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)
71 # define is_migrate_cma_page(_page) (get_pageblock_migratetype(_page) == MIGRATE_CMA)
73 # define is_migrate_cma(migratetype) false
74 # define is_migrate_cma_page(_page) false
77 #define for_each_migratetype_order(order, type) \
78 for (order = 0; order < MAX_ORDER; order++) \
79 for (type = 0; type < MIGRATE_TYPES; type++)
81 extern int page_group_by_mobility_disabled
;
83 #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
84 #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
86 #define get_pageblock_migratetype(page) \
87 get_pfnblock_flags_mask(page, page_to_pfn(page), \
88 PB_migrate_end, MIGRATETYPE_MASK)
91 struct list_head free_list
[MIGRATE_TYPES
];
92 unsigned long nr_free
;
98 * zone->lock and the zone lru_lock are two of the hottest locks in the kernel.
99 * So add a wild amount of padding here to ensure that they fall into separate
100 * cachelines. There are very few zone structures in the machine, so space
101 * consumption is not a concern here.
103 #if defined(CONFIG_SMP)
104 struct zone_padding
{
106 } ____cacheline_internodealigned_in_smp
;
107 #define ZONE_PADDING(name) struct zone_padding name;
109 #define ZONE_PADDING(name)
112 enum zone_stat_item
{
113 /* First 128 byte cacheline (assuming 64 bit words) */
115 NR_ZONE_LRU_BASE
, /* Used only for compaction and reclaim retry */
116 NR_ZONE_INACTIVE_ANON
= NR_ZONE_LRU_BASE
,
118 NR_ZONE_INACTIVE_FILE
,
121 NR_ZONE_WRITE_PENDING
, /* Count of dirty, writeback and unstable pages */
122 NR_MLOCK
, /* mlock()ed pages found and moved off LRU */
124 NR_SLAB_UNRECLAIMABLE
,
125 NR_PAGETABLE
, /* used for pagetables */
126 NR_KERNEL_STACK_KB
, /* measured in KiB */
127 /* Second 128 byte cacheline */
129 #if IS_ENABLED(CONFIG_ZSMALLOC)
130 NR_ZSPAGES
, /* allocated in zsmalloc */
133 NUMA_HIT
, /* allocated in intended node */
134 NUMA_MISS
, /* allocated in non intended node */
135 NUMA_FOREIGN
, /* was intended here, hit elsewhere */
136 NUMA_INTERLEAVE_HIT
, /* interleaver preferred this zone */
137 NUMA_LOCAL
, /* allocation from local node */
138 NUMA_OTHER
, /* allocation from other node */
141 NR_VM_ZONE_STAT_ITEMS
};
143 enum node_stat_item
{
145 NR_INACTIVE_ANON
= NR_LRU_BASE
, /* must match order of LRU_[IN]ACTIVE */
146 NR_ACTIVE_ANON
, /* " " " " " */
147 NR_INACTIVE_FILE
, /* " " " " " */
148 NR_ACTIVE_FILE
, /* " " " " " */
149 NR_UNEVICTABLE
, /* " " " " " */
150 NR_ISOLATED_ANON
, /* Temporary isolated pages from anon lru */
151 NR_ISOLATED_FILE
, /* Temporary isolated pages from file lru */
154 WORKINGSET_NODERECLAIM
,
155 NR_ANON_MAPPED
, /* Mapped anonymous pages */
156 NR_FILE_MAPPED
, /* pagecache pages mapped into pagetables.
157 only modified from process context */
161 NR_WRITEBACK_TEMP
, /* Writeback using temporary buffers */
162 NR_SHMEM
, /* shmem pages (included tmpfs/GEM pages) */
166 NR_UNSTABLE_NFS
, /* NFS unstable pages */
168 NR_VMSCAN_IMMEDIATE
, /* Prioritise for reclaim when writeback ends */
169 NR_DIRTIED
, /* page dirtyings since bootup */
170 NR_WRITTEN
, /* page writings since bootup */
171 NR_VM_NODE_STAT_ITEMS
175 * We do arithmetic on the LRU lists in various places in the code,
176 * so it is important to keep the active lists LRU_ACTIVE higher in
177 * the array than the corresponding inactive lists, and to keep
178 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
180 * This has to be kept in sync with the statistics in zone_stat_item
181 * above and the descriptions in vmstat_text in mm/vmstat.c
188 LRU_INACTIVE_ANON
= LRU_BASE
,
189 LRU_ACTIVE_ANON
= LRU_BASE
+ LRU_ACTIVE
,
190 LRU_INACTIVE_FILE
= LRU_BASE
+ LRU_FILE
,
191 LRU_ACTIVE_FILE
= LRU_BASE
+ LRU_FILE
+ LRU_ACTIVE
,
196 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
198 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
200 static inline int is_file_lru(enum lru_list lru
)
202 return (lru
== LRU_INACTIVE_FILE
|| lru
== LRU_ACTIVE_FILE
);
205 static inline int is_active_lru(enum lru_list lru
)
207 return (lru
== LRU_ACTIVE_ANON
|| lru
== LRU_ACTIVE_FILE
);
210 struct zone_reclaim_stat
{
212 * The pageout code in vmscan.c keeps track of how many of the
213 * mem/swap backed and file backed pages are referenced.
214 * The higher the rotated/scanned ratio, the more valuable
217 * The anon LRU stats live in [0], file LRU stats in [1]
219 unsigned long recent_rotated
[2];
220 unsigned long recent_scanned
[2];
224 struct list_head lists
[NR_LRU_LISTS
];
225 struct zone_reclaim_stat reclaim_stat
;
226 /* Evictions & activations on the inactive file list */
227 atomic_long_t inactive_age
;
229 struct pglist_data
*pgdat
;
233 /* Mask used at gathering information at once (see memcontrol.c) */
234 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
235 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
236 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
238 /* Isolate unmapped file */
239 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
240 /* Isolate for asynchronous migration */
241 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
242 /* Isolate unevictable pages */
243 #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
245 /* LRU Isolation modes. */
246 typedef unsigned __bitwise isolate_mode_t
;
248 enum zone_watermarks
{
255 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
256 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
257 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
259 struct per_cpu_pages
{
260 int count
; /* number of pages in the list */
261 int high
; /* high watermark, emptying needed */
262 int batch
; /* chunk size for buddy add/remove */
264 /* Lists of pages, one per migrate type stored on the pcp-lists */
265 struct list_head lists
[MIGRATE_PCPTYPES
];
268 struct per_cpu_pageset
{
269 struct per_cpu_pages pcp
;
275 s8 vm_stat_diff
[NR_VM_ZONE_STAT_ITEMS
];
279 struct per_cpu_nodestat
{
281 s8 vm_node_stat_diff
[NR_VM_NODE_STAT_ITEMS
];
284 #endif /* !__GENERATING_BOUNDS.H */
287 #ifdef CONFIG_ZONE_DMA
289 * ZONE_DMA is used when there are devices that are not able
290 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
291 * carve out the portion of memory that is needed for these devices.
292 * The range is arch specific.
297 * ---------------------------
298 * parisc, ia64, sparc <4G
301 * alpha Unlimited or 0-16MB.
303 * i386, x86_64 and multiple other arches
308 #ifdef CONFIG_ZONE_DMA32
310 * x86_64 needs two ZONE_DMAs because it supports devices that are
311 * only able to do DMA to the lower 16M but also 32 bit devices that
312 * can only do DMA areas below 4G.
317 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
318 * performed on pages in ZONE_NORMAL if the DMA devices support
319 * transfers to all addressable memory.
322 #ifdef CONFIG_HIGHMEM
324 * A memory area that is only addressable by the kernel through
325 * mapping portions into its own address space. This is for example
326 * used by i386 to allow the kernel to address the memory beyond
327 * 900MB. The kernel will set up special mappings (page
328 * table entries on i386) for each page that the kernel needs to
334 #ifdef CONFIG_ZONE_DEVICE
341 #ifndef __GENERATING_BOUNDS_H
344 /* Read-mostly fields */
346 /* zone watermarks, access with *_wmark_pages(zone) macros */
347 unsigned long watermark
[NR_WMARK
];
349 unsigned long nr_reserved_highatomic
;
352 * We don't know if the memory that we're going to allocate will be
353 * freeable or/and it will be released eventually, so to avoid totally
354 * wasting several GB of ram we must reserve some of the lower zone
355 * memory (otherwise we risk to run OOM on the lower zones despite
356 * there being tons of freeable ram on the higher zones). This array is
357 * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
360 long lowmem_reserve
[MAX_NR_ZONES
];
365 struct pglist_data
*zone_pgdat
;
366 struct per_cpu_pageset __percpu
*pageset
;
368 #ifndef CONFIG_SPARSEMEM
370 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
371 * In SPARSEMEM, this map is stored in struct mem_section
373 unsigned long *pageblock_flags
;
374 #endif /* CONFIG_SPARSEMEM */
376 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
377 unsigned long zone_start_pfn
;
380 * spanned_pages is the total pages spanned by the zone, including
381 * holes, which is calculated as:
382 * spanned_pages = zone_end_pfn - zone_start_pfn;
384 * present_pages is physical pages existing within the zone, which
386 * present_pages = spanned_pages - absent_pages(pages in holes);
388 * managed_pages is present pages managed by the buddy system, which
389 * is calculated as (reserved_pages includes pages allocated by the
390 * bootmem allocator):
391 * managed_pages = present_pages - reserved_pages;
393 * So present_pages may be used by memory hotplug or memory power
394 * management logic to figure out unmanaged pages by checking
395 * (present_pages - managed_pages). And managed_pages should be used
396 * by page allocator and vm scanner to calculate all kinds of watermarks
401 * zone_start_pfn and spanned_pages are protected by span_seqlock.
402 * It is a seqlock because it has to be read outside of zone->lock,
403 * and it is done in the main allocator path. But, it is written
404 * quite infrequently.
406 * The span_seq lock is declared along with zone->lock because it is
407 * frequently read in proximity to zone->lock. It's good to
408 * give them a chance of being in the same cacheline.
410 * Write access to present_pages at runtime should be protected by
411 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
412 * present_pages should get_online_mems() to get a stable value.
414 * Read access to managed_pages should be safe because it's unsigned
415 * long. Write access to zone->managed_pages and totalram_pages are
416 * protected by managed_page_count_lock at runtime. Idealy only
417 * adjust_managed_page_count() should be used instead of directly
418 * touching zone->managed_pages and totalram_pages.
420 unsigned long managed_pages
;
421 unsigned long spanned_pages
;
422 unsigned long present_pages
;
426 #ifdef CONFIG_MEMORY_ISOLATION
428 * Number of isolated pageblock. It is used to solve incorrect
429 * freepage counting problem due to racy retrieving migratetype
430 * of pageblock. Protected by zone->lock.
432 unsigned long nr_isolate_pageblock
;
435 #ifdef CONFIG_MEMORY_HOTPLUG
436 /* see spanned/present_pages for more description */
437 seqlock_t span_seqlock
;
442 /* Write-intensive fields used from the page allocator */
445 /* free areas of different sizes */
446 struct free_area free_area
[MAX_ORDER
];
448 /* zone flags, see below */
451 /* Primarily protects free_area */
454 /* Write-intensive fields used by compaction and vmstats. */
458 * When free pages are below this point, additional steps are taken
459 * when reading the number of free pages to avoid per-cpu counter
460 * drift allowing watermarks to be breached
462 unsigned long percpu_drift_mark
;
464 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
465 /* pfn where compaction free scanner should start */
466 unsigned long compact_cached_free_pfn
;
467 /* pfn where async and sync compaction migration scanner should start */
468 unsigned long compact_cached_migrate_pfn
[2];
471 #ifdef CONFIG_COMPACTION
473 * On compaction failure, 1<<compact_defer_shift compactions
474 * are skipped before trying again. The number attempted since
475 * last failure is tracked with compact_considered.
477 unsigned int compact_considered
;
478 unsigned int compact_defer_shift
;
479 int compact_order_failed
;
482 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
483 /* Set to true when the PG_migrate_skip bits should be cleared */
484 bool compact_blockskip_flush
;
490 /* Zone statistics */
491 atomic_long_t vm_stat
[NR_VM_ZONE_STAT_ITEMS
];
492 } ____cacheline_internodealigned_in_smp
;
495 PGDAT_CONGESTED
, /* pgdat has many dirty pages backed by
498 PGDAT_DIRTY
, /* reclaim scanning has recently found
499 * many dirty file pages at the tail
502 PGDAT_WRITEBACK
, /* reclaim scanning has recently found
503 * many pages under writeback
505 PGDAT_RECLAIM_LOCKED
, /* prevents concurrent reclaim */
508 static inline unsigned long zone_end_pfn(const struct zone
*zone
)
510 return zone
->zone_start_pfn
+ zone
->spanned_pages
;
513 static inline bool zone_spans_pfn(const struct zone
*zone
, unsigned long pfn
)
515 return zone
->zone_start_pfn
<= pfn
&& pfn
< zone_end_pfn(zone
);
518 static inline bool zone_is_initialized(struct zone
*zone
)
520 return zone
->initialized
;
523 static inline bool zone_is_empty(struct zone
*zone
)
525 return zone
->spanned_pages
== 0;
529 * The "priority" of VM scanning is how much of the queues we will scan in one
530 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
531 * queues ("queue_length >> 12") during an aging round.
533 #define DEF_PRIORITY 12
535 /* Maximum number of zones on a zonelist */
536 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
539 ZONELIST_FALLBACK
, /* zonelist with fallback */
542 * The NUMA zonelists are doubled because we need zonelists that
543 * restrict the allocations to a single node for __GFP_THISNODE.
545 ZONELIST_NOFALLBACK
, /* zonelist without fallback (__GFP_THISNODE) */
551 * This struct contains information about a zone in a zonelist. It is stored
552 * here to avoid dereferences into large structures and lookups of tables
555 struct zone
*zone
; /* Pointer to actual zone */
556 int zone_idx
; /* zone_idx(zoneref->zone) */
560 * One allocation request operates on a zonelist. A zonelist
561 * is a list of zones, the first one is the 'goal' of the
562 * allocation, the other zones are fallback zones, in decreasing
565 * To speed the reading of the zonelist, the zonerefs contain the zone index
566 * of the entry being read. Helper functions to access information given
567 * a struct zoneref are
569 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
570 * zonelist_zone_idx() - Return the index of the zone for an entry
571 * zonelist_node_idx() - Return the index of the node for an entry
574 struct zoneref _zonerefs
[MAX_ZONES_PER_ZONELIST
+ 1];
577 #ifndef CONFIG_DISCONTIGMEM
578 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
579 extern struct page
*mem_map
;
583 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
584 * (mostly NUMA machines?) to denote a higher-level memory zone than the
587 * On NUMA machines, each NUMA node would have a pg_data_t to describe
588 * it's memory layout.
590 * Memory statistics and page replacement data structures are maintained on a
594 typedef struct pglist_data
{
595 struct zone node_zones
[MAX_NR_ZONES
];
596 struct zonelist node_zonelists
[MAX_ZONELISTS
];
598 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
599 struct page
*node_mem_map
;
600 #ifdef CONFIG_PAGE_EXTENSION
601 struct page_ext
*node_page_ext
;
604 #ifndef CONFIG_NO_BOOTMEM
605 struct bootmem_data
*bdata
;
607 #ifdef CONFIG_MEMORY_HOTPLUG
609 * Must be held any time you expect node_start_pfn, node_present_pages
610 * or node_spanned_pages stay constant. Holding this will also
611 * guarantee that any pfn_valid() stays that way.
613 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
614 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
616 * Nests above zone->lock and zone->span_seqlock
618 spinlock_t node_size_lock
;
620 unsigned long node_start_pfn
;
621 unsigned long node_present_pages
; /* total number of physical pages */
622 unsigned long node_spanned_pages
; /* total size of physical page
623 range, including holes */
625 wait_queue_head_t kswapd_wait
;
626 wait_queue_head_t pfmemalloc_wait
;
627 struct task_struct
*kswapd
; /* Protected by
628 mem_hotplug_begin/end() */
630 enum zone_type kswapd_classzone_idx
;
632 int kswapd_failures
; /* Number of 'reclaimed == 0' runs */
634 #ifdef CONFIG_COMPACTION
635 int kcompactd_max_order
;
636 enum zone_type kcompactd_classzone_idx
;
637 wait_queue_head_t kcompactd_wait
;
638 struct task_struct
*kcompactd
;
640 #ifdef CONFIG_NUMA_BALANCING
641 /* Lock serializing the migrate rate limiting window */
642 spinlock_t numabalancing_migrate_lock
;
644 /* Rate limiting time interval */
645 unsigned long numabalancing_migrate_next_window
;
647 /* Number of pages migrated during the rate limiting time interval */
648 unsigned long numabalancing_migrate_nr_pages
;
651 * This is a per-node reserve of pages that are not available
652 * to userspace allocations.
654 unsigned long totalreserve_pages
;
658 * zone reclaim becomes active if more unmapped pages exist.
660 unsigned long min_unmapped_pages
;
661 unsigned long min_slab_pages
;
662 #endif /* CONFIG_NUMA */
664 /* Write-intensive fields used by page reclaim */
668 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
670 * If memory initialisation on large machines is deferred then this
671 * is the first PFN that needs to be initialised.
673 unsigned long first_deferred_pfn
;
674 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
676 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
677 spinlock_t split_queue_lock
;
678 struct list_head split_queue
;
679 unsigned long split_queue_len
;
682 /* Fields commonly accessed by the page reclaim scanner */
683 struct lruvec lruvec
;
686 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
687 * this node's LRU. Maintained by the pageout code.
689 unsigned int inactive_ratio
;
695 /* Per-node vmstats */
696 struct per_cpu_nodestat __percpu
*per_cpu_nodestats
;
697 atomic_long_t vm_stat
[NR_VM_NODE_STAT_ITEMS
];
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))
711 static inline spinlock_t
*zone_lru_lock(struct zone
*zone
)
713 return &zone
->zone_pgdat
->lru_lock
;
716 static inline struct lruvec
*node_lruvec(struct pglist_data
*pgdat
)
718 return &pgdat
->lruvec
;
721 static inline unsigned long pgdat_end_pfn(pg_data_t
*pgdat
)
723 return pgdat
->node_start_pfn
+ pgdat
->node_spanned_pages
;
726 static inline bool pgdat_is_empty(pg_data_t
*pgdat
)
728 return !pgdat
->node_start_pfn
&& !pgdat
->node_spanned_pages
;
731 static inline int zone_id(const struct zone
*zone
)
733 struct pglist_data
*pgdat
= zone
->zone_pgdat
;
735 return zone
- pgdat
->node_zones
;
738 #ifdef CONFIG_ZONE_DEVICE
739 static inline bool is_dev_zone(const struct zone
*zone
)
741 return zone_id(zone
) == ZONE_DEVICE
;
744 static inline bool is_dev_zone(const struct zone
*zone
)
750 #include <linux/memory_hotplug.h>
752 extern struct mutex zonelists_mutex
;
753 void build_all_zonelists(pg_data_t
*pgdat
, struct zone
*zone
);
754 void wakeup_kswapd(struct zone
*zone
, int order
, enum zone_type classzone_idx
);
755 bool __zone_watermark_ok(struct zone
*z
, unsigned int order
, unsigned long mark
,
756 int classzone_idx
, unsigned int alloc_flags
,
758 bool zone_watermark_ok(struct zone
*z
, unsigned int order
,
759 unsigned long mark
, int classzone_idx
,
760 unsigned int alloc_flags
);
761 bool zone_watermark_ok_safe(struct zone
*z
, unsigned int order
,
762 unsigned long mark
, int classzone_idx
);
763 enum memmap_context
{
767 extern int init_currently_empty_zone(struct zone
*zone
, unsigned long start_pfn
,
770 extern void lruvec_init(struct lruvec
*lruvec
);
772 static inline struct pglist_data
*lruvec_pgdat(struct lruvec
*lruvec
)
775 return lruvec
->pgdat
;
777 return container_of(lruvec
, struct pglist_data
, lruvec
);
781 extern unsigned long lruvec_lru_size(struct lruvec
*lruvec
, enum lru_list lru
, int zone_idx
);
783 #ifdef CONFIG_HAVE_MEMORY_PRESENT
784 void memory_present(int nid
, unsigned long start
, unsigned long end
);
786 static inline void memory_present(int nid
, unsigned long start
, unsigned long end
) {}
789 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
790 int local_memory_node(int node_id
);
792 static inline int local_memory_node(int node_id
) { return node_id
; };
795 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
796 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
800 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
802 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
805 * Returns true if a zone has pages managed by the buddy allocator.
806 * All the reclaim decisions have to use this function rather than
807 * populated_zone(). If the whole zone is reserved then we can easily
808 * end up with populated_zone() && !managed_zone().
810 static inline bool managed_zone(struct zone
*zone
)
812 return zone
->managed_pages
;
815 /* Returns true if a zone has memory */
816 static inline bool populated_zone(struct zone
*zone
)
818 return zone
->present_pages
;
821 extern int movable_zone
;
823 #ifdef CONFIG_HIGHMEM
824 static inline int zone_movable_is_highmem(void)
826 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
827 return movable_zone
== ZONE_HIGHMEM
;
829 return (ZONE_MOVABLE
- 1) == ZONE_HIGHMEM
;
834 static inline int is_highmem_idx(enum zone_type idx
)
836 #ifdef CONFIG_HIGHMEM
837 return (idx
== ZONE_HIGHMEM
||
838 (idx
== ZONE_MOVABLE
&& zone_movable_is_highmem()));
845 * is_highmem - helper function to quickly check if a struct zone is a
846 * highmem zone or not. This is an attempt to keep references
847 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
848 * @zone - pointer to struct zone variable
850 static inline int is_highmem(struct zone
*zone
)
852 #ifdef CONFIG_HIGHMEM
853 return is_highmem_idx(zone_idx(zone
));
859 /* These two functions are used to setup the per zone pages min values */
861 int min_free_kbytes_sysctl_handler(struct ctl_table
*, int,
862 void __user
*, size_t *, loff_t
*);
863 int watermark_scale_factor_sysctl_handler(struct ctl_table
*, int,
864 void __user
*, size_t *, loff_t
*);
865 extern int sysctl_lowmem_reserve_ratio
[MAX_NR_ZONES
-1];
866 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table
*, int,
867 void __user
*, size_t *, loff_t
*);
868 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table
*, int,
869 void __user
*, size_t *, loff_t
*);
870 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table
*, int,
871 void __user
*, size_t *, loff_t
*);
872 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table
*, int,
873 void __user
*, size_t *, loff_t
*);
875 extern int numa_zonelist_order_handler(struct ctl_table
*, int,
876 void __user
*, size_t *, loff_t
*);
877 extern char numa_zonelist_order
[];
878 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
880 #ifndef CONFIG_NEED_MULTIPLE_NODES
882 extern struct pglist_data contig_page_data
;
883 #define NODE_DATA(nid) (&contig_page_data)
884 #define NODE_MEM_MAP(nid) mem_map
886 #else /* CONFIG_NEED_MULTIPLE_NODES */
888 #include <asm/mmzone.h>
890 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
892 extern struct pglist_data
*first_online_pgdat(void);
893 extern struct pglist_data
*next_online_pgdat(struct pglist_data
*pgdat
);
894 extern struct zone
*next_zone(struct zone
*zone
);
897 * for_each_online_pgdat - helper macro to iterate over all online nodes
898 * @pgdat - pointer to a pg_data_t variable
900 #define for_each_online_pgdat(pgdat) \
901 for (pgdat = first_online_pgdat(); \
903 pgdat = next_online_pgdat(pgdat))
905 * for_each_zone - helper macro to iterate over all memory zones
906 * @zone - pointer to struct zone variable
908 * The user only needs to declare the zone variable, for_each_zone
911 #define for_each_zone(zone) \
912 for (zone = (first_online_pgdat())->node_zones; \
914 zone = next_zone(zone))
916 #define for_each_populated_zone(zone) \
917 for (zone = (first_online_pgdat())->node_zones; \
919 zone = next_zone(zone)) \
920 if (!populated_zone(zone)) \
924 static inline struct zone
*zonelist_zone(struct zoneref
*zoneref
)
926 return zoneref
->zone
;
929 static inline int zonelist_zone_idx(struct zoneref
*zoneref
)
931 return zoneref
->zone_idx
;
934 static inline int zonelist_node_idx(struct zoneref
*zoneref
)
937 /* zone_to_nid not available in this context */
938 return zoneref
->zone
->node
;
941 #endif /* CONFIG_NUMA */
944 struct zoneref
*__next_zones_zonelist(struct zoneref
*z
,
945 enum zone_type highest_zoneidx
,
949 * 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
950 * @z - The cursor used as a starting point for the search
951 * @highest_zoneidx - The zone index of the highest zone to return
952 * @nodes - An optional nodemask to filter the zonelist with
954 * This function returns the next zone at or below a given zone index that is
955 * within the allowed nodemask using a cursor as the starting point for the
956 * search. The zoneref returned is a cursor that represents the current zone
957 * being examined. It should be advanced by one before calling
958 * next_zones_zonelist again.
960 static __always_inline
struct zoneref
*next_zones_zonelist(struct zoneref
*z
,
961 enum zone_type highest_zoneidx
,
964 if (likely(!nodes
&& zonelist_zone_idx(z
) <= highest_zoneidx
))
966 return __next_zones_zonelist(z
, highest_zoneidx
, nodes
);
970 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
971 * @zonelist - The zonelist to search for a suitable zone
972 * @highest_zoneidx - The zone index of the highest zone to return
973 * @nodes - An optional nodemask to filter the zonelist with
974 * @return - Zoneref pointer for the first suitable zone found (see below)
976 * This function returns the first zone at or below a given zone index that is
977 * within the allowed nodemask. The zoneref returned is a cursor that can be
978 * used to iterate the zonelist with next_zones_zonelist by advancing it by
979 * one before calling.
981 * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is
982 * never NULL). This may happen either genuinely, or due to concurrent nodemask
983 * update due to cpuset modification.
985 static inline struct zoneref
*first_zones_zonelist(struct zonelist
*zonelist
,
986 enum zone_type highest_zoneidx
,
989 return next_zones_zonelist(zonelist
->_zonerefs
,
990 highest_zoneidx
, nodes
);
994 * 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
995 * @zone - The current zone in the iterator
996 * @z - The current pointer within zonelist->zones being iterated
997 * @zlist - The zonelist being iterated
998 * @highidx - The zone index of the highest zone to return
999 * @nodemask - Nodemask allowed by the allocator
1001 * This iterator iterates though all zones at or below a given zone index and
1002 * within a given nodemask
1004 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1005 for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \
1007 z = next_zones_zonelist(++z, highidx, nodemask), \
1008 zone = zonelist_zone(z))
1010 #define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1011 for (zone = z->zone; \
1013 z = next_zones_zonelist(++z, highidx, nodemask), \
1014 zone = zonelist_zone(z))
1018 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1019 * @zone - The current zone in the iterator
1020 * @z - The current pointer within zonelist->zones being iterated
1021 * @zlist - The zonelist being iterated
1022 * @highidx - The zone index of the highest zone to return
1024 * This iterator iterates though all zones at or below a given zone index.
1026 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
1027 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1029 #ifdef CONFIG_SPARSEMEM
1030 #include <asm/sparsemem.h>
1033 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
1034 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
1035 static inline unsigned long early_pfn_to_nid(unsigned long pfn
)
1041 #ifdef CONFIG_FLATMEM
1042 #define pfn_to_nid(pfn) (0)
1045 #ifdef CONFIG_SPARSEMEM
1048 * SECTION_SHIFT #bits space required to store a section #
1050 * PA_SECTION_SHIFT physical address to/from section number
1051 * PFN_SECTION_SHIFT pfn to/from section number
1053 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1054 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1056 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1058 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1059 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1061 #define SECTION_BLOCKFLAGS_BITS \
1062 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1064 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1065 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1068 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1069 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1071 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1072 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1076 struct mem_section
{
1078 * This is, logically, a pointer to an array of struct
1079 * pages. However, it is stored with some other magic.
1080 * (see sparse.c::sparse_init_one_section())
1082 * Additionally during early boot we encode node id of
1083 * the location of the section here to guide allocation.
1084 * (see sparse.c::memory_present())
1086 * Making it a UL at least makes someone do a cast
1087 * before using it wrong.
1089 unsigned long section_mem_map
;
1091 /* See declaration of similar field in struct zone */
1092 unsigned long *pageblock_flags
;
1093 #ifdef CONFIG_PAGE_EXTENSION
1095 * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1096 * section. (see page_ext.h about this.)
1098 struct page_ext
*page_ext
;
1102 * WARNING: mem_section must be a power-of-2 in size for the
1103 * calculation and use of SECTION_ROOT_MASK to make sense.
1107 #ifdef CONFIG_SPARSEMEM_EXTREME
1108 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1110 #define SECTIONS_PER_ROOT 1
1113 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1114 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1115 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1117 #ifdef CONFIG_SPARSEMEM_EXTREME
1118 extern struct mem_section
*mem_section
[NR_SECTION_ROOTS
];
1120 extern struct mem_section mem_section
[NR_SECTION_ROOTS
][SECTIONS_PER_ROOT
];
1123 static inline struct mem_section
*__nr_to_section(unsigned long nr
)
1125 if (!mem_section
[SECTION_NR_TO_ROOT(nr
)])
1127 return &mem_section
[SECTION_NR_TO_ROOT(nr
)][nr
& SECTION_ROOT_MASK
];
1129 extern int __section_nr(struct mem_section
* ms
);
1130 extern unsigned long usemap_size(void);
1133 * We use the lower bits of the mem_map pointer to store
1134 * a little bit of information. There should be at least
1135 * 3 bits here due to 32-bit alignment.
1137 #define SECTION_MARKED_PRESENT (1UL<<0)
1138 #define SECTION_HAS_MEM_MAP (1UL<<1)
1139 #define SECTION_MAP_LAST_BIT (1UL<<2)
1140 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1141 #define SECTION_NID_SHIFT 2
1143 static inline struct page
*__section_mem_map_addr(struct mem_section
*section
)
1145 unsigned long map
= section
->section_mem_map
;
1146 map
&= SECTION_MAP_MASK
;
1147 return (struct page
*)map
;
1150 static inline int present_section(struct mem_section
*section
)
1152 return (section
&& (section
->section_mem_map
& SECTION_MARKED_PRESENT
));
1155 static inline int present_section_nr(unsigned long nr
)
1157 return present_section(__nr_to_section(nr
));
1160 static inline int valid_section(struct mem_section
*section
)
1162 return (section
&& (section
->section_mem_map
& SECTION_HAS_MEM_MAP
));
1165 static inline int valid_section_nr(unsigned long nr
)
1167 return valid_section(__nr_to_section(nr
));
1170 static inline struct mem_section
*__pfn_to_section(unsigned long pfn
)
1172 return __nr_to_section(pfn_to_section_nr(pfn
));
1175 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1176 static inline int pfn_valid(unsigned long pfn
)
1178 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1180 return valid_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1184 static inline int pfn_present(unsigned long pfn
)
1186 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1188 return present_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1192 * These are _only_ used during initialisation, therefore they
1193 * can use __initdata ... They could have names to indicate
1197 #define pfn_to_nid(pfn) \
1199 unsigned long __pfn_to_nid_pfn = (pfn); \
1200 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1203 #define pfn_to_nid(pfn) (0)
1206 #define early_pfn_valid(pfn) pfn_valid(pfn)
1207 void sparse_init(void);
1209 #define sparse_init() do {} while (0)
1210 #define sparse_index_init(_sec, _nid) do {} while (0)
1211 #endif /* CONFIG_SPARSEMEM */
1214 * During memory init memblocks map pfns to nids. The search is expensive and
1215 * this caches recent lookups. The implementation of __early_pfn_to_nid
1216 * may treat start/end as pfns or sections.
1218 struct mminit_pfnnid_cache
{
1219 unsigned long last_start
;
1220 unsigned long last_end
;
1224 #ifndef early_pfn_valid
1225 #define early_pfn_valid(pfn) (1)
1228 void memory_present(int nid
, unsigned long start
, unsigned long end
);
1229 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
1232 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1233 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1234 * pfn_valid_within() should be used in this case; we optimise this away
1235 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1237 #ifdef CONFIG_HOLES_IN_ZONE
1238 #define pfn_valid_within(pfn) pfn_valid(pfn)
1240 #define pfn_valid_within(pfn) (1)
1243 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1245 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1246 * associated with it or not. In FLATMEM, it is expected that holes always
1247 * have valid memmap as long as there is valid PFNs either side of the hole.
1248 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1251 * However, an ARM, and maybe other embedded architectures in the future
1252 * free memmap backing holes to save memory on the assumption the memmap is
1253 * never used. The page_zone linkages are then broken even though pfn_valid()
1254 * returns true. A walker of the full memmap must then do this additional
1255 * check to ensure the memmap they are looking at is sane by making sure
1256 * the zone and PFN linkages are still valid. This is expensive, but walkers
1257 * of the full memmap are extremely rare.
1259 bool memmap_valid_within(unsigned long pfn
,
1260 struct page
*page
, struct zone
*zone
);
1262 static inline bool memmap_valid_within(unsigned long pfn
,
1263 struct page
*page
, struct zone
*zone
)
1267 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1269 #endif /* !__GENERATING_BOUNDS.H */
1270 #endif /* !__ASSEMBLY__ */
1271 #endif /* _LINUX_MMZONE_H */