1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
5 #ifndef __GENERATING_BOUNDS_H
7 #include <linux/spinlock.h>
8 #include <linux/list.h>
9 #include <linux/wait.h>
10 #include <linux/bitops.h>
11 #include <linux/cache.h>
12 #include <linux/threads.h>
13 #include <linux/numa.h>
14 #include <linux/init.h>
15 #include <linux/seqlock.h>
16 #include <linux/nodemask.h>
17 #include <linux/pageblock-flags.h>
18 #include <linux/page-flags-layout.h>
19 #include <linux/atomic.h>
22 /* Free memory management - zoned buddy allocator. */
23 #ifndef CONFIG_FORCE_MAX_ZONEORDER
26 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
28 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
31 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
32 * costly to service. That is between allocation orders which should
33 * coalesce naturally under reasonable reclaim pressure and those which
36 #define PAGE_ALLOC_COSTLY_ORDER 3
42 MIGRATE_PCPTYPES, /* the number of types on the pcp lists */
43 MIGRATE_RESERVE = MIGRATE_PCPTYPES,
46 * MIGRATE_CMA migration type is designed to mimic the way
47 * ZONE_MOVABLE works. Only movable pages can be allocated
48 * from MIGRATE_CMA pageblocks and page allocator never
49 * implicitly change migration type of MIGRATE_CMA pageblock.
51 * The way to use it is to change migratetype of a range of
52 * pageblocks to MIGRATE_CMA which can be done by
53 * __free_pageblock_cma() function. What is important though
54 * is that a range of pageblocks must be aligned to
55 * MAX_ORDER_NR_PAGES should biggest page be bigger then
60 MIGRATE_ISOLATE, /* can't allocate from here */
65 # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
66 # define cma_wmark_pages(zone) zone->min_cma_pages
68 # define is_migrate_cma(migratetype) false
69 # define cma_wmark_pages(zone) 0
72 #define for_each_migratetype_order(order, type) \
73 for (order = 0; order < MAX_ORDER; order++) \
74 for (type = 0; type < MIGRATE_TYPES; type++)
76 extern int page_group_by_mobility_disabled;
78 static inline int get_pageblock_migratetype(struct page *page)
80 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
84 struct list_head free_list[MIGRATE_TYPES];
85 unsigned long nr_free;
91 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
92 * So add a wild amount of padding here to ensure that they fall into separate
93 * cachelines. There are very few zone structures in the machine, so space
94 * consumption is not a concern here.
96 #if defined(CONFIG_SMP)
99 } ____cacheline_internodealigned_in_smp;
100 #define ZONE_PADDING(name) struct zone_padding name;
102 #define ZONE_PADDING(name)
105 enum zone_stat_item {
106 /* First 128 byte cacheline (assuming 64 bit words) */
109 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
110 NR_ACTIVE_ANON, /* " " " " " */
111 NR_INACTIVE_FILE, /* " " " " " */
112 NR_ACTIVE_FILE, /* " " " " " */
113 NR_UNEVICTABLE, /* " " " " " */
114 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
115 NR_ANON_PAGES, /* Mapped anonymous pages */
116 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
117 only modified from process context */
122 NR_SLAB_UNRECLAIMABLE,
123 NR_PAGETABLE, /* used for pagetables */
125 /* Second 128 byte cacheline */
126 NR_UNSTABLE_NFS, /* NFS unstable pages */
129 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
130 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
131 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
132 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
133 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
134 NR_DIRTIED, /* page dirtyings since bootup */
135 NR_WRITTEN, /* page writings since bootup */
137 NUMA_HIT, /* allocated in intended node */
138 NUMA_MISS, /* allocated in non intended node */
139 NUMA_FOREIGN, /* was intended here, hit elsewhere */
140 NUMA_LOCAL, /* allocation from local node */
141 NUMA_OTHER, /* allocation from other node */
143 NR_ANON_TRANSPARENT_HUGEPAGES,
145 NR_VM_ZONE_STAT_ITEMS };
148 * We do arithmetic on the LRU lists in various places in the code,
149 * so it is important to keep the active lists LRU_ACTIVE higher in
150 * the array than the corresponding inactive lists, and to keep
151 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
153 * This has to be kept in sync with the statistics in zone_stat_item
154 * above and the descriptions in vmstat_text in mm/vmstat.c
161 LRU_INACTIVE_ANON = LRU_BASE,
162 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
163 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
164 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
169 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
171 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
173 static inline int is_file_lru(enum lru_list lru)
175 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
178 static inline int is_active_lru(enum lru_list lru)
180 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
183 static inline int is_unevictable_lru(enum lru_list lru)
185 return (lru == LRU_UNEVICTABLE);
188 struct zone_reclaim_stat {
190 * The pageout code in vmscan.c keeps track of how many of the
191 * mem/swap backed and file backed pages are referenced.
192 * The higher the rotated/scanned ratio, the more valuable
195 * The anon LRU stats live in [0], file LRU stats in [1]
197 unsigned long recent_rotated[2];
198 unsigned long recent_scanned[2];
202 struct list_head lists[NR_LRU_LISTS];
203 struct zone_reclaim_stat reclaim_stat;
209 /* Mask used at gathering information at once (see memcontrol.c) */
210 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
211 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
212 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
214 /* Isolate clean file */
215 #define ISOLATE_CLEAN ((__force isolate_mode_t)0x1)
216 /* Isolate unmapped file */
217 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
218 /* Isolate for asynchronous migration */
219 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
221 /* LRU Isolation modes. */
222 typedef unsigned __bitwise__ isolate_mode_t;
224 enum zone_watermarks {
231 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
232 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
233 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
235 struct per_cpu_pages {
236 int count; /* number of pages in the list */
237 int high; /* high watermark, emptying needed */
238 int batch; /* chunk size for buddy add/remove */
240 /* Lists of pages, one per migrate type stored on the pcp-lists */
241 struct list_head lists[MIGRATE_PCPTYPES];
244 struct per_cpu_pageset {
245 struct per_cpu_pages pcp;
251 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
255 #endif /* !__GENERATING_BOUNDS.H */
258 #ifdef CONFIG_ZONE_DMA
260 * ZONE_DMA is used when there are devices that are not able
261 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
262 * carve out the portion of memory that is needed for these devices.
263 * The range is arch specific.
268 * ---------------------------
269 * parisc, ia64, sparc <4G
272 * alpha Unlimited or 0-16MB.
274 * i386, x86_64 and multiple other arches
279 #ifdef CONFIG_ZONE_DMA32
281 * x86_64 needs two ZONE_DMAs because it supports devices that are
282 * only able to do DMA to the lower 16M but also 32 bit devices that
283 * can only do DMA areas below 4G.
288 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
289 * performed on pages in ZONE_NORMAL if the DMA devices support
290 * transfers to all addressable memory.
293 #ifdef CONFIG_HIGHMEM
295 * A memory area that is only addressable by the kernel through
296 * mapping portions into its own address space. This is for example
297 * used by i386 to allow the kernel to address the memory beyond
298 * 900MB. The kernel will set up special mappings (page
299 * table entries on i386) for each page that the kernel needs to
308 #ifndef __GENERATING_BOUNDS_H
311 * When a memory allocation must conform to specific limitations (such
312 * as being suitable for DMA) the caller will pass in hints to the
313 * allocator in the gfp_mask, in the zone modifier bits. These bits
314 * are used to select a priority ordered list of memory zones which
315 * match the requested limits. See gfp_zone() in include/linux/gfp.h
319 /* Fields commonly accessed by the page allocator */
321 /* zone watermarks, access with *_wmark_pages(zone) macros */
322 unsigned long watermark[NR_WMARK];
325 * When free pages are below this point, additional steps are taken
326 * when reading the number of free pages to avoid per-cpu counter
327 * drift allowing watermarks to be breached
329 unsigned long percpu_drift_mark;
332 * We don't know if the memory that we're going to allocate will be freeable
333 * or/and it will be released eventually, so to avoid totally wasting several
334 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
335 * to run OOM on the lower zones despite there's tons of freeable ram
336 * on the higher zones). This array is recalculated at runtime if the
337 * sysctl_lowmem_reserve_ratio sysctl changes.
339 unsigned long lowmem_reserve[MAX_NR_ZONES];
342 * This is a per-zone reserve of pages that should not be
343 * considered dirtyable memory.
345 unsigned long dirty_balance_reserve;
350 * zone reclaim becomes active if more unmapped pages exist.
352 unsigned long min_unmapped_pages;
353 unsigned long min_slab_pages;
355 struct per_cpu_pageset __percpu *pageset;
357 * free areas of different sizes
360 int all_unreclaimable; /* All pages pinned */
361 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
362 /* pfn where the last incremental compaction isolated free pages */
363 unsigned long compact_cached_free_pfn;
365 #ifdef CONFIG_MEMORY_HOTPLUG
366 /* see spanned/present_pages for more description */
367 seqlock_t span_seqlock;
371 * CMA needs to increase watermark levels during the allocation
372 * process to make sure that the system is not starved.
374 unsigned long min_cma_pages;
376 struct free_area free_area[MAX_ORDER];
378 #ifndef CONFIG_SPARSEMEM
380 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
381 * In SPARSEMEM, this map is stored in struct mem_section
383 unsigned long *pageblock_flags;
384 #endif /* CONFIG_SPARSEMEM */
386 #ifdef CONFIG_COMPACTION
388 * On compaction failure, 1<<compact_defer_shift compactions
389 * are skipped before trying again. The number attempted since
390 * last failure is tracked with compact_considered.
392 unsigned int compact_considered;
393 unsigned int compact_defer_shift;
394 int compact_order_failed;
399 /* Fields commonly accessed by the page reclaim scanner */
401 struct lruvec lruvec;
403 unsigned long pages_scanned; /* since last reclaim */
404 unsigned long flags; /* zone flags, see below */
406 /* Zone statistics */
407 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
410 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
411 * this zone's LRU. Maintained by the pageout code.
413 unsigned int inactive_ratio;
417 /* Rarely used or read-mostly fields */
420 * wait_table -- the array holding the hash table
421 * wait_table_hash_nr_entries -- the size of the hash table array
422 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
424 * The purpose of all these is to keep track of the people
425 * waiting for a page to become available and make them
426 * runnable again when possible. The trouble is that this
427 * consumes a lot of space, especially when so few things
428 * wait on pages at a given time. So instead of using
429 * per-page waitqueues, we use a waitqueue hash table.
431 * The bucket discipline is to sleep on the same queue when
432 * colliding and wake all in that wait queue when removing.
433 * When something wakes, it must check to be sure its page is
434 * truly available, a la thundering herd. The cost of a
435 * collision is great, but given the expected load of the
436 * table, they should be so rare as to be outweighed by the
437 * benefits from the saved space.
439 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
440 * primary users of these fields, and in mm/page_alloc.c
441 * free_area_init_core() performs the initialization of them.
443 wait_queue_head_t * wait_table;
444 unsigned long wait_table_hash_nr_entries;
445 unsigned long wait_table_bits;
448 * Discontig memory support fields.
450 struct pglist_data *zone_pgdat;
451 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
452 unsigned long zone_start_pfn;
455 * zone_start_pfn, spanned_pages and present_pages are all
456 * protected by span_seqlock. It is a seqlock because it has
457 * to be read outside of zone->lock, and it is done in the main
458 * allocator path. But, it is written quite infrequently.
460 * The lock is declared along with zone->lock because it is
461 * frequently read in proximity to zone->lock. It's good to
462 * give them a chance of being in the same cacheline.
464 unsigned long spanned_pages; /* total size, including holes */
465 unsigned long present_pages; /* amount of memory (excluding holes) */
468 * rarely used fields:
471 #ifdef CONFIG_MEMORY_ISOLATION
473 * the number of MIGRATE_ISOLATE *pageblock*.
474 * We need this for free page counting. Look at zone_watermark_ok_safe.
475 * It's protected by zone->lock
477 int nr_pageblock_isolate;
479 } ____cacheline_internodealigned_in_smp;
482 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
483 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
484 ZONE_CONGESTED, /* zone has many dirty pages backed by
489 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
491 set_bit(flag, &zone->flags);
494 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
496 return test_and_set_bit(flag, &zone->flags);
499 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
501 clear_bit(flag, &zone->flags);
504 static inline int zone_is_reclaim_congested(const struct zone *zone)
506 return test_bit(ZONE_CONGESTED, &zone->flags);
509 static inline int zone_is_reclaim_locked(const struct zone *zone)
511 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
514 static inline int zone_is_oom_locked(const struct zone *zone)
516 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
520 * The "priority" of VM scanning is how much of the queues we will scan in one
521 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
522 * queues ("queue_length >> 12") during an aging round.
524 #define DEF_PRIORITY 12
526 /* Maximum number of zones on a zonelist */
527 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
532 * The NUMA zonelists are doubled because we need zonelists that restrict the
533 * allocations to a single node for GFP_THISNODE.
535 * [0] : Zonelist with fallback
536 * [1] : No fallback (GFP_THISNODE)
538 #define MAX_ZONELISTS 2
542 * We cache key information from each zonelist for smaller cache
543 * footprint when scanning for free pages in get_page_from_freelist().
545 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
546 * up short of free memory since the last time (last_fullzone_zap)
547 * we zero'd fullzones.
548 * 2) The array z_to_n[] maps each zone in the zonelist to its node
549 * id, so that we can efficiently evaluate whether that node is
550 * set in the current tasks mems_allowed.
552 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
553 * indexed by a zones offset in the zonelist zones[] array.
555 * The get_page_from_freelist() routine does two scans. During the
556 * first scan, we skip zones whose corresponding bit in 'fullzones'
557 * is set or whose corresponding node in current->mems_allowed (which
558 * comes from cpusets) is not set. During the second scan, we bypass
559 * this zonelist_cache, to ensure we look methodically at each zone.
561 * Once per second, we zero out (zap) fullzones, forcing us to
562 * reconsider nodes that might have regained more free memory.
563 * The field last_full_zap is the time we last zapped fullzones.
565 * This mechanism reduces the amount of time we waste repeatedly
566 * reexaming zones for free memory when they just came up low on
567 * memory momentarilly ago.
569 * The zonelist_cache struct members logically belong in struct
570 * zonelist. However, the mempolicy zonelists constructed for
571 * MPOL_BIND are intentionally variable length (and usually much
572 * shorter). A general purpose mechanism for handling structs with
573 * multiple variable length members is more mechanism than we want
574 * here. We resort to some special case hackery instead.
576 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
577 * part because they are shorter), so we put the fixed length stuff
578 * at the front of the zonelist struct, ending in a variable length
579 * zones[], as is needed by MPOL_BIND.
581 * Then we put the optional zonelist cache on the end of the zonelist
582 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
583 * the fixed length portion at the front of the struct. This pointer
584 * both enables us to find the zonelist cache, and in the case of
585 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
586 * to know that the zonelist cache is not there.
588 * The end result is that struct zonelists come in two flavors:
589 * 1) The full, fixed length version, shown below, and
590 * 2) The custom zonelists for MPOL_BIND.
591 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
593 * Even though there may be multiple CPU cores on a node modifying
594 * fullzones or last_full_zap in the same zonelist_cache at the same
595 * time, we don't lock it. This is just hint data - if it is wrong now
596 * and then, the allocator will still function, perhaps a bit slower.
600 struct zonelist_cache {
601 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
602 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
603 unsigned long last_full_zap; /* when last zap'd (jiffies) */
606 #define MAX_ZONELISTS 1
607 struct zonelist_cache;
611 * This struct contains information about a zone in a zonelist. It is stored
612 * here to avoid dereferences into large structures and lookups of tables
615 struct zone *zone; /* Pointer to actual zone */
616 int zone_idx; /* zone_idx(zoneref->zone) */
620 * One allocation request operates on a zonelist. A zonelist
621 * is a list of zones, the first one is the 'goal' of the
622 * allocation, the other zones are fallback zones, in decreasing
625 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
626 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
628 * To speed the reading of the zonelist, the zonerefs contain the zone index
629 * of the entry being read. Helper functions to access information given
630 * a struct zoneref are
632 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
633 * zonelist_zone_idx() - Return the index of the zone for an entry
634 * zonelist_node_idx() - Return the index of the node for an entry
637 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
638 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
640 struct zonelist_cache zlcache; // optional ...
644 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
645 struct node_active_region {
646 unsigned long start_pfn;
647 unsigned long end_pfn;
650 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
652 #ifndef CONFIG_DISCONTIGMEM
653 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
654 extern struct page *mem_map;
658 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
659 * (mostly NUMA machines?) to denote a higher-level memory zone than the
662 * On NUMA machines, each NUMA node would have a pg_data_t to describe
663 * it's memory layout.
665 * Memory statistics and page replacement data structures are maintained on a
669 typedef struct pglist_data {
670 struct zone node_zones[MAX_NR_ZONES];
671 struct zonelist node_zonelists[MAX_ZONELISTS];
673 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
674 struct page *node_mem_map;
676 struct page_cgroup *node_page_cgroup;
679 #ifndef CONFIG_NO_BOOTMEM
680 struct bootmem_data *bdata;
682 #ifdef CONFIG_MEMORY_HOTPLUG
684 * Must be held any time you expect node_start_pfn, node_present_pages
685 * or node_spanned_pages stay constant. Holding this will also
686 * guarantee that any pfn_valid() stays that way.
688 * Nests above zone->lock and zone->size_seqlock.
690 spinlock_t node_size_lock;
692 unsigned long node_start_pfn;
693 unsigned long node_present_pages; /* total number of physical pages */
694 unsigned long node_spanned_pages; /* total size of physical page
695 range, including holes */
697 wait_queue_head_t kswapd_wait;
698 wait_queue_head_t pfmemalloc_wait;
699 struct task_struct *kswapd; /* Protected by lock_memory_hotplug() */
700 int kswapd_max_order;
701 enum zone_type classzone_idx;
704 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
705 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
706 #ifdef CONFIG_FLAT_NODE_MEM_MAP
707 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
709 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
711 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
713 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
715 #define node_end_pfn(nid) ({\
716 pg_data_t *__pgdat = NODE_DATA(nid);\
717 __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
720 #include <linux/memory_hotplug.h>
722 extern struct mutex zonelists_mutex;
723 void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
724 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
725 bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
726 int classzone_idx, int alloc_flags);
727 bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
728 int classzone_idx, int alloc_flags);
729 enum memmap_context {
733 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
735 enum memmap_context context);
737 extern void lruvec_init(struct lruvec *lruvec, struct zone *zone);
739 static inline struct zone *lruvec_zone(struct lruvec *lruvec)
744 return container_of(lruvec, struct zone, lruvec);
748 #ifdef CONFIG_HAVE_MEMORY_PRESENT
749 void memory_present(int nid, unsigned long start, unsigned long end);
751 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
754 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
755 int local_memory_node(int node_id);
757 static inline int local_memory_node(int node_id) { return node_id; };
760 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
761 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
765 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
767 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
769 static inline int populated_zone(struct zone *zone)
771 return (!!zone->present_pages);
774 extern int movable_zone;
776 static inline int zone_movable_is_highmem(void)
778 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
779 return movable_zone == ZONE_HIGHMEM;
785 static inline int is_highmem_idx(enum zone_type idx)
787 #ifdef CONFIG_HIGHMEM
788 return (idx == ZONE_HIGHMEM ||
789 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
795 static inline int is_normal_idx(enum zone_type idx)
797 return (idx == ZONE_NORMAL);
801 * is_highmem - helper function to quickly check if a struct zone is a
802 * highmem zone or not. This is an attempt to keep references
803 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
804 * @zone - pointer to struct zone variable
806 static inline int is_highmem(struct zone *zone)
808 #ifdef CONFIG_HIGHMEM
809 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
810 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
811 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
812 zone_movable_is_highmem());
818 static inline int is_normal(struct zone *zone)
820 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
823 static inline int is_dma32(struct zone *zone)
825 #ifdef CONFIG_ZONE_DMA32
826 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
832 static inline int is_dma(struct zone *zone)
834 #ifdef CONFIG_ZONE_DMA
835 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
841 /* These two functions are used to setup the per zone pages min values */
843 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
844 void __user *, size_t *, loff_t *);
845 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
846 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
847 void __user *, size_t *, loff_t *);
848 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
849 void __user *, size_t *, loff_t *);
850 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
851 void __user *, size_t *, loff_t *);
852 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
853 void __user *, size_t *, loff_t *);
855 extern int numa_zonelist_order_handler(struct ctl_table *, int,
856 void __user *, size_t *, loff_t *);
857 extern char numa_zonelist_order[];
858 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
860 #ifndef CONFIG_NEED_MULTIPLE_NODES
862 extern struct pglist_data contig_page_data;
863 #define NODE_DATA(nid) (&contig_page_data)
864 #define NODE_MEM_MAP(nid) mem_map
866 #else /* CONFIG_NEED_MULTIPLE_NODES */
868 #include <asm/mmzone.h>
870 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
872 extern struct pglist_data *first_online_pgdat(void);
873 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
874 extern struct zone *next_zone(struct zone *zone);
877 * for_each_online_pgdat - helper macro to iterate over all online nodes
878 * @pgdat - pointer to a pg_data_t variable
880 #define for_each_online_pgdat(pgdat) \
881 for (pgdat = first_online_pgdat(); \
883 pgdat = next_online_pgdat(pgdat))
885 * for_each_zone - helper macro to iterate over all memory zones
886 * @zone - pointer to struct zone variable
888 * The user only needs to declare the zone variable, for_each_zone
891 #define for_each_zone(zone) \
892 for (zone = (first_online_pgdat())->node_zones; \
894 zone = next_zone(zone))
896 #define for_each_populated_zone(zone) \
897 for (zone = (first_online_pgdat())->node_zones; \
899 zone = next_zone(zone)) \
900 if (!populated_zone(zone)) \
904 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
906 return zoneref->zone;
909 static inline int zonelist_zone_idx(struct zoneref *zoneref)
911 return zoneref->zone_idx;
914 static inline int zonelist_node_idx(struct zoneref *zoneref)
917 /* zone_to_nid not available in this context */
918 return zoneref->zone->node;
921 #endif /* CONFIG_NUMA */
925 * 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
926 * @z - The cursor used as a starting point for the search
927 * @highest_zoneidx - The zone index of the highest zone to return
928 * @nodes - An optional nodemask to filter the zonelist with
929 * @zone - The first suitable zone found is returned via this parameter
931 * This function returns the next zone at or below a given zone index that is
932 * within the allowed nodemask using a cursor as the starting point for the
933 * search. The zoneref returned is a cursor that represents the current zone
934 * being examined. It should be advanced by one before calling
935 * next_zones_zonelist again.
937 struct zoneref *next_zones_zonelist(struct zoneref *z,
938 enum zone_type highest_zoneidx,
943 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
944 * @zonelist - The zonelist to search for a suitable zone
945 * @highest_zoneidx - The zone index of the highest zone to return
946 * @nodes - An optional nodemask to filter the zonelist with
947 * @zone - The first suitable zone found is returned via this parameter
949 * This function returns the first zone at or below a given zone index that is
950 * within the allowed nodemask. The zoneref returned is a cursor that can be
951 * used to iterate the zonelist with next_zones_zonelist by advancing it by
952 * one before calling.
954 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
955 enum zone_type highest_zoneidx,
959 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
964 * 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
965 * @zone - The current zone in the iterator
966 * @z - The current pointer within zonelist->zones being iterated
967 * @zlist - The zonelist being iterated
968 * @highidx - The zone index of the highest zone to return
969 * @nodemask - Nodemask allowed by the allocator
971 * This iterator iterates though all zones at or below a given zone index and
972 * within a given nodemask
974 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
975 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
977 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
980 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
981 * @zone - The current zone in the iterator
982 * @z - The current pointer within zonelist->zones being iterated
983 * @zlist - The zonelist being iterated
984 * @highidx - The zone index of the highest zone to return
986 * This iterator iterates though all zones at or below a given zone index.
988 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
989 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
991 #ifdef CONFIG_SPARSEMEM
992 #include <asm/sparsemem.h>
995 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
996 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
997 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1003 #ifdef CONFIG_FLATMEM
1004 #define pfn_to_nid(pfn) (0)
1007 #ifdef CONFIG_SPARSEMEM
1010 * SECTION_SHIFT #bits space required to store a section #
1012 * PA_SECTION_SHIFT physical address to/from section number
1013 * PFN_SECTION_SHIFT pfn to/from section number
1015 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1016 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1018 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1020 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1021 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1023 #define SECTION_BLOCKFLAGS_BITS \
1024 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1026 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1027 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1030 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1031 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1033 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1034 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1038 struct mem_section {
1040 * This is, logically, a pointer to an array of struct
1041 * pages. However, it is stored with some other magic.
1042 * (see sparse.c::sparse_init_one_section())
1044 * Additionally during early boot we encode node id of
1045 * the location of the section here to guide allocation.
1046 * (see sparse.c::memory_present())
1048 * Making it a UL at least makes someone do a cast
1049 * before using it wrong.
1051 unsigned long section_mem_map;
1053 /* See declaration of similar field in struct zone */
1054 unsigned long *pageblock_flags;
1057 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1058 * section. (see memcontrol.h/page_cgroup.h about this.)
1060 struct page_cgroup *page_cgroup;
1065 #ifdef CONFIG_SPARSEMEM_EXTREME
1066 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1068 #define SECTIONS_PER_ROOT 1
1071 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1072 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1073 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1075 #ifdef CONFIG_SPARSEMEM_EXTREME
1076 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1078 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1081 static inline struct mem_section *__nr_to_section(unsigned long nr)
1083 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1085 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1087 extern int __section_nr(struct mem_section* ms);
1088 extern unsigned long usemap_size(void);
1091 * We use the lower bits of the mem_map pointer to store
1092 * a little bit of information. There should be at least
1093 * 3 bits here due to 32-bit alignment.
1095 #define SECTION_MARKED_PRESENT (1UL<<0)
1096 #define SECTION_HAS_MEM_MAP (1UL<<1)
1097 #define SECTION_MAP_LAST_BIT (1UL<<2)
1098 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1099 #define SECTION_NID_SHIFT 2
1101 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1103 unsigned long map = section->section_mem_map;
1104 map &= SECTION_MAP_MASK;
1105 return (struct page *)map;
1108 static inline int present_section(struct mem_section *section)
1110 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1113 static inline int present_section_nr(unsigned long nr)
1115 return present_section(__nr_to_section(nr));
1118 static inline int valid_section(struct mem_section *section)
1120 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1123 static inline int valid_section_nr(unsigned long nr)
1125 return valid_section(__nr_to_section(nr));
1128 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1130 return __nr_to_section(pfn_to_section_nr(pfn));
1133 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1134 static inline int pfn_valid(unsigned long pfn)
1136 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1138 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1142 static inline int pfn_present(unsigned long pfn)
1144 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1146 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1150 * These are _only_ used during initialisation, therefore they
1151 * can use __initdata ... They could have names to indicate
1155 #define pfn_to_nid(pfn) \
1157 unsigned long __pfn_to_nid_pfn = (pfn); \
1158 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1161 #define pfn_to_nid(pfn) (0)
1164 #define early_pfn_valid(pfn) pfn_valid(pfn)
1165 void sparse_init(void);
1167 #define sparse_init() do {} while (0)
1168 #define sparse_index_init(_sec, _nid) do {} while (0)
1169 #endif /* CONFIG_SPARSEMEM */
1171 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1172 bool early_pfn_in_nid(unsigned long pfn, int nid);
1174 #define early_pfn_in_nid(pfn, nid) (1)
1177 #ifndef early_pfn_valid
1178 #define early_pfn_valid(pfn) (1)
1181 void memory_present(int nid, unsigned long start, unsigned long end);
1182 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1185 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1186 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1187 * pfn_valid_within() should be used in this case; we optimise this away
1188 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1190 #ifdef CONFIG_HOLES_IN_ZONE
1191 #define pfn_valid_within(pfn) pfn_valid(pfn)
1193 #define pfn_valid_within(pfn) (1)
1196 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1198 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1199 * associated with it or not. In FLATMEM, it is expected that holes always
1200 * have valid memmap as long as there is valid PFNs either side of the hole.
1201 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1204 * However, an ARM, and maybe other embedded architectures in the future
1205 * free memmap backing holes to save memory on the assumption the memmap is
1206 * never used. The page_zone linkages are then broken even though pfn_valid()
1207 * returns true. A walker of the full memmap must then do this additional
1208 * check to ensure the memmap they are looking at is sane by making sure
1209 * the zone and PFN linkages are still valid. This is expensive, but walkers
1210 * of the full memmap are extremely rare.
1212 int memmap_valid_within(unsigned long pfn,
1213 struct page *page, struct zone *zone);
1215 static inline int memmap_valid_within(unsigned long pfn,
1216 struct page *page, struct zone *zone)
1220 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1222 #endif /* !__GENERATING_BOUNDS.H */
1223 #endif /* !__ASSEMBLY__ */
1224 #endif /* _LINUX_MMZONE_H */