1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
5 #ifndef __GENERATING_BOUNDS_H
7 #include <linux/spinlock.h>
8 #include <linux/list.h>
9 #include <linux/wait.h>
10 #include <linux/bitops.h>
11 #include <linux/cache.h>
12 #include <linux/threads.h>
13 #include <linux/numa.h>
14 #include <linux/init.h>
15 #include <linux/seqlock.h>
16 #include <linux/nodemask.h>
17 #include <linux/pageblock-flags.h>
18 #include <generated/bounds.h>
19 #include <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 * coelesce naturally under reasonable reclaim pressure and those which
36 #define PAGE_ALLOC_COSTLY_ORDER 3
38 #define MIGRATE_UNMOVABLE 0
39 #define MIGRATE_RECLAIMABLE 1
40 #define MIGRATE_MOVABLE 2
41 #define MIGRATE_PCPTYPES 3 /* the number of types on the pcp lists */
42 #define MIGRATE_RESERVE 3
43 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
44 #define MIGRATE_TYPES 5
46 #define for_each_migratetype_order(order, type) \
47 for (order = 0; order < MAX_ORDER; order++) \
48 for (type = 0; type < MIGRATE_TYPES; type++)
50 extern int page_group_by_mobility_disabled
;
52 static inline int get_pageblock_migratetype(struct page
*page
)
54 return get_pageblock_flags_group(page
, PB_migrate
, PB_migrate_end
);
58 struct list_head free_list
[MIGRATE_TYPES
];
59 unsigned long nr_free
;
65 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
66 * So add a wild amount of padding here to ensure that they fall into separate
67 * cachelines. There are very few zone structures in the machine, so space
68 * consumption is not a concern here.
70 #if defined(CONFIG_SMP)
73 } ____cacheline_internodealigned_in_smp
;
74 #define ZONE_PADDING(name) struct zone_padding name;
76 #define ZONE_PADDING(name)
80 /* First 128 byte cacheline (assuming 64 bit words) */
83 NR_INACTIVE_ANON
= NR_LRU_BASE
, /* must match order of LRU_[IN]ACTIVE */
84 NR_ACTIVE_ANON
, /* " " " " " */
85 NR_INACTIVE_FILE
, /* " " " " " */
86 NR_ACTIVE_FILE
, /* " " " " " */
87 NR_UNEVICTABLE
, /* " " " " " */
88 NR_MLOCK
, /* mlock()ed pages found and moved off LRU */
89 NR_ANON_PAGES
, /* Mapped anonymous pages */
90 NR_FILE_MAPPED
, /* pagecache pages mapped into pagetables.
91 only modified from process context */
96 NR_SLAB_UNRECLAIMABLE
,
97 NR_PAGETABLE
, /* used for pagetables */
99 /* Second 128 byte cacheline */
100 NR_UNSTABLE_NFS
, /* NFS unstable pages */
103 NR_VMSCAN_IMMEDIATE
, /* Prioritise for reclaim when writeback ends */
104 NR_WRITEBACK_TEMP
, /* Writeback using temporary buffers */
105 NR_ISOLATED_ANON
, /* Temporary isolated pages from anon lru */
106 NR_ISOLATED_FILE
, /* Temporary isolated pages from file lru */
107 NR_SHMEM
, /* shmem pages (included tmpfs/GEM pages) */
108 NR_DIRTIED
, /* page dirtyings since bootup */
109 NR_WRITTEN
, /* page writings since bootup */
111 NUMA_HIT
, /* allocated in intended node */
112 NUMA_MISS
, /* allocated in non intended node */
113 NUMA_FOREIGN
, /* was intended here, hit elsewhere */
114 NUMA_INTERLEAVE_HIT
, /* interleaver preferred this zone */
115 NUMA_LOCAL
, /* allocation from local node */
116 NUMA_OTHER
, /* allocation from other node */
118 NR_ANON_TRANSPARENT_HUGEPAGES
,
119 NR_VM_ZONE_STAT_ITEMS
};
122 * We do arithmetic on the LRU lists in various places in the code,
123 * so it is important to keep the active lists LRU_ACTIVE higher in
124 * the array than the corresponding inactive lists, and to keep
125 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
127 * This has to be kept in sync with the statistics in zone_stat_item
128 * above and the descriptions in vmstat_text in mm/vmstat.c
135 LRU_INACTIVE_ANON
= LRU_BASE
,
136 LRU_ACTIVE_ANON
= LRU_BASE
+ LRU_ACTIVE
,
137 LRU_INACTIVE_FILE
= LRU_BASE
+ LRU_FILE
,
138 LRU_ACTIVE_FILE
= LRU_BASE
+ LRU_FILE
+ LRU_ACTIVE
,
143 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
145 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
147 static inline int is_file_lru(enum lru_list l
)
149 return (l
== LRU_INACTIVE_FILE
|| l
== LRU_ACTIVE_FILE
);
152 static inline int is_active_lru(enum lru_list l
)
154 return (l
== LRU_ACTIVE_ANON
|| l
== LRU_ACTIVE_FILE
);
157 static inline int is_unevictable_lru(enum lru_list l
)
159 return (l
== LRU_UNEVICTABLE
);
162 /* Mask used at gathering information at once (see memcontrol.c) */
163 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
164 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
165 #define LRU_ALL_EVICTABLE (LRU_ALL_FILE | LRU_ALL_ANON)
166 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
168 /* Isolate inactive pages */
169 #define ISOLATE_INACTIVE ((__force fmode_t)0x1)
170 /* Isolate active pages */
171 #define ISOLATE_ACTIVE ((__force fmode_t)0x2)
172 /* Isolate clean file */
173 #define ISOLATE_CLEAN ((__force fmode_t)0x4)
174 /* Isolate unmapped file */
175 #define ISOLATE_UNMAPPED ((__force fmode_t)0x8)
177 /* LRU Isolation modes. */
178 typedef unsigned __bitwise__ isolate_mode_t
;
180 enum zone_watermarks
{
187 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
188 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
189 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
191 struct per_cpu_pages
{
192 int count
; /* number of pages in the list */
193 int high
; /* high watermark, emptying needed */
194 int batch
; /* chunk size for buddy add/remove */
196 /* Lists of pages, one per migrate type stored on the pcp-lists */
197 struct list_head lists
[MIGRATE_PCPTYPES
];
200 struct per_cpu_pageset
{
201 struct per_cpu_pages pcp
;
207 s8 vm_stat_diff
[NR_VM_ZONE_STAT_ITEMS
];
211 #endif /* !__GENERATING_BOUNDS.H */
214 #ifdef CONFIG_ZONE_DMA
216 * ZONE_DMA is used when there are devices that are not able
217 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
218 * carve out the portion of memory that is needed for these devices.
219 * The range is arch specific.
224 * ---------------------------
225 * parisc, ia64, sparc <4G
228 * alpha Unlimited or 0-16MB.
230 * i386, x86_64 and multiple other arches
235 #ifdef CONFIG_ZONE_DMA32
237 * x86_64 needs two ZONE_DMAs because it supports devices that are
238 * only able to do DMA to the lower 16M but also 32 bit devices that
239 * can only do DMA areas below 4G.
244 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
245 * performed on pages in ZONE_NORMAL if the DMA devices support
246 * transfers to all addressable memory.
249 #ifdef CONFIG_HIGHMEM
251 * A memory area that is only addressable by the kernel through
252 * mapping portions into its own address space. This is for example
253 * used by i386 to allow the kernel to address the memory beyond
254 * 900MB. The kernel will set up special mappings (page
255 * table entries on i386) for each page that the kernel needs to
264 #ifndef __GENERATING_BOUNDS_H
267 * When a memory allocation must conform to specific limitations (such
268 * as being suitable for DMA) the caller will pass in hints to the
269 * allocator in the gfp_mask, in the zone modifier bits. These bits
270 * are used to select a priority ordered list of memory zones which
271 * match the requested limits. See gfp_zone() in include/linux/gfp.h
275 #define ZONES_SHIFT 0
276 #elif MAX_NR_ZONES <= 2
277 #define ZONES_SHIFT 1
278 #elif MAX_NR_ZONES <= 4
279 #define ZONES_SHIFT 2
281 #error ZONES_SHIFT -- too many zones configured adjust calculation
284 struct zone_reclaim_stat
{
286 * The pageout code in vmscan.c keeps track of how many of the
287 * mem/swap backed and file backed pages are refeferenced.
288 * The higher the rotated/scanned ratio, the more valuable
291 * The anon LRU stats live in [0], file LRU stats in [1]
293 unsigned long recent_rotated
[2];
294 unsigned long recent_scanned
[2];
298 /* Fields commonly accessed by the page allocator */
300 /* zone watermarks, access with *_wmark_pages(zone) macros */
301 unsigned long watermark
[NR_WMARK
];
304 * When free pages are below this point, additional steps are taken
305 * when reading the number of free pages to avoid per-cpu counter
306 * drift allowing watermarks to be breached
308 unsigned long percpu_drift_mark
;
311 * We don't know if the memory that we're going to allocate will be freeable
312 * or/and it will be released eventually, so to avoid totally wasting several
313 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
314 * to run OOM on the lower zones despite there's tons of freeable ram
315 * on the higher zones). This array is recalculated at runtime if the
316 * sysctl_lowmem_reserve_ratio sysctl changes.
318 unsigned long lowmem_reserve
[MAX_NR_ZONES
];
323 * zone reclaim becomes active if more unmapped pages exist.
325 unsigned long min_unmapped_pages
;
326 unsigned long min_slab_pages
;
328 struct per_cpu_pageset __percpu
*pageset
;
330 * free areas of different sizes
333 int all_unreclaimable
; /* All pages pinned */
334 #ifdef CONFIG_MEMORY_HOTPLUG
335 /* see spanned/present_pages for more description */
336 seqlock_t span_seqlock
;
338 struct free_area free_area
[MAX_ORDER
];
340 #ifndef CONFIG_SPARSEMEM
342 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
343 * In SPARSEMEM, this map is stored in struct mem_section
345 unsigned long *pageblock_flags
;
346 #endif /* CONFIG_SPARSEMEM */
348 #ifdef CONFIG_COMPACTION
350 * On compaction failure, 1<<compact_defer_shift compactions
351 * are skipped before trying again. The number attempted since
352 * last failure is tracked with compact_considered.
354 unsigned int compact_considered
;
355 unsigned int compact_defer_shift
;
360 /* Fields commonly accessed by the page reclaim scanner */
363 struct list_head list
;
366 struct zone_reclaim_stat reclaim_stat
;
368 unsigned long pages_scanned
; /* since last reclaim */
369 unsigned long flags
; /* zone flags, see below */
371 /* Zone statistics */
372 atomic_long_t vm_stat
[NR_VM_ZONE_STAT_ITEMS
];
375 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
376 * this zone's LRU. Maintained by the pageout code.
378 unsigned int inactive_ratio
;
382 /* Rarely used or read-mostly fields */
385 * wait_table -- the array holding the hash table
386 * wait_table_hash_nr_entries -- the size of the hash table array
387 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
389 * The purpose of all these is to keep track of the people
390 * waiting for a page to become available and make them
391 * runnable again when possible. The trouble is that this
392 * consumes a lot of space, especially when so few things
393 * wait on pages at a given time. So instead of using
394 * per-page waitqueues, we use a waitqueue hash table.
396 * The bucket discipline is to sleep on the same queue when
397 * colliding and wake all in that wait queue when removing.
398 * When something wakes, it must check to be sure its page is
399 * truly available, a la thundering herd. The cost of a
400 * collision is great, but given the expected load of the
401 * table, they should be so rare as to be outweighed by the
402 * benefits from the saved space.
404 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
405 * primary users of these fields, and in mm/page_alloc.c
406 * free_area_init_core() performs the initialization of them.
408 wait_queue_head_t
* wait_table
;
409 unsigned long wait_table_hash_nr_entries
;
410 unsigned long wait_table_bits
;
413 * Discontig memory support fields.
415 struct pglist_data
*zone_pgdat
;
416 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
417 unsigned long zone_start_pfn
;
420 * zone_start_pfn, spanned_pages and present_pages are all
421 * protected by span_seqlock. It is a seqlock because it has
422 * to be read outside of zone->lock, and it is done in the main
423 * allocator path. But, it is written quite infrequently.
425 * The lock is declared along with zone->lock because it is
426 * frequently read in proximity to zone->lock. It's good to
427 * give them a chance of being in the same cacheline.
429 unsigned long spanned_pages
; /* total size, including holes */
430 unsigned long present_pages
; /* amount of memory (excluding holes) */
433 * rarely used fields:
436 } ____cacheline_internodealigned_in_smp
;
439 ZONE_RECLAIM_LOCKED
, /* prevents concurrent reclaim */
440 ZONE_OOM_LOCKED
, /* zone is in OOM killer zonelist */
441 ZONE_CONGESTED
, /* zone has many dirty pages backed by
446 static inline void zone_set_flag(struct zone
*zone
, zone_flags_t flag
)
448 set_bit(flag
, &zone
->flags
);
451 static inline int zone_test_and_set_flag(struct zone
*zone
, zone_flags_t flag
)
453 return test_and_set_bit(flag
, &zone
->flags
);
456 static inline void zone_clear_flag(struct zone
*zone
, zone_flags_t flag
)
458 clear_bit(flag
, &zone
->flags
);
461 static inline int zone_is_reclaim_congested(const struct zone
*zone
)
463 return test_bit(ZONE_CONGESTED
, &zone
->flags
);
466 static inline int zone_is_reclaim_locked(const struct zone
*zone
)
468 return test_bit(ZONE_RECLAIM_LOCKED
, &zone
->flags
);
471 static inline int zone_is_oom_locked(const struct zone
*zone
)
473 return test_bit(ZONE_OOM_LOCKED
, &zone
->flags
);
477 * The "priority" of VM scanning is how much of the queues we will scan in one
478 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
479 * queues ("queue_length >> 12") during an aging round.
481 #define DEF_PRIORITY 12
483 /* Maximum number of zones on a zonelist */
484 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
489 * The NUMA zonelists are doubled because we need zonelists that restrict the
490 * allocations to a single node for GFP_THISNODE.
492 * [0] : Zonelist with fallback
493 * [1] : No fallback (GFP_THISNODE)
495 #define MAX_ZONELISTS 2
499 * We cache key information from each zonelist for smaller cache
500 * footprint when scanning for free pages in get_page_from_freelist().
502 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
503 * up short of free memory since the last time (last_fullzone_zap)
504 * we zero'd fullzones.
505 * 2) The array z_to_n[] maps each zone in the zonelist to its node
506 * id, so that we can efficiently evaluate whether that node is
507 * set in the current tasks mems_allowed.
509 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
510 * indexed by a zones offset in the zonelist zones[] array.
512 * The get_page_from_freelist() routine does two scans. During the
513 * first scan, we skip zones whose corresponding bit in 'fullzones'
514 * is set or whose corresponding node in current->mems_allowed (which
515 * comes from cpusets) is not set. During the second scan, we bypass
516 * this zonelist_cache, to ensure we look methodically at each zone.
518 * Once per second, we zero out (zap) fullzones, forcing us to
519 * reconsider nodes that might have regained more free memory.
520 * The field last_full_zap is the time we last zapped fullzones.
522 * This mechanism reduces the amount of time we waste repeatedly
523 * reexaming zones for free memory when they just came up low on
524 * memory momentarilly ago.
526 * The zonelist_cache struct members logically belong in struct
527 * zonelist. However, the mempolicy zonelists constructed for
528 * MPOL_BIND are intentionally variable length (and usually much
529 * shorter). A general purpose mechanism for handling structs with
530 * multiple variable length members is more mechanism than we want
531 * here. We resort to some special case hackery instead.
533 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
534 * part because they are shorter), so we put the fixed length stuff
535 * at the front of the zonelist struct, ending in a variable length
536 * zones[], as is needed by MPOL_BIND.
538 * Then we put the optional zonelist cache on the end of the zonelist
539 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
540 * the fixed length portion at the front of the struct. This pointer
541 * both enables us to find the zonelist cache, and in the case of
542 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
543 * to know that the zonelist cache is not there.
545 * The end result is that struct zonelists come in two flavors:
546 * 1) The full, fixed length version, shown below, and
547 * 2) The custom zonelists for MPOL_BIND.
548 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
550 * Even though there may be multiple CPU cores on a node modifying
551 * fullzones or last_full_zap in the same zonelist_cache at the same
552 * time, we don't lock it. This is just hint data - if it is wrong now
553 * and then, the allocator will still function, perhaps a bit slower.
557 struct zonelist_cache
{
558 unsigned short z_to_n
[MAX_ZONES_PER_ZONELIST
]; /* zone->nid */
559 DECLARE_BITMAP(fullzones
, MAX_ZONES_PER_ZONELIST
); /* zone full? */
560 unsigned long last_full_zap
; /* when last zap'd (jiffies) */
563 #define MAX_ZONELISTS 1
564 struct zonelist_cache
;
568 * This struct contains information about a zone in a zonelist. It is stored
569 * here to avoid dereferences into large structures and lookups of tables
572 struct zone
*zone
; /* Pointer to actual zone */
573 int zone_idx
; /* zone_idx(zoneref->zone) */
577 * One allocation request operates on a zonelist. A zonelist
578 * is a list of zones, the first one is the 'goal' of the
579 * allocation, the other zones are fallback zones, in decreasing
582 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
583 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
585 * To speed the reading of the zonelist, the zonerefs contain the zone index
586 * of the entry being read. Helper functions to access information given
587 * a struct zoneref are
589 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
590 * zonelist_zone_idx() - Return the index of the zone for an entry
591 * zonelist_node_idx() - Return the index of the node for an entry
594 struct zonelist_cache
*zlcache_ptr
; // NULL or &zlcache
595 struct zoneref _zonerefs
[MAX_ZONES_PER_ZONELIST
+ 1];
597 struct zonelist_cache zlcache
; // optional ...
601 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
602 struct node_active_region
{
603 unsigned long start_pfn
;
604 unsigned long end_pfn
;
607 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
609 #ifndef CONFIG_DISCONTIGMEM
610 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
611 extern struct page
*mem_map
;
615 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
616 * (mostly NUMA machines?) to denote a higher-level memory zone than the
619 * On NUMA machines, each NUMA node would have a pg_data_t to describe
620 * it's memory layout.
622 * Memory statistics and page replacement data structures are maintained on a
626 typedef struct pglist_data
{
627 struct zone node_zones
[MAX_NR_ZONES
];
628 struct zonelist node_zonelists
[MAX_ZONELISTS
];
630 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
631 struct page
*node_mem_map
;
632 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
633 struct page_cgroup
*node_page_cgroup
;
636 #ifndef CONFIG_NO_BOOTMEM
637 struct bootmem_data
*bdata
;
639 #ifdef CONFIG_MEMORY_HOTPLUG
641 * Must be held any time you expect node_start_pfn, node_present_pages
642 * or node_spanned_pages stay constant. Holding this will also
643 * guarantee that any pfn_valid() stays that way.
645 * Nests above zone->lock and zone->size_seqlock.
647 spinlock_t node_size_lock
;
649 unsigned long node_start_pfn
;
650 unsigned long node_present_pages
; /* total number of physical pages */
651 unsigned long node_spanned_pages
; /* total size of physical page
652 range, including holes */
654 wait_queue_head_t kswapd_wait
;
655 struct task_struct
*kswapd
;
656 int kswapd_max_order
;
657 enum zone_type classzone_idx
;
660 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
661 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
662 #ifdef CONFIG_FLAT_NODE_MEM_MAP
663 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
665 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
667 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
669 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
671 #define node_end_pfn(nid) ({\
672 pg_data_t *__pgdat = NODE_DATA(nid);\
673 __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
676 #include <linux/memory_hotplug.h>
678 extern struct mutex zonelists_mutex
;
679 void build_all_zonelists(void *data
);
680 void wakeup_kswapd(struct zone
*zone
, int order
, enum zone_type classzone_idx
);
681 bool zone_watermark_ok(struct zone
*z
, int order
, unsigned long mark
,
682 int classzone_idx
, int alloc_flags
);
683 bool zone_watermark_ok_safe(struct zone
*z
, int order
, unsigned long mark
,
684 int classzone_idx
, int alloc_flags
);
685 enum memmap_context
{
689 extern int init_currently_empty_zone(struct zone
*zone
, unsigned long start_pfn
,
691 enum memmap_context context
);
693 #ifdef CONFIG_HAVE_MEMORY_PRESENT
694 void memory_present(int nid
, unsigned long start
, unsigned long end
);
696 static inline void memory_present(int nid
, unsigned long start
, unsigned long end
) {}
699 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
700 int local_memory_node(int node_id
);
702 static inline int local_memory_node(int node_id
) { return node_id
; };
705 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
706 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
710 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
712 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
714 static inline int populated_zone(struct zone
*zone
)
716 return (!!zone
->present_pages
);
719 extern int movable_zone
;
721 static inline int zone_movable_is_highmem(void)
723 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
724 return movable_zone
== ZONE_HIGHMEM
;
730 static inline int is_highmem_idx(enum zone_type idx
)
732 #ifdef CONFIG_HIGHMEM
733 return (idx
== ZONE_HIGHMEM
||
734 (idx
== ZONE_MOVABLE
&& zone_movable_is_highmem()));
740 static inline int is_normal_idx(enum zone_type idx
)
742 return (idx
== ZONE_NORMAL
);
746 * is_highmem - helper function to quickly check if a struct zone is a
747 * highmem zone or not. This is an attempt to keep references
748 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
749 * @zone - pointer to struct zone variable
751 static inline int is_highmem(struct zone
*zone
)
753 #ifdef CONFIG_HIGHMEM
754 int zone_off
= (char *)zone
- (char *)zone
->zone_pgdat
->node_zones
;
755 return zone_off
== ZONE_HIGHMEM
* sizeof(*zone
) ||
756 (zone_off
== ZONE_MOVABLE
* sizeof(*zone
) &&
757 zone_movable_is_highmem());
763 static inline int is_normal(struct zone
*zone
)
765 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_NORMAL
;
768 static inline int is_dma32(struct zone
*zone
)
770 #ifdef CONFIG_ZONE_DMA32
771 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_DMA32
;
777 static inline int is_dma(struct zone
*zone
)
779 #ifdef CONFIG_ZONE_DMA
780 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_DMA
;
786 /* These two functions are used to setup the per zone pages min values */
788 int min_free_kbytes_sysctl_handler(struct ctl_table
*, int,
789 void __user
*, size_t *, loff_t
*);
790 extern int sysctl_lowmem_reserve_ratio
[MAX_NR_ZONES
-1];
791 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table
*, int,
792 void __user
*, size_t *, loff_t
*);
793 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table
*, int,
794 void __user
*, size_t *, loff_t
*);
795 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table
*, int,
796 void __user
*, size_t *, loff_t
*);
797 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table
*, int,
798 void __user
*, size_t *, loff_t
*);
800 extern int numa_zonelist_order_handler(struct ctl_table
*, int,
801 void __user
*, size_t *, loff_t
*);
802 extern char numa_zonelist_order
[];
803 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
805 #ifndef CONFIG_NEED_MULTIPLE_NODES
807 extern struct pglist_data contig_page_data
;
808 #define NODE_DATA(nid) (&contig_page_data)
809 #define NODE_MEM_MAP(nid) mem_map
811 #else /* CONFIG_NEED_MULTIPLE_NODES */
813 #include <asm/mmzone.h>
815 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
817 extern struct pglist_data
*first_online_pgdat(void);
818 extern struct pglist_data
*next_online_pgdat(struct pglist_data
*pgdat
);
819 extern struct zone
*next_zone(struct zone
*zone
);
822 * for_each_online_pgdat - helper macro to iterate over all online nodes
823 * @pgdat - pointer to a pg_data_t variable
825 #define for_each_online_pgdat(pgdat) \
826 for (pgdat = first_online_pgdat(); \
828 pgdat = next_online_pgdat(pgdat))
830 * for_each_zone - helper macro to iterate over all memory zones
831 * @zone - pointer to struct zone variable
833 * The user only needs to declare the zone variable, for_each_zone
836 #define for_each_zone(zone) \
837 for (zone = (first_online_pgdat())->node_zones; \
839 zone = next_zone(zone))
841 #define for_each_populated_zone(zone) \
842 for (zone = (first_online_pgdat())->node_zones; \
844 zone = next_zone(zone)) \
845 if (!populated_zone(zone)) \
849 static inline struct zone
*zonelist_zone(struct zoneref
*zoneref
)
851 return zoneref
->zone
;
854 static inline int zonelist_zone_idx(struct zoneref
*zoneref
)
856 return zoneref
->zone_idx
;
859 static inline int zonelist_node_idx(struct zoneref
*zoneref
)
862 /* zone_to_nid not available in this context */
863 return zoneref
->zone
->node
;
866 #endif /* CONFIG_NUMA */
870 * 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
871 * @z - The cursor used as a starting point for the search
872 * @highest_zoneidx - The zone index of the highest zone to return
873 * @nodes - An optional nodemask to filter the zonelist with
874 * @zone - The first suitable zone found is returned via this parameter
876 * This function returns the next zone at or below a given zone index that is
877 * within the allowed nodemask using a cursor as the starting point for the
878 * search. The zoneref returned is a cursor that represents the current zone
879 * being examined. It should be advanced by one before calling
880 * next_zones_zonelist again.
882 struct zoneref
*next_zones_zonelist(struct zoneref
*z
,
883 enum zone_type highest_zoneidx
,
888 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
889 * @zonelist - The zonelist to search for a suitable zone
890 * @highest_zoneidx - The zone index of the highest zone to return
891 * @nodes - An optional nodemask to filter the zonelist with
892 * @zone - The first suitable zone found is returned via this parameter
894 * This function returns the first zone at or below a given zone index that is
895 * within the allowed nodemask. The zoneref returned is a cursor that can be
896 * used to iterate the zonelist with next_zones_zonelist by advancing it by
897 * one before calling.
899 static inline struct zoneref
*first_zones_zonelist(struct zonelist
*zonelist
,
900 enum zone_type highest_zoneidx
,
904 return next_zones_zonelist(zonelist
->_zonerefs
, highest_zoneidx
, nodes
,
909 * 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
910 * @zone - The current zone in the iterator
911 * @z - The current pointer within zonelist->zones being iterated
912 * @zlist - The zonelist being iterated
913 * @highidx - The zone index of the highest zone to return
914 * @nodemask - Nodemask allowed by the allocator
916 * This iterator iterates though all zones at or below a given zone index and
917 * within a given nodemask
919 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
920 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
922 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
925 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
926 * @zone - The current zone in the iterator
927 * @z - The current pointer within zonelist->zones being iterated
928 * @zlist - The zonelist being iterated
929 * @highidx - The zone index of the highest zone to return
931 * This iterator iterates though all zones at or below a given zone index.
933 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
934 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
936 #ifdef CONFIG_SPARSEMEM
937 #include <asm/sparsemem.h>
940 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
941 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
942 static inline unsigned long early_pfn_to_nid(unsigned long pfn
)
948 #ifdef CONFIG_FLATMEM
949 #define pfn_to_nid(pfn) (0)
952 #ifdef CONFIG_SPARSEMEM
955 * SECTION_SHIFT #bits space required to store a section #
957 * PA_SECTION_SHIFT physical address to/from section number
958 * PFN_SECTION_SHIFT pfn to/from section number
960 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
962 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
963 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
965 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
967 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
968 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
970 #define SECTION_BLOCKFLAGS_BITS \
971 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
973 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
974 #error Allocator MAX_ORDER exceeds SECTION_SIZE
977 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
978 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
980 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
981 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
987 * This is, logically, a pointer to an array of struct
988 * pages. However, it is stored with some other magic.
989 * (see sparse.c::sparse_init_one_section())
991 * Additionally during early boot we encode node id of
992 * the location of the section here to guide allocation.
993 * (see sparse.c::memory_present())
995 * Making it a UL at least makes someone do a cast
996 * before using it wrong.
998 unsigned long section_mem_map
;
1000 /* See declaration of similar field in struct zone */
1001 unsigned long *pageblock_flags
;
1002 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1004 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1005 * section. (see memcontrol.h/page_cgroup.h about this.)
1007 struct page_cgroup
*page_cgroup
;
1012 #ifdef CONFIG_SPARSEMEM_EXTREME
1013 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1015 #define SECTIONS_PER_ROOT 1
1018 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1019 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1020 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1022 #ifdef CONFIG_SPARSEMEM_EXTREME
1023 extern struct mem_section
*mem_section
[NR_SECTION_ROOTS
];
1025 extern struct mem_section mem_section
[NR_SECTION_ROOTS
][SECTIONS_PER_ROOT
];
1028 static inline struct mem_section
*__nr_to_section(unsigned long nr
)
1030 if (!mem_section
[SECTION_NR_TO_ROOT(nr
)])
1032 return &mem_section
[SECTION_NR_TO_ROOT(nr
)][nr
& SECTION_ROOT_MASK
];
1034 extern int __section_nr(struct mem_section
* ms
);
1035 extern unsigned long usemap_size(void);
1038 * We use the lower bits of the mem_map pointer to store
1039 * a little bit of information. There should be at least
1040 * 3 bits here due to 32-bit alignment.
1042 #define SECTION_MARKED_PRESENT (1UL<<0)
1043 #define SECTION_HAS_MEM_MAP (1UL<<1)
1044 #define SECTION_MAP_LAST_BIT (1UL<<2)
1045 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1046 #define SECTION_NID_SHIFT 2
1048 static inline struct page
*__section_mem_map_addr(struct mem_section
*section
)
1050 unsigned long map
= section
->section_mem_map
;
1051 map
&= SECTION_MAP_MASK
;
1052 return (struct page
*)map
;
1055 static inline int present_section(struct mem_section
*section
)
1057 return (section
&& (section
->section_mem_map
& SECTION_MARKED_PRESENT
));
1060 static inline int present_section_nr(unsigned long nr
)
1062 return present_section(__nr_to_section(nr
));
1065 static inline int valid_section(struct mem_section
*section
)
1067 return (section
&& (section
->section_mem_map
& SECTION_HAS_MEM_MAP
));
1070 static inline int valid_section_nr(unsigned long nr
)
1072 return valid_section(__nr_to_section(nr
));
1075 static inline struct mem_section
*__pfn_to_section(unsigned long pfn
)
1077 return __nr_to_section(pfn_to_section_nr(pfn
));
1080 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1081 static inline int pfn_valid(unsigned long pfn
)
1083 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1085 return valid_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1089 static inline int pfn_present(unsigned long pfn
)
1091 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1093 return present_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1097 * These are _only_ used during initialisation, therefore they
1098 * can use __initdata ... They could have names to indicate
1102 #define pfn_to_nid(pfn) \
1104 unsigned long __pfn_to_nid_pfn = (pfn); \
1105 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1108 #define pfn_to_nid(pfn) (0)
1111 #define early_pfn_valid(pfn) pfn_valid(pfn)
1112 void sparse_init(void);
1114 #define sparse_init() do {} while (0)
1115 #define sparse_index_init(_sec, _nid) do {} while (0)
1116 #endif /* CONFIG_SPARSEMEM */
1118 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1119 bool early_pfn_in_nid(unsigned long pfn
, int nid
);
1121 #define early_pfn_in_nid(pfn, nid) (1)
1124 #ifndef early_pfn_valid
1125 #define early_pfn_valid(pfn) (1)
1128 void memory_present(int nid
, unsigned long start
, unsigned long end
);
1129 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
1132 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1133 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1134 * pfn_valid_within() should be used in this case; we optimise this away
1135 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1137 #ifdef CONFIG_HOLES_IN_ZONE
1138 #define pfn_valid_within(pfn) pfn_valid(pfn)
1140 #define pfn_valid_within(pfn) (1)
1143 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1145 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1146 * associated with it or not. In FLATMEM, it is expected that holes always
1147 * have valid memmap as long as there is valid PFNs either side of the hole.
1148 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1151 * However, an ARM, and maybe other embedded architectures in the future
1152 * free memmap backing holes to save memory on the assumption the memmap is
1153 * never used. The page_zone linkages are then broken even though pfn_valid()
1154 * returns true. A walker of the full memmap must then do this additional
1155 * check to ensure the memmap they are looking at is sane by making sure
1156 * the zone and PFN linkages are still valid. This is expensive, but walkers
1157 * of the full memmap are extremely rare.
1159 int memmap_valid_within(unsigned long pfn
,
1160 struct page
*page
, struct zone
*zone
);
1162 static inline int memmap_valid_within(unsigned long pfn
,
1163 struct page
*page
, struct zone
*zone
)
1167 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1169 #endif /* !__GENERATING_BOUNDS.H */
1170 #endif /* !__ASSEMBLY__ */
1171 #endif /* _LINUX_MMZONE_H */