1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
6 #ifndef __GENERATING_BOUNDS_H
8 #include <linux/spinlock.h>
9 #include <linux/list.h>
10 #include <linux/wait.h>
11 #include <linux/bitops.h>
12 #include <linux/cache.h>
13 #include <linux/threads.h>
14 #include <linux/numa.h>
15 #include <linux/init.h>
16 #include <linux/seqlock.h>
17 #include <linux/nodemask.h>
18 #include <linux/pageblock-flags.h>
19 #include <linux/bounds.h>
20 #include <asm/atomic.h>
23 /* Free memory management - zoned buddy allocator. */
24 #ifndef CONFIG_FORCE_MAX_ZONEORDER
27 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
29 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
32 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
33 * costly to service. That is between allocation orders which should
34 * coelesce naturally under reasonable reclaim pressure and those which
37 #define PAGE_ALLOC_COSTLY_ORDER 3
39 #define MIGRATE_UNMOVABLE 0
40 #define MIGRATE_RECLAIMABLE 1
41 #define MIGRATE_MOVABLE 2
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 if (unlikely(page_group_by_mobility_disabled))
55 return MIGRATE_UNMOVABLE;
57 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
61 struct list_head free_list[MIGRATE_TYPES];
62 unsigned long nr_free;
68 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
69 * So add a wild amount of padding here to ensure that they fall into separate
70 * cachelines. There are very few zone structures in the machine, so space
71 * consumption is not a concern here.
73 #if defined(CONFIG_SMP)
76 } ____cacheline_internodealigned_in_smp;
77 #define ZONE_PADDING(name) struct zone_padding name;
79 #define ZONE_PADDING(name)
83 /* First 128 byte cacheline (assuming 64 bit words) */
87 NR_ANON_PAGES, /* Mapped anonymous pages */
88 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
89 only modified from process context */
93 /* Second 128 byte cacheline */
95 NR_SLAB_UNRECLAIMABLE,
96 NR_PAGETABLE, /* used for pagetables */
97 NR_UNSTABLE_NFS, /* NFS unstable pages */
100 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
102 NUMA_HIT, /* allocated in intended node */
103 NUMA_MISS, /* allocated in non intended node */
104 NUMA_FOREIGN, /* was intended here, hit elsewhere */
105 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
106 NUMA_LOCAL, /* allocation from local node */
107 NUMA_OTHER, /* allocation from other node */
109 NR_VM_ZONE_STAT_ITEMS };
111 struct per_cpu_pages {
112 int count; /* number of pages in the list */
113 int high; /* high watermark, emptying needed */
114 int batch; /* chunk size for buddy add/remove */
115 struct list_head list; /* the list of pages */
118 struct per_cpu_pageset {
119 struct per_cpu_pages pcp;
125 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
127 } ____cacheline_aligned_in_smp;
130 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
132 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
135 #endif /* !__GENERATING_BOUNDS.H */
138 #ifdef CONFIG_ZONE_DMA
140 * ZONE_DMA is used when there are devices that are not able
141 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
142 * carve out the portion of memory that is needed for these devices.
143 * The range is arch specific.
148 * ---------------------------
149 * parisc, ia64, sparc <4G
152 * alpha Unlimited or 0-16MB.
154 * i386, x86_64 and multiple other arches
159 #ifdef CONFIG_ZONE_DMA32
161 * x86_64 needs two ZONE_DMAs because it supports devices that are
162 * only able to do DMA to the lower 16M but also 32 bit devices that
163 * can only do DMA areas below 4G.
168 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
169 * performed on pages in ZONE_NORMAL if the DMA devices support
170 * transfers to all addressable memory.
173 #ifdef CONFIG_HIGHMEM
175 * A memory area that is only addressable by the kernel through
176 * mapping portions into its own address space. This is for example
177 * used by i386 to allow the kernel to address the memory beyond
178 * 900MB. The kernel will set up special mappings (page
179 * table entries on i386) for each page that the kernel needs to
188 #ifndef __GENERATING_BOUNDS_H
191 * When a memory allocation must conform to specific limitations (such
192 * as being suitable for DMA) the caller will pass in hints to the
193 * allocator in the gfp_mask, in the zone modifier bits. These bits
194 * are used to select a priority ordered list of memory zones which
195 * match the requested limits. See gfp_zone() in include/linux/gfp.h
199 #define ZONES_SHIFT 0
200 #elif MAX_NR_ZONES <= 2
201 #define ZONES_SHIFT 1
202 #elif MAX_NR_ZONES <= 4
203 #define ZONES_SHIFT 2
205 #error ZONES_SHIFT -- too many zones configured adjust calculation
209 /* Fields commonly accessed by the page allocator */
210 unsigned long pages_min, pages_low, pages_high;
212 * We don't know if the memory that we're going to allocate will be freeable
213 * or/and it will be released eventually, so to avoid totally wasting several
214 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
215 * to run OOM on the lower zones despite there's tons of freeable ram
216 * on the higher zones). This array is recalculated at runtime if the
217 * sysctl_lowmem_reserve_ratio sysctl changes.
219 unsigned long lowmem_reserve[MAX_NR_ZONES];
224 * zone reclaim becomes active if more unmapped pages exist.
226 unsigned long min_unmapped_pages;
227 unsigned long min_slab_pages;
228 struct per_cpu_pageset *pageset[NR_CPUS];
230 struct per_cpu_pageset pageset[NR_CPUS];
233 * free areas of different sizes
236 #ifdef CONFIG_MEMORY_HOTPLUG
237 /* see spanned/present_pages for more description */
238 seqlock_t span_seqlock;
240 struct free_area free_area[MAX_ORDER];
242 #ifndef CONFIG_SPARSEMEM
244 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
245 * In SPARSEMEM, this map is stored in struct mem_section
247 unsigned long *pageblock_flags;
248 #endif /* CONFIG_SPARSEMEM */
253 /* Fields commonly accessed by the page reclaim scanner */
255 struct list_head active_list;
256 struct list_head inactive_list;
257 unsigned long nr_scan_active;
258 unsigned long nr_scan_inactive;
259 unsigned long pages_scanned; /* since last reclaim */
260 unsigned long flags; /* zone flags, see below */
262 /* Zone statistics */
263 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
266 * prev_priority holds the scanning priority for this zone. It is
267 * defined as the scanning priority at which we achieved our reclaim
268 * target at the previous try_to_free_pages() or balance_pgdat()
271 * We use prev_priority as a measure of how much stress page reclaim is
272 * under - it drives the swappiness decision: whether to unmap mapped
275 * Access to both this field is quite racy even on uniprocessor. But
276 * it is expected to average out OK.
282 /* Rarely used or read-mostly fields */
285 * wait_table -- the array holding the hash table
286 * wait_table_hash_nr_entries -- the size of the hash table array
287 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
289 * The purpose of all these is to keep track of the people
290 * waiting for a page to become available and make them
291 * runnable again when possible. The trouble is that this
292 * consumes a lot of space, especially when so few things
293 * wait on pages at a given time. So instead of using
294 * per-page waitqueues, we use a waitqueue hash table.
296 * The bucket discipline is to sleep on the same queue when
297 * colliding and wake all in that wait queue when removing.
298 * When something wakes, it must check to be sure its page is
299 * truly available, a la thundering herd. The cost of a
300 * collision is great, but given the expected load of the
301 * table, they should be so rare as to be outweighed by the
302 * benefits from the saved space.
304 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
305 * primary users of these fields, and in mm/page_alloc.c
306 * free_area_init_core() performs the initialization of them.
308 wait_queue_head_t * wait_table;
309 unsigned long wait_table_hash_nr_entries;
310 unsigned long wait_table_bits;
313 * Discontig memory support fields.
315 struct pglist_data *zone_pgdat;
316 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
317 unsigned long zone_start_pfn;
320 * zone_start_pfn, spanned_pages and present_pages are all
321 * protected by span_seqlock. It is a seqlock because it has
322 * to be read outside of zone->lock, and it is done in the main
323 * allocator path. But, it is written quite infrequently.
325 * The lock is declared along with zone->lock because it is
326 * frequently read in proximity to zone->lock. It's good to
327 * give them a chance of being in the same cacheline.
329 unsigned long spanned_pages; /* total size, including holes */
330 unsigned long present_pages; /* amount of memory (excluding holes) */
333 * rarely used fields:
336 } ____cacheline_internodealigned_in_smp;
339 ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */
340 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
341 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
344 static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
346 set_bit(flag, &zone->flags);
349 static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
351 return test_and_set_bit(flag, &zone->flags);
354 static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
356 clear_bit(flag, &zone->flags);
359 static inline int zone_is_all_unreclaimable(const struct zone *zone)
361 return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags);
364 static inline int zone_is_reclaim_locked(const struct zone *zone)
366 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
369 static inline int zone_is_oom_locked(const struct zone *zone)
371 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
375 * The "priority" of VM scanning is how much of the queues we will scan in one
376 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
377 * queues ("queue_length >> 12") during an aging round.
379 #define DEF_PRIORITY 12
381 /* Maximum number of zones on a zonelist */
382 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
387 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
388 * allocations to a single node for GFP_THISNODE.
390 * [0] : Zonelist with fallback
391 * [1] : No fallback (GFP_THISNODE)
393 #define MAX_ZONELISTS 2
397 * We cache key information from each zonelist for smaller cache
398 * footprint when scanning for free pages in get_page_from_freelist().
400 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
401 * up short of free memory since the last time (last_fullzone_zap)
402 * we zero'd fullzones.
403 * 2) The array z_to_n[] maps each zone in the zonelist to its node
404 * id, so that we can efficiently evaluate whether that node is
405 * set in the current tasks mems_allowed.
407 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
408 * indexed by a zones offset in the zonelist zones[] array.
410 * The get_page_from_freelist() routine does two scans. During the
411 * first scan, we skip zones whose corresponding bit in 'fullzones'
412 * is set or whose corresponding node in current->mems_allowed (which
413 * comes from cpusets) is not set. During the second scan, we bypass
414 * this zonelist_cache, to ensure we look methodically at each zone.
416 * Once per second, we zero out (zap) fullzones, forcing us to
417 * reconsider nodes that might have regained more free memory.
418 * The field last_full_zap is the time we last zapped fullzones.
420 * This mechanism reduces the amount of time we waste repeatedly
421 * reexaming zones for free memory when they just came up low on
422 * memory momentarilly ago.
424 * The zonelist_cache struct members logically belong in struct
425 * zonelist. However, the mempolicy zonelists constructed for
426 * MPOL_BIND are intentionally variable length (and usually much
427 * shorter). A general purpose mechanism for handling structs with
428 * multiple variable length members is more mechanism than we want
429 * here. We resort to some special case hackery instead.
431 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
432 * part because they are shorter), so we put the fixed length stuff
433 * at the front of the zonelist struct, ending in a variable length
434 * zones[], as is needed by MPOL_BIND.
436 * Then we put the optional zonelist cache on the end of the zonelist
437 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
438 * the fixed length portion at the front of the struct. This pointer
439 * both enables us to find the zonelist cache, and in the case of
440 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
441 * to know that the zonelist cache is not there.
443 * The end result is that struct zonelists come in two flavors:
444 * 1) The full, fixed length version, shown below, and
445 * 2) The custom zonelists for MPOL_BIND.
446 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
448 * Even though there may be multiple CPU cores on a node modifying
449 * fullzones or last_full_zap in the same zonelist_cache at the same
450 * time, we don't lock it. This is just hint data - if it is wrong now
451 * and then, the allocator will still function, perhaps a bit slower.
455 struct zonelist_cache {
456 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
457 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
458 unsigned long last_full_zap; /* when last zap'd (jiffies) */
461 #define MAX_ZONELISTS 1
462 struct zonelist_cache;
466 * This struct contains information about a zone in a zonelist. It is stored
467 * here to avoid dereferences into large structures and lookups of tables
470 struct zone *zone; /* Pointer to actual zone */
471 int zone_idx; /* zone_idx(zoneref->zone) */
475 * One allocation request operates on a zonelist. A zonelist
476 * is a list of zones, the first one is the 'goal' of the
477 * allocation, the other zones are fallback zones, in decreasing
480 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
481 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
483 * To speed the reading of the zonelist, the zonerefs contain the zone index
484 * of the entry being read. Helper functions to access information given
485 * a struct zoneref are
487 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
488 * zonelist_zone_idx() - Return the index of the zone for an entry
489 * zonelist_node_idx() - Return the index of the node for an entry
492 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
493 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
495 struct zonelist_cache zlcache; // optional ...
499 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
500 struct node_active_region {
501 unsigned long start_pfn;
502 unsigned long end_pfn;
505 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
507 #ifndef CONFIG_DISCONTIGMEM
508 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
509 extern struct page *mem_map;
513 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
514 * (mostly NUMA machines?) to denote a higher-level memory zone than the
517 * On NUMA machines, each NUMA node would have a pg_data_t to describe
518 * it's memory layout.
520 * Memory statistics and page replacement data structures are maintained on a
524 typedef struct pglist_data {
525 struct zone node_zones[MAX_NR_ZONES];
526 struct zonelist node_zonelists[MAX_ZONELISTS];
528 #ifdef CONFIG_FLAT_NODE_MEM_MAP
529 struct page *node_mem_map;
531 struct bootmem_data *bdata;
532 #ifdef CONFIG_MEMORY_HOTPLUG
534 * Must be held any time you expect node_start_pfn, node_present_pages
535 * or node_spanned_pages stay constant. Holding this will also
536 * guarantee that any pfn_valid() stays that way.
538 * Nests above zone->lock and zone->size_seqlock.
540 spinlock_t node_size_lock;
542 unsigned long node_start_pfn;
543 unsigned long node_present_pages; /* total number of physical pages */
544 unsigned long node_spanned_pages; /* total size of physical page
545 range, including holes */
547 wait_queue_head_t kswapd_wait;
548 struct task_struct *kswapd;
549 int kswapd_max_order;
552 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
553 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
554 #ifdef CONFIG_FLAT_NODE_MEM_MAP
555 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
557 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
559 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
561 #include <linux/memory_hotplug.h>
563 void get_zone_counts(unsigned long *active, unsigned long *inactive,
564 unsigned long *free);
565 void build_all_zonelists(void);
566 void wakeup_kswapd(struct zone *zone, int order);
567 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
568 int classzone_idx, int alloc_flags);
569 enum memmap_context {
573 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
575 enum memmap_context context);
577 #ifdef CONFIG_HAVE_MEMORY_PRESENT
578 void memory_present(int nid, unsigned long start, unsigned long end);
580 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
583 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
584 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
588 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
590 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
592 static inline int populated_zone(struct zone *zone)
594 return (!!zone->present_pages);
597 extern int movable_zone;
599 static inline int zone_movable_is_highmem(void)
601 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
602 return movable_zone == ZONE_HIGHMEM;
608 static inline int is_highmem_idx(enum zone_type idx)
610 #ifdef CONFIG_HIGHMEM
611 return (idx == ZONE_HIGHMEM ||
612 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
618 static inline int is_normal_idx(enum zone_type idx)
620 return (idx == ZONE_NORMAL);
624 * is_highmem - helper function to quickly check if a struct zone is a
625 * highmem zone or not. This is an attempt to keep references
626 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
627 * @zone - pointer to struct zone variable
629 static inline int is_highmem(struct zone *zone)
631 #ifdef CONFIG_HIGHMEM
632 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
633 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
634 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
635 zone_movable_is_highmem());
641 static inline int is_normal(struct zone *zone)
643 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
646 static inline int is_dma32(struct zone *zone)
648 #ifdef CONFIG_ZONE_DMA32
649 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
655 static inline int is_dma(struct zone *zone)
657 #ifdef CONFIG_ZONE_DMA
658 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
664 /* These two functions are used to setup the per zone pages min values */
667 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
668 void __user *, size_t *, loff_t *);
669 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
670 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
671 void __user *, size_t *, loff_t *);
672 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
673 void __user *, size_t *, loff_t *);
674 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
675 struct file *, void __user *, size_t *, loff_t *);
676 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
677 struct file *, void __user *, size_t *, loff_t *);
679 extern int numa_zonelist_order_handler(struct ctl_table *, int,
680 struct file *, void __user *, size_t *, loff_t *);
681 extern char numa_zonelist_order[];
682 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
684 #include <linux/topology.h>
685 /* Returns the number of the current Node. */
687 #define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
690 #ifndef CONFIG_NEED_MULTIPLE_NODES
692 extern struct pglist_data contig_page_data;
693 #define NODE_DATA(nid) (&contig_page_data)
694 #define NODE_MEM_MAP(nid) mem_map
696 #else /* CONFIG_NEED_MULTIPLE_NODES */
698 #include <asm/mmzone.h>
700 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
702 extern struct pglist_data *first_online_pgdat(void);
703 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
704 extern struct zone *next_zone(struct zone *zone);
707 * for_each_pgdat - helper macro to iterate over all nodes
708 * @pgdat - pointer to a pg_data_t variable
710 #define for_each_online_pgdat(pgdat) \
711 for (pgdat = first_online_pgdat(); \
713 pgdat = next_online_pgdat(pgdat))
715 * for_each_zone - helper macro to iterate over all memory zones
716 * @zone - pointer to struct zone variable
718 * The user only needs to declare the zone variable, for_each_zone
721 #define for_each_zone(zone) \
722 for (zone = (first_online_pgdat())->node_zones; \
724 zone = next_zone(zone))
726 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
728 return zoneref->zone;
731 static inline int zonelist_zone_idx(struct zoneref *zoneref)
733 return zoneref->zone_idx;
736 static inline int zonelist_node_idx(struct zoneref *zoneref)
739 /* zone_to_nid not available in this context */
740 return zoneref->zone->node;
743 #endif /* CONFIG_NUMA */
747 * 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
748 * @z - The cursor used as a starting point for the search
749 * @highest_zoneidx - The zone index of the highest zone to return
750 * @nodes - An optional nodemask to filter the zonelist with
751 * @zone - The first suitable zone found is returned via this parameter
753 * This function returns the next zone at or below a given zone index that is
754 * within the allowed nodemask using a cursor as the starting point for the
755 * search. The zoneref returned is a cursor that is used as the next starting
756 * point for future calls to next_zones_zonelist().
758 struct zoneref *next_zones_zonelist(struct zoneref *z,
759 enum zone_type highest_zoneidx,
764 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
765 * @zonelist - The zonelist to search for a suitable zone
766 * @highest_zoneidx - The zone index of the highest zone to return
767 * @nodes - An optional nodemask to filter the zonelist with
768 * @zone - The first suitable zone found is returned via this parameter
770 * This function returns the first zone at or below a given zone index that is
771 * within the allowed nodemask. The zoneref returned is a cursor that can be
772 * used to iterate the zonelist with next_zones_zonelist. The cursor should
773 * not be used by the caller as it does not match the value of the zone
776 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
777 enum zone_type highest_zoneidx,
781 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
786 * 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
787 * @zone - The current zone in the iterator
788 * @z - The current pointer within zonelist->zones being iterated
789 * @zlist - The zonelist being iterated
790 * @highidx - The zone index of the highest zone to return
791 * @nodemask - Nodemask allowed by the allocator
793 * This iterator iterates though all zones at or below a given zone index and
794 * within a given nodemask
796 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
797 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
799 z = next_zones_zonelist(z, highidx, nodemask, &zone)) \
802 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
803 * @zone - The current zone in the iterator
804 * @z - The current pointer within zonelist->zones being iterated
805 * @zlist - The zonelist being iterated
806 * @highidx - The zone index of the highest zone to return
808 * This iterator iterates though all zones at or below a given zone index.
810 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
811 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
813 #ifdef CONFIG_SPARSEMEM
814 #include <asm/sparsemem.h>
817 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
818 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
819 static inline unsigned long early_pfn_to_nid(unsigned long pfn)
825 #ifdef CONFIG_FLATMEM
826 #define pfn_to_nid(pfn) (0)
829 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
830 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
832 #ifdef CONFIG_SPARSEMEM
835 * SECTION_SHIFT #bits space required to store a section #
837 * PA_SECTION_SHIFT physical address to/from section number
838 * PFN_SECTION_SHIFT pfn to/from section number
840 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
842 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
843 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
845 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
847 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
848 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
850 #define SECTION_BLOCKFLAGS_BITS \
851 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
853 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
854 #error Allocator MAX_ORDER exceeds SECTION_SIZE
860 * This is, logically, a pointer to an array of struct
861 * pages. However, it is stored with some other magic.
862 * (see sparse.c::sparse_init_one_section())
864 * Additionally during early boot we encode node id of
865 * the location of the section here to guide allocation.
866 * (see sparse.c::memory_present())
868 * Making it a UL at least makes someone do a cast
869 * before using it wrong.
871 unsigned long section_mem_map;
873 /* See declaration of similar field in struct zone */
874 unsigned long *pageblock_flags;
877 #ifdef CONFIG_SPARSEMEM_EXTREME
878 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
880 #define SECTIONS_PER_ROOT 1
883 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
884 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
885 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
887 #ifdef CONFIG_SPARSEMEM_EXTREME
888 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
890 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
893 static inline struct mem_section *__nr_to_section(unsigned long nr)
895 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
897 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
899 extern int __section_nr(struct mem_section* ms);
900 extern unsigned long usemap_size(void);
903 * We use the lower bits of the mem_map pointer to store
904 * a little bit of information. There should be at least
905 * 3 bits here due to 32-bit alignment.
907 #define SECTION_MARKED_PRESENT (1UL<<0)
908 #define SECTION_HAS_MEM_MAP (1UL<<1)
909 #define SECTION_MAP_LAST_BIT (1UL<<2)
910 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
911 #define SECTION_NID_SHIFT 2
913 static inline struct page *__section_mem_map_addr(struct mem_section *section)
915 unsigned long map = section->section_mem_map;
916 map &= SECTION_MAP_MASK;
917 return (struct page *)map;
920 static inline int present_section(struct mem_section *section)
922 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
925 static inline int present_section_nr(unsigned long nr)
927 return present_section(__nr_to_section(nr));
930 static inline int valid_section(struct mem_section *section)
932 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
935 static inline int valid_section_nr(unsigned long nr)
937 return valid_section(__nr_to_section(nr));
940 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
942 return __nr_to_section(pfn_to_section_nr(pfn));
945 static inline int pfn_valid(unsigned long pfn)
947 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
949 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
952 static inline int pfn_present(unsigned long pfn)
954 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
956 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
960 * These are _only_ used during initialisation, therefore they
961 * can use __initdata ... They could have names to indicate
965 #define pfn_to_nid(pfn) \
967 unsigned long __pfn_to_nid_pfn = (pfn); \
968 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
971 #define pfn_to_nid(pfn) (0)
974 #define early_pfn_valid(pfn) pfn_valid(pfn)
975 void sparse_init(void);
977 #define sparse_init() do {} while (0)
978 #define sparse_index_init(_sec, _nid) do {} while (0)
979 #endif /* CONFIG_SPARSEMEM */
981 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
982 #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
984 #define early_pfn_in_nid(pfn, nid) (1)
987 #ifndef early_pfn_valid
988 #define early_pfn_valid(pfn) (1)
991 void memory_present(int nid, unsigned long start, unsigned long end);
992 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
995 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
996 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
997 * pfn_valid_within() should be used in this case; we optimise this away
998 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1000 #ifdef CONFIG_HOLES_IN_ZONE
1001 #define pfn_valid_within(pfn) pfn_valid(pfn)
1003 #define pfn_valid_within(pfn) (1)
1006 #endif /* !__GENERATING_BOUNDS.H */
1007 #endif /* !__ASSEMBLY__ */
1008 #endif /* __KERNEL__ */
1009 #endif /* _LINUX_MMZONE_H */