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.
327 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
328 * this zone's LRU. Maintained by the pageout code.
330 unsigned int inactive_ratio;
334 /* Rarely used or read-mostly fields */
337 * wait_table -- the array holding the hash table
338 * wait_table_hash_nr_entries -- the size of the hash table array
339 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
341 * The purpose of all these is to keep track of the people
342 * waiting for a page to become available and make them
343 * runnable again when possible. The trouble is that this
344 * consumes a lot of space, especially when so few things
345 * wait on pages at a given time. So instead of using
346 * per-page waitqueues, we use a waitqueue hash table.
348 * The bucket discipline is to sleep on the same queue when
349 * colliding and wake all in that wait queue when removing.
350 * When something wakes, it must check to be sure its page is
351 * truly available, a la thundering herd. The cost of a
352 * collision is great, but given the expected load of the
353 * table, they should be so rare as to be outweighed by the
354 * benefits from the saved space.
356 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
357 * primary users of these fields, and in mm/page_alloc.c
358 * free_area_init_core() performs the initialization of them.
360 wait_queue_head_t * wait_table;
361 unsigned long wait_table_hash_nr_entries;
362 unsigned long wait_table_bits;
365 * Discontig memory support fields.
367 struct pglist_data *zone_pgdat;
368 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
369 unsigned long zone_start_pfn;
372 * zone_start_pfn, spanned_pages and present_pages are all
373 * protected by span_seqlock. It is a seqlock because it has
374 * to be read outside of zone->lock, and it is done in the main
375 * allocator path. But, it is written quite infrequently.
377 * The lock is declared along with zone->lock because it is
378 * frequently read in proximity to zone->lock. It's good to
379 * give them a chance of being in the same cacheline.
381 unsigned long spanned_pages; /* total size, including holes */
382 unsigned long present_pages; /* amount of memory (excluding holes) */
385 * rarely used fields:
388 } ____cacheline_internodealigned_in_smp;
391 ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */
392 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
393 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
396 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
398 set_bit(flag, &zone->flags);
401 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
403 return test_and_set_bit(flag, &zone->flags);
406 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
408 clear_bit(flag, &zone->flags);
411 static inline int zone_is_all_unreclaimable(const struct zone *zone)
413 return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags);
416 static inline int zone_is_reclaim_locked(const struct zone *zone)
418 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
421 static inline int zone_is_oom_locked(const struct zone *zone)
423 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
427 * The "priority" of VM scanning is how much of the queues we will scan in one
428 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
429 * queues ("queue_length >> 12") during an aging round.
431 #define DEF_PRIORITY 12
433 /* Maximum number of zones on a zonelist */
434 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
439 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
440 * allocations to a single node for GFP_THISNODE.
442 * [0] : Zonelist with fallback
443 * [1] : No fallback (GFP_THISNODE)
445 #define MAX_ZONELISTS 2
449 * We cache key information from each zonelist for smaller cache
450 * footprint when scanning for free pages in get_page_from_freelist().
452 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
453 * up short of free memory since the last time (last_fullzone_zap)
454 * we zero'd fullzones.
455 * 2) The array z_to_n[] maps each zone in the zonelist to its node
456 * id, so that we can efficiently evaluate whether that node is
457 * set in the current tasks mems_allowed.
459 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
460 * indexed by a zones offset in the zonelist zones[] array.
462 * The get_page_from_freelist() routine does two scans. During the
463 * first scan, we skip zones whose corresponding bit in 'fullzones'
464 * is set or whose corresponding node in current->mems_allowed (which
465 * comes from cpusets) is not set. During the second scan, we bypass
466 * this zonelist_cache, to ensure we look methodically at each zone.
468 * Once per second, we zero out (zap) fullzones, forcing us to
469 * reconsider nodes that might have regained more free memory.
470 * The field last_full_zap is the time we last zapped fullzones.
472 * This mechanism reduces the amount of time we waste repeatedly
473 * reexaming zones for free memory when they just came up low on
474 * memory momentarilly ago.
476 * The zonelist_cache struct members logically belong in struct
477 * zonelist. However, the mempolicy zonelists constructed for
478 * MPOL_BIND are intentionally variable length (and usually much
479 * shorter). A general purpose mechanism for handling structs with
480 * multiple variable length members is more mechanism than we want
481 * here. We resort to some special case hackery instead.
483 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
484 * part because they are shorter), so we put the fixed length stuff
485 * at the front of the zonelist struct, ending in a variable length
486 * zones[], as is needed by MPOL_BIND.
488 * Then we put the optional zonelist cache on the end of the zonelist
489 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
490 * the fixed length portion at the front of the struct. This pointer
491 * both enables us to find the zonelist cache, and in the case of
492 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
493 * to know that the zonelist cache is not there.
495 * The end result is that struct zonelists come in two flavors:
496 * 1) The full, fixed length version, shown below, and
497 * 2) The custom zonelists for MPOL_BIND.
498 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
500 * Even though there may be multiple CPU cores on a node modifying
501 * fullzones or last_full_zap in the same zonelist_cache at the same
502 * time, we don't lock it. This is just hint data - if it is wrong now
503 * and then, the allocator will still function, perhaps a bit slower.
507 struct zonelist_cache {
508 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
509 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
510 unsigned long last_full_zap; /* when last zap'd (jiffies) */
513 #define MAX_ZONELISTS 1
514 struct zonelist_cache;
518 * This struct contains information about a zone in a zonelist. It is stored
519 * here to avoid dereferences into large structures and lookups of tables
522 struct zone *zone; /* Pointer to actual zone */
523 int zone_idx; /* zone_idx(zoneref->zone) */
527 * One allocation request operates on a zonelist. A zonelist
528 * is a list of zones, the first one is the 'goal' of the
529 * allocation, the other zones are fallback zones, in decreasing
532 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
533 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
535 * To speed the reading of the zonelist, the zonerefs contain the zone index
536 * of the entry being read. Helper functions to access information given
537 * a struct zoneref are
539 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
540 * zonelist_zone_idx() - Return the index of the zone for an entry
541 * zonelist_node_idx() - Return the index of the node for an entry
544 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
545 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
547 struct zonelist_cache zlcache; // optional ...
551 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
552 struct node_active_region {
553 unsigned long start_pfn;
554 unsigned long end_pfn;
557 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
559 #ifndef CONFIG_DISCONTIGMEM
560 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
561 extern struct page *mem_map;
565 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
566 * (mostly NUMA machines?) to denote a higher-level memory zone than the
569 * On NUMA machines, each NUMA node would have a pg_data_t to describe
570 * it's memory layout.
572 * Memory statistics and page replacement data structures are maintained on a
576 typedef struct pglist_data {
577 struct zone node_zones[MAX_NR_ZONES];
578 struct zonelist node_zonelists[MAX_ZONELISTS];
580 #ifdef CONFIG_FLAT_NODE_MEM_MAP
581 struct page *node_mem_map;
583 struct bootmem_data *bdata;
584 #ifdef CONFIG_MEMORY_HOTPLUG
586 * Must be held any time you expect node_start_pfn, node_present_pages
587 * or node_spanned_pages stay constant. Holding this will also
588 * guarantee that any pfn_valid() stays that way.
590 * Nests above zone->lock and zone->size_seqlock.
592 spinlock_t node_size_lock;
594 unsigned long node_start_pfn;
595 unsigned long node_present_pages; /* total number of physical pages */
596 unsigned long node_spanned_pages; /* total size of physical page
597 range, including holes */
599 wait_queue_head_t kswapd_wait;
600 struct task_struct *kswapd;
601 int kswapd_max_order;
604 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
605 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
606 #ifdef CONFIG_FLAT_NODE_MEM_MAP
607 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
609 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
611 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
613 #include <linux/memory_hotplug.h>
615 void get_zone_counts(unsigned long *active, unsigned long *inactive,
616 unsigned long *free);
617 void build_all_zonelists(void);
618 void wakeup_kswapd(struct zone *zone, int order);
619 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
620 int classzone_idx, int alloc_flags);
621 enum memmap_context {
625 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
627 enum memmap_context context);
629 #ifdef CONFIG_HAVE_MEMORY_PRESENT
630 void memory_present(int nid, unsigned long start, unsigned long end);
632 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
635 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
636 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
640 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
642 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
644 static inline int populated_zone(struct zone *zone)
646 return (!!zone->present_pages);
649 extern int movable_zone;
651 static inline int zone_movable_is_highmem(void)
653 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
654 return movable_zone == ZONE_HIGHMEM;
660 static inline int is_highmem_idx(enum zone_type idx)
662 #ifdef CONFIG_HIGHMEM
663 return (idx == ZONE_HIGHMEM ||
664 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
670 static inline int is_normal_idx(enum zone_type idx)
672 return (idx == ZONE_NORMAL);
676 * is_highmem - helper function to quickly check if a struct zone is a
677 * highmem zone or not. This is an attempt to keep references
678 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
679 * @zone - pointer to struct zone variable
681 static inline int is_highmem(struct zone *zone)
683 #ifdef CONFIG_HIGHMEM
684 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
685 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
686 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
687 zone_movable_is_highmem());
693 static inline int is_normal(struct zone *zone)
695 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
698 static inline int is_dma32(struct zone *zone)
700 #ifdef CONFIG_ZONE_DMA32
701 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
707 static inline int is_dma(struct zone *zone)
709 #ifdef CONFIG_ZONE_DMA
710 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
716 /* These two functions are used to setup the per zone pages min values */
719 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
720 void __user *, size_t *, loff_t *);
721 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
722 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
723 void __user *, size_t *, loff_t *);
724 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
725 void __user *, size_t *, loff_t *);
726 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
727 struct file *, void __user *, size_t *, loff_t *);
728 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
729 struct file *, void __user *, size_t *, loff_t *);
731 extern int numa_zonelist_order_handler(struct ctl_table *, int,
732 struct file *, void __user *, size_t *, loff_t *);
733 extern char numa_zonelist_order[];
734 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
736 #include <linux/topology.h>
737 /* Returns the number of the current Node. */
739 #define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
742 #ifndef CONFIG_NEED_MULTIPLE_NODES
744 extern struct pglist_data contig_page_data;
745 #define NODE_DATA(nid) (&contig_page_data)
746 #define NODE_MEM_MAP(nid) mem_map
748 #else /* CONFIG_NEED_MULTIPLE_NODES */
750 #include <asm/mmzone.h>
752 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
754 extern struct pglist_data *first_online_pgdat(void);
755 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
756 extern struct zone *next_zone(struct zone *zone);
759 * for_each_online_pgdat - helper macro to iterate over all online nodes
760 * @pgdat - pointer to a pg_data_t variable
762 #define for_each_online_pgdat(pgdat) \
763 for (pgdat = first_online_pgdat(); \
765 pgdat = next_online_pgdat(pgdat))
767 * for_each_zone - helper macro to iterate over all memory zones
768 * @zone - pointer to struct zone variable
770 * The user only needs to declare the zone variable, for_each_zone
773 #define for_each_zone(zone) \
774 for (zone = (first_online_pgdat())->node_zones; \
776 zone = next_zone(zone))
778 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
780 return zoneref->zone;
783 static inline int zonelist_zone_idx(struct zoneref *zoneref)
785 return zoneref->zone_idx;
788 static inline int zonelist_node_idx(struct zoneref *zoneref)
791 /* zone_to_nid not available in this context */
792 return zoneref->zone->node;
795 #endif /* CONFIG_NUMA */
799 * 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
800 * @z - The cursor used as a starting point for the search
801 * @highest_zoneidx - The zone index of the highest zone to return
802 * @nodes - An optional nodemask to filter the zonelist with
803 * @zone - The first suitable zone found is returned via this parameter
805 * This function returns the next zone at or below a given zone index that is
806 * within the allowed nodemask using a cursor as the starting point for the
807 * search. The zoneref returned is a cursor that represents the current zone
808 * being examined. It should be advanced by one before calling
809 * next_zones_zonelist again.
811 struct zoneref *next_zones_zonelist(struct zoneref *z,
812 enum zone_type highest_zoneidx,
817 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
818 * @zonelist - The zonelist to search for a suitable zone
819 * @highest_zoneidx - The zone index of the highest zone to return
820 * @nodes - An optional nodemask to filter the zonelist with
821 * @zone - The first suitable zone found is returned via this parameter
823 * This function returns the first zone at or below a given zone index that is
824 * within the allowed nodemask. The zoneref returned is a cursor that can be
825 * used to iterate the zonelist with next_zones_zonelist by advancing it by
826 * one before calling.
828 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
829 enum zone_type highest_zoneidx,
833 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
838 * 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
839 * @zone - The current zone in the iterator
840 * @z - The current pointer within zonelist->zones being iterated
841 * @zlist - The zonelist being iterated
842 * @highidx - The zone index of the highest zone to return
843 * @nodemask - Nodemask allowed by the allocator
845 * This iterator iterates though all zones at or below a given zone index and
846 * within a given nodemask
848 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
849 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
851 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
854 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
855 * @zone - The current zone in the iterator
856 * @z - The current pointer within zonelist->zones being iterated
857 * @zlist - The zonelist being iterated
858 * @highidx - The zone index of the highest zone to return
860 * This iterator iterates though all zones at or below a given zone index.
862 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
863 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
865 #ifdef CONFIG_SPARSEMEM
866 #include <asm/sparsemem.h>
869 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
870 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
871 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
877 #ifdef CONFIG_FLATMEM
878 #define pfn_to_nid(pfn) (0)
881 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
882 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
884 #ifdef CONFIG_SPARSEMEM
887 * SECTION_SHIFT #bits space required to store a section #
889 * PA_SECTION_SHIFT physical address to/from section number
890 * PFN_SECTION_SHIFT pfn to/from section number
892 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
894 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
895 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
897 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
899 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
900 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
902 #define SECTION_BLOCKFLAGS_BITS \
903 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
905 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
906 #error Allocator MAX_ORDER exceeds SECTION_SIZE
912 * This is, logically, a pointer to an array of struct
913 * pages. However, it is stored with some other magic.
914 * (see sparse.c::sparse_init_one_section())
916 * Additionally during early boot we encode node id of
917 * the location of the section here to guide allocation.
918 * (see sparse.c::memory_present())
920 * Making it a UL at least makes someone do a cast
921 * before using it wrong.
923 unsigned long section_mem_map;
925 /* See declaration of similar field in struct zone */
926 unsigned long *pageblock_flags;
929 #ifdef CONFIG_SPARSEMEM_EXTREME
930 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
932 #define SECTIONS_PER_ROOT 1
935 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
936 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
937 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
939 #ifdef CONFIG_SPARSEMEM_EXTREME
940 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
942 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
945 static inline struct mem_section *__nr_to_section(unsigned long nr)
947 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
949 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
951 extern int __section_nr(struct mem_section* ms);
952 extern unsigned long usemap_size(void);
955 * We use the lower bits of the mem_map pointer to store
956 * a little bit of information. There should be at least
957 * 3 bits here due to 32-bit alignment.
959 #define SECTION_MARKED_PRESENT (1UL<<0)
960 #define SECTION_HAS_MEM_MAP (1UL<<1)
961 #define SECTION_MAP_LAST_BIT (1UL<<2)
962 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
963 #define SECTION_NID_SHIFT 2
965 static inline struct page *__section_mem_map_addr(struct mem_section *section)
967 unsigned long map = section->section_mem_map;
968 map &= SECTION_MAP_MASK;
969 return (struct page *)map;
972 static inline int present_section(struct mem_section *section)
974 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
977 static inline int present_section_nr(unsigned long nr)
979 return present_section(__nr_to_section(nr));
982 static inline int valid_section(struct mem_section *section)
984 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
987 static inline int valid_section_nr(unsigned long nr)
989 return valid_section(__nr_to_section(nr));
992 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
994 return __nr_to_section(pfn_to_section_nr(pfn));
997 static inline int pfn_valid(unsigned long pfn)
999 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1001 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1004 static inline int pfn_present(unsigned long pfn)
1006 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1008 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1012 * These are _only_ used during initialisation, therefore they
1013 * can use __initdata ... They could have names to indicate
1017 #define pfn_to_nid(pfn) \
1019 unsigned long __pfn_to_nid_pfn = (pfn); \
1020 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1023 #define pfn_to_nid(pfn) (0)
1026 #define early_pfn_valid(pfn) pfn_valid(pfn)
1027 void sparse_init(void);
1029 #define sparse_init() do {} while (0)
1030 #define sparse_index_init(_sec, _nid) do {} while (0)
1031 #endif /* CONFIG_SPARSEMEM */
1033 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1034 #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
1036 #define early_pfn_in_nid(pfn, nid) (1)
1039 #ifndef early_pfn_valid
1040 #define early_pfn_valid(pfn) (1)
1043 void memory_present(int nid, unsigned long start, unsigned long end);
1044 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1047 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1048 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1049 * pfn_valid_within() should be used in this case; we optimise this away
1050 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1052 #ifdef CONFIG_HOLES_IN_ZONE
1053 #define pfn_valid_within(pfn) pfn_valid(pfn)
1055 #define pfn_valid_within(pfn) (1)
1058 #endif /* !__GENERATING_BOUNDS.H */
1059 #endif /* !__ASSEMBLY__ */
1060 #endif /* _LINUX_MMZONE_H */