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/bounds.h>
19 #include <asm/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_RESERVE 3
42 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
43 #define MIGRATE_TYPES 5
45 #define for_each_migratetype_order(order, type) \
46 for (order = 0; order < MAX_ORDER; order++) \
47 for (type = 0; type < MIGRATE_TYPES; type++)
49 extern int page_group_by_mobility_disabled
;
51 static inline int get_pageblock_migratetype(struct page
*page
)
53 return get_pageblock_flags_group(page
, PB_migrate
, PB_migrate_end
);
57 struct list_head free_list
[MIGRATE_TYPES
];
58 unsigned long nr_free
;
64 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
65 * So add a wild amount of padding here to ensure that they fall into separate
66 * cachelines. There are very few zone structures in the machine, so space
67 * consumption is not a concern here.
69 #if defined(CONFIG_SMP)
72 } ____cacheline_internodealigned_in_smp
;
73 #define ZONE_PADDING(name) struct zone_padding name;
75 #define ZONE_PADDING(name)
79 /* First 128 byte cacheline (assuming 64 bit words) */
82 NR_INACTIVE_ANON
= NR_LRU_BASE
, /* must match order of LRU_[IN]ACTIVE */
83 NR_ACTIVE_ANON
, /* " " " " " */
84 NR_INACTIVE_FILE
, /* " " " " " */
85 NR_ACTIVE_FILE
, /* " " " " " */
86 NR_UNEVICTABLE
, /* " " " " " */
87 NR_MLOCK
, /* mlock()ed pages found and moved off LRU */
88 NR_ANON_PAGES
, /* Mapped anonymous pages */
89 NR_FILE_MAPPED
, /* pagecache pages mapped into pagetables.
90 only modified from process context */
95 NR_SLAB_UNRECLAIMABLE
,
96 NR_PAGETABLE
, /* used for pagetables */
98 /* Second 128 byte cacheline */
99 NR_UNSTABLE_NFS
, /* NFS unstable pages */
102 NR_WRITEBACK_TEMP
, /* Writeback using temporary buffers */
104 NUMA_HIT
, /* allocated in intended node */
105 NUMA_MISS
, /* allocated in non intended node */
106 NUMA_FOREIGN
, /* was intended here, hit elsewhere */
107 NUMA_INTERLEAVE_HIT
, /* interleaver preferred this zone */
108 NUMA_LOCAL
, /* allocation from local node */
109 NUMA_OTHER
, /* allocation from other node */
111 NR_VM_ZONE_STAT_ITEMS
};
114 * We do arithmetic on the LRU lists in various places in the code,
115 * so it is important to keep the active lists LRU_ACTIVE higher in
116 * the array than the corresponding inactive lists, and to keep
117 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
119 * This has to be kept in sync with the statistics in zone_stat_item
120 * above and the descriptions in vmstat_text in mm/vmstat.c
127 LRU_INACTIVE_ANON
= LRU_BASE
,
128 LRU_ACTIVE_ANON
= LRU_BASE
+ LRU_ACTIVE
,
129 LRU_INACTIVE_FILE
= LRU_BASE
+ LRU_FILE
,
130 LRU_ACTIVE_FILE
= LRU_BASE
+ LRU_FILE
+ LRU_ACTIVE
,
135 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
137 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
139 static inline int is_file_lru(enum lru_list l
)
141 return (l
== LRU_INACTIVE_FILE
|| l
== LRU_ACTIVE_FILE
);
144 static inline int is_active_lru(enum lru_list l
)
146 return (l
== LRU_ACTIVE_ANON
|| l
== LRU_ACTIVE_FILE
);
149 static inline int is_unevictable_lru(enum lru_list l
)
151 return (l
== LRU_UNEVICTABLE
);
154 enum zone_watermarks
{
161 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
162 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
163 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
165 struct per_cpu_pages
{
166 int count
; /* number of pages in the list */
167 int high
; /* high watermark, emptying needed */
168 int batch
; /* chunk size for buddy add/remove */
169 struct list_head list
; /* the list of pages */
172 struct per_cpu_pageset
{
173 struct per_cpu_pages pcp
;
179 s8 vm_stat_diff
[NR_VM_ZONE_STAT_ITEMS
];
181 } ____cacheline_aligned_in_smp
;
184 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
186 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
189 #endif /* !__GENERATING_BOUNDS.H */
192 #ifdef CONFIG_ZONE_DMA
194 * ZONE_DMA is used when there are devices that are not able
195 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
196 * carve out the portion of memory that is needed for these devices.
197 * The range is arch specific.
202 * ---------------------------
203 * parisc, ia64, sparc <4G
206 * alpha Unlimited or 0-16MB.
208 * i386, x86_64 and multiple other arches
213 #ifdef CONFIG_ZONE_DMA32
215 * x86_64 needs two ZONE_DMAs because it supports devices that are
216 * only able to do DMA to the lower 16M but also 32 bit devices that
217 * can only do DMA areas below 4G.
222 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
223 * performed on pages in ZONE_NORMAL if the DMA devices support
224 * transfers to all addressable memory.
227 #ifdef CONFIG_HIGHMEM
229 * A memory area that is only addressable by the kernel through
230 * mapping portions into its own address space. This is for example
231 * used by i386 to allow the kernel to address the memory beyond
232 * 900MB. The kernel will set up special mappings (page
233 * table entries on i386) for each page that the kernel needs to
242 #ifndef __GENERATING_BOUNDS_H
245 * When a memory allocation must conform to specific limitations (such
246 * as being suitable for DMA) the caller will pass in hints to the
247 * allocator in the gfp_mask, in the zone modifier bits. These bits
248 * are used to select a priority ordered list of memory zones which
249 * match the requested limits. See gfp_zone() in include/linux/gfp.h
253 #define ZONES_SHIFT 0
254 #elif MAX_NR_ZONES <= 2
255 #define ZONES_SHIFT 1
256 #elif MAX_NR_ZONES <= 4
257 #define ZONES_SHIFT 2
259 #error ZONES_SHIFT -- too many zones configured adjust calculation
262 struct zone_reclaim_stat
{
264 * The pageout code in vmscan.c keeps track of how many of the
265 * mem/swap backed and file backed pages are refeferenced.
266 * The higher the rotated/scanned ratio, the more valuable
269 * The anon LRU stats live in [0], file LRU stats in [1]
271 unsigned long recent_rotated
[2];
272 unsigned long recent_scanned
[2];
276 /* Fields commonly accessed by the page allocator */
278 /* zone watermarks, access with *_wmark_pages(zone) macros */
279 unsigned long watermark
[NR_WMARK
];
282 * We don't know if the memory that we're going to allocate will be freeable
283 * or/and it will be released eventually, so to avoid totally wasting several
284 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
285 * to run OOM on the lower zones despite there's tons of freeable ram
286 * on the higher zones). This array is recalculated at runtime if the
287 * sysctl_lowmem_reserve_ratio sysctl changes.
289 unsigned long lowmem_reserve
[MAX_NR_ZONES
];
294 * zone reclaim becomes active if more unmapped pages exist.
296 unsigned long min_unmapped_pages
;
297 unsigned long min_slab_pages
;
298 struct per_cpu_pageset
*pageset
[NR_CPUS
];
300 struct per_cpu_pageset pageset
[NR_CPUS
];
303 * free areas of different sizes
306 #ifdef CONFIG_MEMORY_HOTPLUG
307 /* see spanned/present_pages for more description */
308 seqlock_t span_seqlock
;
310 struct free_area free_area
[MAX_ORDER
];
312 #ifndef CONFIG_SPARSEMEM
314 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
315 * In SPARSEMEM, this map is stored in struct mem_section
317 unsigned long *pageblock_flags
;
318 #endif /* CONFIG_SPARSEMEM */
323 /* Fields commonly accessed by the page reclaim scanner */
326 struct list_head list
;
327 unsigned long nr_saved_scan
; /* accumulated for batching */
330 struct zone_reclaim_stat reclaim_stat
;
332 unsigned long pages_scanned
; /* since last reclaim */
333 unsigned long flags
; /* zone flags, see below */
335 /* Zone statistics */
336 atomic_long_t vm_stat
[NR_VM_ZONE_STAT_ITEMS
];
339 * prev_priority holds the scanning priority for this zone. It is
340 * defined as the scanning priority at which we achieved our reclaim
341 * target at the previous try_to_free_pages() or balance_pgdat()
344 * We use prev_priority as a measure of how much stress page reclaim is
345 * under - it drives the swappiness decision: whether to unmap mapped
348 * Access to both this field is quite racy even on uniprocessor. But
349 * it is expected to average out OK.
354 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
355 * this zone's LRU. Maintained by the pageout code.
357 unsigned int inactive_ratio
;
361 /* Rarely used or read-mostly fields */
364 * wait_table -- the array holding the hash table
365 * wait_table_hash_nr_entries -- the size of the hash table array
366 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
368 * The purpose of all these is to keep track of the people
369 * waiting for a page to become available and make them
370 * runnable again when possible. The trouble is that this
371 * consumes a lot of space, especially when so few things
372 * wait on pages at a given time. So instead of using
373 * per-page waitqueues, we use a waitqueue hash table.
375 * The bucket discipline is to sleep on the same queue when
376 * colliding and wake all in that wait queue when removing.
377 * When something wakes, it must check to be sure its page is
378 * truly available, a la thundering herd. The cost of a
379 * collision is great, but given the expected load of the
380 * table, they should be so rare as to be outweighed by the
381 * benefits from the saved space.
383 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
384 * primary users of these fields, and in mm/page_alloc.c
385 * free_area_init_core() performs the initialization of them.
387 wait_queue_head_t
* wait_table
;
388 unsigned long wait_table_hash_nr_entries
;
389 unsigned long wait_table_bits
;
392 * Discontig memory support fields.
394 struct pglist_data
*zone_pgdat
;
395 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
396 unsigned long zone_start_pfn
;
399 * zone_start_pfn, spanned_pages and present_pages are all
400 * protected by span_seqlock. It is a seqlock because it has
401 * to be read outside of zone->lock, and it is done in the main
402 * allocator path. But, it is written quite infrequently.
404 * The lock is declared along with zone->lock because it is
405 * frequently read in proximity to zone->lock. It's good to
406 * give them a chance of being in the same cacheline.
408 unsigned long spanned_pages
; /* total size, including holes */
409 unsigned long present_pages
; /* amount of memory (excluding holes) */
412 * rarely used fields:
415 } ____cacheline_internodealigned_in_smp
;
418 ZONE_ALL_UNRECLAIMABLE
, /* all pages pinned */
419 ZONE_RECLAIM_LOCKED
, /* prevents concurrent reclaim */
420 ZONE_OOM_LOCKED
, /* zone is in OOM killer zonelist */
423 static inline void zone_set_flag(struct zone
*zone
, zone_flags_t flag
)
425 set_bit(flag
, &zone
->flags
);
428 static inline int zone_test_and_set_flag(struct zone
*zone
, zone_flags_t flag
)
430 return test_and_set_bit(flag
, &zone
->flags
);
433 static inline void zone_clear_flag(struct zone
*zone
, zone_flags_t flag
)
435 clear_bit(flag
, &zone
->flags
);
438 static inline int zone_is_all_unreclaimable(const struct zone
*zone
)
440 return test_bit(ZONE_ALL_UNRECLAIMABLE
, &zone
->flags
);
443 static inline int zone_is_reclaim_locked(const struct zone
*zone
)
445 return test_bit(ZONE_RECLAIM_LOCKED
, &zone
->flags
);
448 static inline int zone_is_oom_locked(const struct zone
*zone
)
450 return test_bit(ZONE_OOM_LOCKED
, &zone
->flags
);
454 * The "priority" of VM scanning is how much of the queues we will scan in one
455 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
456 * queues ("queue_length >> 12") during an aging round.
458 #define DEF_PRIORITY 12
460 /* Maximum number of zones on a zonelist */
461 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
466 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
467 * allocations to a single node for GFP_THISNODE.
469 * [0] : Zonelist with fallback
470 * [1] : No fallback (GFP_THISNODE)
472 #define MAX_ZONELISTS 2
476 * We cache key information from each zonelist for smaller cache
477 * footprint when scanning for free pages in get_page_from_freelist().
479 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
480 * up short of free memory since the last time (last_fullzone_zap)
481 * we zero'd fullzones.
482 * 2) The array z_to_n[] maps each zone in the zonelist to its node
483 * id, so that we can efficiently evaluate whether that node is
484 * set in the current tasks mems_allowed.
486 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
487 * indexed by a zones offset in the zonelist zones[] array.
489 * The get_page_from_freelist() routine does two scans. During the
490 * first scan, we skip zones whose corresponding bit in 'fullzones'
491 * is set or whose corresponding node in current->mems_allowed (which
492 * comes from cpusets) is not set. During the second scan, we bypass
493 * this zonelist_cache, to ensure we look methodically at each zone.
495 * Once per second, we zero out (zap) fullzones, forcing us to
496 * reconsider nodes that might have regained more free memory.
497 * The field last_full_zap is the time we last zapped fullzones.
499 * This mechanism reduces the amount of time we waste repeatedly
500 * reexaming zones for free memory when they just came up low on
501 * memory momentarilly ago.
503 * The zonelist_cache struct members logically belong in struct
504 * zonelist. However, the mempolicy zonelists constructed for
505 * MPOL_BIND are intentionally variable length (and usually much
506 * shorter). A general purpose mechanism for handling structs with
507 * multiple variable length members is more mechanism than we want
508 * here. We resort to some special case hackery instead.
510 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
511 * part because they are shorter), so we put the fixed length stuff
512 * at the front of the zonelist struct, ending in a variable length
513 * zones[], as is needed by MPOL_BIND.
515 * Then we put the optional zonelist cache on the end of the zonelist
516 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
517 * the fixed length portion at the front of the struct. This pointer
518 * both enables us to find the zonelist cache, and in the case of
519 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
520 * to know that the zonelist cache is not there.
522 * The end result is that struct zonelists come in two flavors:
523 * 1) The full, fixed length version, shown below, and
524 * 2) The custom zonelists for MPOL_BIND.
525 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
527 * Even though there may be multiple CPU cores on a node modifying
528 * fullzones or last_full_zap in the same zonelist_cache at the same
529 * time, we don't lock it. This is just hint data - if it is wrong now
530 * and then, the allocator will still function, perhaps a bit slower.
534 struct zonelist_cache
{
535 unsigned short z_to_n
[MAX_ZONES_PER_ZONELIST
]; /* zone->nid */
536 DECLARE_BITMAP(fullzones
, MAX_ZONES_PER_ZONELIST
); /* zone full? */
537 unsigned long last_full_zap
; /* when last zap'd (jiffies) */
540 #define MAX_ZONELISTS 1
541 struct zonelist_cache
;
545 * This struct contains information about a zone in a zonelist. It is stored
546 * here to avoid dereferences into large structures and lookups of tables
549 struct zone
*zone
; /* Pointer to actual zone */
550 int zone_idx
; /* zone_idx(zoneref->zone) */
554 * One allocation request operates on a zonelist. A zonelist
555 * is a list of zones, the first one is the 'goal' of the
556 * allocation, the other zones are fallback zones, in decreasing
559 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
560 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
562 * To speed the reading of the zonelist, the zonerefs contain the zone index
563 * of the entry being read. Helper functions to access information given
564 * a struct zoneref are
566 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
567 * zonelist_zone_idx() - Return the index of the zone for an entry
568 * zonelist_node_idx() - Return the index of the node for an entry
571 struct zonelist_cache
*zlcache_ptr
; // NULL or &zlcache
572 struct zoneref _zonerefs
[MAX_ZONES_PER_ZONELIST
+ 1];
574 struct zonelist_cache zlcache
; // optional ...
578 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
579 struct node_active_region
{
580 unsigned long start_pfn
;
581 unsigned long end_pfn
;
584 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
586 #ifndef CONFIG_DISCONTIGMEM
587 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
588 extern struct page
*mem_map
;
592 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
593 * (mostly NUMA machines?) to denote a higher-level memory zone than the
596 * On NUMA machines, each NUMA node would have a pg_data_t to describe
597 * it's memory layout.
599 * Memory statistics and page replacement data structures are maintained on a
603 typedef struct pglist_data
{
604 struct zone node_zones
[MAX_NR_ZONES
];
605 struct zonelist node_zonelists
[MAX_ZONELISTS
];
607 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
608 struct page
*node_mem_map
;
609 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
610 struct page_cgroup
*node_page_cgroup
;
613 struct bootmem_data
*bdata
;
614 #ifdef CONFIG_MEMORY_HOTPLUG
616 * Must be held any time you expect node_start_pfn, node_present_pages
617 * or node_spanned_pages stay constant. Holding this will also
618 * guarantee that any pfn_valid() stays that way.
620 * Nests above zone->lock and zone->size_seqlock.
622 spinlock_t node_size_lock
;
624 unsigned long node_start_pfn
;
625 unsigned long node_present_pages
; /* total number of physical pages */
626 unsigned long node_spanned_pages
; /* total size of physical page
627 range, including holes */
629 wait_queue_head_t kswapd_wait
;
630 struct task_struct
*kswapd
;
631 int kswapd_max_order
;
634 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
635 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
636 #ifdef CONFIG_FLAT_NODE_MEM_MAP
637 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
639 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
641 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
643 #include <linux/memory_hotplug.h>
645 void get_zone_counts(unsigned long *active
, unsigned long *inactive
,
646 unsigned long *free
);
647 void build_all_zonelists(void);
648 void wakeup_kswapd(struct zone
*zone
, int order
);
649 int zone_watermark_ok(struct zone
*z
, int order
, unsigned long mark
,
650 int classzone_idx
, int alloc_flags
);
651 enum memmap_context
{
655 extern int init_currently_empty_zone(struct zone
*zone
, unsigned long start_pfn
,
657 enum memmap_context context
);
659 #ifdef CONFIG_HAVE_MEMORY_PRESENT
660 void memory_present(int nid
, unsigned long start
, unsigned long end
);
662 static inline void memory_present(int nid
, unsigned long start
, unsigned long end
) {}
665 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
666 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
670 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
672 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
674 static inline int populated_zone(struct zone
*zone
)
676 return (!!zone
->present_pages
);
679 extern int movable_zone
;
681 static inline int zone_movable_is_highmem(void)
683 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
684 return movable_zone
== ZONE_HIGHMEM
;
690 static inline int is_highmem_idx(enum zone_type idx
)
692 #ifdef CONFIG_HIGHMEM
693 return (idx
== ZONE_HIGHMEM
||
694 (idx
== ZONE_MOVABLE
&& zone_movable_is_highmem()));
700 static inline int is_normal_idx(enum zone_type idx
)
702 return (idx
== ZONE_NORMAL
);
706 * is_highmem - helper function to quickly check if a struct zone is a
707 * highmem zone or not. This is an attempt to keep references
708 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
709 * @zone - pointer to struct zone variable
711 static inline int is_highmem(struct zone
*zone
)
713 #ifdef CONFIG_HIGHMEM
714 int zone_off
= (char *)zone
- (char *)zone
->zone_pgdat
->node_zones
;
715 return zone_off
== ZONE_HIGHMEM
* sizeof(*zone
) ||
716 (zone_off
== ZONE_MOVABLE
* sizeof(*zone
) &&
717 zone_movable_is_highmem());
723 static inline int is_normal(struct zone
*zone
)
725 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_NORMAL
;
728 static inline int is_dma32(struct zone
*zone
)
730 #ifdef CONFIG_ZONE_DMA32
731 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_DMA32
;
737 static inline int is_dma(struct zone
*zone
)
739 #ifdef CONFIG_ZONE_DMA
740 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_DMA
;
746 /* These two functions are used to setup the per zone pages min values */
749 int min_free_kbytes_sysctl_handler(struct ctl_table
*, int, struct file
*,
750 void __user
*, size_t *, loff_t
*);
751 extern int sysctl_lowmem_reserve_ratio
[MAX_NR_ZONES
-1];
752 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table
*, int, struct file
*,
753 void __user
*, size_t *, loff_t
*);
754 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table
*, int, struct file
*,
755 void __user
*, size_t *, loff_t
*);
756 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table
*, int,
757 struct file
*, void __user
*, size_t *, loff_t
*);
758 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table
*, int,
759 struct file
*, void __user
*, size_t *, loff_t
*);
761 extern int numa_zonelist_order_handler(struct ctl_table
*, int,
762 struct file
*, void __user
*, size_t *, loff_t
*);
763 extern char numa_zonelist_order
[];
764 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
766 #ifndef CONFIG_NEED_MULTIPLE_NODES
768 extern struct pglist_data contig_page_data
;
769 #define NODE_DATA(nid) (&contig_page_data)
770 #define NODE_MEM_MAP(nid) mem_map
772 #else /* CONFIG_NEED_MULTIPLE_NODES */
774 #include <asm/mmzone.h>
776 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
778 extern struct pglist_data
*first_online_pgdat(void);
779 extern struct pglist_data
*next_online_pgdat(struct pglist_data
*pgdat
);
780 extern struct zone
*next_zone(struct zone
*zone
);
783 * for_each_online_pgdat - helper macro to iterate over all online nodes
784 * @pgdat - pointer to a pg_data_t variable
786 #define for_each_online_pgdat(pgdat) \
787 for (pgdat = first_online_pgdat(); \
789 pgdat = next_online_pgdat(pgdat))
791 * for_each_zone - helper macro to iterate over all memory zones
792 * @zone - pointer to struct zone variable
794 * The user only needs to declare the zone variable, for_each_zone
797 #define for_each_zone(zone) \
798 for (zone = (first_online_pgdat())->node_zones; \
800 zone = next_zone(zone))
802 #define for_each_populated_zone(zone) \
803 for (zone = (first_online_pgdat())->node_zones; \
805 zone = next_zone(zone)) \
806 if (!populated_zone(zone)) \
810 static inline struct zone
*zonelist_zone(struct zoneref
*zoneref
)
812 return zoneref
->zone
;
815 static inline int zonelist_zone_idx(struct zoneref
*zoneref
)
817 return zoneref
->zone_idx
;
820 static inline int zonelist_node_idx(struct zoneref
*zoneref
)
823 /* zone_to_nid not available in this context */
824 return zoneref
->zone
->node
;
827 #endif /* CONFIG_NUMA */
831 * 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
832 * @z - The cursor used as a starting point for the search
833 * @highest_zoneidx - The zone index of the highest zone to return
834 * @nodes - An optional nodemask to filter the zonelist with
835 * @zone - The first suitable zone found is returned via this parameter
837 * This function returns the next zone at or below a given zone index that is
838 * within the allowed nodemask using a cursor as the starting point for the
839 * search. The zoneref returned is a cursor that represents the current zone
840 * being examined. It should be advanced by one before calling
841 * next_zones_zonelist again.
843 struct zoneref
*next_zones_zonelist(struct zoneref
*z
,
844 enum zone_type highest_zoneidx
,
849 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
850 * @zonelist - The zonelist to search for a suitable zone
851 * @highest_zoneidx - The zone index of the highest zone to return
852 * @nodes - An optional nodemask to filter the zonelist with
853 * @zone - The first suitable zone found is returned via this parameter
855 * This function returns the first zone at or below a given zone index that is
856 * within the allowed nodemask. The zoneref returned is a cursor that can be
857 * used to iterate the zonelist with next_zones_zonelist by advancing it by
858 * one before calling.
860 static inline struct zoneref
*first_zones_zonelist(struct zonelist
*zonelist
,
861 enum zone_type highest_zoneidx
,
865 return next_zones_zonelist(zonelist
->_zonerefs
, highest_zoneidx
, nodes
,
870 * 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
871 * @zone - The current zone in the iterator
872 * @z - The current pointer within zonelist->zones being iterated
873 * @zlist - The zonelist being iterated
874 * @highidx - The zone index of the highest zone to return
875 * @nodemask - Nodemask allowed by the allocator
877 * This iterator iterates though all zones at or below a given zone index and
878 * within a given nodemask
880 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
881 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
883 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
886 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
887 * @zone - The current zone in the iterator
888 * @z - The current pointer within zonelist->zones being iterated
889 * @zlist - The zonelist being iterated
890 * @highidx - The zone index of the highest zone to return
892 * This iterator iterates though all zones at or below a given zone index.
894 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
895 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
897 #ifdef CONFIG_SPARSEMEM
898 #include <asm/sparsemem.h>
901 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
902 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
903 static inline unsigned long early_pfn_to_nid(unsigned long pfn
)
909 #ifdef CONFIG_FLATMEM
910 #define pfn_to_nid(pfn) (0)
913 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
914 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
916 #ifdef CONFIG_SPARSEMEM
919 * SECTION_SHIFT #bits space required to store a section #
921 * PA_SECTION_SHIFT physical address to/from section number
922 * PFN_SECTION_SHIFT pfn to/from section number
924 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
926 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
927 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
929 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
931 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
932 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
934 #define SECTION_BLOCKFLAGS_BITS \
935 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
937 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
938 #error Allocator MAX_ORDER exceeds SECTION_SIZE
945 * This is, logically, a pointer to an array of struct
946 * pages. However, it is stored with some other magic.
947 * (see sparse.c::sparse_init_one_section())
949 * Additionally during early boot we encode node id of
950 * the location of the section here to guide allocation.
951 * (see sparse.c::memory_present())
953 * Making it a UL at least makes someone do a cast
954 * before using it wrong.
956 unsigned long section_mem_map
;
958 /* See declaration of similar field in struct zone */
959 unsigned long *pageblock_flags
;
960 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
962 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
963 * section. (see memcontrol.h/page_cgroup.h about this.)
965 struct page_cgroup
*page_cgroup
;
970 #ifdef CONFIG_SPARSEMEM_EXTREME
971 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
973 #define SECTIONS_PER_ROOT 1
976 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
977 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
978 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
980 #ifdef CONFIG_SPARSEMEM_EXTREME
981 extern struct mem_section
*mem_section
[NR_SECTION_ROOTS
];
983 extern struct mem_section mem_section
[NR_SECTION_ROOTS
][SECTIONS_PER_ROOT
];
986 static inline struct mem_section
*__nr_to_section(unsigned long nr
)
988 if (!mem_section
[SECTION_NR_TO_ROOT(nr
)])
990 return &mem_section
[SECTION_NR_TO_ROOT(nr
)][nr
& SECTION_ROOT_MASK
];
992 extern int __section_nr(struct mem_section
* ms
);
993 extern unsigned long usemap_size(void);
996 * We use the lower bits of the mem_map pointer to store
997 * a little bit of information. There should be at least
998 * 3 bits here due to 32-bit alignment.
1000 #define SECTION_MARKED_PRESENT (1UL<<0)
1001 #define SECTION_HAS_MEM_MAP (1UL<<1)
1002 #define SECTION_MAP_LAST_BIT (1UL<<2)
1003 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1004 #define SECTION_NID_SHIFT 2
1006 static inline struct page
*__section_mem_map_addr(struct mem_section
*section
)
1008 unsigned long map
= section
->section_mem_map
;
1009 map
&= SECTION_MAP_MASK
;
1010 return (struct page
*)map
;
1013 static inline int present_section(struct mem_section
*section
)
1015 return (section
&& (section
->section_mem_map
& SECTION_MARKED_PRESENT
));
1018 static inline int present_section_nr(unsigned long nr
)
1020 return present_section(__nr_to_section(nr
));
1023 static inline int valid_section(struct mem_section
*section
)
1025 return (section
&& (section
->section_mem_map
& SECTION_HAS_MEM_MAP
));
1028 static inline int valid_section_nr(unsigned long nr
)
1030 return valid_section(__nr_to_section(nr
));
1033 static inline struct mem_section
*__pfn_to_section(unsigned long pfn
)
1035 return __nr_to_section(pfn_to_section_nr(pfn
));
1038 static inline int pfn_valid(unsigned long pfn
)
1040 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1042 return valid_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1045 static inline int pfn_present(unsigned long pfn
)
1047 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1049 return present_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1053 * These are _only_ used during initialisation, therefore they
1054 * can use __initdata ... They could have names to indicate
1058 #define pfn_to_nid(pfn) \
1060 unsigned long __pfn_to_nid_pfn = (pfn); \
1061 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1064 #define pfn_to_nid(pfn) (0)
1067 #define early_pfn_valid(pfn) pfn_valid(pfn)
1068 void sparse_init(void);
1070 #define sparse_init() do {} while (0)
1071 #define sparse_index_init(_sec, _nid) do {} while (0)
1072 #endif /* CONFIG_SPARSEMEM */
1074 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1075 bool early_pfn_in_nid(unsigned long pfn
, int nid
);
1077 #define early_pfn_in_nid(pfn, nid) (1)
1080 #ifndef early_pfn_valid
1081 #define early_pfn_valid(pfn) (1)
1084 void memory_present(int nid
, unsigned long start
, unsigned long end
);
1085 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
1088 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1089 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1090 * pfn_valid_within() should be used in this case; we optimise this away
1091 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1093 #ifdef CONFIG_HOLES_IN_ZONE
1094 #define pfn_valid_within(pfn) pfn_valid(pfn)
1096 #define pfn_valid_within(pfn) (1)
1099 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1101 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1102 * associated with it or not. In FLATMEM, it is expected that holes always
1103 * have valid memmap as long as there is valid PFNs either side of the hole.
1104 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1107 * However, an ARM, and maybe other embedded architectures in the future
1108 * free memmap backing holes to save memory on the assumption the memmap is
1109 * never used. The page_zone linkages are then broken even though pfn_valid()
1110 * returns true. A walker of the full memmap must then do this additional
1111 * check to ensure the memmap they are looking at is sane by making sure
1112 * the zone and PFN linkages are still valid. This is expensive, but walkers
1113 * of the full memmap are extremely rare.
1115 int memmap_valid_within(unsigned long pfn
,
1116 struct page
*page
, struct zone
*zone
);
1118 static inline int memmap_valid_within(unsigned long pfn
,
1119 struct page
*page
, struct zone
*zone
)
1123 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1125 #endif /* !__GENERATING_BOUNDS.H */
1126 #endif /* !__ASSEMBLY__ */
1127 #endif /* _LINUX_MMZONE_H */