]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/mmzone.h
cpumask: reduce stack usage in SD_x_INIT initializers
[mirror_ubuntu-bionic-kernel.git] / include / linux / mmzone.h
CommitLineData
1da177e4
LT
1#ifndef _LINUX_MMZONE_H
2#define _LINUX_MMZONE_H
3
4#ifdef __KERNEL__
5#ifndef __ASSEMBLY__
6
1da177e4
LT
7#include <linux/spinlock.h>
8#include <linux/list.h>
9#include <linux/wait.h>
e815af95 10#include <linux/bitops.h>
1da177e4
LT
11#include <linux/cache.h>
12#include <linux/threads.h>
13#include <linux/numa.h>
14#include <linux/init.h>
bdc8cb98 15#include <linux/seqlock.h>
8357f869 16#include <linux/nodemask.h>
835c134e 17#include <linux/pageblock-flags.h>
1da177e4 18#include <asm/atomic.h>
93ff66bf 19#include <asm/page.h>
1da177e4
LT
20
21/* Free memory management - zoned buddy allocator. */
22#ifndef CONFIG_FORCE_MAX_ZONEORDER
23#define MAX_ORDER 11
24#else
25#define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
26#endif
e984bb43 27#define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
1da177e4 28
5ad333eb
AW
29/*
30 * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed
31 * costly to service. That is between allocation orders which should
32 * coelesce naturally under reasonable reclaim pressure and those which
33 * will not.
34 */
35#define PAGE_ALLOC_COSTLY_ORDER 3
36
b2a0ac88 37#define MIGRATE_UNMOVABLE 0
e12ba74d
MG
38#define MIGRATE_RECLAIMABLE 1
39#define MIGRATE_MOVABLE 2
64c5e135 40#define MIGRATE_RESERVE 3
a5d76b54
KH
41#define MIGRATE_ISOLATE 4 /* can't allocate from here */
42#define MIGRATE_TYPES 5
b2a0ac88
MG
43
44#define for_each_migratetype_order(order, type) \
45 for (order = 0; order < MAX_ORDER; order++) \
46 for (type = 0; type < MIGRATE_TYPES; type++)
47
467c996c
MG
48extern int page_group_by_mobility_disabled;
49
50static inline int get_pageblock_migratetype(struct page *page)
51{
52 if (unlikely(page_group_by_mobility_disabled))
53 return MIGRATE_UNMOVABLE;
54
55 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
56}
57
1da177e4 58struct free_area {
b2a0ac88 59 struct list_head free_list[MIGRATE_TYPES];
1da177e4
LT
60 unsigned long nr_free;
61};
62
63struct pglist_data;
64
65/*
66 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
67 * So add a wild amount of padding here to ensure that they fall into separate
68 * cachelines. There are very few zone structures in the machine, so space
69 * consumption is not a concern here.
70 */
71#if defined(CONFIG_SMP)
72struct zone_padding {
73 char x[0];
22fc6ecc 74} ____cacheline_internodealigned_in_smp;
1da177e4
LT
75#define ZONE_PADDING(name) struct zone_padding name;
76#else
77#define ZONE_PADDING(name)
78#endif
79
2244b95a 80enum zone_stat_item {
51ed4491 81 /* First 128 byte cacheline (assuming 64 bit words) */
d23ad423 82 NR_FREE_PAGES,
c8785385
CL
83 NR_INACTIVE,
84 NR_ACTIVE,
f3dbd344
CL
85 NR_ANON_PAGES, /* Mapped anonymous pages */
86 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
65ba55f5 87 only modified from process context */
347ce434 88 NR_FILE_PAGES,
b1e7a8fd 89 NR_FILE_DIRTY,
ce866b34 90 NR_WRITEBACK,
51ed4491
CL
91 /* Second 128 byte cacheline */
92 NR_SLAB_RECLAIMABLE,
93 NR_SLAB_UNRECLAIMABLE,
94 NR_PAGETABLE, /* used for pagetables */
fd39fc85 95 NR_UNSTABLE_NFS, /* NFS unstable pages */
d2c5e30c 96 NR_BOUNCE,
e129b5c2 97 NR_VMSCAN_WRITE,
ca889e6c
CL
98#ifdef CONFIG_NUMA
99 NUMA_HIT, /* allocated in intended node */
100 NUMA_MISS, /* allocated in non intended node */
101 NUMA_FOREIGN, /* was intended here, hit elsewhere */
102 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
103 NUMA_LOCAL, /* allocation from local node */
104 NUMA_OTHER, /* allocation from other node */
105#endif
2244b95a
CL
106 NR_VM_ZONE_STAT_ITEMS };
107
1da177e4
LT
108struct per_cpu_pages {
109 int count; /* number of pages in the list */
1da177e4
LT
110 int high; /* high watermark, emptying needed */
111 int batch; /* chunk size for buddy add/remove */
112 struct list_head list; /* the list of pages */
113};
114
115struct per_cpu_pageset {
3dfa5721 116 struct per_cpu_pages pcp;
4037d452
CL
117#ifdef CONFIG_NUMA
118 s8 expire;
119#endif
2244b95a 120#ifdef CONFIG_SMP
df9ecaba 121 s8 stat_threshold;
2244b95a
CL
122 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
123#endif
1da177e4
LT
124} ____cacheline_aligned_in_smp;
125
e7c8d5c9
CL
126#ifdef CONFIG_NUMA
127#define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
128#else
129#define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
130#endif
131
2f1b6248 132enum zone_type {
4b51d669 133#ifdef CONFIG_ZONE_DMA
2f1b6248
CL
134 /*
135 * ZONE_DMA is used when there are devices that are not able
136 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
137 * carve out the portion of memory that is needed for these devices.
138 * The range is arch specific.
139 *
140 * Some examples
141 *
142 * Architecture Limit
143 * ---------------------------
144 * parisc, ia64, sparc <4G
145 * s390 <2G
2f1b6248
CL
146 * arm Various
147 * alpha Unlimited or 0-16MB.
148 *
149 * i386, x86_64 and multiple other arches
150 * <16M.
151 */
152 ZONE_DMA,
4b51d669 153#endif
fb0e7942 154#ifdef CONFIG_ZONE_DMA32
2f1b6248
CL
155 /*
156 * x86_64 needs two ZONE_DMAs because it supports devices that are
157 * only able to do DMA to the lower 16M but also 32 bit devices that
158 * can only do DMA areas below 4G.
159 */
160 ZONE_DMA32,
fb0e7942 161#endif
2f1b6248
CL
162 /*
163 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
164 * performed on pages in ZONE_NORMAL if the DMA devices support
165 * transfers to all addressable memory.
166 */
167 ZONE_NORMAL,
e53ef38d 168#ifdef CONFIG_HIGHMEM
2f1b6248
CL
169 /*
170 * A memory area that is only addressable by the kernel through
171 * mapping portions into its own address space. This is for example
172 * used by i386 to allow the kernel to address the memory beyond
173 * 900MB. The kernel will set up special mappings (page
174 * table entries on i386) for each page that the kernel needs to
175 * access.
176 */
177 ZONE_HIGHMEM,
e53ef38d 178#endif
2a1e274a 179 ZONE_MOVABLE,
2f1b6248
CL
180 MAX_NR_ZONES
181};
1da177e4 182
1da177e4
LT
183/*
184 * When a memory allocation must conform to specific limitations (such
185 * as being suitable for DMA) the caller will pass in hints to the
186 * allocator in the gfp_mask, in the zone modifier bits. These bits
187 * are used to select a priority ordered list of memory zones which
19655d34 188 * match the requested limits. See gfp_zone() in include/linux/gfp.h
1da177e4 189 */
fb0e7942 190
4b51d669
CL
191/*
192 * Count the active zones. Note that the use of defined(X) outside
193 * #if and family is not necessarily defined so ensure we cannot use
194 * it later. Use __ZONE_COUNT to work out how many shift bits we need.
195 */
196#define __ZONE_COUNT ( \
197 defined(CONFIG_ZONE_DMA) \
198 + defined(CONFIG_ZONE_DMA32) \
199 + 1 \
200 + defined(CONFIG_HIGHMEM) \
2a1e274a 201 + 1 \
4b51d669
CL
202)
203#if __ZONE_COUNT < 2
204#define ZONES_SHIFT 0
205#elif __ZONE_COUNT <= 2
19655d34 206#define ZONES_SHIFT 1
4b51d669 207#elif __ZONE_COUNT <= 4
19655d34 208#define ZONES_SHIFT 2
4b51d669
CL
209#else
210#error ZONES_SHIFT -- too many zones configured adjust calculation
fb0e7942 211#endif
4b51d669 212#undef __ZONE_COUNT
1da177e4 213
1da177e4
LT
214struct zone {
215 /* Fields commonly accessed by the page allocator */
1da177e4
LT
216 unsigned long pages_min, pages_low, pages_high;
217 /*
218 * We don't know if the memory that we're going to allocate will be freeable
219 * or/and it will be released eventually, so to avoid totally wasting several
220 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
221 * to run OOM on the lower zones despite there's tons of freeable ram
222 * on the higher zones). This array is recalculated at runtime if the
223 * sysctl_lowmem_reserve_ratio sysctl changes.
224 */
225 unsigned long lowmem_reserve[MAX_NR_ZONES];
226
e7c8d5c9 227#ifdef CONFIG_NUMA
d5f541ed 228 int node;
9614634f
CL
229 /*
230 * zone reclaim becomes active if more unmapped pages exist.
231 */
8417bba4 232 unsigned long min_unmapped_pages;
0ff38490 233 unsigned long min_slab_pages;
e7c8d5c9
CL
234 struct per_cpu_pageset *pageset[NR_CPUS];
235#else
1da177e4 236 struct per_cpu_pageset pageset[NR_CPUS];
e7c8d5c9 237#endif
1da177e4
LT
238 /*
239 * free areas of different sizes
240 */
241 spinlock_t lock;
bdc8cb98
DH
242#ifdef CONFIG_MEMORY_HOTPLUG
243 /* see spanned/present_pages for more description */
244 seqlock_t span_seqlock;
245#endif
1da177e4
LT
246 struct free_area free_area[MAX_ORDER];
247
835c134e
MG
248#ifndef CONFIG_SPARSEMEM
249 /*
d9c23400 250 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
835c134e
MG
251 * In SPARSEMEM, this map is stored in struct mem_section
252 */
253 unsigned long *pageblock_flags;
254#endif /* CONFIG_SPARSEMEM */
255
1da177e4
LT
256
257 ZONE_PADDING(_pad1_)
258
259 /* Fields commonly accessed by the page reclaim scanner */
260 spinlock_t lru_lock;
261 struct list_head active_list;
262 struct list_head inactive_list;
263 unsigned long nr_scan_active;
264 unsigned long nr_scan_inactive;
1da177e4 265 unsigned long pages_scanned; /* since last reclaim */
e815af95 266 unsigned long flags; /* zone flags, see below */
753ee728 267
2244b95a
CL
268 /* Zone statistics */
269 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
9eeff239 270
1da177e4
LT
271 /*
272 * prev_priority holds the scanning priority for this zone. It is
273 * defined as the scanning priority at which we achieved our reclaim
274 * target at the previous try_to_free_pages() or balance_pgdat()
275 * invokation.
276 *
277 * We use prev_priority as a measure of how much stress page reclaim is
278 * under - it drives the swappiness decision: whether to unmap mapped
279 * pages.
280 *
3bb1a852 281 * Access to both this field is quite racy even on uniprocessor. But
1da177e4
LT
282 * it is expected to average out OK.
283 */
1da177e4
LT
284 int prev_priority;
285
286
287 ZONE_PADDING(_pad2_)
288 /* Rarely used or read-mostly fields */
289
290 /*
291 * wait_table -- the array holding the hash table
02b694de 292 * wait_table_hash_nr_entries -- the size of the hash table array
1da177e4
LT
293 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
294 *
295 * The purpose of all these is to keep track of the people
296 * waiting for a page to become available and make them
297 * runnable again when possible. The trouble is that this
298 * consumes a lot of space, especially when so few things
299 * wait on pages at a given time. So instead of using
300 * per-page waitqueues, we use a waitqueue hash table.
301 *
302 * The bucket discipline is to sleep on the same queue when
303 * colliding and wake all in that wait queue when removing.
304 * When something wakes, it must check to be sure its page is
305 * truly available, a la thundering herd. The cost of a
306 * collision is great, but given the expected load of the
307 * table, they should be so rare as to be outweighed by the
308 * benefits from the saved space.
309 *
310 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
311 * primary users of these fields, and in mm/page_alloc.c
312 * free_area_init_core() performs the initialization of them.
313 */
314 wait_queue_head_t * wait_table;
02b694de 315 unsigned long wait_table_hash_nr_entries;
1da177e4
LT
316 unsigned long wait_table_bits;
317
318 /*
319 * Discontig memory support fields.
320 */
321 struct pglist_data *zone_pgdat;
1da177e4
LT
322 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
323 unsigned long zone_start_pfn;
324
bdc8cb98
DH
325 /*
326 * zone_start_pfn, spanned_pages and present_pages are all
327 * protected by span_seqlock. It is a seqlock because it has
328 * to be read outside of zone->lock, and it is done in the main
329 * allocator path. But, it is written quite infrequently.
330 *
331 * The lock is declared along with zone->lock because it is
332 * frequently read in proximity to zone->lock. It's good to
333 * give them a chance of being in the same cacheline.
334 */
1da177e4
LT
335 unsigned long spanned_pages; /* total size, including holes */
336 unsigned long present_pages; /* amount of memory (excluding holes) */
337
338 /*
339 * rarely used fields:
340 */
15ad7cdc 341 const char *name;
22fc6ecc 342} ____cacheline_internodealigned_in_smp;
1da177e4 343
e815af95
DR
344typedef enum {
345 ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */
346 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
098d7f12 347 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
e815af95
DR
348} zone_flags_t;
349
350static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
351{
352 set_bit(flag, &zone->flags);
353}
d773ed6b
DR
354
355static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
356{
357 return test_and_set_bit(flag, &zone->flags);
358}
359
e815af95
DR
360static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
361{
362 clear_bit(flag, &zone->flags);
363}
364
365static inline int zone_is_all_unreclaimable(const struct zone *zone)
366{
367 return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags);
368}
d773ed6b 369
e815af95
DR
370static inline int zone_is_reclaim_locked(const struct zone *zone)
371{
372 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
373}
d773ed6b 374
098d7f12
DR
375static inline int zone_is_oom_locked(const struct zone *zone)
376{
377 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
378}
e815af95 379
1da177e4
LT
380/*
381 * The "priority" of VM scanning is how much of the queues we will scan in one
382 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
383 * queues ("queue_length >> 12") during an aging round.
384 */
385#define DEF_PRIORITY 12
386
9276b1bc
PJ
387/* Maximum number of zones on a zonelist */
388#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
389
390#ifdef CONFIG_NUMA
523b9458
CL
391
392/*
393 * The NUMA zonelists are doubled becausse we need zonelists that restrict the
394 * allocations to a single node for GFP_THISNODE.
395 *
396 * [0 .. MAX_NR_ZONES -1] : Zonelists with fallback
397 * [MAZ_NR_ZONES ... MAZ_ZONELISTS -1] : No fallback (GFP_THISNODE)
398 */
399#define MAX_ZONELISTS (2 * MAX_NR_ZONES)
400
401
9276b1bc
PJ
402/*
403 * We cache key information from each zonelist for smaller cache
404 * footprint when scanning for free pages in get_page_from_freelist().
405 *
406 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
407 * up short of free memory since the last time (last_fullzone_zap)
408 * we zero'd fullzones.
409 * 2) The array z_to_n[] maps each zone in the zonelist to its node
410 * id, so that we can efficiently evaluate whether that node is
411 * set in the current tasks mems_allowed.
412 *
413 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
414 * indexed by a zones offset in the zonelist zones[] array.
415 *
416 * The get_page_from_freelist() routine does two scans. During the
417 * first scan, we skip zones whose corresponding bit in 'fullzones'
418 * is set or whose corresponding node in current->mems_allowed (which
419 * comes from cpusets) is not set. During the second scan, we bypass
420 * this zonelist_cache, to ensure we look methodically at each zone.
421 *
422 * Once per second, we zero out (zap) fullzones, forcing us to
423 * reconsider nodes that might have regained more free memory.
424 * The field last_full_zap is the time we last zapped fullzones.
425 *
426 * This mechanism reduces the amount of time we waste repeatedly
427 * reexaming zones for free memory when they just came up low on
428 * memory momentarilly ago.
429 *
430 * The zonelist_cache struct members logically belong in struct
431 * zonelist. However, the mempolicy zonelists constructed for
432 * MPOL_BIND are intentionally variable length (and usually much
433 * shorter). A general purpose mechanism for handling structs with
434 * multiple variable length members is more mechanism than we want
435 * here. We resort to some special case hackery instead.
436 *
437 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
438 * part because they are shorter), so we put the fixed length stuff
439 * at the front of the zonelist struct, ending in a variable length
440 * zones[], as is needed by MPOL_BIND.
441 *
442 * Then we put the optional zonelist cache on the end of the zonelist
443 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
444 * the fixed length portion at the front of the struct. This pointer
445 * both enables us to find the zonelist cache, and in the case of
446 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
447 * to know that the zonelist cache is not there.
448 *
449 * The end result is that struct zonelists come in two flavors:
450 * 1) The full, fixed length version, shown below, and
451 * 2) The custom zonelists for MPOL_BIND.
452 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
453 *
454 * Even though there may be multiple CPU cores on a node modifying
455 * fullzones or last_full_zap in the same zonelist_cache at the same
456 * time, we don't lock it. This is just hint data - if it is wrong now
457 * and then, the allocator will still function, perhaps a bit slower.
458 */
459
460
461struct zonelist_cache {
9276b1bc 462 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
7253f4ef 463 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
9276b1bc
PJ
464 unsigned long last_full_zap; /* when last zap'd (jiffies) */
465};
466#else
523b9458 467#define MAX_ZONELISTS MAX_NR_ZONES
9276b1bc
PJ
468struct zonelist_cache;
469#endif
470
1da177e4
LT
471/*
472 * One allocation request operates on a zonelist. A zonelist
473 * is a list of zones, the first one is the 'goal' of the
474 * allocation, the other zones are fallback zones, in decreasing
475 * priority.
476 *
9276b1bc
PJ
477 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
478 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
1da177e4 479 */
9276b1bc 480
1da177e4 481struct zonelist {
9276b1bc
PJ
482 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
483 struct zone *zones[MAX_ZONES_PER_ZONELIST + 1]; // NULL delimited
484#ifdef CONFIG_NUMA
485 struct zonelist_cache zlcache; // optional ...
486#endif
1da177e4
LT
487};
488
b377fd39
MG
489#ifdef CONFIG_NUMA
490/*
491 * Only custom zonelists like MPOL_BIND need to be filtered as part of
492 * policies. As described in the comment for struct zonelist_cache, these
493 * zonelists will not have a zlcache so zlcache_ptr will not be set. Use
494 * that to determine if the zonelists needs to be filtered or not.
495 */
496static inline int alloc_should_filter_zonelist(struct zonelist *zonelist)
497{
498 return !zonelist->zlcache_ptr;
499}
500#else
501static inline int alloc_should_filter_zonelist(struct zonelist *zonelist)
502{
503 return 0;
504}
505#endif /* CONFIG_NUMA */
506
c713216d
MG
507#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
508struct node_active_region {
509 unsigned long start_pfn;
510 unsigned long end_pfn;
511 int nid;
512};
513#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1da177e4 514
5b99cd0e
HC
515#ifndef CONFIG_DISCONTIGMEM
516/* The array of struct pages - for discontigmem use pgdat->lmem_map */
517extern struct page *mem_map;
518#endif
519
1da177e4
LT
520/*
521 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
522 * (mostly NUMA machines?) to denote a higher-level memory zone than the
523 * zone denotes.
524 *
525 * On NUMA machines, each NUMA node would have a pg_data_t to describe
526 * it's memory layout.
527 *
528 * Memory statistics and page replacement data structures are maintained on a
529 * per-zone basis.
530 */
531struct bootmem_data;
532typedef struct pglist_data {
533 struct zone node_zones[MAX_NR_ZONES];
523b9458 534 struct zonelist node_zonelists[MAX_ZONELISTS];
1da177e4 535 int nr_zones;
d41dee36 536#ifdef CONFIG_FLAT_NODE_MEM_MAP
1da177e4 537 struct page *node_mem_map;
d41dee36 538#endif
1da177e4 539 struct bootmem_data *bdata;
208d54e5
DH
540#ifdef CONFIG_MEMORY_HOTPLUG
541 /*
542 * Must be held any time you expect node_start_pfn, node_present_pages
543 * or node_spanned_pages stay constant. Holding this will also
544 * guarantee that any pfn_valid() stays that way.
545 *
546 * Nests above zone->lock and zone->size_seqlock.
547 */
548 spinlock_t node_size_lock;
549#endif
1da177e4
LT
550 unsigned long node_start_pfn;
551 unsigned long node_present_pages; /* total number of physical pages */
552 unsigned long node_spanned_pages; /* total size of physical page
553 range, including holes */
554 int node_id;
1da177e4
LT
555 wait_queue_head_t kswapd_wait;
556 struct task_struct *kswapd;
557 int kswapd_max_order;
558} pg_data_t;
559
560#define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
561#define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
d41dee36 562#ifdef CONFIG_FLAT_NODE_MEM_MAP
408fde81 563#define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
d41dee36
AW
564#else
565#define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
566#endif
408fde81 567#define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
1da177e4 568
208d54e5
DH
569#include <linux/memory_hotplug.h>
570
1da177e4
LT
571void get_zone_counts(unsigned long *active, unsigned long *inactive,
572 unsigned long *free);
573void build_all_zonelists(void);
574void wakeup_kswapd(struct zone *zone, int order);
575int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
7fb1d9fc 576 int classzone_idx, int alloc_flags);
a2f3aa02
DH
577enum memmap_context {
578 MEMMAP_EARLY,
579 MEMMAP_HOTPLUG,
580};
718127cc 581extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
a2f3aa02
DH
582 unsigned long size,
583 enum memmap_context context);
718127cc 584
1da177e4
LT
585#ifdef CONFIG_HAVE_MEMORY_PRESENT
586void memory_present(int nid, unsigned long start, unsigned long end);
587#else
588static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
589#endif
590
591#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
592unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
593#endif
594
595/*
596 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
597 */
598#define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
599
f3fe6512
CK
600static inline int populated_zone(struct zone *zone)
601{
602 return (!!zone->present_pages);
603}
604
2a1e274a
MG
605extern int movable_zone;
606
607static inline int zone_movable_is_highmem(void)
608{
609#if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP)
610 return movable_zone == ZONE_HIGHMEM;
611#else
612 return 0;
613#endif
614}
615
2f1b6248 616static inline int is_highmem_idx(enum zone_type idx)
1da177e4 617{
e53ef38d 618#ifdef CONFIG_HIGHMEM
2a1e274a
MG
619 return (idx == ZONE_HIGHMEM ||
620 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
e53ef38d
CL
621#else
622 return 0;
623#endif
1da177e4
LT
624}
625
2f1b6248 626static inline int is_normal_idx(enum zone_type idx)
1da177e4
LT
627{
628 return (idx == ZONE_NORMAL);
629}
9328b8fa 630
1da177e4
LT
631/**
632 * is_highmem - helper function to quickly check if a struct zone is a
633 * highmem zone or not. This is an attempt to keep references
634 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
635 * @zone - pointer to struct zone variable
636 */
637static inline int is_highmem(struct zone *zone)
638{
e53ef38d 639#ifdef CONFIG_HIGHMEM
2a1e274a
MG
640 int zone_idx = zone - zone->zone_pgdat->node_zones;
641 return zone_idx == ZONE_HIGHMEM ||
642 (zone_idx == ZONE_MOVABLE && zone_movable_is_highmem());
e53ef38d
CL
643#else
644 return 0;
645#endif
1da177e4
LT
646}
647
648static inline int is_normal(struct zone *zone)
649{
650 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
651}
652
9328b8fa
NP
653static inline int is_dma32(struct zone *zone)
654{
fb0e7942 655#ifdef CONFIG_ZONE_DMA32
9328b8fa 656 return zone == zone->zone_pgdat->node_zones + ZONE_DMA32;
fb0e7942
CL
657#else
658 return 0;
659#endif
9328b8fa
NP
660}
661
662static inline int is_dma(struct zone *zone)
663{
4b51d669 664#ifdef CONFIG_ZONE_DMA
9328b8fa 665 return zone == zone->zone_pgdat->node_zones + ZONE_DMA;
4b51d669
CL
666#else
667 return 0;
668#endif
9328b8fa
NP
669}
670
1da177e4
LT
671/* These two functions are used to setup the per zone pages min values */
672struct ctl_table;
673struct file;
674int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
675 void __user *, size_t *, loff_t *);
676extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
677int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
678 void __user *, size_t *, loff_t *);
8ad4b1fb
RS
679int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *,
680 void __user *, size_t *, loff_t *);
9614634f
CL
681int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
682 struct file *, void __user *, size_t *, loff_t *);
0ff38490
CL
683int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
684 struct file *, void __user *, size_t *, loff_t *);
1da177e4 685
f0c0b2b8
KH
686extern int numa_zonelist_order_handler(struct ctl_table *, int,
687 struct file *, void __user *, size_t *, loff_t *);
688extern char numa_zonelist_order[];
689#define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
690
1da177e4
LT
691#include <linux/topology.h>
692/* Returns the number of the current Node. */
69d81fcd 693#ifndef numa_node_id
39c715b7 694#define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
69d81fcd 695#endif
1da177e4 696
93b7504e 697#ifndef CONFIG_NEED_MULTIPLE_NODES
1da177e4
LT
698
699extern struct pglist_data contig_page_data;
700#define NODE_DATA(nid) (&contig_page_data)
701#define NODE_MEM_MAP(nid) mem_map
702#define MAX_NODES_SHIFT 1
1da177e4 703
93b7504e 704#else /* CONFIG_NEED_MULTIPLE_NODES */
1da177e4
LT
705
706#include <asm/mmzone.h>
707
93b7504e 708#endif /* !CONFIG_NEED_MULTIPLE_NODES */
348f8b6c 709
95144c78
KH
710extern struct pglist_data *first_online_pgdat(void);
711extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
712extern struct zone *next_zone(struct zone *zone);
8357f869
KH
713
714/**
715 * for_each_pgdat - helper macro to iterate over all nodes
716 * @pgdat - pointer to a pg_data_t variable
717 */
718#define for_each_online_pgdat(pgdat) \
719 for (pgdat = first_online_pgdat(); \
720 pgdat; \
721 pgdat = next_online_pgdat(pgdat))
8357f869
KH
722/**
723 * for_each_zone - helper macro to iterate over all memory zones
724 * @zone - pointer to struct zone variable
725 *
726 * The user only needs to declare the zone variable, for_each_zone
727 * fills it in.
728 */
729#define for_each_zone(zone) \
730 for (zone = (first_online_pgdat())->node_zones; \
731 zone; \
732 zone = next_zone(zone))
733
d41dee36
AW
734#ifdef CONFIG_SPARSEMEM
735#include <asm/sparsemem.h>
736#endif
737
07808b74 738#if BITS_PER_LONG == 32
1da177e4 739/*
a2f1b424
AK
740 * with 32 bit page->flags field, we reserve 9 bits for node/zone info.
741 * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes.
1da177e4 742 */
a2f1b424 743#define FLAGS_RESERVED 9
348f8b6c 744
1da177e4
LT
745#elif BITS_PER_LONG == 64
746/*
747 * with 64 bit flags field, there's plenty of room.
748 */
348f8b6c 749#define FLAGS_RESERVED 32
1da177e4 750
348f8b6c 751#else
1da177e4 752
348f8b6c 753#error BITS_PER_LONG not defined
1da177e4 754
1da177e4
LT
755#endif
756
c713216d
MG
757#if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
758 !defined(CONFIG_ARCH_POPULATES_NODE_MAP)
b159d43f
AW
759#define early_pfn_to_nid(nid) (0UL)
760#endif
761
2bdaf115
AW
762#ifdef CONFIG_FLATMEM
763#define pfn_to_nid(pfn) (0)
764#endif
765
d41dee36
AW
766#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
767#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
768
769#ifdef CONFIG_SPARSEMEM
770
771/*
772 * SECTION_SHIFT #bits space required to store a section #
773 *
774 * PA_SECTION_SHIFT physical address to/from section number
775 * PFN_SECTION_SHIFT pfn to/from section number
776 */
777#define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
778
779#define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
780#define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
781
782#define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
783
784#define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
785#define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
786
835c134e 787#define SECTION_BLOCKFLAGS_BITS \
d9c23400 788 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
835c134e 789
d41dee36
AW
790#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
791#error Allocator MAX_ORDER exceeds SECTION_SIZE
792#endif
793
794struct page;
795struct mem_section {
29751f69
AW
796 /*
797 * This is, logically, a pointer to an array of struct
798 * pages. However, it is stored with some other magic.
799 * (see sparse.c::sparse_init_one_section())
800 *
30c253e6
AW
801 * Additionally during early boot we encode node id of
802 * the location of the section here to guide allocation.
803 * (see sparse.c::memory_present())
804 *
29751f69
AW
805 * Making it a UL at least makes someone do a cast
806 * before using it wrong.
807 */
808 unsigned long section_mem_map;
5c0e3066
MG
809
810 /* See declaration of similar field in struct zone */
811 unsigned long *pageblock_flags;
d41dee36
AW
812};
813
3e347261
BP
814#ifdef CONFIG_SPARSEMEM_EXTREME
815#define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
816#else
817#define SECTIONS_PER_ROOT 1
818#endif
802f192e 819
3e347261
BP
820#define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
821#define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
822#define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
802f192e 823
3e347261
BP
824#ifdef CONFIG_SPARSEMEM_EXTREME
825extern struct mem_section *mem_section[NR_SECTION_ROOTS];
802f192e 826#else
3e347261
BP
827extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
828#endif
d41dee36 829
29751f69
AW
830static inline struct mem_section *__nr_to_section(unsigned long nr)
831{
3e347261
BP
832 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
833 return NULL;
834 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
29751f69 835}
4ca644d9 836extern int __section_nr(struct mem_section* ms);
29751f69
AW
837
838/*
839 * We use the lower bits of the mem_map pointer to store
840 * a little bit of information. There should be at least
841 * 3 bits here due to 32-bit alignment.
842 */
843#define SECTION_MARKED_PRESENT (1UL<<0)
844#define SECTION_HAS_MEM_MAP (1UL<<1)
845#define SECTION_MAP_LAST_BIT (1UL<<2)
846#define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
30c253e6 847#define SECTION_NID_SHIFT 2
29751f69
AW
848
849static inline struct page *__section_mem_map_addr(struct mem_section *section)
850{
851 unsigned long map = section->section_mem_map;
852 map &= SECTION_MAP_MASK;
853 return (struct page *)map;
854}
855
540557b9 856static inline int present_section(struct mem_section *section)
29751f69 857{
802f192e 858 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
29751f69
AW
859}
860
540557b9
AW
861static inline int present_section_nr(unsigned long nr)
862{
863 return present_section(__nr_to_section(nr));
864}
865
866static inline int valid_section(struct mem_section *section)
29751f69 867{
802f192e 868 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
29751f69
AW
869}
870
871static inline int valid_section_nr(unsigned long nr)
872{
873 return valid_section(__nr_to_section(nr));
874}
875
d41dee36
AW
876static inline struct mem_section *__pfn_to_section(unsigned long pfn)
877{
29751f69 878 return __nr_to_section(pfn_to_section_nr(pfn));
d41dee36
AW
879}
880
d41dee36
AW
881static inline int pfn_valid(unsigned long pfn)
882{
883 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
884 return 0;
29751f69 885 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
d41dee36
AW
886}
887
540557b9
AW
888static inline int pfn_present(unsigned long pfn)
889{
890 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
891 return 0;
892 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
893}
894
d41dee36
AW
895/*
896 * These are _only_ used during initialisation, therefore they
897 * can use __initdata ... They could have names to indicate
898 * this restriction.
899 */
900#ifdef CONFIG_NUMA
161599ff
AW
901#define pfn_to_nid(pfn) \
902({ \
903 unsigned long __pfn_to_nid_pfn = (pfn); \
904 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
905})
2bdaf115
AW
906#else
907#define pfn_to_nid(pfn) (0)
d41dee36
AW
908#endif
909
d41dee36
AW
910#define early_pfn_valid(pfn) pfn_valid(pfn)
911void sparse_init(void);
912#else
913#define sparse_init() do {} while (0)
28ae55c9 914#define sparse_index_init(_sec, _nid) do {} while (0)
d41dee36
AW
915#endif /* CONFIG_SPARSEMEM */
916
75167957
AW
917#ifdef CONFIG_NODES_SPAN_OTHER_NODES
918#define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
919#else
920#define early_pfn_in_nid(pfn, nid) (1)
921#endif
922
d41dee36
AW
923#ifndef early_pfn_valid
924#define early_pfn_valid(pfn) (1)
925#endif
926
927void memory_present(int nid, unsigned long start, unsigned long end);
928unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
929
14e07298
AW
930/*
931 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
932 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
933 * pfn_valid_within() should be used in this case; we optimise this away
934 * when we have no holes within a MAX_ORDER_NR_PAGES block.
935 */
936#ifdef CONFIG_HOLES_IN_ZONE
937#define pfn_valid_within(pfn) pfn_valid(pfn)
938#else
939#define pfn_valid_within(pfn) (1)
940#endif
941
1da177e4
LT
942#endif /* !__ASSEMBLY__ */
943#endif /* __KERNEL__ */
944#endif /* _LINUX_MMZONE_H */