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 <generated/bounds.h>
19 #include <asm/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 * coelesce naturally under reasonable reclaim pressure and those which
36 #define PAGE_ALLOC_COSTLY_ORDER 3
38 #define MIGRATE_UNMOVABLE 0
39 #define MIGRATE_RECLAIMABLE 1
40 #define MIGRATE_MOVABLE 2
41 #define MIGRATE_PCPTYPES 3 /* the number of types on the pcp lists */
42 #define MIGRATE_RESERVE 3
43 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
44 #define MIGRATE_TYPES 5
46 #define for_each_migratetype_order(order, type) \
47 for (order = 0; order < MAX_ORDER; order++) \
48 for (type = 0; type < MIGRATE_TYPES; type++)
50 extern int page_group_by_mobility_disabled;
52 static inline int get_pageblock_migratetype(struct page *page)
54 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
58 struct list_head free_list[MIGRATE_TYPES];
59 unsigned long nr_free;
65 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
66 * So add a wild amount of padding here to ensure that they fall into separate
67 * cachelines. There are very few zone structures in the machine, so space
68 * consumption is not a concern here.
70 #if defined(CONFIG_SMP)
73 } ____cacheline_internodealigned_in_smp;
74 #define ZONE_PADDING(name) struct zone_padding name;
76 #define ZONE_PADDING(name)
80 /* First 128 byte cacheline (assuming 64 bit words) */
83 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
84 NR_ACTIVE_ANON, /* " " " " " */
85 NR_INACTIVE_FILE, /* " " " " " */
86 NR_ACTIVE_FILE, /* " " " " " */
87 NR_UNEVICTABLE, /* " " " " " */
88 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
89 NR_ANON_PAGES, /* Mapped anonymous pages */
90 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
91 only modified from process context */
96 NR_SLAB_UNRECLAIMABLE,
97 NR_PAGETABLE, /* used for pagetables */
99 /* Second 128 byte cacheline */
100 NR_UNSTABLE_NFS, /* NFS unstable pages */
103 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
104 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
105 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
106 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
108 NUMA_HIT, /* allocated in intended node */
109 NUMA_MISS, /* allocated in non intended node */
110 NUMA_FOREIGN, /* was intended here, hit elsewhere */
111 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
112 NUMA_LOCAL, /* allocation from local node */
113 NUMA_OTHER, /* allocation from other node */
115 NR_VM_ZONE_STAT_ITEMS };
118 * We do arithmetic on the LRU lists in various places in the code,
119 * so it is important to keep the active lists LRU_ACTIVE higher in
120 * the array than the corresponding inactive lists, and to keep
121 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
123 * This has to be kept in sync with the statistics in zone_stat_item
124 * above and the descriptions in vmstat_text in mm/vmstat.c
131 LRU_INACTIVE_ANON = LRU_BASE,
132 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
133 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
134 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
139 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
141 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
143 static inline int is_file_lru(enum lru_list l)
145 return (l == LRU_INACTIVE_FILE || l == LRU_ACTIVE_FILE);
148 static inline int is_active_lru(enum lru_list l)
150 return (l == LRU_ACTIVE_ANON || l == LRU_ACTIVE_FILE);
153 static inline int is_unevictable_lru(enum lru_list l)
155 return (l == LRU_UNEVICTABLE);
158 enum zone_watermarks {
165 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
166 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
167 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
169 struct per_cpu_pages {
170 int count; /* number of pages in the list */
171 int high; /* high watermark, emptying needed */
172 int batch; /* chunk size for buddy add/remove */
174 /* Lists of pages, one per migrate type stored on the pcp-lists */
175 struct list_head lists[MIGRATE_PCPTYPES];
178 struct per_cpu_pageset {
179 struct per_cpu_pages pcp;
185 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
187 } ____cacheline_aligned_in_smp;
190 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
192 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
195 #endif /* !__GENERATING_BOUNDS.H */
198 #ifdef CONFIG_ZONE_DMA
200 * ZONE_DMA is used when there are devices that are not able
201 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
202 * carve out the portion of memory that is needed for these devices.
203 * The range is arch specific.
208 * ---------------------------
209 * parisc, ia64, sparc <4G
212 * alpha Unlimited or 0-16MB.
214 * i386, x86_64 and multiple other arches
219 #ifdef CONFIG_ZONE_DMA32
221 * x86_64 needs two ZONE_DMAs because it supports devices that are
222 * only able to do DMA to the lower 16M but also 32 bit devices that
223 * can only do DMA areas below 4G.
228 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
229 * performed on pages in ZONE_NORMAL if the DMA devices support
230 * transfers to all addressable memory.
233 #ifdef CONFIG_HIGHMEM
235 * A memory area that is only addressable by the kernel through
236 * mapping portions into its own address space. This is for example
237 * used by i386 to allow the kernel to address the memory beyond
238 * 900MB. The kernel will set up special mappings (page
239 * table entries on i386) for each page that the kernel needs to
248 #ifndef __GENERATING_BOUNDS_H
251 * When a memory allocation must conform to specific limitations (such
252 * as being suitable for DMA) the caller will pass in hints to the
253 * allocator in the gfp_mask, in the zone modifier bits. These bits
254 * are used to select a priority ordered list of memory zones which
255 * match the requested limits. See gfp_zone() in include/linux/gfp.h
259 #define ZONES_SHIFT 0
260 #elif MAX_NR_ZONES <= 2
261 #define ZONES_SHIFT 1
262 #elif MAX_NR_ZONES <= 4
263 #define ZONES_SHIFT 2
265 #error ZONES_SHIFT -- too many zones configured adjust calculation
268 struct zone_reclaim_stat {
270 * The pageout code in vmscan.c keeps track of how many of the
271 * mem/swap backed and file backed pages are refeferenced.
272 * The higher the rotated/scanned ratio, the more valuable
275 * The anon LRU stats live in [0], file LRU stats in [1]
277 unsigned long recent_rotated[2];
278 unsigned long recent_scanned[2];
281 * accumulated for batching
283 unsigned long nr_saved_scan[NR_LRU_LISTS];
287 /* Fields commonly accessed by the page allocator */
289 /* zone watermarks, access with *_wmark_pages(zone) macros */
290 unsigned long watermark[NR_WMARK];
293 * We don't know if the memory that we're going to allocate will be freeable
294 * or/and it will be released eventually, so to avoid totally wasting several
295 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
296 * to run OOM on the lower zones despite there's tons of freeable ram
297 * on the higher zones). This array is recalculated at runtime if the
298 * sysctl_lowmem_reserve_ratio sysctl changes.
300 unsigned long lowmem_reserve[MAX_NR_ZONES];
305 * zone reclaim becomes active if more unmapped pages exist.
307 unsigned long min_unmapped_pages;
308 unsigned long min_slab_pages;
309 struct per_cpu_pageset *pageset[NR_CPUS];
311 struct per_cpu_pageset pageset[NR_CPUS];
314 * free areas of different sizes
317 #ifdef CONFIG_MEMORY_HOTPLUG
318 /* see spanned/present_pages for more description */
319 seqlock_t span_seqlock;
321 struct free_area free_area[MAX_ORDER];
323 #ifndef CONFIG_SPARSEMEM
325 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
326 * In SPARSEMEM, this map is stored in struct mem_section
328 unsigned long *pageblock_flags;
329 #endif /* CONFIG_SPARSEMEM */
334 /* Fields commonly accessed by the page reclaim scanner */
337 struct list_head list;
340 struct zone_reclaim_stat reclaim_stat;
342 unsigned long pages_scanned; /* since last reclaim */
343 unsigned long flags; /* zone flags, see below */
345 /* Zone statistics */
346 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
349 * prev_priority holds the scanning priority for this zone. It is
350 * defined as the scanning priority at which we achieved our reclaim
351 * target at the previous try_to_free_pages() or balance_pgdat()
354 * We use prev_priority as a measure of how much stress page reclaim is
355 * under - it drives the swappiness decision: whether to unmap mapped
358 * Access to both this field is quite racy even on uniprocessor. But
359 * it is expected to average out OK.
364 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
365 * this zone's LRU. Maintained by the pageout code.
367 unsigned int inactive_ratio;
371 /* Rarely used or read-mostly fields */
374 * wait_table -- the array holding the hash table
375 * wait_table_hash_nr_entries -- the size of the hash table array
376 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
378 * The purpose of all these is to keep track of the people
379 * waiting for a page to become available and make them
380 * runnable again when possible. The trouble is that this
381 * consumes a lot of space, especially when so few things
382 * wait on pages at a given time. So instead of using
383 * per-page waitqueues, we use a waitqueue hash table.
385 * The bucket discipline is to sleep on the same queue when
386 * colliding and wake all in that wait queue when removing.
387 * When something wakes, it must check to be sure its page is
388 * truly available, a la thundering herd. The cost of a
389 * collision is great, but given the expected load of the
390 * table, they should be so rare as to be outweighed by the
391 * benefits from the saved space.
393 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
394 * primary users of these fields, and in mm/page_alloc.c
395 * free_area_init_core() performs the initialization of them.
397 wait_queue_head_t * wait_table;
398 unsigned long wait_table_hash_nr_entries;
399 unsigned long wait_table_bits;
402 * Discontig memory support fields.
404 struct pglist_data *zone_pgdat;
405 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
406 unsigned long zone_start_pfn;
409 * zone_start_pfn, spanned_pages and present_pages are all
410 * protected by span_seqlock. It is a seqlock because it has
411 * to be read outside of zone->lock, and it is done in the main
412 * allocator path. But, it is written quite infrequently.
414 * The lock is declared along with zone->lock because it is
415 * frequently read in proximity to zone->lock. It's good to
416 * give them a chance of being in the same cacheline.
418 unsigned long spanned_pages; /* total size, including holes */
419 unsigned long present_pages; /* amount of memory (excluding holes) */
422 * rarely used fields:
425 } ____cacheline_internodealigned_in_smp;
428 ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */
429 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
430 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
433 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
435 set_bit(flag, &zone->flags);
438 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
440 return test_and_set_bit(flag, &zone->flags);
443 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
445 clear_bit(flag, &zone->flags);
448 static inline int zone_is_all_unreclaimable(const struct zone *zone)
450 return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags);
453 static inline int zone_is_reclaim_locked(const struct zone *zone)
455 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
458 static inline int zone_is_oom_locked(const struct zone *zone)
460 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
464 * The "priority" of VM scanning is how much of the queues we will scan in one
465 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
466 * queues ("queue_length >> 12") during an aging round.
468 #define DEF_PRIORITY 12
470 /* Maximum number of zones on a zonelist */
471 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
476 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
477 * allocations to a single node for GFP_THISNODE.
479 * [0] : Zonelist with fallback
480 * [1] : No fallback (GFP_THISNODE)
482 #define MAX_ZONELISTS 2
486 * We cache key information from each zonelist for smaller cache
487 * footprint when scanning for free pages in get_page_from_freelist().
489 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
490 * up short of free memory since the last time (last_fullzone_zap)
491 * we zero'd fullzones.
492 * 2) The array z_to_n[] maps each zone in the zonelist to its node
493 * id, so that we can efficiently evaluate whether that node is
494 * set in the current tasks mems_allowed.
496 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
497 * indexed by a zones offset in the zonelist zones[] array.
499 * The get_page_from_freelist() routine does two scans. During the
500 * first scan, we skip zones whose corresponding bit in 'fullzones'
501 * is set or whose corresponding node in current->mems_allowed (which
502 * comes from cpusets) is not set. During the second scan, we bypass
503 * this zonelist_cache, to ensure we look methodically at each zone.
505 * Once per second, we zero out (zap) fullzones, forcing us to
506 * reconsider nodes that might have regained more free memory.
507 * The field last_full_zap is the time we last zapped fullzones.
509 * This mechanism reduces the amount of time we waste repeatedly
510 * reexaming zones for free memory when they just came up low on
511 * memory momentarilly ago.
513 * The zonelist_cache struct members logically belong in struct
514 * zonelist. However, the mempolicy zonelists constructed for
515 * MPOL_BIND are intentionally variable length (and usually much
516 * shorter). A general purpose mechanism for handling structs with
517 * multiple variable length members is more mechanism than we want
518 * here. We resort to some special case hackery instead.
520 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
521 * part because they are shorter), so we put the fixed length stuff
522 * at the front of the zonelist struct, ending in a variable length
523 * zones[], as is needed by MPOL_BIND.
525 * Then we put the optional zonelist cache on the end of the zonelist
526 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
527 * the fixed length portion at the front of the struct. This pointer
528 * both enables us to find the zonelist cache, and in the case of
529 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
530 * to know that the zonelist cache is not there.
532 * The end result is that struct zonelists come in two flavors:
533 * 1) The full, fixed length version, shown below, and
534 * 2) The custom zonelists for MPOL_BIND.
535 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
537 * Even though there may be multiple CPU cores on a node modifying
538 * fullzones or last_full_zap in the same zonelist_cache at the same
539 * time, we don't lock it. This is just hint data - if it is wrong now
540 * and then, the allocator will still function, perhaps a bit slower.
544 struct zonelist_cache {
545 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
546 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
547 unsigned long last_full_zap; /* when last zap'd (jiffies) */
550 #define MAX_ZONELISTS 1
551 struct zonelist_cache;
555 * This struct contains information about a zone in a zonelist. It is stored
556 * here to avoid dereferences into large structures and lookups of tables
559 struct zone *zone; /* Pointer to actual zone */
560 int zone_idx; /* zone_idx(zoneref->zone) */
564 * One allocation request operates on a zonelist. A zonelist
565 * is a list of zones, the first one is the 'goal' of the
566 * allocation, the other zones are fallback zones, in decreasing
569 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
570 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
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
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
581 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
582 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
584 struct zonelist_cache zlcache; // optional ...
588 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
589 struct node_active_region {
590 unsigned long start_pfn;
591 unsigned long end_pfn;
594 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
596 #ifndef CONFIG_DISCONTIGMEM
597 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
598 extern struct page *mem_map;
602 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
603 * (mostly NUMA machines?) to denote a higher-level memory zone than the
606 * On NUMA machines, each NUMA node would have a pg_data_t to describe
607 * it's memory layout.
609 * Memory statistics and page replacement data structures are maintained on a
613 typedef struct pglist_data {
614 struct zone node_zones[MAX_NR_ZONES];
615 struct zonelist node_zonelists[MAX_ZONELISTS];
617 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
618 struct page *node_mem_map;
619 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
620 struct page_cgroup *node_page_cgroup;
623 #ifndef CONFIG_NO_BOOTMEM
624 struct bootmem_data *bdata;
626 #ifdef CONFIG_MEMORY_HOTPLUG
628 * Must be held any time you expect node_start_pfn, node_present_pages
629 * or node_spanned_pages stay constant. Holding this will also
630 * guarantee that any pfn_valid() stays that way.
632 * Nests above zone->lock and zone->size_seqlock.
634 spinlock_t node_size_lock;
636 unsigned long node_start_pfn;
637 unsigned long node_present_pages; /* total number of physical pages */
638 unsigned long node_spanned_pages; /* total size of physical page
639 range, including holes */
641 wait_queue_head_t kswapd_wait;
642 struct task_struct *kswapd;
643 int kswapd_max_order;
646 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
647 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
648 #ifdef CONFIG_FLAT_NODE_MEM_MAP
649 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
651 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
653 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
655 #include <linux/memory_hotplug.h>
657 void get_zone_counts(unsigned long *active, unsigned long *inactive,
658 unsigned long *free);
659 void build_all_zonelists(void);
660 void wakeup_kswapd(struct zone *zone, int order);
661 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
662 int classzone_idx, int alloc_flags);
663 enum memmap_context {
667 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
669 enum memmap_context context);
671 #ifdef CONFIG_HAVE_MEMORY_PRESENT
672 void memory_present(int nid, unsigned long start, unsigned long end);
674 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
677 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
678 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
682 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
684 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
686 static inline int populated_zone(struct zone *zone)
688 return (!!zone->present_pages);
691 extern int movable_zone;
693 static inline int zone_movable_is_highmem(void)
695 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
696 return movable_zone == ZONE_HIGHMEM;
702 static inline int is_highmem_idx(enum zone_type idx)
704 #ifdef CONFIG_HIGHMEM
705 return (idx == ZONE_HIGHMEM ||
706 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
712 static inline int is_normal_idx(enum zone_type idx)
714 return (idx == ZONE_NORMAL);
718 * is_highmem - helper function to quickly check if a struct zone is a
719 * highmem zone or not. This is an attempt to keep references
720 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
721 * @zone - pointer to struct zone variable
723 static inline int is_highmem(struct zone *zone)
725 #ifdef CONFIG_HIGHMEM
726 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
727 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
728 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
729 zone_movable_is_highmem());
735 static inline int is_normal(struct zone *zone)
737 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
740 static inline int is_dma32(struct zone *zone)
742 #ifdef CONFIG_ZONE_DMA32
743 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
749 static inline int is_dma(struct zone *zone)
751 #ifdef CONFIG_ZONE_DMA
752 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
758 /* These two functions are used to setup the per zone pages min values */
760 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
761 void __user *, size_t *, loff_t *);
762 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
763 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
764 void __user *, size_t *, loff_t *);
765 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
766 void __user *, size_t *, loff_t *);
767 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
768 void __user *, size_t *, loff_t *);
769 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
770 void __user *, size_t *, loff_t *);
772 extern int numa_zonelist_order_handler(struct ctl_table *, int,
773 void __user *, size_t *, loff_t *);
774 extern char numa_zonelist_order[];
775 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
777 #ifndef CONFIG_NEED_MULTIPLE_NODES
779 extern struct pglist_data contig_page_data;
780 #define NODE_DATA(nid) (&contig_page_data)
781 #define NODE_MEM_MAP(nid) mem_map
783 #else /* CONFIG_NEED_MULTIPLE_NODES */
785 #include <asm/mmzone.h>
787 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
789 extern struct pglist_data *first_online_pgdat(void);
790 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
791 extern struct zone *next_zone(struct zone *zone);
794 * for_each_online_pgdat - helper macro to iterate over all online nodes
795 * @pgdat - pointer to a pg_data_t variable
797 #define for_each_online_pgdat(pgdat) \
798 for (pgdat = first_online_pgdat(); \
800 pgdat = next_online_pgdat(pgdat))
802 * for_each_zone - helper macro to iterate over all memory zones
803 * @zone - pointer to struct zone variable
805 * The user only needs to declare the zone variable, for_each_zone
808 #define for_each_zone(zone) \
809 for (zone = (first_online_pgdat())->node_zones; \
811 zone = next_zone(zone))
813 #define for_each_populated_zone(zone) \
814 for (zone = (first_online_pgdat())->node_zones; \
816 zone = next_zone(zone)) \
817 if (!populated_zone(zone)) \
821 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
823 return zoneref->zone;
826 static inline int zonelist_zone_idx(struct zoneref *zoneref)
828 return zoneref->zone_idx;
831 static inline int zonelist_node_idx(struct zoneref *zoneref)
834 /* zone_to_nid not available in this context */
835 return zoneref->zone->node;
838 #endif /* CONFIG_NUMA */
842 * 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
843 * @z - The cursor used as a starting point for the search
844 * @highest_zoneidx - The zone index of the highest zone to return
845 * @nodes - An optional nodemask to filter the zonelist with
846 * @zone - The first suitable zone found is returned via this parameter
848 * This function returns the next zone at or below a given zone index that is
849 * within the allowed nodemask using a cursor as the starting point for the
850 * search. The zoneref returned is a cursor that represents the current zone
851 * being examined. It should be advanced by one before calling
852 * next_zones_zonelist again.
854 struct zoneref *next_zones_zonelist(struct zoneref *z,
855 enum zone_type highest_zoneidx,
860 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
861 * @zonelist - The zonelist to search for a suitable zone
862 * @highest_zoneidx - The zone index of the highest zone to return
863 * @nodes - An optional nodemask to filter the zonelist with
864 * @zone - The first suitable zone found is returned via this parameter
866 * This function returns the first zone at or below a given zone index that is
867 * within the allowed nodemask. The zoneref returned is a cursor that can be
868 * used to iterate the zonelist with next_zones_zonelist by advancing it by
869 * one before calling.
871 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
872 enum zone_type highest_zoneidx,
876 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
881 * 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
882 * @zone - The current zone in the iterator
883 * @z - The current pointer within zonelist->zones being iterated
884 * @zlist - The zonelist being iterated
885 * @highidx - The zone index of the highest zone to return
886 * @nodemask - Nodemask allowed by the allocator
888 * This iterator iterates though all zones at or below a given zone index and
889 * within a given nodemask
891 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
892 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
894 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
897 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
898 * @zone - The current zone in the iterator
899 * @z - The current pointer within zonelist->zones being iterated
900 * @zlist - The zonelist being iterated
901 * @highidx - The zone index of the highest zone to return
903 * This iterator iterates though all zones at or below a given zone index.
905 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
906 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
908 #ifdef CONFIG_SPARSEMEM
909 #include <asm/sparsemem.h>
912 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
913 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
914 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
920 #ifdef CONFIG_FLATMEM
921 #define pfn_to_nid(pfn) (0)
924 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
925 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
927 #ifdef CONFIG_SPARSEMEM
930 * SECTION_SHIFT #bits space required to store a section #
932 * PA_SECTION_SHIFT physical address to/from section number
933 * PFN_SECTION_SHIFT pfn to/from section number
935 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
937 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
938 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
940 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
942 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
943 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
945 #define SECTION_BLOCKFLAGS_BITS \
946 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
948 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
949 #error Allocator MAX_ORDER exceeds SECTION_SIZE
956 * This is, logically, a pointer to an array of struct
957 * pages. However, it is stored with some other magic.
958 * (see sparse.c::sparse_init_one_section())
960 * Additionally during early boot we encode node id of
961 * the location of the section here to guide allocation.
962 * (see sparse.c::memory_present())
964 * Making it a UL at least makes someone do a cast
965 * before using it wrong.
967 unsigned long section_mem_map;
969 /* See declaration of similar field in struct zone */
970 unsigned long *pageblock_flags;
971 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
973 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
974 * section. (see memcontrol.h/page_cgroup.h about this.)
976 struct page_cgroup *page_cgroup;
981 #ifdef CONFIG_SPARSEMEM_EXTREME
982 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
984 #define SECTIONS_PER_ROOT 1
987 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
988 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
989 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
991 #ifdef CONFIG_SPARSEMEM_EXTREME
992 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
994 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
997 static inline struct mem_section *__nr_to_section(unsigned long nr)
999 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1001 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1003 extern int __section_nr(struct mem_section* ms);
1004 extern unsigned long usemap_size(void);
1007 * We use the lower bits of the mem_map pointer to store
1008 * a little bit of information. There should be at least
1009 * 3 bits here due to 32-bit alignment.
1011 #define SECTION_MARKED_PRESENT (1UL<<0)
1012 #define SECTION_HAS_MEM_MAP (1UL<<1)
1013 #define SECTION_MAP_LAST_BIT (1UL<<2)
1014 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1015 #define SECTION_NID_SHIFT 2
1017 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1019 unsigned long map = section->section_mem_map;
1020 map &= SECTION_MAP_MASK;
1021 return (struct page *)map;
1024 static inline int present_section(struct mem_section *section)
1026 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1029 static inline int present_section_nr(unsigned long nr)
1031 return present_section(__nr_to_section(nr));
1034 static inline int valid_section(struct mem_section *section)
1036 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1039 static inline int valid_section_nr(unsigned long nr)
1041 return valid_section(__nr_to_section(nr));
1044 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1046 return __nr_to_section(pfn_to_section_nr(pfn));
1049 static inline int pfn_valid(unsigned long pfn)
1051 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1053 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1056 static inline int pfn_present(unsigned long pfn)
1058 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1060 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1064 * These are _only_ used during initialisation, therefore they
1065 * can use __initdata ... They could have names to indicate
1069 #define pfn_to_nid(pfn) \
1071 unsigned long __pfn_to_nid_pfn = (pfn); \
1072 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1075 #define pfn_to_nid(pfn) (0)
1078 #define early_pfn_valid(pfn) pfn_valid(pfn)
1079 void sparse_init(void);
1081 #define sparse_init() do {} while (0)
1082 #define sparse_index_init(_sec, _nid) do {} while (0)
1083 #endif /* CONFIG_SPARSEMEM */
1085 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1086 bool early_pfn_in_nid(unsigned long pfn, int nid);
1088 #define early_pfn_in_nid(pfn, nid) (1)
1091 #ifndef early_pfn_valid
1092 #define early_pfn_valid(pfn) (1)
1095 void memory_present(int nid, unsigned long start, unsigned long end);
1096 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1099 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1100 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1101 * pfn_valid_within() should be used in this case; we optimise this away
1102 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1104 #ifdef CONFIG_HOLES_IN_ZONE
1105 #define pfn_valid_within(pfn) pfn_valid(pfn)
1107 #define pfn_valid_within(pfn) (1)
1110 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1112 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1113 * associated with it or not. In FLATMEM, it is expected that holes always
1114 * have valid memmap as long as there is valid PFNs either side of the hole.
1115 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1118 * However, an ARM, and maybe other embedded architectures in the future
1119 * free memmap backing holes to save memory on the assumption the memmap is
1120 * never used. The page_zone linkages are then broken even though pfn_valid()
1121 * returns true. A walker of the full memmap must then do this additional
1122 * check to ensure the memmap they are looking at is sane by making sure
1123 * the zone and PFN linkages are still valid. This is expensive, but walkers
1124 * of the full memmap are extremely rare.
1126 int memmap_valid_within(unsigned long pfn,
1127 struct page *page, struct zone *zone);
1129 static inline int memmap_valid_within(unsigned long pfn,
1130 struct page *page, struct zone *zone)
1134 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1136 #endif /* !__GENERATING_BOUNDS.H */
1137 #endif /* !__ASSEMBLY__ */
1138 #endif /* _LINUX_MMZONE_H */