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_HIGHATOMIC = 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 #ifdef CONFIG_MEMORY_ISOLATION
61 MIGRATE_ISOLATE, /* can't allocate from here */
66 /* In mm/page_alloc.c; keep in sync also with show_migration_types() there */
67 extern char * const migratetype_names[MIGRATE_TYPES];
70 # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
72 # define is_migrate_cma(migratetype) false
75 #define for_each_migratetype_order(order, type) \
76 for (order = 0; order < MAX_ORDER; order++) \
77 for (type = 0; type < MIGRATE_TYPES; type++)
79 extern int page_group_by_mobility_disabled;
81 #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
82 #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
84 #define get_pageblock_migratetype(page) \
85 get_pfnblock_flags_mask(page, page_to_pfn(page), \
86 PB_migrate_end, MIGRATETYPE_MASK)
89 struct list_head free_list[MIGRATE_TYPES];
90 unsigned long nr_free;
96 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
97 * So add a wild amount of padding here to ensure that they fall into separate
98 * cachelines. There are very few zone structures in the machine, so space
99 * consumption is not a concern here.
101 #if defined(CONFIG_SMP)
102 struct zone_padding {
104 } ____cacheline_internodealigned_in_smp;
105 #define ZONE_PADDING(name) struct zone_padding name;
107 #define ZONE_PADDING(name)
110 enum zone_stat_item {
111 /* First 128 byte cacheline (assuming 64 bit words) */
115 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
116 NR_ACTIVE_ANON, /* " " " " " */
117 NR_INACTIVE_FILE, /* " " " " " */
118 NR_ACTIVE_FILE, /* " " " " " */
119 NR_UNEVICTABLE, /* " " " " " */
120 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
121 NR_ANON_PAGES, /* Mapped anonymous pages */
122 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
123 only modified from process context */
128 NR_SLAB_UNRECLAIMABLE,
129 NR_PAGETABLE, /* used for pagetables */
131 /* Second 128 byte cacheline */
132 NR_UNSTABLE_NFS, /* NFS unstable pages */
135 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
136 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
137 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
138 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
139 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
140 NR_DIRTIED, /* page dirtyings since bootup */
141 NR_WRITTEN, /* page writings since bootup */
142 NR_PAGES_SCANNED, /* pages scanned since last reclaim */
143 #if IS_ENABLED(CONFIG_ZSMALLOC)
144 NR_ZSPAGES, /* allocated in zsmalloc */
147 NUMA_HIT, /* allocated in intended node */
148 NUMA_MISS, /* allocated in non intended node */
149 NUMA_FOREIGN, /* was intended here, hit elsewhere */
150 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
151 NUMA_LOCAL, /* allocation from local node */
152 NUMA_OTHER, /* allocation from other node */
156 WORKINGSET_NODERECLAIM,
161 NR_VM_ZONE_STAT_ITEMS };
164 * We do arithmetic on the LRU lists in various places in the code,
165 * so it is important to keep the active lists LRU_ACTIVE higher in
166 * the array than the corresponding inactive lists, and to keep
167 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
169 * This has to be kept in sync with the statistics in zone_stat_item
170 * above and the descriptions in vmstat_text in mm/vmstat.c
177 LRU_INACTIVE_ANON = LRU_BASE,
178 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
179 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
180 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
185 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
187 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
189 static inline int is_file_lru(enum lru_list lru)
191 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
194 static inline int is_active_lru(enum lru_list lru)
196 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
199 struct zone_reclaim_stat {
201 * The pageout code in vmscan.c keeps track of how many of the
202 * mem/swap backed and file backed pages are referenced.
203 * The higher the rotated/scanned ratio, the more valuable
206 * The anon LRU stats live in [0], file LRU stats in [1]
208 unsigned long recent_rotated[2];
209 unsigned long recent_scanned[2];
213 struct list_head lists[NR_LRU_LISTS];
214 struct zone_reclaim_stat reclaim_stat;
215 /* Evictions & activations on the inactive file list */
216 atomic_long_t inactive_age;
222 /* Mask used at gathering information at once (see memcontrol.c) */
223 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
224 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
225 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
227 /* Isolate clean file */
228 #define ISOLATE_CLEAN ((__force isolate_mode_t)0x1)
229 /* Isolate unmapped file */
230 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
231 /* Isolate for asynchronous migration */
232 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
233 /* Isolate unevictable pages */
234 #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
236 /* LRU Isolation modes. */
237 typedef unsigned __bitwise__ isolate_mode_t;
239 enum zone_watermarks {
246 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
247 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
248 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
250 struct per_cpu_pages {
251 int count; /* number of pages in the list */
252 int high; /* high watermark, emptying needed */
253 int batch; /* chunk size for buddy add/remove */
255 /* Lists of pages, one per migrate type stored on the pcp-lists */
256 struct list_head lists[MIGRATE_PCPTYPES];
259 struct per_cpu_pageset {
260 struct per_cpu_pages pcp;
266 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
270 #endif /* !__GENERATING_BOUNDS.H */
273 #ifdef CONFIG_ZONE_DMA
275 * ZONE_DMA is used when there are devices that are not able
276 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
277 * carve out the portion of memory that is needed for these devices.
278 * The range is arch specific.
283 * ---------------------------
284 * parisc, ia64, sparc <4G
287 * alpha Unlimited or 0-16MB.
289 * i386, x86_64 and multiple other arches
294 #ifdef CONFIG_ZONE_DMA32
296 * x86_64 needs two ZONE_DMAs because it supports devices that are
297 * only able to do DMA to the lower 16M but also 32 bit devices that
298 * can only do DMA areas below 4G.
303 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
304 * performed on pages in ZONE_NORMAL if the DMA devices support
305 * transfers to all addressable memory.
308 #ifdef CONFIG_HIGHMEM
310 * A memory area that is only addressable by the kernel through
311 * mapping portions into its own address space. This is for example
312 * used by i386 to allow the kernel to address the memory beyond
313 * 900MB. The kernel will set up special mappings (page
314 * table entries on i386) for each page that the kernel needs to
320 #ifdef CONFIG_ZONE_DEVICE
327 #ifndef __GENERATING_BOUNDS_H
330 /* Read-mostly fields */
332 /* zone watermarks, access with *_wmark_pages(zone) macros */
333 unsigned long watermark[NR_WMARK];
335 unsigned long nr_reserved_highatomic;
338 * We don't know if the memory that we're going to allocate will be
339 * freeable or/and it will be released eventually, so to avoid totally
340 * wasting several GB of ram we must reserve some of the lower zone
341 * memory (otherwise we risk to run OOM on the lower zones despite
342 * there being tons of freeable ram on the higher zones). This array is
343 * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
346 long lowmem_reserve[MAX_NR_ZONES];
353 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
354 * this zone's LRU. Maintained by the pageout code.
356 unsigned int inactive_ratio;
358 struct pglist_data *zone_pgdat;
359 struct per_cpu_pageset __percpu *pageset;
362 * This is a per-zone reserve of pages that are not available
363 * to userspace allocations.
365 unsigned long totalreserve_pages;
367 #ifndef CONFIG_SPARSEMEM
369 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
370 * In SPARSEMEM, this map is stored in struct mem_section
372 unsigned long *pageblock_flags;
373 #endif /* CONFIG_SPARSEMEM */
377 * zone reclaim becomes active if more unmapped pages exist.
379 unsigned long min_unmapped_pages;
380 unsigned long min_slab_pages;
381 #endif /* CONFIG_NUMA */
383 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
384 unsigned long zone_start_pfn;
387 * spanned_pages is the total pages spanned by the zone, including
388 * holes, which is calculated as:
389 * spanned_pages = zone_end_pfn - zone_start_pfn;
391 * present_pages is physical pages existing within the zone, which
393 * present_pages = spanned_pages - absent_pages(pages in holes);
395 * managed_pages is present pages managed by the buddy system, which
396 * is calculated as (reserved_pages includes pages allocated by the
397 * bootmem allocator):
398 * managed_pages = present_pages - reserved_pages;
400 * So present_pages may be used by memory hotplug or memory power
401 * management logic to figure out unmanaged pages by checking
402 * (present_pages - managed_pages). And managed_pages should be used
403 * by page allocator and vm scanner to calculate all kinds of watermarks
408 * zone_start_pfn and spanned_pages are protected by span_seqlock.
409 * It is a seqlock because it has to be read outside of zone->lock,
410 * and it is done in the main allocator path. But, it is written
411 * quite infrequently.
413 * The span_seq lock is declared along with zone->lock because it is
414 * frequently read in proximity to zone->lock. It's good to
415 * give them a chance of being in the same cacheline.
417 * Write access to present_pages at runtime should be protected by
418 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
419 * present_pages should get_online_mems() to get a stable value.
421 * Read access to managed_pages should be safe because it's unsigned
422 * long. Write access to zone->managed_pages and totalram_pages are
423 * protected by managed_page_count_lock at runtime. Idealy only
424 * adjust_managed_page_count() should be used instead of directly
425 * touching zone->managed_pages and totalram_pages.
427 unsigned long managed_pages;
428 unsigned long spanned_pages;
429 unsigned long present_pages;
433 #ifdef CONFIG_MEMORY_ISOLATION
435 * Number of isolated pageblock. It is used to solve incorrect
436 * freepage counting problem due to racy retrieving migratetype
437 * of pageblock. Protected by zone->lock.
439 unsigned long nr_isolate_pageblock;
442 #ifdef CONFIG_MEMORY_HOTPLUG
443 /* see spanned/present_pages for more description */
444 seqlock_t span_seqlock;
448 * wait_table -- the array holding the hash table
449 * wait_table_hash_nr_entries -- the size of the hash table array
450 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
452 * The purpose of all these is to keep track of the people
453 * waiting for a page to become available and make them
454 * runnable again when possible. The trouble is that this
455 * consumes a lot of space, especially when so few things
456 * wait on pages at a given time. So instead of using
457 * per-page waitqueues, we use a waitqueue hash table.
459 * The bucket discipline is to sleep on the same queue when
460 * colliding and wake all in that wait queue when removing.
461 * When something wakes, it must check to be sure its page is
462 * truly available, a la thundering herd. The cost of a
463 * collision is great, but given the expected load of the
464 * table, they should be so rare as to be outweighed by the
465 * benefits from the saved space.
467 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
468 * primary users of these fields, and in mm/page_alloc.c
469 * free_area_init_core() performs the initialization of them.
471 wait_queue_head_t *wait_table;
472 unsigned long wait_table_hash_nr_entries;
473 unsigned long wait_table_bits;
476 /* free areas of different sizes */
477 struct free_area free_area[MAX_ORDER];
479 /* zone flags, see below */
482 /* Write-intensive fields used from the page allocator */
487 /* Write-intensive fields used by page reclaim */
489 /* Fields commonly accessed by the page reclaim scanner */
491 struct lruvec lruvec;
494 * When free pages are below this point, additional steps are taken
495 * when reading the number of free pages to avoid per-cpu counter
496 * drift allowing watermarks to be breached
498 unsigned long percpu_drift_mark;
500 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
501 /* pfn where compaction free scanner should start */
502 unsigned long compact_cached_free_pfn;
503 /* pfn where async and sync compaction migration scanner should start */
504 unsigned long compact_cached_migrate_pfn[2];
507 #ifdef CONFIG_COMPACTION
509 * On compaction failure, 1<<compact_defer_shift compactions
510 * are skipped before trying again. The number attempted since
511 * last failure is tracked with compact_considered.
513 unsigned int compact_considered;
514 unsigned int compact_defer_shift;
515 int compact_order_failed;
518 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
519 /* Set to true when the PG_migrate_skip bits should be cleared */
520 bool compact_blockskip_flush;
526 /* Zone statistics */
527 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
528 } ____cacheline_internodealigned_in_smp;
531 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
532 ZONE_CONGESTED, /* zone has many dirty pages backed by
535 ZONE_DIRTY, /* reclaim scanning has recently found
536 * many dirty file pages at the tail
539 ZONE_WRITEBACK, /* reclaim scanning has recently found
540 * many pages under writeback
542 ZONE_FAIR_DEPLETED, /* fair zone policy batch depleted */
545 static inline unsigned long zone_end_pfn(const struct zone *zone)
547 return zone->zone_start_pfn + zone->spanned_pages;
550 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
552 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
555 static inline bool zone_is_initialized(struct zone *zone)
557 return !!zone->wait_table;
560 static inline bool zone_is_empty(struct zone *zone)
562 return zone->spanned_pages == 0;
566 * The "priority" of VM scanning is how much of the queues we will scan in one
567 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
568 * queues ("queue_length >> 12") during an aging round.
570 #define DEF_PRIORITY 12
572 /* Maximum number of zones on a zonelist */
573 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
576 ZONELIST_FALLBACK, /* zonelist with fallback */
579 * The NUMA zonelists are doubled because we need zonelists that
580 * restrict the allocations to a single node for __GFP_THISNODE.
582 ZONELIST_NOFALLBACK, /* zonelist without fallback (__GFP_THISNODE) */
588 * This struct contains information about a zone in a zonelist. It is stored
589 * here to avoid dereferences into large structures and lookups of tables
592 struct zone *zone; /* Pointer to actual zone */
593 int zone_idx; /* zone_idx(zoneref->zone) */
597 * One allocation request operates on a zonelist. A zonelist
598 * is a list of zones, the first one is the 'goal' of the
599 * allocation, the other zones are fallback zones, in decreasing
602 * To speed the reading of the zonelist, the zonerefs contain the zone index
603 * of the entry being read. Helper functions to access information given
604 * a struct zoneref are
606 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
607 * zonelist_zone_idx() - Return the index of the zone for an entry
608 * zonelist_node_idx() - Return the index of the node for an entry
611 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
614 #ifndef CONFIG_DISCONTIGMEM
615 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
616 extern struct page *mem_map;
620 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
621 * (mostly NUMA machines?) to denote a higher-level memory zone than the
624 * On NUMA machines, each NUMA node would have a pg_data_t to describe
625 * it's memory layout.
627 * Memory statistics and page replacement data structures are maintained on a
631 typedef struct pglist_data {
632 struct zone node_zones[MAX_NR_ZONES];
633 struct zonelist node_zonelists[MAX_ZONELISTS];
635 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
636 struct page *node_mem_map;
637 #ifdef CONFIG_PAGE_EXTENSION
638 struct page_ext *node_page_ext;
641 #ifndef CONFIG_NO_BOOTMEM
642 struct bootmem_data *bdata;
644 #ifdef CONFIG_MEMORY_HOTPLUG
646 * Must be held any time you expect node_start_pfn, node_present_pages
647 * or node_spanned_pages stay constant. Holding this will also
648 * guarantee that any pfn_valid() stays that way.
650 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
651 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
653 * Nests above zone->lock and zone->span_seqlock
655 spinlock_t node_size_lock;
657 unsigned long node_start_pfn;
658 unsigned long node_present_pages; /* total number of physical pages */
659 unsigned long node_spanned_pages; /* total size of physical page
660 range, including holes */
662 wait_queue_head_t kswapd_wait;
663 wait_queue_head_t pfmemalloc_wait;
664 struct task_struct *kswapd; /* Protected by
665 mem_hotplug_begin/end() */
666 int kswapd_max_order;
667 enum zone_type classzone_idx;
668 #ifdef CONFIG_COMPACTION
669 int kcompactd_max_order;
670 enum zone_type kcompactd_classzone_idx;
671 wait_queue_head_t kcompactd_wait;
672 struct task_struct *kcompactd;
674 #ifdef CONFIG_NUMA_BALANCING
675 /* Lock serializing the migrate rate limiting window */
676 spinlock_t numabalancing_migrate_lock;
678 /* Rate limiting time interval */
679 unsigned long numabalancing_migrate_next_window;
681 /* Number of pages migrated during the rate limiting time interval */
682 unsigned long numabalancing_migrate_nr_pages;
685 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
687 * If memory initialisation on large machines is deferred then this
688 * is the first PFN that needs to be initialised.
690 unsigned long first_deferred_pfn;
691 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
693 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
694 spinlock_t split_queue_lock;
695 struct list_head split_queue;
696 unsigned long split_queue_len;
700 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
701 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
702 #ifdef CONFIG_FLAT_NODE_MEM_MAP
703 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
705 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
707 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
709 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
710 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
712 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
714 return pgdat->node_start_pfn + pgdat->node_spanned_pages;
717 static inline bool pgdat_is_empty(pg_data_t *pgdat)
719 return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
722 static inline int zone_id(const struct zone *zone)
724 struct pglist_data *pgdat = zone->zone_pgdat;
726 return zone - pgdat->node_zones;
729 #ifdef CONFIG_ZONE_DEVICE
730 static inline bool is_dev_zone(const struct zone *zone)
732 return zone_id(zone) == ZONE_DEVICE;
735 static inline bool is_dev_zone(const struct zone *zone)
741 #include <linux/memory_hotplug.h>
743 extern struct mutex zonelists_mutex;
744 void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
745 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
746 bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
747 int classzone_idx, unsigned int alloc_flags,
749 bool zone_watermark_ok(struct zone *z, unsigned int order,
750 unsigned long mark, int classzone_idx,
751 unsigned int alloc_flags);
752 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
753 unsigned long mark, int classzone_idx);
754 enum memmap_context {
758 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
761 extern void lruvec_init(struct lruvec *lruvec);
763 static inline struct zone *lruvec_zone(struct lruvec *lruvec)
768 return container_of(lruvec, struct zone, lruvec);
772 extern unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru);
774 #ifdef CONFIG_HAVE_MEMORY_PRESENT
775 void memory_present(int nid, unsigned long start, unsigned long end);
777 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
780 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
781 int local_memory_node(int node_id);
783 static inline int local_memory_node(int node_id) { return node_id; };
786 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
787 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
791 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
793 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
795 static inline int populated_zone(struct zone *zone)
797 return (!!zone->present_pages);
800 extern int movable_zone;
802 #ifdef CONFIG_HIGHMEM
803 static inline int zone_movable_is_highmem(void)
805 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
806 return movable_zone == ZONE_HIGHMEM;
808 return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
813 static inline int is_highmem_idx(enum zone_type idx)
815 #ifdef CONFIG_HIGHMEM
816 return (idx == ZONE_HIGHMEM ||
817 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
824 * is_highmem - helper function to quickly check if a struct zone is a
825 * highmem zone or not. This is an attempt to keep references
826 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
827 * @zone - pointer to struct zone variable
829 static inline int is_highmem(struct zone *zone)
831 #ifdef CONFIG_HIGHMEM
832 return is_highmem_idx(zone_idx(zone));
838 /* These two functions are used to setup the per zone pages min values */
840 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
841 void __user *, size_t *, loff_t *);
842 int watermark_scale_factor_sysctl_handler(struct ctl_table *, int,
843 void __user *, size_t *, loff_t *);
844 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
845 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
846 void __user *, size_t *, loff_t *);
847 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
848 void __user *, size_t *, loff_t *);
849 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
850 void __user *, size_t *, loff_t *);
851 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
852 void __user *, size_t *, loff_t *);
854 extern int numa_zonelist_order_handler(struct ctl_table *, int,
855 void __user *, size_t *, loff_t *);
856 extern char numa_zonelist_order[];
857 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
859 #ifndef CONFIG_NEED_MULTIPLE_NODES
861 extern struct pglist_data contig_page_data;
862 #define NODE_DATA(nid) (&contig_page_data)
863 #define NODE_MEM_MAP(nid) mem_map
865 #else /* CONFIG_NEED_MULTIPLE_NODES */
867 #include <asm/mmzone.h>
869 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
871 extern struct pglist_data *first_online_pgdat(void);
872 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
873 extern struct zone *next_zone(struct zone *zone);
876 * for_each_online_pgdat - helper macro to iterate over all online nodes
877 * @pgdat - pointer to a pg_data_t variable
879 #define for_each_online_pgdat(pgdat) \
880 for (pgdat = first_online_pgdat(); \
882 pgdat = next_online_pgdat(pgdat))
884 * for_each_zone - helper macro to iterate over all memory zones
885 * @zone - pointer to struct zone variable
887 * The user only needs to declare the zone variable, for_each_zone
890 #define for_each_zone(zone) \
891 for (zone = (first_online_pgdat())->node_zones; \
893 zone = next_zone(zone))
895 #define for_each_populated_zone(zone) \
896 for (zone = (first_online_pgdat())->node_zones; \
898 zone = next_zone(zone)) \
899 if (!populated_zone(zone)) \
903 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
905 return zoneref->zone;
908 static inline int zonelist_zone_idx(struct zoneref *zoneref)
910 return zoneref->zone_idx;
913 static inline int zonelist_node_idx(struct zoneref *zoneref)
916 /* zone_to_nid not available in this context */
917 return zoneref->zone->node;
920 #endif /* CONFIG_NUMA */
923 struct zoneref *__next_zones_zonelist(struct zoneref *z,
924 enum zone_type highest_zoneidx,
928 * 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
929 * @z - The cursor used as a starting point for the search
930 * @highest_zoneidx - The zone index of the highest zone to return
931 * @nodes - An optional nodemask to filter the zonelist with
933 * This function returns the next zone at or below a given zone index that is
934 * within the allowed nodemask using a cursor as the starting point for the
935 * search. The zoneref returned is a cursor that represents the current zone
936 * being examined. It should be advanced by one before calling
937 * next_zones_zonelist again.
939 static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,
940 enum zone_type highest_zoneidx,
943 if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))
945 return __next_zones_zonelist(z, highest_zoneidx, nodes);
949 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
950 * @zonelist - The zonelist to search for a suitable zone
951 * @highest_zoneidx - The zone index of the highest zone to return
952 * @nodes - An optional nodemask to filter the zonelist with
953 * @zone - The first suitable zone found is returned via this parameter
955 * This function returns the first zone at or below a given zone index that is
956 * within the allowed nodemask. The zoneref returned is a cursor that can be
957 * used to iterate the zonelist with next_zones_zonelist by advancing it by
958 * one before calling.
960 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
961 enum zone_type highest_zoneidx,
964 return next_zones_zonelist(zonelist->_zonerefs,
965 highest_zoneidx, nodes);
969 * 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
970 * @zone - The current zone in the iterator
971 * @z - The current pointer within zonelist->zones being iterated
972 * @zlist - The zonelist being iterated
973 * @highidx - The zone index of the highest zone to return
974 * @nodemask - Nodemask allowed by the allocator
976 * This iterator iterates though all zones at or below a given zone index and
977 * within a given nodemask
979 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
980 for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \
982 z = next_zones_zonelist(++z, highidx, nodemask), \
983 zone = zonelist_zone(z))
985 #define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
986 for (zone = z->zone; \
988 z = next_zones_zonelist(++z, highidx, nodemask), \
989 zone = zonelist_zone(z))
993 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
994 * @zone - The current zone in the iterator
995 * @z - The current pointer within zonelist->zones being iterated
996 * @zlist - The zonelist being iterated
997 * @highidx - The zone index of the highest zone to return
999 * This iterator iterates though all zones at or below a given zone index.
1001 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
1002 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1004 #ifdef CONFIG_SPARSEMEM
1005 #include <asm/sparsemem.h>
1008 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1009 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1010 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1016 #ifdef CONFIG_FLATMEM
1017 #define pfn_to_nid(pfn) (0)
1020 #ifdef CONFIG_SPARSEMEM
1023 * SECTION_SHIFT #bits space required to store a section #
1025 * PA_SECTION_SHIFT physical address to/from section number
1026 * PFN_SECTION_SHIFT pfn to/from section number
1028 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1029 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1031 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1033 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1034 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1036 #define SECTION_BLOCKFLAGS_BITS \
1037 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1039 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1040 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1043 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1044 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1046 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1047 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1051 struct mem_section {
1053 * This is, logically, a pointer to an array of struct
1054 * pages. However, it is stored with some other magic.
1055 * (see sparse.c::sparse_init_one_section())
1057 * Additionally during early boot we encode node id of
1058 * the location of the section here to guide allocation.
1059 * (see sparse.c::memory_present())
1061 * Making it a UL at least makes someone do a cast
1062 * before using it wrong.
1064 unsigned long section_mem_map;
1066 /* See declaration of similar field in struct zone */
1067 unsigned long *pageblock_flags;
1068 #ifdef CONFIG_PAGE_EXTENSION
1070 * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1071 * section. (see page_ext.h about this.)
1073 struct page_ext *page_ext;
1077 * WARNING: mem_section must be a power-of-2 in size for the
1078 * calculation and use of SECTION_ROOT_MASK to make sense.
1082 #ifdef CONFIG_SPARSEMEM_EXTREME
1083 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1085 #define SECTIONS_PER_ROOT 1
1088 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1089 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1090 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1092 #ifdef CONFIG_SPARSEMEM_EXTREME
1093 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1095 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1098 static inline struct mem_section *__nr_to_section(unsigned long nr)
1100 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1102 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1104 extern int __section_nr(struct mem_section* ms);
1105 extern unsigned long usemap_size(void);
1108 * We use the lower bits of the mem_map pointer to store
1109 * a little bit of information. There should be at least
1110 * 3 bits here due to 32-bit alignment.
1112 #define SECTION_MARKED_PRESENT (1UL<<0)
1113 #define SECTION_HAS_MEM_MAP (1UL<<1)
1114 #define SECTION_MAP_LAST_BIT (1UL<<2)
1115 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1116 #define SECTION_NID_SHIFT 2
1118 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1120 unsigned long map = section->section_mem_map;
1121 map &= SECTION_MAP_MASK;
1122 return (struct page *)map;
1125 static inline int present_section(struct mem_section *section)
1127 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1130 static inline int present_section_nr(unsigned long nr)
1132 return present_section(__nr_to_section(nr));
1135 static inline int valid_section(struct mem_section *section)
1137 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1140 static inline int valid_section_nr(unsigned long nr)
1142 return valid_section(__nr_to_section(nr));
1145 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1147 return __nr_to_section(pfn_to_section_nr(pfn));
1150 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1151 static inline int pfn_valid(unsigned long pfn)
1153 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1155 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1159 static inline int pfn_present(unsigned long pfn)
1161 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1163 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1167 * These are _only_ used during initialisation, therefore they
1168 * can use __initdata ... They could have names to indicate
1172 #define pfn_to_nid(pfn) \
1174 unsigned long __pfn_to_nid_pfn = (pfn); \
1175 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1178 #define pfn_to_nid(pfn) (0)
1181 #define early_pfn_valid(pfn) pfn_valid(pfn)
1182 void sparse_init(void);
1184 #define sparse_init() do {} while (0)
1185 #define sparse_index_init(_sec, _nid) do {} while (0)
1186 #endif /* CONFIG_SPARSEMEM */
1189 * During memory init memblocks map pfns to nids. The search is expensive and
1190 * this caches recent lookups. The implementation of __early_pfn_to_nid
1191 * may treat start/end as pfns or sections.
1193 struct mminit_pfnnid_cache {
1194 unsigned long last_start;
1195 unsigned long last_end;
1199 #ifndef early_pfn_valid
1200 #define early_pfn_valid(pfn) (1)
1203 void memory_present(int nid, unsigned long start, unsigned long end);
1204 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1207 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1208 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1209 * pfn_valid_within() should be used in this case; we optimise this away
1210 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1212 #ifdef CONFIG_HOLES_IN_ZONE
1213 #define pfn_valid_within(pfn) pfn_valid(pfn)
1215 #define pfn_valid_within(pfn) (1)
1218 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1220 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1221 * associated with it or not. In FLATMEM, it is expected that holes always
1222 * have valid memmap as long as there is valid PFNs either side of the hole.
1223 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1226 * However, an ARM, and maybe other embedded architectures in the future
1227 * free memmap backing holes to save memory on the assumption the memmap is
1228 * never used. The page_zone linkages are then broken even though pfn_valid()
1229 * returns true. A walker of the full memmap must then do this additional
1230 * check to ensure the memmap they are looking at is sane by making sure
1231 * the zone and PFN linkages are still valid. This is expensive, but walkers
1232 * of the full memmap are extremely rare.
1234 bool memmap_valid_within(unsigned long pfn,
1235 struct page *page, struct zone *zone);
1237 static inline bool memmap_valid_within(unsigned long pfn,
1238 struct page *page, struct zone *zone)
1242 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1244 #endif /* !__GENERATING_BOUNDS.H */
1245 #endif /* !__ASSEMBLY__ */
1246 #endif /* _LINUX_MMZONE_H */