1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
5 #ifndef __GENERATING_BOUNDS_H
7 #include <linux/spinlock.h>
8 #include <linux/list.h>
9 #include <linux/wait.h>
10 #include <linux/bitops.h>
11 #include <linux/cache.h>
12 #include <linux/threads.h>
13 #include <linux/numa.h>
14 #include <linux/init.h>
15 #include <linux/seqlock.h>
16 #include <linux/nodemask.h>
17 #include <linux/pageblock-flags.h>
18 #include <generated/bounds.h>
19 #include <linux/atomic.h>
22 /* Free memory management - zoned buddy allocator. */
23 #ifndef CONFIG_FORCE_MAX_ZONEORDER
26 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
28 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
31 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
32 * costly to service. That is between allocation orders which should
33 * coelesce naturally under reasonable reclaim pressure and those which
36 #define PAGE_ALLOC_COSTLY_ORDER 3
38 #define MIGRATE_UNMOVABLE 0
39 #define MIGRATE_RECLAIMABLE 1
40 #define MIGRATE_MOVABLE 2
41 #define MIGRATE_PCPTYPES 3 /* the number of types on the pcp lists */
42 #define MIGRATE_RESERVE 3
43 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
44 #define MIGRATE_TYPES 5
46 #define for_each_migratetype_order(order, type) \
47 for (order = 0; order < MAX_ORDER; order++) \
48 for (type = 0; type < MIGRATE_TYPES; type++)
50 extern int page_group_by_mobility_disabled;
52 static inline int get_pageblock_migratetype(struct page *page)
54 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
58 struct list_head free_list[MIGRATE_TYPES];
59 unsigned long nr_free;
65 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
66 * So add a wild amount of padding here to ensure that they fall into separate
67 * cachelines. There are very few zone structures in the machine, so space
68 * consumption is not a concern here.
70 #if defined(CONFIG_SMP)
73 } ____cacheline_internodealigned_in_smp;
74 #define ZONE_PADDING(name) struct zone_padding name;
76 #define ZONE_PADDING(name)
80 /* First 128 byte cacheline (assuming 64 bit words) */
83 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
84 NR_ACTIVE_ANON, /* " " " " " */
85 NR_INACTIVE_FILE, /* " " " " " */
86 NR_ACTIVE_FILE, /* " " " " " */
87 NR_UNEVICTABLE, /* " " " " " */
88 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
89 NR_ANON_PAGES, /* Mapped anonymous pages */
90 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
91 only modified from process context */
96 NR_SLAB_UNRECLAIMABLE,
97 NR_PAGETABLE, /* used for pagetables */
99 /* Second 128 byte cacheline */
100 NR_UNSTABLE_NFS, /* NFS unstable pages */
103 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
104 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
105 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
106 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
107 NR_DIRTIED, /* page dirtyings since bootup */
108 NR_WRITTEN, /* page writings since bootup */
110 NUMA_HIT, /* allocated in intended node */
111 NUMA_MISS, /* allocated in non intended node */
112 NUMA_FOREIGN, /* was intended here, hit elsewhere */
113 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
114 NUMA_LOCAL, /* allocation from local node */
115 NUMA_OTHER, /* allocation from other node */
117 NR_ANON_TRANSPARENT_HUGEPAGES,
118 NR_VM_ZONE_STAT_ITEMS };
121 * We do arithmetic on the LRU lists in various places in the code,
122 * so it is important to keep the active lists LRU_ACTIVE higher in
123 * the array than the corresponding inactive lists, and to keep
124 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
126 * This has to be kept in sync with the statistics in zone_stat_item
127 * above and the descriptions in vmstat_text in mm/vmstat.c
134 LRU_INACTIVE_ANON = LRU_BASE,
135 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
136 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
137 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
142 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
144 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
146 static inline int is_file_lru(enum lru_list l)
148 return (l == LRU_INACTIVE_FILE || l == LRU_ACTIVE_FILE);
151 static inline int is_active_lru(enum lru_list l)
153 return (l == LRU_ACTIVE_ANON || l == LRU_ACTIVE_FILE);
156 static inline int is_unevictable_lru(enum lru_list l)
158 return (l == LRU_UNEVICTABLE);
161 /* Mask used at gathering information at once (see memcontrol.c) */
162 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
163 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
164 #define LRU_ALL_EVICTABLE (LRU_ALL_FILE | LRU_ALL_ANON)
165 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
167 /* Isolate inactive pages */
168 #define ISOLATE_INACTIVE ((__force isolate_mode_t)0x1)
169 /* Isolate active pages */
170 #define ISOLATE_ACTIVE ((__force isolate_mode_t)0x2)
172 /* LRU Isolation modes. */
173 typedef unsigned __bitwise__ isolate_mode_t;
175 enum zone_watermarks {
182 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
183 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
184 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
186 struct per_cpu_pages {
187 int count; /* number of pages in the list */
188 int high; /* high watermark, emptying needed */
189 int batch; /* chunk size for buddy add/remove */
191 /* Lists of pages, one per migrate type stored on the pcp-lists */
192 struct list_head lists[MIGRATE_PCPTYPES];
195 struct per_cpu_pageset {
196 struct per_cpu_pages pcp;
202 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
206 #endif /* !__GENERATING_BOUNDS.H */
209 #ifdef CONFIG_ZONE_DMA
211 * ZONE_DMA is used when there are devices that are not able
212 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
213 * carve out the portion of memory that is needed for these devices.
214 * The range is arch specific.
219 * ---------------------------
220 * parisc, ia64, sparc <4G
223 * alpha Unlimited or 0-16MB.
225 * i386, x86_64 and multiple other arches
230 #ifdef CONFIG_ZONE_DMA32
232 * x86_64 needs two ZONE_DMAs because it supports devices that are
233 * only able to do DMA to the lower 16M but also 32 bit devices that
234 * can only do DMA areas below 4G.
239 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
240 * performed on pages in ZONE_NORMAL if the DMA devices support
241 * transfers to all addressable memory.
244 #ifdef CONFIG_HIGHMEM
246 * A memory area that is only addressable by the kernel through
247 * mapping portions into its own address space. This is for example
248 * used by i386 to allow the kernel to address the memory beyond
249 * 900MB. The kernel will set up special mappings (page
250 * table entries on i386) for each page that the kernel needs to
259 #ifndef __GENERATING_BOUNDS_H
262 * When a memory allocation must conform to specific limitations (such
263 * as being suitable for DMA) the caller will pass in hints to the
264 * allocator in the gfp_mask, in the zone modifier bits. These bits
265 * are used to select a priority ordered list of memory zones which
266 * match the requested limits. See gfp_zone() in include/linux/gfp.h
270 #define ZONES_SHIFT 0
271 #elif MAX_NR_ZONES <= 2
272 #define ZONES_SHIFT 1
273 #elif MAX_NR_ZONES <= 4
274 #define ZONES_SHIFT 2
276 #error ZONES_SHIFT -- too many zones configured adjust calculation
279 struct zone_reclaim_stat {
281 * The pageout code in vmscan.c keeps track of how many of the
282 * mem/swap backed and file backed pages are refeferenced.
283 * The higher the rotated/scanned ratio, the more valuable
286 * The anon LRU stats live in [0], file LRU stats in [1]
288 unsigned long recent_rotated[2];
289 unsigned long recent_scanned[2];
293 /* Fields commonly accessed by the page allocator */
295 /* zone watermarks, access with *_wmark_pages(zone) macros */
296 unsigned long watermark[NR_WMARK];
299 * When free pages are below this point, additional steps are taken
300 * when reading the number of free pages to avoid per-cpu counter
301 * drift allowing watermarks to be breached
303 unsigned long percpu_drift_mark;
306 * We don't know if the memory that we're going to allocate will be freeable
307 * or/and it will be released eventually, so to avoid totally wasting several
308 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
309 * to run OOM on the lower zones despite there's tons of freeable ram
310 * on the higher zones). This array is recalculated at runtime if the
311 * sysctl_lowmem_reserve_ratio sysctl changes.
313 unsigned long lowmem_reserve[MAX_NR_ZONES];
318 * zone reclaim becomes active if more unmapped pages exist.
320 unsigned long min_unmapped_pages;
321 unsigned long min_slab_pages;
323 struct per_cpu_pageset __percpu *pageset;
325 * free areas of different sizes
328 int all_unreclaimable; /* All pages pinned */
329 #ifdef CONFIG_MEMORY_HOTPLUG
330 /* see spanned/present_pages for more description */
331 seqlock_t span_seqlock;
333 struct free_area free_area[MAX_ORDER];
335 #ifndef CONFIG_SPARSEMEM
337 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
338 * In SPARSEMEM, this map is stored in struct mem_section
340 unsigned long *pageblock_flags;
341 #endif /* CONFIG_SPARSEMEM */
343 #ifdef CONFIG_COMPACTION
345 * On compaction failure, 1<<compact_defer_shift compactions
346 * are skipped before trying again. The number attempted since
347 * last failure is tracked with compact_considered.
349 unsigned int compact_considered;
350 unsigned int compact_defer_shift;
355 /* Fields commonly accessed by the page reclaim scanner */
358 struct list_head list;
361 struct zone_reclaim_stat reclaim_stat;
363 unsigned long pages_scanned; /* since last reclaim */
364 unsigned long flags; /* zone flags, see below */
366 /* Zone statistics */
367 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
370 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
371 * this zone's LRU. Maintained by the pageout code.
373 unsigned int inactive_ratio;
377 /* Rarely used or read-mostly fields */
380 * wait_table -- the array holding the hash table
381 * wait_table_hash_nr_entries -- the size of the hash table array
382 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
384 * The purpose of all these is to keep track of the people
385 * waiting for a page to become available and make them
386 * runnable again when possible. The trouble is that this
387 * consumes a lot of space, especially when so few things
388 * wait on pages at a given time. So instead of using
389 * per-page waitqueues, we use a waitqueue hash table.
391 * The bucket discipline is to sleep on the same queue when
392 * colliding and wake all in that wait queue when removing.
393 * When something wakes, it must check to be sure its page is
394 * truly available, a la thundering herd. The cost of a
395 * collision is great, but given the expected load of the
396 * table, they should be so rare as to be outweighed by the
397 * benefits from the saved space.
399 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
400 * primary users of these fields, and in mm/page_alloc.c
401 * free_area_init_core() performs the initialization of them.
403 wait_queue_head_t * wait_table;
404 unsigned long wait_table_hash_nr_entries;
405 unsigned long wait_table_bits;
408 * Discontig memory support fields.
410 struct pglist_data *zone_pgdat;
411 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
412 unsigned long zone_start_pfn;
415 * zone_start_pfn, spanned_pages and present_pages are all
416 * protected by span_seqlock. It is a seqlock because it has
417 * to be read outside of zone->lock, and it is done in the main
418 * allocator path. But, it is written quite infrequently.
420 * The lock is declared along with zone->lock because it is
421 * frequently read in proximity to zone->lock. It's good to
422 * give them a chance of being in the same cacheline.
424 unsigned long spanned_pages; /* total size, including holes */
425 unsigned long present_pages; /* amount of memory (excluding holes) */
428 * rarely used fields:
431 } ____cacheline_internodealigned_in_smp;
434 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
435 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
436 ZONE_CONGESTED, /* zone has many dirty pages backed by
441 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
443 set_bit(flag, &zone->flags);
446 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
448 return test_and_set_bit(flag, &zone->flags);
451 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
453 clear_bit(flag, &zone->flags);
456 static inline int zone_is_reclaim_congested(const struct zone *zone)
458 return test_bit(ZONE_CONGESTED, &zone->flags);
461 static inline int zone_is_reclaim_locked(const struct zone *zone)
463 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
466 static inline int zone_is_oom_locked(const struct zone *zone)
468 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
472 * The "priority" of VM scanning is how much of the queues we will scan in one
473 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
474 * queues ("queue_length >> 12") during an aging round.
476 #define DEF_PRIORITY 12
478 /* Maximum number of zones on a zonelist */
479 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
484 * The NUMA zonelists are doubled because we need zonelists that restrict the
485 * allocations to a single node for GFP_THISNODE.
487 * [0] : Zonelist with fallback
488 * [1] : No fallback (GFP_THISNODE)
490 #define MAX_ZONELISTS 2
494 * We cache key information from each zonelist for smaller cache
495 * footprint when scanning for free pages in get_page_from_freelist().
497 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
498 * up short of free memory since the last time (last_fullzone_zap)
499 * we zero'd fullzones.
500 * 2) The array z_to_n[] maps each zone in the zonelist to its node
501 * id, so that we can efficiently evaluate whether that node is
502 * set in the current tasks mems_allowed.
504 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
505 * indexed by a zones offset in the zonelist zones[] array.
507 * The get_page_from_freelist() routine does two scans. During the
508 * first scan, we skip zones whose corresponding bit in 'fullzones'
509 * is set or whose corresponding node in current->mems_allowed (which
510 * comes from cpusets) is not set. During the second scan, we bypass
511 * this zonelist_cache, to ensure we look methodically at each zone.
513 * Once per second, we zero out (zap) fullzones, forcing us to
514 * reconsider nodes that might have regained more free memory.
515 * The field last_full_zap is the time we last zapped fullzones.
517 * This mechanism reduces the amount of time we waste repeatedly
518 * reexaming zones for free memory when they just came up low on
519 * memory momentarilly ago.
521 * The zonelist_cache struct members logically belong in struct
522 * zonelist. However, the mempolicy zonelists constructed for
523 * MPOL_BIND are intentionally variable length (and usually much
524 * shorter). A general purpose mechanism for handling structs with
525 * multiple variable length members is more mechanism than we want
526 * here. We resort to some special case hackery instead.
528 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
529 * part because they are shorter), so we put the fixed length stuff
530 * at the front of the zonelist struct, ending in a variable length
531 * zones[], as is needed by MPOL_BIND.
533 * Then we put the optional zonelist cache on the end of the zonelist
534 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
535 * the fixed length portion at the front of the struct. This pointer
536 * both enables us to find the zonelist cache, and in the case of
537 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
538 * to know that the zonelist cache is not there.
540 * The end result is that struct zonelists come in two flavors:
541 * 1) The full, fixed length version, shown below, and
542 * 2) The custom zonelists for MPOL_BIND.
543 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
545 * Even though there may be multiple CPU cores on a node modifying
546 * fullzones or last_full_zap in the same zonelist_cache at the same
547 * time, we don't lock it. This is just hint data - if it is wrong now
548 * and then, the allocator will still function, perhaps a bit slower.
552 struct zonelist_cache {
553 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
554 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
555 unsigned long last_full_zap; /* when last zap'd (jiffies) */
558 #define MAX_ZONELISTS 1
559 struct zonelist_cache;
563 * This struct contains information about a zone in a zonelist. It is stored
564 * here to avoid dereferences into large structures and lookups of tables
567 struct zone *zone; /* Pointer to actual zone */
568 int zone_idx; /* zone_idx(zoneref->zone) */
572 * One allocation request operates on a zonelist. A zonelist
573 * is a list of zones, the first one is the 'goal' of the
574 * allocation, the other zones are fallback zones, in decreasing
577 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
578 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
580 * To speed the reading of the zonelist, the zonerefs contain the zone index
581 * of the entry being read. Helper functions to access information given
582 * a struct zoneref are
584 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
585 * zonelist_zone_idx() - Return the index of the zone for an entry
586 * zonelist_node_idx() - Return the index of the node for an entry
589 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
590 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
592 struct zonelist_cache zlcache; // optional ...
596 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
597 struct node_active_region {
598 unsigned long start_pfn;
599 unsigned long end_pfn;
602 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
604 #ifndef CONFIG_DISCONTIGMEM
605 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
606 extern struct page *mem_map;
610 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
611 * (mostly NUMA machines?) to denote a higher-level memory zone than the
614 * On NUMA machines, each NUMA node would have a pg_data_t to describe
615 * it's memory layout.
617 * Memory statistics and page replacement data structures are maintained on a
621 typedef struct pglist_data {
622 struct zone node_zones[MAX_NR_ZONES];
623 struct zonelist node_zonelists[MAX_ZONELISTS];
625 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
626 struct page *node_mem_map;
627 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
628 struct page_cgroup *node_page_cgroup;
631 #ifndef CONFIG_NO_BOOTMEM
632 struct bootmem_data *bdata;
634 #ifdef CONFIG_MEMORY_HOTPLUG
636 * Must be held any time you expect node_start_pfn, node_present_pages
637 * or node_spanned_pages stay constant. Holding this will also
638 * guarantee that any pfn_valid() stays that way.
640 * Nests above zone->lock and zone->size_seqlock.
642 spinlock_t node_size_lock;
644 unsigned long node_start_pfn;
645 unsigned long node_present_pages; /* total number of physical pages */
646 unsigned long node_spanned_pages; /* total size of physical page
647 range, including holes */
649 wait_queue_head_t kswapd_wait;
650 struct task_struct *kswapd;
651 int kswapd_max_order;
652 enum zone_type classzone_idx;
655 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
656 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
657 #ifdef CONFIG_FLAT_NODE_MEM_MAP
658 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
660 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
662 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
664 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
666 #define node_end_pfn(nid) ({\
667 pg_data_t *__pgdat = NODE_DATA(nid);\
668 __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
671 #include <linux/memory_hotplug.h>
673 extern struct mutex zonelists_mutex;
674 void build_all_zonelists(void *data);
675 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
676 bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
677 int classzone_idx, int alloc_flags);
678 bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
679 int classzone_idx, int alloc_flags);
680 enum memmap_context {
684 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
686 enum memmap_context context);
688 #ifdef CONFIG_HAVE_MEMORY_PRESENT
689 void memory_present(int nid, unsigned long start, unsigned long end);
691 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
694 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
695 int local_memory_node(int node_id);
697 static inline int local_memory_node(int node_id) { return node_id; };
700 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
701 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
705 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
707 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
709 static inline int populated_zone(struct zone *zone)
711 return (!!zone->present_pages);
714 extern int movable_zone;
716 static inline int zone_movable_is_highmem(void)
718 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
719 return movable_zone == ZONE_HIGHMEM;
725 static inline int is_highmem_idx(enum zone_type idx)
727 #ifdef CONFIG_HIGHMEM
728 return (idx == ZONE_HIGHMEM ||
729 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
735 static inline int is_normal_idx(enum zone_type idx)
737 return (idx == ZONE_NORMAL);
741 * is_highmem - helper function to quickly check if a struct zone is a
742 * highmem zone or not. This is an attempt to keep references
743 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
744 * @zone - pointer to struct zone variable
746 static inline int is_highmem(struct zone *zone)
748 #ifdef CONFIG_HIGHMEM
749 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
750 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
751 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
752 zone_movable_is_highmem());
758 static inline int is_normal(struct zone *zone)
760 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
763 static inline int is_dma32(struct zone *zone)
765 #ifdef CONFIG_ZONE_DMA32
766 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
772 static inline int is_dma(struct zone *zone)
774 #ifdef CONFIG_ZONE_DMA
775 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
781 /* These two functions are used to setup the per zone pages min values */
783 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
784 void __user *, size_t *, loff_t *);
785 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
786 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
787 void __user *, size_t *, loff_t *);
788 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
789 void __user *, size_t *, loff_t *);
790 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
791 void __user *, size_t *, loff_t *);
792 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
793 void __user *, size_t *, loff_t *);
795 extern int numa_zonelist_order_handler(struct ctl_table *, int,
796 void __user *, size_t *, loff_t *);
797 extern char numa_zonelist_order[];
798 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
800 #ifndef CONFIG_NEED_MULTIPLE_NODES
802 extern struct pglist_data contig_page_data;
803 #define NODE_DATA(nid) (&contig_page_data)
804 #define NODE_MEM_MAP(nid) mem_map
806 #else /* CONFIG_NEED_MULTIPLE_NODES */
808 #include <asm/mmzone.h>
810 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
812 extern struct pglist_data *first_online_pgdat(void);
813 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
814 extern struct zone *next_zone(struct zone *zone);
817 * for_each_online_pgdat - helper macro to iterate over all online nodes
818 * @pgdat - pointer to a pg_data_t variable
820 #define for_each_online_pgdat(pgdat) \
821 for (pgdat = first_online_pgdat(); \
823 pgdat = next_online_pgdat(pgdat))
825 * for_each_zone - helper macro to iterate over all memory zones
826 * @zone - pointer to struct zone variable
828 * The user only needs to declare the zone variable, for_each_zone
831 #define for_each_zone(zone) \
832 for (zone = (first_online_pgdat())->node_zones; \
834 zone = next_zone(zone))
836 #define for_each_populated_zone(zone) \
837 for (zone = (first_online_pgdat())->node_zones; \
839 zone = next_zone(zone)) \
840 if (!populated_zone(zone)) \
844 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
846 return zoneref->zone;
849 static inline int zonelist_zone_idx(struct zoneref *zoneref)
851 return zoneref->zone_idx;
854 static inline int zonelist_node_idx(struct zoneref *zoneref)
857 /* zone_to_nid not available in this context */
858 return zoneref->zone->node;
861 #endif /* CONFIG_NUMA */
865 * 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
866 * @z - The cursor used as a starting point for the search
867 * @highest_zoneidx - The zone index of the highest zone to return
868 * @nodes - An optional nodemask to filter the zonelist with
869 * @zone - The first suitable zone found is returned via this parameter
871 * This function returns the next zone at or below a given zone index that is
872 * within the allowed nodemask using a cursor as the starting point for the
873 * search. The zoneref returned is a cursor that represents the current zone
874 * being examined. It should be advanced by one before calling
875 * next_zones_zonelist again.
877 struct zoneref *next_zones_zonelist(struct zoneref *z,
878 enum zone_type highest_zoneidx,
883 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
884 * @zonelist - The zonelist to search for a suitable zone
885 * @highest_zoneidx - The zone index of the highest zone to return
886 * @nodes - An optional nodemask to filter the zonelist with
887 * @zone - The first suitable zone found is returned via this parameter
889 * This function returns the first zone at or below a given zone index that is
890 * within the allowed nodemask. The zoneref returned is a cursor that can be
891 * used to iterate the zonelist with next_zones_zonelist by advancing it by
892 * one before calling.
894 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
895 enum zone_type highest_zoneidx,
899 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
904 * 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
905 * @zone - The current zone in the iterator
906 * @z - The current pointer within zonelist->zones being iterated
907 * @zlist - The zonelist being iterated
908 * @highidx - The zone index of the highest zone to return
909 * @nodemask - Nodemask allowed by the allocator
911 * This iterator iterates though all zones at or below a given zone index and
912 * within a given nodemask
914 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
915 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
917 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
920 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
921 * @zone - The current zone in the iterator
922 * @z - The current pointer within zonelist->zones being iterated
923 * @zlist - The zonelist being iterated
924 * @highidx - The zone index of the highest zone to return
926 * This iterator iterates though all zones at or below a given zone index.
928 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
929 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
931 #ifdef CONFIG_SPARSEMEM
932 #include <asm/sparsemem.h>
935 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
936 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
937 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
943 #ifdef CONFIG_FLATMEM
944 #define pfn_to_nid(pfn) (0)
947 #ifdef CONFIG_SPARSEMEM
950 * SECTION_SHIFT #bits space required to store a section #
952 * PA_SECTION_SHIFT physical address to/from section number
953 * PFN_SECTION_SHIFT pfn to/from section number
955 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
957 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
958 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
960 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
962 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
963 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
965 #define SECTION_BLOCKFLAGS_BITS \
966 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
968 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
969 #error Allocator MAX_ORDER exceeds SECTION_SIZE
972 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
973 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
975 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
976 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
982 * This is, logically, a pointer to an array of struct
983 * pages. However, it is stored with some other magic.
984 * (see sparse.c::sparse_init_one_section())
986 * Additionally during early boot we encode node id of
987 * the location of the section here to guide allocation.
988 * (see sparse.c::memory_present())
990 * Making it a UL at least makes someone do a cast
991 * before using it wrong.
993 unsigned long section_mem_map;
995 /* See declaration of similar field in struct zone */
996 unsigned long *pageblock_flags;
997 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
999 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1000 * section. (see memcontrol.h/page_cgroup.h about this.)
1002 struct page_cgroup *page_cgroup;
1007 #ifdef CONFIG_SPARSEMEM_EXTREME
1008 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1010 #define SECTIONS_PER_ROOT 1
1013 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1014 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1015 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1017 #ifdef CONFIG_SPARSEMEM_EXTREME
1018 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1020 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1023 static inline struct mem_section *__nr_to_section(unsigned long nr)
1025 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1027 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1029 extern int __section_nr(struct mem_section* ms);
1030 extern unsigned long usemap_size(void);
1033 * We use the lower bits of the mem_map pointer to store
1034 * a little bit of information. There should be at least
1035 * 3 bits here due to 32-bit alignment.
1037 #define SECTION_MARKED_PRESENT (1UL<<0)
1038 #define SECTION_HAS_MEM_MAP (1UL<<1)
1039 #define SECTION_MAP_LAST_BIT (1UL<<2)
1040 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1041 #define SECTION_NID_SHIFT 2
1043 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1045 unsigned long map = section->section_mem_map;
1046 map &= SECTION_MAP_MASK;
1047 return (struct page *)map;
1050 static inline int present_section(struct mem_section *section)
1052 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1055 static inline int present_section_nr(unsigned long nr)
1057 return present_section(__nr_to_section(nr));
1060 static inline int valid_section(struct mem_section *section)
1062 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1065 static inline int valid_section_nr(unsigned long nr)
1067 return valid_section(__nr_to_section(nr));
1070 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1072 return __nr_to_section(pfn_to_section_nr(pfn));
1075 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1076 static inline int pfn_valid(unsigned long pfn)
1078 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1080 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1084 static inline int pfn_present(unsigned long pfn)
1086 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1088 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1092 * These are _only_ used during initialisation, therefore they
1093 * can use __initdata ... They could have names to indicate
1097 #define pfn_to_nid(pfn) \
1099 unsigned long __pfn_to_nid_pfn = (pfn); \
1100 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1103 #define pfn_to_nid(pfn) (0)
1106 #define early_pfn_valid(pfn) pfn_valid(pfn)
1107 void sparse_init(void);
1109 #define sparse_init() do {} while (0)
1110 #define sparse_index_init(_sec, _nid) do {} while (0)
1111 #endif /* CONFIG_SPARSEMEM */
1113 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1114 bool early_pfn_in_nid(unsigned long pfn, int nid);
1116 #define early_pfn_in_nid(pfn, nid) (1)
1119 #ifndef early_pfn_valid
1120 #define early_pfn_valid(pfn) (1)
1123 void memory_present(int nid, unsigned long start, unsigned long end);
1124 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1127 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1128 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1129 * pfn_valid_within() should be used in this case; we optimise this away
1130 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1132 #ifdef CONFIG_HOLES_IN_ZONE
1133 #define pfn_valid_within(pfn) pfn_valid(pfn)
1135 #define pfn_valid_within(pfn) (1)
1138 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1140 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1141 * associated with it or not. In FLATMEM, it is expected that holes always
1142 * have valid memmap as long as there is valid PFNs either side of the hole.
1143 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1146 * However, an ARM, and maybe other embedded architectures in the future
1147 * free memmap backing holes to save memory on the assumption the memmap is
1148 * never used. The page_zone linkages are then broken even though pfn_valid()
1149 * returns true. A walker of the full memmap must then do this additional
1150 * check to ensure the memmap they are looking at is sane by making sure
1151 * the zone and PFN linkages are still valid. This is expensive, but walkers
1152 * of the full memmap are extremely rare.
1154 int memmap_valid_within(unsigned long pfn,
1155 struct page *page, struct zone *zone);
1157 static inline int memmap_valid_within(unsigned long pfn,
1158 struct page *page, struct zone *zone)
1162 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1164 #endif /* !__GENERATING_BOUNDS.H */
1165 #endif /* !__ASSEMBLY__ */
1166 #endif /* _LINUX_MMZONE_H */