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
42 MIGRATE_PCPTYPES, /* the number of types on the pcp lists */
43 MIGRATE_RESERVE = MIGRATE_PCPTYPES,
46 * MIGRATE_CMA migration type is designed to mimic the way
47 * ZONE_MOVABLE works. Only movable pages can be allocated
48 * from MIGRATE_CMA pageblocks and page allocator never
49 * implicitly change migration type of MIGRATE_CMA pageblock.
51 * The way to use it is to change migratetype of a range of
52 * pageblocks to MIGRATE_CMA which can be done by
53 * __free_pageblock_cma() function. What is important though
54 * is that a range of pageblocks must be aligned to
55 * MAX_ORDER_NR_PAGES should biggest page be bigger then
60 MIGRATE_ISOLATE, /* can't allocate from here */
65 # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
66 # define cma_wmark_pages(zone) zone->min_cma_pages
68 # define is_migrate_cma(migratetype) false
69 # define cma_wmark_pages(zone) 0
72 #define for_each_migratetype_order(order, type) \
73 for (order = 0; order < MAX_ORDER; order++) \
74 for (type = 0; type < MIGRATE_TYPES; type++)
76 extern int page_group_by_mobility_disabled;
78 static inline int get_pageblock_migratetype(struct page *page)
80 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
84 struct list_head free_list[MIGRATE_TYPES];
85 unsigned long nr_free;
91 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
92 * So add a wild amount of padding here to ensure that they fall into separate
93 * cachelines. There are very few zone structures in the machine, so space
94 * consumption is not a concern here.
96 #if defined(CONFIG_SMP)
99 } ____cacheline_internodealigned_in_smp;
100 #define ZONE_PADDING(name) struct zone_padding name;
102 #define ZONE_PADDING(name)
105 enum zone_stat_item {
106 /* First 128 byte cacheline (assuming 64 bit words) */
109 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
110 NR_ACTIVE_ANON, /* " " " " " */
111 NR_INACTIVE_FILE, /* " " " " " */
112 NR_ACTIVE_FILE, /* " " " " " */
113 NR_UNEVICTABLE, /* " " " " " */
114 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
115 NR_ANON_PAGES, /* Mapped anonymous pages */
116 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
117 only modified from process context */
122 NR_SLAB_UNRECLAIMABLE,
123 NR_PAGETABLE, /* used for pagetables */
125 /* Second 128 byte cacheline */
126 NR_UNSTABLE_NFS, /* NFS unstable pages */
129 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
130 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
131 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
132 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
133 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
134 NR_DIRTIED, /* page dirtyings since bootup */
135 NR_WRITTEN, /* page writings since bootup */
137 NUMA_HIT, /* allocated in intended node */
138 NUMA_MISS, /* allocated in non intended node */
139 NUMA_FOREIGN, /* was intended here, hit elsewhere */
140 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
141 NUMA_LOCAL, /* allocation from local node */
142 NUMA_OTHER, /* allocation from other node */
144 NR_ANON_TRANSPARENT_HUGEPAGES,
145 NR_VM_ZONE_STAT_ITEMS };
148 * We do arithmetic on the LRU lists in various places in the code,
149 * so it is important to keep the active lists LRU_ACTIVE higher in
150 * the array than the corresponding inactive lists, and to keep
151 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
153 * This has to be kept in sync with the statistics in zone_stat_item
154 * above and the descriptions in vmstat_text in mm/vmstat.c
161 LRU_INACTIVE_ANON = LRU_BASE,
162 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
163 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
164 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
169 #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++)
171 #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++)
173 static inline int is_file_lru(enum lru_list l)
175 return (l == LRU_INACTIVE_FILE || l == LRU_ACTIVE_FILE);
178 static inline int is_active_lru(enum lru_list l)
180 return (l == LRU_ACTIVE_ANON || l == LRU_ACTIVE_FILE);
183 static inline int is_unevictable_lru(enum lru_list l)
185 return (l == LRU_UNEVICTABLE);
188 /* Mask used at gathering information at once (see memcontrol.c) */
189 #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
190 #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
191 #define LRU_ALL_EVICTABLE (LRU_ALL_FILE | LRU_ALL_ANON)
192 #define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
194 /* Isolate inactive pages */
195 #define ISOLATE_INACTIVE ((__force isolate_mode_t)0x1)
196 /* Isolate active pages */
197 #define ISOLATE_ACTIVE ((__force isolate_mode_t)0x2)
198 /* Isolate clean file */
199 #define ISOLATE_CLEAN ((__force isolate_mode_t)0x4)
200 /* Isolate unmapped file */
201 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x8)
202 /* Isolate for asynchronous migration */
203 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x10)
205 /* LRU Isolation modes. */
206 typedef unsigned __bitwise__ isolate_mode_t;
208 enum zone_watermarks {
215 #define min_wmark_pages(z) (z->watermark[WMARK_MIN])
216 #define low_wmark_pages(z) (z->watermark[WMARK_LOW])
217 #define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
219 struct per_cpu_pages {
220 int count; /* number of pages in the list */
221 int high; /* high watermark, emptying needed */
222 int batch; /* chunk size for buddy add/remove */
224 /* Lists of pages, one per migrate type stored on the pcp-lists */
225 struct list_head lists[MIGRATE_PCPTYPES];
228 struct per_cpu_pageset {
229 struct per_cpu_pages pcp;
235 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
239 #endif /* !__GENERATING_BOUNDS.H */
242 #ifdef CONFIG_ZONE_DMA
244 * ZONE_DMA is used when there are devices that are not able
245 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
246 * carve out the portion of memory that is needed for these devices.
247 * The range is arch specific.
252 * ---------------------------
253 * parisc, ia64, sparc <4G
256 * alpha Unlimited or 0-16MB.
258 * i386, x86_64 and multiple other arches
263 #ifdef CONFIG_ZONE_DMA32
265 * x86_64 needs two ZONE_DMAs because it supports devices that are
266 * only able to do DMA to the lower 16M but also 32 bit devices that
267 * can only do DMA areas below 4G.
272 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
273 * performed on pages in ZONE_NORMAL if the DMA devices support
274 * transfers to all addressable memory.
277 #ifdef CONFIG_HIGHMEM
279 * A memory area that is only addressable by the kernel through
280 * mapping portions into its own address space. This is for example
281 * used by i386 to allow the kernel to address the memory beyond
282 * 900MB. The kernel will set up special mappings (page
283 * table entries on i386) for each page that the kernel needs to
292 #ifndef __GENERATING_BOUNDS_H
295 * When a memory allocation must conform to specific limitations (such
296 * as being suitable for DMA) the caller will pass in hints to the
297 * allocator in the gfp_mask, in the zone modifier bits. These bits
298 * are used to select a priority ordered list of memory zones which
299 * match the requested limits. See gfp_zone() in include/linux/gfp.h
303 #define ZONES_SHIFT 0
304 #elif MAX_NR_ZONES <= 2
305 #define ZONES_SHIFT 1
306 #elif MAX_NR_ZONES <= 4
307 #define ZONES_SHIFT 2
309 #error ZONES_SHIFT -- too many zones configured adjust calculation
312 struct zone_reclaim_stat {
314 * The pageout code in vmscan.c keeps track of how many of the
315 * mem/swap backed and file backed pages are refeferenced.
316 * The higher the rotated/scanned ratio, the more valuable
319 * The anon LRU stats live in [0], file LRU stats in [1]
321 unsigned long recent_rotated[2];
322 unsigned long recent_scanned[2];
326 /* Fields commonly accessed by the page allocator */
328 /* zone watermarks, access with *_wmark_pages(zone) macros */
329 unsigned long watermark[NR_WMARK];
332 * When free pages are below this point, additional steps are taken
333 * when reading the number of free pages to avoid per-cpu counter
334 * drift allowing watermarks to be breached
336 unsigned long percpu_drift_mark;
339 * We don't know if the memory that we're going to allocate will be freeable
340 * or/and it will be released eventually, so to avoid totally wasting several
341 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
342 * to run OOM on the lower zones despite there's tons of freeable ram
343 * on the higher zones). This array is recalculated at runtime if the
344 * sysctl_lowmem_reserve_ratio sysctl changes.
346 unsigned long lowmem_reserve[MAX_NR_ZONES];
349 * This is a per-zone reserve of pages that should not be
350 * considered dirtyable memory.
352 unsigned long dirty_balance_reserve;
357 * zone reclaim becomes active if more unmapped pages exist.
359 unsigned long min_unmapped_pages;
360 unsigned long min_slab_pages;
362 struct per_cpu_pageset __percpu *pageset;
364 * free areas of different sizes
367 int all_unreclaimable; /* All pages pinned */
368 #ifdef CONFIG_MEMORY_HOTPLUG
369 /* see spanned/present_pages for more description */
370 seqlock_t span_seqlock;
374 * CMA needs to increase watermark levels during the allocation
375 * process to make sure that the system is not starved.
377 unsigned long min_cma_pages;
379 struct free_area free_area[MAX_ORDER];
381 #ifndef CONFIG_SPARSEMEM
383 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
384 * In SPARSEMEM, this map is stored in struct mem_section
386 unsigned long *pageblock_flags;
387 #endif /* CONFIG_SPARSEMEM */
389 #ifdef CONFIG_COMPACTION
391 * On compaction failure, 1<<compact_defer_shift compactions
392 * are skipped before trying again. The number attempted since
393 * last failure is tracked with compact_considered.
395 unsigned int compact_considered;
396 unsigned int compact_defer_shift;
401 /* Fields commonly accessed by the page reclaim scanner */
404 struct list_head list;
407 struct zone_reclaim_stat reclaim_stat;
409 unsigned long pages_scanned; /* since last reclaim */
410 unsigned long flags; /* zone flags, see below */
412 /* Zone statistics */
413 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
416 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
417 * this zone's LRU. Maintained by the pageout code.
419 unsigned int inactive_ratio;
423 /* Rarely used or read-mostly fields */
426 * wait_table -- the array holding the hash table
427 * wait_table_hash_nr_entries -- the size of the hash table array
428 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
430 * The purpose of all these is to keep track of the people
431 * waiting for a page to become available and make them
432 * runnable again when possible. The trouble is that this
433 * consumes a lot of space, especially when so few things
434 * wait on pages at a given time. So instead of using
435 * per-page waitqueues, we use a waitqueue hash table.
437 * The bucket discipline is to sleep on the same queue when
438 * colliding and wake all in that wait queue when removing.
439 * When something wakes, it must check to be sure its page is
440 * truly available, a la thundering herd. The cost of a
441 * collision is great, but given the expected load of the
442 * table, they should be so rare as to be outweighed by the
443 * benefits from the saved space.
445 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
446 * primary users of these fields, and in mm/page_alloc.c
447 * free_area_init_core() performs the initialization of them.
449 wait_queue_head_t * wait_table;
450 unsigned long wait_table_hash_nr_entries;
451 unsigned long wait_table_bits;
454 * Discontig memory support fields.
456 struct pglist_data *zone_pgdat;
457 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
458 unsigned long zone_start_pfn;
461 * zone_start_pfn, spanned_pages and present_pages are all
462 * protected by span_seqlock. It is a seqlock because it has
463 * to be read outside of zone->lock, and it is done in the main
464 * allocator path. But, it is written quite infrequently.
466 * The lock is declared along with zone->lock because it is
467 * frequently read in proximity to zone->lock. It's good to
468 * give them a chance of being in the same cacheline.
470 unsigned long spanned_pages; /* total size, including holes */
471 unsigned long present_pages; /* amount of memory (excluding holes) */
474 * rarely used fields:
477 } ____cacheline_internodealigned_in_smp;
480 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
481 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
482 ZONE_CONGESTED, /* zone has many dirty pages backed by
487 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
489 set_bit(flag, &zone->flags);
492 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
494 return test_and_set_bit(flag, &zone->flags);
497 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
499 clear_bit(flag, &zone->flags);
502 static inline int zone_is_reclaim_congested(const struct zone *zone)
504 return test_bit(ZONE_CONGESTED, &zone->flags);
507 static inline int zone_is_reclaim_locked(const struct zone *zone)
509 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
512 static inline int zone_is_oom_locked(const struct zone *zone)
514 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
518 * The "priority" of VM scanning is how much of the queues we will scan in one
519 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
520 * queues ("queue_length >> 12") during an aging round.
522 #define DEF_PRIORITY 12
524 /* Maximum number of zones on a zonelist */
525 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
530 * The NUMA zonelists are doubled because we need zonelists that restrict the
531 * allocations to a single node for GFP_THISNODE.
533 * [0] : Zonelist with fallback
534 * [1] : No fallback (GFP_THISNODE)
536 #define MAX_ZONELISTS 2
540 * We cache key information from each zonelist for smaller cache
541 * footprint when scanning for free pages in get_page_from_freelist().
543 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
544 * up short of free memory since the last time (last_fullzone_zap)
545 * we zero'd fullzones.
546 * 2) The array z_to_n[] maps each zone in the zonelist to its node
547 * id, so that we can efficiently evaluate whether that node is
548 * set in the current tasks mems_allowed.
550 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
551 * indexed by a zones offset in the zonelist zones[] array.
553 * The get_page_from_freelist() routine does two scans. During the
554 * first scan, we skip zones whose corresponding bit in 'fullzones'
555 * is set or whose corresponding node in current->mems_allowed (which
556 * comes from cpusets) is not set. During the second scan, we bypass
557 * this zonelist_cache, to ensure we look methodically at each zone.
559 * Once per second, we zero out (zap) fullzones, forcing us to
560 * reconsider nodes that might have regained more free memory.
561 * The field last_full_zap is the time we last zapped fullzones.
563 * This mechanism reduces the amount of time we waste repeatedly
564 * reexaming zones for free memory when they just came up low on
565 * memory momentarilly ago.
567 * The zonelist_cache struct members logically belong in struct
568 * zonelist. However, the mempolicy zonelists constructed for
569 * MPOL_BIND are intentionally variable length (and usually much
570 * shorter). A general purpose mechanism for handling structs with
571 * multiple variable length members is more mechanism than we want
572 * here. We resort to some special case hackery instead.
574 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
575 * part because they are shorter), so we put the fixed length stuff
576 * at the front of the zonelist struct, ending in a variable length
577 * zones[], as is needed by MPOL_BIND.
579 * Then we put the optional zonelist cache on the end of the zonelist
580 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
581 * the fixed length portion at the front of the struct. This pointer
582 * both enables us to find the zonelist cache, and in the case of
583 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
584 * to know that the zonelist cache is not there.
586 * The end result is that struct zonelists come in two flavors:
587 * 1) The full, fixed length version, shown below, and
588 * 2) The custom zonelists for MPOL_BIND.
589 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
591 * Even though there may be multiple CPU cores on a node modifying
592 * fullzones or last_full_zap in the same zonelist_cache at the same
593 * time, we don't lock it. This is just hint data - if it is wrong now
594 * and then, the allocator will still function, perhaps a bit slower.
598 struct zonelist_cache {
599 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
600 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
601 unsigned long last_full_zap; /* when last zap'd (jiffies) */
604 #define MAX_ZONELISTS 1
605 struct zonelist_cache;
609 * This struct contains information about a zone in a zonelist. It is stored
610 * here to avoid dereferences into large structures and lookups of tables
613 struct zone *zone; /* Pointer to actual zone */
614 int zone_idx; /* zone_idx(zoneref->zone) */
618 * One allocation request operates on a zonelist. A zonelist
619 * is a list of zones, the first one is the 'goal' of the
620 * allocation, the other zones are fallback zones, in decreasing
623 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
624 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
626 * To speed the reading of the zonelist, the zonerefs contain the zone index
627 * of the entry being read. Helper functions to access information given
628 * a struct zoneref are
630 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
631 * zonelist_zone_idx() - Return the index of the zone for an entry
632 * zonelist_node_idx() - Return the index of the node for an entry
635 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
636 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
638 struct zonelist_cache zlcache; // optional ...
642 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
643 struct node_active_region {
644 unsigned long start_pfn;
645 unsigned long end_pfn;
648 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
650 #ifndef CONFIG_DISCONTIGMEM
651 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
652 extern struct page *mem_map;
656 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
657 * (mostly NUMA machines?) to denote a higher-level memory zone than the
660 * On NUMA machines, each NUMA node would have a pg_data_t to describe
661 * it's memory layout.
663 * Memory statistics and page replacement data structures are maintained on a
667 typedef struct pglist_data {
668 struct zone node_zones[MAX_NR_ZONES];
669 struct zonelist node_zonelists[MAX_ZONELISTS];
671 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
672 struct page *node_mem_map;
673 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
674 struct page_cgroup *node_page_cgroup;
677 #ifndef CONFIG_NO_BOOTMEM
678 struct bootmem_data *bdata;
680 #ifdef CONFIG_MEMORY_HOTPLUG
682 * Must be held any time you expect node_start_pfn, node_present_pages
683 * or node_spanned_pages stay constant. Holding this will also
684 * guarantee that any pfn_valid() stays that way.
686 * Nests above zone->lock and zone->size_seqlock.
688 spinlock_t node_size_lock;
690 unsigned long node_start_pfn;
691 unsigned long node_present_pages; /* total number of physical pages */
692 unsigned long node_spanned_pages; /* total size of physical page
693 range, including holes */
695 wait_queue_head_t kswapd_wait;
696 struct task_struct *kswapd; /* Protected by lock_memory_hotplug() */
697 int kswapd_max_order;
698 enum zone_type classzone_idx;
701 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
702 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
703 #ifdef CONFIG_FLAT_NODE_MEM_MAP
704 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
706 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
708 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
710 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
712 #define node_end_pfn(nid) ({\
713 pg_data_t *__pgdat = NODE_DATA(nid);\
714 __pgdat->node_start_pfn + __pgdat->node_spanned_pages;\
717 #include <linux/memory_hotplug.h>
719 extern struct mutex zonelists_mutex;
720 void build_all_zonelists(void *data);
721 void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
722 bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
723 int classzone_idx, int alloc_flags);
724 bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
725 int classzone_idx, int alloc_flags);
726 enum memmap_context {
730 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
732 enum memmap_context context);
734 #ifdef CONFIG_HAVE_MEMORY_PRESENT
735 void memory_present(int nid, unsigned long start, unsigned long end);
737 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
740 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
741 int local_memory_node(int node_id);
743 static inline int local_memory_node(int node_id) { return node_id; };
746 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
747 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
751 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
753 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
755 static inline int populated_zone(struct zone *zone)
757 return (!!zone->present_pages);
760 extern int movable_zone;
762 static inline int zone_movable_is_highmem(void)
764 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
765 return movable_zone == ZONE_HIGHMEM;
771 static inline int is_highmem_idx(enum zone_type idx)
773 #ifdef CONFIG_HIGHMEM
774 return (idx == ZONE_HIGHMEM ||
775 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
781 static inline int is_normal_idx(enum zone_type idx)
783 return (idx == ZONE_NORMAL);
787 * is_highmem - helper function to quickly check if a struct zone is a
788 * highmem zone or not. This is an attempt to keep references
789 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
790 * @zone - pointer to struct zone variable
792 static inline int is_highmem(struct zone *zone)
794 #ifdef CONFIG_HIGHMEM
795 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
796 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
797 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
798 zone_movable_is_highmem());
804 static inline int is_normal(struct zone *zone)
806 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
809 static inline int is_dma32(struct zone *zone)
811 #ifdef CONFIG_ZONE_DMA32
812 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
818 static inline int is_dma(struct zone *zone)
820 #ifdef CONFIG_ZONE_DMA
821 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
827 /* These two functions are used to setup the per zone pages min values */
829 int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
830 void __user *, size_t *, loff_t *);
831 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
832 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
833 void __user *, size_t *, loff_t *);
834 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
835 void __user *, size_t *, loff_t *);
836 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
837 void __user *, size_t *, loff_t *);
838 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
839 void __user *, size_t *, loff_t *);
841 extern int numa_zonelist_order_handler(struct ctl_table *, int,
842 void __user *, size_t *, loff_t *);
843 extern char numa_zonelist_order[];
844 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
846 #ifndef CONFIG_NEED_MULTIPLE_NODES
848 extern struct pglist_data contig_page_data;
849 #define NODE_DATA(nid) (&contig_page_data)
850 #define NODE_MEM_MAP(nid) mem_map
852 #else /* CONFIG_NEED_MULTIPLE_NODES */
854 #include <asm/mmzone.h>
856 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
858 extern struct pglist_data *first_online_pgdat(void);
859 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
860 extern struct zone *next_zone(struct zone *zone);
863 * for_each_online_pgdat - helper macro to iterate over all online nodes
864 * @pgdat - pointer to a pg_data_t variable
866 #define for_each_online_pgdat(pgdat) \
867 for (pgdat = first_online_pgdat(); \
869 pgdat = next_online_pgdat(pgdat))
871 * for_each_zone - helper macro to iterate over all memory zones
872 * @zone - pointer to struct zone variable
874 * The user only needs to declare the zone variable, for_each_zone
877 #define for_each_zone(zone) \
878 for (zone = (first_online_pgdat())->node_zones; \
880 zone = next_zone(zone))
882 #define for_each_populated_zone(zone) \
883 for (zone = (first_online_pgdat())->node_zones; \
885 zone = next_zone(zone)) \
886 if (!populated_zone(zone)) \
890 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
892 return zoneref->zone;
895 static inline int zonelist_zone_idx(struct zoneref *zoneref)
897 return zoneref->zone_idx;
900 static inline int zonelist_node_idx(struct zoneref *zoneref)
903 /* zone_to_nid not available in this context */
904 return zoneref->zone->node;
907 #endif /* CONFIG_NUMA */
911 * 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
912 * @z - The cursor used as a starting point for the search
913 * @highest_zoneidx - The zone index of the highest zone to return
914 * @nodes - An optional nodemask to filter the zonelist with
915 * @zone - The first suitable zone found is returned via this parameter
917 * This function returns the next zone at or below a given zone index that is
918 * within the allowed nodemask using a cursor as the starting point for the
919 * search. The zoneref returned is a cursor that represents the current zone
920 * being examined. It should be advanced by one before calling
921 * next_zones_zonelist again.
923 struct zoneref *next_zones_zonelist(struct zoneref *z,
924 enum zone_type highest_zoneidx,
929 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
930 * @zonelist - The zonelist to search for a suitable zone
931 * @highest_zoneidx - The zone index of the highest zone to return
932 * @nodes - An optional nodemask to filter the zonelist with
933 * @zone - The first suitable zone found is returned via this parameter
935 * This function returns the first zone at or below a given zone index that is
936 * within the allowed nodemask. The zoneref returned is a cursor that can be
937 * used to iterate the zonelist with next_zones_zonelist by advancing it by
938 * one before calling.
940 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
941 enum zone_type highest_zoneidx,
945 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
950 * 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
951 * @zone - The current zone in the iterator
952 * @z - The current pointer within zonelist->zones being iterated
953 * @zlist - The zonelist being iterated
954 * @highidx - The zone index of the highest zone to return
955 * @nodemask - Nodemask allowed by the allocator
957 * This iterator iterates though all zones at or below a given zone index and
958 * within a given nodemask
960 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
961 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
963 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
966 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
967 * @zone - The current zone in the iterator
968 * @z - The current pointer within zonelist->zones being iterated
969 * @zlist - The zonelist being iterated
970 * @highidx - The zone index of the highest zone to return
972 * This iterator iterates though all zones at or below a given zone index.
974 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
975 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
977 #ifdef CONFIG_SPARSEMEM
978 #include <asm/sparsemem.h>
981 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
982 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
983 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
989 #ifdef CONFIG_FLATMEM
990 #define pfn_to_nid(pfn) (0)
993 #ifdef CONFIG_SPARSEMEM
996 * SECTION_SHIFT #bits space required to store a section #
998 * PA_SECTION_SHIFT physical address to/from section number
999 * PFN_SECTION_SHIFT pfn to/from section number
1001 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
1003 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1004 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1006 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1008 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1009 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1011 #define SECTION_BLOCKFLAGS_BITS \
1012 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1014 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1015 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1018 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1019 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1021 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1022 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1026 struct mem_section {
1028 * This is, logically, a pointer to an array of struct
1029 * pages. However, it is stored with some other magic.
1030 * (see sparse.c::sparse_init_one_section())
1032 * Additionally during early boot we encode node id of
1033 * the location of the section here to guide allocation.
1034 * (see sparse.c::memory_present())
1036 * Making it a UL at least makes someone do a cast
1037 * before using it wrong.
1039 unsigned long section_mem_map;
1041 /* See declaration of similar field in struct zone */
1042 unsigned long *pageblock_flags;
1043 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1045 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1046 * section. (see memcontrol.h/page_cgroup.h about this.)
1048 struct page_cgroup *page_cgroup;
1053 #ifdef CONFIG_SPARSEMEM_EXTREME
1054 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1056 #define SECTIONS_PER_ROOT 1
1059 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1060 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1061 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1063 #ifdef CONFIG_SPARSEMEM_EXTREME
1064 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1066 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1069 static inline struct mem_section *__nr_to_section(unsigned long nr)
1071 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1073 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1075 extern int __section_nr(struct mem_section* ms);
1076 extern unsigned long usemap_size(void);
1079 * We use the lower bits of the mem_map pointer to store
1080 * a little bit of information. There should be at least
1081 * 3 bits here due to 32-bit alignment.
1083 #define SECTION_MARKED_PRESENT (1UL<<0)
1084 #define SECTION_HAS_MEM_MAP (1UL<<1)
1085 #define SECTION_MAP_LAST_BIT (1UL<<2)
1086 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1087 #define SECTION_NID_SHIFT 2
1089 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1091 unsigned long map = section->section_mem_map;
1092 map &= SECTION_MAP_MASK;
1093 return (struct page *)map;
1096 static inline int present_section(struct mem_section *section)
1098 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1101 static inline int present_section_nr(unsigned long nr)
1103 return present_section(__nr_to_section(nr));
1106 static inline int valid_section(struct mem_section *section)
1108 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1111 static inline int valid_section_nr(unsigned long nr)
1113 return valid_section(__nr_to_section(nr));
1116 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1118 return __nr_to_section(pfn_to_section_nr(pfn));
1121 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1122 static inline int pfn_valid(unsigned long pfn)
1124 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1126 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
1130 static inline int pfn_present(unsigned long pfn)
1132 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1134 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1138 * These are _only_ used during initialisation, therefore they
1139 * can use __initdata ... They could have names to indicate
1143 #define pfn_to_nid(pfn) \
1145 unsigned long __pfn_to_nid_pfn = (pfn); \
1146 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1149 #define pfn_to_nid(pfn) (0)
1152 #define early_pfn_valid(pfn) pfn_valid(pfn)
1153 void sparse_init(void);
1155 #define sparse_init() do {} while (0)
1156 #define sparse_index_init(_sec, _nid) do {} while (0)
1157 #endif /* CONFIG_SPARSEMEM */
1159 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
1160 bool early_pfn_in_nid(unsigned long pfn, int nid);
1162 #define early_pfn_in_nid(pfn, nid) (1)
1165 #ifndef early_pfn_valid
1166 #define early_pfn_valid(pfn) (1)
1169 void memory_present(int nid, unsigned long start, unsigned long end);
1170 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1173 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1174 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1175 * pfn_valid_within() should be used in this case; we optimise this away
1176 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1178 #ifdef CONFIG_HOLES_IN_ZONE
1179 #define pfn_valid_within(pfn) pfn_valid(pfn)
1181 #define pfn_valid_within(pfn) (1)
1184 #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1186 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1187 * associated with it or not. In FLATMEM, it is expected that holes always
1188 * have valid memmap as long as there is valid PFNs either side of the hole.
1189 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1192 * However, an ARM, and maybe other embedded architectures in the future
1193 * free memmap backing holes to save memory on the assumption the memmap is
1194 * never used. The page_zone linkages are then broken even though pfn_valid()
1195 * returns true. A walker of the full memmap must then do this additional
1196 * check to ensure the memmap they are looking at is sane by making sure
1197 * the zone and PFN linkages are still valid. This is expensive, but walkers
1198 * of the full memmap are extremely rare.
1200 int memmap_valid_within(unsigned long pfn,
1201 struct page *page, struct zone *zone);
1203 static inline int memmap_valid_within(unsigned long pfn,
1204 struct page *page, struct zone *zone)
1208 #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1210 #endif /* !__GENERATING_BOUNDS.H */
1211 #endif /* !__ASSEMBLY__ */
1212 #endif /* _LINUX_MMZONE_H */