1 #ifndef _LINUX_MMZONE_H
2 #define _LINUX_MMZONE_H
7 #include <linux/spinlock.h>
8 #include <linux/list.h>
9 #include <linux/wait.h>
10 #include <linux/cache.h>
11 #include <linux/threads.h>
12 #include <linux/numa.h>
13 #include <linux/init.h>
14 #include <linux/seqlock.h>
15 #include <linux/nodemask.h>
16 #include <linux/pageblock-flags.h>
17 #include <asm/atomic.h>
20 /* Free memory management - zoned buddy allocator. */
21 #ifndef CONFIG_FORCE_MAX_ZONEORDER
24 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
26 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
29 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
30 * costly to service. That is between allocation orders which should
31 * coelesce naturally under reasonable reclaim pressure and those which
34 #define PAGE_ALLOC_COSTLY_ORDER 3
36 #define MIGRATE_UNMOVABLE 0
37 #define MIGRATE_RECLAIMABLE 1
38 #define MIGRATE_MOVABLE 2
39 #define MIGRATE_RESERVE 3
40 #define MIGRATE_ISOLATE 4 /* can't allocate from here */
41 #define MIGRATE_TYPES 5
43 #define for_each_migratetype_order(order, type) \
44 for (order = 0; order < MAX_ORDER; order++) \
45 for (type = 0; type < MIGRATE_TYPES; type++)
47 extern int page_group_by_mobility_disabled;
49 static inline int get_pageblock_migratetype(struct page *page)
51 if (unlikely(page_group_by_mobility_disabled))
52 return MIGRATE_UNMOVABLE;
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) */
84 NR_ANON_PAGES, /* Mapped anonymous pages */
85 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
86 only modified from process context */
90 /* Second 128 byte cacheline */
92 NR_SLAB_UNRECLAIMABLE,
93 NR_PAGETABLE, /* used for pagetables */
94 NR_UNSTABLE_NFS, /* NFS unstable pages */
98 NUMA_HIT, /* allocated in intended node */
99 NUMA_MISS, /* allocated in non intended node */
100 NUMA_FOREIGN, /* was intended here, hit elsewhere */
101 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
102 NUMA_LOCAL, /* allocation from local node */
103 NUMA_OTHER, /* allocation from other node */
105 NR_VM_ZONE_STAT_ITEMS };
107 struct per_cpu_pages {
108 int count; /* number of pages in the list */
109 int high; /* high watermark, emptying needed */
110 int batch; /* chunk size for buddy add/remove */
111 struct list_head list; /* the list of pages */
114 struct per_cpu_pageset {
115 struct per_cpu_pages pcp[2]; /* 0: hot. 1: cold */
121 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
123 } ____cacheline_aligned_in_smp;
126 #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
128 #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
132 #ifdef CONFIG_ZONE_DMA
134 * ZONE_DMA is used when there are devices that are not able
135 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
136 * carve out the portion of memory that is needed for these devices.
137 * The range is arch specific.
142 * ---------------------------
143 * parisc, ia64, sparc <4G
146 * alpha Unlimited or 0-16MB.
148 * i386, x86_64 and multiple other arches
153 #ifdef CONFIG_ZONE_DMA32
155 * x86_64 needs two ZONE_DMAs because it supports devices that are
156 * only able to do DMA to the lower 16M but also 32 bit devices that
157 * can only do DMA areas below 4G.
162 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
163 * performed on pages in ZONE_NORMAL if the DMA devices support
164 * transfers to all addressable memory.
167 #ifdef CONFIG_HIGHMEM
169 * A memory area that is only addressable by the kernel through
170 * mapping portions into its own address space. This is for example
171 * used by i386 to allow the kernel to address the memory beyond
172 * 900MB. The kernel will set up special mappings (page
173 * table entries on i386) for each page that the kernel needs to
183 * When a memory allocation must conform to specific limitations (such
184 * as being suitable for DMA) the caller will pass in hints to the
185 * allocator in the gfp_mask, in the zone modifier bits. These bits
186 * are used to select a priority ordered list of memory zones which
187 * match the requested limits. See gfp_zone() in include/linux/gfp.h
191 * Count the active zones. Note that the use of defined(X) outside
192 * #if and family is not necessarily defined so ensure we cannot use
193 * it later. Use __ZONE_COUNT to work out how many shift bits we need.
195 #define __ZONE_COUNT ( \
196 defined(CONFIG_ZONE_DMA) \
197 + defined(CONFIG_ZONE_DMA32) \
199 + defined(CONFIG_HIGHMEM) \
203 #define ZONES_SHIFT 0
204 #elif __ZONE_COUNT <= 2
205 #define ZONES_SHIFT 1
206 #elif __ZONE_COUNT <= 4
207 #define ZONES_SHIFT 2
209 #error ZONES_SHIFT -- too many zones configured adjust calculation
214 /* Fields commonly accessed by the page allocator */
215 unsigned long pages_min, pages_low, pages_high;
217 * We don't know if the memory that we're going to allocate will be freeable
218 * or/and it will be released eventually, so to avoid totally wasting several
219 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
220 * to run OOM on the lower zones despite there's tons of freeable ram
221 * on the higher zones). This array is recalculated at runtime if the
222 * sysctl_lowmem_reserve_ratio sysctl changes.
224 unsigned long lowmem_reserve[MAX_NR_ZONES];
229 * zone reclaim becomes active if more unmapped pages exist.
231 unsigned long min_unmapped_pages;
232 unsigned long min_slab_pages;
233 struct per_cpu_pageset *pageset[NR_CPUS];
235 struct per_cpu_pageset pageset[NR_CPUS];
238 * free areas of different sizes
241 #ifdef CONFIG_MEMORY_HOTPLUG
242 /* see spanned/present_pages for more description */
243 seqlock_t span_seqlock;
245 struct free_area free_area[MAX_ORDER];
247 #ifndef CONFIG_SPARSEMEM
249 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
250 * In SPARSEMEM, this map is stored in struct mem_section
252 unsigned long *pageblock_flags;
253 #endif /* CONFIG_SPARSEMEM */
258 /* Fields commonly accessed by the page reclaim scanner */
260 struct list_head active_list;
261 struct list_head inactive_list;
262 unsigned long nr_scan_active;
263 unsigned long nr_scan_inactive;
264 unsigned long pages_scanned; /* since last reclaim */
265 int all_unreclaimable; /* All pages pinned */
267 /* A count of how many reclaimers are scanning this zone */
268 atomic_t reclaim_in_progress;
270 /* Zone statistics */
271 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
274 * prev_priority holds the scanning priority for this zone. It is
275 * defined as the scanning priority at which we achieved our reclaim
276 * target at the previous try_to_free_pages() or balance_pgdat()
279 * We use prev_priority as a measure of how much stress page reclaim is
280 * under - it drives the swappiness decision: whether to unmap mapped
283 * Access to both this field is quite racy even on uniprocessor. But
284 * it is expected to average out OK.
290 /* Rarely used or read-mostly fields */
293 * wait_table -- the array holding the hash table
294 * wait_table_hash_nr_entries -- the size of the hash table array
295 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
297 * The purpose of all these is to keep track of the people
298 * waiting for a page to become available and make them
299 * runnable again when possible. The trouble is that this
300 * consumes a lot of space, especially when so few things
301 * wait on pages at a given time. So instead of using
302 * per-page waitqueues, we use a waitqueue hash table.
304 * The bucket discipline is to sleep on the same queue when
305 * colliding and wake all in that wait queue when removing.
306 * When something wakes, it must check to be sure its page is
307 * truly available, a la thundering herd. The cost of a
308 * collision is great, but given the expected load of the
309 * table, they should be so rare as to be outweighed by the
310 * benefits from the saved space.
312 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
313 * primary users of these fields, and in mm/page_alloc.c
314 * free_area_init_core() performs the initialization of them.
316 wait_queue_head_t * wait_table;
317 unsigned long wait_table_hash_nr_entries;
318 unsigned long wait_table_bits;
321 * Discontig memory support fields.
323 struct pglist_data *zone_pgdat;
324 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
325 unsigned long zone_start_pfn;
328 * zone_start_pfn, spanned_pages and present_pages are all
329 * protected by span_seqlock. It is a seqlock because it has
330 * to be read outside of zone->lock, and it is done in the main
331 * allocator path. But, it is written quite infrequently.
333 * The lock is declared along with zone->lock because it is
334 * frequently read in proximity to zone->lock. It's good to
335 * give them a chance of being in the same cacheline.
337 unsigned long spanned_pages; /* total size, including holes */
338 unsigned long present_pages; /* amount of memory (excluding holes) */
341 * rarely used fields:
344 } ____cacheline_internodealigned_in_smp;
347 * The "priority" of VM scanning is how much of the queues we will scan in one
348 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
349 * queues ("queue_length >> 12") during an aging round.
351 #define DEF_PRIORITY 12
353 /* Maximum number of zones on a zonelist */
354 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
359 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
360 * allocations to a single node for GFP_THISNODE.
362 * [0 .. MAX_NR_ZONES -1] : Zonelists with fallback
363 * [MAZ_NR_ZONES ... MAZ_ZONELISTS -1] : No fallback (GFP_THISNODE)
365 #define MAX_ZONELISTS (2 * MAX_NR_ZONES)
369 * We cache key information from each zonelist for smaller cache
370 * footprint when scanning for free pages in get_page_from_freelist().
372 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
373 * up short of free memory since the last time (last_fullzone_zap)
374 * we zero'd fullzones.
375 * 2) The array z_to_n[] maps each zone in the zonelist to its node
376 * id, so that we can efficiently evaluate whether that node is
377 * set in the current tasks mems_allowed.
379 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
380 * indexed by a zones offset in the zonelist zones[] array.
382 * The get_page_from_freelist() routine does two scans. During the
383 * first scan, we skip zones whose corresponding bit in 'fullzones'
384 * is set or whose corresponding node in current->mems_allowed (which
385 * comes from cpusets) is not set. During the second scan, we bypass
386 * this zonelist_cache, to ensure we look methodically at each zone.
388 * Once per second, we zero out (zap) fullzones, forcing us to
389 * reconsider nodes that might have regained more free memory.
390 * The field last_full_zap is the time we last zapped fullzones.
392 * This mechanism reduces the amount of time we waste repeatedly
393 * reexaming zones for free memory when they just came up low on
394 * memory momentarilly ago.
396 * The zonelist_cache struct members logically belong in struct
397 * zonelist. However, the mempolicy zonelists constructed for
398 * MPOL_BIND are intentionally variable length (and usually much
399 * shorter). A general purpose mechanism for handling structs with
400 * multiple variable length members is more mechanism than we want
401 * here. We resort to some special case hackery instead.
403 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
404 * part because they are shorter), so we put the fixed length stuff
405 * at the front of the zonelist struct, ending in a variable length
406 * zones[], as is needed by MPOL_BIND.
408 * Then we put the optional zonelist cache on the end of the zonelist
409 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
410 * the fixed length portion at the front of the struct. This pointer
411 * both enables us to find the zonelist cache, and in the case of
412 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
413 * to know that the zonelist cache is not there.
415 * The end result is that struct zonelists come in two flavors:
416 * 1) The full, fixed length version, shown below, and
417 * 2) The custom zonelists for MPOL_BIND.
418 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
420 * Even though there may be multiple CPU cores on a node modifying
421 * fullzones or last_full_zap in the same zonelist_cache at the same
422 * time, we don't lock it. This is just hint data - if it is wrong now
423 * and then, the allocator will still function, perhaps a bit slower.
427 struct zonelist_cache {
428 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
429 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
430 unsigned long last_full_zap; /* when last zap'd (jiffies) */
433 #define MAX_ZONELISTS MAX_NR_ZONES
434 struct zonelist_cache;
438 * One allocation request operates on a zonelist. A zonelist
439 * is a list of zones, the first one is the 'goal' of the
440 * allocation, the other zones are fallback zones, in decreasing
443 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
444 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
448 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
449 struct zone *zones[MAX_ZONES_PER_ZONELIST + 1]; // NULL delimited
451 struct zonelist_cache zlcache; // optional ...
457 * Only custom zonelists like MPOL_BIND need to be filtered as part of
458 * policies. As described in the comment for struct zonelist_cache, these
459 * zonelists will not have a zlcache so zlcache_ptr will not be set. Use
460 * that to determine if the zonelists needs to be filtered or not.
462 static inline int alloc_should_filter_zonelist(struct zonelist *zonelist)
464 return !zonelist->zlcache_ptr;
467 static inline int alloc_should_filter_zonelist(struct zonelist *zonelist)
471 #endif /* CONFIG_NUMA */
473 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
474 struct node_active_region {
475 unsigned long start_pfn;
476 unsigned long end_pfn;
479 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
481 #ifndef CONFIG_DISCONTIGMEM
482 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
483 extern struct page *mem_map;
487 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
488 * (mostly NUMA machines?) to denote a higher-level memory zone than the
491 * On NUMA machines, each NUMA node would have a pg_data_t to describe
492 * it's memory layout.
494 * Memory statistics and page replacement data structures are maintained on a
498 typedef struct pglist_data {
499 struct zone node_zones[MAX_NR_ZONES];
500 struct zonelist node_zonelists[MAX_ZONELISTS];
502 #ifdef CONFIG_FLAT_NODE_MEM_MAP
503 struct page *node_mem_map;
505 struct bootmem_data *bdata;
506 #ifdef CONFIG_MEMORY_HOTPLUG
508 * Must be held any time you expect node_start_pfn, node_present_pages
509 * or node_spanned_pages stay constant. Holding this will also
510 * guarantee that any pfn_valid() stays that way.
512 * Nests above zone->lock and zone->size_seqlock.
514 spinlock_t node_size_lock;
516 unsigned long node_start_pfn;
517 unsigned long node_present_pages; /* total number of physical pages */
518 unsigned long node_spanned_pages; /* total size of physical page
519 range, including holes */
521 wait_queue_head_t kswapd_wait;
522 struct task_struct *kswapd;
523 int kswapd_max_order;
526 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
527 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
528 #ifdef CONFIG_FLAT_NODE_MEM_MAP
529 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
531 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
533 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
535 #include <linux/memory_hotplug.h>
537 void get_zone_counts(unsigned long *active, unsigned long *inactive,
538 unsigned long *free);
539 void build_all_zonelists(void);
540 void wakeup_kswapd(struct zone *zone, int order);
541 int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
542 int classzone_idx, int alloc_flags);
543 enum memmap_context {
547 extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
549 enum memmap_context context);
551 #ifdef CONFIG_HAVE_MEMORY_PRESENT
552 void memory_present(int nid, unsigned long start, unsigned long end);
554 static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
557 #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
558 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
562 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
564 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
566 static inline int populated_zone(struct zone *zone)
568 return (!!zone->present_pages);
571 extern int movable_zone;
573 static inline int zone_movable_is_highmem(void)
575 #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
576 return movable_zone == ZONE_HIGHMEM;
582 static inline int is_highmem_idx(enum zone_type idx)
584 #ifdef CONFIG_HIGHMEM
585 return (idx == ZONE_HIGHMEM ||
586 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
592 static inline int is_normal_idx(enum zone_type idx)
594 return (idx == ZONE_NORMAL);
598 * is_highmem - helper function to quickly check if a struct zone is a
599 * highmem zone or not. This is an attempt to keep references
600 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
601 * @zone - pointer to struct zone variable
603 static inline int is_highmem(struct zone *zone)
605 #ifdef CONFIG_HIGHMEM
606 int zone_idx = zone - zone->zone_pgdat->node_zones;
607 return zone_idx == ZONE_HIGHMEM ||
608 (zone_idx == ZONE_MOVABLE && zone_movable_is_highmem());
614 static inline int is_normal(struct zone *zone)
616 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
619 static inline int is_dma32(struct zone *zone)
621 #ifdef CONFIG_ZONE_DMA32
622 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
628 static inline int is_dma(struct zone *zone)
630 #ifdef CONFIG_ZONE_DMA
631 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
637 /* These two functions are used to setup the per zone pages min values */
640 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
641 void __user *, size_t *, loff_t *);
642 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
643 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
644 void __user *, size_t *, loff_t *);
645 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
646 void __user *, size_t *, loff_t *);
647 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
648 struct file *, void __user *, size_t *, loff_t *);
649 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
650 struct file *, void __user *, size_t *, loff_t *);
652 extern int numa_zonelist_order_handler(struct ctl_table *, int,
653 struct file *, void __user *, size_t *, loff_t *);
654 extern char numa_zonelist_order[];
655 #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
657 #include <linux/topology.h>
658 /* Returns the number of the current Node. */
660 #define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
663 #ifndef CONFIG_NEED_MULTIPLE_NODES
665 extern struct pglist_data contig_page_data;
666 #define NODE_DATA(nid) (&contig_page_data)
667 #define NODE_MEM_MAP(nid) mem_map
668 #define MAX_NODES_SHIFT 1
670 #else /* CONFIG_NEED_MULTIPLE_NODES */
672 #include <asm/mmzone.h>
674 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
676 extern struct pglist_data *first_online_pgdat(void);
677 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
678 extern struct zone *next_zone(struct zone *zone);
681 * for_each_pgdat - helper macro to iterate over all nodes
682 * @pgdat - pointer to a pg_data_t variable
684 #define for_each_online_pgdat(pgdat) \
685 for (pgdat = first_online_pgdat(); \
687 pgdat = next_online_pgdat(pgdat))
689 * for_each_zone - helper macro to iterate over all memory zones
690 * @zone - pointer to struct zone variable
692 * The user only needs to declare the zone variable, for_each_zone
695 #define for_each_zone(zone) \
696 for (zone = (first_online_pgdat())->node_zones; \
698 zone = next_zone(zone))
700 #ifdef CONFIG_SPARSEMEM
701 #include <asm/sparsemem.h>
704 #if BITS_PER_LONG == 32
706 * with 32 bit page->flags field, we reserve 9 bits for node/zone info.
707 * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes.
709 #define FLAGS_RESERVED 9
711 #elif BITS_PER_LONG == 64
713 * with 64 bit flags field, there's plenty of room.
715 #define FLAGS_RESERVED 32
719 #error BITS_PER_LONG not defined
723 #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
724 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
725 #define early_pfn_to_nid(nid) (0UL)
728 #ifdef CONFIG_FLATMEM
729 #define pfn_to_nid(pfn) (0)
732 #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
733 #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
735 #ifdef CONFIG_SPARSEMEM
738 * SECTION_SHIFT #bits space required to store a section #
740 * PA_SECTION_SHIFT physical address to/from section number
741 * PFN_SECTION_SHIFT pfn to/from section number
743 #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
745 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
746 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
748 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
750 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
751 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
753 #define SECTION_BLOCKFLAGS_BITS \
754 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
756 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
757 #error Allocator MAX_ORDER exceeds SECTION_SIZE
763 * This is, logically, a pointer to an array of struct
764 * pages. However, it is stored with some other magic.
765 * (see sparse.c::sparse_init_one_section())
767 * Additionally during early boot we encode node id of
768 * the location of the section here to guide allocation.
769 * (see sparse.c::memory_present())
771 * Making it a UL at least makes someone do a cast
772 * before using it wrong.
774 unsigned long section_mem_map;
776 /* See declaration of similar field in struct zone */
777 unsigned long *pageblock_flags;
780 #ifdef CONFIG_SPARSEMEM_EXTREME
781 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
783 #define SECTIONS_PER_ROOT 1
786 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
787 #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
788 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
790 #ifdef CONFIG_SPARSEMEM_EXTREME
791 extern struct mem_section *mem_section[NR_SECTION_ROOTS];
793 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
796 static inline struct mem_section *__nr_to_section(unsigned long nr)
798 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
800 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
802 extern int __section_nr(struct mem_section* ms);
805 * We use the lower bits of the mem_map pointer to store
806 * a little bit of information. There should be at least
807 * 3 bits here due to 32-bit alignment.
809 #define SECTION_MARKED_PRESENT (1UL<<0)
810 #define SECTION_HAS_MEM_MAP (1UL<<1)
811 #define SECTION_MAP_LAST_BIT (1UL<<2)
812 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
813 #define SECTION_NID_SHIFT 2
815 static inline struct page *__section_mem_map_addr(struct mem_section *section)
817 unsigned long map = section->section_mem_map;
818 map &= SECTION_MAP_MASK;
819 return (struct page *)map;
822 static inline int present_section(struct mem_section *section)
824 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
827 static inline int present_section_nr(unsigned long nr)
829 return present_section(__nr_to_section(nr));
832 static inline int valid_section(struct mem_section *section)
834 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
837 static inline int valid_section_nr(unsigned long nr)
839 return valid_section(__nr_to_section(nr));
842 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
844 return __nr_to_section(pfn_to_section_nr(pfn));
847 static inline int pfn_valid(unsigned long pfn)
849 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
851 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
854 static inline int pfn_present(unsigned long pfn)
856 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
858 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
862 * These are _only_ used during initialisation, therefore they
863 * can use __initdata ... They could have names to indicate
867 #define pfn_to_nid(pfn) \
869 unsigned long __pfn_to_nid_pfn = (pfn); \
870 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
873 #define pfn_to_nid(pfn) (0)
876 #define early_pfn_valid(pfn) pfn_valid(pfn)
877 void sparse_init(void);
879 #define sparse_init() do {} while (0)
880 #define sparse_index_init(_sec, _nid) do {} while (0)
881 #endif /* CONFIG_SPARSEMEM */
883 #ifdef CONFIG_NODES_SPAN_OTHER_NODES
884 #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
886 #define early_pfn_in_nid(pfn, nid) (1)
889 #ifndef early_pfn_valid
890 #define early_pfn_valid(pfn) (1)
893 void memory_present(int nid, unsigned long start, unsigned long end);
894 unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
897 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
898 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
899 * pfn_valid_within() should be used in this case; we optimise this away
900 * when we have no holes within a MAX_ORDER_NR_PAGES block.
902 #ifdef CONFIG_HOLES_IN_ZONE
903 #define pfn_valid_within(pfn) pfn_valid(pfn)
905 #define pfn_valid_within(pfn) (1)
908 #endif /* !__ASSEMBLY__ */
909 #endif /* __KERNEL__ */
910 #endif /* _LINUX_MMZONE_H */