172 #ifndef early_pfn_valid
1173 #define early_pfn_valid(pfn) (1)
1176 void memory_present(int nid, unsigned long start, unsigned long end);
1177 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1180 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1181 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1182 * pfn_valid_within() should be used in this case; we optimise this away
1183 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1185 #ifdef CONFIG_HOLES_IN_ZONE
1186 #define pfn_valid_within(pfn) pfn_valid(pfn)
1188 #define pfn_valid_within(pfn) (1)
1191 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1193 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1194 * associated with it or not. In FLATMEM, it is expected that holes always
1195 * have valid memmap as long as there is valid PFNs either side of the hole.
1196 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1199 * However, an ARM, and maybe other embedded architectures in the future
1200 * free memmap backing holes to save memory on the assumption the memmap is
1201 * never used. The page_zone linkages are then broken even though pfn_valid()
1202 * returns true. A walker of the full memmap must then do this additional
1203 * check to ensure the memmap they are looking at is sane by making sure
1204 * the zone and PFN linkages are still valid. This is expensive, but walkers
1205 * of the full memmap are extremely rare.
1207 int memmap_valid_within(unsigned long pfn,
1208 struct page *page, struct zone *zone);
1210 static inline int memmap_valid_within(unsigned long pfn,
1211 struct page *page, struct zone *zone)
1215 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1217 #endif /* !__GENERATING_BOUNDS.H */
1218 #endif /* !__ASSEMBLY__ */
1219 #endif /* _LINUX_MMZONE_H */