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)
88 static inline int get_pfnblock_migratetype(struct page
*page
, unsigned long pfn
)
90 BUILD_BUG_ON(PB_migrate_end
- PB_migrate
!= 2);
91 return get_pfnblock_flags_mask(page
, pfn
, PB_migrate_end
,
96 struct list_head free_list
[MIGRATE_TYPES
];
97 unsigned long nr_free
;
103 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
104 * So add a wild amount of padding here to ensure that they fall into separate
105 * cachelines. There are very few zone structures in the machine, so space
106 * consumption is not a concern here.
108 #if defined(CONFIG_SMP)
109 struct zone_padding
{
111 } ____cacheline_internodealigned_in_smp
;
112 #define ZONE_PADDING(name) struct zone_padding name;
114 #define ZONE_PADDING(name)
117 enum zone_stat_item
{
118 /* First 128 byte cacheline (assuming 64 bit words) */
122 NR_INACTIVE_ANON
= NR_LRU_BASE
, /* must match order of LRU_[IN]ACTIVE */
123 NR_ACTIVE_ANON
, /* " " " " " */
124 NR_INACTIVE_FILE
, /* " " " " " */
125 NR_ACTIVE_FILE
, /* " " " " " */
126 NR_UNEVICTABLE
, /* " " " " " */
127 NR_MLOCK
, /* mlock()ed pages found and moved off LRU */
128 NR_ANON_PAGES
, /* Mapped anonymous pages */
129 NR_FILE_MAPPED
, /* pagecache pages mapped into pagetables.
130 only modified from process context */
135 NR_SLAB_UNRECLAIMABLE
,
136 NR_PAGETABLE
, /* used for pagetables */
138 /* Second 128 byte cacheline */
139 NR_UNSTABLE_NFS
, /* NFS unstable pages */
142 NR_VMSCAN_IMMEDIATE
, /* Prioritise for reclaim when writeback ends */
143 NR_WRITEBACK_TEMP
, /* Writeback using temporary buffers */
144 NR_ISOLATED_ANON
, /* Temporary isolated pages from anon lru */
145 NR_ISOLATED_FILE
, /* Temporary isolated pages from file lru */
146 NR_SHMEM
, /* shmem pages (included tmpfs/GEM pages) */
147 NR_DIRTIED
, /* page dirtyings since bootup */
148 NR_WRITTEN
, /* page writings since bootup */
149 NR_PAGES_SCANNED
, /* pages scanned since last reclaim */
151 NUMA_HIT
, /* allocated in intended node */
152 NUMA_MISS
, /* allocated in non intended node */
153 NUMA_FOREIGN
, /* was intended here, hit elsewhere */
154 NUMA_INTERLEAVE_HIT
, /* interleaver preferred this zone */
155 NUMA_LOCAL
, /* allocation from local node */
156 NUMA_OTHER
, /* allocation from other node */
160 WORKINGSET_NODERECLAIM
,
161 NR_ANON_TRANSPARENT_HUGEPAGES
,
163 NR_VM_ZONE_STAT_ITEMS
};
166 * We do arithmetic on the LRU lists in various places in the code,
167 * so it is important to keep the active lists LRU_ACTIVE higher in
168 * the array than the corresponding inactive lists, and to keep
169 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
171 * This has to be kept in sync with the statistics in zone_stat_item
172 * above and the descriptions in vmstat_text in mm/vmstat.c
179 LRU_INACTIVE_ANON
= LRU_BASE
,
180 LRU_ACTIVE_ANON
= LRU_BASE
+ LRU_ACTIVE
,
181 LRU_INACTIVE_FILE
= LRU_BASE
+ LRU_FILE
,
182 LRU_ACTIVE_FILE
= LRU_BASE
+ LRU_FILE
+ LRU_ACTIVE
,
187 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
189 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
191 static inline int is_file_lru(enum lru_list lru
)
193 return (lru
== LRU_INACTIVE_FILE
|| lru
== LRU_ACTIVE_FILE
);
196 static inline int is_active_lru(enum lru_list lru
)
198 return (lru
== LRU_ACTIVE_ANON
|| lru
== LRU_ACTIVE_FILE
);
201 struct zone_reclaim_stat
{
203 * The pageout code in vmscan.c keeps track of how many of the
204 * mem/swap backed and file backed pages are referenced.
205 * The higher the rotated/scanned ratio, the more valuable
208 * The anon LRU stats live in [0], file LRU stats in [1]
210 unsigned long recent_rotated
[2];
211 unsigned long recent_scanned
[2];
215 struct list_head lists
[NR_LRU_LISTS
];
216 struct zone_reclaim_stat reclaim_stat
;
217 /* Evictions & activations on the inactive file list */
218 atomic_long_t inactive_age
;
224 /* Mask used at gathering information at once (see memcontrol.c) */
225 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
226 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
227 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
229 /* Isolate clean file */
230 #define ISOLATE_CLEAN ((__force isolate_mode_t)0x1)
231 /* Isolate unmapped file */
232 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
233 /* Isolate for asynchronous migration */
234 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
235 /* Isolate unevictable pages */
236 #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
238 /* LRU Isolation modes. */
239 typedef unsigned __bitwise__ isolate_mode_t
;
241 enum zone_watermarks
{
248 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
249 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
250 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
252 struct per_cpu_pages
{
253 int count
; /* number of pages in the list */
254 int high
; /* high watermark, emptying needed */
255 int batch
; /* chunk size for buddy add/remove */
257 /* Lists of pages, one per migrate type stored on the pcp-lists */
258 struct list_head lists
[MIGRATE_PCPTYPES
];
261 struct per_cpu_pageset
{
262 struct per_cpu_pages pcp
;
268 s8 vm_stat_diff
[NR_VM_ZONE_STAT_ITEMS
];
272 #endif /* !__GENERATING_BOUNDS.H */
275 #ifdef CONFIG_ZONE_DMA
277 * ZONE_DMA is used when there are devices that are not able
278 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
279 * carve out the portion of memory that is needed for these devices.
280 * The range is arch specific.
285 * ---------------------------
286 * parisc, ia64, sparc <4G
289 * alpha Unlimited or 0-16MB.
291 * i386, x86_64 and multiple other arches
296 #ifdef CONFIG_ZONE_DMA32
298 * x86_64 needs two ZONE_DMAs because it supports devices that are
299 * only able to do DMA to the lower 16M but also 32 bit devices that
300 * can only do DMA areas below 4G.
305 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
306 * performed on pages in ZONE_NORMAL if the DMA devices support
307 * transfers to all addressable memory.
310 #ifdef CONFIG_HIGHMEM
312 * A memory area that is only addressable by the kernel through
313 * mapping portions into its own address space. This is for example
314 * used by i386 to allow the kernel to address the memory beyond
315 * 900MB. The kernel will set up special mappings (page
316 * table entries on i386) for each page that the kernel needs to
322 #ifdef CONFIG_ZONE_DEVICE
329 #ifndef __GENERATING_BOUNDS_H
332 /* Read-mostly fields */
334 /* zone watermarks, access with *_wmark_pages(zone) macros */
335 unsigned long watermark
[NR_WMARK
];
337 unsigned long nr_reserved_highatomic
;
340 * We don't know if the memory that we're going to allocate will be
341 * freeable or/and it will be released eventually, so to avoid totally
342 * wasting several GB of ram we must reserve some of the lower zone
343 * memory (otherwise we risk to run OOM on the lower zones despite
344 * there being tons of freeable ram on the higher zones). This array is
345 * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
348 long lowmem_reserve
[MAX_NR_ZONES
];
355 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
356 * this zone's LRU. Maintained by the pageout code.
358 unsigned int inactive_ratio
;
360 struct pglist_data
*zone_pgdat
;
361 struct per_cpu_pageset __percpu
*pageset
;
364 * This is a per-zone reserve of pages that are not available
365 * to userspace allocations.
367 unsigned long totalreserve_pages
;
369 #ifndef CONFIG_SPARSEMEM
371 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
372 * In SPARSEMEM, this map is stored in struct mem_section
374 unsigned long *pageblock_flags
;
375 #endif /* CONFIG_SPARSEMEM */
379 * zone reclaim becomes active if more unmapped pages exist.
381 unsigned long min_unmapped_pages
;
382 unsigned long min_slab_pages
;
383 #endif /* CONFIG_NUMA */
385 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
386 unsigned long zone_start_pfn
;
389 * spanned_pages is the total pages spanned by the zone, including
390 * holes, which is calculated as:
391 * spanned_pages = zone_end_pfn - zone_start_pfn;
393 * present_pages is physical pages existing within the zone, which
395 * present_pages = spanned_pages - absent_pages(pages in holes);
397 * managed_pages is present pages managed by the buddy system, which
398 * is calculated as (reserved_pages includes pages allocated by the
399 * bootmem allocator):
400 * managed_pages = present_pages - reserved_pages;
402 * So present_pages may be used by memory hotplug or memory power
403 * management logic to figure out unmanaged pages by checking
404 * (present_pages - managed_pages). And managed_pages should be used
405 * by page allocator and vm scanner to calculate all kinds of watermarks
410 * zone_start_pfn and spanned_pages are protected by span_seqlock.
411 * It is a seqlock because it has to be read outside of zone->lock,
412 * and it is done in the main allocator path. But, it is written
413 * quite infrequently.
415 * The span_seq lock is declared along with zone->lock because it is
416 * frequently read in proximity to zone->lock. It's good to
417 * give them a chance of being in the same cacheline.
419 * Write access to present_pages at runtime should be protected by
420 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
421 * present_pages should get_online_mems() to get a stable value.
423 * Read access to managed_pages should be safe because it's unsigned
424 * long. Write access to zone->managed_pages and totalram_pages are
425 * protected by managed_page_count_lock at runtime. Idealy only
426 * adjust_managed_page_count() should be used instead of directly
427 * touching zone->managed_pages and totalram_pages.
429 unsigned long managed_pages
;
430 unsigned long spanned_pages
;
431 unsigned long present_pages
;
435 #ifdef CONFIG_MEMORY_ISOLATION
437 * Number of isolated pageblock. It is used to solve incorrect
438 * freepage counting problem due to racy retrieving migratetype
439 * of pageblock. Protected by zone->lock.
441 unsigned long nr_isolate_pageblock
;
444 #ifdef CONFIG_MEMORY_HOTPLUG
445 /* see spanned/present_pages for more description */
446 seqlock_t span_seqlock
;
450 * wait_table -- the array holding the hash table
451 * wait_table_hash_nr_entries -- the size of the hash table array
452 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
454 * The purpose of all these is to keep track of the people
455 * waiting for a page to become available and make them
456 * runnable again when possible. The trouble is that this
457 * consumes a lot of space, especially when so few things
458 * wait on pages at a given time. So instead of using
459 * per-page waitqueues, we use a waitqueue hash table.
461 * The bucket discipline is to sleep on the same queue when
462 * colliding and wake all in that wait queue when removing.
463 * When something wakes, it must check to be sure its page is
464 * truly available, a la thundering herd. The cost of a
465 * collision is great, but given the expected load of the
466 * table, they should be so rare as to be outweighed by the
467 * benefits from the saved space.
469 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
470 * primary users of these fields, and in mm/page_alloc.c
471 * free_area_init_core() performs the initialization of them.
473 wait_queue_head_t
*wait_table
;
474 unsigned long wait_table_hash_nr_entries
;
475 unsigned long wait_table_bits
;
478 /* free areas of different sizes */
479 struct free_area free_area
[MAX_ORDER
];
481 /* zone flags, see below */
484 /* Write-intensive fields used from the page allocator */
489 /* Write-intensive fields used by page reclaim */
491 /* Fields commonly accessed by the page reclaim scanner */
493 struct lruvec lruvec
;
496 * When free pages are below this point, additional steps are taken
497 * when reading the number of free pages to avoid per-cpu counter
498 * drift allowing watermarks to be breached
500 unsigned long percpu_drift_mark
;
502 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
503 /* pfn where compaction free scanner should start */
504 unsigned long compact_cached_free_pfn
;
505 /* pfn where async and sync compaction migration scanner should start */
506 unsigned long compact_cached_migrate_pfn
[2];
509 #ifdef CONFIG_COMPACTION
511 * On compaction failure, 1<<compact_defer_shift compactions
512 * are skipped before trying again. The number attempted since
513 * last failure is tracked with compact_considered.
515 unsigned int compact_considered
;
516 unsigned int compact_defer_shift
;
517 int compact_order_failed
;
520 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
521 /* Set to true when the PG_migrate_skip bits should be cleared */
522 bool compact_blockskip_flush
;
528 /* Zone statistics */
529 atomic_long_t vm_stat
[NR_VM_ZONE_STAT_ITEMS
];
530 } ____cacheline_internodealigned_in_smp
;
533 ZONE_RECLAIM_LOCKED
, /* prevents concurrent reclaim */
534 ZONE_OOM_LOCKED
, /* zone is in OOM killer zonelist */
535 ZONE_CONGESTED
, /* zone has many dirty pages backed by
538 ZONE_DIRTY
, /* reclaim scanning has recently found
539 * many dirty file pages at the tail
542 ZONE_WRITEBACK
, /* reclaim scanning has recently found
543 * many pages under writeback
545 ZONE_FAIR_DEPLETED
, /* fair zone policy batch depleted */
548 static inline unsigned long zone_end_pfn(const struct zone
*zone
)
550 return zone
->zone_start_pfn
+ zone
->spanned_pages
;
553 static inline bool zone_spans_pfn(const struct zone
*zone
, unsigned long pfn
)
555 return zone
->zone_start_pfn
<= pfn
&& pfn
< zone_end_pfn(zone
);
558 static inline bool zone_is_initialized(struct zone
*zone
)
560 return !!zone
->wait_table
;
563 static inline bool zone_is_empty(struct zone
*zone
)
565 return zone
->spanned_pages
== 0;
569 * The "priority" of VM scanning is how much of the queues we will scan in one
570 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
571 * queues ("queue_length >> 12") during an aging round.
573 #define DEF_PRIORITY 12
575 /* Maximum number of zones on a zonelist */
576 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
579 ZONELIST_FALLBACK
, /* zonelist with fallback */
582 * The NUMA zonelists are doubled because we need zonelists that
583 * restrict the allocations to a single node for __GFP_THISNODE.
585 ZONELIST_NOFALLBACK
, /* zonelist without fallback (__GFP_THISNODE) */
591 * This struct contains information about a zone in a zonelist. It is stored
592 * here to avoid dereferences into large structures and lookups of tables
595 struct zone
*zone
; /* Pointer to actual zone */
596 int zone_idx
; /* zone_idx(zoneref->zone) */
600 * One allocation request operates on a zonelist. A zonelist
601 * is a list of zones, the first one is the 'goal' of the
602 * allocation, the other zones are fallback zones, in decreasing
605 * To speed the reading of the zonelist, the zonerefs contain the zone index
606 * of the entry being read. Helper functions to access information given
607 * a struct zoneref are
609 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
610 * zonelist_zone_idx() - Return the index of the zone for an entry
611 * zonelist_node_idx() - Return the index of the node for an entry
614 struct zoneref _zonerefs
[MAX_ZONES_PER_ZONELIST
+ 1];
617 #ifndef CONFIG_DISCONTIGMEM
618 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
619 extern struct page
*mem_map
;
623 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
624 * (mostly NUMA machines?) to denote a higher-level memory zone than the
627 * On NUMA machines, each NUMA node would have a pg_data_t to describe
628 * it's memory layout.
630 * Memory statistics and page replacement data structures are maintained on a
634 typedef struct pglist_data
{
635 struct zone node_zones
[MAX_NR_ZONES
];
636 struct zonelist node_zonelists
[MAX_ZONELISTS
];
638 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
639 struct page
*node_mem_map
;
640 #ifdef CONFIG_PAGE_EXTENSION
641 struct page_ext
*node_page_ext
;
644 #ifndef CONFIG_NO_BOOTMEM
645 struct bootmem_data
*bdata
;
647 #ifdef CONFIG_MEMORY_HOTPLUG
649 * Must be held any time you expect node_start_pfn, node_present_pages
650 * or node_spanned_pages stay constant. Holding this will also
651 * guarantee that any pfn_valid() stays that way.
653 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
654 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
656 * Nests above zone->lock and zone->span_seqlock
658 spinlock_t node_size_lock
;
660 unsigned long node_start_pfn
;
661 unsigned long node_present_pages
; /* total number of physical pages */
662 unsigned long node_spanned_pages
; /* total size of physical page
663 range, including holes */
665 wait_queue_head_t kswapd_wait
;
666 wait_queue_head_t pfmemalloc_wait
;
667 struct task_struct
*kswapd
; /* Protected by
668 mem_hotplug_begin/end() */
669 int kswapd_max_order
;
670 enum zone_type classzone_idx
;
671 #ifdef CONFIG_COMPACTION
672 int kcompactd_max_order
;
673 enum zone_type kcompactd_classzone_idx
;
674 wait_queue_head_t kcompactd_wait
;
675 struct task_struct
*kcompactd
;
677 #ifdef CONFIG_NUMA_BALANCING
678 /* Lock serializing the migrate rate limiting window */
679 spinlock_t numabalancing_migrate_lock
;
681 /* Rate limiting time interval */
682 unsigned long numabalancing_migrate_next_window
;
684 /* Number of pages migrated during the rate limiting time interval */
685 unsigned long numabalancing_migrate_nr_pages
;
688 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
690 * If memory initialisation on large machines is deferred then this
691 * is the first PFN that needs to be initialised.
693 unsigned long first_deferred_pfn
;
694 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
696 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
697 spinlock_t split_queue_lock
;
698 struct list_head split_queue
;
699 unsigned long split_queue_len
;
703 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
704 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
705 #ifdef CONFIG_FLAT_NODE_MEM_MAP
706 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
708 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
710 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
712 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
713 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
715 static inline unsigned long pgdat_end_pfn(pg_data_t
*pgdat
)
717 return pgdat
->node_start_pfn
+ pgdat
->node_spanned_pages
;
720 static inline bool pgdat_is_empty(pg_data_t
*pgdat
)
722 return !pgdat
->node_start_pfn
&& !pgdat
->node_spanned_pages
;
725 static inline int zone_id(const struct zone
*zone
)
727 struct pglist_data
*pgdat
= zone
->zone_pgdat
;
729 return zone
- pgdat
->node_zones
;
732 #ifdef CONFIG_ZONE_DEVICE
733 static inline bool is_dev_zone(const struct zone
*zone
)
735 return zone_id(zone
) == ZONE_DEVICE
;
738 static inline bool is_dev_zone(const struct zone
*zone
)
744 #include <linux/memory_hotplug.h>
746 extern struct mutex zonelists_mutex
;
747 void build_all_zonelists(pg_data_t
*pgdat
, struct zone
*zone
);
748 void wakeup_kswapd(struct zone
*zone
, int order
, enum zone_type classzone_idx
);
749 bool zone_watermark_ok(struct zone
*z
, unsigned int order
,
750 unsigned long mark
, int classzone_idx
, int alloc_flags
);
751 bool zone_watermark_ok_safe(struct zone
*z
, unsigned int order
,
752 unsigned long mark
, int classzone_idx
);
753 enum memmap_context
{
757 extern int init_currently_empty_zone(struct zone
*zone
, unsigned long start_pfn
,
760 extern void lruvec_init(struct lruvec
*lruvec
);
762 static inline struct zone
*lruvec_zone(struct lruvec
*lruvec
)
767 return container_of(lruvec
, struct zone
, lruvec
);
771 extern unsigned long lruvec_lru_size(struct lruvec
*lruvec
, enum lru_list lru
);
773 #ifdef CONFIG_HAVE_MEMORY_PRESENT
774 void memory_present(int nid
, unsigned long start
, unsigned long end
);
776 static inline void memory_present(int nid
, unsigned long start
, unsigned long end
) {}
779 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
780 int local_memory_node(int node_id
);
782 static inline int local_memory_node(int node_id
) { return node_id
; };
785 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
786 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
790 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
792 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
794 static inline int populated_zone(struct zone
*zone
)
796 return (!!zone
->present_pages
);
799 extern int movable_zone
;
801 #ifdef CONFIG_HIGHMEM
802 static inline int zone_movable_is_highmem(void)
804 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
805 return movable_zone
== ZONE_HIGHMEM
;
807 return (ZONE_MOVABLE
- 1) == ZONE_HIGHMEM
;
812 static inline int is_highmem_idx(enum zone_type idx
)
814 #ifdef CONFIG_HIGHMEM
815 return (idx
== ZONE_HIGHMEM
||
816 (idx
== ZONE_MOVABLE
&& zone_movable_is_highmem()));
823 * is_highmem - helper function to quickly check if a struct zone is a
824 * highmem zone or not. This is an attempt to keep references
825 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
826 * @zone - pointer to struct zone variable
828 static inline int is_highmem(struct zone
*zone
)
830 #ifdef CONFIG_HIGHMEM
831 int zone_off
= (char *)zone
- (char *)zone
->zone_pgdat
->node_zones
;
832 return zone_off
== ZONE_HIGHMEM
* sizeof(*zone
) ||
833 (zone_off
== ZONE_MOVABLE
* sizeof(*zone
) &&
834 zone_movable_is_highmem());
840 /* These two functions are used to setup the per zone pages min values */
842 int min_free_kbytes_sysctl_handler(struct ctl_table
*, int,
843 void __user
*, size_t *, loff_t
*);
844 int watermark_scale_factor_sysctl_handler(struct ctl_table
*, int,
845 void __user
*, size_t *, loff_t
*);
846 extern int sysctl_lowmem_reserve_ratio
[MAX_NR_ZONES
-1];
847 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table
*, int,
848 void __user
*, size_t *, loff_t
*);
849 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table
*, int,
850 void __user
*, size_t *, loff_t
*);
851 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table
*, int,
852 void __user
*, size_t *, loff_t
*);
853 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table
*, int,
854 void __user
*, size_t *, loff_t
*);
856 extern int numa_zonelist_order_handler(struct ctl_table
*, int,
857 void __user
*, size_t *, loff_t
*);
858 extern char numa_zonelist_order
[];
859 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
861 #ifndef CONFIG_NEED_MULTIPLE_NODES
863 extern struct pglist_data contig_page_data
;
864 #define NODE_DATA(nid) (&contig_page_data)
865 #define NODE_MEM_MAP(nid) mem_map
867 #else /* CONFIG_NEED_MULTIPLE_NODES */
869 #include <asm/mmzone.h>
871 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
873 extern struct pglist_data
*first_online_pgdat(void);
874 extern struct pglist_data
*next_online_pgdat(struct pglist_data
*pgdat
);
875 extern struct zone
*next_zone(struct zone
*zone
);
878 * for_each_online_pgdat - helper macro to iterate over all online nodes
879 * @pgdat - pointer to a pg_data_t variable
881 #define for_each_online_pgdat(pgdat) \
882 for (pgdat = first_online_pgdat(); \
884 pgdat = next_online_pgdat(pgdat))
886 * for_each_zone - helper macro to iterate over all memory zones
887 * @zone - pointer to struct zone variable
889 * The user only needs to declare the zone variable, for_each_zone
892 #define for_each_zone(zone) \
893 for (zone = (first_online_pgdat())->node_zones; \
895 zone = next_zone(zone))
897 #define for_each_populated_zone(zone) \
898 for (zone = (first_online_pgdat())->node_zones; \
900 zone = next_zone(zone)) \
901 if (!populated_zone(zone)) \
905 static inline struct zone
*zonelist_zone(struct zoneref
*zoneref
)
907 return zoneref
->zone
;
910 static inline int zonelist_zone_idx(struct zoneref
*zoneref
)
912 return zoneref
->zone_idx
;
915 static inline int zonelist_node_idx(struct zoneref
*zoneref
)
918 /* zone_to_nid not available in this context */
919 return zoneref
->zone
->node
;
922 #endif /* CONFIG_NUMA */
926 * 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
927 * @z - The cursor used as a starting point for the search
928 * @highest_zoneidx - The zone index of the highest zone to return
929 * @nodes - An optional nodemask to filter the zonelist with
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
,
942 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
943 * @zonelist - The zonelist to search for a suitable zone
944 * @highest_zoneidx - The zone index of the highest zone to return
945 * @nodes - An optional nodemask to filter the zonelist with
946 * @zone - The first suitable zone found is returned via this parameter
948 * This function returns the first zone at or below a given zone index that is
949 * within the allowed nodemask. The zoneref returned is a cursor that can be
950 * used to iterate the zonelist with next_zones_zonelist by advancing it by
951 * one before calling.
953 static inline struct zoneref
*first_zones_zonelist(struct zonelist
*zonelist
,
954 enum zone_type highest_zoneidx
,
958 struct zoneref
*z
= next_zones_zonelist(zonelist
->_zonerefs
,
959 highest_zoneidx
, nodes
);
960 *zone
= zonelist_zone(z
);
965 * 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
966 * @zone - The current zone in the iterator
967 * @z - The current pointer within zonelist->zones being iterated
968 * @zlist - The zonelist being iterated
969 * @highidx - The zone index of the highest zone to return
970 * @nodemask - Nodemask allowed by the allocator
972 * This iterator iterates though all zones at or below a given zone index and
973 * within a given nodemask
975 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
976 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
978 z = next_zones_zonelist(++z, highidx, nodemask), \
979 zone = zonelist_zone(z)) \
982 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
983 * @zone - The current zone in the iterator
984 * @z - The current pointer within zonelist->zones being iterated
985 * @zlist - The zonelist being iterated
986 * @highidx - The zone index of the highest zone to return
988 * This iterator iterates though all zones at or below a given zone index.
990 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
991 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
993 #ifdef CONFIG_SPARSEMEM
994 #include <asm/sparsemem.h>
997 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
998 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
999 static inline unsigned long early_pfn_to_nid(unsigned long pfn
)
1005 #ifdef CONFIG_FLATMEM
1006 #define pfn_to_nid(pfn) (0)
1009 #ifdef CONFIG_SPARSEMEM
1012 * SECTION_SHIFT #bits space required to store a section #
1014 * PA_SECTION_SHIFT physical address to/from section number
1015 * PFN_SECTION_SHIFT pfn to/from section number
1017 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1018 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1020 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1022 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1023 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1025 #define SECTION_BLOCKFLAGS_BITS \
1026 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1028 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1029 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1032 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1033 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1035 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1036 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1040 struct mem_section
{
1042 * This is, logically, a pointer to an array of struct
1043 * pages. However, it is stored with some other magic.
1044 * (see sparse.c::sparse_init_one_section())
1046 * Additionally during early boot we encode node id of
1047 * the location of the section here to guide allocation.
1048 * (see sparse.c::memory_present())
1050 * Making it a UL at least makes someone do a cast
1051 * before using it wrong.
1053 unsigned long section_mem_map
;
1055 /* See declaration of similar field in struct zone */
1056 unsigned long *pageblock_flags
;
1057 #ifdef CONFIG_PAGE_EXTENSION
1059 * If !SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1060 * section. (see page_ext.h about this.)
1062 struct page_ext
*page_ext
;
1066 * WARNING: mem_section must be a power-of-2 in size for the
1067 * calculation and use of SECTION_ROOT_MASK to make sense.
1071 #ifdef CONFIG_SPARSEMEM_EXTREME
1072 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1074 #define SECTIONS_PER_ROOT 1
1077 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1078 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1079 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1081 #ifdef CONFIG_SPARSEMEM_EXTREME
1082 extern struct mem_section
*mem_section
[NR_SECTION_ROOTS
];
1084 extern struct mem_section mem_section
[NR_SECTION_ROOTS
][SECTIONS_PER_ROOT
];
1087 static inline struct mem_section
*__nr_to_section(unsigned long nr
)
1089 if (!mem_section
[SECTION_NR_TO_ROOT(nr
)])
1091 return &mem_section
[SECTION_NR_TO_ROOT(nr
)][nr
& SECTION_ROOT_MASK
];
1093 extern int __section_nr(struct mem_section
* ms
);
1094 extern unsigned long usemap_size(void);
1097 * We use the lower bits of the mem_map pointer to store
1098 * a little bit of information. There should be at least
1099 * 3 bits here due to 32-bit alignment.
1101 #define SECTION_MARKED_PRESENT (1UL<<0)
1102 #define SECTION_HAS_MEM_MAP (1UL<<1)
1103 #define SECTION_MAP_LAST_BIT (1UL<<2)
1104 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1105 #define SECTION_NID_SHIFT 2
1107 static inline struct page
*__section_mem_map_addr(struct mem_section
*section
)
1109 unsigned long map
= section
->section_mem_map
;
1110 map
&= SECTION_MAP_MASK
;
1111 return (struct page
*)map
;
1114 static inline int present_section(struct mem_section
*section
)
1116 return (section
&& (section
->section_mem_map
& SECTION_MARKED_PRESENT
));
1119 static inline int present_section_nr(unsigned long nr
)
1121 return present_section(__nr_to_section(nr
));
1124 static inline int valid_section(struct mem_section
*section
)
1126 return (section
&& (section
->section_mem_map
& SECTION_HAS_MEM_MAP
));
1129 static inline int valid_section_nr(unsigned long nr
)
1131 return valid_section(__nr_to_section(nr
));
1134 static inline struct mem_section
*__pfn_to_section(unsigned long pfn
)
1136 return __nr_to_section(pfn_to_section_nr(pfn
));
1139 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1140 static inline int pfn_valid(unsigned long pfn
)
1142 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1144 return valid_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1148 static inline int pfn_present(unsigned long pfn
)
1150 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1152 return present_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1156 * These are _only_ used during initialisation, therefore they
1157 * can use __initdata ... They could have names to indicate
1161 #define pfn_to_nid(pfn) \
1163 unsigned long __pfn_to_nid_pfn = (pfn); \
1164 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1167 #define pfn_to_nid(pfn) (0)
1170 #define early_pfn_valid(pfn) pfn_valid(pfn)
1171 void sparse_init(void);
1173 #define sparse_init() do {} while (0)
1174 #define sparse_index_init(_sec, _nid) do {} while (0)
1175 #endif /* CONFIG_SPARSEMEM */
1178 * During memory init memblocks map pfns to nids. The search is expensive and
1179 * this caches recent lookups. The implementation of __early_pfn_to_nid
1180 * may treat start/end as pfns or sections.
1182 struct mminit_pfnnid_cache
{
1183 unsigned long last_start
;
1184 unsigned long last_end
;
1188 #ifndef early_pfn_valid
1189 #define early_pfn_valid(pfn) (1)
1192 void memory_present(int nid
, unsigned long start
, unsigned long end
);
1193 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
1196 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1197 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1198 * pfn_valid_within() should be used in this case; we optimise this away
1199 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1201 #ifdef CONFIG_HOLES_IN_ZONE
1202 #define pfn_valid_within(pfn) pfn_valid(pfn)
1204 #define pfn_valid_within(pfn) (1)
1207 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1209 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1210 * associated with it or not. In FLATMEM, it is expected that holes always
1211 * have valid memmap as long as there is valid PFNs either side of the hole.
1212 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1215 * However, an ARM, and maybe other embedded architectures in the future
1216 * free memmap backing holes to save memory on the assumption the memmap is
1217 * never used. The page_zone linkages are then broken even though pfn_valid()
1218 * returns true. A walker of the full memmap must then do this additional
1219 * check to ensure the memmap they are looking at is sane by making sure
1220 * the zone and PFN linkages are still valid. This is expensive, but walkers
1221 * of the full memmap are extremely rare.
1223 bool memmap_valid_within(unsigned long pfn
,
1224 struct page
*page
, struct zone
*zone
);
1226 static inline bool memmap_valid_within(unsigned long pfn
,
1227 struct page
*page
, struct zone
*zone
)
1231 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1233 #endif /* !__GENERATING_BOUNDS.H */
1234 #endif /* !__ASSEMBLY__ */
1235 #endif /* _LINUX_MMZONE_H */