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];
189 #endif /* !__GENERATING_BOUNDS.H */
192 #ifdef CONFIG_ZONE_DMA
194 * ZONE_DMA is used when there are devices that are not able
195 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
196 * carve out the portion of memory that is needed for these devices.
197 * The range is arch specific.
202 * ---------------------------
203 * parisc, ia64, sparc <4G
206 * alpha Unlimited or 0-16MB.
208 * i386, x86_64 and multiple other arches
213 #ifdef CONFIG_ZONE_DMA32
215 * x86_64 needs two ZONE_DMAs because it supports devices that are
216 * only able to do DMA to the lower 16M but also 32 bit devices that
217 * can only do DMA areas below 4G.
222 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
223 * performed on pages in ZONE_NORMAL if the DMA devices support
224 * transfers to all addressable memory.
227 #ifdef CONFIG_HIGHMEM
229 * A memory area that is only addressable by the kernel through
230 * mapping portions into its own address space. This is for example
231 * used by i386 to allow the kernel to address the memory beyond
232 * 900MB. The kernel will set up special mappings (page
233 * table entries on i386) for each page that the kernel needs to
242 #ifndef __GENERATING_BOUNDS_H
245 * When a memory allocation must conform to specific limitations (such
246 * as being suitable for DMA) the caller will pass in hints to the
247 * allocator in the gfp_mask, in the zone modifier bits. These bits
248 * are used to select a priority ordered list of memory zones which
249 * match the requested limits. See gfp_zone() in include/linux/gfp.h
253 #define ZONES_SHIFT 0
254 #elif MAX_NR_ZONES <= 2
255 #define ZONES_SHIFT 1
256 #elif MAX_NR_ZONES <= 4
257 #define ZONES_SHIFT 2
259 #error ZONES_SHIFT -- too many zones configured adjust calculation
262 struct zone_reclaim_stat {
264 * The pageout code in vmscan.c keeps track of how many of the
265 * mem/swap backed and file backed pages are refeferenced.
266 * The higher the rotated/scanned ratio, the more valuable
269 * The anon LRU stats live in [0], file LRU stats in [1]
271 unsigned long recent_rotated[2];
272 unsigned long recent_scanned[2];
275 * accumulated for batching
277 unsigned long nr_saved_scan[NR_LRU_LISTS];
281 /* Fields commonly accessed by the page allocator */
283 /* zone watermarks, access with *_wmark_pages(zone) macros */
284 unsigned long watermark[NR_WMARK];
287 * We don't know if the memory that we're going to allocate will be freeable
288 * or/and it will be released eventually, so to avoid totally wasting several
289 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
290 * to run OOM on the lower zones despite there's tons of freeable ram
291 * on the higher zones). This array is recalculated at runtime if the
292 * sysctl_lowmem_reserve_ratio sysctl changes.
294 unsigned long lowmem_reserve[MAX_NR_ZONES];
299 * zone reclaim becomes active if more unmapped pages exist.
301 unsigned long min_unmapped_pages;
302 unsigned long min_slab_pages;
304 struct per_cpu_pageset __percpu *pageset;
306 * free areas of different sizes
309 int all_unreclaimable; /* All pages pinned */
310 #ifdef CONFIG_MEMORY_HOTPLUG
311 /* see spanned/present_pages for more description */
312 seqlock_t span_seqlock;
314 struct free_area free_area[MAX_ORDER];
316 #ifndef CONFIG_SPARSEMEM
318 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
319 * In SPARSEMEM, this map is stored in struct mem_section
321 unsigned long *pageblock_flags;
322 #endif /* CONFIG_SPARSEMEM */
324 #ifdef CONFIG_COMPACTION
326 * On compaction failure, 1<<compact_defer_shift compactions
327 * are skipped before trying again. The number attempted since
328 * last failure is tracked with compact_considered.
330 unsigned int compact_considered;
331 unsigned int compact_defer_shift;
336 /* Fields commonly accessed by the page reclaim scanner */
339 struct list_head list;
342 struct zone_reclaim_stat reclaim_stat;
344 unsigned long pages_scanned; /* since last reclaim */
345 unsigned long flags; /* zone flags, see below */
347 /* Zone statistics */
348 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
351 * prev_priority holds the scanning priority for this zone. It is
352 * defined as the scanning priority at which we achieved our reclaim
353 * target at the previous try_to_free_pages() or balance_pgdat()
356 * We use prev_priority as a measure of how much stress page reclaim is
357 * under - it drives the swappiness decision: whether to unmap mapped
360 * Access to both this field is quite racy even on uniprocessor. But
361 * it is expected to average out OK.
366 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
367 * this zone's LRU. Maintained by the pageout code.
369 unsigned int inactive_ratio;
373 /* Rarely used or read-mostly fields */
376 * wait_table -- the array holding the hash table
377 * wait_table_hash_nr_entries -- the size of the hash table array
378 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
380 * The purpose of all these is to keep track of the people
381 * waiting for a page to become available and make them
382 * runnable again when possible. The trouble is that this
383 * consumes a lot of space, especially when so few things
384 * wait on pages at a given time. So instead of using
385 * per-page waitqueues, we use a waitqueue hash table.
387 * The bucket discipline is to sleep on the same queue when
388 * colliding and wake all in that wait queue when removing.
389 * When something wakes, it must check to be sure its page is
390 * truly available, a la thundering herd. The cost of a
391 * collision is great, but given the expected load of the
392 * table, they should be so rare as to be outweighed by the
393 * benefits from the saved space.
395 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
396 * primary users of these fields, and in mm/page_alloc.c
397 * free_area_init_core() performs the initialization of them.
399 wait_queue_head_t * wait_table;
400 unsigned long wait_table_hash_nr_entries;
401 unsigned long wait_table_bits;
404 * Discontig memory support fields.
406 struct pglist_data *zone_pgdat;
407 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
408 unsigned long zone_start_pfn;
411 * zone_start_pfn, spanned_pages and present_pages are all
412 * protected by span_seqlock. It is a seqlock because it has
413 * to be read outside of zone->lock, and it is done in the main
414 * allocator path. But, it is written quite infrequently.
416 * The lock is declared along with zone->lock because it is
417 * frequently read in proximity to zone->lock. It's good to
418 * give them a chance of being in the same cacheline.
420 unsigned long spanned_pages; /* total size, including holes */
421 unsigned long present_pages; /* amount of memory (excluding holes) */
424 * rarely used fields:
427 } ____cacheline_internodealigned_in_smp;
430 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
431 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
434 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
436 set_bit(flag, &zone->flags);
439 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
441 return test_and_set_bit(flag, &zone->flags);
444 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
446 clear_bit(flag, &zone->flags);
449 static inline int zone_is_reclaim_locked(const struct zone *zone)
451 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
454 static inline int zone_is_oom_locked(const struct zone *zone)
456 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
460 * The "priority" of VM scanning is how much of the queues we will scan in one
461 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
462 * queues ("queue_length >> 12") during an aging round.
464 #define DEF_PRIORITY 12
466 /* Maximum number of zones on a zonelist */
467 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
472 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
473 * allocations to a single node for GFP_THISNODE.
475 * [0] : Zonelist with fallback
476 * [1] : No fallback (GFP_THISNODE)
478 #define MAX_ZONELISTS 2
482 * We cache key information from each zonelist for smaller cache
483 * footprint when scanning for free pages in get_page_from_freelist().
485 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
486 * up short of free memory since the last time (last_fullzone_zap)
487 * we zero'd fullzones.
488 * 2) The array z_to_n[] maps each zone in the zonelist to its node
489 * id, so that we can efficiently evaluate whether that node is
490 * set in the current tasks mems_allowed.
492 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
493 * indexed by a zones offset in the zonelist zones[] array.
495 * The get_page_from_freelist() routine does two scans. During the
496 * first scan, we skip zones whose corresponding bit in 'fullzones'
497 * is set or whose corresponding node in current->mems_allowed (which
498 * comes from cpusets) is not set. During the second scan, we bypass
499 * this zonelist_cache, to ensure we look methodically at each zone.
501 * Once per second, we zero out (zap) fullzones, forcing us to
502 * reconsider nodes that might have regained more free memory.
503 * The field last_full_zap is the time we last zapped fullzones.
505 * This mechanism reduces the amount of time we waste repeatedly
506 * reexaming zones for free memory when they just came up low on
507 * memory momentarilly ago.
509 * The zonelist_cache struct members logically belong in struct
510 * zonelist. However, the mempolicy zonelists constructed for
511 * MPOL_BIND are intentionally variable length (and usually much
512 * shorter). A general purpose mechanism for handling structs with
513 * multiple variable length members is more mechanism than we want
514 * here. We resort to some special case hackery instead.
516 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
517 * part because they are shorter), so we put the fixed length stuff
518 * at the front of the zonelist struct, ending in a variable length
519 * zones[], as is needed by MPOL_BIND.
521 * Then we put the optional zonelist cache on the end of the zonelist
522 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
523 * the fixed length portion at the front of the struct. This pointer
524 * both enables us to find the zonelist cache, and in the case of
525 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
526 * to know that the zonelist cache is not there.
528 * The end result is that struct zonelists come in two flavors:
529 * 1) The full, fixed length version, shown below, and
530 * 2) The custom zonelists for MPOL_BIND.
531 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
533 * Even though there may be multiple CPU cores on a node modifying
534 * fullzones or last_full_zap in the same zonelist_cache at the same
535 * time, we don't lock it. This is just hint data - if it is wrong now
536 * and then, the allocator will still function, perhaps a bit slower.
540 struct zonelist_cache {
541 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
542 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
543 unsigned long last_full_zap; /* when last zap'd (jiffies) */
546 #define MAX_ZONELISTS 1
547 struct zonelist_cache;
551 * This struct contains information about a zone in a zonelist. It is stored
552 * here to avoid dereferences into large structures and lookups of tables
555 struct zone *zone; /* Pointer to actual zone */
556 int zone_idx; /* zone_idx(zoneref->zone) */
560 * One allocation request operates on a zonelist. A zonelist
561 * is a list of zones, the first one is the 'goal' of the
562 * allocation, the other zones are fallback zones, in decreasing
565 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
566 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
568 * To speed the reading of the zonelist, the zonerefs contain the zone index
569 * of the entry being read. Helper functions to access information given
570 * a struct zoneref are
572 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
573 * zonelist_zone_idx() - Return the index of the zone for an entry
574 * zonelist_node_idx() - Return the index of the node for an entry
577 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
578 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
580 struct zonelist_cache zlcache; // optional ...
584 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
585 struct node_active_region {
586 unsigned long start_pfn;
587 unsigned long end_pfn;
590 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
592 #ifndef CONFIG_DISCONTIGMEM
593 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
594 extern struct page *mem_map;
598 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
599 * (mostly NUMA machines?) to denote a higher-level memory zone than the
602 * On NUMA machines, each NUMA node would have a pg_data_t to describe
603 * it's memory layout.
605 * Memory statistics and page replacement data structures are maintained on a
609 typedef struct pglist_data {
610 struct zone node_zones[MAX_NR_ZONES];
611 struct zonelist node_zonelists[MAX_ZONELISTS];
613 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
614 struct page *node_mem_map;
615 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
616 struct page_cgroup *node_page_cgroup;
619 #ifndef CONFIG_NO_BOOTMEM
620 struct bootmem_data *bdata;
622 #ifdef CONFIG_MEMORY_HOTPLUG
624 * Must be held any time you expect node_start_pfn, node_present_pages
625 * or node_spanned_pages stay constant. Holding this will also
626 * guarantee that any pfn_valid() stays that way.
628 * Nests above zone->lock and zone->size_seqlock.
630 spinlock_t node_size_lock;
632 unsigned long node_start_pfn;
633 unsigned long node_present_pages; /* total number of physical pages */
634 unsigned long node_spanned_pages; /* total size of physical page
635 range, including holes */
637 wait_queue_head_t kswapd_wait;
638 struct task_struct *kswapd;
639 int kswapd_max_order;
642 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
643 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
644 #ifdef CONFIG_FLAT_NODE_MEM_MAP
645 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
647 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
649 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
651 #include <linux/memory_hotplug.h>
653 void get_zone_counts(unsigned long *active, unsigned long *inactive,
654 unsigned long *free);
655 void build_all_zonelists(void *data);
656 void wakeup_kswapd(struct zone *zone, int order);
657 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
658 int classzone_idx, int alloc_flags);
659 enum memmap_context {
663 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
665 enum memmap_context context);
667 #ifdef CONFIG_HAVE_MEMORY_PRESENT
668 void memory_present(int nid, unsigned long start, unsigned long end);
670 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
673 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
674 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
678 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
680 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
682 static inline int populated_zone(struct zone *zone)
684 return (!!zone->present_pages);
687 extern int movable_zone;
689 static inline int zone_movable_is_highmem(void)
691 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
692 return movable_zone == ZONE_HIGHMEM;
698 static inline int is_highmem_idx(enum zone_type idx)
700 #ifdef CONFIG_HIGHMEM
701 return (idx == ZONE_HIGHMEM ||
702 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
708 static inline int is_normal_idx(enum zone_type idx)
710 return (idx == ZONE_NORMAL);
714 * is_highmem - helper function to quickly check if a struct zone is a
715 * highmem zone or not. This is an attempt to keep references
716 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
717 * @zone - pointer to struct zone variable
719 static inline int is_highmem(struct zone *zone)
721 #ifdef CONFIG_HIGHMEM
722 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
723 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
724 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
725 zone_movable_is_highmem());
731 static inline int is_normal(struct zone *zone)
733 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
736 static inline int is_dma32(struct zone *zone)
738 #ifdef CONFIG_ZONE_DMA32
739 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
745 static inline int is_dma(struct zone *zone)
747 #ifdef CONFIG_ZONE_DMA
748 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
754 /* These two functions are used to setup the per zone pages min values */
756 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
757 void __user *, size_t *, loff_t *);
758 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
759 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
760 void __user *, size_t *, loff_t *);
761 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
762 void __user *, size_t *, loff_t *);
763 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
764 void __user *, size_t *, loff_t *);
765 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
766 void __user *, size_t *, loff_t *);
768 extern int numa_zonelist_order_handler(struct ctl_table *, int,
769 void __user *, size_t *, loff_t *);
770 extern char numa_zonelist_order[];
771 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
773 #ifndef CONFIG_NEED_MULTIPLE_NODES
775 extern struct pglist_data contig_page_data;
776 #define NODE_DATA(nid) (&contig_page_data)
777 #define NODE_MEM_MAP(nid) mem_map
779 #else /* CONFIG_NEED_MULTIPLE_NODES */
781 #include <asm/mmzone.h>
783 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
785 extern struct pglist_data *first_online_pgdat(void);
786 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
787 extern struct zone *next_zone(struct zone *zone);
790 * for_each_online_pgdat - helper macro to iterate over all online nodes
791 * @pgdat - pointer to a pg_data_t variable
793 #define for_each_online_pgdat(pgdat) \
794 for (pgdat = first_online_pgdat(); \
796 pgdat = next_online_pgdat(pgdat))
798 * for_each_zone - helper macro to iterate over all memory zones
799 * @zone - pointer to struct zone variable
801 * The user only needs to declare the zone variable, for_each_zone
804 #define for_each_zone(zone) \
805 for (zone = (first_online_pgdat())->node_zones; \
807 zone = next_zone(zone))
809 #define for_each_populated_zone(zone) \
810 for (zone = (first_online_pgdat())->node_zones; \
812 zone = next_zone(zone)) \
813 if (!populated_zone(zone)) \
817 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
819 return zoneref->zone;
822 static inline int zonelist_zone_idx(struct zoneref *zoneref)
824 return zoneref->zone_idx;
827 static inline int zonelist_node_idx(struct zoneref *zoneref)
830 /* zone_to_nid not available in this context */
831 return zoneref->zone->node;
834 #endif /* CONFIG_NUMA */
838 * 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
839 * @z - The cursor used as a starting point for the search
840 * @highest_zoneidx - The zone index of the highest zone to return
841 * @nodes - An optional nodemask to filter the zonelist with
842 * @zone - The first suitable zone found is returned via this parameter
844 * This function returns the next zone at or below a given zone index that is
845 * within the allowed nodemask using a cursor as the starting point for the
846 * search. The zoneref returned is a cursor that represents the current zone
847 * being examined. It should be advanced by one before calling
848 * next_zones_zonelist again.
850 struct zoneref *next_zones_zonelist(struct zoneref *z,
851 enum zone_type highest_zoneidx,
856 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
857 * @zonelist - The zonelist to search for a suitable zone
858 * @highest_zoneidx - The zone index of the highest zone to return
859 * @nodes - An optional nodemask to filter the zonelist with
860 * @zone - The first suitable zone found is returned via this parameter
862 * This function returns the first zone at or below a given zone index that is
863 * within the allowed nodemask. The zoneref returned is a cursor that can be
864 * used to iterate the zonelist with next_zones_zonelist by advancing it by
865 * one before calling.
867 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
868 enum zone_type highest_zoneidx,
872 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
877 * 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
878 * @zone - The current zone in the iterator
879 * @z - The current pointer within zonelist->zones being iterated
880 * @zlist - The zonelist being iterated
881 * @highidx - The zone index of the highest zone to return
882 * @nodemask - Nodemask allowed by the allocator
884 * This iterator iterates though all zones at or below a given zone index and
885 * within a given nodemask
887 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
888 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
890 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
893 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
894 * @zone - The current zone in the iterator
895 * @z - The current pointer within zonelist->zones being iterated
896 * @zlist - The zonelist being iterated
897 * @highidx - The zone index of the highest zone to return
899 * This iterator iterates though all zones at or below a given zone index.
901 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
902 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
904 #ifdef CONFIG_SPARSEMEM
905 #include <asm/sparsemem.h>
908 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
909 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
910 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
916 #ifdef CONFIG_FLATMEM
917 #define pfn_to_nid(pfn) (0)
920 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
921 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
923 #ifdef CONFIG_SPARSEMEM
926 * SECTION_SHIFT #bits space required to store a section #
928 * PA_SECTION_SHIFT physical address to/from section number
929 * PFN_SECTION_SHIFT pfn to/from section number
931 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
933 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
934 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
936 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
938 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
939 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
941 #define SECTION_BLOCKFLAGS_BITS \
942 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
944 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
945 #error Allocator MAX_ORDER exceeds SECTION_SIZE
952 * This is, logically, a pointer to an array of struct
953 * pages. However, it is stored with some other magic.
954 * (see sparse.c::sparse_init_one_section())
956 * Additionally during early boot we encode node id of
957 * the location of the section here to guide allocation.
958 * (see sparse.c::memory_present())
960 * Making it a UL at least makes someone do a cast
961 * before using it wrong.
963 unsigned long section_mem_map;
965 /* See declaration of similar field in struct zone */
966 unsigned long *pageblock_flags;
967 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
969 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
970 * section. (see memcontrol.h/page_cgroup.h about this.)
972 struct page_cgroup *page_cgroup;
977 #ifdef CONFIG_SPARSEMEM_EXTREME
978 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
980 #define SECTIONS_PER_ROOT 1
983 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
984 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
985 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
987 #ifdef CONFIG_SPARSEMEM_EXTREME
988 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
990 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
993 static inline struct mem_section *__nr_to_section(unsigned long nr)
995 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
997 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
999 extern int __section_nr(struct mem_section* ms);
1000 extern unsigned long usemap_size(void);
1003 * We use the lower bits of the mem_map pointer to store
1004 * a little bit of information. There should be at least
1005 * 3 bits here due to 32-bit alignment.
1007 #define SECTION_MARKED_PRESENT (1UL<<0)
1008 #define SECTION_HAS_MEM_MAP (1UL<<1)
1009 #define SECTION_MAP_LAST_BIT (1UL<<2)
1010 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1011 #define SECTION_NID_SHIFT 2
1013 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1015 unsigned long map = section->section_mem_map;
1016 map &= SECTION_MAP_MASK;
1017 return (struct page *)map;
1020 static inline int present_section(struct mem_section *section)
1022 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1025 static inline int present_section_nr(unsigned long nr)
1027 return present_section(__nr_to_section(nr));
1030 static inline int valid_section(struct mem_section *section)
1032 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1035 static inline int valid_section_nr(unsigned long nr)
1037 return valid_section(__nr_to_section(nr));
1040 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1042 return __nr_to_section(pfn_to_section_nr(pfn));
1045 static inline int pfn_valid(unsigned long pfn)
1047 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1049 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1052 static inline int pfn_present(unsigned long pfn)
1054 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1056 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1060 * These are _only_ used during initialisation, therefore they
1061 * can use __initdata ... They could have names to indicate
1065 #define pfn_to_nid(pfn) \
1067 unsigned long __pfn_to_nid_pfn = (pfn); \
1068 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1071 #define pfn_to_nid(pfn) (0)
1074 #define early_pfn_valid(pfn) pfn_valid(pfn)
1075 void sparse_init(void);
1077 #define sparse_init() do {} while (0)
1078 #define sparse_index_init(_sec, _nid) do {} while (0)
1079 #endif /* CONFIG_SPARSEMEM */
1081 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1082 bool early_pfn_in_nid(unsigned long pfn, int nid);
1084 #define early_pfn_in_nid(pfn, nid) (1)
1087 #ifndef early_pfn_valid
1088 #define early_pfn_valid(pfn) (1)
1091 void memory_present(int nid, unsigned long start, unsigned long end);
1092 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1095 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1096 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1097 * pfn_valid_within() should be used in this case; we optimise this away
1098 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1100 #ifdef CONFIG_HOLES_IN_ZONE
1101 #define pfn_valid_within(pfn) pfn_valid(pfn)
1103 #define pfn_valid_within(pfn) (1)
1106 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1108 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1109 * associated with it or not. In FLATMEM, it is expected that holes always
1110 * have valid memmap as long as there is valid PFNs either side of the hole.
1111 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1114 * However, an ARM, and maybe other embedded architectures in the future
1115 * free memmap backing holes to save memory on the assumption the memmap is
1116 * never used. The page_zone linkages are then broken even though pfn_valid()
1117 * returns true. A walker of the full memmap must then do this additional
1118 * check to ensure the memmap they are looking at is sane by making sure
1119 * the zone and PFN linkages are still valid. This is expensive, but walkers
1120 * of the full memmap are extremely rare.
1122 int memmap_valid_within(unsigned long pfn,
1123 struct page *page, struct zone *zone);
1125 static inline int memmap_valid_within(unsigned long pfn,
1126 struct page *page, struct zone *zone)
1130 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1132 #endif /* !__GENERATING_BOUNDS.H */
1133 #endif /* !__ASSEMBLY__ */
1134 #endif /* _LINUX_MMZONE_H */