1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MMZONE_H
3 #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/page-flags-layout.h>
20 #include <linux/atomic.h>
21 #include <linux/mm_types.h>
22 #include <linux/page-flags.h>
23 #include <linux/local_lock.h>
26 /* Free memory management - zoned buddy allocator. */
27 #ifndef CONFIG_FORCE_MAX_ZONEORDER
30 #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
32 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
35 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
36 * costly to service. That is between allocation orders which should
37 * coalesce naturally under reasonable reclaim pressure and those which
40 #define PAGE_ALLOC_COSTLY_ORDER 3
46 MIGRATE_PCPTYPES, /* the number of types on the pcp lists */
47 MIGRATE_HIGHATOMIC = MIGRATE_PCPTYPES,
50 * MIGRATE_CMA migration type is designed to mimic the way
51 * ZONE_MOVABLE works. Only movable pages can be allocated
52 * from MIGRATE_CMA pageblocks and page allocator never
53 * implicitly change migration type of MIGRATE_CMA pageblock.
55 * The way to use it is to change migratetype of a range of
56 * pageblocks to MIGRATE_CMA which can be done by
57 * __free_pageblock_cma() function. What is important though
58 * is that a range of pageblocks must be aligned to
59 * MAX_ORDER_NR_PAGES should biggest page be bigger than
64 #ifdef CONFIG_MEMORY_ISOLATION
65 MIGRATE_ISOLATE, /* can't allocate from here */
70 /* In mm/page_alloc.c; keep in sync also with show_migration_types() there */
71 extern const char * const migratetype_names[MIGRATE_TYPES];
74 # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
75 # define is_migrate_cma_page(_page) (get_pageblock_migratetype(_page) == MIGRATE_CMA)
77 # define is_migrate_cma(migratetype) false
78 # define is_migrate_cma_page(_page) false
81 static inline bool is_migrate_movable(int mt)
83 return is_migrate_cma(mt) || mt == MIGRATE_MOVABLE;
86 #define for_each_migratetype_order(order, type) \
87 for (order = 0; order < MAX_ORDER; order++) \
88 for (type = 0; type < MIGRATE_TYPES; type++)
90 extern int page_group_by_mobility_disabled;
92 #define MIGRATETYPE_MASK ((1UL << PB_migratetype_bits) - 1)
94 #define get_pageblock_migratetype(page) \
95 get_pfnblock_flags_mask(page, page_to_pfn(page), MIGRATETYPE_MASK)
98 struct list_head free_list[MIGRATE_TYPES];
99 unsigned long nr_free;
102 static inline struct page *get_page_from_free_area(struct free_area *area,
105 return list_first_entry_or_null(&area->free_list[migratetype],
109 static inline bool free_area_empty(struct free_area *area, int migratetype)
111 return list_empty(&area->free_list[migratetype]);
117 * Add a wild amount of padding here to ensure datas fall into separate
118 * cachelines. There are very few zone structures in the machine, so space
119 * consumption is not a concern here.
121 #if defined(CONFIG_SMP)
122 struct zone_padding {
124 } ____cacheline_internodealigned_in_smp;
125 #define ZONE_PADDING(name) struct zone_padding name;
127 #define ZONE_PADDING(name)
131 enum numa_stat_item {
132 NUMA_HIT, /* allocated in intended node */
133 NUMA_MISS, /* allocated in non intended node */
134 NUMA_FOREIGN, /* was intended here, hit elsewhere */
135 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
136 NUMA_LOCAL, /* allocation from local node */
137 NUMA_OTHER, /* allocation from other node */
138 NR_VM_NUMA_STAT_ITEMS
141 #define NR_VM_NUMA_STAT_ITEMS 0
144 enum zone_stat_item {
145 /* First 128 byte cacheline (assuming 64 bit words) */
147 NR_ZONE_LRU_BASE, /* Used only for compaction and reclaim retry */
148 NR_ZONE_INACTIVE_ANON = NR_ZONE_LRU_BASE,
150 NR_ZONE_INACTIVE_FILE,
153 NR_ZONE_WRITE_PENDING, /* Count of dirty, writeback and unstable pages */
154 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
155 /* Second 128 byte cacheline */
157 #if IS_ENABLED(CONFIG_ZSMALLOC)
158 NR_ZSPAGES, /* allocated in zsmalloc */
161 NR_VM_ZONE_STAT_ITEMS };
163 enum node_stat_item {
165 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
166 NR_ACTIVE_ANON, /* " " " " " */
167 NR_INACTIVE_FILE, /* " " " " " */
168 NR_ACTIVE_FILE, /* " " " " " */
169 NR_UNEVICTABLE, /* " " " " " */
170 NR_SLAB_RECLAIMABLE_B,
171 NR_SLAB_UNRECLAIMABLE_B,
172 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
173 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
175 WORKINGSET_REFAULT_BASE,
176 WORKINGSET_REFAULT_ANON = WORKINGSET_REFAULT_BASE,
177 WORKINGSET_REFAULT_FILE,
178 WORKINGSET_ACTIVATE_BASE,
179 WORKINGSET_ACTIVATE_ANON = WORKINGSET_ACTIVATE_BASE,
180 WORKINGSET_ACTIVATE_FILE,
181 WORKINGSET_RESTORE_BASE,
182 WORKINGSET_RESTORE_ANON = WORKINGSET_RESTORE_BASE,
183 WORKINGSET_RESTORE_FILE,
184 WORKINGSET_NODERECLAIM,
185 NR_ANON_MAPPED, /* Mapped anonymous pages */
186 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
187 only modified from process context */
191 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
192 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
199 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
200 NR_DIRTIED, /* page dirtyings since bootup */
201 NR_WRITTEN, /* page writings since bootup */
202 NR_KERNEL_MISC_RECLAIMABLE, /* reclaimable non-slab kernel pages */
203 NR_FOLL_PIN_ACQUIRED, /* via: pin_user_page(), gup flag: FOLL_PIN */
204 NR_FOLL_PIN_RELEASED, /* pages returned via unpin_user_page() */
205 NR_KERNEL_STACK_KB, /* measured in KiB */
206 #if IS_ENABLED(CONFIG_SHADOW_CALL_STACK)
207 NR_KERNEL_SCS_KB, /* measured in KiB */
209 NR_PAGETABLE, /* used for pagetables */
213 NR_VM_NODE_STAT_ITEMS
217 * Returns true if the item should be printed in THPs (/proc/vmstat
218 * currently prints number of anon, file and shmem THPs. But the item
219 * is charged in pages).
221 static __always_inline bool vmstat_item_print_in_thp(enum node_stat_item item)
223 if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
226 return item == NR_ANON_THPS ||
227 item == NR_FILE_THPS ||
228 item == NR_SHMEM_THPS ||
229 item == NR_SHMEM_PMDMAPPED ||
230 item == NR_FILE_PMDMAPPED;
234 * Returns true if the value is measured in bytes (most vmstat values are
235 * measured in pages). This defines the API part, the internal representation
236 * might be different.
238 static __always_inline bool vmstat_item_in_bytes(int idx)
241 * Global and per-node slab counters track slab pages.
242 * It's expected that changes are multiples of PAGE_SIZE.
243 * Internally values are stored in pages.
245 * Per-memcg and per-lruvec counters track memory, consumed
246 * by individual slab objects. These counters are actually
249 return (idx == NR_SLAB_RECLAIMABLE_B ||
250 idx == NR_SLAB_UNRECLAIMABLE_B);
254 * We do arithmetic on the LRU lists in various places in the code,
255 * so it is important to keep the active lists LRU_ACTIVE higher in
256 * the array than the corresponding inactive lists, and to keep
257 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
259 * This has to be kept in sync with the statistics in zone_stat_item
260 * above and the descriptions in vmstat_text in mm/vmstat.c
267 LRU_INACTIVE_ANON = LRU_BASE,
268 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
269 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
270 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
275 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
277 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
279 static inline bool is_file_lru(enum lru_list lru)
281 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
284 static inline bool is_active_lru(enum lru_list lru)
286 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
289 #define ANON_AND_FILE 2
292 LRUVEC_CONGESTED, /* lruvec has many dirty pages
293 * backed by a congested BDI
298 struct list_head lists[NR_LRU_LISTS];
299 /* per lruvec lru_lock for memcg */
302 * These track the cost of reclaiming one LRU - file or anon -
303 * over the other. As the observed cost of reclaiming one LRU
304 * increases, the reclaim scan balance tips toward the other.
306 unsigned long anon_cost;
307 unsigned long file_cost;
308 /* Non-resident age, driven by LRU movement */
309 atomic_long_t nonresident_age;
310 /* Refaults at the time of last reclaim cycle */
311 unsigned long refaults[ANON_AND_FILE];
312 /* Various lruvec state flags (enum lruvec_flags) */
315 struct pglist_data *pgdat;
319 /* Isolate unmapped pages */
320 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
321 /* Isolate for asynchronous migration */
322 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
323 /* Isolate unevictable pages */
324 #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
326 /* LRU Isolation modes. */
327 typedef unsigned __bitwise isolate_mode_t;
329 enum zone_watermarks {
336 #define min_wmark_pages(z) (z->_watermark[WMARK_MIN] + z->watermark_boost)
337 #define low_wmark_pages(z) (z->_watermark[WMARK_LOW] + z->watermark_boost)
338 #define high_wmark_pages(z) (z->_watermark[WMARK_HIGH] + z->watermark_boost)
339 #define wmark_pages(z, i) (z->_watermark[i] + z->watermark_boost)
341 /* Fields and list protected by pagesets local_lock in page_alloc.c */
342 struct per_cpu_pages {
343 int count; /* number of pages in the list */
344 int high; /* high watermark, emptying needed */
345 int batch; /* chunk size for buddy add/remove */
347 int expire; /* When 0, remote pagesets are drained */
350 /* Lists of pages, one per migrate type stored on the pcp-lists */
351 struct list_head lists[MIGRATE_PCPTYPES];
354 struct per_cpu_zonestat {
356 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
360 u16 vm_numa_stat_diff[NR_VM_NUMA_STAT_ITEMS];
364 struct per_cpu_nodestat {
366 s8 vm_node_stat_diff[NR_VM_NODE_STAT_ITEMS];
369 #endif /* !__GENERATING_BOUNDS.H */
373 * ZONE_DMA and ZONE_DMA32 are used when there are peripherals not able
374 * to DMA to all of the addressable memory (ZONE_NORMAL).
375 * On architectures where this area covers the whole 32 bit address
376 * space ZONE_DMA32 is used. ZONE_DMA is left for the ones with smaller
377 * DMA addressing constraints. This distinction is important as a 32bit
378 * DMA mask is assumed when ZONE_DMA32 is defined. Some 64-bit
379 * platforms may need both zones as they support peripherals with
380 * different DMA addressing limitations.
382 #ifdef CONFIG_ZONE_DMA
385 #ifdef CONFIG_ZONE_DMA32
389 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
390 * performed on pages in ZONE_NORMAL if the DMA devices support
391 * transfers to all addressable memory.
394 #ifdef CONFIG_HIGHMEM
396 * A memory area that is only addressable by the kernel through
397 * mapping portions into its own address space. This is for example
398 * used by i386 to allow the kernel to address the memory beyond
399 * 900MB. The kernel will set up special mappings (page
400 * table entries on i386) for each page that the kernel needs to
406 * ZONE_MOVABLE is similar to ZONE_NORMAL, except that it contains
407 * movable pages with few exceptional cases described below. Main use
408 * cases for ZONE_MOVABLE are to make memory offlining/unplug more
409 * likely to succeed, and to locally limit unmovable allocations - e.g.,
410 * to increase the number of THP/huge pages. Notable special cases are:
412 * 1. Pinned pages: (long-term) pinning of movable pages might
413 * essentially turn such pages unmovable. Therefore, we do not allow
414 * pinning long-term pages in ZONE_MOVABLE. When pages are pinned and
415 * faulted, they come from the right zone right away. However, it is
416 * still possible that address space already has pages in
417 * ZONE_MOVABLE at the time when pages are pinned (i.e. user has
418 * touches that memory before pinning). In such case we migrate them
419 * to a different zone. When migration fails - pinning fails.
420 * 2. memblock allocations: kernelcore/movablecore setups might create
421 * situations where ZONE_MOVABLE contains unmovable allocations
422 * after boot. Memory offlining and allocations fail early.
423 * 3. Memory holes: kernelcore/movablecore setups might create very rare
424 * situations where ZONE_MOVABLE contains memory holes after boot,
425 * for example, if we have sections that are only partially
426 * populated. Memory offlining and allocations fail early.
427 * 4. PG_hwpoison pages: while poisoned pages can be skipped during
428 * memory offlining, such pages cannot be allocated.
429 * 5. Unmovable PG_offline pages: in paravirtualized environments,
430 * hotplugged memory blocks might only partially be managed by the
431 * buddy (e.g., via XEN-balloon, Hyper-V balloon, virtio-mem). The
432 * parts not manged by the buddy are unmovable PG_offline pages. In
433 * some cases (virtio-mem), such pages can be skipped during
434 * memory offlining, however, cannot be moved/allocated. These
435 * techniques might use alloc_contig_range() to hide previously
436 * exposed pages from the buddy again (e.g., to implement some sort
437 * of memory unplug in virtio-mem).
438 * 6. ZERO_PAGE(0), kernelcore/movablecore setups might create
439 * situations where ZERO_PAGE(0) which is allocated differently
440 * on different platforms may end up in a movable zone. ZERO_PAGE(0)
441 * cannot be migrated.
442 * 7. Memory-hotplug: when using memmap_on_memory and onlining the
443 * memory to the MOVABLE zone, the vmemmap pages are also placed in
444 * such zone. Such pages cannot be really moved around as they are
445 * self-stored in the range, but they are treated as movable when
446 * the range they describe is about to be offlined.
448 * In general, no unmovable allocations that degrade memory offlining
449 * should end up in ZONE_MOVABLE. Allocators (like alloc_contig_range())
450 * have to expect that migrating pages in ZONE_MOVABLE can fail (even
451 * if has_unmovable_pages() states that there are no unmovable pages,
452 * there can be false negatives).
455 #ifdef CONFIG_ZONE_DEVICE
462 #ifndef __GENERATING_BOUNDS_H
464 #define ASYNC_AND_SYNC 2
467 /* Read-mostly fields */
469 /* zone watermarks, access with *_wmark_pages(zone) macros */
470 unsigned long _watermark[NR_WMARK];
471 unsigned long watermark_boost;
473 unsigned long nr_reserved_highatomic;
476 * We don't know if the memory that we're going to allocate will be
477 * freeable or/and it will be released eventually, so to avoid totally
478 * wasting several GB of ram we must reserve some of the lower zone
479 * memory (otherwise we risk to run OOM on the lower zones despite
480 * there being tons of freeable ram on the higher zones). This array is
481 * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
484 long lowmem_reserve[MAX_NR_ZONES];
489 struct pglist_data *zone_pgdat;
490 struct per_cpu_pages __percpu *per_cpu_pageset;
491 struct per_cpu_zonestat __percpu *per_cpu_zonestats;
493 * the high and batch values are copied to individual pagesets for
499 #ifndef CONFIG_SPARSEMEM
501 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
502 * In SPARSEMEM, this map is stored in struct mem_section
504 unsigned long *pageblock_flags;
505 #endif /* CONFIG_SPARSEMEM */
507 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
508 unsigned long zone_start_pfn;
511 * spanned_pages is the total pages spanned by the zone, including
512 * holes, which is calculated as:
513 * spanned_pages = zone_end_pfn - zone_start_pfn;
515 * present_pages is physical pages existing within the zone, which
517 * present_pages = spanned_pages - absent_pages(pages in holes);
519 * managed_pages is present pages managed by the buddy system, which
520 * is calculated as (reserved_pages includes pages allocated by the
521 * bootmem allocator):
522 * managed_pages = present_pages - reserved_pages;
524 * cma pages is present pages that are assigned for CMA use
527 * So present_pages may be used by memory hotplug or memory power
528 * management logic to figure out unmanaged pages by checking
529 * (present_pages - managed_pages). And managed_pages should be used
530 * by page allocator and vm scanner to calculate all kinds of watermarks
535 * zone_start_pfn and spanned_pages are protected by span_seqlock.
536 * It is a seqlock because it has to be read outside of zone->lock,
537 * and it is done in the main allocator path. But, it is written
538 * quite infrequently.
540 * The span_seq lock is declared along with zone->lock because it is
541 * frequently read in proximity to zone->lock. It's good to
542 * give them a chance of being in the same cacheline.
544 * Write access to present_pages at runtime should be protected by
545 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
546 * present_pages should get_online_mems() to get a stable value.
548 atomic_long_t managed_pages;
549 unsigned long spanned_pages;
550 unsigned long present_pages;
552 unsigned long cma_pages;
557 #ifdef CONFIG_MEMORY_ISOLATION
559 * Number of isolated pageblock. It is used to solve incorrect
560 * freepage counting problem due to racy retrieving migratetype
561 * of pageblock. Protected by zone->lock.
563 unsigned long nr_isolate_pageblock;
566 #ifdef CONFIG_MEMORY_HOTPLUG
567 /* see spanned/present_pages for more description */
568 seqlock_t span_seqlock;
573 /* Write-intensive fields used from the page allocator */
576 /* free areas of different sizes */
577 struct free_area free_area[MAX_ORDER];
579 /* zone flags, see below */
582 /* Primarily protects free_area */
585 /* Write-intensive fields used by compaction and vmstats. */
589 * When free pages are below this point, additional steps are taken
590 * when reading the number of free pages to avoid per-cpu counter
591 * drift allowing watermarks to be breached
593 unsigned long percpu_drift_mark;
595 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
596 /* pfn where compaction free scanner should start */
597 unsigned long compact_cached_free_pfn;
598 /* pfn where compaction migration scanner should start */
599 unsigned long compact_cached_migrate_pfn[ASYNC_AND_SYNC];
600 unsigned long compact_init_migrate_pfn;
601 unsigned long compact_init_free_pfn;
604 #ifdef CONFIG_COMPACTION
606 * On compaction failure, 1<<compact_defer_shift compactions
607 * are skipped before trying again. The number attempted since
608 * last failure is tracked with compact_considered.
609 * compact_order_failed is the minimum compaction failed order.
611 unsigned int compact_considered;
612 unsigned int compact_defer_shift;
613 int compact_order_failed;
616 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
617 /* Set to true when the PG_migrate_skip bits should be cleared */
618 bool compact_blockskip_flush;
624 /* Zone statistics */
625 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
626 atomic_long_t vm_numa_stat[NR_VM_NUMA_STAT_ITEMS];
627 } ____cacheline_internodealigned_in_smp;
630 PGDAT_DIRTY, /* reclaim scanning has recently found
631 * many dirty file pages at the tail
634 PGDAT_WRITEBACK, /* reclaim scanning has recently found
635 * many pages under writeback
637 PGDAT_RECLAIM_LOCKED, /* prevents concurrent reclaim */
641 ZONE_BOOSTED_WATERMARK, /* zone recently boosted watermarks.
642 * Cleared when kswapd is woken.
646 static inline unsigned long zone_managed_pages(struct zone *zone)
648 return (unsigned long)atomic_long_read(&zone->managed_pages);
651 static inline unsigned long zone_cma_pages(struct zone *zone)
654 return zone->cma_pages;
660 static inline unsigned long zone_end_pfn(const struct zone *zone)
662 return zone->zone_start_pfn + zone->spanned_pages;
665 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
667 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
670 static inline bool zone_is_initialized(struct zone *zone)
672 return zone->initialized;
675 static inline bool zone_is_empty(struct zone *zone)
677 return zone->spanned_pages == 0;
681 * Return true if [start_pfn, start_pfn + nr_pages) range has a non-empty
682 * intersection with the given zone
684 static inline bool zone_intersects(struct zone *zone,
685 unsigned long start_pfn, unsigned long nr_pages)
687 if (zone_is_empty(zone))
689 if (start_pfn >= zone_end_pfn(zone) ||
690 start_pfn + nr_pages <= zone->zone_start_pfn)
697 * The "priority" of VM scanning is how much of the queues we will scan in one
698 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
699 * queues ("queue_length >> 12") during an aging round.
701 #define DEF_PRIORITY 12
703 /* Maximum number of zones on a zonelist */
704 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
707 ZONELIST_FALLBACK, /* zonelist with fallback */
710 * The NUMA zonelists are doubled because we need zonelists that
711 * restrict the allocations to a single node for __GFP_THISNODE.
713 ZONELIST_NOFALLBACK, /* zonelist without fallback (__GFP_THISNODE) */
719 * This struct contains information about a zone in a zonelist. It is stored
720 * here to avoid dereferences into large structures and lookups of tables
723 struct zone *zone; /* Pointer to actual zone */
724 int zone_idx; /* zone_idx(zoneref->zone) */
728 * One allocation request operates on a zonelist. A zonelist
729 * is a list of zones, the first one is the 'goal' of the
730 * allocation, the other zones are fallback zones, in decreasing
733 * To speed the reading of the zonelist, the zonerefs contain the zone index
734 * of the entry being read. Helper functions to access information given
735 * a struct zoneref are
737 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
738 * zonelist_zone_idx() - Return the index of the zone for an entry
739 * zonelist_node_idx() - Return the index of the node for an entry
742 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
745 #ifndef CONFIG_DISCONTIGMEM
746 /* The array of struct pages - for discontigmem use pgdat->lmem_map */
747 extern struct page *mem_map;
750 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
751 struct deferred_split {
752 spinlock_t split_queue_lock;
753 struct list_head split_queue;
754 unsigned long split_queue_len;
759 * On NUMA machines, each NUMA node would have a pg_data_t to describe
760 * it's memory layout. On UMA machines there is a single pglist_data which
761 * describes the whole memory.
763 * Memory statistics and page replacement data structures are maintained on a
766 typedef struct pglist_data {
768 * node_zones contains just the zones for THIS node. Not all of the
769 * zones may be populated, but it is the full list. It is referenced by
770 * this node's node_zonelists as well as other node's node_zonelists.
772 struct zone node_zones[MAX_NR_ZONES];
775 * node_zonelists contains references to all zones in all nodes.
776 * Generally the first zones will be references to this node's
779 struct zonelist node_zonelists[MAX_ZONELISTS];
781 int nr_zones; /* number of populated zones in this node */
782 #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
783 struct page *node_mem_map;
784 #ifdef CONFIG_PAGE_EXTENSION
785 struct page_ext *node_page_ext;
788 #if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)
790 * Must be held any time you expect node_start_pfn,
791 * node_present_pages, node_spanned_pages or nr_zones to stay constant.
792 * Also synchronizes pgdat->first_deferred_pfn during deferred page
795 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
796 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG
797 * or CONFIG_DEFERRED_STRUCT_PAGE_INIT.
799 * Nests above zone->lock and zone->span_seqlock
801 spinlock_t node_size_lock;
803 unsigned long node_start_pfn;
804 unsigned long node_present_pages; /* total number of physical pages */
805 unsigned long node_spanned_pages; /* total size of physical page
806 range, including holes */
808 wait_queue_head_t kswapd_wait;
809 wait_queue_head_t pfmemalloc_wait;
810 struct task_struct *kswapd; /* Protected by
811 mem_hotplug_begin/end() */
813 enum zone_type kswapd_highest_zoneidx;
815 int kswapd_failures; /* Number of 'reclaimed == 0' runs */
817 #ifdef CONFIG_COMPACTION
818 int kcompactd_max_order;
819 enum zone_type kcompactd_highest_zoneidx;
820 wait_queue_head_t kcompactd_wait;
821 struct task_struct *kcompactd;
824 * This is a per-node reserve of pages that are not available
825 * to userspace allocations.
827 unsigned long totalreserve_pages;
831 * node reclaim becomes active if more unmapped pages exist.
833 unsigned long min_unmapped_pages;
834 unsigned long min_slab_pages;
835 #endif /* CONFIG_NUMA */
837 /* Write-intensive fields used by page reclaim */
840 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
842 * If memory initialisation on large machines is deferred then this
843 * is the first PFN that needs to be initialised.
845 unsigned long first_deferred_pfn;
846 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
848 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
849 struct deferred_split deferred_split_queue;
852 /* Fields commonly accessed by the page reclaim scanner */
855 * NOTE: THIS IS UNUSED IF MEMCG IS ENABLED.
857 * Use mem_cgroup_lruvec() to look up lruvecs.
859 struct lruvec __lruvec;
865 /* Per-node vmstats */
866 struct per_cpu_nodestat __percpu *per_cpu_nodestats;
867 atomic_long_t vm_stat[NR_VM_NODE_STAT_ITEMS];
870 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
871 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
872 #ifdef CONFIG_FLAT_NODE_MEM_MAP
873 #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
875 #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
877 #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
879 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
880 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
882 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
884 return pgdat->node_start_pfn + pgdat->node_spanned_pages;
887 static inline bool pgdat_is_empty(pg_data_t *pgdat)
889 return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
892 #include <linux/memory_hotplug.h>
894 void build_all_zonelists(pg_data_t *pgdat);
895 void wakeup_kswapd(struct zone *zone, gfp_t gfp_mask, int order,
896 enum zone_type highest_zoneidx);
897 bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
898 int highest_zoneidx, unsigned int alloc_flags,
900 bool zone_watermark_ok(struct zone *z, unsigned int order,
901 unsigned long mark, int highest_zoneidx,
902 unsigned int alloc_flags);
903 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
904 unsigned long mark, int highest_zoneidx);
906 * Memory initialization context, use to differentiate memory added by
907 * the platform statically or via memory hotplug interface.
909 enum meminit_context {
914 extern void init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
917 extern void lruvec_init(struct lruvec *lruvec);
919 static inline struct pglist_data *lruvec_pgdat(struct lruvec *lruvec)
922 return lruvec->pgdat;
924 return container_of(lruvec, struct pglist_data, __lruvec);
928 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
929 int local_memory_node(int node_id);
931 static inline int local_memory_node(int node_id) { return node_id; };
935 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
937 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
939 #ifdef CONFIG_ZONE_DEVICE
940 static inline bool zone_is_zone_device(struct zone *zone)
942 return zone_idx(zone) == ZONE_DEVICE;
945 static inline bool zone_is_zone_device(struct zone *zone)
952 * Returns true if a zone has pages managed by the buddy allocator.
953 * All the reclaim decisions have to use this function rather than
954 * populated_zone(). If the whole zone is reserved then we can easily
955 * end up with populated_zone() && !managed_zone().
957 static inline bool managed_zone(struct zone *zone)
959 return zone_managed_pages(zone);
962 /* Returns true if a zone has memory */
963 static inline bool populated_zone(struct zone *zone)
965 return zone->present_pages;
969 static inline int zone_to_nid(struct zone *zone)
974 static inline void zone_set_nid(struct zone *zone, int nid)
979 static inline int zone_to_nid(struct zone *zone)
984 static inline void zone_set_nid(struct zone *zone, int nid) {}
987 extern int movable_zone;
989 static inline int is_highmem_idx(enum zone_type idx)
991 #ifdef CONFIG_HIGHMEM
992 return (idx == ZONE_HIGHMEM ||
993 (idx == ZONE_MOVABLE && movable_zone == ZONE_HIGHMEM));
1000 * is_highmem - helper function to quickly check if a struct zone is a
1001 * highmem zone or not. This is an attempt to keep references
1002 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
1003 * @zone: pointer to struct zone variable
1004 * Return: 1 for a highmem zone, 0 otherwise
1006 static inline int is_highmem(struct zone *zone)
1008 #ifdef CONFIG_HIGHMEM
1009 return is_highmem_idx(zone_idx(zone));
1015 /* These two functions are used to setup the per zone pages min values */
1018 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, void *, size_t *,
1020 int watermark_scale_factor_sysctl_handler(struct ctl_table *, int, void *,
1021 size_t *, loff_t *);
1022 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES];
1023 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, void *,
1024 size_t *, loff_t *);
1025 int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
1026 void *, size_t *, loff_t *);
1027 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
1028 void *, size_t *, loff_t *);
1029 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
1030 void *, size_t *, loff_t *);
1031 int numa_zonelist_order_handler(struct ctl_table *, int,
1032 void *, size_t *, loff_t *);
1033 extern int percpu_pagelist_fraction;
1034 extern char numa_zonelist_order[];
1035 #define NUMA_ZONELIST_ORDER_LEN 16
1037 #ifndef CONFIG_NEED_MULTIPLE_NODES
1039 extern struct pglist_data contig_page_data;
1040 #define NODE_DATA(nid) (&contig_page_data)
1041 #define NODE_MEM_MAP(nid) mem_map
1043 #else /* CONFIG_NEED_MULTIPLE_NODES */
1045 #include <asm/mmzone.h>
1047 #endif /* !CONFIG_NEED_MULTIPLE_NODES */
1049 extern struct pglist_data *first_online_pgdat(void);
1050 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
1051 extern struct zone *next_zone(struct zone *zone);
1054 * for_each_online_pgdat - helper macro to iterate over all online nodes
1055 * @pgdat: pointer to a pg_data_t variable
1057 #define for_each_online_pgdat(pgdat) \
1058 for (pgdat = first_online_pgdat(); \
1060 pgdat = next_online_pgdat(pgdat))
1062 * for_each_zone - helper macro to iterate over all memory zones
1063 * @zone: pointer to struct zone variable
1065 * The user only needs to declare the zone variable, for_each_zone
1068 #define for_each_zone(zone) \
1069 for (zone = (first_online_pgdat())->node_zones; \
1071 zone = next_zone(zone))
1073 #define for_each_populated_zone(zone) \
1074 for (zone = (first_online_pgdat())->node_zones; \
1076 zone = next_zone(zone)) \
1077 if (!populated_zone(zone)) \
1078 ; /* do nothing */ \
1081 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
1083 return zoneref->zone;
1086 static inline int zonelist_zone_idx(struct zoneref *zoneref)
1088 return zoneref->zone_idx;
1091 static inline int zonelist_node_idx(struct zoneref *zoneref)
1093 return zone_to_nid(zoneref->zone);
1096 struct zoneref *__next_zones_zonelist(struct zoneref *z,
1097 enum zone_type highest_zoneidx,
1101 * 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
1102 * @z: The cursor used as a starting point for the search
1103 * @highest_zoneidx: The zone index of the highest zone to return
1104 * @nodes: An optional nodemask to filter the zonelist with
1106 * This function returns the next zone at or below a given zone index that is
1107 * within the allowed nodemask using a cursor as the starting point for the
1108 * search. The zoneref returned is a cursor that represents the current zone
1109 * being examined. It should be advanced by one before calling
1110 * next_zones_zonelist again.
1112 * Return: the next zone at or below highest_zoneidx within the allowed
1113 * nodemask using a cursor within a zonelist as a starting point
1115 static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,
1116 enum zone_type highest_zoneidx,
1119 if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))
1121 return __next_zones_zonelist(z, highest_zoneidx, nodes);
1125 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
1126 * @zonelist: The zonelist to search for a suitable zone
1127 * @highest_zoneidx: The zone index of the highest zone to return
1128 * @nodes: An optional nodemask to filter the zonelist with
1130 * This function returns the first zone at or below a given zone index that is
1131 * within the allowed nodemask. The zoneref returned is a cursor that can be
1132 * used to iterate the zonelist with next_zones_zonelist by advancing it by
1133 * one before calling.
1135 * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is
1136 * never NULL). This may happen either genuinely, or due to concurrent nodemask
1137 * update due to cpuset modification.
1139 * Return: Zoneref pointer for the first suitable zone found
1141 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
1142 enum zone_type highest_zoneidx,
1145 return next_zones_zonelist(zonelist->_zonerefs,
1146 highest_zoneidx, nodes);
1150 * 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
1151 * @zone: The current zone in the iterator
1152 * @z: The current pointer within zonelist->_zonerefs being iterated
1153 * @zlist: The zonelist being iterated
1154 * @highidx: The zone index of the highest zone to return
1155 * @nodemask: Nodemask allowed by the allocator
1157 * This iterator iterates though all zones at or below a given zone index and
1158 * within a given nodemask
1160 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1161 for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \
1163 z = next_zones_zonelist(++z, highidx, nodemask), \
1164 zone = zonelist_zone(z))
1166 #define for_next_zone_zonelist_nodemask(zone, z, highidx, nodemask) \
1167 for (zone = z->zone; \
1169 z = next_zones_zonelist(++z, highidx, nodemask), \
1170 zone = zonelist_zone(z))
1174 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1175 * @zone: The current zone in the iterator
1176 * @z: The current pointer within zonelist->zones being iterated
1177 * @zlist: The zonelist being iterated
1178 * @highidx: The zone index of the highest zone to return
1180 * This iterator iterates though all zones at or below a given zone index.
1182 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
1183 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1185 #ifdef CONFIG_SPARSEMEM
1186 #include <asm/sparsemem.h>
1189 #ifdef CONFIG_FLATMEM
1190 #define pfn_to_nid(pfn) (0)
1193 #ifdef CONFIG_SPARSEMEM
1196 * SECTION_SHIFT #bits space required to store a section #
1198 * PA_SECTION_SHIFT physical address to/from section number
1199 * PFN_SECTION_SHIFT pfn to/from section number
1201 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1202 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1204 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1206 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1207 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1209 #define SECTION_BLOCKFLAGS_BITS \
1210 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1212 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1213 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1216 static inline unsigned long pfn_to_section_nr(unsigned long pfn)
1218 return pfn >> PFN_SECTION_SHIFT;
1220 static inline unsigned long section_nr_to_pfn(unsigned long sec)
1222 return sec << PFN_SECTION_SHIFT;
1225 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1226 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1228 #define SUBSECTION_SHIFT 21
1229 #define SUBSECTION_SIZE (1UL << SUBSECTION_SHIFT)
1231 #define PFN_SUBSECTION_SHIFT (SUBSECTION_SHIFT - PAGE_SHIFT)
1232 #define PAGES_PER_SUBSECTION (1UL << PFN_SUBSECTION_SHIFT)
1233 #define PAGE_SUBSECTION_MASK (~(PAGES_PER_SUBSECTION-1))
1235 #if SUBSECTION_SHIFT > SECTION_SIZE_BITS
1236 #error Subsection size exceeds section size
1238 #define SUBSECTIONS_PER_SECTION (1UL << (SECTION_SIZE_BITS - SUBSECTION_SHIFT))
1241 #define SUBSECTION_ALIGN_UP(pfn) ALIGN((pfn), PAGES_PER_SUBSECTION)
1242 #define SUBSECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SUBSECTION_MASK)
1244 struct mem_section_usage {
1245 #ifdef CONFIG_SPARSEMEM_VMEMMAP
1246 DECLARE_BITMAP(subsection_map, SUBSECTIONS_PER_SECTION);
1248 /* See declaration of similar field in struct zone */
1249 unsigned long pageblock_flags[0];
1252 void subsection_map_init(unsigned long pfn, unsigned long nr_pages);
1256 struct mem_section {
1258 * This is, logically, a pointer to an array of struct
1259 * pages. However, it is stored with some other magic.
1260 * (see sparse.c::sparse_init_one_section())
1262 * Additionally during early boot we encode node id of
1263 * the location of the section here to guide allocation.
1264 * (see sparse.c::memory_present())
1266 * Making it a UL at least makes someone do a cast
1267 * before using it wrong.
1269 unsigned long section_mem_map;
1271 struct mem_section_usage *usage;
1272 #ifdef CONFIG_PAGE_EXTENSION
1274 * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1275 * section. (see page_ext.h about this.)
1277 struct page_ext *page_ext;
1281 * WARNING: mem_section must be a power-of-2 in size for the
1282 * calculation and use of SECTION_ROOT_MASK to make sense.
1286 #ifdef CONFIG_SPARSEMEM_EXTREME
1287 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1289 #define SECTIONS_PER_ROOT 1
1292 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1293 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1294 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1296 #ifdef CONFIG_SPARSEMEM_EXTREME
1297 extern struct mem_section **mem_section;
1299 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1302 static inline unsigned long *section_to_usemap(struct mem_section *ms)
1304 return ms->usage->pageblock_flags;
1307 static inline struct mem_section *__nr_to_section(unsigned long nr)
1309 #ifdef CONFIG_SPARSEMEM_EXTREME
1313 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1315 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1317 extern unsigned long __section_nr(struct mem_section *ms);
1318 extern size_t mem_section_usage_size(void);
1321 * We use the lower bits of the mem_map pointer to store
1322 * a little bit of information. The pointer is calculated
1323 * as mem_map - section_nr_to_pfn(pnum). The result is
1324 * aligned to the minimum alignment of the two values:
1325 * 1. All mem_map arrays are page-aligned.
1326 * 2. section_nr_to_pfn() always clears PFN_SECTION_SHIFT
1327 * lowest bits. PFN_SECTION_SHIFT is arch-specific
1328 * (equal SECTION_SIZE_BITS - PAGE_SHIFT), and the
1329 * worst combination is powerpc with 256k pages,
1330 * which results in PFN_SECTION_SHIFT equal 6.
1331 * To sum it up, at least 6 bits are available.
1333 #define SECTION_MARKED_PRESENT (1UL<<0)
1334 #define SECTION_HAS_MEM_MAP (1UL<<1)
1335 #define SECTION_IS_ONLINE (1UL<<2)
1336 #define SECTION_IS_EARLY (1UL<<3)
1337 #define SECTION_TAINT_ZONE_DEVICE (1UL<<4)
1338 #define SECTION_MAP_LAST_BIT (1UL<<5)
1339 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1340 #define SECTION_NID_SHIFT 3
1342 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1344 unsigned long map = section->section_mem_map;
1345 map &= SECTION_MAP_MASK;
1346 return (struct page *)map;
1349 static inline int present_section(struct mem_section *section)
1351 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1354 static inline int present_section_nr(unsigned long nr)
1356 return present_section(__nr_to_section(nr));
1359 static inline int valid_section(struct mem_section *section)
1361 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1364 static inline int early_section(struct mem_section *section)
1366 return (section && (section->section_mem_map & SECTION_IS_EARLY));
1369 static inline int valid_section_nr(unsigned long nr)
1371 return valid_section(__nr_to_section(nr));
1374 static inline int online_section(struct mem_section *section)
1376 return (section && (section->section_mem_map & SECTION_IS_ONLINE));
1379 static inline int online_device_section(struct mem_section *section)
1381 unsigned long flags = SECTION_IS_ONLINE | SECTION_TAINT_ZONE_DEVICE;
1383 return section && ((section->section_mem_map & flags) == flags);
1386 static inline int online_section_nr(unsigned long nr)
1388 return online_section(__nr_to_section(nr));
1391 #ifdef CONFIG_MEMORY_HOTPLUG
1392 void online_mem_sections(unsigned long start_pfn, unsigned long end_pfn);
1393 void offline_mem_sections(unsigned long start_pfn, unsigned long end_pfn);
1396 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1398 return __nr_to_section(pfn_to_section_nr(pfn));
1401 extern unsigned long __highest_present_section_nr;
1403 static inline int subsection_map_index(unsigned long pfn)
1405 return (pfn & ~(PAGE_SECTION_MASK)) / PAGES_PER_SUBSECTION;
1408 #ifdef CONFIG_SPARSEMEM_VMEMMAP
1409 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
1411 int idx = subsection_map_index(pfn);
1413 return test_bit(idx, ms->usage->subsection_map);
1416 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
1422 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1423 static inline int pfn_valid(unsigned long pfn)
1425 struct mem_section *ms;
1427 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1429 ms = __nr_to_section(pfn_to_section_nr(pfn));
1430 if (!valid_section(ms))
1433 * Traditionally early sections always returned pfn_valid() for
1434 * the entire section-sized span.
1436 return early_section(ms) || pfn_section_valid(ms, pfn);
1440 static inline int pfn_in_present_section(unsigned long pfn)
1442 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1444 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1447 static inline unsigned long next_present_section_nr(unsigned long section_nr)
1449 while (++section_nr <= __highest_present_section_nr) {
1450 if (present_section_nr(section_nr))
1458 * These are _only_ used during initialisation, therefore they
1459 * can use __initdata ... They could have names to indicate
1463 #define pfn_to_nid(pfn) \
1465 unsigned long __pfn_to_nid_pfn = (pfn); \
1466 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1469 #define pfn_to_nid(pfn) (0)
1472 void sparse_init(void);
1474 #define sparse_init() do {} while (0)
1475 #define sparse_index_init(_sec, _nid) do {} while (0)
1476 #define pfn_in_present_section pfn_valid
1477 #define subsection_map_init(_pfn, _nr_pages) do {} while (0)
1478 #endif /* CONFIG_SPARSEMEM */
1481 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1482 * need to check pfn validity within that MAX_ORDER_NR_PAGES block.
1483 * pfn_valid_within() should be used in this case; we optimise this away
1484 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1486 #ifdef CONFIG_HOLES_IN_ZONE
1487 #define pfn_valid_within(pfn) pfn_valid(pfn)
1489 #define pfn_valid_within(pfn) (1)
1492 #endif /* !__GENERATING_BOUNDS.H */
1493 #endif /* !__ASSEMBLY__ */
1494 #endif /* _LINUX_MMZONE_H */