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/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_RESERVE 3
42 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
43 #define MIGRATE_TYPES 5
45 #define for_each_migratetype_order(order, type) \
46 for (order = 0; order < MAX_ORDER; order++) \
47 for (type = 0; type < MIGRATE_TYPES; type++)
49 extern int page_group_by_mobility_disabled;
51 static inline int get_pageblock_migratetype(struct page *page)
53 if (unlikely(page_group_by_mobility_disabled))
54 return MIGRATE_UNMOVABLE;
56 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
60 struct list_head free_list[MIGRATE_TYPES];
61 unsigned long nr_free;
67 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
68 * So add a wild amount of padding here to ensure that they fall into separate
69 * cachelines. There are very few zone structures in the machine, so space
70 * consumption is not a concern here.
72 #if defined(CONFIG_SMP)
75 } ____cacheline_internodealigned_in_smp;
76 #define ZONE_PADDING(name) struct zone_padding name;
78 #define ZONE_PADDING(name)
82 /* First 128 byte cacheline (assuming 64 bit words) */
85 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
86 NR_ACTIVE_ANON, /* " " " " " */
87 NR_INACTIVE_FILE, /* " " " " " */
88 NR_ACTIVE_FILE, /* " " " " " */
89 #ifdef CONFIG_UNEVICTABLE_LRU
90 NR_UNEVICTABLE, /* " " " " " */
92 NR_UNEVICTABLE = NR_ACTIVE_FILE, /* avoid compiler errors in dead code */
94 NR_ANON_PAGES, /* Mapped anonymous pages */
95 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
96 only modified from process context */
101 NR_SLAB_UNRECLAIMABLE,
102 NR_PAGETABLE, /* used for pagetables */
103 NR_UNSTABLE_NFS, /* NFS unstable pages */
106 /* Second 128 byte cacheline */
107 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
109 NUMA_HIT, /* allocated in intended node */
110 NUMA_MISS, /* allocated in non intended node */
111 NUMA_FOREIGN, /* was intended here, hit elsewhere */
112 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
113 NUMA_LOCAL, /* allocation from local node */
114 NUMA_OTHER, /* allocation from other node */
116 NR_VM_ZONE_STAT_ITEMS };
119 * We do arithmetic on the LRU lists in various places in the code,
120 * so it is important to keep the active lists LRU_ACTIVE higher in
121 * the array than the corresponding inactive lists, and to keep
122 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
124 * This has to be kept in sync with the statistics in zone_stat_item
125 * above and the descriptions in vmstat_text in mm/vmstat.c
132 LRU_INACTIVE_ANON = LRU_BASE,
133 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
134 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
135 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
136 #ifdef CONFIG_UNEVICTABLE_LRU
139 LRU_UNEVICTABLE = LRU_ACTIVE_FILE, /* avoid compiler errors in dead code */
144 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
146 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
148 static inline int is_file_lru(enum lru_list l)
150 return (l == LRU_INACTIVE_FILE || l == LRU_ACTIVE_FILE);
153 static inline int is_active_lru(enum lru_list l)
155 return (l == LRU_ACTIVE_ANON || l == LRU_ACTIVE_FILE);
158 static inline int is_unevictable_lru(enum lru_list l)
160 #ifdef CONFIG_UNEVICTABLE_LRU
161 return (l == LRU_UNEVICTABLE);
167 struct per_cpu_pages {
168 int count; /* number of pages in the list */
169 int high; /* high watermark, emptying needed */
170 int batch; /* chunk size for buddy add/remove */
171 struct list_head list; /* the list of pages */
174 struct per_cpu_pageset {
175 struct per_cpu_pages pcp;
181 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
183 } ____cacheline_aligned_in_smp;
186 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
188 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
191 #endif /* !__GENERATING_BOUNDS.H */
194 #ifdef CONFIG_ZONE_DMA
196 * ZONE_DMA is used when there are devices that are not able
197 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
198 * carve out the portion of memory that is needed for these devices.
199 * The range is arch specific.
204 * ---------------------------
205 * parisc, ia64, sparc <4G
208 * alpha Unlimited or 0-16MB.
210 * i386, x86_64 and multiple other arches
215 #ifdef CONFIG_ZONE_DMA32
217 * x86_64 needs two ZONE_DMAs because it supports devices that are
218 * only able to do DMA to the lower 16M but also 32 bit devices that
219 * can only do DMA areas below 4G.
224 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
225 * performed on pages in ZONE_NORMAL if the DMA devices support
226 * transfers to all addressable memory.
229 #ifdef CONFIG_HIGHMEM
231 * A memory area that is only addressable by the kernel through
232 * mapping portions into its own address space. This is for example
233 * used by i386 to allow the kernel to address the memory beyond
234 * 900MB. The kernel will set up special mappings (page
235 * table entries on i386) for each page that the kernel needs to
244 #ifndef __GENERATING_BOUNDS_H
247 * When a memory allocation must conform to specific limitations (such
248 * as being suitable for DMA) the caller will pass in hints to the
249 * allocator in the gfp_mask, in the zone modifier bits. These bits
250 * are used to select a priority ordered list of memory zones which
251 * match the requested limits. See gfp_zone() in include/linux/gfp.h
255 #define ZONES_SHIFT 0
256 #elif MAX_NR_ZONES <= 2
257 #define ZONES_SHIFT 1
258 #elif MAX_NR_ZONES <= 4
259 #define ZONES_SHIFT 2
261 #error ZONES_SHIFT -- too many zones configured adjust calculation
265 /* Fields commonly accessed by the page allocator */
266 unsigned long pages_min, pages_low, pages_high;
268 * We don't know if the memory that we're going to allocate will be freeable
269 * or/and it will be released eventually, so to avoid totally wasting several
270 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
271 * to run OOM on the lower zones despite there's tons of freeable ram
272 * on the higher zones). This array is recalculated at runtime if the
273 * sysctl_lowmem_reserve_ratio sysctl changes.
275 unsigned long lowmem_reserve[MAX_NR_ZONES];
280 * zone reclaim becomes active if more unmapped pages exist.
282 unsigned long min_unmapped_pages;
283 unsigned long min_slab_pages;
284 struct per_cpu_pageset *pageset[NR_CPUS];
286 struct per_cpu_pageset pageset[NR_CPUS];
289 * free areas of different sizes
292 #ifdef CONFIG_MEMORY_HOTPLUG
293 /* see spanned/present_pages for more description */
294 seqlock_t span_seqlock;
296 struct free_area free_area[MAX_ORDER];
298 #ifndef CONFIG_SPARSEMEM
300 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
301 * In SPARSEMEM, this map is stored in struct mem_section
303 unsigned long *pageblock_flags;
304 #endif /* CONFIG_SPARSEMEM */
309 /* Fields commonly accessed by the page reclaim scanner */
312 struct list_head list;
313 unsigned long nr_scan;
317 * The pageout code in vmscan.c keeps track of how many of the
318 * mem/swap backed and file backed pages are refeferenced.
319 * The higher the rotated/scanned ratio, the more valuable
322 * The anon LRU stats live in [0], file LRU stats in [1]
324 unsigned long recent_rotated[2];
325 unsigned long recent_scanned[2];
327 unsigned long pages_scanned; /* since last reclaim */
328 unsigned long flags; /* zone flags, see below */
330 /* Zone statistics */
331 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
334 * prev_priority holds the scanning priority for this zone. It is
335 * defined as the scanning priority at which we achieved our reclaim
336 * target at the previous try_to_free_pages() or balance_pgdat()
339 * We use prev_priority as a measure of how much stress page reclaim is
340 * under - it drives the swappiness decision: whether to unmap mapped
343 * Access to both this field is quite racy even on uniprocessor. But
344 * it is expected to average out OK.
349 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
350 * this zone's LRU. Maintained by the pageout code.
352 unsigned int inactive_ratio;
356 /* Rarely used or read-mostly fields */
359 * wait_table -- the array holding the hash table
360 * wait_table_hash_nr_entries -- the size of the hash table array
361 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
363 * The purpose of all these is to keep track of the people
364 * waiting for a page to become available and make them
365 * runnable again when possible. The trouble is that this
366 * consumes a lot of space, especially when so few things
367 * wait on pages at a given time. So instead of using
368 * per-page waitqueues, we use a waitqueue hash table.
370 * The bucket discipline is to sleep on the same queue when
371 * colliding and wake all in that wait queue when removing.
372 * When something wakes, it must check to be sure its page is
373 * truly available, a la thundering herd. The cost of a
374 * collision is great, but given the expected load of the
375 * table, they should be so rare as to be outweighed by the
376 * benefits from the saved space.
378 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
379 * primary users of these fields, and in mm/page_alloc.c
380 * free_area_init_core() performs the initialization of them.
382 wait_queue_head_t * wait_table;
383 unsigned long wait_table_hash_nr_entries;
384 unsigned long wait_table_bits;
387 * Discontig memory support fields.
389 struct pglist_data *zone_pgdat;
390 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
391 unsigned long zone_start_pfn;
394 * zone_start_pfn, spanned_pages and present_pages are all
395 * protected by span_seqlock. It is a seqlock because it has
396 * to be read outside of zone->lock, and it is done in the main
397 * allocator path. But, it is written quite infrequently.
399 * The lock is declared along with zone->lock because it is
400 * frequently read in proximity to zone->lock. It's good to
401 * give them a chance of being in the same cacheline.
403 unsigned long spanned_pages; /* total size, including holes */
404 unsigned long present_pages; /* amount of memory (excluding holes) */
407 * rarely used fields:
410 } ____cacheline_internodealigned_in_smp;
413 ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */
414 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
415 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
418 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
420 set_bit(flag, &zone->flags);
423 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
425 return test_and_set_bit(flag, &zone->flags);
428 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
430 clear_bit(flag, &zone->flags);
433 static inline int zone_is_all_unreclaimable(const struct zone *zone)
435 return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags);
438 static inline int zone_is_reclaim_locked(const struct zone *zone)
440 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
443 static inline int zone_is_oom_locked(const struct zone *zone)
445 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
449 * The "priority" of VM scanning is how much of the queues we will scan in one
450 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
451 * queues ("queue_length >> 12") during an aging round.
453 #define DEF_PRIORITY 12
455 /* Maximum number of zones on a zonelist */
456 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
461 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
462 * allocations to a single node for GFP_THISNODE.
464 * [0] : Zonelist with fallback
465 * [1] : No fallback (GFP_THISNODE)
467 #define MAX_ZONELISTS 2
471 * We cache key information from each zonelist for smaller cache
472 * footprint when scanning for free pages in get_page_from_freelist().
474 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
475 * up short of free memory since the last time (last_fullzone_zap)
476 * we zero'd fullzones.
477 * 2) The array z_to_n[] maps each zone in the zonelist to its node
478 * id, so that we can efficiently evaluate whether that node is
479 * set in the current tasks mems_allowed.
481 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
482 * indexed by a zones offset in the zonelist zones[] array.
484 * The get_page_from_freelist() routine does two scans. During the
485 * first scan, we skip zones whose corresponding bit in 'fullzones'
486 * is set or whose corresponding node in current->mems_allowed (which
487 * comes from cpusets) is not set. During the second scan, we bypass
488 * this zonelist_cache, to ensure we look methodically at each zone.
490 * Once per second, we zero out (zap) fullzones, forcing us to
491 * reconsider nodes that might have regained more free memory.
492 * The field last_full_zap is the time we last zapped fullzones.
494 * This mechanism reduces the amount of time we waste repeatedly
495 * reexaming zones for free memory when they just came up low on
496 * memory momentarilly ago.
498 * The zonelist_cache struct members logically belong in struct
499 * zonelist. However, the mempolicy zonelists constructed for
500 * MPOL_BIND are intentionally variable length (and usually much
501 * shorter). A general purpose mechanism for handling structs with
502 * multiple variable length members is more mechanism than we want
503 * here. We resort to some special case hackery instead.
505 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
506 * part because they are shorter), so we put the fixed length stuff
507 * at the front of the zonelist struct, ending in a variable length
508 * zones[], as is needed by MPOL_BIND.
510 * Then we put the optional zonelist cache on the end of the zonelist
511 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
512 * the fixed length portion at the front of the struct. This pointer
513 * both enables us to find the zonelist cache, and in the case of
514 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
515 * to know that the zonelist cache is not there.
517 * The end result is that struct zonelists come in two flavors:
518 * 1) The full, fixed length version, shown below, and
519 * 2) The custom zonelists for MPOL_BIND.
520 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
522 * Even though there may be multiple CPU cores on a node modifying
523 * fullzones or last_full_zap in the same zonelist_cache at the same
524 * time, we don't lock it. This is just hint data - if it is wrong now
525 * and then, the allocator will still function, perhaps a bit slower.
529 struct zonelist_cache {
530 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
531 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
532 unsigned long last_full_zap; /* when last zap'd (jiffies) */
535 #define MAX_ZONELISTS 1
536 struct zonelist_cache;
540 * This struct contains information about a zone in a zonelist. It is stored
541 * here to avoid dereferences into large structures and lookups of tables
544 struct zone *zone; /* Pointer to actual zone */
545 int zone_idx; /* zone_idx(zoneref->zone) */
549 * One allocation request operates on a zonelist. A zonelist
550 * is a list of zones, the first one is the 'goal' of the
551 * allocation, the other zones are fallback zones, in decreasing
554 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
555 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
557 * To speed the reading of the zonelist, the zonerefs contain the zone index
558 * of the entry being read. Helper functions to access information given
559 * a struct zoneref are
561 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
562 * zonelist_zone_idx() - Return the index of the zone for an entry
563 * zonelist_node_idx() - Return the index of the node for an entry
566 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
567 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
569 struct zonelist_cache zlcache; // optional ...
573 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
574 struct node_active_region {
575 unsigned long start_pfn;
576 unsigned long end_pfn;
579 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
581 #ifndef CONFIG_DISCONTIGMEM
582 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
583 extern struct page *mem_map;
587 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
588 * (mostly NUMA machines?) to denote a higher-level memory zone than the
591 * On NUMA machines, each NUMA node would have a pg_data_t to describe
592 * it's memory layout.
594 * Memory statistics and page replacement data structures are maintained on a
598 typedef struct pglist_data {
599 struct zone node_zones[MAX_NR_ZONES];
600 struct zonelist node_zonelists[MAX_ZONELISTS];
602 #ifdef CONFIG_FLAT_NODE_MEM_MAP
603 struct page *node_mem_map;
605 struct bootmem_data *bdata;
606 #ifdef CONFIG_MEMORY_HOTPLUG
608 * Must be held any time you expect node_start_pfn, node_present_pages
609 * or node_spanned_pages stay constant. Holding this will also
610 * guarantee that any pfn_valid() stays that way.
612 * Nests above zone->lock and zone->size_seqlock.
614 spinlock_t node_size_lock;
616 unsigned long node_start_pfn;
617 unsigned long node_present_pages; /* total number of physical pages */
618 unsigned long node_spanned_pages; /* total size of physical page
619 range, including holes */
621 wait_queue_head_t kswapd_wait;
622 struct task_struct *kswapd;
623 int kswapd_max_order;
626 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
627 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
628 #ifdef CONFIG_FLAT_NODE_MEM_MAP
629 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
631 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
633 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
635 #include <linux/memory_hotplug.h>
637 void get_zone_counts(unsigned long *active, unsigned long *inactive,
638 unsigned long *free);
639 void build_all_zonelists(void);
640 void wakeup_kswapd(struct zone *zone, int order);
641 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
642 int classzone_idx, int alloc_flags);
643 enum memmap_context {
647 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
649 enum memmap_context context);
651 #ifdef CONFIG_HAVE_MEMORY_PRESENT
652 void memory_present(int nid, unsigned long start, unsigned long end);
654 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
657 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
658 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
662 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
664 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
666 static inline int populated_zone(struct zone *zone)
668 return (!!zone->present_pages);
671 extern int movable_zone;
673 static inline int zone_movable_is_highmem(void)
675 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
676 return movable_zone == ZONE_HIGHMEM;
682 static inline int is_highmem_idx(enum zone_type idx)
684 #ifdef CONFIG_HIGHMEM
685 return (idx == ZONE_HIGHMEM ||
686 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
692 static inline int is_normal_idx(enum zone_type idx)
694 return (idx == ZONE_NORMAL);
698 * is_highmem - helper function to quickly check if a struct zone is a
699 * highmem zone or not. This is an attempt to keep references
700 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
701 * @zone - pointer to struct zone variable
703 static inline int is_highmem(struct zone *zone)
705 #ifdef CONFIG_HIGHMEM
706 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
707 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
708 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
709 zone_movable_is_highmem());
715 static inline int is_normal(struct zone *zone)
717 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
720 static inline int is_dma32(struct zone *zone)
722 #ifdef CONFIG_ZONE_DMA32
723 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
729 static inline int is_dma(struct zone *zone)
731 #ifdef CONFIG_ZONE_DMA
732 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
738 /* These two functions are used to setup the per zone pages min values */
741 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
742 void __user *, size_t *, loff_t *);
743 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
744 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
745 void __user *, size_t *, loff_t *);
746 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
747 void __user *, size_t *, loff_t *);
748 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
749 struct file *, void __user *, size_t *, loff_t *);
750 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
751 struct file *, void __user *, size_t *, loff_t *);
753 extern int numa_zonelist_order_handler(struct ctl_table *, int,
754 struct file *, void __user *, size_t *, loff_t *);
755 extern char numa_zonelist_order[];
756 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
758 #include <linux/topology.h>
759 /* Returns the number of the current Node. */
761 #define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
764 #ifndef CONFIG_NEED_MULTIPLE_NODES
766 extern struct pglist_data contig_page_data;
767 #define NODE_DATA(nid) (&contig_page_data)
768 #define NODE_MEM_MAP(nid) mem_map
770 #else /* CONFIG_NEED_MULTIPLE_NODES */
772 #include <asm/mmzone.h>
774 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
776 extern struct pglist_data *first_online_pgdat(void);
777 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
778 extern struct zone *next_zone(struct zone *zone);
781 * for_each_online_pgdat - helper macro to iterate over all online nodes
782 * @pgdat - pointer to a pg_data_t variable
784 #define for_each_online_pgdat(pgdat) \
785 for (pgdat = first_online_pgdat(); \
787 pgdat = next_online_pgdat(pgdat))
789 * for_each_zone - helper macro to iterate over all memory zones
790 * @zone - pointer to struct zone variable
792 * The user only needs to declare the zone variable, for_each_zone
795 #define for_each_zone(zone) \
796 for (zone = (first_online_pgdat())->node_zones; \
798 zone = next_zone(zone))
800 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
802 return zoneref->zone;
805 static inline int zonelist_zone_idx(struct zoneref *zoneref)
807 return zoneref->zone_idx;
810 static inline int zonelist_node_idx(struct zoneref *zoneref)
813 /* zone_to_nid not available in this context */
814 return zoneref->zone->node;
817 #endif /* CONFIG_NUMA */
821 * 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
822 * @z - The cursor used as a starting point for the search
823 * @highest_zoneidx - The zone index of the highest zone to return
824 * @nodes - An optional nodemask to filter the zonelist with
825 * @zone - The first suitable zone found is returned via this parameter
827 * This function returns the next zone at or below a given zone index that is
828 * within the allowed nodemask using a cursor as the starting point for the
829 * search. The zoneref returned is a cursor that represents the current zone
830 * being examined. It should be advanced by one before calling
831 * next_zones_zonelist again.
833 struct zoneref *next_zones_zonelist(struct zoneref *z,
834 enum zone_type highest_zoneidx,
839 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
840 * @zonelist - The zonelist to search for a suitable zone
841 * @highest_zoneidx - The zone index of the highest zone to return
842 * @nodes - An optional nodemask to filter the zonelist with
843 * @zone - The first suitable zone found is returned via this parameter
845 * This function returns the first zone at or below a given zone index that is
846 * within the allowed nodemask. The zoneref returned is a cursor that can be
847 * used to iterate the zonelist with next_zones_zonelist by advancing it by
848 * one before calling.
850 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
851 enum zone_type highest_zoneidx,
855 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
860 * 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
861 * @zone - The current zone in the iterator
862 * @z - The current pointer within zonelist->zones being iterated
863 * @zlist - The zonelist being iterated
864 * @highidx - The zone index of the highest zone to return
865 * @nodemask - Nodemask allowed by the allocator
867 * This iterator iterates though all zones at or below a given zone index and
868 * within a given nodemask
870 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
871 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
873 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
876 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
877 * @zone - The current zone in the iterator
878 * @z - The current pointer within zonelist->zones being iterated
879 * @zlist - The zonelist being iterated
880 * @highidx - The zone index of the highest zone to return
882 * This iterator iterates though all zones at or below a given zone index.
884 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
885 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
887 #ifdef CONFIG_SPARSEMEM
888 #include <asm/sparsemem.h>
891 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
892 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
893 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
899 #ifdef CONFIG_FLATMEM
900 #define pfn_to_nid(pfn) (0)
903 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
904 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
906 #ifdef CONFIG_SPARSEMEM
909 * SECTION_SHIFT #bits space required to store a section #
911 * PA_SECTION_SHIFT physical address to/from section number
912 * PFN_SECTION_SHIFT pfn to/from section number
914 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
916 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
917 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
919 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
921 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
922 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
924 #define SECTION_BLOCKFLAGS_BITS \
925 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
927 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
928 #error Allocator MAX_ORDER exceeds SECTION_SIZE
934 * This is, logically, a pointer to an array of struct
935 * pages. However, it is stored with some other magic.
936 * (see sparse.c::sparse_init_one_section())
938 * Additionally during early boot we encode node id of
939 * the location of the section here to guide allocation.
940 * (see sparse.c::memory_present())
942 * Making it a UL at least makes someone do a cast
943 * before using it wrong.
945 unsigned long section_mem_map;
947 /* See declaration of similar field in struct zone */
948 unsigned long *pageblock_flags;
951 #ifdef CONFIG_SPARSEMEM_EXTREME
952 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
954 #define SECTIONS_PER_ROOT 1
957 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
958 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
959 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
961 #ifdef CONFIG_SPARSEMEM_EXTREME
962 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
964 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
967 static inline struct mem_section *__nr_to_section(unsigned long nr)
969 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
971 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
973 extern int __section_nr(struct mem_section* ms);
974 extern unsigned long usemap_size(void);
977 * We use the lower bits of the mem_map pointer to store
978 * a little bit of information. There should be at least
979 * 3 bits here due to 32-bit alignment.
981 #define SECTION_MARKED_PRESENT (1UL<<0)
982 #define SECTION_HAS_MEM_MAP (1UL<<1)
983 #define SECTION_MAP_LAST_BIT (1UL<<2)
984 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
985 #define SECTION_NID_SHIFT 2
987 static inline struct page *__section_mem_map_addr(struct mem_section *section)
989 unsigned long map = section->section_mem_map;
990 map &= SECTION_MAP_MASK;
991 return (struct page *)map;
994 static inline int present_section(struct mem_section *section)
996 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
999 static inline int present_section_nr(unsigned long nr)
1001 return present_section(__nr_to_section(nr));
1004 static inline int valid_section(struct mem_section *section)
1006 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1009 static inline int valid_section_nr(unsigned long nr)
1011 return valid_section(__nr_to_section(nr));
1014 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1016 return __nr_to_section(pfn_to_section_nr(pfn));
1019 static inline int pfn_valid(unsigned long pfn)
1021 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1023 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1026 static inline int pfn_present(unsigned long pfn)
1028 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1030 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1034 * These are _only_ used during initialisation, therefore they
1035 * can use __initdata ... They could have names to indicate
1039 #define pfn_to_nid(pfn) \
1041 unsigned long __pfn_to_nid_pfn = (pfn); \
1042 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1045 #define pfn_to_nid(pfn) (0)
1048 #define early_pfn_valid(pfn) pfn_valid(pfn)
1049 void sparse_init(void);
1051 #define sparse_init() do {} while (0)
1052 #define sparse_index_init(_sec, _nid) do {} while (0)
1053 #endif /* CONFIG_SPARSEMEM */
1055 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1056 #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
1058 #define early_pfn_in_nid(pfn, nid) (1)
1061 #ifndef early_pfn_valid
1062 #define early_pfn_valid(pfn) (1)
1065 void memory_present(int nid, unsigned long start, unsigned long end);
1066 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1069 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1070 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1071 * pfn_valid_within() should be used in this case; we optimise this away
1072 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1074 #ifdef CONFIG_HOLES_IN_ZONE
1075 #define pfn_valid_within(pfn) pfn_valid(pfn)
1077 #define pfn_valid_within(pfn) (1)
1080 #endif /* !__GENERATING_BOUNDS.H */
1081 #endif /* !__ASSEMBLY__ */
1082 #endif /* _LINUX_MMZONE_H */