Commit | Line | Data |
---|---|---|
2874c5fd | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
1da177e4 LT |
2 | /* internal.h: mm/ internal definitions |
3 | * | |
4 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. | |
5 | * Written by David Howells (dhowells@redhat.com) | |
1da177e4 | 6 | */ |
0f8053a5 NP |
7 | #ifndef __MM_INTERNAL_H |
8 | #define __MM_INTERNAL_H | |
9 | ||
29f175d1 | 10 | #include <linux/fs.h> |
0f8053a5 | 11 | #include <linux/mm.h> |
e9b61f19 | 12 | #include <linux/pagemap.h> |
edf14cdb | 13 | #include <linux/tracepoint-defs.h> |
1da177e4 | 14 | |
dd56b046 MG |
15 | /* |
16 | * The set of flags that only affect watermark checking and reclaim | |
17 | * behaviour. This is used by the MM to obey the caller constraints | |
18 | * about IO, FS and watermark checking while ignoring placement | |
19 | * hints such as HIGHMEM usage. | |
20 | */ | |
21 | #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\ | |
dcda9b04 | 22 | __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\ |
e838a45f MG |
23 | __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\ |
24 | __GFP_ATOMIC) | |
dd56b046 MG |
25 | |
26 | /* The GFP flags allowed during early boot */ | |
27 | #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS)) | |
28 | ||
29 | /* Control allocation cpuset and node placement constraints */ | |
30 | #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) | |
31 | ||
32 | /* Do not use these with a slab allocator */ | |
33 | #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) | |
34 | ||
62906027 NP |
35 | void page_writeback_init(void); |
36 | ||
64601000 MWO |
37 | static inline void *folio_raw_mapping(struct folio *folio) |
38 | { | |
39 | unsigned long mapping = (unsigned long)folio->mapping; | |
40 | ||
41 | return (void *)(mapping & ~PAGE_MAPPING_FLAGS); | |
42 | } | |
43 | ||
512b7931 | 44 | void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, |
8cd7c588 | 45 | int nr_throttled); |
512b7931 | 46 | static inline void acct_reclaim_writeback(struct folio *folio) |
8cd7c588 | 47 | { |
512b7931 | 48 | pg_data_t *pgdat = folio_pgdat(folio); |
8cd7c588 MG |
49 | int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled); |
50 | ||
51 | if (nr_throttled) | |
512b7931 | 52 | __acct_reclaim_writeback(pgdat, folio, nr_throttled); |
8cd7c588 MG |
53 | } |
54 | ||
d818fca1 MG |
55 | static inline void wake_throttle_isolated(pg_data_t *pgdat) |
56 | { | |
57 | wait_queue_head_t *wqh; | |
58 | ||
59 | wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED]; | |
60 | if (waitqueue_active(wqh)) | |
61 | wake_up(wqh); | |
62 | } | |
63 | ||
2b740303 | 64 | vm_fault_t do_swap_page(struct vm_fault *vmf); |
575ced1c | 65 | void folio_rotate_reclaimable(struct folio *folio); |
269ccca3 | 66 | bool __folio_end_writeback(struct folio *folio); |
8a966ed7 | 67 | |
42b77728 JB |
68 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
69 | unsigned long floor, unsigned long ceiling); | |
03c4f204 | 70 | void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); |
42b77728 | 71 | |
9c276cc6 | 72 | static inline bool can_madv_lru_vma(struct vm_area_struct *vma) |
23519073 KS |
73 | { |
74 | return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP)); | |
75 | } | |
76 | ||
aac45363 MH |
77 | void unmap_page_range(struct mmu_gather *tlb, |
78 | struct vm_area_struct *vma, | |
79 | unsigned long addr, unsigned long end, | |
80 | struct zap_details *details); | |
81 | ||
7b3df3b9 DH |
82 | void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read, |
83 | unsigned long lookahead_size); | |
fcd9ae4f | 84 | void force_page_cache_ra(struct readahead_control *, unsigned long nr); |
7b3df3b9 DH |
85 | static inline void force_page_cache_readahead(struct address_space *mapping, |
86 | struct file *file, pgoff_t index, unsigned long nr_to_read) | |
87 | { | |
fcd9ae4f MWO |
88 | DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index); |
89 | force_page_cache_ra(&ractl, nr_to_read); | |
7b3df3b9 | 90 | } |
29f175d1 | 91 | |
5c211ba2 MWO |
92 | unsigned find_lock_entries(struct address_space *mapping, pgoff_t start, |
93 | pgoff_t end, struct pagevec *pvec, pgoff_t *indices); | |
94 | ||
1eb6234e | 95 | /** |
3eed3ef5 MWO |
96 | * folio_evictable - Test whether a folio is evictable. |
97 | * @folio: The folio to test. | |
1eb6234e | 98 | * |
3eed3ef5 MWO |
99 | * Test whether @folio is evictable -- i.e., should be placed on |
100 | * active/inactive lists vs unevictable list. | |
1eb6234e | 101 | * |
3eed3ef5 MWO |
102 | * Reasons folio might not be evictable: |
103 | * 1. folio's mapping marked unevictable | |
104 | * 2. One of the pages in the folio is part of an mlocked VMA | |
1eb6234e | 105 | */ |
3eed3ef5 MWO |
106 | static inline bool folio_evictable(struct folio *folio) |
107 | { | |
108 | bool ret; | |
109 | ||
110 | /* Prevent address_space of inode and swap cache from being freed */ | |
111 | rcu_read_lock(); | |
112 | ret = !mapping_unevictable(folio_mapping(folio)) && | |
113 | !folio_test_mlocked(folio); | |
114 | rcu_read_unlock(); | |
115 | return ret; | |
116 | } | |
117 | ||
1eb6234e YS |
118 | static inline bool page_evictable(struct page *page) |
119 | { | |
120 | bool ret; | |
121 | ||
122 | /* Prevent address_space of inode and swap cache from being freed */ | |
123 | rcu_read_lock(); | |
124 | ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page); | |
125 | rcu_read_unlock(); | |
126 | return ret; | |
127 | } | |
128 | ||
7835e98b | 129 | /* |
0139aa7b | 130 | * Turn a non-refcounted page (->_refcount == 0) into refcounted with |
7835e98b NP |
131 | * a count of one. |
132 | */ | |
133 | static inline void set_page_refcounted(struct page *page) | |
134 | { | |
309381fe | 135 | VM_BUG_ON_PAGE(PageTail(page), page); |
fe896d18 | 136 | VM_BUG_ON_PAGE(page_ref_count(page), page); |
77a8a788 | 137 | set_page_count(page, 1); |
77a8a788 NP |
138 | } |
139 | ||
03f6462a HD |
140 | extern unsigned long highest_memmap_pfn; |
141 | ||
c73322d0 JW |
142 | /* |
143 | * Maximum number of reclaim retries without progress before the OOM | |
144 | * killer is consider the only way forward. | |
145 | */ | |
146 | #define MAX_RECLAIM_RETRIES 16 | |
147 | ||
894bc310 LS |
148 | /* |
149 | * in mm/vmscan.c: | |
150 | */ | |
62695a84 | 151 | extern int isolate_lru_page(struct page *page); |
894bc310 | 152 | extern void putback_lru_page(struct page *page); |
c3f4a9a2 | 153 | extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); |
62695a84 | 154 | |
6219049a BL |
155 | /* |
156 | * in mm/rmap.c: | |
157 | */ | |
158 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); | |
159 | ||
894bc310 LS |
160 | /* |
161 | * in mm/page_alloc.c | |
162 | */ | |
3c605096 | 163 | |
1a6d53a1 VB |
164 | /* |
165 | * Structure for holding the mostly immutable allocation parameters passed | |
166 | * between functions involved in allocations, including the alloc_pages* | |
167 | * family of functions. | |
168 | * | |
97a225e6 | 169 | * nodemask, migratetype and highest_zoneidx are initialized only once in |
84172f4b | 170 | * __alloc_pages() and then never change. |
1a6d53a1 | 171 | * |
97a225e6 | 172 | * zonelist, preferred_zone and highest_zoneidx are set first in |
84172f4b | 173 | * __alloc_pages() for the fast path, and might be later changed |
68956ccb | 174 | * in __alloc_pages_slowpath(). All other functions pass the whole structure |
1a6d53a1 VB |
175 | * by a const pointer. |
176 | */ | |
177 | struct alloc_context { | |
178 | struct zonelist *zonelist; | |
179 | nodemask_t *nodemask; | |
c33d6c06 | 180 | struct zoneref *preferred_zoneref; |
1a6d53a1 | 181 | int migratetype; |
97a225e6 JK |
182 | |
183 | /* | |
184 | * highest_zoneidx represents highest usable zone index of | |
185 | * the allocation request. Due to the nature of the zone, | |
186 | * memory on lower zone than the highest_zoneidx will be | |
187 | * protected by lowmem_reserve[highest_zoneidx]. | |
188 | * | |
189 | * highest_zoneidx is also used by reclaim/compaction to limit | |
190 | * the target zone since higher zone than this index cannot be | |
191 | * usable for this allocation request. | |
192 | */ | |
193 | enum zone_type highest_zoneidx; | |
c9ab0c4f | 194 | bool spread_dirty_pages; |
1a6d53a1 VB |
195 | }; |
196 | ||
3c605096 JK |
197 | /* |
198 | * Locate the struct page for both the matching buddy in our | |
199 | * pair (buddy1) and the combined O(n+1) page they form (page). | |
200 | * | |
201 | * 1) Any buddy B1 will have an order O twin B2 which satisfies | |
202 | * the following equation: | |
203 | * B2 = B1 ^ (1 << O) | |
204 | * For example, if the starting buddy (buddy2) is #8 its order | |
205 | * 1 buddy is #10: | |
206 | * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 | |
207 | * | |
208 | * 2) Any buddy B will have an order O+1 parent P which | |
209 | * satisfies the following equation: | |
210 | * P = B & ~(1 << O) | |
211 | * | |
212 | * Assumption: *_mem_map is contiguous at least up to MAX_ORDER | |
213 | */ | |
214 | static inline unsigned long | |
76741e77 | 215 | __find_buddy_pfn(unsigned long page_pfn, unsigned int order) |
3c605096 | 216 | { |
76741e77 | 217 | return page_pfn ^ (1 << order); |
3c605096 JK |
218 | } |
219 | ||
7cf91a98 JK |
220 | extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, |
221 | unsigned long end_pfn, struct zone *zone); | |
222 | ||
223 | static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, | |
224 | unsigned long end_pfn, struct zone *zone) | |
225 | { | |
226 | if (zone->contiguous) | |
227 | return pfn_to_page(start_pfn); | |
228 | ||
229 | return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); | |
230 | } | |
231 | ||
3c605096 | 232 | extern int __isolate_free_page(struct page *page, unsigned int order); |
624f58d8 AD |
233 | extern void __putback_isolated_page(struct page *page, unsigned int order, |
234 | int mt); | |
7c2ee349 | 235 | extern void memblock_free_pages(struct page *page, unsigned long pfn, |
d70ddd7a | 236 | unsigned int order); |
a9cd410a | 237 | extern void __free_pages_core(struct page *page, unsigned int order); |
d00181b9 | 238 | extern void prep_compound_page(struct page *page, unsigned int order); |
46f24fd8 JK |
239 | extern void post_alloc_hook(struct page *page, unsigned int order, |
240 | gfp_t gfp_flags); | |
42aa83cb | 241 | extern int user_min_free_kbytes; |
20a0307c | 242 | |
44042b44 | 243 | extern void free_unref_page(struct page *page, unsigned int order); |
0966aeb4 MWO |
244 | extern void free_unref_page_list(struct list_head *list); |
245 | ||
04f8cfea | 246 | extern void zone_pcp_update(struct zone *zone, int cpu_online); |
68265390 | 247 | extern void zone_pcp_reset(struct zone *zone); |
ec6e8c7e VB |
248 | extern void zone_pcp_disable(struct zone *zone); |
249 | extern void zone_pcp_enable(struct zone *zone); | |
68265390 | 250 | |
c803b3c8 MR |
251 | extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, |
252 | phys_addr_t min_addr, | |
253 | int nid, bool exact_nid); | |
254 | ||
ff9543fd MN |
255 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
256 | ||
257 | /* | |
258 | * in mm/compaction.c | |
259 | */ | |
260 | /* | |
261 | * compact_control is used to track pages being migrated and the free pages | |
262 | * they are being migrated to during memory compaction. The free_pfn starts | |
263 | * at the end of a zone and migrate_pfn begins at the start. Movable pages | |
264 | * are moved to the end of a zone during a compaction run and the run | |
265 | * completes when free_pfn <= migrate_pfn | |
266 | */ | |
267 | struct compact_control { | |
268 | struct list_head freepages; /* List of free pages to migrate to */ | |
269 | struct list_head migratepages; /* List of pages being migrated */ | |
c5fbd937 MG |
270 | unsigned int nr_freepages; /* Number of isolated free pages */ |
271 | unsigned int nr_migratepages; /* Number of pages to migrate */ | |
ff9543fd | 272 | unsigned long free_pfn; /* isolate_freepages search base */ |
c2ad7a1f OS |
273 | /* |
274 | * Acts as an in/out parameter to page isolation for migration. | |
275 | * isolate_migratepages uses it as a search base. | |
276 | * isolate_migratepages_block will update the value to the next pfn | |
277 | * after the last isolated one. | |
278 | */ | |
279 | unsigned long migrate_pfn; | |
70b44595 | 280 | unsigned long fast_start_pfn; /* a pfn to start linear scan from */ |
c5943b9c MG |
281 | struct zone *zone; |
282 | unsigned long total_migrate_scanned; | |
283 | unsigned long total_free_scanned; | |
dbe2d4e4 MG |
284 | unsigned short fast_search_fail;/* failures to use free list searches */ |
285 | short search_order; /* order to start a fast search at */ | |
f25ba6dc VB |
286 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
287 | int order; /* order a direct compactor needs */ | |
d39773a0 | 288 | int migratetype; /* migratetype of direct compactor */ |
f25ba6dc | 289 | const unsigned int alloc_flags; /* alloc flags of a direct compactor */ |
97a225e6 | 290 | const int highest_zoneidx; /* zone index of a direct compactor */ |
e0b9daeb | 291 | enum migrate_mode mode; /* Async or sync migration mode */ |
bb13ffeb | 292 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
2583d671 | 293 | bool no_set_skip_hint; /* Don't mark blocks for skipping */ |
9f7e3387 | 294 | bool ignore_block_suitable; /* Scan blocks considered unsuitable */ |
accf6242 | 295 | bool direct_compaction; /* False from kcompactd or /proc/... */ |
facdaa91 | 296 | bool proactive_compaction; /* kcompactd proactive compaction */ |
06ed2998 | 297 | bool whole_zone; /* Whole zone should/has been scanned */ |
c3486f53 | 298 | bool contended; /* Signal lock or sched contention */ |
804d3121 | 299 | bool rescan; /* Rescanning the same pageblock */ |
b06eda09 | 300 | bool alloc_contig; /* alloc_contig_range allocation */ |
ff9543fd MN |
301 | }; |
302 | ||
5e1f0f09 MG |
303 | /* |
304 | * Used in direct compaction when a page should be taken from the freelists | |
305 | * immediately when one is created during the free path. | |
306 | */ | |
307 | struct capture_control { | |
308 | struct compact_control *cc; | |
309 | struct page *page; | |
310 | }; | |
311 | ||
ff9543fd | 312 | unsigned long |
bb13ffeb MG |
313 | isolate_freepages_range(struct compact_control *cc, |
314 | unsigned long start_pfn, unsigned long end_pfn); | |
c2ad7a1f | 315 | int |
edc2ca61 VB |
316 | isolate_migratepages_range(struct compact_control *cc, |
317 | unsigned long low_pfn, unsigned long end_pfn); | |
ffd8f251 | 318 | #endif |
2149cdae JK |
319 | int find_suitable_fallback(struct free_area *area, unsigned int order, |
320 | int migratetype, bool only_stealable, bool *can_steal); | |
ff9543fd | 321 | |
48f13bf3 | 322 | /* |
6c14466c MG |
323 | * This function returns the order of a free page in the buddy system. In |
324 | * general, page_zone(page)->lock must be held by the caller to prevent the | |
325 | * page from being allocated in parallel and returning garbage as the order. | |
326 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the | |
99c0fd5e | 327 | * page cannot be allocated or merged in parallel. Alternatively, it must |
ab130f91 | 328 | * handle invalid values gracefully, and use buddy_order_unsafe() below. |
48f13bf3 | 329 | */ |
ab130f91 | 330 | static inline unsigned int buddy_order(struct page *page) |
48f13bf3 | 331 | { |
572438f9 | 332 | /* PageBuddy() must be checked by the caller */ |
48f13bf3 MG |
333 | return page_private(page); |
334 | } | |
b5a0e011 | 335 | |
99c0fd5e | 336 | /* |
ab130f91 | 337 | * Like buddy_order(), but for callers who cannot afford to hold the zone lock. |
99c0fd5e VB |
338 | * PageBuddy() should be checked first by the caller to minimize race window, |
339 | * and invalid values must be handled gracefully. | |
340 | * | |
4db0c3c2 | 341 | * READ_ONCE is used so that if the caller assigns the result into a local |
99c0fd5e VB |
342 | * variable and e.g. tests it for valid range before using, the compiler cannot |
343 | * decide to remove the variable and inline the page_private(page) multiple | |
344 | * times, potentially observing different values in the tests and the actual | |
345 | * use of the result. | |
346 | */ | |
ab130f91 | 347 | #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) |
99c0fd5e | 348 | |
30bdbb78 KK |
349 | /* |
350 | * These three helpers classifies VMAs for virtual memory accounting. | |
351 | */ | |
352 | ||
353 | /* | |
354 | * Executable code area - executable, not writable, not stack | |
355 | */ | |
d977d56c KK |
356 | static inline bool is_exec_mapping(vm_flags_t flags) |
357 | { | |
30bdbb78 | 358 | return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; |
d977d56c KK |
359 | } |
360 | ||
30bdbb78 | 361 | /* |
f0953a1b | 362 | * Stack area - automatically grows in one direction |
30bdbb78 KK |
363 | * |
364 | * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: | |
365 | * do_mmap() forbids all other combinations. | |
366 | */ | |
d977d56c KK |
367 | static inline bool is_stack_mapping(vm_flags_t flags) |
368 | { | |
30bdbb78 | 369 | return (flags & VM_STACK) == VM_STACK; |
d977d56c KK |
370 | } |
371 | ||
30bdbb78 KK |
372 | /* |
373 | * Data area - private, writable, not stack | |
374 | */ | |
d977d56c KK |
375 | static inline bool is_data_mapping(vm_flags_t flags) |
376 | { | |
30bdbb78 | 377 | return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; |
d977d56c KK |
378 | } |
379 | ||
6038def0 NK |
380 | /* mm/util.c */ |
381 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, | |
aba6dfb7 | 382 | struct vm_area_struct *prev); |
1b9fc5b2 | 383 | void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma); |
6038def0 | 384 | |
af8e3354 | 385 | #ifdef CONFIG_MMU |
fc05f566 | 386 | extern long populate_vma_page_range(struct vm_area_struct *vma, |
a78f1ccd | 387 | unsigned long start, unsigned long end, int *locked); |
4ca9b385 DH |
388 | extern long faultin_vma_page_range(struct vm_area_struct *vma, |
389 | unsigned long start, unsigned long end, | |
390 | bool write, int *locked); | |
af8e3354 HD |
391 | extern void munlock_vma_pages_range(struct vm_area_struct *vma, |
392 | unsigned long start, unsigned long end); | |
393 | static inline void munlock_vma_pages_all(struct vm_area_struct *vma) | |
394 | { | |
395 | munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end); | |
396 | } | |
397 | ||
b291f000 | 398 | /* |
c1e8d7c6 | 399 | * must be called with vma's mmap_lock held for read or write, and page locked. |
b291f000 NP |
400 | */ |
401 | extern void mlock_vma_page(struct page *page); | |
ff6a6da6 | 402 | extern unsigned int munlock_vma_page(struct page *page); |
b291f000 | 403 | |
6aeb2542 MR |
404 | extern int mlock_future_check(struct mm_struct *mm, unsigned long flags, |
405 | unsigned long len); | |
406 | ||
b291f000 NP |
407 | /* |
408 | * Clear the page's PageMlocked(). This can be useful in a situation where | |
409 | * we want to unconditionally remove a page from the pagecache -- e.g., | |
410 | * on truncation or freeing. | |
411 | * | |
412 | * It is legal to call this function for any page, mlocked or not. | |
413 | * If called for a page that is still mapped by mlocked vmas, all we do | |
414 | * is revert to lazy LRU behaviour -- semantics are not broken. | |
415 | */ | |
e6c509f8 | 416 | extern void clear_page_mlock(struct page *page); |
b291f000 | 417 | |
f55e1014 | 418 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
b32967ff | 419 | |
e9b61f19 | 420 | /* |
494334e4 HD |
421 | * At what user virtual address is page expected in vma? |
422 | * Returns -EFAULT if all of the page is outside the range of vma. | |
423 | * If page is a compound head, the entire compound page is considered. | |
e9b61f19 KS |
424 | */ |
425 | static inline unsigned long | |
494334e4 | 426 | vma_address(struct page *page, struct vm_area_struct *vma) |
e9b61f19 | 427 | { |
494334e4 HD |
428 | pgoff_t pgoff; |
429 | unsigned long address; | |
430 | ||
431 | VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ | |
432 | pgoff = page_to_pgoff(page); | |
433 | if (pgoff >= vma->vm_pgoff) { | |
434 | address = vma->vm_start + | |
435 | ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); | |
436 | /* Check for address beyond vma (or wrapped through 0?) */ | |
437 | if (address < vma->vm_start || address >= vma->vm_end) | |
438 | address = -EFAULT; | |
439 | } else if (PageHead(page) && | |
440 | pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) { | |
441 | /* Test above avoids possibility of wrap to 0 on 32-bit */ | |
442 | address = vma->vm_start; | |
443 | } else { | |
444 | address = -EFAULT; | |
445 | } | |
446 | return address; | |
e9b61f19 KS |
447 | } |
448 | ||
494334e4 HD |
449 | /* |
450 | * Then at what user virtual address will none of the page be found in vma? | |
451 | * Assumes that vma_address() already returned a good starting address. | |
452 | * If page is a compound head, the entire compound page is considered. | |
453 | */ | |
e9b61f19 | 454 | static inline unsigned long |
494334e4 | 455 | vma_address_end(struct page *page, struct vm_area_struct *vma) |
e9b61f19 | 456 | { |
494334e4 HD |
457 | pgoff_t pgoff; |
458 | unsigned long address; | |
459 | ||
460 | VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ | |
461 | pgoff = page_to_pgoff(page) + compound_nr(page); | |
462 | address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); | |
463 | /* Check for address beyond vma (or wrapped through 0?) */ | |
464 | if (address < vma->vm_start || address > vma->vm_end) | |
465 | address = vma->vm_end; | |
466 | return address; | |
e9b61f19 KS |
467 | } |
468 | ||
89b15332 JW |
469 | static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf, |
470 | struct file *fpin) | |
471 | { | |
472 | int flags = vmf->flags; | |
473 | ||
474 | if (fpin) | |
475 | return fpin; | |
476 | ||
477 | /* | |
478 | * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or | |
c1e8d7c6 | 479 | * anything, so we only pin the file and drop the mmap_lock if only |
4064b982 | 480 | * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. |
89b15332 | 481 | */ |
4064b982 PX |
482 | if (fault_flag_allow_retry_first(flags) && |
483 | !(flags & FAULT_FLAG_RETRY_NOWAIT)) { | |
89b15332 | 484 | fpin = get_file(vmf->vma->vm_file); |
d8ed45c5 | 485 | mmap_read_unlock(vmf->vma->vm_mm); |
89b15332 JW |
486 | } |
487 | return fpin; | |
488 | } | |
489 | ||
af8e3354 | 490 | #else /* !CONFIG_MMU */ |
b291f000 NP |
491 | static inline void clear_page_mlock(struct page *page) { } |
492 | static inline void mlock_vma_page(struct page *page) { } | |
4ad0ae8c NP |
493 | static inline void vunmap_range_noflush(unsigned long start, unsigned long end) |
494 | { | |
495 | } | |
af8e3354 | 496 | #endif /* !CONFIG_MMU */ |
894bc310 | 497 | |
69d177c2 AW |
498 | /* |
499 | * Return the mem_map entry representing the 'offset' subpage within | |
500 | * the maximally aligned gigantic page 'base'. Handle any discontiguity | |
501 | * in the mem_map at MAX_ORDER_NR_PAGES boundaries. | |
502 | */ | |
503 | static inline struct page *mem_map_offset(struct page *base, int offset) | |
504 | { | |
505 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) | |
bc7f84c0 | 506 | return nth_page(base, offset); |
69d177c2 AW |
507 | return base + offset; |
508 | } | |
509 | ||
510 | /* | |
25985edc | 511 | * Iterator over all subpages within the maximally aligned gigantic |
69d177c2 AW |
512 | * page 'base'. Handle any discontiguity in the mem_map. |
513 | */ | |
514 | static inline struct page *mem_map_next(struct page *iter, | |
515 | struct page *base, int offset) | |
516 | { | |
517 | if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { | |
518 | unsigned long pfn = page_to_pfn(base) + offset; | |
519 | if (!pfn_valid(pfn)) | |
520 | return NULL; | |
521 | return pfn_to_page(pfn); | |
522 | } | |
523 | return iter + 1; | |
524 | } | |
525 | ||
6b74ab97 MG |
526 | /* Memory initialisation debug and verification */ |
527 | enum mminit_level { | |
528 | MMINIT_WARNING, | |
529 | MMINIT_VERIFY, | |
530 | MMINIT_TRACE | |
531 | }; | |
532 | ||
533 | #ifdef CONFIG_DEBUG_MEMORY_INIT | |
534 | ||
535 | extern int mminit_loglevel; | |
536 | ||
537 | #define mminit_dprintk(level, prefix, fmt, arg...) \ | |
538 | do { \ | |
539 | if (level < mminit_loglevel) { \ | |
fc5199d1 | 540 | if (level <= MMINIT_WARNING) \ |
1170532b | 541 | pr_warn("mminit::" prefix " " fmt, ##arg); \ |
fc5199d1 RV |
542 | else \ |
543 | printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ | |
6b74ab97 MG |
544 | } \ |
545 | } while (0) | |
546 | ||
708614e6 | 547 | extern void mminit_verify_pageflags_layout(void); |
68ad8df4 | 548 | extern void mminit_verify_zonelist(void); |
6b74ab97 MG |
549 | #else |
550 | ||
551 | static inline void mminit_dprintk(enum mminit_level level, | |
552 | const char *prefix, const char *fmt, ...) | |
553 | { | |
554 | } | |
555 | ||
708614e6 MG |
556 | static inline void mminit_verify_pageflags_layout(void) |
557 | { | |
558 | } | |
559 | ||
68ad8df4 MG |
560 | static inline void mminit_verify_zonelist(void) |
561 | { | |
562 | } | |
6b74ab97 | 563 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
2dbb51c4 MG |
564 | |
565 | /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ | |
566 | #if defined(CONFIG_SPARSEMEM) | |
567 | extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, | |
568 | unsigned long *end_pfn); | |
569 | #else | |
570 | static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, | |
571 | unsigned long *end_pfn) | |
572 | { | |
573 | } | |
574 | #endif /* CONFIG_SPARSEMEM */ | |
575 | ||
a5f5f91d MG |
576 | #define NODE_RECLAIM_NOSCAN -2 |
577 | #define NODE_RECLAIM_FULL -1 | |
578 | #define NODE_RECLAIM_SOME 0 | |
579 | #define NODE_RECLAIM_SUCCESS 1 | |
7c116f2b | 580 | |
8b09549c WY |
581 | #ifdef CONFIG_NUMA |
582 | extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); | |
79c28a41 | 583 | extern int find_next_best_node(int node, nodemask_t *used_node_mask); |
8b09549c WY |
584 | #else |
585 | static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, | |
586 | unsigned int order) | |
587 | { | |
588 | return NODE_RECLAIM_NOSCAN; | |
589 | } | |
79c28a41 DH |
590 | static inline int find_next_best_node(int node, nodemask_t *used_node_mask) |
591 | { | |
592 | return NUMA_NO_NODE; | |
593 | } | |
8b09549c WY |
594 | #endif |
595 | ||
31d3d348 WF |
596 | extern int hwpoison_filter(struct page *p); |
597 | ||
7c116f2b WF |
598 | extern u32 hwpoison_filter_dev_major; |
599 | extern u32 hwpoison_filter_dev_minor; | |
478c5ffc WF |
600 | extern u64 hwpoison_filter_flags_mask; |
601 | extern u64 hwpoison_filter_flags_value; | |
4fd466eb | 602 | extern u64 hwpoison_filter_memcg; |
1bfe5feb | 603 | extern u32 hwpoison_filter_enable; |
eb36c587 | 604 | |
dc0ef0df | 605 | extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, |
eb36c587 | 606 | unsigned long, unsigned long, |
9fbeb5ab | 607 | unsigned long, unsigned long); |
ca57df79 XQ |
608 | |
609 | extern void set_pageblock_order(void); | |
730ec8c0 | 610 | unsigned int reclaim_clean_pages_from_list(struct zone *zone, |
02c6de8d | 611 | struct list_head *page_list); |
d95ea5d1 BZ |
612 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
613 | #define ALLOC_WMARK_MIN WMARK_MIN | |
614 | #define ALLOC_WMARK_LOW WMARK_LOW | |
615 | #define ALLOC_WMARK_HIGH WMARK_HIGH | |
616 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ | |
617 | ||
618 | /* Mask to get the watermark bits */ | |
619 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) | |
620 | ||
cd04ae1e MH |
621 | /* |
622 | * Only MMU archs have async oom victim reclaim - aka oom_reaper so we | |
623 | * cannot assume a reduced access to memory reserves is sufficient for | |
624 | * !MMU | |
625 | */ | |
626 | #ifdef CONFIG_MMU | |
627 | #define ALLOC_OOM 0x08 | |
628 | #else | |
629 | #define ALLOC_OOM ALLOC_NO_WATERMARKS | |
630 | #endif | |
631 | ||
6bb15450 MG |
632 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ |
633 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ | |
634 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ | |
635 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ | |
636 | #ifdef CONFIG_ZONE_DMA32 | |
637 | #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */ | |
638 | #else | |
639 | #define ALLOC_NOFRAGMENT 0x0 | |
640 | #endif | |
736838e9 | 641 | #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ |
d95ea5d1 | 642 | |
72b252ae MG |
643 | enum ttu_flags; |
644 | struct tlbflush_unmap_batch; | |
645 | ||
ce612879 MH |
646 | |
647 | /* | |
648 | * only for MM internal work items which do not depend on | |
649 | * any allocations or locks which might depend on allocations | |
650 | */ | |
651 | extern struct workqueue_struct *mm_percpu_wq; | |
652 | ||
72b252ae MG |
653 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
654 | void try_to_unmap_flush(void); | |
d950c947 | 655 | void try_to_unmap_flush_dirty(void); |
3ea27719 | 656 | void flush_tlb_batched_pending(struct mm_struct *mm); |
72b252ae MG |
657 | #else |
658 | static inline void try_to_unmap_flush(void) | |
659 | { | |
660 | } | |
d950c947 MG |
661 | static inline void try_to_unmap_flush_dirty(void) |
662 | { | |
663 | } | |
3ea27719 MG |
664 | static inline void flush_tlb_batched_pending(struct mm_struct *mm) |
665 | { | |
666 | } | |
72b252ae | 667 | #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ |
edf14cdb VB |
668 | |
669 | extern const struct trace_print_flags pageflag_names[]; | |
670 | extern const struct trace_print_flags vmaflag_names[]; | |
671 | extern const struct trace_print_flags gfpflag_names[]; | |
672 | ||
a6ffdc07 XQ |
673 | static inline bool is_migrate_highatomic(enum migratetype migratetype) |
674 | { | |
675 | return migratetype == MIGRATE_HIGHATOMIC; | |
676 | } | |
677 | ||
678 | static inline bool is_migrate_highatomic_page(struct page *page) | |
679 | { | |
680 | return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC; | |
681 | } | |
682 | ||
72675e13 | 683 | void setup_zone_pageset(struct zone *zone); |
19fc7bed JK |
684 | |
685 | struct migration_target_control { | |
686 | int nid; /* preferred node id */ | |
687 | nodemask_t *nmask; | |
688 | gfp_t gfp_mask; | |
689 | }; | |
690 | ||
b67177ec NP |
691 | /* |
692 | * mm/vmalloc.c | |
693 | */ | |
4ad0ae8c | 694 | #ifdef CONFIG_MMU |
b67177ec NP |
695 | int vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
696 | pgprot_t prot, struct page **pages, unsigned int page_shift); | |
4ad0ae8c NP |
697 | #else |
698 | static inline | |
699 | int vmap_pages_range_noflush(unsigned long addr, unsigned long end, | |
700 | pgprot_t prot, struct page **pages, unsigned int page_shift) | |
701 | { | |
702 | return -EINVAL; | |
703 | } | |
704 | #endif | |
705 | ||
706 | void vunmap_range_noflush(unsigned long start, unsigned long end); | |
b67177ec | 707 | |
f4c0d836 YS |
708 | int numa_migrate_prep(struct page *page, struct vm_area_struct *vma, |
709 | unsigned long addr, int page_nid, int *flags); | |
710 | ||
db971418 | 711 | #endif /* __MM_INTERNAL_H */ |