]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
2 | #ifndef _LINUX_MMZONE_H |
3 | #define _LINUX_MMZONE_H | |
4 | ||
1da177e4 | 5 | #ifndef __ASSEMBLY__ |
97965478 | 6 | #ifndef __GENERATING_BOUNDS_H |
1da177e4 | 7 | |
1da177e4 LT |
8 | #include <linux/spinlock.h> |
9 | #include <linux/list.h> | |
10 | #include <linux/wait.h> | |
e815af95 | 11 | #include <linux/bitops.h> |
1da177e4 LT |
12 | #include <linux/cache.h> |
13 | #include <linux/threads.h> | |
14 | #include <linux/numa.h> | |
15 | #include <linux/init.h> | |
bdc8cb98 | 16 | #include <linux/seqlock.h> |
8357f869 | 17 | #include <linux/nodemask.h> |
835c134e | 18 | #include <linux/pageblock-flags.h> |
bbeae5b0 | 19 | #include <linux/page-flags-layout.h> |
60063497 | 20 | #include <linux/atomic.h> |
b03641af DW |
21 | #include <linux/mm_types.h> |
22 | #include <linux/page-flags.h> | |
93ff66bf | 23 | #include <asm/page.h> |
1da177e4 LT |
24 | |
25 | /* Free memory management - zoned buddy allocator. */ | |
26 | #ifndef CONFIG_FORCE_MAX_ZONEORDER | |
27 | #define MAX_ORDER 11 | |
28 | #else | |
29 | #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER | |
30 | #endif | |
e984bb43 | 31 | #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1)) |
1da177e4 | 32 | |
5ad333eb AW |
33 | /* |
34 | * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed | |
35 | * costly to service. That is between allocation orders which should | |
35fca53e | 36 | * coalesce naturally under reasonable reclaim pressure and those which |
5ad333eb AW |
37 | * will not. |
38 | */ | |
39 | #define PAGE_ALLOC_COSTLY_ORDER 3 | |
40 | ||
a6ffdc07 | 41 | enum migratetype { |
47118af0 | 42 | MIGRATE_UNMOVABLE, |
47118af0 | 43 | MIGRATE_MOVABLE, |
016c13da | 44 | MIGRATE_RECLAIMABLE, |
0aaa29a5 MG |
45 | MIGRATE_PCPTYPES, /* the number of types on the pcp lists */ |
46 | MIGRATE_HIGHATOMIC = MIGRATE_PCPTYPES, | |
47118af0 MN |
47 | #ifdef CONFIG_CMA |
48 | /* | |
49 | * MIGRATE_CMA migration type is designed to mimic the way | |
50 | * ZONE_MOVABLE works. Only movable pages can be allocated | |
51 | * from MIGRATE_CMA pageblocks and page allocator never | |
52 | * implicitly change migration type of MIGRATE_CMA pageblock. | |
53 | * | |
54 | * The way to use it is to change migratetype of a range of | |
55 | * pageblocks to MIGRATE_CMA which can be done by | |
56 | * __free_pageblock_cma() function. What is important though | |
57 | * is that a range of pageblocks must be aligned to | |
58 | * MAX_ORDER_NR_PAGES should biggest page be bigger then | |
59 | * a single pageblock. | |
60 | */ | |
61 | MIGRATE_CMA, | |
62 | #endif | |
194159fb | 63 | #ifdef CONFIG_MEMORY_ISOLATION |
47118af0 | 64 | MIGRATE_ISOLATE, /* can't allocate from here */ |
194159fb | 65 | #endif |
47118af0 MN |
66 | MIGRATE_TYPES |
67 | }; | |
68 | ||
60f30350 | 69 | /* In mm/page_alloc.c; keep in sync also with show_migration_types() there */ |
c999fbd3 | 70 | extern const char * const migratetype_names[MIGRATE_TYPES]; |
60f30350 | 71 | |
47118af0 MN |
72 | #ifdef CONFIG_CMA |
73 | # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA) | |
7c15d9bb | 74 | # define is_migrate_cma_page(_page) (get_pageblock_migratetype(_page) == MIGRATE_CMA) |
47118af0 MN |
75 | #else |
76 | # define is_migrate_cma(migratetype) false | |
7c15d9bb | 77 | # define is_migrate_cma_page(_page) false |
47118af0 | 78 | #endif |
b2a0ac88 | 79 | |
b682debd VB |
80 | static inline bool is_migrate_movable(int mt) |
81 | { | |
82 | return is_migrate_cma(mt) || mt == MIGRATE_MOVABLE; | |
83 | } | |
84 | ||
b2a0ac88 MG |
85 | #define for_each_migratetype_order(order, type) \ |
86 | for (order = 0; order < MAX_ORDER; order++) \ | |
87 | for (type = 0; type < MIGRATE_TYPES; type++) | |
88 | ||
467c996c MG |
89 | extern int page_group_by_mobility_disabled; |
90 | ||
e58469ba MG |
91 | #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1) |
92 | #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1) | |
93 | ||
dc4b0caf MG |
94 | #define get_pageblock_migratetype(page) \ |
95 | get_pfnblock_flags_mask(page, page_to_pfn(page), \ | |
96 | PB_migrate_end, MIGRATETYPE_MASK) | |
97 | ||
1da177e4 | 98 | struct free_area { |
b2a0ac88 | 99 | struct list_head free_list[MIGRATE_TYPES]; |
1da177e4 LT |
100 | unsigned long nr_free; |
101 | }; | |
102 | ||
b03641af DW |
103 | /* Used for pages not on another list */ |
104 | static inline void add_to_free_area(struct page *page, struct free_area *area, | |
105 | int migratetype) | |
106 | { | |
107 | list_add(&page->lru, &area->free_list[migratetype]); | |
108 | area->nr_free++; | |
109 | } | |
110 | ||
111 | /* Used for pages not on another list */ | |
112 | static inline void add_to_free_area_tail(struct page *page, struct free_area *area, | |
113 | int migratetype) | |
114 | { | |
115 | list_add_tail(&page->lru, &area->free_list[migratetype]); | |
116 | area->nr_free++; | |
117 | } | |
118 | ||
97500a4a DW |
119 | #ifdef CONFIG_SHUFFLE_PAGE_ALLOCATOR |
120 | /* Used to preserve page allocation order entropy */ | |
121 | void add_to_free_area_random(struct page *page, struct free_area *area, | |
122 | int migratetype); | |
123 | #else | |
124 | static inline void add_to_free_area_random(struct page *page, | |
125 | struct free_area *area, int migratetype) | |
126 | { | |
127 | add_to_free_area(page, area, migratetype); | |
128 | } | |
129 | #endif | |
130 | ||
b03641af DW |
131 | /* Used for pages which are on another list */ |
132 | static inline void move_to_free_area(struct page *page, struct free_area *area, | |
133 | int migratetype) | |
134 | { | |
135 | list_move(&page->lru, &area->free_list[migratetype]); | |
136 | } | |
137 | ||
138 | static inline struct page *get_page_from_free_area(struct free_area *area, | |
139 | int migratetype) | |
140 | { | |
141 | return list_first_entry_or_null(&area->free_list[migratetype], | |
142 | struct page, lru); | |
143 | } | |
144 | ||
145 | static inline void del_page_from_free_area(struct page *page, | |
146 | struct free_area *area) | |
147 | { | |
148 | list_del(&page->lru); | |
149 | __ClearPageBuddy(page); | |
150 | set_page_private(page, 0); | |
151 | area->nr_free--; | |
152 | } | |
153 | ||
154 | static inline bool free_area_empty(struct free_area *area, int migratetype) | |
155 | { | |
156 | return list_empty(&area->free_list[migratetype]); | |
157 | } | |
158 | ||
1da177e4 LT |
159 | struct pglist_data; |
160 | ||
161 | /* | |
a52633d8 | 162 | * zone->lock and the zone lru_lock are two of the hottest locks in the kernel. |
1da177e4 LT |
163 | * So add a wild amount of padding here to ensure that they fall into separate |
164 | * cachelines. There are very few zone structures in the machine, so space | |
165 | * consumption is not a concern here. | |
166 | */ | |
167 | #if defined(CONFIG_SMP) | |
168 | struct zone_padding { | |
169 | char x[0]; | |
22fc6ecc | 170 | } ____cacheline_internodealigned_in_smp; |
1da177e4 LT |
171 | #define ZONE_PADDING(name) struct zone_padding name; |
172 | #else | |
173 | #define ZONE_PADDING(name) | |
174 | #endif | |
175 | ||
3a321d2a KW |
176 | #ifdef CONFIG_NUMA |
177 | enum numa_stat_item { | |
178 | NUMA_HIT, /* allocated in intended node */ | |
179 | NUMA_MISS, /* allocated in non intended node */ | |
180 | NUMA_FOREIGN, /* was intended here, hit elsewhere */ | |
181 | NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */ | |
182 | NUMA_LOCAL, /* allocation from local node */ | |
183 | NUMA_OTHER, /* allocation from other node */ | |
184 | NR_VM_NUMA_STAT_ITEMS | |
185 | }; | |
186 | #else | |
187 | #define NR_VM_NUMA_STAT_ITEMS 0 | |
188 | #endif | |
189 | ||
2244b95a | 190 | enum zone_stat_item { |
51ed4491 | 191 | /* First 128 byte cacheline (assuming 64 bit words) */ |
d23ad423 | 192 | NR_FREE_PAGES, |
71c799f4 MK |
193 | NR_ZONE_LRU_BASE, /* Used only for compaction and reclaim retry */ |
194 | NR_ZONE_INACTIVE_ANON = NR_ZONE_LRU_BASE, | |
195 | NR_ZONE_ACTIVE_ANON, | |
196 | NR_ZONE_INACTIVE_FILE, | |
197 | NR_ZONE_ACTIVE_FILE, | |
198 | NR_ZONE_UNEVICTABLE, | |
5a1c84b4 | 199 | NR_ZONE_WRITE_PENDING, /* Count of dirty, writeback and unstable pages */ |
5344b7e6 | 200 | NR_MLOCK, /* mlock()ed pages found and moved off LRU */ |
51ed4491 | 201 | NR_PAGETABLE, /* used for pagetables */ |
d30dd8be | 202 | NR_KERNEL_STACK_KB, /* measured in KiB */ |
c6a7f572 | 203 | /* Second 128 byte cacheline */ |
d2c5e30c | 204 | NR_BOUNCE, |
91537fee MK |
205 | #if IS_ENABLED(CONFIG_ZSMALLOC) |
206 | NR_ZSPAGES, /* allocated in zsmalloc */ | |
ca889e6c | 207 | #endif |
d1ce749a | 208 | NR_FREE_CMA_PAGES, |
2244b95a CL |
209 | NR_VM_ZONE_STAT_ITEMS }; |
210 | ||
75ef7184 | 211 | enum node_stat_item { |
599d0c95 MG |
212 | NR_LRU_BASE, |
213 | NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */ | |
214 | NR_ACTIVE_ANON, /* " " " " " */ | |
215 | NR_INACTIVE_FILE, /* " " " " " */ | |
216 | NR_ACTIVE_FILE, /* " " " " " */ | |
217 | NR_UNEVICTABLE, /* " " " " " */ | |
4a87e2a2 RG |
218 | NR_SLAB_RECLAIMABLE, |
219 | NR_SLAB_UNRECLAIMABLE, | |
599d0c95 MG |
220 | NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */ |
221 | NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */ | |
68d48e6a | 222 | WORKINGSET_NODES, |
1e6b1085 MG |
223 | WORKINGSET_REFAULT, |
224 | WORKINGSET_ACTIVATE, | |
1899ad18 | 225 | WORKINGSET_RESTORE, |
1e6b1085 | 226 | WORKINGSET_NODERECLAIM, |
4b9d0fab | 227 | NR_ANON_MAPPED, /* Mapped anonymous pages */ |
50658e2e MG |
228 | NR_FILE_MAPPED, /* pagecache pages mapped into pagetables. |
229 | only modified from process context */ | |
11fb9989 MG |
230 | NR_FILE_PAGES, |
231 | NR_FILE_DIRTY, | |
232 | NR_WRITEBACK, | |
233 | NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */ | |
234 | NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */ | |
235 | NR_SHMEM_THPS, | |
236 | NR_SHMEM_PMDMAPPED, | |
60fbf0ab SL |
237 | NR_FILE_THPS, |
238 | NR_FILE_PMDMAPPED, | |
11fb9989 MG |
239 | NR_ANON_THPS, |
240 | NR_UNSTABLE_NFS, /* NFS unstable pages */ | |
c4a25635 MG |
241 | NR_VMSCAN_WRITE, |
242 | NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */ | |
243 | NR_DIRTIED, /* page dirtyings since bootup */ | |
244 | NR_WRITTEN, /* page writings since bootup */ | |
b29940c1 | 245 | NR_KERNEL_MISC_RECLAIMABLE, /* reclaimable non-slab kernel pages */ |
75ef7184 MG |
246 | NR_VM_NODE_STAT_ITEMS |
247 | }; | |
248 | ||
4f98a2fe RR |
249 | /* |
250 | * We do arithmetic on the LRU lists in various places in the code, | |
251 | * so it is important to keep the active lists LRU_ACTIVE higher in | |
252 | * the array than the corresponding inactive lists, and to keep | |
253 | * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists. | |
254 | * | |
255 | * This has to be kept in sync with the statistics in zone_stat_item | |
256 | * above and the descriptions in vmstat_text in mm/vmstat.c | |
257 | */ | |
258 | #define LRU_BASE 0 | |
259 | #define LRU_ACTIVE 1 | |
260 | #define LRU_FILE 2 | |
261 | ||
b69408e8 | 262 | enum lru_list { |
4f98a2fe RR |
263 | LRU_INACTIVE_ANON = LRU_BASE, |
264 | LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE, | |
265 | LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE, | |
266 | LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE, | |
894bc310 | 267 | LRU_UNEVICTABLE, |
894bc310 LS |
268 | NR_LRU_LISTS |
269 | }; | |
b69408e8 | 270 | |
4111304d | 271 | #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++) |
b69408e8 | 272 | |
4111304d | 273 | #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++) |
894bc310 | 274 | |
b91ac374 | 275 | static inline bool is_file_lru(enum lru_list lru) |
4f98a2fe | 276 | { |
4111304d | 277 | return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE); |
4f98a2fe RR |
278 | } |
279 | ||
b91ac374 | 280 | static inline bool is_active_lru(enum lru_list lru) |
b69408e8 | 281 | { |
4111304d | 282 | return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE); |
b69408e8 CL |
283 | } |
284 | ||
89abfab1 HD |
285 | struct zone_reclaim_stat { |
286 | /* | |
287 | * The pageout code in vmscan.c keeps track of how many of the | |
59f91e5d | 288 | * mem/swap backed and file backed pages are referenced. |
89abfab1 HD |
289 | * The higher the rotated/scanned ratio, the more valuable |
290 | * that cache is. | |
291 | * | |
292 | * The anon LRU stats live in [0], file LRU stats in [1] | |
293 | */ | |
294 | unsigned long recent_rotated[2]; | |
295 | unsigned long recent_scanned[2]; | |
296 | }; | |
297 | ||
1b05117d JW |
298 | enum lruvec_flags { |
299 | LRUVEC_CONGESTED, /* lruvec has many dirty pages | |
300 | * backed by a congested BDI | |
301 | */ | |
302 | }; | |
303 | ||
6290df54 | 304 | struct lruvec { |
23047a96 JW |
305 | struct list_head lists[NR_LRU_LISTS]; |
306 | struct zone_reclaim_stat reclaim_stat; | |
307 | /* Evictions & activations on the inactive file list */ | |
308 | atomic_long_t inactive_age; | |
2a2e4885 JW |
309 | /* Refaults at the time of last reclaim cycle */ |
310 | unsigned long refaults; | |
1b05117d JW |
311 | /* Various lruvec state flags (enum lruvec_flags) */ |
312 | unsigned long flags; | |
c255a458 | 313 | #ifdef CONFIG_MEMCG |
599d0c95 | 314 | struct pglist_data *pgdat; |
7f5e86c2 | 315 | #endif |
6290df54 JW |
316 | }; |
317 | ||
653e003d | 318 | /* Isolate unmapped pages */ |
f3fd4a61 | 319 | #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2) |
c8244935 | 320 | /* Isolate for asynchronous migration */ |
f3fd4a61 | 321 | #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4) |
e46a2879 MK |
322 | /* Isolate unevictable pages */ |
323 | #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8) | |
4356f21d MK |
324 | |
325 | /* LRU Isolation modes. */ | |
9efeccac | 326 | typedef unsigned __bitwise isolate_mode_t; |
4356f21d | 327 | |
41858966 MG |
328 | enum zone_watermarks { |
329 | WMARK_MIN, | |
330 | WMARK_LOW, | |
331 | WMARK_HIGH, | |
332 | NR_WMARK | |
333 | }; | |
334 | ||
1c30844d MG |
335 | #define min_wmark_pages(z) (z->_watermark[WMARK_MIN] + z->watermark_boost) |
336 | #define low_wmark_pages(z) (z->_watermark[WMARK_LOW] + z->watermark_boost) | |
337 | #define high_wmark_pages(z) (z->_watermark[WMARK_HIGH] + z->watermark_boost) | |
338 | #define wmark_pages(z, i) (z->_watermark[i] + z->watermark_boost) | |
41858966 | 339 | |
1da177e4 LT |
340 | struct per_cpu_pages { |
341 | int count; /* number of pages in the list */ | |
1da177e4 LT |
342 | int high; /* high watermark, emptying needed */ |
343 | int batch; /* chunk size for buddy add/remove */ | |
5f8dcc21 MG |
344 | |
345 | /* Lists of pages, one per migrate type stored on the pcp-lists */ | |
346 | struct list_head lists[MIGRATE_PCPTYPES]; | |
1da177e4 LT |
347 | }; |
348 | ||
349 | struct per_cpu_pageset { | |
3dfa5721 | 350 | struct per_cpu_pages pcp; |
4037d452 CL |
351 | #ifdef CONFIG_NUMA |
352 | s8 expire; | |
1d90ca89 | 353 | u16 vm_numa_stat_diff[NR_VM_NUMA_STAT_ITEMS]; |
4037d452 | 354 | #endif |
2244b95a | 355 | #ifdef CONFIG_SMP |
df9ecaba | 356 | s8 stat_threshold; |
2244b95a CL |
357 | s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS]; |
358 | #endif | |
99dcc3e5 | 359 | }; |
e7c8d5c9 | 360 | |
75ef7184 MG |
361 | struct per_cpu_nodestat { |
362 | s8 stat_threshold; | |
363 | s8 vm_node_stat_diff[NR_VM_NODE_STAT_ITEMS]; | |
364 | }; | |
365 | ||
97965478 CL |
366 | #endif /* !__GENERATING_BOUNDS.H */ |
367 | ||
2f1b6248 CL |
368 | enum zone_type { |
369 | /* | |
734f9246 NSJ |
370 | * ZONE_DMA and ZONE_DMA32 are used when there are peripherals not able |
371 | * to DMA to all of the addressable memory (ZONE_NORMAL). | |
372 | * On architectures where this area covers the whole 32 bit address | |
373 | * space ZONE_DMA32 is used. ZONE_DMA is left for the ones with smaller | |
374 | * DMA addressing constraints. This distinction is important as a 32bit | |
375 | * DMA mask is assumed when ZONE_DMA32 is defined. Some 64-bit | |
376 | * platforms may need both zones as they support peripherals with | |
377 | * different DMA addressing limitations. | |
378 | * | |
379 | * Some examples: | |
380 | * | |
381 | * - i386 and x86_64 have a fixed 16M ZONE_DMA and ZONE_DMA32 for the | |
382 | * rest of the lower 4G. | |
383 | * | |
384 | * - arm only uses ZONE_DMA, the size, up to 4G, may vary depending on | |
385 | * the specific device. | |
386 | * | |
387 | * - arm64 has a fixed 1G ZONE_DMA and ZONE_DMA32 for the rest of the | |
388 | * lower 4G. | |
2f1b6248 | 389 | * |
734f9246 NSJ |
390 | * - powerpc only uses ZONE_DMA, the size, up to 2G, may vary |
391 | * depending on the specific device. | |
2f1b6248 | 392 | * |
734f9246 | 393 | * - s390 uses ZONE_DMA fixed to the lower 2G. |
2f1b6248 | 394 | * |
734f9246 NSJ |
395 | * - ia64 and riscv only use ZONE_DMA32. |
396 | * | |
397 | * - parisc uses neither. | |
2f1b6248 | 398 | */ |
734f9246 | 399 | #ifdef CONFIG_ZONE_DMA |
2f1b6248 | 400 | ZONE_DMA, |
4b51d669 | 401 | #endif |
fb0e7942 | 402 | #ifdef CONFIG_ZONE_DMA32 |
2f1b6248 | 403 | ZONE_DMA32, |
fb0e7942 | 404 | #endif |
2f1b6248 CL |
405 | /* |
406 | * Normal addressable memory is in ZONE_NORMAL. DMA operations can be | |
407 | * performed on pages in ZONE_NORMAL if the DMA devices support | |
408 | * transfers to all addressable memory. | |
409 | */ | |
410 | ZONE_NORMAL, | |
e53ef38d | 411 | #ifdef CONFIG_HIGHMEM |
2f1b6248 CL |
412 | /* |
413 | * A memory area that is only addressable by the kernel through | |
414 | * mapping portions into its own address space. This is for example | |
415 | * used by i386 to allow the kernel to address the memory beyond | |
416 | * 900MB. The kernel will set up special mappings (page | |
417 | * table entries on i386) for each page that the kernel needs to | |
418 | * access. | |
419 | */ | |
420 | ZONE_HIGHMEM, | |
e53ef38d | 421 | #endif |
2a1e274a | 422 | ZONE_MOVABLE, |
033fbae9 DW |
423 | #ifdef CONFIG_ZONE_DEVICE |
424 | ZONE_DEVICE, | |
425 | #endif | |
97965478 | 426 | __MAX_NR_ZONES |
033fbae9 | 427 | |
2f1b6248 | 428 | }; |
1da177e4 | 429 | |
97965478 CL |
430 | #ifndef __GENERATING_BOUNDS_H |
431 | ||
1da177e4 | 432 | struct zone { |
3484b2de | 433 | /* Read-mostly fields */ |
41858966 MG |
434 | |
435 | /* zone watermarks, access with *_wmark_pages(zone) macros */ | |
a9214443 | 436 | unsigned long _watermark[NR_WMARK]; |
1c30844d | 437 | unsigned long watermark_boost; |
41858966 | 438 | |
0aaa29a5 MG |
439 | unsigned long nr_reserved_highatomic; |
440 | ||
1da177e4 | 441 | /* |
89903327 AM |
442 | * We don't know if the memory that we're going to allocate will be |
443 | * freeable or/and it will be released eventually, so to avoid totally | |
444 | * wasting several GB of ram we must reserve some of the lower zone | |
445 | * memory (otherwise we risk to run OOM on the lower zones despite | |
446 | * there being tons of freeable ram on the higher zones). This array is | |
447 | * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl | |
448 | * changes. | |
1da177e4 | 449 | */ |
3484b2de | 450 | long lowmem_reserve[MAX_NR_ZONES]; |
ab8fabd4 | 451 | |
e7c8d5c9 | 452 | #ifdef CONFIG_NUMA |
d5f541ed | 453 | int node; |
3484b2de | 454 | #endif |
3484b2de | 455 | struct pglist_data *zone_pgdat; |
43cf38eb | 456 | struct per_cpu_pageset __percpu *pageset; |
3484b2de | 457 | |
835c134e MG |
458 | #ifndef CONFIG_SPARSEMEM |
459 | /* | |
d9c23400 | 460 | * Flags for a pageblock_nr_pages block. See pageblock-flags.h. |
835c134e MG |
461 | * In SPARSEMEM, this map is stored in struct mem_section |
462 | */ | |
463 | unsigned long *pageblock_flags; | |
464 | #endif /* CONFIG_SPARSEMEM */ | |
465 | ||
1da177e4 LT |
466 | /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */ |
467 | unsigned long zone_start_pfn; | |
468 | ||
bdc8cb98 | 469 | /* |
9feedc9d JL |
470 | * spanned_pages is the total pages spanned by the zone, including |
471 | * holes, which is calculated as: | |
472 | * spanned_pages = zone_end_pfn - zone_start_pfn; | |
bdc8cb98 | 473 | * |
9feedc9d JL |
474 | * present_pages is physical pages existing within the zone, which |
475 | * is calculated as: | |
8761e31c | 476 | * present_pages = spanned_pages - absent_pages(pages in holes); |
9feedc9d JL |
477 | * |
478 | * managed_pages is present pages managed by the buddy system, which | |
479 | * is calculated as (reserved_pages includes pages allocated by the | |
480 | * bootmem allocator): | |
481 | * managed_pages = present_pages - reserved_pages; | |
482 | * | |
483 | * So present_pages may be used by memory hotplug or memory power | |
484 | * management logic to figure out unmanaged pages by checking | |
485 | * (present_pages - managed_pages). And managed_pages should be used | |
486 | * by page allocator and vm scanner to calculate all kinds of watermarks | |
487 | * and thresholds. | |
488 | * | |
489 | * Locking rules: | |
490 | * | |
491 | * zone_start_pfn and spanned_pages are protected by span_seqlock. | |
492 | * It is a seqlock because it has to be read outside of zone->lock, | |
493 | * and it is done in the main allocator path. But, it is written | |
494 | * quite infrequently. | |
495 | * | |
496 | * The span_seq lock is declared along with zone->lock because it is | |
bdc8cb98 DH |
497 | * frequently read in proximity to zone->lock. It's good to |
498 | * give them a chance of being in the same cacheline. | |
9feedc9d | 499 | * |
c3d5f5f0 | 500 | * Write access to present_pages at runtime should be protected by |
bfc8c901 VD |
501 | * mem_hotplug_begin/end(). Any reader who can't tolerant drift of |
502 | * present_pages should get_online_mems() to get a stable value. | |
bdc8cb98 | 503 | */ |
9705bea5 | 504 | atomic_long_t managed_pages; |
9feedc9d JL |
505 | unsigned long spanned_pages; |
506 | unsigned long present_pages; | |
3484b2de MG |
507 | |
508 | const char *name; | |
1da177e4 | 509 | |
ad53f92e JK |
510 | #ifdef CONFIG_MEMORY_ISOLATION |
511 | /* | |
512 | * Number of isolated pageblock. It is used to solve incorrect | |
513 | * freepage counting problem due to racy retrieving migratetype | |
514 | * of pageblock. Protected by zone->lock. | |
515 | */ | |
516 | unsigned long nr_isolate_pageblock; | |
517 | #endif | |
518 | ||
3484b2de MG |
519 | #ifdef CONFIG_MEMORY_HOTPLUG |
520 | /* see spanned/present_pages for more description */ | |
521 | seqlock_t span_seqlock; | |
522 | #endif | |
523 | ||
9dcb8b68 | 524 | int initialized; |
3484b2de | 525 | |
0f661148 | 526 | /* Write-intensive fields used from the page allocator */ |
3484b2de | 527 | ZONE_PADDING(_pad1_) |
0f661148 | 528 | |
3484b2de MG |
529 | /* free areas of different sizes */ |
530 | struct free_area free_area[MAX_ORDER]; | |
531 | ||
532 | /* zone flags, see below */ | |
533 | unsigned long flags; | |
534 | ||
0f661148 | 535 | /* Primarily protects free_area */ |
a368ab67 MG |
536 | spinlock_t lock; |
537 | ||
0f661148 | 538 | /* Write-intensive fields used by compaction and vmstats. */ |
3484b2de MG |
539 | ZONE_PADDING(_pad2_) |
540 | ||
3484b2de MG |
541 | /* |
542 | * When free pages are below this point, additional steps are taken | |
543 | * when reading the number of free pages to avoid per-cpu counter | |
544 | * drift allowing watermarks to be breached | |
545 | */ | |
546 | unsigned long percpu_drift_mark; | |
547 | ||
548 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA | |
549 | /* pfn where compaction free scanner should start */ | |
550 | unsigned long compact_cached_free_pfn; | |
551 | /* pfn where async and sync compaction migration scanner should start */ | |
552 | unsigned long compact_cached_migrate_pfn[2]; | |
e332f741 MG |
553 | unsigned long compact_init_migrate_pfn; |
554 | unsigned long compact_init_free_pfn; | |
3484b2de MG |
555 | #endif |
556 | ||
557 | #ifdef CONFIG_COMPACTION | |
558 | /* | |
559 | * On compaction failure, 1<<compact_defer_shift compactions | |
560 | * are skipped before trying again. The number attempted since | |
561 | * last failure is tracked with compact_considered. | |
562 | */ | |
563 | unsigned int compact_considered; | |
564 | unsigned int compact_defer_shift; | |
565 | int compact_order_failed; | |
566 | #endif | |
567 | ||
568 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA | |
569 | /* Set to true when the PG_migrate_skip bits should be cleared */ | |
570 | bool compact_blockskip_flush; | |
571 | #endif | |
572 | ||
7cf91a98 JK |
573 | bool contiguous; |
574 | ||
3484b2de MG |
575 | ZONE_PADDING(_pad3_) |
576 | /* Zone statistics */ | |
577 | atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS]; | |
3a321d2a | 578 | atomic_long_t vm_numa_stat[NR_VM_NUMA_STAT_ITEMS]; |
22fc6ecc | 579 | } ____cacheline_internodealigned_in_smp; |
1da177e4 | 580 | |
599d0c95 | 581 | enum pgdat_flags { |
599d0c95 | 582 | PGDAT_DIRTY, /* reclaim scanning has recently found |
d43006d5 MG |
583 | * many dirty file pages at the tail |
584 | * of the LRU. | |
585 | */ | |
599d0c95 | 586 | PGDAT_WRITEBACK, /* reclaim scanning has recently found |
283aba9f MG |
587 | * many pages under writeback |
588 | */ | |
a5f5f91d | 589 | PGDAT_RECLAIM_LOCKED, /* prevents concurrent reclaim */ |
57054651 | 590 | }; |
e815af95 | 591 | |
73444bc4 MG |
592 | enum zone_flags { |
593 | ZONE_BOOSTED_WATERMARK, /* zone recently boosted watermarks. | |
594 | * Cleared when kswapd is woken. | |
595 | */ | |
596 | }; | |
597 | ||
9705bea5 AK |
598 | static inline unsigned long zone_managed_pages(struct zone *zone) |
599 | { | |
600 | return (unsigned long)atomic_long_read(&zone->managed_pages); | |
601 | } | |
602 | ||
f9228b20 | 603 | static inline unsigned long zone_end_pfn(const struct zone *zone) |
108bcc96 CS |
604 | { |
605 | return zone->zone_start_pfn + zone->spanned_pages; | |
606 | } | |
607 | ||
608 | static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn) | |
609 | { | |
610 | return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone); | |
611 | } | |
612 | ||
2a6e3ebe CS |
613 | static inline bool zone_is_initialized(struct zone *zone) |
614 | { | |
9dcb8b68 | 615 | return zone->initialized; |
2a6e3ebe CS |
616 | } |
617 | ||
618 | static inline bool zone_is_empty(struct zone *zone) | |
619 | { | |
620 | return zone->spanned_pages == 0; | |
621 | } | |
622 | ||
f1dd2cd1 MH |
623 | /* |
624 | * Return true if [start_pfn, start_pfn + nr_pages) range has a non-empty | |
625 | * intersection with the given zone | |
626 | */ | |
627 | static inline bool zone_intersects(struct zone *zone, | |
628 | unsigned long start_pfn, unsigned long nr_pages) | |
629 | { | |
630 | if (zone_is_empty(zone)) | |
631 | return false; | |
632 | if (start_pfn >= zone_end_pfn(zone) || | |
633 | start_pfn + nr_pages <= zone->zone_start_pfn) | |
634 | return false; | |
635 | ||
636 | return true; | |
637 | } | |
638 | ||
1da177e4 LT |
639 | /* |
640 | * The "priority" of VM scanning is how much of the queues we will scan in one | |
641 | * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the | |
642 | * queues ("queue_length >> 12") during an aging round. | |
643 | */ | |
644 | #define DEF_PRIORITY 12 | |
645 | ||
9276b1bc PJ |
646 | /* Maximum number of zones on a zonelist */ |
647 | #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES) | |
648 | ||
c00eb15a YB |
649 | enum { |
650 | ZONELIST_FALLBACK, /* zonelist with fallback */ | |
9276b1bc | 651 | #ifdef CONFIG_NUMA |
c00eb15a YB |
652 | /* |
653 | * The NUMA zonelists are doubled because we need zonelists that | |
654 | * restrict the allocations to a single node for __GFP_THISNODE. | |
655 | */ | |
656 | ZONELIST_NOFALLBACK, /* zonelist without fallback (__GFP_THISNODE) */ | |
9276b1bc | 657 | #endif |
c00eb15a YB |
658 | MAX_ZONELISTS |
659 | }; | |
9276b1bc | 660 | |
dd1a239f MG |
661 | /* |
662 | * This struct contains information about a zone in a zonelist. It is stored | |
663 | * here to avoid dereferences into large structures and lookups of tables | |
664 | */ | |
665 | struct zoneref { | |
666 | struct zone *zone; /* Pointer to actual zone */ | |
667 | int zone_idx; /* zone_idx(zoneref->zone) */ | |
668 | }; | |
669 | ||
1da177e4 LT |
670 | /* |
671 | * One allocation request operates on a zonelist. A zonelist | |
672 | * is a list of zones, the first one is the 'goal' of the | |
673 | * allocation, the other zones are fallback zones, in decreasing | |
674 | * priority. | |
675 | * | |
dd1a239f MG |
676 | * To speed the reading of the zonelist, the zonerefs contain the zone index |
677 | * of the entry being read. Helper functions to access information given | |
678 | * a struct zoneref are | |
679 | * | |
680 | * zonelist_zone() - Return the struct zone * for an entry in _zonerefs | |
681 | * zonelist_zone_idx() - Return the index of the zone for an entry | |
682 | * zonelist_node_idx() - Return the index of the node for an entry | |
1da177e4 LT |
683 | */ |
684 | struct zonelist { | |
dd1a239f | 685 | struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1]; |
1da177e4 LT |
686 | }; |
687 | ||
5b99cd0e HC |
688 | #ifndef CONFIG_DISCONTIGMEM |
689 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ | |
690 | extern struct page *mem_map; | |
691 | #endif | |
692 | ||
364c1eeb YS |
693 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
694 | struct deferred_split { | |
695 | spinlock_t split_queue_lock; | |
696 | struct list_head split_queue; | |
697 | unsigned long split_queue_len; | |
698 | }; | |
699 | #endif | |
700 | ||
1da177e4 | 701 | /* |
1da177e4 | 702 | * On NUMA machines, each NUMA node would have a pg_data_t to describe |
618b8c20 NB |
703 | * it's memory layout. On UMA machines there is a single pglist_data which |
704 | * describes the whole memory. | |
1da177e4 LT |
705 | * |
706 | * Memory statistics and page replacement data structures are maintained on a | |
707 | * per-zone basis. | |
708 | */ | |
709 | struct bootmem_data; | |
710 | typedef struct pglist_data { | |
711 | struct zone node_zones[MAX_NR_ZONES]; | |
523b9458 | 712 | struct zonelist node_zonelists[MAX_ZONELISTS]; |
1da177e4 | 713 | int nr_zones; |
52d4b9ac | 714 | #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */ |
1da177e4 | 715 | struct page *node_mem_map; |
eefa864b JK |
716 | #ifdef CONFIG_PAGE_EXTENSION |
717 | struct page_ext *node_page_ext; | |
718 | #endif | |
d41dee36 | 719 | #endif |
3a2d7fa8 | 720 | #if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT) |
208d54e5 | 721 | /* |
fa004ab7 WY |
722 | * Must be held any time you expect node_start_pfn, |
723 | * node_present_pages, node_spanned_pages or nr_zones to stay constant. | |
208d54e5 | 724 | * |
114d4b79 | 725 | * pgdat_resize_lock() and pgdat_resize_unlock() are provided to |
3a2d7fa8 PT |
726 | * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG |
727 | * or CONFIG_DEFERRED_STRUCT_PAGE_INIT. | |
114d4b79 | 728 | * |
72c3b51b | 729 | * Nests above zone->lock and zone->span_seqlock |
208d54e5 DH |
730 | */ |
731 | spinlock_t node_size_lock; | |
732 | #endif | |
1da177e4 LT |
733 | unsigned long node_start_pfn; |
734 | unsigned long node_present_pages; /* total number of physical pages */ | |
735 | unsigned long node_spanned_pages; /* total size of physical page | |
736 | range, including holes */ | |
737 | int node_id; | |
1da177e4 | 738 | wait_queue_head_t kswapd_wait; |
5515061d | 739 | wait_queue_head_t pfmemalloc_wait; |
bfc8c901 VD |
740 | struct task_struct *kswapd; /* Protected by |
741 | mem_hotplug_begin/end() */ | |
38087d9b MG |
742 | int kswapd_order; |
743 | enum zone_type kswapd_classzone_idx; | |
744 | ||
c73322d0 JW |
745 | int kswapd_failures; /* Number of 'reclaimed == 0' runs */ |
746 | ||
698b1b30 VB |
747 | #ifdef CONFIG_COMPACTION |
748 | int kcompactd_max_order; | |
749 | enum zone_type kcompactd_classzone_idx; | |
750 | wait_queue_head_t kcompactd_wait; | |
751 | struct task_struct *kcompactd; | |
8177a420 | 752 | #endif |
281e3726 MG |
753 | /* |
754 | * This is a per-node reserve of pages that are not available | |
755 | * to userspace allocations. | |
756 | */ | |
757 | unsigned long totalreserve_pages; | |
758 | ||
a5f5f91d MG |
759 | #ifdef CONFIG_NUMA |
760 | /* | |
0a3c5772 | 761 | * node reclaim becomes active if more unmapped pages exist. |
a5f5f91d MG |
762 | */ |
763 | unsigned long min_unmapped_pages; | |
764 | unsigned long min_slab_pages; | |
765 | #endif /* CONFIG_NUMA */ | |
766 | ||
a52633d8 MG |
767 | /* Write-intensive fields used by page reclaim */ |
768 | ZONE_PADDING(_pad1_) | |
769 | spinlock_t lru_lock; | |
3a80a7fa MG |
770 | |
771 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT | |
772 | /* | |
773 | * If memory initialisation on large machines is deferred then this | |
774 | * is the first PFN that needs to be initialised. | |
775 | */ | |
776 | unsigned long first_deferred_pfn; | |
777 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
a3d0a918 KS |
778 | |
779 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
364c1eeb | 780 | struct deferred_split deferred_split_queue; |
a3d0a918 | 781 | #endif |
75ef7184 | 782 | |
599d0c95 | 783 | /* Fields commonly accessed by the page reclaim scanner */ |
867e5e1d JW |
784 | |
785 | /* | |
786 | * NOTE: THIS IS UNUSED IF MEMCG IS ENABLED. | |
787 | * | |
788 | * Use mem_cgroup_lruvec() to look up lruvecs. | |
789 | */ | |
790 | struct lruvec __lruvec; | |
599d0c95 | 791 | |
599d0c95 MG |
792 | unsigned long flags; |
793 | ||
794 | ZONE_PADDING(_pad2_) | |
795 | ||
75ef7184 MG |
796 | /* Per-node vmstats */ |
797 | struct per_cpu_nodestat __percpu *per_cpu_nodestats; | |
798 | atomic_long_t vm_stat[NR_VM_NODE_STAT_ITEMS]; | |
1da177e4 LT |
799 | } pg_data_t; |
800 | ||
801 | #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) | |
802 | #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) | |
d41dee36 | 803 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
408fde81 | 804 | #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) |
d41dee36 AW |
805 | #else |
806 | #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) | |
807 | #endif | |
408fde81 | 808 | #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr)) |
1da177e4 | 809 | |
c6830c22 | 810 | #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn) |
da3649e1 | 811 | #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid)) |
c6830c22 | 812 | |
da3649e1 CS |
813 | static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat) |
814 | { | |
815 | return pgdat->node_start_pfn + pgdat->node_spanned_pages; | |
816 | } | |
817 | ||
818 | static inline bool pgdat_is_empty(pg_data_t *pgdat) | |
819 | { | |
820 | return !pgdat->node_start_pfn && !pgdat->node_spanned_pages; | |
821 | } | |
c6830c22 | 822 | |
208d54e5 DH |
823 | #include <linux/memory_hotplug.h> |
824 | ||
72675e13 | 825 | void build_all_zonelists(pg_data_t *pgdat); |
5ecd9d40 DR |
826 | void wakeup_kswapd(struct zone *zone, gfp_t gfp_mask, int order, |
827 | enum zone_type classzone_idx); | |
86a294a8 MH |
828 | bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark, |
829 | int classzone_idx, unsigned int alloc_flags, | |
830 | long free_pages); | |
7aeb09f9 | 831 | bool zone_watermark_ok(struct zone *z, unsigned int order, |
c603844b MG |
832 | unsigned long mark, int classzone_idx, |
833 | unsigned int alloc_flags); | |
7aeb09f9 | 834 | bool zone_watermark_ok_safe(struct zone *z, unsigned int order, |
e2b19197 | 835 | unsigned long mark, int classzone_idx); |
a2f3aa02 DH |
836 | enum memmap_context { |
837 | MEMMAP_EARLY, | |
838 | MEMMAP_HOTPLUG, | |
839 | }; | |
dc0bbf3b | 840 | extern void init_currently_empty_zone(struct zone *zone, unsigned long start_pfn, |
b171e409 | 841 | unsigned long size); |
718127cc | 842 | |
bea8c150 | 843 | extern void lruvec_init(struct lruvec *lruvec); |
7f5e86c2 | 844 | |
599d0c95 | 845 | static inline struct pglist_data *lruvec_pgdat(struct lruvec *lruvec) |
7f5e86c2 | 846 | { |
c255a458 | 847 | #ifdef CONFIG_MEMCG |
599d0c95 | 848 | return lruvec->pgdat; |
7f5e86c2 | 849 | #else |
867e5e1d | 850 | return container_of(lruvec, struct pglist_data, __lruvec); |
7f5e86c2 KK |
851 | #endif |
852 | } | |
853 | ||
fd538803 | 854 | extern unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru, int zone_idx); |
23047a96 | 855 | |
1da177e4 LT |
856 | #ifdef CONFIG_HAVE_MEMORY_PRESENT |
857 | void memory_present(int nid, unsigned long start, unsigned long end); | |
858 | #else | |
859 | static inline void memory_present(int nid, unsigned long start, unsigned long end) {} | |
860 | #endif | |
861 | ||
9def36e0 LG |
862 | #if defined(CONFIG_SPARSEMEM) |
863 | void memblocks_present(void); | |
864 | #else | |
865 | static inline void memblocks_present(void) {} | |
866 | #endif | |
867 | ||
7aac7898 LS |
868 | #ifdef CONFIG_HAVE_MEMORYLESS_NODES |
869 | int local_memory_node(int node_id); | |
870 | #else | |
871 | static inline int local_memory_node(int node_id) { return node_id; }; | |
872 | #endif | |
873 | ||
1da177e4 LT |
874 | /* |
875 | * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc. | |
876 | */ | |
877 | #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones) | |
878 | ||
6aa303de MG |
879 | /* |
880 | * Returns true if a zone has pages managed by the buddy allocator. | |
881 | * All the reclaim decisions have to use this function rather than | |
882 | * populated_zone(). If the whole zone is reserved then we can easily | |
883 | * end up with populated_zone() && !managed_zone(). | |
884 | */ | |
885 | static inline bool managed_zone(struct zone *zone) | |
886 | { | |
9705bea5 | 887 | return zone_managed_pages(zone); |
6aa303de MG |
888 | } |
889 | ||
890 | /* Returns true if a zone has memory */ | |
891 | static inline bool populated_zone(struct zone *zone) | |
f3fe6512 | 892 | { |
6aa303de | 893 | return zone->present_pages; |
f3fe6512 CK |
894 | } |
895 | ||
c1093b74 PT |
896 | #ifdef CONFIG_NUMA |
897 | static inline int zone_to_nid(struct zone *zone) | |
898 | { | |
899 | return zone->node; | |
900 | } | |
901 | ||
902 | static inline void zone_set_nid(struct zone *zone, int nid) | |
903 | { | |
904 | zone->node = nid; | |
905 | } | |
906 | #else | |
907 | static inline int zone_to_nid(struct zone *zone) | |
908 | { | |
909 | return 0; | |
910 | } | |
911 | ||
912 | static inline void zone_set_nid(struct zone *zone, int nid) {} | |
913 | #endif | |
914 | ||
2a1e274a MG |
915 | extern int movable_zone; |
916 | ||
d7e4a2ea | 917 | #ifdef CONFIG_HIGHMEM |
2a1e274a MG |
918 | static inline int zone_movable_is_highmem(void) |
919 | { | |
d7e4a2ea | 920 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
2a1e274a MG |
921 | return movable_zone == ZONE_HIGHMEM; |
922 | #else | |
d7e4a2ea | 923 | return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM; |
2a1e274a MG |
924 | #endif |
925 | } | |
d7e4a2ea | 926 | #endif |
2a1e274a | 927 | |
2f1b6248 | 928 | static inline int is_highmem_idx(enum zone_type idx) |
1da177e4 | 929 | { |
e53ef38d | 930 | #ifdef CONFIG_HIGHMEM |
2a1e274a MG |
931 | return (idx == ZONE_HIGHMEM || |
932 | (idx == ZONE_MOVABLE && zone_movable_is_highmem())); | |
e53ef38d CL |
933 | #else |
934 | return 0; | |
935 | #endif | |
1da177e4 LT |
936 | } |
937 | ||
1da177e4 | 938 | /** |
b4a991ec | 939 | * is_highmem - helper function to quickly check if a struct zone is a |
1da177e4 LT |
940 | * highmem zone or not. This is an attempt to keep references |
941 | * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum. | |
942 | * @zone - pointer to struct zone variable | |
943 | */ | |
944 | static inline int is_highmem(struct zone *zone) | |
945 | { | |
e53ef38d | 946 | #ifdef CONFIG_HIGHMEM |
29f9cb53 | 947 | return is_highmem_idx(zone_idx(zone)); |
e53ef38d CL |
948 | #else |
949 | return 0; | |
950 | #endif | |
1da177e4 LT |
951 | } |
952 | ||
1da177e4 LT |
953 | /* These two functions are used to setup the per zone pages min values */ |
954 | struct ctl_table; | |
8d65af78 | 955 | int min_free_kbytes_sysctl_handler(struct ctl_table *, int, |
1da177e4 | 956 | void __user *, size_t *, loff_t *); |
1c30844d MG |
957 | int watermark_boost_factor_sysctl_handler(struct ctl_table *, int, |
958 | void __user *, size_t *, loff_t *); | |
795ae7a0 JW |
959 | int watermark_scale_factor_sysctl_handler(struct ctl_table *, int, |
960 | void __user *, size_t *, loff_t *); | |
d3cda233 | 961 | extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES]; |
8d65af78 | 962 | int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, |
1da177e4 | 963 | void __user *, size_t *, loff_t *); |
8d65af78 | 964 | int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, |
8ad4b1fb | 965 | void __user *, size_t *, loff_t *); |
9614634f | 966 | int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int, |
8d65af78 | 967 | void __user *, size_t *, loff_t *); |
0ff38490 | 968 | int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int, |
8d65af78 | 969 | void __user *, size_t *, loff_t *); |
1da177e4 | 970 | |
f0c0b2b8 | 971 | extern int numa_zonelist_order_handler(struct ctl_table *, int, |
8d65af78 | 972 | void __user *, size_t *, loff_t *); |
f0c0b2b8 | 973 | extern char numa_zonelist_order[]; |
c9bff3ee | 974 | #define NUMA_ZONELIST_ORDER_LEN 16 |
f0c0b2b8 | 975 | |
93b7504e | 976 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
1da177e4 LT |
977 | |
978 | extern struct pglist_data contig_page_data; | |
979 | #define NODE_DATA(nid) (&contig_page_data) | |
980 | #define NODE_MEM_MAP(nid) mem_map | |
1da177e4 | 981 | |
93b7504e | 982 | #else /* CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 LT |
983 | |
984 | #include <asm/mmzone.h> | |
985 | ||
93b7504e | 986 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
348f8b6c | 987 | |
95144c78 KH |
988 | extern struct pglist_data *first_online_pgdat(void); |
989 | extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat); | |
990 | extern struct zone *next_zone(struct zone *zone); | |
8357f869 KH |
991 | |
992 | /** | |
12d15f0d | 993 | * for_each_online_pgdat - helper macro to iterate over all online nodes |
8357f869 KH |
994 | * @pgdat - pointer to a pg_data_t variable |
995 | */ | |
996 | #define for_each_online_pgdat(pgdat) \ | |
997 | for (pgdat = first_online_pgdat(); \ | |
998 | pgdat; \ | |
999 | pgdat = next_online_pgdat(pgdat)) | |
8357f869 KH |
1000 | /** |
1001 | * for_each_zone - helper macro to iterate over all memory zones | |
1002 | * @zone - pointer to struct zone variable | |
1003 | * | |
1004 | * The user only needs to declare the zone variable, for_each_zone | |
1005 | * fills it in. | |
1006 | */ | |
1007 | #define for_each_zone(zone) \ | |
1008 | for (zone = (first_online_pgdat())->node_zones; \ | |
1009 | zone; \ | |
1010 | zone = next_zone(zone)) | |
1011 | ||
ee99c71c KM |
1012 | #define for_each_populated_zone(zone) \ |
1013 | for (zone = (first_online_pgdat())->node_zones; \ | |
1014 | zone; \ | |
1015 | zone = next_zone(zone)) \ | |
1016 | if (!populated_zone(zone)) \ | |
1017 | ; /* do nothing */ \ | |
1018 | else | |
1019 | ||
dd1a239f MG |
1020 | static inline struct zone *zonelist_zone(struct zoneref *zoneref) |
1021 | { | |
1022 | return zoneref->zone; | |
1023 | } | |
1024 | ||
1025 | static inline int zonelist_zone_idx(struct zoneref *zoneref) | |
1026 | { | |
1027 | return zoneref->zone_idx; | |
1028 | } | |
1029 | ||
1030 | static inline int zonelist_node_idx(struct zoneref *zoneref) | |
1031 | { | |
c1093b74 | 1032 | return zone_to_nid(zoneref->zone); |
dd1a239f MG |
1033 | } |
1034 | ||
682a3385 MG |
1035 | struct zoneref *__next_zones_zonelist(struct zoneref *z, |
1036 | enum zone_type highest_zoneidx, | |
1037 | nodemask_t *nodes); | |
1038 | ||
19770b32 MG |
1039 | /** |
1040 | * 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 | |
1041 | * @z - The cursor used as a starting point for the search | |
1042 | * @highest_zoneidx - The zone index of the highest zone to return | |
1043 | * @nodes - An optional nodemask to filter the zonelist with | |
19770b32 MG |
1044 | * |
1045 | * This function returns the next zone at or below a given zone index that is | |
1046 | * within the allowed nodemask using a cursor as the starting point for the | |
5bead2a0 MG |
1047 | * search. The zoneref returned is a cursor that represents the current zone |
1048 | * being examined. It should be advanced by one before calling | |
1049 | * next_zones_zonelist again. | |
19770b32 | 1050 | */ |
682a3385 | 1051 | static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z, |
19770b32 | 1052 | enum zone_type highest_zoneidx, |
682a3385 MG |
1053 | nodemask_t *nodes) |
1054 | { | |
1055 | if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx)) | |
1056 | return z; | |
1057 | return __next_zones_zonelist(z, highest_zoneidx, nodes); | |
1058 | } | |
dd1a239f | 1059 | |
19770b32 MG |
1060 | /** |
1061 | * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist | |
1062 | * @zonelist - The zonelist to search for a suitable zone | |
1063 | * @highest_zoneidx - The zone index of the highest zone to return | |
1064 | * @nodes - An optional nodemask to filter the zonelist with | |
ea57485a | 1065 | * @return - Zoneref pointer for the first suitable zone found (see below) |
19770b32 MG |
1066 | * |
1067 | * This function returns the first zone at or below a given zone index that is | |
1068 | * within the allowed nodemask. The zoneref returned is a cursor that can be | |
5bead2a0 MG |
1069 | * used to iterate the zonelist with next_zones_zonelist by advancing it by |
1070 | * one before calling. | |
ea57485a VB |
1071 | * |
1072 | * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is | |
1073 | * never NULL). This may happen either genuinely, or due to concurrent nodemask | |
1074 | * update due to cpuset modification. | |
19770b32 | 1075 | */ |
dd1a239f | 1076 | static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist, |
19770b32 | 1077 | enum zone_type highest_zoneidx, |
c33d6c06 | 1078 | nodemask_t *nodes) |
54a6eb5c | 1079 | { |
c33d6c06 | 1080 | return next_zones_zonelist(zonelist->_zonerefs, |
05891fb0 | 1081 | highest_zoneidx, nodes); |
54a6eb5c MG |
1082 | } |
1083 | ||
19770b32 MG |
1084 | /** |
1085 | * 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 | |
1086 | * @zone - The current zone in the iterator | |
84218b55 | 1087 | * @z - The current pointer within zonelist->_zonerefs being iterated |
19770b32 MG |
1088 | * @zlist - The zonelist being iterated |
1089 | * @highidx - The zone index of the highest zone to return | |
1090 | * @nodemask - Nodemask allowed by the allocator | |
1091 | * | |
1092 | * This iterator iterates though all zones at or below a given zone index and | |
1093 | * within a given nodemask | |
1094 | */ | |
1095 | #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \ | |
c33d6c06 | 1096 | for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \ |
19770b32 | 1097 | zone; \ |
05891fb0 | 1098 | z = next_zones_zonelist(++z, highidx, nodemask), \ |
c33d6c06 MG |
1099 | zone = zonelist_zone(z)) |
1100 | ||
1101 | #define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \ | |
1102 | for (zone = z->zone; \ | |
1103 | zone; \ | |
1104 | z = next_zones_zonelist(++z, highidx, nodemask), \ | |
1105 | zone = zonelist_zone(z)) | |
1106 | ||
54a6eb5c MG |
1107 | |
1108 | /** | |
1109 | * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index | |
1110 | * @zone - The current zone in the iterator | |
1111 | * @z - The current pointer within zonelist->zones being iterated | |
1112 | * @zlist - The zonelist being iterated | |
1113 | * @highidx - The zone index of the highest zone to return | |
1114 | * | |
1115 | * This iterator iterates though all zones at or below a given zone index. | |
1116 | */ | |
1117 | #define for_each_zone_zonelist(zone, z, zlist, highidx) \ | |
19770b32 | 1118 | for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL) |
54a6eb5c | 1119 | |
d41dee36 AW |
1120 | #ifdef CONFIG_SPARSEMEM |
1121 | #include <asm/sparsemem.h> | |
1122 | #endif | |
1123 | ||
c713216d | 1124 | #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \ |
0ee332c1 | 1125 | !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) |
b4544568 AM |
1126 | static inline unsigned long early_pfn_to_nid(unsigned long pfn) |
1127 | { | |
9d1f4b3f | 1128 | BUILD_BUG_ON(IS_ENABLED(CONFIG_NUMA)); |
b4544568 AM |
1129 | return 0; |
1130 | } | |
b159d43f AW |
1131 | #endif |
1132 | ||
2bdaf115 AW |
1133 | #ifdef CONFIG_FLATMEM |
1134 | #define pfn_to_nid(pfn) (0) | |
1135 | #endif | |
1136 | ||
d41dee36 AW |
1137 | #ifdef CONFIG_SPARSEMEM |
1138 | ||
1139 | /* | |
1140 | * SECTION_SHIFT #bits space required to store a section # | |
1141 | * | |
1142 | * PA_SECTION_SHIFT physical address to/from section number | |
1143 | * PFN_SECTION_SHIFT pfn to/from section number | |
1144 | */ | |
d41dee36 AW |
1145 | #define PA_SECTION_SHIFT (SECTION_SIZE_BITS) |
1146 | #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) | |
1147 | ||
1148 | #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) | |
1149 | ||
1150 | #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) | |
1151 | #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) | |
1152 | ||
835c134e | 1153 | #define SECTION_BLOCKFLAGS_BITS \ |
d9c23400 | 1154 | ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS) |
835c134e | 1155 | |
d41dee36 AW |
1156 | #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS |
1157 | #error Allocator MAX_ORDER exceeds SECTION_SIZE | |
1158 | #endif | |
1159 | ||
1dd2bfc8 YI |
1160 | static inline unsigned long pfn_to_section_nr(unsigned long pfn) |
1161 | { | |
1162 | return pfn >> PFN_SECTION_SHIFT; | |
1163 | } | |
1164 | static inline unsigned long section_nr_to_pfn(unsigned long sec) | |
1165 | { | |
1166 | return sec << PFN_SECTION_SHIFT; | |
1167 | } | |
e3c40f37 | 1168 | |
a539f353 DK |
1169 | #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK) |
1170 | #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK) | |
1171 | ||
f1eca35a DW |
1172 | #define SUBSECTION_SHIFT 21 |
1173 | ||
1174 | #define PFN_SUBSECTION_SHIFT (SUBSECTION_SHIFT - PAGE_SHIFT) | |
1175 | #define PAGES_PER_SUBSECTION (1UL << PFN_SUBSECTION_SHIFT) | |
1176 | #define PAGE_SUBSECTION_MASK (~(PAGES_PER_SUBSECTION-1)) | |
1177 | ||
1178 | #if SUBSECTION_SHIFT > SECTION_SIZE_BITS | |
1179 | #error Subsection size exceeds section size | |
1180 | #else | |
1181 | #define SUBSECTIONS_PER_SECTION (1UL << (SECTION_SIZE_BITS - SUBSECTION_SHIFT)) | |
1182 | #endif | |
1183 | ||
a3619190 DW |
1184 | #define SUBSECTION_ALIGN_UP(pfn) ALIGN((pfn), PAGES_PER_SUBSECTION) |
1185 | #define SUBSECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SUBSECTION_MASK) | |
1186 | ||
f1eca35a DW |
1187 | struct mem_section_usage { |
1188 | DECLARE_BITMAP(subsection_map, SUBSECTIONS_PER_SECTION); | |
1189 | /* See declaration of similar field in struct zone */ | |
1190 | unsigned long pageblock_flags[0]; | |
1191 | }; | |
1192 | ||
f46edbd1 DW |
1193 | void subsection_map_init(unsigned long pfn, unsigned long nr_pages); |
1194 | ||
d41dee36 | 1195 | struct page; |
eefa864b | 1196 | struct page_ext; |
d41dee36 | 1197 | struct mem_section { |
29751f69 AW |
1198 | /* |
1199 | * This is, logically, a pointer to an array of struct | |
1200 | * pages. However, it is stored with some other magic. | |
1201 | * (see sparse.c::sparse_init_one_section()) | |
1202 | * | |
30c253e6 AW |
1203 | * Additionally during early boot we encode node id of |
1204 | * the location of the section here to guide allocation. | |
1205 | * (see sparse.c::memory_present()) | |
1206 | * | |
29751f69 AW |
1207 | * Making it a UL at least makes someone do a cast |
1208 | * before using it wrong. | |
1209 | */ | |
1210 | unsigned long section_mem_map; | |
5c0e3066 | 1211 | |
f1eca35a | 1212 | struct mem_section_usage *usage; |
eefa864b JK |
1213 | #ifdef CONFIG_PAGE_EXTENSION |
1214 | /* | |
0c9ad804 | 1215 | * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use |
eefa864b JK |
1216 | * section. (see page_ext.h about this.) |
1217 | */ | |
1218 | struct page_ext *page_ext; | |
1219 | unsigned long pad; | |
1220 | #endif | |
55878e88 CS |
1221 | /* |
1222 | * WARNING: mem_section must be a power-of-2 in size for the | |
1223 | * calculation and use of SECTION_ROOT_MASK to make sense. | |
1224 | */ | |
d41dee36 AW |
1225 | }; |
1226 | ||
3e347261 BP |
1227 | #ifdef CONFIG_SPARSEMEM_EXTREME |
1228 | #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section)) | |
1229 | #else | |
1230 | #define SECTIONS_PER_ROOT 1 | |
1231 | #endif | |
802f192e | 1232 | |
3e347261 | 1233 | #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT) |
0faa5638 | 1234 | #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT) |
3e347261 | 1235 | #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1) |
802f192e | 1236 | |
3e347261 | 1237 | #ifdef CONFIG_SPARSEMEM_EXTREME |
83e3c487 | 1238 | extern struct mem_section **mem_section; |
802f192e | 1239 | #else |
3e347261 BP |
1240 | extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT]; |
1241 | #endif | |
d41dee36 | 1242 | |
f1eca35a DW |
1243 | static inline unsigned long *section_to_usemap(struct mem_section *ms) |
1244 | { | |
1245 | return ms->usage->pageblock_flags; | |
1246 | } | |
1247 | ||
29751f69 AW |
1248 | static inline struct mem_section *__nr_to_section(unsigned long nr) |
1249 | { | |
83e3c487 KS |
1250 | #ifdef CONFIG_SPARSEMEM_EXTREME |
1251 | if (!mem_section) | |
1252 | return NULL; | |
1253 | #endif | |
3e347261 BP |
1254 | if (!mem_section[SECTION_NR_TO_ROOT(nr)]) |
1255 | return NULL; | |
1256 | return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK]; | |
29751f69 | 1257 | } |
2491f0a2 | 1258 | extern unsigned long __section_nr(struct mem_section *ms); |
f1eca35a | 1259 | extern size_t mem_section_usage_size(void); |
29751f69 AW |
1260 | |
1261 | /* | |
1262 | * We use the lower bits of the mem_map pointer to store | |
def9b71e PT |
1263 | * a little bit of information. The pointer is calculated |
1264 | * as mem_map - section_nr_to_pfn(pnum). The result is | |
1265 | * aligned to the minimum alignment of the two values: | |
1266 | * 1. All mem_map arrays are page-aligned. | |
1267 | * 2. section_nr_to_pfn() always clears PFN_SECTION_SHIFT | |
1268 | * lowest bits. PFN_SECTION_SHIFT is arch-specific | |
1269 | * (equal SECTION_SIZE_BITS - PAGE_SHIFT), and the | |
1270 | * worst combination is powerpc with 256k pages, | |
1271 | * which results in PFN_SECTION_SHIFT equal 6. | |
1272 | * To sum it up, at least 6 bits are available. | |
29751f69 AW |
1273 | */ |
1274 | #define SECTION_MARKED_PRESENT (1UL<<0) | |
1275 | #define SECTION_HAS_MEM_MAP (1UL<<1) | |
2d070eab | 1276 | #define SECTION_IS_ONLINE (1UL<<2) |
326e1b8f DW |
1277 | #define SECTION_IS_EARLY (1UL<<3) |
1278 | #define SECTION_MAP_LAST_BIT (1UL<<4) | |
29751f69 | 1279 | #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1)) |
2d070eab | 1280 | #define SECTION_NID_SHIFT 3 |
29751f69 AW |
1281 | |
1282 | static inline struct page *__section_mem_map_addr(struct mem_section *section) | |
1283 | { | |
1284 | unsigned long map = section->section_mem_map; | |
1285 | map &= SECTION_MAP_MASK; | |
1286 | return (struct page *)map; | |
1287 | } | |
1288 | ||
540557b9 | 1289 | static inline int present_section(struct mem_section *section) |
29751f69 | 1290 | { |
802f192e | 1291 | return (section && (section->section_mem_map & SECTION_MARKED_PRESENT)); |
29751f69 AW |
1292 | } |
1293 | ||
540557b9 AW |
1294 | static inline int present_section_nr(unsigned long nr) |
1295 | { | |
1296 | return present_section(__nr_to_section(nr)); | |
1297 | } | |
1298 | ||
1299 | static inline int valid_section(struct mem_section *section) | |
29751f69 | 1300 | { |
802f192e | 1301 | return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP)); |
29751f69 AW |
1302 | } |
1303 | ||
326e1b8f DW |
1304 | static inline int early_section(struct mem_section *section) |
1305 | { | |
1306 | return (section && (section->section_mem_map & SECTION_IS_EARLY)); | |
1307 | } | |
1308 | ||
29751f69 AW |
1309 | static inline int valid_section_nr(unsigned long nr) |
1310 | { | |
1311 | return valid_section(__nr_to_section(nr)); | |
1312 | } | |
1313 | ||
2d070eab MH |
1314 | static inline int online_section(struct mem_section *section) |
1315 | { | |
1316 | return (section && (section->section_mem_map & SECTION_IS_ONLINE)); | |
1317 | } | |
1318 | ||
1319 | static inline int online_section_nr(unsigned long nr) | |
1320 | { | |
1321 | return online_section(__nr_to_section(nr)); | |
1322 | } | |
1323 | ||
1324 | #ifdef CONFIG_MEMORY_HOTPLUG | |
1325 | void online_mem_sections(unsigned long start_pfn, unsigned long end_pfn); | |
1326 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
1327 | void offline_mem_sections(unsigned long start_pfn, unsigned long end_pfn); | |
1328 | #endif | |
1329 | #endif | |
1330 | ||
d41dee36 AW |
1331 | static inline struct mem_section *__pfn_to_section(unsigned long pfn) |
1332 | { | |
29751f69 | 1333 | return __nr_to_section(pfn_to_section_nr(pfn)); |
d41dee36 AW |
1334 | } |
1335 | ||
2491f0a2 | 1336 | extern unsigned long __highest_present_section_nr; |
c4e1be9e | 1337 | |
f46edbd1 DW |
1338 | static inline int subsection_map_index(unsigned long pfn) |
1339 | { | |
1340 | return (pfn & ~(PAGE_SECTION_MASK)) / PAGES_PER_SUBSECTION; | |
1341 | } | |
1342 | ||
1343 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | |
1344 | static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn) | |
1345 | { | |
1346 | int idx = subsection_map_index(pfn); | |
1347 | ||
1348 | return test_bit(idx, ms->usage->subsection_map); | |
1349 | } | |
1350 | #else | |
1351 | static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn) | |
1352 | { | |
1353 | return 1; | |
1354 | } | |
1355 | #endif | |
1356 | ||
7b7bf499 | 1357 | #ifndef CONFIG_HAVE_ARCH_PFN_VALID |
d41dee36 AW |
1358 | static inline int pfn_valid(unsigned long pfn) |
1359 | { | |
f46edbd1 DW |
1360 | struct mem_section *ms; |
1361 | ||
d41dee36 AW |
1362 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) |
1363 | return 0; | |
f46edbd1 DW |
1364 | ms = __nr_to_section(pfn_to_section_nr(pfn)); |
1365 | if (!valid_section(ms)) | |
1366 | return 0; | |
1367 | /* | |
1368 | * Traditionally early sections always returned pfn_valid() for | |
1369 | * the entire section-sized span. | |
1370 | */ | |
1371 | return early_section(ms) || pfn_section_valid(ms, pfn); | |
d41dee36 | 1372 | } |
7b7bf499 | 1373 | #endif |
d41dee36 | 1374 | |
540557b9 AW |
1375 | static inline int pfn_present(unsigned long pfn) |
1376 | { | |
1377 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) | |
1378 | return 0; | |
1379 | return present_section(__nr_to_section(pfn_to_section_nr(pfn))); | |
1380 | } | |
1381 | ||
d41dee36 AW |
1382 | /* |
1383 | * These are _only_ used during initialisation, therefore they | |
1384 | * can use __initdata ... They could have names to indicate | |
1385 | * this restriction. | |
1386 | */ | |
1387 | #ifdef CONFIG_NUMA | |
161599ff AW |
1388 | #define pfn_to_nid(pfn) \ |
1389 | ({ \ | |
1390 | unsigned long __pfn_to_nid_pfn = (pfn); \ | |
1391 | page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \ | |
1392 | }) | |
2bdaf115 AW |
1393 | #else |
1394 | #define pfn_to_nid(pfn) (0) | |
d41dee36 AW |
1395 | #endif |
1396 | ||
d41dee36 AW |
1397 | #define early_pfn_valid(pfn) pfn_valid(pfn) |
1398 | void sparse_init(void); | |
1399 | #else | |
1400 | #define sparse_init() do {} while (0) | |
28ae55c9 | 1401 | #define sparse_index_init(_sec, _nid) do {} while (0) |
e900a918 | 1402 | #define pfn_present pfn_valid |
f46edbd1 | 1403 | #define subsection_map_init(_pfn, _nr_pages) do {} while (0) |
d41dee36 AW |
1404 | #endif /* CONFIG_SPARSEMEM */ |
1405 | ||
8a942fde MG |
1406 | /* |
1407 | * During memory init memblocks map pfns to nids. The search is expensive and | |
1408 | * this caches recent lookups. The implementation of __early_pfn_to_nid | |
1409 | * may treat start/end as pfns or sections. | |
1410 | */ | |
1411 | struct mminit_pfnnid_cache { | |
1412 | unsigned long last_start; | |
1413 | unsigned long last_end; | |
1414 | int last_nid; | |
1415 | }; | |
1416 | ||
d41dee36 AW |
1417 | #ifndef early_pfn_valid |
1418 | #define early_pfn_valid(pfn) (1) | |
1419 | #endif | |
1420 | ||
1421 | void memory_present(int nid, unsigned long start, unsigned long end); | |
d41dee36 | 1422 | |
14e07298 AW |
1423 | /* |
1424 | * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we | |
8bb4e7a2 | 1425 | * need to check pfn validity within that MAX_ORDER_NR_PAGES block. |
14e07298 AW |
1426 | * pfn_valid_within() should be used in this case; we optimise this away |
1427 | * when we have no holes within a MAX_ORDER_NR_PAGES block. | |
1428 | */ | |
1429 | #ifdef CONFIG_HOLES_IN_ZONE | |
1430 | #define pfn_valid_within(pfn) pfn_valid(pfn) | |
1431 | #else | |
1432 | #define pfn_valid_within(pfn) (1) | |
1433 | #endif | |
1434 | ||
eb33575c MG |
1435 | #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL |
1436 | /* | |
1437 | * pfn_valid() is meant to be able to tell if a given PFN has valid memmap | |
2d070eab MH |
1438 | * associated with it or not. This means that a struct page exists for this |
1439 | * pfn. The caller cannot assume the page is fully initialized in general. | |
1440 | * Hotplugable pages might not have been onlined yet. pfn_to_online_page() | |
1441 | * will ensure the struct page is fully online and initialized. Special pages | |
1442 | * (e.g. ZONE_DEVICE) are never onlined and should be treated accordingly. | |
1443 | * | |
1444 | * In FLATMEM, it is expected that holes always have valid memmap as long as | |
1445 | * there is valid PFNs either side of the hole. In SPARSEMEM, it is assumed | |
1446 | * that a valid section has a memmap for the entire section. | |
eb33575c MG |
1447 | * |
1448 | * However, an ARM, and maybe other embedded architectures in the future | |
1449 | * free memmap backing holes to save memory on the assumption the memmap is | |
1450 | * never used. The page_zone linkages are then broken even though pfn_valid() | |
1451 | * returns true. A walker of the full memmap must then do this additional | |
1452 | * check to ensure the memmap they are looking at is sane by making sure | |
1453 | * the zone and PFN linkages are still valid. This is expensive, but walkers | |
1454 | * of the full memmap are extremely rare. | |
1455 | */ | |
5b80287a | 1456 | bool memmap_valid_within(unsigned long pfn, |
eb33575c MG |
1457 | struct page *page, struct zone *zone); |
1458 | #else | |
5b80287a | 1459 | static inline bool memmap_valid_within(unsigned long pfn, |
eb33575c MG |
1460 | struct page *page, struct zone *zone) |
1461 | { | |
5b80287a | 1462 | return true; |
eb33575c MG |
1463 | } |
1464 | #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */ | |
1465 | ||
97965478 | 1466 | #endif /* !__GENERATING_BOUNDS.H */ |
1da177e4 | 1467 | #endif /* !__ASSEMBLY__ */ |
1da177e4 | 1468 | #endif /* _LINUX_MMZONE_H */ |