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;
87 * Check whether a migratetype can be merged with another migratetype.
89 * It is only mergeable when it can fall back to other migratetypes for
90 * allocation. See fallbacks[MIGRATE_TYPES][3] in page_alloc.c.
92 static inline bool migratetype_is_mergeable(int mt)
94 return mt < MIGRATE_PCPTYPES;
97 #define for_each_migratetype_order(order, type) \
98 for (order = 0; order < MAX_ORDER; order++) \
99 for (type = 0; type < MIGRATE_TYPES; type++)
101 extern int page_group_by_mobility_disabled;
103 #define MIGRATETYPE_MASK ((1UL << PB_migratetype_bits) - 1)
105 #define get_pageblock_migratetype(page) \
106 get_pfnblock_flags_mask(page, page_to_pfn(page), MIGRATETYPE_MASK)
109 struct list_head free_list[MIGRATE_TYPES];
110 unsigned long nr_free;
113 static inline struct page *get_page_from_free_area(struct free_area *area,
116 return list_first_entry_or_null(&area->free_list[migratetype],
120 static inline bool free_area_empty(struct free_area *area, int migratetype)
122 return list_empty(&area->free_list[migratetype]);
128 * Add a wild amount of padding here to ensure data fall into separate
129 * cachelines. There are very few zone structures in the machine, so space
130 * consumption is not a concern here.
132 #if defined(CONFIG_SMP)
133 struct zone_padding {
135 } ____cacheline_internodealigned_in_smp;
136 #define ZONE_PADDING(name) struct zone_padding name;
138 #define ZONE_PADDING(name)
142 enum numa_stat_item {
143 NUMA_HIT, /* allocated in intended node */
144 NUMA_MISS, /* allocated in non intended node */
145 NUMA_FOREIGN, /* was intended here, hit elsewhere */
146 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
147 NUMA_LOCAL, /* allocation from local node */
148 NUMA_OTHER, /* allocation from other node */
149 NR_VM_NUMA_EVENT_ITEMS
152 #define NR_VM_NUMA_EVENT_ITEMS 0
155 enum zone_stat_item {
156 /* First 128 byte cacheline (assuming 64 bit words) */
158 NR_ZONE_LRU_BASE, /* Used only for compaction and reclaim retry */
159 NR_ZONE_INACTIVE_ANON = NR_ZONE_LRU_BASE,
161 NR_ZONE_INACTIVE_FILE,
164 NR_ZONE_WRITE_PENDING, /* Count of dirty, writeback and unstable pages */
165 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
166 /* Second 128 byte cacheline */
168 #if IS_ENABLED(CONFIG_ZSMALLOC)
169 NR_ZSPAGES, /* allocated in zsmalloc */
172 NR_VM_ZONE_STAT_ITEMS };
174 enum node_stat_item {
176 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
177 NR_ACTIVE_ANON, /* " " " " " */
178 NR_INACTIVE_FILE, /* " " " " " */
179 NR_ACTIVE_FILE, /* " " " " " */
180 NR_UNEVICTABLE, /* " " " " " */
181 NR_SLAB_RECLAIMABLE_B,
182 NR_SLAB_UNRECLAIMABLE_B,
183 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
184 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
186 WORKINGSET_REFAULT_BASE,
187 WORKINGSET_REFAULT_ANON = WORKINGSET_REFAULT_BASE,
188 WORKINGSET_REFAULT_FILE,
189 WORKINGSET_ACTIVATE_BASE,
190 WORKINGSET_ACTIVATE_ANON = WORKINGSET_ACTIVATE_BASE,
191 WORKINGSET_ACTIVATE_FILE,
192 WORKINGSET_RESTORE_BASE,
193 WORKINGSET_RESTORE_ANON = WORKINGSET_RESTORE_BASE,
194 WORKINGSET_RESTORE_FILE,
195 WORKINGSET_NODERECLAIM,
196 NR_ANON_MAPPED, /* Mapped anonymous pages */
197 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
198 only modified from process context */
202 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
203 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
210 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
211 NR_DIRTIED, /* page dirtyings since bootup */
212 NR_WRITTEN, /* page writings since bootup */
213 NR_THROTTLED_WRITTEN, /* NR_WRITTEN while reclaim throttled */
214 NR_KERNEL_MISC_RECLAIMABLE, /* reclaimable non-slab kernel pages */
215 NR_FOLL_PIN_ACQUIRED, /* via: pin_user_page(), gup flag: FOLL_PIN */
216 NR_FOLL_PIN_RELEASED, /* pages returned via unpin_user_page() */
217 NR_KERNEL_STACK_KB, /* measured in KiB */
218 #if IS_ENABLED(CONFIG_SHADOW_CALL_STACK)
219 NR_KERNEL_SCS_KB, /* measured in KiB */
221 NR_PAGETABLE, /* used for pagetables */
225 #ifdef CONFIG_NUMA_BALANCING
226 PGPROMOTE_SUCCESS, /* promote successfully */
228 NR_VM_NODE_STAT_ITEMS
232 * Returns true if the item should be printed in THPs (/proc/vmstat
233 * currently prints number of anon, file and shmem THPs. But the item
234 * is charged in pages).
236 static __always_inline bool vmstat_item_print_in_thp(enum node_stat_item item)
238 if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
241 return item == NR_ANON_THPS ||
242 item == NR_FILE_THPS ||
243 item == NR_SHMEM_THPS ||
244 item == NR_SHMEM_PMDMAPPED ||
245 item == NR_FILE_PMDMAPPED;
249 * Returns true if the value is measured in bytes (most vmstat values are
250 * measured in pages). This defines the API part, the internal representation
251 * might be different.
253 static __always_inline bool vmstat_item_in_bytes(int idx)
256 * Global and per-node slab counters track slab pages.
257 * It's expected that changes are multiples of PAGE_SIZE.
258 * Internally values are stored in pages.
260 * Per-memcg and per-lruvec counters track memory, consumed
261 * by individual slab objects. These counters are actually
264 return (idx == NR_SLAB_RECLAIMABLE_B ||
265 idx == NR_SLAB_UNRECLAIMABLE_B);
269 * We do arithmetic on the LRU lists in various places in the code,
270 * so it is important to keep the active lists LRU_ACTIVE higher in
271 * the array than the corresponding inactive lists, and to keep
272 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
274 * This has to be kept in sync with the statistics in zone_stat_item
275 * above and the descriptions in vmstat_text in mm/vmstat.c
282 LRU_INACTIVE_ANON = LRU_BASE,
283 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
284 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
285 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
290 enum vmscan_throttle_state {
291 VMSCAN_THROTTLE_WRITEBACK,
292 VMSCAN_THROTTLE_ISOLATED,
293 VMSCAN_THROTTLE_NOPROGRESS,
294 VMSCAN_THROTTLE_CONGESTED,
298 #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
300 #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
302 static inline bool is_file_lru(enum lru_list lru)
304 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
307 static inline bool is_active_lru(enum lru_list lru)
309 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
312 #define ANON_AND_FILE 2
315 LRUVEC_CONGESTED, /* lruvec has many dirty pages
316 * backed by a congested BDI
321 struct list_head lists[NR_LRU_LISTS];
322 /* per lruvec lru_lock for memcg */
325 * These track the cost of reclaiming one LRU - file or anon -
326 * over the other. As the observed cost of reclaiming one LRU
327 * increases, the reclaim scan balance tips toward the other.
329 unsigned long anon_cost;
330 unsigned long file_cost;
331 /* Non-resident age, driven by LRU movement */
332 atomic_long_t nonresident_age;
333 /* Refaults at the time of last reclaim cycle */
334 unsigned long refaults[ANON_AND_FILE];
335 /* Various lruvec state flags (enum lruvec_flags) */
338 struct pglist_data *pgdat;
342 /* Isolate unmapped pages */
343 #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
344 /* Isolate for asynchronous migration */
345 #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
346 /* Isolate unevictable pages */
347 #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
349 /* LRU Isolation modes. */
350 typedef unsigned __bitwise isolate_mode_t;
352 enum zone_watermarks {
360 * One per migratetype for each PAGE_ALLOC_COSTLY_ORDER plus one additional
361 * for pageblock size for THP if configured.
363 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
368 #define NR_PCP_LISTS (MIGRATE_PCPTYPES * (PAGE_ALLOC_COSTLY_ORDER + 1 + NR_PCP_THP))
371 * Shift to encode migratetype and order in the same integer, with order
372 * in the least significant bits.
374 #define NR_PCP_ORDER_WIDTH 8
375 #define NR_PCP_ORDER_MASK ((1<<NR_PCP_ORDER_WIDTH) - 1)
377 #define min_wmark_pages(z) (z->_watermark[WMARK_MIN] + z->watermark_boost)
378 #define low_wmark_pages(z) (z->_watermark[WMARK_LOW] + z->watermark_boost)
379 #define high_wmark_pages(z) (z->_watermark[WMARK_HIGH] + z->watermark_boost)
380 #define wmark_pages(z, i) (z->_watermark[i] + z->watermark_boost)
382 /* Fields and list protected by pagesets local_lock in page_alloc.c */
383 struct per_cpu_pages {
384 int count; /* number of pages in the list */
385 int high; /* high watermark, emptying needed */
386 int batch; /* chunk size for buddy add/remove */
387 short free_factor; /* batch scaling factor during free */
389 short expire; /* When 0, remote pagesets are drained */
392 /* Lists of pages, one per migrate type stored on the pcp-lists */
393 struct list_head lists[NR_PCP_LISTS];
396 struct per_cpu_zonestat {
398 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
403 * Low priority inaccurate counters that are only folded
404 * on demand. Use a large type to avoid the overhead of
405 * folding during refresh_cpu_vm_stats.
407 unsigned long vm_numa_event[NR_VM_NUMA_EVENT_ITEMS];
411 struct per_cpu_nodestat {
413 s8 vm_node_stat_diff[NR_VM_NODE_STAT_ITEMS];
416 #endif /* !__GENERATING_BOUNDS.H */
420 * ZONE_DMA and ZONE_DMA32 are used when there are peripherals not able
421 * to DMA to all of the addressable memory (ZONE_NORMAL).
422 * On architectures where this area covers the whole 32 bit address
423 * space ZONE_DMA32 is used. ZONE_DMA is left for the ones with smaller
424 * DMA addressing constraints. This distinction is important as a 32bit
425 * DMA mask is assumed when ZONE_DMA32 is defined. Some 64-bit
426 * platforms may need both zones as they support peripherals with
427 * different DMA addressing limitations.
429 #ifdef CONFIG_ZONE_DMA
432 #ifdef CONFIG_ZONE_DMA32
436 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
437 * performed on pages in ZONE_NORMAL if the DMA devices support
438 * transfers to all addressable memory.
441 #ifdef CONFIG_HIGHMEM
443 * A memory area that is only addressable by the kernel through
444 * mapping portions into its own address space. This is for example
445 * used by i386 to allow the kernel to address the memory beyond
446 * 900MB. The kernel will set up special mappings (page
447 * table entries on i386) for each page that the kernel needs to
453 * ZONE_MOVABLE is similar to ZONE_NORMAL, except that it contains
454 * movable pages with few exceptional cases described below. Main use
455 * cases for ZONE_MOVABLE are to make memory offlining/unplug more
456 * likely to succeed, and to locally limit unmovable allocations - e.g.,
457 * to increase the number of THP/huge pages. Notable special cases are:
459 * 1. Pinned pages: (long-term) pinning of movable pages might
460 * essentially turn such pages unmovable. Therefore, we do not allow
461 * pinning long-term pages in ZONE_MOVABLE. When pages are pinned and
462 * faulted, they come from the right zone right away. However, it is
463 * still possible that address space already has pages in
464 * ZONE_MOVABLE at the time when pages are pinned (i.e. user has
465 * touches that memory before pinning). In such case we migrate them
466 * to a different zone. When migration fails - pinning fails.
467 * 2. memblock allocations: kernelcore/movablecore setups might create
468 * situations where ZONE_MOVABLE contains unmovable allocations
469 * after boot. Memory offlining and allocations fail early.
470 * 3. Memory holes: kernelcore/movablecore setups might create very rare
471 * situations where ZONE_MOVABLE contains memory holes after boot,
472 * for example, if we have sections that are only partially
473 * populated. Memory offlining and allocations fail early.
474 * 4. PG_hwpoison pages: while poisoned pages can be skipped during
475 * memory offlining, such pages cannot be allocated.
476 * 5. Unmovable PG_offline pages: in paravirtualized environments,
477 * hotplugged memory blocks might only partially be managed by the
478 * buddy (e.g., via XEN-balloon, Hyper-V balloon, virtio-mem). The
479 * parts not manged by the buddy are unmovable PG_offline pages. In
480 * some cases (virtio-mem), such pages can be skipped during
481 * memory offlining, however, cannot be moved/allocated. These
482 * techniques might use alloc_contig_range() to hide previously
483 * exposed pages from the buddy again (e.g., to implement some sort
484 * of memory unplug in virtio-mem).
485 * 6. ZERO_PAGE(0), kernelcore/movablecore setups might create
486 * situations where ZERO_PAGE(0) which is allocated differently
487 * on different platforms may end up in a movable zone. ZERO_PAGE(0)
488 * cannot be migrated.
489 * 7. Memory-hotplug: when using memmap_on_memory and onlining the
490 * memory to the MOVABLE zone, the vmemmap pages are also placed in
491 * such zone. Such pages cannot be really moved around as they are
492 * self-stored in the range, but they are treated as movable when
493 * the range they describe is about to be offlined.
495 * In general, no unmovable allocations that degrade memory offlining
496 * should end up in ZONE_MOVABLE. Allocators (like alloc_contig_range())
497 * have to expect that migrating pages in ZONE_MOVABLE can fail (even
498 * if has_unmovable_pages() states that there are no unmovable pages,
499 * there can be false negatives).
502 #ifdef CONFIG_ZONE_DEVICE
509 #ifndef __GENERATING_BOUNDS_H
511 #define ASYNC_AND_SYNC 2
514 /* Read-mostly fields */
516 /* zone watermarks, access with *_wmark_pages(zone) macros */
517 unsigned long _watermark[NR_WMARK];
518 unsigned long watermark_boost;
520 unsigned long nr_reserved_highatomic;
523 * We don't know if the memory that we're going to allocate will be
524 * freeable or/and it will be released eventually, so to avoid totally
525 * wasting several GB of ram we must reserve some of the lower zone
526 * memory (otherwise we risk to run OOM on the lower zones despite
527 * there being tons of freeable ram on the higher zones). This array is
528 * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
531 long lowmem_reserve[MAX_NR_ZONES];
536 struct pglist_data *zone_pgdat;
537 struct per_cpu_pages __percpu *per_cpu_pageset;
538 struct per_cpu_zonestat __percpu *per_cpu_zonestats;
540 * the high and batch values are copied to individual pagesets for
546 #ifndef CONFIG_SPARSEMEM
548 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
549 * In SPARSEMEM, this map is stored in struct mem_section
551 unsigned long *pageblock_flags;
552 #endif /* CONFIG_SPARSEMEM */
554 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
555 unsigned long zone_start_pfn;
558 * spanned_pages is the total pages spanned by the zone, including
559 * holes, which is calculated as:
560 * spanned_pages = zone_end_pfn - zone_start_pfn;
562 * present_pages is physical pages existing within the zone, which
564 * present_pages = spanned_pages - absent_pages(pages in holes);
566 * present_early_pages is present pages existing within the zone
567 * located on memory available since early boot, excluding hotplugged
570 * managed_pages is present pages managed by the buddy system, which
571 * is calculated as (reserved_pages includes pages allocated by the
572 * bootmem allocator):
573 * managed_pages = present_pages - reserved_pages;
575 * cma pages is present pages that are assigned for CMA use
578 * So present_pages may be used by memory hotplug or memory power
579 * management logic to figure out unmanaged pages by checking
580 * (present_pages - managed_pages). And managed_pages should be used
581 * by page allocator and vm scanner to calculate all kinds of watermarks
586 * zone_start_pfn and spanned_pages are protected by span_seqlock.
587 * It is a seqlock because it has to be read outside of zone->lock,
588 * and it is done in the main allocator path. But, it is written
589 * quite infrequently.
591 * The span_seq lock is declared along with zone->lock because it is
592 * frequently read in proximity to zone->lock. It's good to
593 * give them a chance of being in the same cacheline.
595 * Write access to present_pages at runtime should be protected by
596 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
597 * present_pages should get_online_mems() to get a stable value.
599 atomic_long_t managed_pages;
600 unsigned long spanned_pages;
601 unsigned long present_pages;
602 #if defined(CONFIG_MEMORY_HOTPLUG)
603 unsigned long present_early_pages;
606 unsigned long cma_pages;
611 #ifdef CONFIG_MEMORY_ISOLATION
613 * Number of isolated pageblock. It is used to solve incorrect
614 * freepage counting problem due to racy retrieving migratetype
615 * of pageblock. Protected by zone->lock.
617 unsigned long nr_isolate_pageblock;
620 #ifdef CONFIG_MEMORY_HOTPLUG
621 /* see spanned/present_pages for more description */
622 seqlock_t span_seqlock;
627 /* Write-intensive fields used from the page allocator */
630 /* free areas of different sizes */
631 struct free_area free_area[MAX_ORDER];
633 /* zone flags, see below */
636 /* Primarily protects free_area */
639 /* Write-intensive fields used by compaction and vmstats. */
643 * When free pages are below this point, additional steps are taken
644 * when reading the number of free pages to avoid per-cpu counter
645 * drift allowing watermarks to be breached
647 unsigned long percpu_drift_mark;
649 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
650 /* pfn where compaction free scanner should start */
651 unsigned long compact_cached_free_pfn;
652 /* pfn where compaction migration scanner should start */
653 unsigned long compact_cached_migrate_pfn[ASYNC_AND_SYNC];
654 unsigned long compact_init_migrate_pfn;
655 unsigned long compact_init_free_pfn;
658 #ifdef CONFIG_COMPACTION
660 * On compaction failure, 1<<compact_defer_shift compactions
661 * are skipped before trying again. The number attempted since
662 * last failure is tracked with compact_considered.
663 * compact_order_failed is the minimum compaction failed order.
665 unsigned int compact_considered;
666 unsigned int compact_defer_shift;
667 int compact_order_failed;
670 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
671 /* Set to true when the PG_migrate_skip bits should be cleared */
672 bool compact_blockskip_flush;
678 /* Zone statistics */
679 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
680 atomic_long_t vm_numa_event[NR_VM_NUMA_EVENT_ITEMS];
681 } ____cacheline_internodealigned_in_smp;
684 PGDAT_DIRTY, /* reclaim scanning has recently found
685 * many dirty file pages at the tail
688 PGDAT_WRITEBACK, /* reclaim scanning has recently found
689 * many pages under writeback
691 PGDAT_RECLAIM_LOCKED, /* prevents concurrent reclaim */
695 ZONE_BOOSTED_WATERMARK, /* zone recently boosted watermarks.
696 * Cleared when kswapd is woken.
698 ZONE_RECLAIM_ACTIVE, /* kswapd may be scanning the zone. */
701 static inline unsigned long zone_managed_pages(struct zone *zone)
703 return (unsigned long)atomic_long_read(&zone->managed_pages);
706 static inline unsigned long zone_cma_pages(struct zone *zone)
709 return zone->cma_pages;
715 static inline unsigned long zone_end_pfn(const struct zone *zone)
717 return zone->zone_start_pfn + zone->spanned_pages;
720 static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
722 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
725 static inline bool zone_is_initialized(struct zone *zone)
727 return zone->initialized;
730 static inline bool zone_is_empty(struct zone *zone)
732 return zone->spanned_pages == 0;
736 * Return true if [start_pfn, start_pfn + nr_pages) range has a non-empty
737 * intersection with the given zone
739 static inline bool zone_intersects(struct zone *zone,
740 unsigned long start_pfn, unsigned long nr_pages)
742 if (zone_is_empty(zone))
744 if (start_pfn >= zone_end_pfn(zone) ||
745 start_pfn + nr_pages <= zone->zone_start_pfn)
752 * The "priority" of VM scanning is how much of the queues we will scan in one
753 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
754 * queues ("queue_length >> 12") during an aging round.
756 #define DEF_PRIORITY 12
758 /* Maximum number of zones on a zonelist */
759 #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
762 ZONELIST_FALLBACK, /* zonelist with fallback */
765 * The NUMA zonelists are doubled because we need zonelists that
766 * restrict the allocations to a single node for __GFP_THISNODE.
768 ZONELIST_NOFALLBACK, /* zonelist without fallback (__GFP_THISNODE) */
774 * This struct contains information about a zone in a zonelist. It is stored
775 * here to avoid dereferences into large structures and lookups of tables
778 struct zone *zone; /* Pointer to actual zone */
779 int zone_idx; /* zone_idx(zoneref->zone) */
783 * One allocation request operates on a zonelist. A zonelist
784 * is a list of zones, the first one is the 'goal' of the
785 * allocation, the other zones are fallback zones, in decreasing
788 * To speed the reading of the zonelist, the zonerefs contain the zone index
789 * of the entry being read. Helper functions to access information given
790 * a struct zoneref are
792 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
793 * zonelist_zone_idx() - Return the index of the zone for an entry
794 * zonelist_node_idx() - Return the index of the node for an entry
797 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
801 * The array of struct pages for flatmem.
802 * It must be declared for SPARSEMEM as well because there are configurations
805 extern struct page *mem_map;
807 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
808 struct deferred_split {
809 spinlock_t split_queue_lock;
810 struct list_head split_queue;
811 unsigned long split_queue_len;
816 * On NUMA machines, each NUMA node would have a pg_data_t to describe
817 * it's memory layout. On UMA machines there is a single pglist_data which
818 * describes the whole memory.
820 * Memory statistics and page replacement data structures are maintained on a
823 typedef struct pglist_data {
825 * node_zones contains just the zones for THIS node. Not all of the
826 * zones may be populated, but it is the full list. It is referenced by
827 * this node's node_zonelists as well as other node's node_zonelists.
829 struct zone node_zones[MAX_NR_ZONES];
832 * node_zonelists contains references to all zones in all nodes.
833 * Generally the first zones will be references to this node's
836 struct zonelist node_zonelists[MAX_ZONELISTS];
838 int nr_zones; /* number of populated zones in this node */
839 #ifdef CONFIG_FLATMEM /* means !SPARSEMEM */
840 struct page *node_mem_map;
841 #ifdef CONFIG_PAGE_EXTENSION
842 struct page_ext *node_page_ext;
845 #if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)
847 * Must be held any time you expect node_start_pfn,
848 * node_present_pages, node_spanned_pages or nr_zones to stay constant.
849 * Also synchronizes pgdat->first_deferred_pfn during deferred page
852 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
853 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG
854 * or CONFIG_DEFERRED_STRUCT_PAGE_INIT.
856 * Nests above zone->lock and zone->span_seqlock
858 spinlock_t node_size_lock;
860 unsigned long node_start_pfn;
861 unsigned long node_present_pages; /* total number of physical pages */
862 unsigned long node_spanned_pages; /* total size of physical page
863 range, including holes */
865 wait_queue_head_t kswapd_wait;
866 wait_queue_head_t pfmemalloc_wait;
868 /* workqueues for throttling reclaim for different reasons. */
869 wait_queue_head_t reclaim_wait[NR_VMSCAN_THROTTLE];
871 atomic_t nr_writeback_throttled;/* nr of writeback-throttled tasks */
872 unsigned long nr_reclaim_start; /* nr pages written while throttled
873 * when throttling started. */
874 struct task_struct *kswapd; /* Protected by
875 mem_hotplug_begin/end() */
877 enum zone_type kswapd_highest_zoneidx;
879 int kswapd_failures; /* Number of 'reclaimed == 0' runs */
881 #ifdef CONFIG_COMPACTION
882 int kcompactd_max_order;
883 enum zone_type kcompactd_highest_zoneidx;
884 wait_queue_head_t kcompactd_wait;
885 struct task_struct *kcompactd;
886 bool proactive_compact_trigger;
889 * This is a per-node reserve of pages that are not available
890 * to userspace allocations.
892 unsigned long totalreserve_pages;
896 * node reclaim becomes active if more unmapped pages exist.
898 unsigned long min_unmapped_pages;
899 unsigned long min_slab_pages;
900 #endif /* CONFIG_NUMA */
902 /* Write-intensive fields used by page reclaim */
905 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
907 * If memory initialisation on large machines is deferred then this
908 * is the first PFN that needs to be initialised.
910 unsigned long first_deferred_pfn;
911 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
913 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
914 struct deferred_split deferred_split_queue;
917 /* Fields commonly accessed by the page reclaim scanner */
920 * NOTE: THIS IS UNUSED IF MEMCG IS ENABLED.
922 * Use mem_cgroup_lruvec() to look up lruvecs.
924 struct lruvec __lruvec;
930 /* Per-node vmstats */
931 struct per_cpu_nodestat __percpu *per_cpu_nodestats;
932 atomic_long_t vm_stat[NR_VM_NODE_STAT_ITEMS];
935 #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
936 #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
938 #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
939 #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
941 static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
943 return pgdat->node_start_pfn + pgdat->node_spanned_pages;
946 static inline bool pgdat_is_empty(pg_data_t *pgdat)
948 return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
951 #include <linux/memory_hotplug.h>
953 void build_all_zonelists(pg_data_t *pgdat);
954 void wakeup_kswapd(struct zone *zone, gfp_t gfp_mask, int order,
955 enum zone_type highest_zoneidx);
956 bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
957 int highest_zoneidx, unsigned int alloc_flags,
959 bool zone_watermark_ok(struct zone *z, unsigned int order,
960 unsigned long mark, int highest_zoneidx,
961 unsigned int alloc_flags);
962 bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
963 unsigned long mark, int highest_zoneidx);
965 * Memory initialization context, use to differentiate memory added by
966 * the platform statically or via memory hotplug interface.
968 enum meminit_context {
973 extern void init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
976 extern void lruvec_init(struct lruvec *lruvec);
978 static inline struct pglist_data *lruvec_pgdat(struct lruvec *lruvec)
981 return lruvec->pgdat;
983 return container_of(lruvec, struct pglist_data, __lruvec);
987 #ifdef CONFIG_HAVE_MEMORYLESS_NODES
988 int local_memory_node(int node_id);
990 static inline int local_memory_node(int node_id) { return node_id; };
994 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
996 #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
998 #ifdef CONFIG_ZONE_DEVICE
999 static inline bool zone_is_zone_device(struct zone *zone)
1001 return zone_idx(zone) == ZONE_DEVICE;
1004 static inline bool zone_is_zone_device(struct zone *zone)
1011 * Returns true if a zone has pages managed by the buddy allocator.
1012 * All the reclaim decisions have to use this function rather than
1013 * populated_zone(). If the whole zone is reserved then we can easily
1014 * end up with populated_zone() && !managed_zone().
1016 static inline bool managed_zone(struct zone *zone)
1018 return zone_managed_pages(zone);
1021 /* Returns true if a zone has memory */
1022 static inline bool populated_zone(struct zone *zone)
1024 return zone->present_pages;
1028 static inline int zone_to_nid(struct zone *zone)
1033 static inline void zone_set_nid(struct zone *zone, int nid)
1038 static inline int zone_to_nid(struct zone *zone)
1043 static inline void zone_set_nid(struct zone *zone, int nid) {}
1046 extern int movable_zone;
1048 static inline int is_highmem_idx(enum zone_type idx)
1050 #ifdef CONFIG_HIGHMEM
1051 return (idx == ZONE_HIGHMEM ||
1052 (idx == ZONE_MOVABLE && movable_zone == ZONE_HIGHMEM));
1058 #ifdef CONFIG_ZONE_DMA
1059 bool has_managed_dma(void);
1061 static inline bool has_managed_dma(void)
1068 * is_highmem - helper function to quickly check if a struct zone is a
1069 * highmem zone or not. This is an attempt to keep references
1070 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
1071 * @zone: pointer to struct zone variable
1072 * Return: 1 for a highmem zone, 0 otherwise
1074 static inline int is_highmem(struct zone *zone)
1076 #ifdef CONFIG_HIGHMEM
1077 return is_highmem_idx(zone_idx(zone));
1083 /* These two functions are used to setup the per zone pages min values */
1086 int min_free_kbytes_sysctl_handler(struct ctl_table *, int, void *, size_t *,
1088 int watermark_scale_factor_sysctl_handler(struct ctl_table *, int, void *,
1089 size_t *, loff_t *);
1090 extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES];
1091 int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, void *,
1092 size_t *, loff_t *);
1093 int percpu_pagelist_high_fraction_sysctl_handler(struct ctl_table *, int,
1094 void *, size_t *, loff_t *);
1095 int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
1096 void *, size_t *, loff_t *);
1097 int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
1098 void *, size_t *, loff_t *);
1099 int numa_zonelist_order_handler(struct ctl_table *, int,
1100 void *, size_t *, loff_t *);
1101 extern int percpu_pagelist_high_fraction;
1102 extern char numa_zonelist_order[];
1103 #define NUMA_ZONELIST_ORDER_LEN 16
1107 extern struct pglist_data contig_page_data;
1108 static inline struct pglist_data *NODE_DATA(int nid)
1110 return &contig_page_data;
1113 #else /* CONFIG_NUMA */
1115 #include <asm/mmzone.h>
1117 #endif /* !CONFIG_NUMA */
1119 extern struct pglist_data *first_online_pgdat(void);
1120 extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
1121 extern struct zone *next_zone(struct zone *zone);
1124 * for_each_online_pgdat - helper macro to iterate over all online nodes
1125 * @pgdat: pointer to a pg_data_t variable
1127 #define for_each_online_pgdat(pgdat) \
1128 for (pgdat = first_online_pgdat(); \
1130 pgdat = next_online_pgdat(pgdat))
1132 * for_each_zone - helper macro to iterate over all memory zones
1133 * @zone: pointer to struct zone variable
1135 * The user only needs to declare the zone variable, for_each_zone
1138 #define for_each_zone(zone) \
1139 for (zone = (first_online_pgdat())->node_zones; \
1141 zone = next_zone(zone))
1143 #define for_each_populated_zone(zone) \
1144 for (zone = (first_online_pgdat())->node_zones; \
1146 zone = next_zone(zone)) \
1147 if (!populated_zone(zone)) \
1148 ; /* do nothing */ \
1151 static inline struct zone *zonelist_zone(struct zoneref *zoneref)
1153 return zoneref->zone;
1156 static inline int zonelist_zone_idx(struct zoneref *zoneref)
1158 return zoneref->zone_idx;
1161 static inline int zonelist_node_idx(struct zoneref *zoneref)
1163 return zone_to_nid(zoneref->zone);
1166 struct zoneref *__next_zones_zonelist(struct zoneref *z,
1167 enum zone_type highest_zoneidx,
1171 * 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
1172 * @z: The cursor used as a starting point for the search
1173 * @highest_zoneidx: The zone index of the highest zone to return
1174 * @nodes: An optional nodemask to filter the zonelist with
1176 * This function returns the next zone at or below a given zone index that is
1177 * within the allowed nodemask using a cursor as the starting point for the
1178 * search. The zoneref returned is a cursor that represents the current zone
1179 * being examined. It should be advanced by one before calling
1180 * next_zones_zonelist again.
1182 * Return: the next zone at or below highest_zoneidx within the allowed
1183 * nodemask using a cursor within a zonelist as a starting point
1185 static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,
1186 enum zone_type highest_zoneidx,
1189 if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))
1191 return __next_zones_zonelist(z, highest_zoneidx, nodes);
1195 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
1196 * @zonelist: The zonelist to search for a suitable zone
1197 * @highest_zoneidx: The zone index of the highest zone to return
1198 * @nodes: An optional nodemask to filter the zonelist with
1200 * This function returns the first zone at or below a given zone index that is
1201 * within the allowed nodemask. The zoneref returned is a cursor that can be
1202 * used to iterate the zonelist with next_zones_zonelist by advancing it by
1203 * one before calling.
1205 * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is
1206 * never NULL). This may happen either genuinely, or due to concurrent nodemask
1207 * update due to cpuset modification.
1209 * Return: Zoneref pointer for the first suitable zone found
1211 static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
1212 enum zone_type highest_zoneidx,
1215 return next_zones_zonelist(zonelist->_zonerefs,
1216 highest_zoneidx, nodes);
1220 * 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
1221 * @zone: The current zone in the iterator
1222 * @z: The current pointer within zonelist->_zonerefs being iterated
1223 * @zlist: The zonelist being iterated
1224 * @highidx: The zone index of the highest zone to return
1225 * @nodemask: Nodemask allowed by the allocator
1227 * This iterator iterates though all zones at or below a given zone index and
1228 * within a given nodemask
1230 #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1231 for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \
1233 z = next_zones_zonelist(++z, highidx, nodemask), \
1234 zone = zonelist_zone(z))
1236 #define for_next_zone_zonelist_nodemask(zone, z, highidx, nodemask) \
1237 for (zone = z->zone; \
1239 z = next_zones_zonelist(++z, highidx, nodemask), \
1240 zone = zonelist_zone(z))
1244 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1245 * @zone: The current zone in the iterator
1246 * @z: The current pointer within zonelist->zones being iterated
1247 * @zlist: The zonelist being iterated
1248 * @highidx: The zone index of the highest zone to return
1250 * This iterator iterates though all zones at or below a given zone index.
1252 #define for_each_zone_zonelist(zone, z, zlist, highidx) \
1253 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
1255 /* Whether the 'nodes' are all movable nodes */
1256 static inline bool movable_only_nodes(nodemask_t *nodes)
1258 struct zonelist *zonelist;
1262 if (nodes_empty(*nodes))
1266 * We can chose arbitrary node from the nodemask to get a
1267 * zonelist as they are interlinked. We just need to find
1268 * at least one zone that can satisfy kernel allocations.
1270 nid = first_node(*nodes);
1271 zonelist = &NODE_DATA(nid)->node_zonelists[ZONELIST_FALLBACK];
1272 z = first_zones_zonelist(zonelist, ZONE_NORMAL, nodes);
1273 return (!z->zone) ? true : false;
1277 #ifdef CONFIG_SPARSEMEM
1278 #include <asm/sparsemem.h>
1281 #ifdef CONFIG_FLATMEM
1282 #define pfn_to_nid(pfn) (0)
1285 #ifdef CONFIG_SPARSEMEM
1288 * PA_SECTION_SHIFT physical address to/from section number
1289 * PFN_SECTION_SHIFT pfn to/from section number
1291 #define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1292 #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1294 #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1296 #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1297 #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1299 #define SECTION_BLOCKFLAGS_BITS \
1300 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
1302 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1303 #error Allocator MAX_ORDER exceeds SECTION_SIZE
1306 static inline unsigned long pfn_to_section_nr(unsigned long pfn)
1308 return pfn >> PFN_SECTION_SHIFT;
1310 static inline unsigned long section_nr_to_pfn(unsigned long sec)
1312 return sec << PFN_SECTION_SHIFT;
1315 #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1316 #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1318 #define SUBSECTION_SHIFT 21
1319 #define SUBSECTION_SIZE (1UL << SUBSECTION_SHIFT)
1321 #define PFN_SUBSECTION_SHIFT (SUBSECTION_SHIFT - PAGE_SHIFT)
1322 #define PAGES_PER_SUBSECTION (1UL << PFN_SUBSECTION_SHIFT)
1323 #define PAGE_SUBSECTION_MASK (~(PAGES_PER_SUBSECTION-1))
1325 #if SUBSECTION_SHIFT > SECTION_SIZE_BITS
1326 #error Subsection size exceeds section size
1328 #define SUBSECTIONS_PER_SECTION (1UL << (SECTION_SIZE_BITS - SUBSECTION_SHIFT))
1331 #define SUBSECTION_ALIGN_UP(pfn) ALIGN((pfn), PAGES_PER_SUBSECTION)
1332 #define SUBSECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SUBSECTION_MASK)
1334 struct mem_section_usage {
1335 #ifdef CONFIG_SPARSEMEM_VMEMMAP
1336 DECLARE_BITMAP(subsection_map, SUBSECTIONS_PER_SECTION);
1338 /* See declaration of similar field in struct zone */
1339 unsigned long pageblock_flags[0];
1342 void subsection_map_init(unsigned long pfn, unsigned long nr_pages);
1346 struct mem_section {
1348 * This is, logically, a pointer to an array of struct
1349 * pages. However, it is stored with some other magic.
1350 * (see sparse.c::sparse_init_one_section())
1352 * Additionally during early boot we encode node id of
1353 * the location of the section here to guide allocation.
1354 * (see sparse.c::memory_present())
1356 * Making it a UL at least makes someone do a cast
1357 * before using it wrong.
1359 unsigned long section_mem_map;
1361 struct mem_section_usage *usage;
1362 #ifdef CONFIG_PAGE_EXTENSION
1364 * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1365 * section. (see page_ext.h about this.)
1367 struct page_ext *page_ext;
1371 * WARNING: mem_section must be a power-of-2 in size for the
1372 * calculation and use of SECTION_ROOT_MASK to make sense.
1376 #ifdef CONFIG_SPARSEMEM_EXTREME
1377 #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
1379 #define SECTIONS_PER_ROOT 1
1382 #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
1383 #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
1384 #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1386 #ifdef CONFIG_SPARSEMEM_EXTREME
1387 extern struct mem_section **mem_section;
1389 extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1392 static inline unsigned long *section_to_usemap(struct mem_section *ms)
1394 return ms->usage->pageblock_flags;
1397 static inline struct mem_section *__nr_to_section(unsigned long nr)
1399 #ifdef CONFIG_SPARSEMEM_EXTREME
1403 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1405 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
1407 extern size_t mem_section_usage_size(void);
1410 * We use the lower bits of the mem_map pointer to store
1411 * a little bit of information. The pointer is calculated
1412 * as mem_map - section_nr_to_pfn(pnum). The result is
1413 * aligned to the minimum alignment of the two values:
1414 * 1. All mem_map arrays are page-aligned.
1415 * 2. section_nr_to_pfn() always clears PFN_SECTION_SHIFT
1416 * lowest bits. PFN_SECTION_SHIFT is arch-specific
1417 * (equal SECTION_SIZE_BITS - PAGE_SHIFT), and the
1418 * worst combination is powerpc with 256k pages,
1419 * which results in PFN_SECTION_SHIFT equal 6.
1420 * To sum it up, at least 6 bits are available.
1422 #define SECTION_MARKED_PRESENT (1UL<<0)
1423 #define SECTION_HAS_MEM_MAP (1UL<<1)
1424 #define SECTION_IS_ONLINE (1UL<<2)
1425 #define SECTION_IS_EARLY (1UL<<3)
1426 #define SECTION_TAINT_ZONE_DEVICE (1UL<<4)
1427 #define SECTION_MAP_LAST_BIT (1UL<<5)
1428 #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
1429 #define SECTION_NID_SHIFT 6
1431 static inline struct page *__section_mem_map_addr(struct mem_section *section)
1433 unsigned long map = section->section_mem_map;
1434 map &= SECTION_MAP_MASK;
1435 return (struct page *)map;
1438 static inline int present_section(struct mem_section *section)
1440 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
1443 static inline int present_section_nr(unsigned long nr)
1445 return present_section(__nr_to_section(nr));
1448 static inline int valid_section(struct mem_section *section)
1450 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
1453 static inline int early_section(struct mem_section *section)
1455 return (section && (section->section_mem_map & SECTION_IS_EARLY));
1458 static inline int valid_section_nr(unsigned long nr)
1460 return valid_section(__nr_to_section(nr));
1463 static inline int online_section(struct mem_section *section)
1465 return (section && (section->section_mem_map & SECTION_IS_ONLINE));
1468 static inline int online_device_section(struct mem_section *section)
1470 unsigned long flags = SECTION_IS_ONLINE | SECTION_TAINT_ZONE_DEVICE;
1472 return section && ((section->section_mem_map & flags) == flags);
1475 static inline int online_section_nr(unsigned long nr)
1477 return online_section(__nr_to_section(nr));
1480 #ifdef CONFIG_MEMORY_HOTPLUG
1481 void online_mem_sections(unsigned long start_pfn, unsigned long end_pfn);
1482 void offline_mem_sections(unsigned long start_pfn, unsigned long end_pfn);
1485 static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1487 return __nr_to_section(pfn_to_section_nr(pfn));
1490 extern unsigned long __highest_present_section_nr;
1492 static inline int subsection_map_index(unsigned long pfn)
1494 return (pfn & ~(PAGE_SECTION_MASK)) / PAGES_PER_SUBSECTION;
1497 #ifdef CONFIG_SPARSEMEM_VMEMMAP
1498 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
1500 int idx = subsection_map_index(pfn);
1502 return test_bit(idx, ms->usage->subsection_map);
1505 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
1511 #ifndef CONFIG_HAVE_ARCH_PFN_VALID
1513 * pfn_valid - check if there is a valid memory map entry for a PFN
1514 * @pfn: the page frame number to check
1516 * Check if there is a valid memory map entry aka struct page for the @pfn.
1517 * Note, that availability of the memory map entry does not imply that
1518 * there is actual usable memory at that @pfn. The struct page may
1519 * represent a hole or an unusable page frame.
1521 * Return: 1 for PFNs that have memory map entries and 0 otherwise
1523 static inline int pfn_valid(unsigned long pfn)
1525 struct mem_section *ms;
1528 * Ensure the upper PAGE_SHIFT bits are clear in the
1529 * pfn. Else it might lead to false positives when
1530 * some of the upper bits are set, but the lower bits
1531 * match a valid pfn.
1533 if (PHYS_PFN(PFN_PHYS(pfn)) != pfn)
1536 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1538 ms = __pfn_to_section(pfn);
1539 if (!valid_section(ms))
1542 * Traditionally early sections always returned pfn_valid() for
1543 * the entire section-sized span.
1545 return early_section(ms) || pfn_section_valid(ms, pfn);
1549 static inline int pfn_in_present_section(unsigned long pfn)
1551 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1553 return present_section(__pfn_to_section(pfn));
1556 static inline unsigned long next_present_section_nr(unsigned long section_nr)
1558 while (++section_nr <= __highest_present_section_nr) {
1559 if (present_section_nr(section_nr))
1567 * These are _only_ used during initialisation, therefore they
1568 * can use __initdata ... They could have names to indicate
1572 #define pfn_to_nid(pfn) \
1574 unsigned long __pfn_to_nid_pfn = (pfn); \
1575 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1578 #define pfn_to_nid(pfn) (0)
1581 void sparse_init(void);
1583 #define sparse_init() do {} while (0)
1584 #define sparse_index_init(_sec, _nid) do {} while (0)
1585 #define pfn_in_present_section pfn_valid
1586 #define subsection_map_init(_pfn, _nr_pages) do {} while (0)
1587 #endif /* CONFIG_SPARSEMEM */
1589 #endif /* !__GENERATING_BOUNDS.H */
1590 #endif /* !__ASSEMBLY__ */
1591 #endif /* _LINUX_MMZONE_H */