Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #ifndef _LINUX_MMZONE_H |
2 | #define _LINUX_MMZONE_H | |
3 | ||
1da177e4 | 4 | #ifndef __ASSEMBLY__ |
97965478 | 5 | #ifndef __GENERATING_BOUNDS_H |
1da177e4 | 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> |
01fc0ac1 | 18 | #include <generated/bounds.h> |
1da177e4 | 19 | #include <asm/atomic.h> |
93ff66bf | 20 | #include <asm/page.h> |
1da177e4 LT |
21 | |
22 | /* Free memory management - zoned buddy allocator. */ | |
23 | #ifndef CONFIG_FORCE_MAX_ZONEORDER | |
24 | #define MAX_ORDER 11 | |
25 | #else | |
26 | #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER | |
27 | #endif | |
e984bb43 | 28 | #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1)) |
1da177e4 | 29 | |
5ad333eb AW |
30 | /* |
31 | * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed | |
32 | * costly to service. That is between allocation orders which should | |
33 | * coelesce naturally under reasonable reclaim pressure and those which | |
34 | * will not. | |
35 | */ | |
36 | #define PAGE_ALLOC_COSTLY_ORDER 3 | |
37 | ||
b2a0ac88 | 38 | #define MIGRATE_UNMOVABLE 0 |
e12ba74d MG |
39 | #define MIGRATE_RECLAIMABLE 1 |
40 | #define MIGRATE_MOVABLE 2 | |
5f8dcc21 | 41 | #define MIGRATE_PCPTYPES 3 /* the number of types on the pcp lists */ |
64c5e135 | 42 | #define MIGRATE_RESERVE 3 |
a5d76b54 KH |
43 | #define MIGRATE_ISOLATE 4 /* can't allocate from here */ |
44 | #define MIGRATE_TYPES 5 | |
b2a0ac88 MG |
45 | |
46 | #define for_each_migratetype_order(order, type) \ | |
47 | for (order = 0; order < MAX_ORDER; order++) \ | |
48 | for (type = 0; type < MIGRATE_TYPES; type++) | |
49 | ||
467c996c MG |
50 | extern int page_group_by_mobility_disabled; |
51 | ||
52 | static inline int get_pageblock_migratetype(struct page *page) | |
53 | { | |
467c996c MG |
54 | return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end); |
55 | } | |
56 | ||
1da177e4 | 57 | struct free_area { |
b2a0ac88 | 58 | struct list_head free_list[MIGRATE_TYPES]; |
1da177e4 LT |
59 | unsigned long nr_free; |
60 | }; | |
61 | ||
62 | struct pglist_data; | |
63 | ||
64 | /* | |
65 | * zone->lock and zone->lru_lock are two of the hottest locks in the kernel. | |
66 | * So add a wild amount of padding here to ensure that they fall into separate | |
67 | * cachelines. There are very few zone structures in the machine, so space | |
68 | * consumption is not a concern here. | |
69 | */ | |
70 | #if defined(CONFIG_SMP) | |
71 | struct zone_padding { | |
72 | char x[0]; | |
22fc6ecc | 73 | } ____cacheline_internodealigned_in_smp; |
1da177e4 LT |
74 | #define ZONE_PADDING(name) struct zone_padding name; |
75 | #else | |
76 | #define ZONE_PADDING(name) | |
77 | #endif | |
78 | ||
2244b95a | 79 | enum zone_stat_item { |
51ed4491 | 80 | /* First 128 byte cacheline (assuming 64 bit words) */ |
d23ad423 | 81 | NR_FREE_PAGES, |
b69408e8 | 82 | NR_LRU_BASE, |
4f98a2fe RR |
83 | NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */ |
84 | NR_ACTIVE_ANON, /* " " " " " */ | |
85 | NR_INACTIVE_FILE, /* " " " " " */ | |
86 | NR_ACTIVE_FILE, /* " " " " " */ | |
894bc310 | 87 | NR_UNEVICTABLE, /* " " " " " */ |
5344b7e6 | 88 | NR_MLOCK, /* mlock()ed pages found and moved off LRU */ |
f3dbd344 CL |
89 | NR_ANON_PAGES, /* Mapped anonymous pages */ |
90 | NR_FILE_MAPPED, /* pagecache pages mapped into pagetables. | |
65ba55f5 | 91 | only modified from process context */ |
347ce434 | 92 | NR_FILE_PAGES, |
b1e7a8fd | 93 | NR_FILE_DIRTY, |
ce866b34 | 94 | NR_WRITEBACK, |
51ed4491 CL |
95 | NR_SLAB_RECLAIMABLE, |
96 | NR_SLAB_UNRECLAIMABLE, | |
97 | NR_PAGETABLE, /* used for pagetables */ | |
c6a7f572 KM |
98 | NR_KERNEL_STACK, |
99 | /* Second 128 byte cacheline */ | |
fd39fc85 | 100 | NR_UNSTABLE_NFS, /* NFS unstable pages */ |
d2c5e30c | 101 | NR_BOUNCE, |
e129b5c2 | 102 | NR_VMSCAN_WRITE, |
fc3ba692 | 103 | NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */ |
a731286d KM |
104 | NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */ |
105 | NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */ | |
4b02108a | 106 | NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */ |
ca889e6c CL |
107 | #ifdef CONFIG_NUMA |
108 | NUMA_HIT, /* allocated in intended node */ | |
109 | NUMA_MISS, /* allocated in non intended node */ | |
110 | NUMA_FOREIGN, /* was intended here, hit elsewhere */ | |
111 | NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */ | |
112 | NUMA_LOCAL, /* allocation from local node */ | |
113 | NUMA_OTHER, /* allocation from other node */ | |
114 | #endif | |
2244b95a CL |
115 | NR_VM_ZONE_STAT_ITEMS }; |
116 | ||
4f98a2fe RR |
117 | /* |
118 | * We do arithmetic on the LRU lists in various places in the code, | |
119 | * so it is important to keep the active lists LRU_ACTIVE higher in | |
120 | * the array than the corresponding inactive lists, and to keep | |
121 | * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists. | |
122 | * | |
123 | * This has to be kept in sync with the statistics in zone_stat_item | |
124 | * above and the descriptions in vmstat_text in mm/vmstat.c | |
125 | */ | |
126 | #define LRU_BASE 0 | |
127 | #define LRU_ACTIVE 1 | |
128 | #define LRU_FILE 2 | |
129 | ||
b69408e8 | 130 | enum lru_list { |
4f98a2fe RR |
131 | LRU_INACTIVE_ANON = LRU_BASE, |
132 | LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE, | |
133 | LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE, | |
134 | LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE, | |
894bc310 | 135 | LRU_UNEVICTABLE, |
894bc310 LS |
136 | NR_LRU_LISTS |
137 | }; | |
b69408e8 CL |
138 | |
139 | #define for_each_lru(l) for (l = 0; l < NR_LRU_LISTS; l++) | |
140 | ||
894bc310 LS |
141 | #define for_each_evictable_lru(l) for (l = 0; l <= LRU_ACTIVE_FILE; l++) |
142 | ||
4f98a2fe RR |
143 | static inline int is_file_lru(enum lru_list l) |
144 | { | |
145 | return (l == LRU_INACTIVE_FILE || l == LRU_ACTIVE_FILE); | |
146 | } | |
147 | ||
b69408e8 CL |
148 | static inline int is_active_lru(enum lru_list l) |
149 | { | |
4f98a2fe | 150 | return (l == LRU_ACTIVE_ANON || l == LRU_ACTIVE_FILE); |
b69408e8 CL |
151 | } |
152 | ||
894bc310 LS |
153 | static inline int is_unevictable_lru(enum lru_list l) |
154 | { | |
894bc310 | 155 | return (l == LRU_UNEVICTABLE); |
894bc310 LS |
156 | } |
157 | ||
41858966 MG |
158 | enum zone_watermarks { |
159 | WMARK_MIN, | |
160 | WMARK_LOW, | |
161 | WMARK_HIGH, | |
162 | NR_WMARK | |
163 | }; | |
164 | ||
165 | #define min_wmark_pages(z) (z->watermark[WMARK_MIN]) | |
166 | #define low_wmark_pages(z) (z->watermark[WMARK_LOW]) | |
167 | #define high_wmark_pages(z) (z->watermark[WMARK_HIGH]) | |
168 | ||
1da177e4 LT |
169 | struct per_cpu_pages { |
170 | int count; /* number of pages in the list */ | |
1da177e4 LT |
171 | int high; /* high watermark, emptying needed */ |
172 | int batch; /* chunk size for buddy add/remove */ | |
5f8dcc21 MG |
173 | |
174 | /* Lists of pages, one per migrate type stored on the pcp-lists */ | |
175 | struct list_head lists[MIGRATE_PCPTYPES]; | |
1da177e4 LT |
176 | }; |
177 | ||
178 | struct per_cpu_pageset { | |
3dfa5721 | 179 | struct per_cpu_pages pcp; |
4037d452 CL |
180 | #ifdef CONFIG_NUMA |
181 | s8 expire; | |
182 | #endif | |
2244b95a | 183 | #ifdef CONFIG_SMP |
df9ecaba | 184 | s8 stat_threshold; |
2244b95a CL |
185 | s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS]; |
186 | #endif | |
99dcc3e5 | 187 | }; |
e7c8d5c9 | 188 | |
97965478 CL |
189 | #endif /* !__GENERATING_BOUNDS.H */ |
190 | ||
2f1b6248 | 191 | enum zone_type { |
4b51d669 | 192 | #ifdef CONFIG_ZONE_DMA |
2f1b6248 CL |
193 | /* |
194 | * ZONE_DMA is used when there are devices that are not able | |
195 | * to do DMA to all of addressable memory (ZONE_NORMAL). Then we | |
196 | * carve out the portion of memory that is needed for these devices. | |
197 | * The range is arch specific. | |
198 | * | |
199 | * Some examples | |
200 | * | |
201 | * Architecture Limit | |
202 | * --------------------------- | |
203 | * parisc, ia64, sparc <4G | |
204 | * s390 <2G | |
2f1b6248 CL |
205 | * arm Various |
206 | * alpha Unlimited or 0-16MB. | |
207 | * | |
208 | * i386, x86_64 and multiple other arches | |
209 | * <16M. | |
210 | */ | |
211 | ZONE_DMA, | |
4b51d669 | 212 | #endif |
fb0e7942 | 213 | #ifdef CONFIG_ZONE_DMA32 |
2f1b6248 CL |
214 | /* |
215 | * x86_64 needs two ZONE_DMAs because it supports devices that are | |
216 | * only able to do DMA to the lower 16M but also 32 bit devices that | |
217 | * can only do DMA areas below 4G. | |
218 | */ | |
219 | ZONE_DMA32, | |
fb0e7942 | 220 | #endif |
2f1b6248 CL |
221 | /* |
222 | * Normal addressable memory is in ZONE_NORMAL. DMA operations can be | |
223 | * performed on pages in ZONE_NORMAL if the DMA devices support | |
224 | * transfers to all addressable memory. | |
225 | */ | |
226 | ZONE_NORMAL, | |
e53ef38d | 227 | #ifdef CONFIG_HIGHMEM |
2f1b6248 CL |
228 | /* |
229 | * A memory area that is only addressable by the kernel through | |
230 | * mapping portions into its own address space. This is for example | |
231 | * used by i386 to allow the kernel to address the memory beyond | |
232 | * 900MB. The kernel will set up special mappings (page | |
233 | * table entries on i386) for each page that the kernel needs to | |
234 | * access. | |
235 | */ | |
236 | ZONE_HIGHMEM, | |
e53ef38d | 237 | #endif |
2a1e274a | 238 | ZONE_MOVABLE, |
97965478 | 239 | __MAX_NR_ZONES |
2f1b6248 | 240 | }; |
1da177e4 | 241 | |
97965478 CL |
242 | #ifndef __GENERATING_BOUNDS_H |
243 | ||
1da177e4 LT |
244 | /* |
245 | * When a memory allocation must conform to specific limitations (such | |
246 | * as being suitable for DMA) the caller will pass in hints to the | |
247 | * allocator in the gfp_mask, in the zone modifier bits. These bits | |
248 | * are used to select a priority ordered list of memory zones which | |
19655d34 | 249 | * match the requested limits. See gfp_zone() in include/linux/gfp.h |
1da177e4 | 250 | */ |
fb0e7942 | 251 | |
97965478 | 252 | #if MAX_NR_ZONES < 2 |
4b51d669 | 253 | #define ZONES_SHIFT 0 |
97965478 | 254 | #elif MAX_NR_ZONES <= 2 |
19655d34 | 255 | #define ZONES_SHIFT 1 |
97965478 | 256 | #elif MAX_NR_ZONES <= 4 |
19655d34 | 257 | #define ZONES_SHIFT 2 |
4b51d669 CL |
258 | #else |
259 | #error ZONES_SHIFT -- too many zones configured adjust calculation | |
fb0e7942 | 260 | #endif |
1da177e4 | 261 | |
6e901571 KM |
262 | struct zone_reclaim_stat { |
263 | /* | |
264 | * The pageout code in vmscan.c keeps track of how many of the | |
265 | * mem/swap backed and file backed pages are refeferenced. | |
266 | * The higher the rotated/scanned ratio, the more valuable | |
267 | * that cache is. | |
268 | * | |
269 | * The anon LRU stats live in [0], file LRU stats in [1] | |
270 | */ | |
271 | unsigned long recent_rotated[2]; | |
272 | unsigned long recent_scanned[2]; | |
f8629631 WF |
273 | |
274 | /* | |
275 | * accumulated for batching | |
276 | */ | |
277 | unsigned long nr_saved_scan[NR_LRU_LISTS]; | |
6e901571 KM |
278 | }; |
279 | ||
1da177e4 LT |
280 | struct zone { |
281 | /* Fields commonly accessed by the page allocator */ | |
41858966 MG |
282 | |
283 | /* zone watermarks, access with *_wmark_pages(zone) macros */ | |
284 | unsigned long watermark[NR_WMARK]; | |
285 | ||
1da177e4 LT |
286 | /* |
287 | * We don't know if the memory that we're going to allocate will be freeable | |
288 | * or/and it will be released eventually, so to avoid totally wasting several | |
289 | * GB of ram we must reserve some of the lower zone memory (otherwise we risk | |
290 | * to run OOM on the lower zones despite there's tons of freeable ram | |
291 | * on the higher zones). This array is recalculated at runtime if the | |
292 | * sysctl_lowmem_reserve_ratio sysctl changes. | |
293 | */ | |
294 | unsigned long lowmem_reserve[MAX_NR_ZONES]; | |
295 | ||
e7c8d5c9 | 296 | #ifdef CONFIG_NUMA |
d5f541ed | 297 | int node; |
9614634f CL |
298 | /* |
299 | * zone reclaim becomes active if more unmapped pages exist. | |
300 | */ | |
8417bba4 | 301 | unsigned long min_unmapped_pages; |
0ff38490 | 302 | unsigned long min_slab_pages; |
e7c8d5c9 | 303 | #endif |
43cf38eb | 304 | struct per_cpu_pageset __percpu *pageset; |
1da177e4 LT |
305 | /* |
306 | * free areas of different sizes | |
307 | */ | |
308 | spinlock_t lock; | |
93e4a89a | 309 | int all_unreclaimable; /* All pages pinned */ |
bdc8cb98 DH |
310 | #ifdef CONFIG_MEMORY_HOTPLUG |
311 | /* see spanned/present_pages for more description */ | |
312 | seqlock_t span_seqlock; | |
313 | #endif | |
1da177e4 LT |
314 | struct free_area free_area[MAX_ORDER]; |
315 | ||
835c134e MG |
316 | #ifndef CONFIG_SPARSEMEM |
317 | /* | |
d9c23400 | 318 | * Flags for a pageblock_nr_pages block. See pageblock-flags.h. |
835c134e MG |
319 | * In SPARSEMEM, this map is stored in struct mem_section |
320 | */ | |
321 | unsigned long *pageblock_flags; | |
322 | #endif /* CONFIG_SPARSEMEM */ | |
323 | ||
4f92e258 MG |
324 | #ifdef CONFIG_COMPACTION |
325 | /* | |
326 | * On compaction failure, 1<<compact_defer_shift compactions | |
327 | * are skipped before trying again. The number attempted since | |
328 | * last failure is tracked with compact_considered. | |
329 | */ | |
330 | unsigned int compact_considered; | |
331 | unsigned int compact_defer_shift; | |
332 | #endif | |
1da177e4 LT |
333 | |
334 | ZONE_PADDING(_pad1_) | |
335 | ||
336 | /* Fields commonly accessed by the page reclaim scanner */ | |
337 | spinlock_t lru_lock; | |
6e08a369 | 338 | struct zone_lru { |
b69408e8 | 339 | struct list_head list; |
b69408e8 | 340 | } lru[NR_LRU_LISTS]; |
4f98a2fe | 341 | |
6e901571 | 342 | struct zone_reclaim_stat reclaim_stat; |
4f98a2fe | 343 | |
1da177e4 | 344 | unsigned long pages_scanned; /* since last reclaim */ |
e815af95 | 345 | unsigned long flags; /* zone flags, see below */ |
753ee728 | 346 | |
2244b95a CL |
347 | /* Zone statistics */ |
348 | atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS]; | |
9eeff239 | 349 | |
1da177e4 LT |
350 | /* |
351 | * prev_priority holds the scanning priority for this zone. It is | |
352 | * defined as the scanning priority at which we achieved our reclaim | |
353 | * target at the previous try_to_free_pages() or balance_pgdat() | |
2a61aa40 | 354 | * invocation. |
1da177e4 LT |
355 | * |
356 | * We use prev_priority as a measure of how much stress page reclaim is | |
357 | * under - it drives the swappiness decision: whether to unmap mapped | |
358 | * pages. | |
359 | * | |
3bb1a852 | 360 | * Access to both this field is quite racy even on uniprocessor. But |
1da177e4 LT |
361 | * it is expected to average out OK. |
362 | */ | |
1da177e4 LT |
363 | int prev_priority; |
364 | ||
556adecb RR |
365 | /* |
366 | * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on | |
367 | * this zone's LRU. Maintained by the pageout code. | |
368 | */ | |
369 | unsigned int inactive_ratio; | |
370 | ||
1da177e4 LT |
371 | |
372 | ZONE_PADDING(_pad2_) | |
373 | /* Rarely used or read-mostly fields */ | |
374 | ||
375 | /* | |
376 | * wait_table -- the array holding the hash table | |
02b694de | 377 | * wait_table_hash_nr_entries -- the size of the hash table array |
1da177e4 LT |
378 | * wait_table_bits -- wait_table_size == (1 << wait_table_bits) |
379 | * | |
380 | * The purpose of all these is to keep track of the people | |
381 | * waiting for a page to become available and make them | |
382 | * runnable again when possible. The trouble is that this | |
383 | * consumes a lot of space, especially when so few things | |
384 | * wait on pages at a given time. So instead of using | |
385 | * per-page waitqueues, we use a waitqueue hash table. | |
386 | * | |
387 | * The bucket discipline is to sleep on the same queue when | |
388 | * colliding and wake all in that wait queue when removing. | |
389 | * When something wakes, it must check to be sure its page is | |
390 | * truly available, a la thundering herd. The cost of a | |
391 | * collision is great, but given the expected load of the | |
392 | * table, they should be so rare as to be outweighed by the | |
393 | * benefits from the saved space. | |
394 | * | |
395 | * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the | |
396 | * primary users of these fields, and in mm/page_alloc.c | |
397 | * free_area_init_core() performs the initialization of them. | |
398 | */ | |
399 | wait_queue_head_t * wait_table; | |
02b694de | 400 | unsigned long wait_table_hash_nr_entries; |
1da177e4 LT |
401 | unsigned long wait_table_bits; |
402 | ||
403 | /* | |
404 | * Discontig memory support fields. | |
405 | */ | |
406 | struct pglist_data *zone_pgdat; | |
1da177e4 LT |
407 | /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */ |
408 | unsigned long zone_start_pfn; | |
409 | ||
bdc8cb98 DH |
410 | /* |
411 | * zone_start_pfn, spanned_pages and present_pages are all | |
412 | * protected by span_seqlock. It is a seqlock because it has | |
413 | * to be read outside of zone->lock, and it is done in the main | |
414 | * allocator path. But, it is written quite infrequently. | |
415 | * | |
416 | * The lock is declared along with zone->lock because it is | |
417 | * frequently read in proximity to zone->lock. It's good to | |
418 | * give them a chance of being in the same cacheline. | |
419 | */ | |
1da177e4 LT |
420 | unsigned long spanned_pages; /* total size, including holes */ |
421 | unsigned long present_pages; /* amount of memory (excluding holes) */ | |
422 | ||
423 | /* | |
424 | * rarely used fields: | |
425 | */ | |
15ad7cdc | 426 | const char *name; |
22fc6ecc | 427 | } ____cacheline_internodealigned_in_smp; |
1da177e4 | 428 | |
e815af95 | 429 | typedef enum { |
e815af95 | 430 | ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */ |
098d7f12 | 431 | ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */ |
e815af95 DR |
432 | } zone_flags_t; |
433 | ||
434 | static inline void zone_set_flag(struct zone *zone, zone_flags_t flag) | |
435 | { | |
436 | set_bit(flag, &zone->flags); | |
437 | } | |
d773ed6b DR |
438 | |
439 | static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag) | |
440 | { | |
441 | return test_and_set_bit(flag, &zone->flags); | |
442 | } | |
443 | ||
e815af95 DR |
444 | static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag) |
445 | { | |
446 | clear_bit(flag, &zone->flags); | |
447 | } | |
448 | ||
e815af95 DR |
449 | static inline int zone_is_reclaim_locked(const struct zone *zone) |
450 | { | |
451 | return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags); | |
452 | } | |
d773ed6b | 453 | |
098d7f12 DR |
454 | static inline int zone_is_oom_locked(const struct zone *zone) |
455 | { | |
456 | return test_bit(ZONE_OOM_LOCKED, &zone->flags); | |
457 | } | |
e815af95 | 458 | |
1da177e4 LT |
459 | /* |
460 | * The "priority" of VM scanning is how much of the queues we will scan in one | |
461 | * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the | |
462 | * queues ("queue_length >> 12") during an aging round. | |
463 | */ | |
464 | #define DEF_PRIORITY 12 | |
465 | ||
9276b1bc PJ |
466 | /* Maximum number of zones on a zonelist */ |
467 | #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES) | |
468 | ||
469 | #ifdef CONFIG_NUMA | |
523b9458 CL |
470 | |
471 | /* | |
472 | * The NUMA zonelists are doubled becausse we need zonelists that restrict the | |
473 | * allocations to a single node for GFP_THISNODE. | |
474 | * | |
54a6eb5c MG |
475 | * [0] : Zonelist with fallback |
476 | * [1] : No fallback (GFP_THISNODE) | |
523b9458 | 477 | */ |
54a6eb5c | 478 | #define MAX_ZONELISTS 2 |
523b9458 CL |
479 | |
480 | ||
9276b1bc PJ |
481 | /* |
482 | * We cache key information from each zonelist for smaller cache | |
483 | * footprint when scanning for free pages in get_page_from_freelist(). | |
484 | * | |
485 | * 1) The BITMAP fullzones tracks which zones in a zonelist have come | |
486 | * up short of free memory since the last time (last_fullzone_zap) | |
487 | * we zero'd fullzones. | |
488 | * 2) The array z_to_n[] maps each zone in the zonelist to its node | |
489 | * id, so that we can efficiently evaluate whether that node is | |
490 | * set in the current tasks mems_allowed. | |
491 | * | |
492 | * Both fullzones and z_to_n[] are one-to-one with the zonelist, | |
493 | * indexed by a zones offset in the zonelist zones[] array. | |
494 | * | |
495 | * The get_page_from_freelist() routine does two scans. During the | |
496 | * first scan, we skip zones whose corresponding bit in 'fullzones' | |
497 | * is set or whose corresponding node in current->mems_allowed (which | |
498 | * comes from cpusets) is not set. During the second scan, we bypass | |
499 | * this zonelist_cache, to ensure we look methodically at each zone. | |
500 | * | |
501 | * Once per second, we zero out (zap) fullzones, forcing us to | |
502 | * reconsider nodes that might have regained more free memory. | |
503 | * The field last_full_zap is the time we last zapped fullzones. | |
504 | * | |
505 | * This mechanism reduces the amount of time we waste repeatedly | |
506 | * reexaming zones for free memory when they just came up low on | |
507 | * memory momentarilly ago. | |
508 | * | |
509 | * The zonelist_cache struct members logically belong in struct | |
510 | * zonelist. However, the mempolicy zonelists constructed for | |
511 | * MPOL_BIND are intentionally variable length (and usually much | |
512 | * shorter). A general purpose mechanism for handling structs with | |
513 | * multiple variable length members is more mechanism than we want | |
514 | * here. We resort to some special case hackery instead. | |
515 | * | |
516 | * The MPOL_BIND zonelists don't need this zonelist_cache (in good | |
517 | * part because they are shorter), so we put the fixed length stuff | |
518 | * at the front of the zonelist struct, ending in a variable length | |
519 | * zones[], as is needed by MPOL_BIND. | |
520 | * | |
521 | * Then we put the optional zonelist cache on the end of the zonelist | |
522 | * struct. This optional stuff is found by a 'zlcache_ptr' pointer in | |
523 | * the fixed length portion at the front of the struct. This pointer | |
524 | * both enables us to find the zonelist cache, and in the case of | |
525 | * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL) | |
526 | * to know that the zonelist cache is not there. | |
527 | * | |
528 | * The end result is that struct zonelists come in two flavors: | |
529 | * 1) The full, fixed length version, shown below, and | |
530 | * 2) The custom zonelists for MPOL_BIND. | |
531 | * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache. | |
532 | * | |
533 | * Even though there may be multiple CPU cores on a node modifying | |
534 | * fullzones or last_full_zap in the same zonelist_cache at the same | |
535 | * time, we don't lock it. This is just hint data - if it is wrong now | |
536 | * and then, the allocator will still function, perhaps a bit slower. | |
537 | */ | |
538 | ||
539 | ||
540 | struct zonelist_cache { | |
9276b1bc | 541 | unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */ |
7253f4ef | 542 | DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */ |
9276b1bc PJ |
543 | unsigned long last_full_zap; /* when last zap'd (jiffies) */ |
544 | }; | |
545 | #else | |
54a6eb5c | 546 | #define MAX_ZONELISTS 1 |
9276b1bc PJ |
547 | struct zonelist_cache; |
548 | #endif | |
549 | ||
dd1a239f MG |
550 | /* |
551 | * This struct contains information about a zone in a zonelist. It is stored | |
552 | * here to avoid dereferences into large structures and lookups of tables | |
553 | */ | |
554 | struct zoneref { | |
555 | struct zone *zone; /* Pointer to actual zone */ | |
556 | int zone_idx; /* zone_idx(zoneref->zone) */ | |
557 | }; | |
558 | ||
1da177e4 LT |
559 | /* |
560 | * One allocation request operates on a zonelist. A zonelist | |
561 | * is a list of zones, the first one is the 'goal' of the | |
562 | * allocation, the other zones are fallback zones, in decreasing | |
563 | * priority. | |
564 | * | |
9276b1bc PJ |
565 | * If zlcache_ptr is not NULL, then it is just the address of zlcache, |
566 | * as explained above. If zlcache_ptr is NULL, there is no zlcache. | |
dd1a239f MG |
567 | * * |
568 | * To speed the reading of the zonelist, the zonerefs contain the zone index | |
569 | * of the entry being read. Helper functions to access information given | |
570 | * a struct zoneref are | |
571 | * | |
572 | * zonelist_zone() - Return the struct zone * for an entry in _zonerefs | |
573 | * zonelist_zone_idx() - Return the index of the zone for an entry | |
574 | * zonelist_node_idx() - Return the index of the node for an entry | |
1da177e4 LT |
575 | */ |
576 | struct zonelist { | |
9276b1bc | 577 | struct zonelist_cache *zlcache_ptr; // NULL or &zlcache |
dd1a239f | 578 | struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1]; |
9276b1bc PJ |
579 | #ifdef CONFIG_NUMA |
580 | struct zonelist_cache zlcache; // optional ... | |
581 | #endif | |
1da177e4 LT |
582 | }; |
583 | ||
c713216d MG |
584 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
585 | struct node_active_region { | |
586 | unsigned long start_pfn; | |
587 | unsigned long end_pfn; | |
588 | int nid; | |
589 | }; | |
590 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ | |
1da177e4 | 591 | |
5b99cd0e HC |
592 | #ifndef CONFIG_DISCONTIGMEM |
593 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ | |
594 | extern struct page *mem_map; | |
595 | #endif | |
596 | ||
1da177e4 LT |
597 | /* |
598 | * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM | |
599 | * (mostly NUMA machines?) to denote a higher-level memory zone than the | |
600 | * zone denotes. | |
601 | * | |
602 | * On NUMA machines, each NUMA node would have a pg_data_t to describe | |
603 | * it's memory layout. | |
604 | * | |
605 | * Memory statistics and page replacement data structures are maintained on a | |
606 | * per-zone basis. | |
607 | */ | |
608 | struct bootmem_data; | |
609 | typedef struct pglist_data { | |
610 | struct zone node_zones[MAX_NR_ZONES]; | |
523b9458 | 611 | struct zonelist node_zonelists[MAX_ZONELISTS]; |
1da177e4 | 612 | int nr_zones; |
52d4b9ac | 613 | #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */ |
1da177e4 | 614 | struct page *node_mem_map; |
52d4b9ac KH |
615 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR |
616 | struct page_cgroup *node_page_cgroup; | |
617 | #endif | |
d41dee36 | 618 | #endif |
08677214 | 619 | #ifndef CONFIG_NO_BOOTMEM |
1da177e4 | 620 | struct bootmem_data *bdata; |
08677214 | 621 | #endif |
208d54e5 DH |
622 | #ifdef CONFIG_MEMORY_HOTPLUG |
623 | /* | |
624 | * Must be held any time you expect node_start_pfn, node_present_pages | |
625 | * or node_spanned_pages stay constant. Holding this will also | |
626 | * guarantee that any pfn_valid() stays that way. | |
627 | * | |
628 | * Nests above zone->lock and zone->size_seqlock. | |
629 | */ | |
630 | spinlock_t node_size_lock; | |
631 | #endif | |
1da177e4 LT |
632 | unsigned long node_start_pfn; |
633 | unsigned long node_present_pages; /* total number of physical pages */ | |
634 | unsigned long node_spanned_pages; /* total size of physical page | |
635 | range, including holes */ | |
636 | int node_id; | |
1da177e4 LT |
637 | wait_queue_head_t kswapd_wait; |
638 | struct task_struct *kswapd; | |
639 | int kswapd_max_order; | |
640 | } pg_data_t; | |
641 | ||
642 | #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) | |
643 | #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) | |
d41dee36 | 644 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
408fde81 | 645 | #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) |
d41dee36 AW |
646 | #else |
647 | #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) | |
648 | #endif | |
408fde81 | 649 | #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr)) |
1da177e4 | 650 | |
208d54e5 DH |
651 | #include <linux/memory_hotplug.h> |
652 | ||
1da177e4 LT |
653 | void get_zone_counts(unsigned long *active, unsigned long *inactive, |
654 | unsigned long *free); | |
655 | void build_all_zonelists(void); | |
656 | void wakeup_kswapd(struct zone *zone, int order); | |
657 | int zone_watermark_ok(struct zone *z, int order, unsigned long mark, | |
7fb1d9fc | 658 | int classzone_idx, int alloc_flags); |
a2f3aa02 DH |
659 | enum memmap_context { |
660 | MEMMAP_EARLY, | |
661 | MEMMAP_HOTPLUG, | |
662 | }; | |
718127cc | 663 | extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn, |
a2f3aa02 DH |
664 | unsigned long size, |
665 | enum memmap_context context); | |
718127cc | 666 | |
1da177e4 LT |
667 | #ifdef CONFIG_HAVE_MEMORY_PRESENT |
668 | void memory_present(int nid, unsigned long start, unsigned long end); | |
669 | #else | |
670 | static inline void memory_present(int nid, unsigned long start, unsigned long end) {} | |
671 | #endif | |
672 | ||
673 | #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE | |
674 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); | |
675 | #endif | |
676 | ||
677 | /* | |
678 | * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc. | |
679 | */ | |
680 | #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones) | |
681 | ||
f3fe6512 CK |
682 | static inline int populated_zone(struct zone *zone) |
683 | { | |
684 | return (!!zone->present_pages); | |
685 | } | |
686 | ||
2a1e274a MG |
687 | extern int movable_zone; |
688 | ||
689 | static inline int zone_movable_is_highmem(void) | |
690 | { | |
691 | #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP) | |
692 | return movable_zone == ZONE_HIGHMEM; | |
693 | #else | |
694 | return 0; | |
695 | #endif | |
696 | } | |
697 | ||
2f1b6248 | 698 | static inline int is_highmem_idx(enum zone_type idx) |
1da177e4 | 699 | { |
e53ef38d | 700 | #ifdef CONFIG_HIGHMEM |
2a1e274a MG |
701 | return (idx == ZONE_HIGHMEM || |
702 | (idx == ZONE_MOVABLE && zone_movable_is_highmem())); | |
e53ef38d CL |
703 | #else |
704 | return 0; | |
705 | #endif | |
1da177e4 LT |
706 | } |
707 | ||
2f1b6248 | 708 | static inline int is_normal_idx(enum zone_type idx) |
1da177e4 LT |
709 | { |
710 | return (idx == ZONE_NORMAL); | |
711 | } | |
9328b8fa | 712 | |
1da177e4 LT |
713 | /** |
714 | * is_highmem - helper function to quickly check if a struct zone is a | |
715 | * highmem zone or not. This is an attempt to keep references | |
716 | * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum. | |
717 | * @zone - pointer to struct zone variable | |
718 | */ | |
719 | static inline int is_highmem(struct zone *zone) | |
720 | { | |
e53ef38d | 721 | #ifdef CONFIG_HIGHMEM |
ddc81ed2 HH |
722 | int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones; |
723 | return zone_off == ZONE_HIGHMEM * sizeof(*zone) || | |
724 | (zone_off == ZONE_MOVABLE * sizeof(*zone) && | |
725 | zone_movable_is_highmem()); | |
e53ef38d CL |
726 | #else |
727 | return 0; | |
728 | #endif | |
1da177e4 LT |
729 | } |
730 | ||
731 | static inline int is_normal(struct zone *zone) | |
732 | { | |
733 | return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL; | |
734 | } | |
735 | ||
9328b8fa NP |
736 | static inline int is_dma32(struct zone *zone) |
737 | { | |
fb0e7942 | 738 | #ifdef CONFIG_ZONE_DMA32 |
9328b8fa | 739 | return zone == zone->zone_pgdat->node_zones + ZONE_DMA32; |
fb0e7942 CL |
740 | #else |
741 | return 0; | |
742 | #endif | |
9328b8fa NP |
743 | } |
744 | ||
745 | static inline int is_dma(struct zone *zone) | |
746 | { | |
4b51d669 | 747 | #ifdef CONFIG_ZONE_DMA |
9328b8fa | 748 | return zone == zone->zone_pgdat->node_zones + ZONE_DMA; |
4b51d669 CL |
749 | #else |
750 | return 0; | |
751 | #endif | |
9328b8fa NP |
752 | } |
753 | ||
1da177e4 LT |
754 | /* These two functions are used to setup the per zone pages min values */ |
755 | struct ctl_table; | |
8d65af78 | 756 | int min_free_kbytes_sysctl_handler(struct ctl_table *, int, |
1da177e4 LT |
757 | void __user *, size_t *, loff_t *); |
758 | extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1]; | |
8d65af78 | 759 | int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, |
1da177e4 | 760 | void __user *, size_t *, loff_t *); |
8d65af78 | 761 | int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, |
8ad4b1fb | 762 | void __user *, size_t *, loff_t *); |
9614634f | 763 | int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int, |
8d65af78 | 764 | void __user *, size_t *, loff_t *); |
0ff38490 | 765 | int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int, |
8d65af78 | 766 | void __user *, size_t *, loff_t *); |
1da177e4 | 767 | |
f0c0b2b8 | 768 | extern int numa_zonelist_order_handler(struct ctl_table *, int, |
8d65af78 | 769 | void __user *, size_t *, loff_t *); |
f0c0b2b8 KH |
770 | extern char numa_zonelist_order[]; |
771 | #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */ | |
772 | ||
93b7504e | 773 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
1da177e4 LT |
774 | |
775 | extern struct pglist_data contig_page_data; | |
776 | #define NODE_DATA(nid) (&contig_page_data) | |
777 | #define NODE_MEM_MAP(nid) mem_map | |
1da177e4 | 778 | |
93b7504e | 779 | #else /* CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 LT |
780 | |
781 | #include <asm/mmzone.h> | |
782 | ||
93b7504e | 783 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
348f8b6c | 784 | |
95144c78 KH |
785 | extern struct pglist_data *first_online_pgdat(void); |
786 | extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat); | |
787 | extern struct zone *next_zone(struct zone *zone); | |
8357f869 KH |
788 | |
789 | /** | |
12d15f0d | 790 | * for_each_online_pgdat - helper macro to iterate over all online nodes |
8357f869 KH |
791 | * @pgdat - pointer to a pg_data_t variable |
792 | */ | |
793 | #define for_each_online_pgdat(pgdat) \ | |
794 | for (pgdat = first_online_pgdat(); \ | |
795 | pgdat; \ | |
796 | pgdat = next_online_pgdat(pgdat)) | |
8357f869 KH |
797 | /** |
798 | * for_each_zone - helper macro to iterate over all memory zones | |
799 | * @zone - pointer to struct zone variable | |
800 | * | |
801 | * The user only needs to declare the zone variable, for_each_zone | |
802 | * fills it in. | |
803 | */ | |
804 | #define for_each_zone(zone) \ | |
805 | for (zone = (first_online_pgdat())->node_zones; \ | |
806 | zone; \ | |
807 | zone = next_zone(zone)) | |
808 | ||
ee99c71c KM |
809 | #define for_each_populated_zone(zone) \ |
810 | for (zone = (first_online_pgdat())->node_zones; \ | |
811 | zone; \ | |
812 | zone = next_zone(zone)) \ | |
813 | if (!populated_zone(zone)) \ | |
814 | ; /* do nothing */ \ | |
815 | else | |
816 | ||
dd1a239f MG |
817 | static inline struct zone *zonelist_zone(struct zoneref *zoneref) |
818 | { | |
819 | return zoneref->zone; | |
820 | } | |
821 | ||
822 | static inline int zonelist_zone_idx(struct zoneref *zoneref) | |
823 | { | |
824 | return zoneref->zone_idx; | |
825 | } | |
826 | ||
827 | static inline int zonelist_node_idx(struct zoneref *zoneref) | |
828 | { | |
829 | #ifdef CONFIG_NUMA | |
830 | /* zone_to_nid not available in this context */ | |
831 | return zoneref->zone->node; | |
832 | #else | |
833 | return 0; | |
834 | #endif /* CONFIG_NUMA */ | |
835 | } | |
836 | ||
19770b32 MG |
837 | /** |
838 | * 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 | |
839 | * @z - The cursor used as a starting point for the search | |
840 | * @highest_zoneidx - The zone index of the highest zone to return | |
841 | * @nodes - An optional nodemask to filter the zonelist with | |
842 | * @zone - The first suitable zone found is returned via this parameter | |
843 | * | |
844 | * This function returns the next zone at or below a given zone index that is | |
845 | * within the allowed nodemask using a cursor as the starting point for the | |
5bead2a0 MG |
846 | * search. The zoneref returned is a cursor that represents the current zone |
847 | * being examined. It should be advanced by one before calling | |
848 | * next_zones_zonelist again. | |
19770b32 MG |
849 | */ |
850 | struct zoneref *next_zones_zonelist(struct zoneref *z, | |
851 | enum zone_type highest_zoneidx, | |
852 | nodemask_t *nodes, | |
853 | struct zone **zone); | |
dd1a239f | 854 | |
19770b32 MG |
855 | /** |
856 | * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist | |
857 | * @zonelist - The zonelist to search for a suitable zone | |
858 | * @highest_zoneidx - The zone index of the highest zone to return | |
859 | * @nodes - An optional nodemask to filter the zonelist with | |
860 | * @zone - The first suitable zone found is returned via this parameter | |
861 | * | |
862 | * This function returns the first zone at or below a given zone index that is | |
863 | * within the allowed nodemask. The zoneref returned is a cursor that can be | |
5bead2a0 MG |
864 | * used to iterate the zonelist with next_zones_zonelist by advancing it by |
865 | * one before calling. | |
19770b32 | 866 | */ |
dd1a239f | 867 | static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist, |
19770b32 MG |
868 | enum zone_type highest_zoneidx, |
869 | nodemask_t *nodes, | |
870 | struct zone **zone) | |
54a6eb5c | 871 | { |
19770b32 MG |
872 | return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes, |
873 | zone); | |
54a6eb5c MG |
874 | } |
875 | ||
19770b32 MG |
876 | /** |
877 | * 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 | |
878 | * @zone - The current zone in the iterator | |
879 | * @z - The current pointer within zonelist->zones being iterated | |
880 | * @zlist - The zonelist being iterated | |
881 | * @highidx - The zone index of the highest zone to return | |
882 | * @nodemask - Nodemask allowed by the allocator | |
883 | * | |
884 | * This iterator iterates though all zones at or below a given zone index and | |
885 | * within a given nodemask | |
886 | */ | |
887 | #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \ | |
888 | for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \ | |
889 | zone; \ | |
5bead2a0 | 890 | z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \ |
54a6eb5c MG |
891 | |
892 | /** | |
893 | * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index | |
894 | * @zone - The current zone in the iterator | |
895 | * @z - The current pointer within zonelist->zones being iterated | |
896 | * @zlist - The zonelist being iterated | |
897 | * @highidx - The zone index of the highest zone to return | |
898 | * | |
899 | * This iterator iterates though all zones at or below a given zone index. | |
900 | */ | |
901 | #define for_each_zone_zonelist(zone, z, zlist, highidx) \ | |
19770b32 | 902 | for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL) |
54a6eb5c | 903 | |
d41dee36 AW |
904 | #ifdef CONFIG_SPARSEMEM |
905 | #include <asm/sparsemem.h> | |
906 | #endif | |
907 | ||
c713216d MG |
908 | #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \ |
909 | !defined(CONFIG_ARCH_POPULATES_NODE_MAP) | |
b4544568 AM |
910 | static inline unsigned long early_pfn_to_nid(unsigned long pfn) |
911 | { | |
912 | return 0; | |
913 | } | |
b159d43f AW |
914 | #endif |
915 | ||
2bdaf115 AW |
916 | #ifdef CONFIG_FLATMEM |
917 | #define pfn_to_nid(pfn) (0) | |
918 | #endif | |
919 | ||
d41dee36 AW |
920 | #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT) |
921 | #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT) | |
922 | ||
923 | #ifdef CONFIG_SPARSEMEM | |
924 | ||
925 | /* | |
926 | * SECTION_SHIFT #bits space required to store a section # | |
927 | * | |
928 | * PA_SECTION_SHIFT physical address to/from section number | |
929 | * PFN_SECTION_SHIFT pfn to/from section number | |
930 | */ | |
931 | #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS) | |
932 | ||
933 | #define PA_SECTION_SHIFT (SECTION_SIZE_BITS) | |
934 | #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) | |
935 | ||
936 | #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) | |
937 | ||
938 | #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) | |
939 | #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) | |
940 | ||
835c134e | 941 | #define SECTION_BLOCKFLAGS_BITS \ |
d9c23400 | 942 | ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS) |
835c134e | 943 | |
d41dee36 AW |
944 | #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS |
945 | #error Allocator MAX_ORDER exceeds SECTION_SIZE | |
946 | #endif | |
947 | ||
948 | struct page; | |
52d4b9ac | 949 | struct page_cgroup; |
d41dee36 | 950 | struct mem_section { |
29751f69 AW |
951 | /* |
952 | * This is, logically, a pointer to an array of struct | |
953 | * pages. However, it is stored with some other magic. | |
954 | * (see sparse.c::sparse_init_one_section()) | |
955 | * | |
30c253e6 AW |
956 | * Additionally during early boot we encode node id of |
957 | * the location of the section here to guide allocation. | |
958 | * (see sparse.c::memory_present()) | |
959 | * | |
29751f69 AW |
960 | * Making it a UL at least makes someone do a cast |
961 | * before using it wrong. | |
962 | */ | |
963 | unsigned long section_mem_map; | |
5c0e3066 MG |
964 | |
965 | /* See declaration of similar field in struct zone */ | |
966 | unsigned long *pageblock_flags; | |
52d4b9ac KH |
967 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR |
968 | /* | |
969 | * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use | |
970 | * section. (see memcontrol.h/page_cgroup.h about this.) | |
971 | */ | |
972 | struct page_cgroup *page_cgroup; | |
973 | unsigned long pad; | |
974 | #endif | |
d41dee36 AW |
975 | }; |
976 | ||
3e347261 BP |
977 | #ifdef CONFIG_SPARSEMEM_EXTREME |
978 | #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section)) | |
979 | #else | |
980 | #define SECTIONS_PER_ROOT 1 | |
981 | #endif | |
802f192e | 982 | |
3e347261 | 983 | #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT) |
0faa5638 | 984 | #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT) |
3e347261 | 985 | #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1) |
802f192e | 986 | |
3e347261 BP |
987 | #ifdef CONFIG_SPARSEMEM_EXTREME |
988 | extern struct mem_section *mem_section[NR_SECTION_ROOTS]; | |
802f192e | 989 | #else |
3e347261 BP |
990 | extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT]; |
991 | #endif | |
d41dee36 | 992 | |
29751f69 AW |
993 | static inline struct mem_section *__nr_to_section(unsigned long nr) |
994 | { | |
3e347261 BP |
995 | if (!mem_section[SECTION_NR_TO_ROOT(nr)]) |
996 | return NULL; | |
997 | return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK]; | |
29751f69 | 998 | } |
4ca644d9 | 999 | extern int __section_nr(struct mem_section* ms); |
04753278 | 1000 | extern unsigned long usemap_size(void); |
29751f69 AW |
1001 | |
1002 | /* | |
1003 | * We use the lower bits of the mem_map pointer to store | |
1004 | * a little bit of information. There should be at least | |
1005 | * 3 bits here due to 32-bit alignment. | |
1006 | */ | |
1007 | #define SECTION_MARKED_PRESENT (1UL<<0) | |
1008 | #define SECTION_HAS_MEM_MAP (1UL<<1) | |
1009 | #define SECTION_MAP_LAST_BIT (1UL<<2) | |
1010 | #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1)) | |
30c253e6 | 1011 | #define SECTION_NID_SHIFT 2 |
29751f69 AW |
1012 | |
1013 | static inline struct page *__section_mem_map_addr(struct mem_section *section) | |
1014 | { | |
1015 | unsigned long map = section->section_mem_map; | |
1016 | map &= SECTION_MAP_MASK; | |
1017 | return (struct page *)map; | |
1018 | } | |
1019 | ||
540557b9 | 1020 | static inline int present_section(struct mem_section *section) |
29751f69 | 1021 | { |
802f192e | 1022 | return (section && (section->section_mem_map & SECTION_MARKED_PRESENT)); |
29751f69 AW |
1023 | } |
1024 | ||
540557b9 AW |
1025 | static inline int present_section_nr(unsigned long nr) |
1026 | { | |
1027 | return present_section(__nr_to_section(nr)); | |
1028 | } | |
1029 | ||
1030 | static inline int valid_section(struct mem_section *section) | |
29751f69 | 1031 | { |
802f192e | 1032 | return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP)); |
29751f69 AW |
1033 | } |
1034 | ||
1035 | static inline int valid_section_nr(unsigned long nr) | |
1036 | { | |
1037 | return valid_section(__nr_to_section(nr)); | |
1038 | } | |
1039 | ||
d41dee36 AW |
1040 | static inline struct mem_section *__pfn_to_section(unsigned long pfn) |
1041 | { | |
29751f69 | 1042 | return __nr_to_section(pfn_to_section_nr(pfn)); |
d41dee36 AW |
1043 | } |
1044 | ||
d41dee36 AW |
1045 | static inline int pfn_valid(unsigned long pfn) |
1046 | { | |
1047 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) | |
1048 | return 0; | |
29751f69 | 1049 | return valid_section(__nr_to_section(pfn_to_section_nr(pfn))); |
d41dee36 AW |
1050 | } |
1051 | ||
540557b9 AW |
1052 | static inline int pfn_present(unsigned long pfn) |
1053 | { | |
1054 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) | |
1055 | return 0; | |
1056 | return present_section(__nr_to_section(pfn_to_section_nr(pfn))); | |
1057 | } | |
1058 | ||
d41dee36 AW |
1059 | /* |
1060 | * These are _only_ used during initialisation, therefore they | |
1061 | * can use __initdata ... They could have names to indicate | |
1062 | * this restriction. | |
1063 | */ | |
1064 | #ifdef CONFIG_NUMA | |
161599ff AW |
1065 | #define pfn_to_nid(pfn) \ |
1066 | ({ \ | |
1067 | unsigned long __pfn_to_nid_pfn = (pfn); \ | |
1068 | page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \ | |
1069 | }) | |
2bdaf115 AW |
1070 | #else |
1071 | #define pfn_to_nid(pfn) (0) | |
d41dee36 AW |
1072 | #endif |
1073 | ||
d41dee36 AW |
1074 | #define early_pfn_valid(pfn) pfn_valid(pfn) |
1075 | void sparse_init(void); | |
1076 | #else | |
1077 | #define sparse_init() do {} while (0) | |
28ae55c9 | 1078 | #define sparse_index_init(_sec, _nid) do {} while (0) |
d41dee36 AW |
1079 | #endif /* CONFIG_SPARSEMEM */ |
1080 | ||
75167957 | 1081 | #ifdef CONFIG_NODES_SPAN_OTHER_NODES |
cc2559bc | 1082 | bool early_pfn_in_nid(unsigned long pfn, int nid); |
75167957 AW |
1083 | #else |
1084 | #define early_pfn_in_nid(pfn, nid) (1) | |
1085 | #endif | |
1086 | ||
d41dee36 AW |
1087 | #ifndef early_pfn_valid |
1088 | #define early_pfn_valid(pfn) (1) | |
1089 | #endif | |
1090 | ||
1091 | void memory_present(int nid, unsigned long start, unsigned long end); | |
1092 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); | |
1093 | ||
14e07298 AW |
1094 | /* |
1095 | * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we | |
1096 | * need to check pfn validility within that MAX_ORDER_NR_PAGES block. | |
1097 | * pfn_valid_within() should be used in this case; we optimise this away | |
1098 | * when we have no holes within a MAX_ORDER_NR_PAGES block. | |
1099 | */ | |
1100 | #ifdef CONFIG_HOLES_IN_ZONE | |
1101 | #define pfn_valid_within(pfn) pfn_valid(pfn) | |
1102 | #else | |
1103 | #define pfn_valid_within(pfn) (1) | |
1104 | #endif | |
1105 | ||
eb33575c MG |
1106 | #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL |
1107 | /* | |
1108 | * pfn_valid() is meant to be able to tell if a given PFN has valid memmap | |
1109 | * associated with it or not. In FLATMEM, it is expected that holes always | |
1110 | * have valid memmap as long as there is valid PFNs either side of the hole. | |
1111 | * In SPARSEMEM, it is assumed that a valid section has a memmap for the | |
1112 | * entire section. | |
1113 | * | |
1114 | * However, an ARM, and maybe other embedded architectures in the future | |
1115 | * free memmap backing holes to save memory on the assumption the memmap is | |
1116 | * never used. The page_zone linkages are then broken even though pfn_valid() | |
1117 | * returns true. A walker of the full memmap must then do this additional | |
1118 | * check to ensure the memmap they are looking at is sane by making sure | |
1119 | * the zone and PFN linkages are still valid. This is expensive, but walkers | |
1120 | * of the full memmap are extremely rare. | |
1121 | */ | |
1122 | int memmap_valid_within(unsigned long pfn, | |
1123 | struct page *page, struct zone *zone); | |
1124 | #else | |
1125 | static inline int memmap_valid_within(unsigned long pfn, | |
1126 | struct page *page, struct zone *zone) | |
1127 | { | |
1128 | return 1; | |
1129 | } | |
1130 | #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */ | |
1131 | ||
97965478 | 1132 | #endif /* !__GENERATING_BOUNDS.H */ |
1da177e4 | 1133 | #endif /* !__ASSEMBLY__ */ |
1da177e4 | 1134 | #endif /* _LINUX_MMZONE_H */ |