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 if (unlikely(page_group_by_mobility_disabled
))
54 return MIGRATE_UNMOVABLE
;
56 return get_pageblock_flags_group(page
, PB_migrate
, PB_migrate_end
);
60 struct list_head free_list
[MIGRATE_TYPES
];
61 unsigned long nr_free
;
67 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
68 * So add a wild amount of padding here to ensure that they fall into separate
69 * cachelines. There are very few zone structures in the machine, so space
70 * consumption is not a concern here.
72 #if defined(CONFIG_SMP)
75 } ____cacheline_internodealigned_in_smp
;
76 #define ZONE_PADDING(name) struct zone_padding name;
78 #define ZONE_PADDING(name)
82 /* First 128 byte cacheline (assuming 64 bit words) */
85 NR_INACTIVE_ANON
= NR_LRU_BASE
, /* must match order of LRU_[IN]ACTIVE */
86 NR_ACTIVE_ANON
, /* " " " " " */
87 NR_INACTIVE_FILE
, /* " " " " " */
88 NR_ACTIVE_FILE
, /* " " " " " */
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 */
98 NR_UNSTABLE_NFS
, /* NFS unstable pages */
101 /* Second 128 byte cacheline */
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
,
133 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
135 static inline int is_file_lru(enum lru_list l
)
137 return (l
== LRU_INACTIVE_FILE
|| l
== LRU_ACTIVE_FILE
);
140 static inline int is_active_lru(enum lru_list l
)
142 return (l
== LRU_ACTIVE_ANON
|| l
== LRU_ACTIVE_FILE
);
145 struct per_cpu_pages
{
146 int count
; /* number of pages in the list */
147 int high
; /* high watermark, emptying needed */
148 int batch
; /* chunk size for buddy add/remove */
149 struct list_head list
; /* the list of pages */
152 struct per_cpu_pageset
{
153 struct per_cpu_pages pcp
;
159 s8 vm_stat_diff
[NR_VM_ZONE_STAT_ITEMS
];
161 } ____cacheline_aligned_in_smp
;
164 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
166 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
169 #endif /* !__GENERATING_BOUNDS.H */
172 #ifdef CONFIG_ZONE_DMA
174 * ZONE_DMA is used when there are devices that are not able
175 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
176 * carve out the portion of memory that is needed for these devices.
177 * The range is arch specific.
182 * ---------------------------
183 * parisc, ia64, sparc <4G
186 * alpha Unlimited or 0-16MB.
188 * i386, x86_64 and multiple other arches
193 #ifdef CONFIG_ZONE_DMA32
195 * x86_64 needs two ZONE_DMAs because it supports devices that are
196 * only able to do DMA to the lower 16M but also 32 bit devices that
197 * can only do DMA areas below 4G.
202 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
203 * performed on pages in ZONE_NORMAL if the DMA devices support
204 * transfers to all addressable memory.
207 #ifdef CONFIG_HIGHMEM
209 * A memory area that is only addressable by the kernel through
210 * mapping portions into its own address space. This is for example
211 * used by i386 to allow the kernel to address the memory beyond
212 * 900MB. The kernel will set up special mappings (page
213 * table entries on i386) for each page that the kernel needs to
222 #ifndef __GENERATING_BOUNDS_H
225 * When a memory allocation must conform to specific limitations (such
226 * as being suitable for DMA) the caller will pass in hints to the
227 * allocator in the gfp_mask, in the zone modifier bits. These bits
228 * are used to select a priority ordered list of memory zones which
229 * match the requested limits. See gfp_zone() in include/linux/gfp.h
233 #define ZONES_SHIFT 0
234 #elif MAX_NR_ZONES <= 2
235 #define ZONES_SHIFT 1
236 #elif MAX_NR_ZONES <= 4
237 #define ZONES_SHIFT 2
239 #error ZONES_SHIFT -- too many zones configured adjust calculation
243 /* Fields commonly accessed by the page allocator */
244 unsigned long pages_min
, pages_low
, pages_high
;
246 * We don't know if the memory that we're going to allocate will be freeable
247 * or/and it will be released eventually, so to avoid totally wasting several
248 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
249 * to run OOM on the lower zones despite there's tons of freeable ram
250 * on the higher zones). This array is recalculated at runtime if the
251 * sysctl_lowmem_reserve_ratio sysctl changes.
253 unsigned long lowmem_reserve
[MAX_NR_ZONES
];
258 * zone reclaim becomes active if more unmapped pages exist.
260 unsigned long min_unmapped_pages
;
261 unsigned long min_slab_pages
;
262 struct per_cpu_pageset
*pageset
[NR_CPUS
];
264 struct per_cpu_pageset pageset
[NR_CPUS
];
267 * free areas of different sizes
270 #ifdef CONFIG_MEMORY_HOTPLUG
271 /* see spanned/present_pages for more description */
272 seqlock_t span_seqlock
;
274 struct free_area free_area
[MAX_ORDER
];
276 #ifndef CONFIG_SPARSEMEM
278 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
279 * In SPARSEMEM, this map is stored in struct mem_section
281 unsigned long *pageblock_flags
;
282 #endif /* CONFIG_SPARSEMEM */
287 /* Fields commonly accessed by the page reclaim scanner */
290 struct list_head list
;
291 unsigned long nr_scan
;
295 * The pageout code in vmscan.c keeps track of how many of the
296 * mem/swap backed and file backed pages are refeferenced.
297 * The higher the rotated/scanned ratio, the more valuable
300 * The anon LRU stats live in [0], file LRU stats in [1]
302 unsigned long recent_rotated
[2];
303 unsigned long recent_scanned
[2];
305 unsigned long pages_scanned
; /* since last reclaim */
306 unsigned long flags
; /* zone flags, see below */
308 /* Zone statistics */
309 atomic_long_t vm_stat
[NR_VM_ZONE_STAT_ITEMS
];
312 * prev_priority holds the scanning priority for this zone. It is
313 * defined as the scanning priority at which we achieved our reclaim
314 * target at the previous try_to_free_pages() or balance_pgdat()
317 * We use prev_priority as a measure of how much stress page reclaim is
318 * under - it drives the swappiness decision: whether to unmap mapped
321 * Access to both this field is quite racy even on uniprocessor. But
322 * it is expected to average out OK.
328 /* Rarely used or read-mostly fields */
331 * wait_table -- the array holding the hash table
332 * wait_table_hash_nr_entries -- the size of the hash table array
333 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
335 * The purpose of all these is to keep track of the people
336 * waiting for a page to become available and make them
337 * runnable again when possible. The trouble is that this
338 * consumes a lot of space, especially when so few things
339 * wait on pages at a given time. So instead of using
340 * per-page waitqueues, we use a waitqueue hash table.
342 * The bucket discipline is to sleep on the same queue when
343 * colliding and wake all in that wait queue when removing.
344 * When something wakes, it must check to be sure its page is
345 * truly available, a la thundering herd. The cost of a
346 * collision is great, but given the expected load of the
347 * table, they should be so rare as to be outweighed by the
348 * benefits from the saved space.
350 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
351 * primary users of these fields, and in mm/page_alloc.c
352 * free_area_init_core() performs the initialization of them.
354 wait_queue_head_t
* wait_table
;
355 unsigned long wait_table_hash_nr_entries
;
356 unsigned long wait_table_bits
;
359 * Discontig memory support fields.
361 struct pglist_data
*zone_pgdat
;
362 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
363 unsigned long zone_start_pfn
;
366 * zone_start_pfn, spanned_pages and present_pages are all
367 * protected by span_seqlock. It is a seqlock because it has
368 * to be read outside of zone->lock, and it is done in the main
369 * allocator path. But, it is written quite infrequently.
371 * The lock is declared along with zone->lock because it is
372 * frequently read in proximity to zone->lock. It's good to
373 * give them a chance of being in the same cacheline.
375 unsigned long spanned_pages
; /* total size, including holes */
376 unsigned long present_pages
; /* amount of memory (excluding holes) */
379 * rarely used fields:
382 } ____cacheline_internodealigned_in_smp
;
385 ZONE_ALL_UNRECLAIMABLE
, /* all pages pinned */
386 ZONE_RECLAIM_LOCKED
, /* prevents concurrent reclaim */
387 ZONE_OOM_LOCKED
, /* zone is in OOM killer zonelist */
390 static inline void zone_set_flag(struct zone
*zone
, zone_flags_t flag
)
392 set_bit(flag
, &zone
->flags
);
395 static inline int zone_test_and_set_flag(struct zone
*zone
, zone_flags_t flag
)
397 return test_and_set_bit(flag
, &zone
->flags
);
400 static inline void zone_clear_flag(struct zone
*zone
, zone_flags_t flag
)
402 clear_bit(flag
, &zone
->flags
);
405 static inline int zone_is_all_unreclaimable(const struct zone
*zone
)
407 return test_bit(ZONE_ALL_UNRECLAIMABLE
, &zone
->flags
);
410 static inline int zone_is_reclaim_locked(const struct zone
*zone
)
412 return test_bit(ZONE_RECLAIM_LOCKED
, &zone
->flags
);
415 static inline int zone_is_oom_locked(const struct zone
*zone
)
417 return test_bit(ZONE_OOM_LOCKED
, &zone
->flags
);
421 * The "priority" of VM scanning is how much of the queues we will scan in one
422 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
423 * queues ("queue_length >> 12") during an aging round.
425 #define DEF_PRIORITY 12
427 /* Maximum number of zones on a zonelist */
428 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
433 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
434 * allocations to a single node for GFP_THISNODE.
436 * [0] : Zonelist with fallback
437 * [1] : No fallback (GFP_THISNODE)
439 #define MAX_ZONELISTS 2
443 * We cache key information from each zonelist for smaller cache
444 * footprint when scanning for free pages in get_page_from_freelist().
446 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
447 * up short of free memory since the last time (last_fullzone_zap)
448 * we zero'd fullzones.
449 * 2) The array z_to_n[] maps each zone in the zonelist to its node
450 * id, so that we can efficiently evaluate whether that node is
451 * set in the current tasks mems_allowed.
453 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
454 * indexed by a zones offset in the zonelist zones[] array.
456 * The get_page_from_freelist() routine does two scans. During the
457 * first scan, we skip zones whose corresponding bit in 'fullzones'
458 * is set or whose corresponding node in current->mems_allowed (which
459 * comes from cpusets) is not set. During the second scan, we bypass
460 * this zonelist_cache, to ensure we look methodically at each zone.
462 * Once per second, we zero out (zap) fullzones, forcing us to
463 * reconsider nodes that might have regained more free memory.
464 * The field last_full_zap is the time we last zapped fullzones.
466 * This mechanism reduces the amount of time we waste repeatedly
467 * reexaming zones for free memory when they just came up low on
468 * memory momentarilly ago.
470 * The zonelist_cache struct members logically belong in struct
471 * zonelist. However, the mempolicy zonelists constructed for
472 * MPOL_BIND are intentionally variable length (and usually much
473 * shorter). A general purpose mechanism for handling structs with
474 * multiple variable length members is more mechanism than we want
475 * here. We resort to some special case hackery instead.
477 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
478 * part because they are shorter), so we put the fixed length stuff
479 * at the front of the zonelist struct, ending in a variable length
480 * zones[], as is needed by MPOL_BIND.
482 * Then we put the optional zonelist cache on the end of the zonelist
483 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
484 * the fixed length portion at the front of the struct. This pointer
485 * both enables us to find the zonelist cache, and in the case of
486 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
487 * to know that the zonelist cache is not there.
489 * The end result is that struct zonelists come in two flavors:
490 * 1) The full, fixed length version, shown below, and
491 * 2) The custom zonelists for MPOL_BIND.
492 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
494 * Even though there may be multiple CPU cores on a node modifying
495 * fullzones or last_full_zap in the same zonelist_cache at the same
496 * time, we don't lock it. This is just hint data - if it is wrong now
497 * and then, the allocator will still function, perhaps a bit slower.
501 struct zonelist_cache
{
502 unsigned short z_to_n
[MAX_ZONES_PER_ZONELIST
]; /* zone->nid */
503 DECLARE_BITMAP(fullzones
, MAX_ZONES_PER_ZONELIST
); /* zone full? */
504 unsigned long last_full_zap
; /* when last zap'd (jiffies) */
507 #define MAX_ZONELISTS 1
508 struct zonelist_cache
;
512 * This struct contains information about a zone in a zonelist. It is stored
513 * here to avoid dereferences into large structures and lookups of tables
516 struct zone
*zone
; /* Pointer to actual zone */
517 int zone_idx
; /* zone_idx(zoneref->zone) */
521 * One allocation request operates on a zonelist. A zonelist
522 * is a list of zones, the first one is the 'goal' of the
523 * allocation, the other zones are fallback zones, in decreasing
526 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
527 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
529 * To speed the reading of the zonelist, the zonerefs contain the zone index
530 * of the entry being read. Helper functions to access information given
531 * a struct zoneref are
533 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
534 * zonelist_zone_idx() - Return the index of the zone for an entry
535 * zonelist_node_idx() - Return the index of the node for an entry
538 struct zonelist_cache
*zlcache_ptr
; // NULL or &zlcache
539 struct zoneref _zonerefs
[MAX_ZONES_PER_ZONELIST
+ 1];
541 struct zonelist_cache zlcache
; // optional ...
545 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
546 struct node_active_region
{
547 unsigned long start_pfn
;
548 unsigned long end_pfn
;
551 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
553 #ifndef CONFIG_DISCONTIGMEM
554 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
555 extern struct page
*mem_map
;
559 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
560 * (mostly NUMA machines?) to denote a higher-level memory zone than the
563 * On NUMA machines, each NUMA node would have a pg_data_t to describe
564 * it's memory layout.
566 * Memory statistics and page replacement data structures are maintained on a
570 typedef struct pglist_data
{
571 struct zone node_zones
[MAX_NR_ZONES
];
572 struct zonelist node_zonelists
[MAX_ZONELISTS
];
574 #ifdef CONFIG_FLAT_NODE_MEM_MAP
575 struct page
*node_mem_map
;
577 struct bootmem_data
*bdata
;
578 #ifdef CONFIG_MEMORY_HOTPLUG
580 * Must be held any time you expect node_start_pfn, node_present_pages
581 * or node_spanned_pages stay constant. Holding this will also
582 * guarantee that any pfn_valid() stays that way.
584 * Nests above zone->lock and zone->size_seqlock.
586 spinlock_t node_size_lock
;
588 unsigned long node_start_pfn
;
589 unsigned long node_present_pages
; /* total number of physical pages */
590 unsigned long node_spanned_pages
; /* total size of physical page
591 range, including holes */
593 wait_queue_head_t kswapd_wait
;
594 struct task_struct
*kswapd
;
595 int kswapd_max_order
;
598 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
599 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
600 #ifdef CONFIG_FLAT_NODE_MEM_MAP
601 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
603 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
605 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
607 #include <linux/memory_hotplug.h>
609 void get_zone_counts(unsigned long *active
, unsigned long *inactive
,
610 unsigned long *free
);
611 void build_all_zonelists(void);
612 void wakeup_kswapd(struct zone
*zone
, int order
);
613 int zone_watermark_ok(struct zone
*z
, int order
, unsigned long mark
,
614 int classzone_idx
, int alloc_flags
);
615 enum memmap_context
{
619 extern int init_currently_empty_zone(struct zone
*zone
, unsigned long start_pfn
,
621 enum memmap_context context
);
623 #ifdef CONFIG_HAVE_MEMORY_PRESENT
624 void memory_present(int nid
, unsigned long start
, unsigned long end
);
626 static inline void memory_present(int nid
, unsigned long start
, unsigned long end
) {}
629 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
630 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
634 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
636 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
638 static inline int populated_zone(struct zone
*zone
)
640 return (!!zone
->present_pages
);
643 extern int movable_zone
;
645 static inline int zone_movable_is_highmem(void)
647 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
648 return movable_zone
== ZONE_HIGHMEM
;
654 static inline int is_highmem_idx(enum zone_type idx
)
656 #ifdef CONFIG_HIGHMEM
657 return (idx
== ZONE_HIGHMEM
||
658 (idx
== ZONE_MOVABLE
&& zone_movable_is_highmem()));
664 static inline int is_normal_idx(enum zone_type idx
)
666 return (idx
== ZONE_NORMAL
);
670 * is_highmem - helper function to quickly check if a struct zone is a
671 * highmem zone or not. This is an attempt to keep references
672 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
673 * @zone - pointer to struct zone variable
675 static inline int is_highmem(struct zone
*zone
)
677 #ifdef CONFIG_HIGHMEM
678 int zone_off
= (char *)zone
- (char *)zone
->zone_pgdat
->node_zones
;
679 return zone_off
== ZONE_HIGHMEM
* sizeof(*zone
) ||
680 (zone_off
== ZONE_MOVABLE
* sizeof(*zone
) &&
681 zone_movable_is_highmem());
687 static inline int is_normal(struct zone
*zone
)
689 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_NORMAL
;
692 static inline int is_dma32(struct zone
*zone
)
694 #ifdef CONFIG_ZONE_DMA32
695 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_DMA32
;
701 static inline int is_dma(struct zone
*zone
)
703 #ifdef CONFIG_ZONE_DMA
704 return zone
== zone
->zone_pgdat
->node_zones
+ ZONE_DMA
;
710 /* These two functions are used to setup the per zone pages min values */
713 int min_free_kbytes_sysctl_handler(struct ctl_table
*, int, struct file
*,
714 void __user
*, size_t *, loff_t
*);
715 extern int sysctl_lowmem_reserve_ratio
[MAX_NR_ZONES
-1];
716 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table
*, int, struct file
*,
717 void __user
*, size_t *, loff_t
*);
718 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table
*, int, struct file
*,
719 void __user
*, size_t *, loff_t
*);
720 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table
*, int,
721 struct file
*, void __user
*, size_t *, loff_t
*);
722 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table
*, int,
723 struct file
*, void __user
*, size_t *, loff_t
*);
725 extern int numa_zonelist_order_handler(struct ctl_table
*, int,
726 struct file
*, void __user
*, size_t *, loff_t
*);
727 extern char numa_zonelist_order
[];
728 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
730 #include <linux/topology.h>
731 /* Returns the number of the current Node. */
733 #define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
736 #ifndef CONFIG_NEED_MULTIPLE_NODES
738 extern struct pglist_data contig_page_data
;
739 #define NODE_DATA(nid) (&contig_page_data)
740 #define NODE_MEM_MAP(nid) mem_map
742 #else /* CONFIG_NEED_MULTIPLE_NODES */
744 #include <asm/mmzone.h>
746 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
748 extern struct pglist_data
*first_online_pgdat(void);
749 extern struct pglist_data
*next_online_pgdat(struct pglist_data
*pgdat
);
750 extern struct zone
*next_zone(struct zone
*zone
);
753 * for_each_online_pgdat - helper macro to iterate over all online nodes
754 * @pgdat - pointer to a pg_data_t variable
756 #define for_each_online_pgdat(pgdat) \
757 for (pgdat = first_online_pgdat(); \
759 pgdat = next_online_pgdat(pgdat))
761 * for_each_zone - helper macro to iterate over all memory zones
762 * @zone - pointer to struct zone variable
764 * The user only needs to declare the zone variable, for_each_zone
767 #define for_each_zone(zone) \
768 for (zone = (first_online_pgdat())->node_zones; \
770 zone = next_zone(zone))
772 static inline struct zone
*zonelist_zone(struct zoneref
*zoneref
)
774 return zoneref
->zone
;
777 static inline int zonelist_zone_idx(struct zoneref
*zoneref
)
779 return zoneref
->zone_idx
;
782 static inline int zonelist_node_idx(struct zoneref
*zoneref
)
785 /* zone_to_nid not available in this context */
786 return zoneref
->zone
->node
;
789 #endif /* CONFIG_NUMA */
793 * 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
794 * @z - The cursor used as a starting point for the search
795 * @highest_zoneidx - The zone index of the highest zone to return
796 * @nodes - An optional nodemask to filter the zonelist with
797 * @zone - The first suitable zone found is returned via this parameter
799 * This function returns the next zone at or below a given zone index that is
800 * within the allowed nodemask using a cursor as the starting point for the
801 * search. The zoneref returned is a cursor that represents the current zone
802 * being examined. It should be advanced by one before calling
803 * next_zones_zonelist again.
805 struct zoneref
*next_zones_zonelist(struct zoneref
*z
,
806 enum zone_type highest_zoneidx
,
811 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
812 * @zonelist - The zonelist to search for a suitable zone
813 * @highest_zoneidx - The zone index of the highest zone to return
814 * @nodes - An optional nodemask to filter the zonelist with
815 * @zone - The first suitable zone found is returned via this parameter
817 * This function returns the first zone at or below a given zone index that is
818 * within the allowed nodemask. The zoneref returned is a cursor that can be
819 * used to iterate the zonelist with next_zones_zonelist by advancing it by
820 * one before calling.
822 static inline struct zoneref
*first_zones_zonelist(struct zonelist
*zonelist
,
823 enum zone_type highest_zoneidx
,
827 return next_zones_zonelist(zonelist
->_zonerefs
, highest_zoneidx
, nodes
,
832 * 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
833 * @zone - The current zone in the iterator
834 * @z - The current pointer within zonelist->zones being iterated
835 * @zlist - The zonelist being iterated
836 * @highidx - The zone index of the highest zone to return
837 * @nodemask - Nodemask allowed by the allocator
839 * This iterator iterates though all zones at or below a given zone index and
840 * within a given nodemask
842 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
843 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
845 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
848 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
849 * @zone - The current zone in the iterator
850 * @z - The current pointer within zonelist->zones being iterated
851 * @zlist - The zonelist being iterated
852 * @highidx - The zone index of the highest zone to return
854 * This iterator iterates though all zones at or below a given zone index.
856 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
857 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
859 #ifdef CONFIG_SPARSEMEM
860 #include <asm/sparsemem.h>
863 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
864 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
865 static inline unsigned long early_pfn_to_nid(unsigned long pfn
)
871 #ifdef CONFIG_FLATMEM
872 #define pfn_to_nid(pfn) (0)
875 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
876 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
878 #ifdef CONFIG_SPARSEMEM
881 * SECTION_SHIFT #bits space required to store a section #
883 * PA_SECTION_SHIFT physical address to/from section number
884 * PFN_SECTION_SHIFT pfn to/from section number
886 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
888 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
889 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
891 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
893 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
894 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
896 #define SECTION_BLOCKFLAGS_BITS \
897 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
899 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
900 #error Allocator MAX_ORDER exceeds SECTION_SIZE
906 * This is, logically, a pointer to an array of struct
907 * pages. However, it is stored with some other magic.
908 * (see sparse.c::sparse_init_one_section())
910 * Additionally during early boot we encode node id of
911 * the location of the section here to guide allocation.
912 * (see sparse.c::memory_present())
914 * Making it a UL at least makes someone do a cast
915 * before using it wrong.
917 unsigned long section_mem_map
;
919 /* See declaration of similar field in struct zone */
920 unsigned long *pageblock_flags
;
923 #ifdef CONFIG_SPARSEMEM_EXTREME
924 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
926 #define SECTIONS_PER_ROOT 1
929 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
930 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
931 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
933 #ifdef CONFIG_SPARSEMEM_EXTREME
934 extern struct mem_section
*mem_section
[NR_SECTION_ROOTS
];
936 extern struct mem_section mem_section
[NR_SECTION_ROOTS
][SECTIONS_PER_ROOT
];
939 static inline struct mem_section
*__nr_to_section(unsigned long nr
)
941 if (!mem_section
[SECTION_NR_TO_ROOT(nr
)])
943 return &mem_section
[SECTION_NR_TO_ROOT(nr
)][nr
& SECTION_ROOT_MASK
];
945 extern int __section_nr(struct mem_section
* ms
);
946 extern unsigned long usemap_size(void);
949 * We use the lower bits of the mem_map pointer to store
950 * a little bit of information. There should be at least
951 * 3 bits here due to 32-bit alignment.
953 #define SECTION_MARKED_PRESENT (1UL<<0)
954 #define SECTION_HAS_MEM_MAP (1UL<<1)
955 #define SECTION_MAP_LAST_BIT (1UL<<2)
956 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
957 #define SECTION_NID_SHIFT 2
959 static inline struct page
*__section_mem_map_addr(struct mem_section
*section
)
961 unsigned long map
= section
->section_mem_map
;
962 map
&= SECTION_MAP_MASK
;
963 return (struct page
*)map
;
966 static inline int present_section(struct mem_section
*section
)
968 return (section
&& (section
->section_mem_map
& SECTION_MARKED_PRESENT
));
971 static inline int present_section_nr(unsigned long nr
)
973 return present_section(__nr_to_section(nr
));
976 static inline int valid_section(struct mem_section
*section
)
978 return (section
&& (section
->section_mem_map
& SECTION_HAS_MEM_MAP
));
981 static inline int valid_section_nr(unsigned long nr
)
983 return valid_section(__nr_to_section(nr
));
986 static inline struct mem_section
*__pfn_to_section(unsigned long pfn
)
988 return __nr_to_section(pfn_to_section_nr(pfn
));
991 static inline int pfn_valid(unsigned long pfn
)
993 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
995 return valid_section(__nr_to_section(pfn_to_section_nr(pfn
)));
998 static inline int pfn_present(unsigned long pfn
)
1000 if (pfn_to_section_nr(pfn
) >= NR_MEM_SECTIONS
)
1002 return present_section(__nr_to_section(pfn_to_section_nr(pfn
)));
1006 * These are _only_ used during initialisation, therefore they
1007 * can use __initdata ... They could have names to indicate
1011 #define pfn_to_nid(pfn) \
1013 unsigned long __pfn_to_nid_pfn = (pfn); \
1014 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1017 #define pfn_to_nid(pfn) (0)
1020 #define early_pfn_valid(pfn) pfn_valid(pfn)
1021 void sparse_init(void);
1023 #define sparse_init() do {} while (0)
1024 #define sparse_index_init(_sec, _nid) do {} while (0)
1025 #endif /* CONFIG_SPARSEMEM */
1027 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1028 #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
1030 #define early_pfn_in_nid(pfn, nid) (1)
1033 #ifndef early_pfn_valid
1034 #define early_pfn_valid(pfn) (1)
1037 void memory_present(int nid
, unsigned long start
, unsigned long end
);
1038 unsigned long __init
node_memmap_size_bytes(int, unsigned long, unsigned long);
1041 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1042 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1043 * pfn_valid_within() should be used in this case; we optimise this away
1044 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1046 #ifdef CONFIG_HOLES_IN_ZONE
1047 #define pfn_valid_within(pfn) pfn_valid(pfn)
1049 #define pfn_valid_within(pfn) (1)
1052 #endif /* !__GENERATING_BOUNDS.H */
1053 #endif /* !__ASSEMBLY__ */
1054 #endif /* _LINUX_MMZONE_H */