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 | ||
d38ac97f | 91 | #define MIGRATETYPE_MASK ((1UL << PB_migratetype_bits) - 1) |
e58469ba | 92 | |
dc4b0caf | 93 | #define get_pageblock_migratetype(page) \ |
535b81e2 | 94 | get_pfnblock_flags_mask(page, page_to_pfn(page), MIGRATETYPE_MASK) |
dc4b0caf | 95 | |
1da177e4 | 96 | struct free_area { |
b2a0ac88 | 97 | struct list_head free_list[MIGRATE_TYPES]; |
1da177e4 LT |
98 | unsigned long nr_free; |
99 | }; | |
100 | ||
b03641af DW |
101 | static inline struct page *get_page_from_free_area(struct free_area *area, |
102 | int migratetype) | |
103 | { | |
104 | return list_first_entry_or_null(&area->free_list[migratetype], | |
105 | struct page, lru); | |
106 | } | |
107 | ||
b03641af DW |
108 | static inline bool free_area_empty(struct free_area *area, int migratetype) |
109 | { | |
110 | return list_empty(&area->free_list[migratetype]); | |
111 | } | |
112 | ||
1da177e4 LT |
113 | struct pglist_data; |
114 | ||
115 | /* | |
a52633d8 | 116 | * zone->lock and the zone lru_lock are two of the hottest locks in the kernel. |
1da177e4 LT |
117 | * So add a wild amount of padding here to ensure that they fall into separate |
118 | * cachelines. There are very few zone structures in the machine, so space | |
119 | * consumption is not a concern here. | |
120 | */ | |
121 | #if defined(CONFIG_SMP) | |
122 | struct zone_padding { | |
123 | char x[0]; | |
22fc6ecc | 124 | } ____cacheline_internodealigned_in_smp; |
1da177e4 LT |
125 | #define ZONE_PADDING(name) struct zone_padding name; |
126 | #else | |
127 | #define ZONE_PADDING(name) | |
128 | #endif | |
129 | ||
3a321d2a KW |
130 | #ifdef CONFIG_NUMA |
131 | enum numa_stat_item { | |
132 | NUMA_HIT, /* allocated in intended node */ | |
133 | NUMA_MISS, /* allocated in non intended node */ | |
134 | NUMA_FOREIGN, /* was intended here, hit elsewhere */ | |
135 | NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */ | |
136 | NUMA_LOCAL, /* allocation from local node */ | |
137 | NUMA_OTHER, /* allocation from other node */ | |
138 | NR_VM_NUMA_STAT_ITEMS | |
139 | }; | |
140 | #else | |
141 | #define NR_VM_NUMA_STAT_ITEMS 0 | |
142 | #endif | |
143 | ||
2244b95a | 144 | enum zone_stat_item { |
51ed4491 | 145 | /* First 128 byte cacheline (assuming 64 bit words) */ |
d23ad423 | 146 | NR_FREE_PAGES, |
71c799f4 MK |
147 | NR_ZONE_LRU_BASE, /* Used only for compaction and reclaim retry */ |
148 | NR_ZONE_INACTIVE_ANON = NR_ZONE_LRU_BASE, | |
149 | NR_ZONE_ACTIVE_ANON, | |
150 | NR_ZONE_INACTIVE_FILE, | |
151 | NR_ZONE_ACTIVE_FILE, | |
152 | NR_ZONE_UNEVICTABLE, | |
5a1c84b4 | 153 | NR_ZONE_WRITE_PENDING, /* Count of dirty, writeback and unstable pages */ |
5344b7e6 | 154 | NR_MLOCK, /* mlock()ed pages found and moved off LRU */ |
51ed4491 | 155 | NR_PAGETABLE, /* used for pagetables */ |
c6a7f572 | 156 | /* Second 128 byte cacheline */ |
d2c5e30c | 157 | NR_BOUNCE, |
91537fee MK |
158 | #if IS_ENABLED(CONFIG_ZSMALLOC) |
159 | NR_ZSPAGES, /* allocated in zsmalloc */ | |
ca889e6c | 160 | #endif |
d1ce749a | 161 | NR_FREE_CMA_PAGES, |
2244b95a CL |
162 | NR_VM_ZONE_STAT_ITEMS }; |
163 | ||
75ef7184 | 164 | enum node_stat_item { |
599d0c95 MG |
165 | NR_LRU_BASE, |
166 | NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */ | |
167 | NR_ACTIVE_ANON, /* " " " " " */ | |
168 | NR_INACTIVE_FILE, /* " " " " " */ | |
169 | NR_ACTIVE_FILE, /* " " " " " */ | |
170 | NR_UNEVICTABLE, /* " " " " " */ | |
d42f3245 RG |
171 | NR_SLAB_RECLAIMABLE_B, |
172 | NR_SLAB_UNRECLAIMABLE_B, | |
599d0c95 MG |
173 | NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */ |
174 | NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */ | |
68d48e6a | 175 | WORKINGSET_NODES, |
170b04b7 JK |
176 | WORKINGSET_REFAULT_BASE, |
177 | WORKINGSET_REFAULT_ANON = WORKINGSET_REFAULT_BASE, | |
178 | WORKINGSET_REFAULT_FILE, | |
179 | WORKINGSET_ACTIVATE_BASE, | |
180 | WORKINGSET_ACTIVATE_ANON = WORKINGSET_ACTIVATE_BASE, | |
181 | WORKINGSET_ACTIVATE_FILE, | |
182 | WORKINGSET_RESTORE_BASE, | |
183 | WORKINGSET_RESTORE_ANON = WORKINGSET_RESTORE_BASE, | |
184 | WORKINGSET_RESTORE_FILE, | |
1e6b1085 | 185 | WORKINGSET_NODERECLAIM, |
4b9d0fab | 186 | NR_ANON_MAPPED, /* Mapped anonymous pages */ |
50658e2e MG |
187 | NR_FILE_MAPPED, /* pagecache pages mapped into pagetables. |
188 | only modified from process context */ | |
11fb9989 MG |
189 | NR_FILE_PAGES, |
190 | NR_FILE_DIRTY, | |
191 | NR_WRITEBACK, | |
192 | NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */ | |
193 | NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */ | |
194 | NR_SHMEM_THPS, | |
195 | NR_SHMEM_PMDMAPPED, | |
60fbf0ab SL |
196 | NR_FILE_THPS, |
197 | NR_FILE_PMDMAPPED, | |
11fb9989 | 198 | NR_ANON_THPS, |
c4a25635 MG |
199 | NR_VMSCAN_WRITE, |
200 | NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */ | |
201 | NR_DIRTIED, /* page dirtyings since bootup */ | |
202 | NR_WRITTEN, /* page writings since bootup */ | |
b29940c1 | 203 | NR_KERNEL_MISC_RECLAIMABLE, /* reclaimable non-slab kernel pages */ |
1970dc6f JH |
204 | NR_FOLL_PIN_ACQUIRED, /* via: pin_user_page(), gup flag: FOLL_PIN */ |
205 | NR_FOLL_PIN_RELEASED, /* pages returned via unpin_user_page() */ | |
991e7673 SB |
206 | NR_KERNEL_STACK_KB, /* measured in KiB */ |
207 | #if IS_ENABLED(CONFIG_SHADOW_CALL_STACK) | |
208 | NR_KERNEL_SCS_KB, /* measured in KiB */ | |
209 | #endif | |
75ef7184 MG |
210 | NR_VM_NODE_STAT_ITEMS |
211 | }; | |
212 | ||
ea426c2a RG |
213 | /* |
214 | * Returns true if the value is measured in bytes (most vmstat values are | |
215 | * measured in pages). This defines the API part, the internal representation | |
216 | * might be different. | |
217 | */ | |
218 | static __always_inline bool vmstat_item_in_bytes(int idx) | |
219 | { | |
d42f3245 RG |
220 | /* |
221 | * Global and per-node slab counters track slab pages. | |
222 | * It's expected that changes are multiples of PAGE_SIZE. | |
223 | * Internally values are stored in pages. | |
224 | * | |
225 | * Per-memcg and per-lruvec counters track memory, consumed | |
226 | * by individual slab objects. These counters are actually | |
227 | * byte-precise. | |
228 | */ | |
229 | return (idx == NR_SLAB_RECLAIMABLE_B || | |
230 | idx == NR_SLAB_UNRECLAIMABLE_B); | |
ea426c2a RG |
231 | } |
232 | ||
4f98a2fe RR |
233 | /* |
234 | * We do arithmetic on the LRU lists in various places in the code, | |
235 | * so it is important to keep the active lists LRU_ACTIVE higher in | |
236 | * the array than the corresponding inactive lists, and to keep | |
237 | * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists. | |
238 | * | |
239 | * This has to be kept in sync with the statistics in zone_stat_item | |
240 | * above and the descriptions in vmstat_text in mm/vmstat.c | |
241 | */ | |
242 | #define LRU_BASE 0 | |
243 | #define LRU_ACTIVE 1 | |
244 | #define LRU_FILE 2 | |
245 | ||
b69408e8 | 246 | enum lru_list { |
4f98a2fe RR |
247 | LRU_INACTIVE_ANON = LRU_BASE, |
248 | LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE, | |
249 | LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE, | |
250 | LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE, | |
894bc310 | 251 | LRU_UNEVICTABLE, |
894bc310 LS |
252 | NR_LRU_LISTS |
253 | }; | |
b69408e8 | 254 | |
4111304d | 255 | #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++) |
b69408e8 | 256 | |
4111304d | 257 | #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++) |
894bc310 | 258 | |
b91ac374 | 259 | static inline bool is_file_lru(enum lru_list lru) |
4f98a2fe | 260 | { |
4111304d | 261 | return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE); |
4f98a2fe RR |
262 | } |
263 | ||
b91ac374 | 264 | static inline bool is_active_lru(enum lru_list lru) |
b69408e8 | 265 | { |
4111304d | 266 | return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE); |
b69408e8 CL |
267 | } |
268 | ||
1b05117d JW |
269 | enum lruvec_flags { |
270 | LRUVEC_CONGESTED, /* lruvec has many dirty pages | |
271 | * backed by a congested BDI | |
272 | */ | |
273 | }; | |
274 | ||
6290df54 | 275 | struct lruvec { |
23047a96 | 276 | struct list_head lists[NR_LRU_LISTS]; |
1431d4d1 JW |
277 | /* |
278 | * These track the cost of reclaiming one LRU - file or anon - | |
279 | * over the other. As the observed cost of reclaiming one LRU | |
280 | * increases, the reclaim scan balance tips toward the other. | |
281 | */ | |
282 | unsigned long anon_cost; | |
283 | unsigned long file_cost; | |
31d8fcac JW |
284 | /* Non-resident age, driven by LRU movement */ |
285 | atomic_long_t nonresident_age; | |
170b04b7 JK |
286 | /* Refaults at the time of last reclaim cycle, anon=0, file=1 */ |
287 | unsigned long refaults[2]; | |
1b05117d JW |
288 | /* Various lruvec state flags (enum lruvec_flags) */ |
289 | unsigned long flags; | |
c255a458 | 290 | #ifdef CONFIG_MEMCG |
599d0c95 | 291 | struct pglist_data *pgdat; |
7f5e86c2 | 292 | #endif |
6290df54 JW |
293 | }; |
294 | ||
653e003d | 295 | /* Isolate unmapped pages */ |
f3fd4a61 | 296 | #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2) |
c8244935 | 297 | /* Isolate for asynchronous migration */ |
f3fd4a61 | 298 | #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4) |
e46a2879 MK |
299 | /* Isolate unevictable pages */ |
300 | #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8) | |
4356f21d MK |
301 | |
302 | /* LRU Isolation modes. */ | |
9efeccac | 303 | typedef unsigned __bitwise isolate_mode_t; |
4356f21d | 304 | |
41858966 MG |
305 | enum zone_watermarks { |
306 | WMARK_MIN, | |
307 | WMARK_LOW, | |
308 | WMARK_HIGH, | |
309 | NR_WMARK | |
310 | }; | |
311 | ||
1c30844d MG |
312 | #define min_wmark_pages(z) (z->_watermark[WMARK_MIN] + z->watermark_boost) |
313 | #define low_wmark_pages(z) (z->_watermark[WMARK_LOW] + z->watermark_boost) | |
314 | #define high_wmark_pages(z) (z->_watermark[WMARK_HIGH] + z->watermark_boost) | |
315 | #define wmark_pages(z, i) (z->_watermark[i] + z->watermark_boost) | |
41858966 | 316 | |
1da177e4 LT |
317 | struct per_cpu_pages { |
318 | int count; /* number of pages in the list */ | |
1da177e4 LT |
319 | int high; /* high watermark, emptying needed */ |
320 | int batch; /* chunk size for buddy add/remove */ | |
5f8dcc21 MG |
321 | |
322 | /* Lists of pages, one per migrate type stored on the pcp-lists */ | |
323 | struct list_head lists[MIGRATE_PCPTYPES]; | |
1da177e4 LT |
324 | }; |
325 | ||
326 | struct per_cpu_pageset { | |
3dfa5721 | 327 | struct per_cpu_pages pcp; |
4037d452 CL |
328 | #ifdef CONFIG_NUMA |
329 | s8 expire; | |
1d90ca89 | 330 | u16 vm_numa_stat_diff[NR_VM_NUMA_STAT_ITEMS]; |
4037d452 | 331 | #endif |
2244b95a | 332 | #ifdef CONFIG_SMP |
df9ecaba | 333 | s8 stat_threshold; |
2244b95a CL |
334 | s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS]; |
335 | #endif | |
99dcc3e5 | 336 | }; |
e7c8d5c9 | 337 | |
75ef7184 MG |
338 | struct per_cpu_nodestat { |
339 | s8 stat_threshold; | |
340 | s8 vm_node_stat_diff[NR_VM_NODE_STAT_ITEMS]; | |
341 | }; | |
342 | ||
97965478 CL |
343 | #endif /* !__GENERATING_BOUNDS.H */ |
344 | ||
2f1b6248 CL |
345 | enum zone_type { |
346 | /* | |
734f9246 NSJ |
347 | * ZONE_DMA and ZONE_DMA32 are used when there are peripherals not able |
348 | * to DMA to all of the addressable memory (ZONE_NORMAL). | |
349 | * On architectures where this area covers the whole 32 bit address | |
350 | * space ZONE_DMA32 is used. ZONE_DMA is left for the ones with smaller | |
351 | * DMA addressing constraints. This distinction is important as a 32bit | |
352 | * DMA mask is assumed when ZONE_DMA32 is defined. Some 64-bit | |
353 | * platforms may need both zones as they support peripherals with | |
354 | * different DMA addressing limitations. | |
355 | * | |
356 | * Some examples: | |
357 | * | |
358 | * - i386 and x86_64 have a fixed 16M ZONE_DMA and ZONE_DMA32 for the | |
359 | * rest of the lower 4G. | |
360 | * | |
361 | * - arm only uses ZONE_DMA, the size, up to 4G, may vary depending on | |
362 | * the specific device. | |
363 | * | |
364 | * - arm64 has a fixed 1G ZONE_DMA and ZONE_DMA32 for the rest of the | |
365 | * lower 4G. | |
2f1b6248 | 366 | * |
734f9246 NSJ |
367 | * - powerpc only uses ZONE_DMA, the size, up to 2G, may vary |
368 | * depending on the specific device. | |
2f1b6248 | 369 | * |
734f9246 | 370 | * - s390 uses ZONE_DMA fixed to the lower 2G. |
2f1b6248 | 371 | * |
734f9246 NSJ |
372 | * - ia64 and riscv only use ZONE_DMA32. |
373 | * | |
374 | * - parisc uses neither. | |
2f1b6248 | 375 | */ |
734f9246 | 376 | #ifdef CONFIG_ZONE_DMA |
2f1b6248 | 377 | ZONE_DMA, |
4b51d669 | 378 | #endif |
fb0e7942 | 379 | #ifdef CONFIG_ZONE_DMA32 |
2f1b6248 | 380 | ZONE_DMA32, |
fb0e7942 | 381 | #endif |
2f1b6248 CL |
382 | /* |
383 | * Normal addressable memory is in ZONE_NORMAL. DMA operations can be | |
384 | * performed on pages in ZONE_NORMAL if the DMA devices support | |
385 | * transfers to all addressable memory. | |
386 | */ | |
387 | ZONE_NORMAL, | |
e53ef38d | 388 | #ifdef CONFIG_HIGHMEM |
2f1b6248 CL |
389 | /* |
390 | * A memory area that is only addressable by the kernel through | |
391 | * mapping portions into its own address space. This is for example | |
392 | * used by i386 to allow the kernel to address the memory beyond | |
393 | * 900MB. The kernel will set up special mappings (page | |
394 | * table entries on i386) for each page that the kernel needs to | |
395 | * access. | |
396 | */ | |
397 | ZONE_HIGHMEM, | |
e53ef38d | 398 | #endif |
9181a980 DH |
399 | /* |
400 | * ZONE_MOVABLE is similar to ZONE_NORMAL, except that it contains | |
401 | * movable pages with few exceptional cases described below. Main use | |
402 | * cases for ZONE_MOVABLE are to make memory offlining/unplug more | |
403 | * likely to succeed, and to locally limit unmovable allocations - e.g., | |
404 | * to increase the number of THP/huge pages. Notable special cases are: | |
405 | * | |
406 | * 1. Pinned pages: (long-term) pinning of movable pages might | |
407 | * essentially turn such pages unmovable. Memory offlining might | |
408 | * retry a long time. | |
409 | * 2. memblock allocations: kernelcore/movablecore setups might create | |
410 | * situations where ZONE_MOVABLE contains unmovable allocations | |
411 | * after boot. Memory offlining and allocations fail early. | |
412 | * 3. Memory holes: kernelcore/movablecore setups might create very rare | |
413 | * situations where ZONE_MOVABLE contains memory holes after boot, | |
414 | * for example, if we have sections that are only partially | |
415 | * populated. Memory offlining and allocations fail early. | |
416 | * 4. PG_hwpoison pages: while poisoned pages can be skipped during | |
417 | * memory offlining, such pages cannot be allocated. | |
418 | * 5. Unmovable PG_offline pages: in paravirtualized environments, | |
419 | * hotplugged memory blocks might only partially be managed by the | |
420 | * buddy (e.g., via XEN-balloon, Hyper-V balloon, virtio-mem). The | |
421 | * parts not manged by the buddy are unmovable PG_offline pages. In | |
422 | * some cases (virtio-mem), such pages can be skipped during | |
423 | * memory offlining, however, cannot be moved/allocated. These | |
424 | * techniques might use alloc_contig_range() to hide previously | |
425 | * exposed pages from the buddy again (e.g., to implement some sort | |
426 | * of memory unplug in virtio-mem). | |
427 | * | |
428 | * In general, no unmovable allocations that degrade memory offlining | |
429 | * should end up in ZONE_MOVABLE. Allocators (like alloc_contig_range()) | |
430 | * have to expect that migrating pages in ZONE_MOVABLE can fail (even | |
431 | * if has_unmovable_pages() states that there are no unmovable pages, | |
432 | * there can be false negatives). | |
433 | */ | |
2a1e274a | 434 | ZONE_MOVABLE, |
033fbae9 DW |
435 | #ifdef CONFIG_ZONE_DEVICE |
436 | ZONE_DEVICE, | |
437 | #endif | |
97965478 | 438 | __MAX_NR_ZONES |
033fbae9 | 439 | |
2f1b6248 | 440 | }; |
1da177e4 | 441 | |
97965478 CL |
442 | #ifndef __GENERATING_BOUNDS_H |
443 | ||
1da177e4 | 444 | struct zone { |
3484b2de | 445 | /* Read-mostly fields */ |
41858966 MG |
446 | |
447 | /* zone watermarks, access with *_wmark_pages(zone) macros */ | |
a9214443 | 448 | unsigned long _watermark[NR_WMARK]; |
1c30844d | 449 | unsigned long watermark_boost; |
41858966 | 450 | |
0aaa29a5 MG |
451 | unsigned long nr_reserved_highatomic; |
452 | ||
1da177e4 | 453 | /* |
89903327 AM |
454 | * We don't know if the memory that we're going to allocate will be |
455 | * freeable or/and it will be released eventually, so to avoid totally | |
456 | * wasting several GB of ram we must reserve some of the lower zone | |
457 | * memory (otherwise we risk to run OOM on the lower zones despite | |
458 | * there being tons of freeable ram on the higher zones). This array is | |
459 | * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl | |
460 | * changes. | |
1da177e4 | 461 | */ |
3484b2de | 462 | long lowmem_reserve[MAX_NR_ZONES]; |
ab8fabd4 | 463 | |
e7c8d5c9 | 464 | #ifdef CONFIG_NUMA |
d5f541ed | 465 | int node; |
3484b2de | 466 | #endif |
3484b2de | 467 | struct pglist_data *zone_pgdat; |
43cf38eb | 468 | struct per_cpu_pageset __percpu *pageset; |
3484b2de | 469 | |
835c134e MG |
470 | #ifndef CONFIG_SPARSEMEM |
471 | /* | |
d9c23400 | 472 | * Flags for a pageblock_nr_pages block. See pageblock-flags.h. |
835c134e MG |
473 | * In SPARSEMEM, this map is stored in struct mem_section |
474 | */ | |
475 | unsigned long *pageblock_flags; | |
476 | #endif /* CONFIG_SPARSEMEM */ | |
477 | ||
1da177e4 LT |
478 | /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */ |
479 | unsigned long zone_start_pfn; | |
480 | ||
bdc8cb98 | 481 | /* |
9feedc9d JL |
482 | * spanned_pages is the total pages spanned by the zone, including |
483 | * holes, which is calculated as: | |
484 | * spanned_pages = zone_end_pfn - zone_start_pfn; | |
bdc8cb98 | 485 | * |
9feedc9d JL |
486 | * present_pages is physical pages existing within the zone, which |
487 | * is calculated as: | |
8761e31c | 488 | * present_pages = spanned_pages - absent_pages(pages in holes); |
9feedc9d JL |
489 | * |
490 | * managed_pages is present pages managed by the buddy system, which | |
491 | * is calculated as (reserved_pages includes pages allocated by the | |
492 | * bootmem allocator): | |
493 | * managed_pages = present_pages - reserved_pages; | |
494 | * | |
495 | * So present_pages may be used by memory hotplug or memory power | |
496 | * management logic to figure out unmanaged pages by checking | |
497 | * (present_pages - managed_pages). And managed_pages should be used | |
498 | * by page allocator and vm scanner to calculate all kinds of watermarks | |
499 | * and thresholds. | |
500 | * | |
501 | * Locking rules: | |
502 | * | |
503 | * zone_start_pfn and spanned_pages are protected by span_seqlock. | |
504 | * It is a seqlock because it has to be read outside of zone->lock, | |
505 | * and it is done in the main allocator path. But, it is written | |
506 | * quite infrequently. | |
507 | * | |
508 | * The span_seq lock is declared along with zone->lock because it is | |
bdc8cb98 DH |
509 | * frequently read in proximity to zone->lock. It's good to |
510 | * give them a chance of being in the same cacheline. | |
9feedc9d | 511 | * |
c3d5f5f0 | 512 | * Write access to present_pages at runtime should be protected by |
bfc8c901 VD |
513 | * mem_hotplug_begin/end(). Any reader who can't tolerant drift of |
514 | * present_pages should get_online_mems() to get a stable value. | |
bdc8cb98 | 515 | */ |
9705bea5 | 516 | atomic_long_t managed_pages; |
9feedc9d JL |
517 | unsigned long spanned_pages; |
518 | unsigned long present_pages; | |
3484b2de MG |
519 | |
520 | const char *name; | |
1da177e4 | 521 | |
ad53f92e JK |
522 | #ifdef CONFIG_MEMORY_ISOLATION |
523 | /* | |
524 | * Number of isolated pageblock. It is used to solve incorrect | |
525 | * freepage counting problem due to racy retrieving migratetype | |
526 | * of pageblock. Protected by zone->lock. | |
527 | */ | |
528 | unsigned long nr_isolate_pageblock; | |
529 | #endif | |
530 | ||
3484b2de MG |
531 | #ifdef CONFIG_MEMORY_HOTPLUG |
532 | /* see spanned/present_pages for more description */ | |
533 | seqlock_t span_seqlock; | |
534 | #endif | |
535 | ||
9dcb8b68 | 536 | int initialized; |
3484b2de | 537 | |
0f661148 | 538 | /* Write-intensive fields used from the page allocator */ |
3484b2de | 539 | ZONE_PADDING(_pad1_) |
0f661148 | 540 | |
3484b2de MG |
541 | /* free areas of different sizes */ |
542 | struct free_area free_area[MAX_ORDER]; | |
543 | ||
544 | /* zone flags, see below */ | |
545 | unsigned long flags; | |
546 | ||
0f661148 | 547 | /* Primarily protects free_area */ |
a368ab67 MG |
548 | spinlock_t lock; |
549 | ||
0f661148 | 550 | /* Write-intensive fields used by compaction and vmstats. */ |
3484b2de MG |
551 | ZONE_PADDING(_pad2_) |
552 | ||
3484b2de MG |
553 | /* |
554 | * When free pages are below this point, additional steps are taken | |
555 | * when reading the number of free pages to avoid per-cpu counter | |
556 | * drift allowing watermarks to be breached | |
557 | */ | |
558 | unsigned long percpu_drift_mark; | |
559 | ||
560 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA | |
561 | /* pfn where compaction free scanner should start */ | |
562 | unsigned long compact_cached_free_pfn; | |
563 | /* pfn where async and sync compaction migration scanner should start */ | |
564 | unsigned long compact_cached_migrate_pfn[2]; | |
e332f741 MG |
565 | unsigned long compact_init_migrate_pfn; |
566 | unsigned long compact_init_free_pfn; | |
3484b2de MG |
567 | #endif |
568 | ||
569 | #ifdef CONFIG_COMPACTION | |
570 | /* | |
571 | * On compaction failure, 1<<compact_defer_shift compactions | |
572 | * are skipped before trying again. The number attempted since | |
573 | * last failure is tracked with compact_considered. | |
860b3272 | 574 | * compact_order_failed is the minimum compaction failed order. |
3484b2de MG |
575 | */ |
576 | unsigned int compact_considered; | |
577 | unsigned int compact_defer_shift; | |
578 | int compact_order_failed; | |
579 | #endif | |
580 | ||
581 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA | |
582 | /* Set to true when the PG_migrate_skip bits should be cleared */ | |
583 | bool compact_blockskip_flush; | |
584 | #endif | |
585 | ||
7cf91a98 JK |
586 | bool contiguous; |
587 | ||
3484b2de MG |
588 | ZONE_PADDING(_pad3_) |
589 | /* Zone statistics */ | |
590 | atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS]; | |
3a321d2a | 591 | atomic_long_t vm_numa_stat[NR_VM_NUMA_STAT_ITEMS]; |
22fc6ecc | 592 | } ____cacheline_internodealigned_in_smp; |
1da177e4 | 593 | |
599d0c95 | 594 | enum pgdat_flags { |
599d0c95 | 595 | PGDAT_DIRTY, /* reclaim scanning has recently found |
d43006d5 MG |
596 | * many dirty file pages at the tail |
597 | * of the LRU. | |
598 | */ | |
599d0c95 | 599 | PGDAT_WRITEBACK, /* reclaim scanning has recently found |
283aba9f MG |
600 | * many pages under writeback |
601 | */ | |
a5f5f91d | 602 | PGDAT_RECLAIM_LOCKED, /* prevents concurrent reclaim */ |
57054651 | 603 | }; |
e815af95 | 604 | |
73444bc4 MG |
605 | enum zone_flags { |
606 | ZONE_BOOSTED_WATERMARK, /* zone recently boosted watermarks. | |
607 | * Cleared when kswapd is woken. | |
608 | */ | |
609 | }; | |
610 | ||
9705bea5 AK |
611 | static inline unsigned long zone_managed_pages(struct zone *zone) |
612 | { | |
613 | return (unsigned long)atomic_long_read(&zone->managed_pages); | |
614 | } | |
615 | ||
f9228b20 | 616 | static inline unsigned long zone_end_pfn(const struct zone *zone) |
108bcc96 CS |
617 | { |
618 | return zone->zone_start_pfn + zone->spanned_pages; | |
619 | } | |
620 | ||
621 | static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn) | |
622 | { | |
623 | return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone); | |
624 | } | |
625 | ||
2a6e3ebe CS |
626 | static inline bool zone_is_initialized(struct zone *zone) |
627 | { | |
9dcb8b68 | 628 | return zone->initialized; |
2a6e3ebe CS |
629 | } |
630 | ||
631 | static inline bool zone_is_empty(struct zone *zone) | |
632 | { | |
633 | return zone->spanned_pages == 0; | |
634 | } | |
635 | ||
f1dd2cd1 MH |
636 | /* |
637 | * Return true if [start_pfn, start_pfn + nr_pages) range has a non-empty | |
638 | * intersection with the given zone | |
639 | */ | |
640 | static inline bool zone_intersects(struct zone *zone, | |
641 | unsigned long start_pfn, unsigned long nr_pages) | |
642 | { | |
643 | if (zone_is_empty(zone)) | |
644 | return false; | |
645 | if (start_pfn >= zone_end_pfn(zone) || | |
646 | start_pfn + nr_pages <= zone->zone_start_pfn) | |
647 | return false; | |
648 | ||
649 | return true; | |
650 | } | |
651 | ||
1da177e4 LT |
652 | /* |
653 | * The "priority" of VM scanning is how much of the queues we will scan in one | |
654 | * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the | |
655 | * queues ("queue_length >> 12") during an aging round. | |
656 | */ | |
657 | #define DEF_PRIORITY 12 | |
658 | ||
9276b1bc PJ |
659 | /* Maximum number of zones on a zonelist */ |
660 | #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES) | |
661 | ||
c00eb15a YB |
662 | enum { |
663 | ZONELIST_FALLBACK, /* zonelist with fallback */ | |
9276b1bc | 664 | #ifdef CONFIG_NUMA |
c00eb15a YB |
665 | /* |
666 | * The NUMA zonelists are doubled because we need zonelists that | |
667 | * restrict the allocations to a single node for __GFP_THISNODE. | |
668 | */ | |
669 | ZONELIST_NOFALLBACK, /* zonelist without fallback (__GFP_THISNODE) */ | |
9276b1bc | 670 | #endif |
c00eb15a YB |
671 | MAX_ZONELISTS |
672 | }; | |
9276b1bc | 673 | |
dd1a239f MG |
674 | /* |
675 | * This struct contains information about a zone in a zonelist. It is stored | |
676 | * here to avoid dereferences into large structures and lookups of tables | |
677 | */ | |
678 | struct zoneref { | |
679 | struct zone *zone; /* Pointer to actual zone */ | |
680 | int zone_idx; /* zone_idx(zoneref->zone) */ | |
681 | }; | |
682 | ||
1da177e4 LT |
683 | /* |
684 | * One allocation request operates on a zonelist. A zonelist | |
685 | * is a list of zones, the first one is the 'goal' of the | |
686 | * allocation, the other zones are fallback zones, in decreasing | |
687 | * priority. | |
688 | * | |
dd1a239f MG |
689 | * To speed the reading of the zonelist, the zonerefs contain the zone index |
690 | * of the entry being read. Helper functions to access information given | |
691 | * a struct zoneref are | |
692 | * | |
693 | * zonelist_zone() - Return the struct zone * for an entry in _zonerefs | |
694 | * zonelist_zone_idx() - Return the index of the zone for an entry | |
695 | * zonelist_node_idx() - Return the index of the node for an entry | |
1da177e4 LT |
696 | */ |
697 | struct zonelist { | |
dd1a239f | 698 | struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1]; |
1da177e4 LT |
699 | }; |
700 | ||
5b99cd0e HC |
701 | #ifndef CONFIG_DISCONTIGMEM |
702 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ | |
703 | extern struct page *mem_map; | |
704 | #endif | |
705 | ||
364c1eeb YS |
706 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
707 | struct deferred_split { | |
708 | spinlock_t split_queue_lock; | |
709 | struct list_head split_queue; | |
710 | unsigned long split_queue_len; | |
711 | }; | |
712 | #endif | |
713 | ||
1da177e4 | 714 | /* |
1da177e4 | 715 | * On NUMA machines, each NUMA node would have a pg_data_t to describe |
618b8c20 NB |
716 | * it's memory layout. On UMA machines there is a single pglist_data which |
717 | * describes the whole memory. | |
1da177e4 LT |
718 | * |
719 | * Memory statistics and page replacement data structures are maintained on a | |
720 | * per-zone basis. | |
721 | */ | |
1da177e4 | 722 | typedef struct pglist_data { |
496df3d3 BW |
723 | /* |
724 | * node_zones contains just the zones for THIS node. Not all of the | |
725 | * zones may be populated, but it is the full list. It is referenced by | |
726 | * this node's node_zonelists as well as other node's node_zonelists. | |
727 | */ | |
1da177e4 | 728 | struct zone node_zones[MAX_NR_ZONES]; |
496df3d3 BW |
729 | |
730 | /* | |
731 | * node_zonelists contains references to all zones in all nodes. | |
732 | * Generally the first zones will be references to this node's | |
733 | * node_zones. | |
734 | */ | |
523b9458 | 735 | struct zonelist node_zonelists[MAX_ZONELISTS]; |
496df3d3 BW |
736 | |
737 | int nr_zones; /* number of populated zones in this node */ | |
52d4b9ac | 738 | #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */ |
1da177e4 | 739 | struct page *node_mem_map; |
eefa864b JK |
740 | #ifdef CONFIG_PAGE_EXTENSION |
741 | struct page_ext *node_page_ext; | |
742 | #endif | |
d41dee36 | 743 | #endif |
3a2d7fa8 | 744 | #if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT) |
208d54e5 | 745 | /* |
fa004ab7 WY |
746 | * Must be held any time you expect node_start_pfn, |
747 | * node_present_pages, node_spanned_pages or nr_zones to stay constant. | |
3d060856 PT |
748 | * Also synchronizes pgdat->first_deferred_pfn during deferred page |
749 | * init. | |
208d54e5 | 750 | * |
114d4b79 | 751 | * pgdat_resize_lock() and pgdat_resize_unlock() are provided to |
3a2d7fa8 PT |
752 | * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG |
753 | * or CONFIG_DEFERRED_STRUCT_PAGE_INIT. | |
114d4b79 | 754 | * |
72c3b51b | 755 | * Nests above zone->lock and zone->span_seqlock |
208d54e5 DH |
756 | */ |
757 | spinlock_t node_size_lock; | |
758 | #endif | |
1da177e4 LT |
759 | unsigned long node_start_pfn; |
760 | unsigned long node_present_pages; /* total number of physical pages */ | |
761 | unsigned long node_spanned_pages; /* total size of physical page | |
762 | range, including holes */ | |
763 | int node_id; | |
1da177e4 | 764 | wait_queue_head_t kswapd_wait; |
5515061d | 765 | wait_queue_head_t pfmemalloc_wait; |
bfc8c901 VD |
766 | struct task_struct *kswapd; /* Protected by |
767 | mem_hotplug_begin/end() */ | |
38087d9b | 768 | int kswapd_order; |
97a225e6 | 769 | enum zone_type kswapd_highest_zoneidx; |
38087d9b | 770 | |
c73322d0 JW |
771 | int kswapd_failures; /* Number of 'reclaimed == 0' runs */ |
772 | ||
698b1b30 VB |
773 | #ifdef CONFIG_COMPACTION |
774 | int kcompactd_max_order; | |
97a225e6 | 775 | enum zone_type kcompactd_highest_zoneidx; |
698b1b30 VB |
776 | wait_queue_head_t kcompactd_wait; |
777 | struct task_struct *kcompactd; | |
8177a420 | 778 | #endif |
281e3726 MG |
779 | /* |
780 | * This is a per-node reserve of pages that are not available | |
781 | * to userspace allocations. | |
782 | */ | |
783 | unsigned long totalreserve_pages; | |
784 | ||
a5f5f91d MG |
785 | #ifdef CONFIG_NUMA |
786 | /* | |
0a3c5772 | 787 | * node reclaim becomes active if more unmapped pages exist. |
a5f5f91d MG |
788 | */ |
789 | unsigned long min_unmapped_pages; | |
790 | unsigned long min_slab_pages; | |
791 | #endif /* CONFIG_NUMA */ | |
792 | ||
a52633d8 MG |
793 | /* Write-intensive fields used by page reclaim */ |
794 | ZONE_PADDING(_pad1_) | |
795 | spinlock_t lru_lock; | |
3a80a7fa MG |
796 | |
797 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT | |
798 | /* | |
799 | * If memory initialisation on large machines is deferred then this | |
800 | * is the first PFN that needs to be initialised. | |
801 | */ | |
802 | unsigned long first_deferred_pfn; | |
803 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
a3d0a918 KS |
804 | |
805 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
364c1eeb | 806 | struct deferred_split deferred_split_queue; |
a3d0a918 | 807 | #endif |
75ef7184 | 808 | |
599d0c95 | 809 | /* Fields commonly accessed by the page reclaim scanner */ |
867e5e1d JW |
810 | |
811 | /* | |
812 | * NOTE: THIS IS UNUSED IF MEMCG IS ENABLED. | |
813 | * | |
814 | * Use mem_cgroup_lruvec() to look up lruvecs. | |
815 | */ | |
816 | struct lruvec __lruvec; | |
599d0c95 | 817 | |
599d0c95 MG |
818 | unsigned long flags; |
819 | ||
820 | ZONE_PADDING(_pad2_) | |
821 | ||
75ef7184 MG |
822 | /* Per-node vmstats */ |
823 | struct per_cpu_nodestat __percpu *per_cpu_nodestats; | |
824 | atomic_long_t vm_stat[NR_VM_NODE_STAT_ITEMS]; | |
1da177e4 LT |
825 | } pg_data_t; |
826 | ||
827 | #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) | |
828 | #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) | |
d41dee36 | 829 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
408fde81 | 830 | #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) |
d41dee36 AW |
831 | #else |
832 | #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) | |
833 | #endif | |
408fde81 | 834 | #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr)) |
1da177e4 | 835 | |
c6830c22 | 836 | #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn) |
da3649e1 | 837 | #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid)) |
c6830c22 | 838 | |
da3649e1 CS |
839 | static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat) |
840 | { | |
841 | return pgdat->node_start_pfn + pgdat->node_spanned_pages; | |
842 | } | |
843 | ||
844 | static inline bool pgdat_is_empty(pg_data_t *pgdat) | |
845 | { | |
846 | return !pgdat->node_start_pfn && !pgdat->node_spanned_pages; | |
847 | } | |
c6830c22 | 848 | |
208d54e5 DH |
849 | #include <linux/memory_hotplug.h> |
850 | ||
72675e13 | 851 | void build_all_zonelists(pg_data_t *pgdat); |
5ecd9d40 | 852 | void wakeup_kswapd(struct zone *zone, gfp_t gfp_mask, int order, |
97a225e6 | 853 | enum zone_type highest_zoneidx); |
86a294a8 | 854 | bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark, |
97a225e6 | 855 | int highest_zoneidx, unsigned int alloc_flags, |
86a294a8 | 856 | long free_pages); |
7aeb09f9 | 857 | bool zone_watermark_ok(struct zone *z, unsigned int order, |
97a225e6 | 858 | unsigned long mark, int highest_zoneidx, |
c603844b | 859 | unsigned int alloc_flags); |
7aeb09f9 | 860 | bool zone_watermark_ok_safe(struct zone *z, unsigned int order, |
97a225e6 | 861 | unsigned long mark, int highest_zoneidx); |
c1d0da83 LD |
862 | /* |
863 | * Memory initialization context, use to differentiate memory added by | |
864 | * the platform statically or via memory hotplug interface. | |
865 | */ | |
866 | enum meminit_context { | |
867 | MEMINIT_EARLY, | |
868 | MEMINIT_HOTPLUG, | |
a2f3aa02 | 869 | }; |
c1d0da83 | 870 | |
dc0bbf3b | 871 | extern void init_currently_empty_zone(struct zone *zone, unsigned long start_pfn, |
b171e409 | 872 | unsigned long size); |
718127cc | 873 | |
bea8c150 | 874 | extern void lruvec_init(struct lruvec *lruvec); |
7f5e86c2 | 875 | |
599d0c95 | 876 | static inline struct pglist_data *lruvec_pgdat(struct lruvec *lruvec) |
7f5e86c2 | 877 | { |
c255a458 | 878 | #ifdef CONFIG_MEMCG |
599d0c95 | 879 | return lruvec->pgdat; |
7f5e86c2 | 880 | #else |
867e5e1d | 881 | return container_of(lruvec, struct pglist_data, __lruvec); |
7f5e86c2 KK |
882 | #endif |
883 | } | |
884 | ||
fd538803 | 885 | extern unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru, int zone_idx); |
23047a96 | 886 | |
7aac7898 LS |
887 | #ifdef CONFIG_HAVE_MEMORYLESS_NODES |
888 | int local_memory_node(int node_id); | |
889 | #else | |
890 | static inline int local_memory_node(int node_id) { return node_id; }; | |
891 | #endif | |
892 | ||
1da177e4 LT |
893 | /* |
894 | * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc. | |
895 | */ | |
896 | #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones) | |
897 | ||
6aa303de MG |
898 | /* |
899 | * Returns true if a zone has pages managed by the buddy allocator. | |
900 | * All the reclaim decisions have to use this function rather than | |
901 | * populated_zone(). If the whole zone is reserved then we can easily | |
902 | * end up with populated_zone() && !managed_zone(). | |
903 | */ | |
904 | static inline bool managed_zone(struct zone *zone) | |
905 | { | |
9705bea5 | 906 | return zone_managed_pages(zone); |
6aa303de MG |
907 | } |
908 | ||
909 | /* Returns true if a zone has memory */ | |
910 | static inline bool populated_zone(struct zone *zone) | |
f3fe6512 | 911 | { |
6aa303de | 912 | return zone->present_pages; |
f3fe6512 CK |
913 | } |
914 | ||
c1093b74 PT |
915 | #ifdef CONFIG_NUMA |
916 | static inline int zone_to_nid(struct zone *zone) | |
917 | { | |
918 | return zone->node; | |
919 | } | |
920 | ||
921 | static inline void zone_set_nid(struct zone *zone, int nid) | |
922 | { | |
923 | zone->node = nid; | |
924 | } | |
925 | #else | |
926 | static inline int zone_to_nid(struct zone *zone) | |
927 | { | |
928 | return 0; | |
929 | } | |
930 | ||
931 | static inline void zone_set_nid(struct zone *zone, int nid) {} | |
932 | #endif | |
933 | ||
2a1e274a MG |
934 | extern int movable_zone; |
935 | ||
d7e4a2ea | 936 | #ifdef CONFIG_HIGHMEM |
2a1e274a MG |
937 | static inline int zone_movable_is_highmem(void) |
938 | { | |
3f08a302 | 939 | #ifdef CONFIG_NEED_MULTIPLE_NODES |
2a1e274a MG |
940 | return movable_zone == ZONE_HIGHMEM; |
941 | #else | |
d7e4a2ea | 942 | return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM; |
2a1e274a MG |
943 | #endif |
944 | } | |
d7e4a2ea | 945 | #endif |
2a1e274a | 946 | |
2f1b6248 | 947 | static inline int is_highmem_idx(enum zone_type idx) |
1da177e4 | 948 | { |
e53ef38d | 949 | #ifdef CONFIG_HIGHMEM |
2a1e274a MG |
950 | return (idx == ZONE_HIGHMEM || |
951 | (idx == ZONE_MOVABLE && zone_movable_is_highmem())); | |
e53ef38d CL |
952 | #else |
953 | return 0; | |
954 | #endif | |
1da177e4 LT |
955 | } |
956 | ||
1da177e4 | 957 | /** |
b4a991ec | 958 | * is_highmem - helper function to quickly check if a struct zone is a |
1da177e4 LT |
959 | * highmem zone or not. This is an attempt to keep references |
960 | * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum. | |
961 | * @zone - pointer to struct zone variable | |
962 | */ | |
963 | static inline int is_highmem(struct zone *zone) | |
964 | { | |
e53ef38d | 965 | #ifdef CONFIG_HIGHMEM |
29f9cb53 | 966 | return is_highmem_idx(zone_idx(zone)); |
e53ef38d CL |
967 | #else |
968 | return 0; | |
969 | #endif | |
1da177e4 LT |
970 | } |
971 | ||
1da177e4 LT |
972 | /* These two functions are used to setup the per zone pages min values */ |
973 | struct ctl_table; | |
2374c09b | 974 | |
32927393 CH |
975 | int min_free_kbytes_sysctl_handler(struct ctl_table *, int, void *, size_t *, |
976 | loff_t *); | |
977 | int watermark_scale_factor_sysctl_handler(struct ctl_table *, int, void *, | |
978 | size_t *, loff_t *); | |
d3cda233 | 979 | extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES]; |
32927393 CH |
980 | int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, void *, |
981 | size_t *, loff_t *); | |
8d65af78 | 982 | int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, |
32927393 | 983 | void *, size_t *, loff_t *); |
9614634f | 984 | int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int, |
32927393 | 985 | void *, size_t *, loff_t *); |
0ff38490 | 986 | int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int, |
32927393 CH |
987 | void *, size_t *, loff_t *); |
988 | int numa_zonelist_order_handler(struct ctl_table *, int, | |
989 | void *, size_t *, loff_t *); | |
2374c09b | 990 | extern int percpu_pagelist_fraction; |
f0c0b2b8 | 991 | extern char numa_zonelist_order[]; |
c9bff3ee | 992 | #define NUMA_ZONELIST_ORDER_LEN 16 |
f0c0b2b8 | 993 | |
93b7504e | 994 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
1da177e4 LT |
995 | |
996 | extern struct pglist_data contig_page_data; | |
997 | #define NODE_DATA(nid) (&contig_page_data) | |
998 | #define NODE_MEM_MAP(nid) mem_map | |
1da177e4 | 999 | |
93b7504e | 1000 | #else /* CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 LT |
1001 | |
1002 | #include <asm/mmzone.h> | |
1003 | ||
93b7504e | 1004 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
348f8b6c | 1005 | |
95144c78 KH |
1006 | extern struct pglist_data *first_online_pgdat(void); |
1007 | extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat); | |
1008 | extern struct zone *next_zone(struct zone *zone); | |
8357f869 KH |
1009 | |
1010 | /** | |
12d15f0d | 1011 | * for_each_online_pgdat - helper macro to iterate over all online nodes |
8357f869 KH |
1012 | * @pgdat - pointer to a pg_data_t variable |
1013 | */ | |
1014 | #define for_each_online_pgdat(pgdat) \ | |
1015 | for (pgdat = first_online_pgdat(); \ | |
1016 | pgdat; \ | |
1017 | pgdat = next_online_pgdat(pgdat)) | |
8357f869 KH |
1018 | /** |
1019 | * for_each_zone - helper macro to iterate over all memory zones | |
1020 | * @zone - pointer to struct zone variable | |
1021 | * | |
1022 | * The user only needs to declare the zone variable, for_each_zone | |
1023 | * fills it in. | |
1024 | */ | |
1025 | #define for_each_zone(zone) \ | |
1026 | for (zone = (first_online_pgdat())->node_zones; \ | |
1027 | zone; \ | |
1028 | zone = next_zone(zone)) | |
1029 | ||
ee99c71c KM |
1030 | #define for_each_populated_zone(zone) \ |
1031 | for (zone = (first_online_pgdat())->node_zones; \ | |
1032 | zone; \ | |
1033 | zone = next_zone(zone)) \ | |
1034 | if (!populated_zone(zone)) \ | |
1035 | ; /* do nothing */ \ | |
1036 | else | |
1037 | ||
dd1a239f MG |
1038 | static inline struct zone *zonelist_zone(struct zoneref *zoneref) |
1039 | { | |
1040 | return zoneref->zone; | |
1041 | } | |
1042 | ||
1043 | static inline int zonelist_zone_idx(struct zoneref *zoneref) | |
1044 | { | |
1045 | return zoneref->zone_idx; | |
1046 | } | |
1047 | ||
1048 | static inline int zonelist_node_idx(struct zoneref *zoneref) | |
1049 | { | |
c1093b74 | 1050 | return zone_to_nid(zoneref->zone); |
dd1a239f MG |
1051 | } |
1052 | ||
682a3385 MG |
1053 | struct zoneref *__next_zones_zonelist(struct zoneref *z, |
1054 | enum zone_type highest_zoneidx, | |
1055 | nodemask_t *nodes); | |
1056 | ||
19770b32 MG |
1057 | /** |
1058 | * 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 | |
1059 | * @z - The cursor used as a starting point for the search | |
1060 | * @highest_zoneidx - The zone index of the highest zone to return | |
1061 | * @nodes - An optional nodemask to filter the zonelist with | |
19770b32 MG |
1062 | * |
1063 | * This function returns the next zone at or below a given zone index that is | |
1064 | * within the allowed nodemask using a cursor as the starting point for the | |
5bead2a0 MG |
1065 | * search. The zoneref returned is a cursor that represents the current zone |
1066 | * being examined. It should be advanced by one before calling | |
1067 | * next_zones_zonelist again. | |
19770b32 | 1068 | */ |
682a3385 | 1069 | static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z, |
19770b32 | 1070 | enum zone_type highest_zoneidx, |
682a3385 MG |
1071 | nodemask_t *nodes) |
1072 | { | |
1073 | if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx)) | |
1074 | return z; | |
1075 | return __next_zones_zonelist(z, highest_zoneidx, nodes); | |
1076 | } | |
dd1a239f | 1077 | |
19770b32 MG |
1078 | /** |
1079 | * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist | |
1080 | * @zonelist - The zonelist to search for a suitable zone | |
1081 | * @highest_zoneidx - The zone index of the highest zone to return | |
1082 | * @nodes - An optional nodemask to filter the zonelist with | |
ea57485a | 1083 | * @return - Zoneref pointer for the first suitable zone found (see below) |
19770b32 MG |
1084 | * |
1085 | * This function returns the first zone at or below a given zone index that is | |
1086 | * within the allowed nodemask. The zoneref returned is a cursor that can be | |
5bead2a0 MG |
1087 | * used to iterate the zonelist with next_zones_zonelist by advancing it by |
1088 | * one before calling. | |
ea57485a VB |
1089 | * |
1090 | * When no eligible zone is found, zoneref->zone is NULL (zoneref itself is | |
1091 | * never NULL). This may happen either genuinely, or due to concurrent nodemask | |
1092 | * update due to cpuset modification. | |
19770b32 | 1093 | */ |
dd1a239f | 1094 | static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist, |
19770b32 | 1095 | enum zone_type highest_zoneidx, |
c33d6c06 | 1096 | nodemask_t *nodes) |
54a6eb5c | 1097 | { |
c33d6c06 | 1098 | return next_zones_zonelist(zonelist->_zonerefs, |
05891fb0 | 1099 | highest_zoneidx, nodes); |
54a6eb5c MG |
1100 | } |
1101 | ||
19770b32 MG |
1102 | /** |
1103 | * 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 | |
1104 | * @zone - The current zone in the iterator | |
84218b55 | 1105 | * @z - The current pointer within zonelist->_zonerefs being iterated |
19770b32 MG |
1106 | * @zlist - The zonelist being iterated |
1107 | * @highidx - The zone index of the highest zone to return | |
1108 | * @nodemask - Nodemask allowed by the allocator | |
1109 | * | |
1110 | * This iterator iterates though all zones at or below a given zone index and | |
1111 | * within a given nodemask | |
1112 | */ | |
1113 | #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \ | |
c33d6c06 | 1114 | for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \ |
19770b32 | 1115 | zone; \ |
05891fb0 | 1116 | z = next_zones_zonelist(++z, highidx, nodemask), \ |
c33d6c06 MG |
1117 | zone = zonelist_zone(z)) |
1118 | ||
1119 | #define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \ | |
1120 | for (zone = z->zone; \ | |
1121 | zone; \ | |
1122 | z = next_zones_zonelist(++z, highidx, nodemask), \ | |
1123 | zone = zonelist_zone(z)) | |
1124 | ||
54a6eb5c MG |
1125 | |
1126 | /** | |
1127 | * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index | |
1128 | * @zone - The current zone in the iterator | |
1129 | * @z - The current pointer within zonelist->zones being iterated | |
1130 | * @zlist - The zonelist being iterated | |
1131 | * @highidx - The zone index of the highest zone to return | |
1132 | * | |
1133 | * This iterator iterates though all zones at or below a given zone index. | |
1134 | */ | |
1135 | #define for_each_zone_zonelist(zone, z, zlist, highidx) \ | |
19770b32 | 1136 | for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL) |
54a6eb5c | 1137 | |
d41dee36 AW |
1138 | #ifdef CONFIG_SPARSEMEM |
1139 | #include <asm/sparsemem.h> | |
1140 | #endif | |
1141 | ||
2bdaf115 AW |
1142 | #ifdef CONFIG_FLATMEM |
1143 | #define pfn_to_nid(pfn) (0) | |
1144 | #endif | |
1145 | ||
d41dee36 AW |
1146 | #ifdef CONFIG_SPARSEMEM |
1147 | ||
1148 | /* | |
1149 | * SECTION_SHIFT #bits space required to store a section # | |
1150 | * | |
1151 | * PA_SECTION_SHIFT physical address to/from section number | |
1152 | * PFN_SECTION_SHIFT pfn to/from section number | |
1153 | */ | |
d41dee36 AW |
1154 | #define PA_SECTION_SHIFT (SECTION_SIZE_BITS) |
1155 | #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) | |
1156 | ||
1157 | #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) | |
1158 | ||
1159 | #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) | |
1160 | #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) | |
1161 | ||
835c134e | 1162 | #define SECTION_BLOCKFLAGS_BITS \ |
d9c23400 | 1163 | ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS) |
835c134e | 1164 | |
d41dee36 AW |
1165 | #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS |
1166 | #error Allocator MAX_ORDER exceeds SECTION_SIZE | |
1167 | #endif | |
1168 | ||
1dd2bfc8 YI |
1169 | static inline unsigned long pfn_to_section_nr(unsigned long pfn) |
1170 | { | |
1171 | return pfn >> PFN_SECTION_SHIFT; | |
1172 | } | |
1173 | static inline unsigned long section_nr_to_pfn(unsigned long sec) | |
1174 | { | |
1175 | return sec << PFN_SECTION_SHIFT; | |
1176 | } | |
e3c40f37 | 1177 | |
a539f353 DK |
1178 | #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK) |
1179 | #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK) | |
1180 | ||
f1eca35a | 1181 | #define SUBSECTION_SHIFT 21 |
9ffc1d19 | 1182 | #define SUBSECTION_SIZE (1UL << SUBSECTION_SHIFT) |
f1eca35a DW |
1183 | |
1184 | #define PFN_SUBSECTION_SHIFT (SUBSECTION_SHIFT - PAGE_SHIFT) | |
1185 | #define PAGES_PER_SUBSECTION (1UL << PFN_SUBSECTION_SHIFT) | |
1186 | #define PAGE_SUBSECTION_MASK (~(PAGES_PER_SUBSECTION-1)) | |
1187 | ||
1188 | #if SUBSECTION_SHIFT > SECTION_SIZE_BITS | |
1189 | #error Subsection size exceeds section size | |
1190 | #else | |
1191 | #define SUBSECTIONS_PER_SECTION (1UL << (SECTION_SIZE_BITS - SUBSECTION_SHIFT)) | |
1192 | #endif | |
1193 | ||
a3619190 DW |
1194 | #define SUBSECTION_ALIGN_UP(pfn) ALIGN((pfn), PAGES_PER_SUBSECTION) |
1195 | #define SUBSECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SUBSECTION_MASK) | |
1196 | ||
f1eca35a | 1197 | struct mem_section_usage { |
0a9f9f62 | 1198 | #ifdef CONFIG_SPARSEMEM_VMEMMAP |
f1eca35a | 1199 | DECLARE_BITMAP(subsection_map, SUBSECTIONS_PER_SECTION); |
0a9f9f62 | 1200 | #endif |
f1eca35a DW |
1201 | /* See declaration of similar field in struct zone */ |
1202 | unsigned long pageblock_flags[0]; | |
1203 | }; | |
1204 | ||
f46edbd1 DW |
1205 | void subsection_map_init(unsigned long pfn, unsigned long nr_pages); |
1206 | ||
d41dee36 | 1207 | struct page; |
eefa864b | 1208 | struct page_ext; |
d41dee36 | 1209 | struct mem_section { |
29751f69 AW |
1210 | /* |
1211 | * This is, logically, a pointer to an array of struct | |
1212 | * pages. However, it is stored with some other magic. | |
1213 | * (see sparse.c::sparse_init_one_section()) | |
1214 | * | |
30c253e6 AW |
1215 | * Additionally during early boot we encode node id of |
1216 | * the location of the section here to guide allocation. | |
1217 | * (see sparse.c::memory_present()) | |
1218 | * | |
29751f69 AW |
1219 | * Making it a UL at least makes someone do a cast |
1220 | * before using it wrong. | |
1221 | */ | |
1222 | unsigned long section_mem_map; | |
5c0e3066 | 1223 | |
f1eca35a | 1224 | struct mem_section_usage *usage; |
eefa864b JK |
1225 | #ifdef CONFIG_PAGE_EXTENSION |
1226 | /* | |
0c9ad804 | 1227 | * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use |
eefa864b JK |
1228 | * section. (see page_ext.h about this.) |
1229 | */ | |
1230 | struct page_ext *page_ext; | |
1231 | unsigned long pad; | |
1232 | #endif | |
55878e88 CS |
1233 | /* |
1234 | * WARNING: mem_section must be a power-of-2 in size for the | |
1235 | * calculation and use of SECTION_ROOT_MASK to make sense. | |
1236 | */ | |
d41dee36 AW |
1237 | }; |
1238 | ||
3e347261 BP |
1239 | #ifdef CONFIG_SPARSEMEM_EXTREME |
1240 | #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section)) | |
1241 | #else | |
1242 | #define SECTIONS_PER_ROOT 1 | |
1243 | #endif | |
802f192e | 1244 | |
3e347261 | 1245 | #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT) |
0faa5638 | 1246 | #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT) |
3e347261 | 1247 | #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1) |
802f192e | 1248 | |
3e347261 | 1249 | #ifdef CONFIG_SPARSEMEM_EXTREME |
83e3c487 | 1250 | extern struct mem_section **mem_section; |
802f192e | 1251 | #else |
3e347261 BP |
1252 | extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT]; |
1253 | #endif | |
d41dee36 | 1254 | |
f1eca35a DW |
1255 | static inline unsigned long *section_to_usemap(struct mem_section *ms) |
1256 | { | |
1257 | return ms->usage->pageblock_flags; | |
1258 | } | |
1259 | ||
29751f69 AW |
1260 | static inline struct mem_section *__nr_to_section(unsigned long nr) |
1261 | { | |
83e3c487 KS |
1262 | #ifdef CONFIG_SPARSEMEM_EXTREME |
1263 | if (!mem_section) | |
1264 | return NULL; | |
1265 | #endif | |
3e347261 BP |
1266 | if (!mem_section[SECTION_NR_TO_ROOT(nr)]) |
1267 | return NULL; | |
1268 | return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK]; | |
29751f69 | 1269 | } |
2491f0a2 | 1270 | extern unsigned long __section_nr(struct mem_section *ms); |
f1eca35a | 1271 | extern size_t mem_section_usage_size(void); |
29751f69 AW |
1272 | |
1273 | /* | |
1274 | * We use the lower bits of the mem_map pointer to store | |
def9b71e PT |
1275 | * a little bit of information. The pointer is calculated |
1276 | * as mem_map - section_nr_to_pfn(pnum). The result is | |
1277 | * aligned to the minimum alignment of the two values: | |
1278 | * 1. All mem_map arrays are page-aligned. | |
1279 | * 2. section_nr_to_pfn() always clears PFN_SECTION_SHIFT | |
1280 | * lowest bits. PFN_SECTION_SHIFT is arch-specific | |
1281 | * (equal SECTION_SIZE_BITS - PAGE_SHIFT), and the | |
1282 | * worst combination is powerpc with 256k pages, | |
1283 | * which results in PFN_SECTION_SHIFT equal 6. | |
1284 | * To sum it up, at least 6 bits are available. | |
29751f69 AW |
1285 | */ |
1286 | #define SECTION_MARKED_PRESENT (1UL<<0) | |
1287 | #define SECTION_HAS_MEM_MAP (1UL<<1) | |
2d070eab | 1288 | #define SECTION_IS_ONLINE (1UL<<2) |
326e1b8f DW |
1289 | #define SECTION_IS_EARLY (1UL<<3) |
1290 | #define SECTION_MAP_LAST_BIT (1UL<<4) | |
29751f69 | 1291 | #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1)) |
2d070eab | 1292 | #define SECTION_NID_SHIFT 3 |
29751f69 AW |
1293 | |
1294 | static inline struct page *__section_mem_map_addr(struct mem_section *section) | |
1295 | { | |
1296 | unsigned long map = section->section_mem_map; | |
1297 | map &= SECTION_MAP_MASK; | |
1298 | return (struct page *)map; | |
1299 | } | |
1300 | ||
540557b9 | 1301 | static inline int present_section(struct mem_section *section) |
29751f69 | 1302 | { |
802f192e | 1303 | return (section && (section->section_mem_map & SECTION_MARKED_PRESENT)); |
29751f69 AW |
1304 | } |
1305 | ||
540557b9 AW |
1306 | static inline int present_section_nr(unsigned long nr) |
1307 | { | |
1308 | return present_section(__nr_to_section(nr)); | |
1309 | } | |
1310 | ||
1311 | static inline int valid_section(struct mem_section *section) | |
29751f69 | 1312 | { |
802f192e | 1313 | return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP)); |
29751f69 AW |
1314 | } |
1315 | ||
326e1b8f DW |
1316 | static inline int early_section(struct mem_section *section) |
1317 | { | |
1318 | return (section && (section->section_mem_map & SECTION_IS_EARLY)); | |
1319 | } | |
1320 | ||
29751f69 AW |
1321 | static inline int valid_section_nr(unsigned long nr) |
1322 | { | |
1323 | return valid_section(__nr_to_section(nr)); | |
1324 | } | |
1325 | ||
2d070eab MH |
1326 | static inline int online_section(struct mem_section *section) |
1327 | { | |
1328 | return (section && (section->section_mem_map & SECTION_IS_ONLINE)); | |
1329 | } | |
1330 | ||
1331 | static inline int online_section_nr(unsigned long nr) | |
1332 | { | |
1333 | return online_section(__nr_to_section(nr)); | |
1334 | } | |
1335 | ||
1336 | #ifdef CONFIG_MEMORY_HOTPLUG | |
1337 | void online_mem_sections(unsigned long start_pfn, unsigned long end_pfn); | |
1338 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
1339 | void offline_mem_sections(unsigned long start_pfn, unsigned long end_pfn); | |
1340 | #endif | |
1341 | #endif | |
1342 | ||
d41dee36 AW |
1343 | static inline struct mem_section *__pfn_to_section(unsigned long pfn) |
1344 | { | |
29751f69 | 1345 | return __nr_to_section(pfn_to_section_nr(pfn)); |
d41dee36 AW |
1346 | } |
1347 | ||
2491f0a2 | 1348 | extern unsigned long __highest_present_section_nr; |
c4e1be9e | 1349 | |
f46edbd1 DW |
1350 | static inline int subsection_map_index(unsigned long pfn) |
1351 | { | |
1352 | return (pfn & ~(PAGE_SECTION_MASK)) / PAGES_PER_SUBSECTION; | |
1353 | } | |
1354 | ||
1355 | #ifdef CONFIG_SPARSEMEM_VMEMMAP | |
1356 | static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn) | |
1357 | { | |
1358 | int idx = subsection_map_index(pfn); | |
1359 | ||
1360 | return test_bit(idx, ms->usage->subsection_map); | |
1361 | } | |
1362 | #else | |
1363 | static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn) | |
1364 | { | |
1365 | return 1; | |
1366 | } | |
1367 | #endif | |
1368 | ||
7b7bf499 | 1369 | #ifndef CONFIG_HAVE_ARCH_PFN_VALID |
d41dee36 AW |
1370 | static inline int pfn_valid(unsigned long pfn) |
1371 | { | |
f46edbd1 DW |
1372 | struct mem_section *ms; |
1373 | ||
d41dee36 AW |
1374 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) |
1375 | return 0; | |
f46edbd1 DW |
1376 | ms = __nr_to_section(pfn_to_section_nr(pfn)); |
1377 | if (!valid_section(ms)) | |
1378 | return 0; | |
1379 | /* | |
1380 | * Traditionally early sections always returned pfn_valid() for | |
1381 | * the entire section-sized span. | |
1382 | */ | |
1383 | return early_section(ms) || pfn_section_valid(ms, pfn); | |
d41dee36 | 1384 | } |
7b7bf499 | 1385 | #endif |
d41dee36 | 1386 | |
e03d1f78 | 1387 | static inline int pfn_in_present_section(unsigned long pfn) |
540557b9 AW |
1388 | { |
1389 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) | |
1390 | return 0; | |
1391 | return present_section(__nr_to_section(pfn_to_section_nr(pfn))); | |
1392 | } | |
1393 | ||
4c605881 DH |
1394 | static inline unsigned long next_present_section_nr(unsigned long section_nr) |
1395 | { | |
1396 | while (++section_nr <= __highest_present_section_nr) { | |
1397 | if (present_section_nr(section_nr)) | |
1398 | return section_nr; | |
1399 | } | |
1400 | ||
1401 | return -1; | |
1402 | } | |
1403 | ||
d41dee36 AW |
1404 | /* |
1405 | * These are _only_ used during initialisation, therefore they | |
1406 | * can use __initdata ... They could have names to indicate | |
1407 | * this restriction. | |
1408 | */ | |
1409 | #ifdef CONFIG_NUMA | |
161599ff AW |
1410 | #define pfn_to_nid(pfn) \ |
1411 | ({ \ | |
1412 | unsigned long __pfn_to_nid_pfn = (pfn); \ | |
1413 | page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \ | |
1414 | }) | |
2bdaf115 AW |
1415 | #else |
1416 | #define pfn_to_nid(pfn) (0) | |
d41dee36 AW |
1417 | #endif |
1418 | ||
d41dee36 AW |
1419 | #define early_pfn_valid(pfn) pfn_valid(pfn) |
1420 | void sparse_init(void); | |
1421 | #else | |
1422 | #define sparse_init() do {} while (0) | |
28ae55c9 | 1423 | #define sparse_index_init(_sec, _nid) do {} while (0) |
e03d1f78 | 1424 | #define pfn_in_present_section pfn_valid |
f46edbd1 | 1425 | #define subsection_map_init(_pfn, _nr_pages) do {} while (0) |
d41dee36 AW |
1426 | #endif /* CONFIG_SPARSEMEM */ |
1427 | ||
8a942fde MG |
1428 | /* |
1429 | * During memory init memblocks map pfns to nids. The search is expensive and | |
1430 | * this caches recent lookups. The implementation of __early_pfn_to_nid | |
1431 | * may treat start/end as pfns or sections. | |
1432 | */ | |
1433 | struct mminit_pfnnid_cache { | |
1434 | unsigned long last_start; | |
1435 | unsigned long last_end; | |
1436 | int last_nid; | |
1437 | }; | |
1438 | ||
d41dee36 AW |
1439 | #ifndef early_pfn_valid |
1440 | #define early_pfn_valid(pfn) (1) | |
1441 | #endif | |
1442 | ||
14e07298 AW |
1443 | /* |
1444 | * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we | |
8bb4e7a2 | 1445 | * need to check pfn validity within that MAX_ORDER_NR_PAGES block. |
14e07298 AW |
1446 | * pfn_valid_within() should be used in this case; we optimise this away |
1447 | * when we have no holes within a MAX_ORDER_NR_PAGES block. | |
1448 | */ | |
1449 | #ifdef CONFIG_HOLES_IN_ZONE | |
1450 | #define pfn_valid_within(pfn) pfn_valid(pfn) | |
1451 | #else | |
1452 | #define pfn_valid_within(pfn) (1) | |
1453 | #endif | |
1454 | ||
eb33575c MG |
1455 | #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL |
1456 | /* | |
1457 | * pfn_valid() is meant to be able to tell if a given PFN has valid memmap | |
2d070eab MH |
1458 | * associated with it or not. This means that a struct page exists for this |
1459 | * pfn. The caller cannot assume the page is fully initialized in general. | |
1460 | * Hotplugable pages might not have been onlined yet. pfn_to_online_page() | |
1461 | * will ensure the struct page is fully online and initialized. Special pages | |
1462 | * (e.g. ZONE_DEVICE) are never onlined and should be treated accordingly. | |
1463 | * | |
1464 | * In FLATMEM, it is expected that holes always have valid memmap as long as | |
1465 | * there is valid PFNs either side of the hole. In SPARSEMEM, it is assumed | |
1466 | * that a valid section has a memmap for the entire section. | |
eb33575c MG |
1467 | * |
1468 | * However, an ARM, and maybe other embedded architectures in the future | |
1469 | * free memmap backing holes to save memory on the assumption the memmap is | |
1470 | * never used. The page_zone linkages are then broken even though pfn_valid() | |
1471 | * returns true. A walker of the full memmap must then do this additional | |
1472 | * check to ensure the memmap they are looking at is sane by making sure | |
1473 | * the zone and PFN linkages are still valid. This is expensive, but walkers | |
1474 | * of the full memmap are extremely rare. | |
1475 | */ | |
5b80287a | 1476 | bool memmap_valid_within(unsigned long pfn, |
eb33575c MG |
1477 | struct page *page, struct zone *zone); |
1478 | #else | |
5b80287a | 1479 | static inline bool memmap_valid_within(unsigned long pfn, |
eb33575c MG |
1480 | struct page *page, struct zone *zone) |
1481 | { | |
5b80287a | 1482 | return true; |
eb33575c MG |
1483 | } |
1484 | #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */ | |
1485 | ||
97965478 | 1486 | #endif /* !__GENERATING_BOUNDS.H */ |
1da177e4 | 1487 | #endif /* !__ASSEMBLY__ */ |
1da177e4 | 1488 | #endif /* _LINUX_MMZONE_H */ |