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> |
2aff7a47 | 13 | #include <linux/rmap.h> |
edf14cdb | 14 | #include <linux/tracepoint-defs.h> |
1da177e4 | 15 | |
0e499ed3 MWO |
16 | struct folio_batch; |
17 | ||
dd56b046 MG |
18 | /* |
19 | * The set of flags that only affect watermark checking and reclaim | |
20 | * behaviour. This is used by the MM to obey the caller constraints | |
21 | * about IO, FS and watermark checking while ignoring placement | |
22 | * hints such as HIGHMEM usage. | |
23 | */ | |
24 | #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\ | |
dcda9b04 | 25 | __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\ |
e838a45f | 26 | __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\ |
704687de | 27 | __GFP_ATOMIC|__GFP_NOLOCKDEP) |
dd56b046 MG |
28 | |
29 | /* The GFP flags allowed during early boot */ | |
30 | #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS)) | |
31 | ||
32 | /* Control allocation cpuset and node placement constraints */ | |
33 | #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) | |
34 | ||
35 | /* Do not use these with a slab allocator */ | |
36 | #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) | |
37 | ||
62906027 NP |
38 | void page_writeback_init(void); |
39 | ||
64601000 MWO |
40 | static inline void *folio_raw_mapping(struct folio *folio) |
41 | { | |
42 | unsigned long mapping = (unsigned long)folio->mapping; | |
43 | ||
44 | return (void *)(mapping & ~PAGE_MAPPING_FLAGS); | |
45 | } | |
46 | ||
512b7931 | 47 | void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, |
8cd7c588 | 48 | int nr_throttled); |
512b7931 | 49 | static inline void acct_reclaim_writeback(struct folio *folio) |
8cd7c588 | 50 | { |
512b7931 | 51 | pg_data_t *pgdat = folio_pgdat(folio); |
8cd7c588 MG |
52 | int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled); |
53 | ||
54 | if (nr_throttled) | |
512b7931 | 55 | __acct_reclaim_writeback(pgdat, folio, nr_throttled); |
8cd7c588 MG |
56 | } |
57 | ||
d818fca1 MG |
58 | static inline void wake_throttle_isolated(pg_data_t *pgdat) |
59 | { | |
60 | wait_queue_head_t *wqh; | |
61 | ||
62 | wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED]; | |
63 | if (waitqueue_active(wqh)) | |
64 | wake_up(wqh); | |
65 | } | |
66 | ||
2b740303 | 67 | vm_fault_t do_swap_page(struct vm_fault *vmf); |
575ced1c | 68 | void folio_rotate_reclaimable(struct folio *folio); |
269ccca3 | 69 | bool __folio_end_writeback(struct folio *folio); |
261b6840 | 70 | void deactivate_file_folio(struct folio *folio); |
8a966ed7 | 71 | |
42b77728 JB |
72 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
73 | unsigned long floor, unsigned long ceiling); | |
03c4f204 | 74 | void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); |
42b77728 | 75 | |
3506659e | 76 | struct zap_details; |
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 | ||
56a4d67c MWO |
82 | void page_cache_ra_order(struct readahead_control *, struct file_ra_state *, |
83 | unsigned int order); | |
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 | 92 | unsigned find_lock_entries(struct address_space *mapping, pgoff_t start, |
51dcbdac | 93 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
0e499ed3 MWO |
94 | unsigned find_get_entries(struct address_space *mapping, pgoff_t start, |
95 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); | |
78f42660 | 96 | void filemap_free_folio(struct address_space *mapping, struct folio *folio); |
1e84a3d9 | 97 | int truncate_inode_folio(struct address_space *mapping, struct folio *folio); |
b9a8a419 MWO |
98 | bool truncate_inode_partial_folio(struct folio *folio, loff_t start, |
99 | loff_t end); | |
d6c75dc2 | 100 | long invalidate_inode_page(struct page *page); |
c56109dd MWO |
101 | unsigned long invalidate_mapping_pagevec(struct address_space *mapping, |
102 | pgoff_t start, pgoff_t end, unsigned long *nr_pagevec); | |
5c211ba2 | 103 | |
1eb6234e | 104 | /** |
3eed3ef5 MWO |
105 | * folio_evictable - Test whether a folio is evictable. |
106 | * @folio: The folio to test. | |
1eb6234e | 107 | * |
3eed3ef5 MWO |
108 | * Test whether @folio is evictable -- i.e., should be placed on |
109 | * active/inactive lists vs unevictable list. | |
1eb6234e | 110 | * |
3eed3ef5 MWO |
111 | * Reasons folio might not be evictable: |
112 | * 1. folio's mapping marked unevictable | |
113 | * 2. One of the pages in the folio is part of an mlocked VMA | |
1eb6234e | 114 | */ |
3eed3ef5 MWO |
115 | static inline bool folio_evictable(struct folio *folio) |
116 | { | |
117 | bool ret; | |
118 | ||
119 | /* Prevent address_space of inode and swap cache from being freed */ | |
120 | rcu_read_lock(); | |
121 | ret = !mapping_unevictable(folio_mapping(folio)) && | |
122 | !folio_test_mlocked(folio); | |
123 | rcu_read_unlock(); | |
124 | return ret; | |
125 | } | |
126 | ||
1eb6234e YS |
127 | static inline bool page_evictable(struct page *page) |
128 | { | |
129 | bool ret; | |
130 | ||
131 | /* Prevent address_space of inode and swap cache from being freed */ | |
132 | rcu_read_lock(); | |
133 | ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page); | |
134 | rcu_read_unlock(); | |
135 | return ret; | |
136 | } | |
137 | ||
7835e98b | 138 | /* |
0139aa7b | 139 | * Turn a non-refcounted page (->_refcount == 0) into refcounted with |
7835e98b NP |
140 | * a count of one. |
141 | */ | |
142 | static inline void set_page_refcounted(struct page *page) | |
143 | { | |
309381fe | 144 | VM_BUG_ON_PAGE(PageTail(page), page); |
fe896d18 | 145 | VM_BUG_ON_PAGE(page_ref_count(page), page); |
77a8a788 | 146 | set_page_count(page, 1); |
77a8a788 NP |
147 | } |
148 | ||
03f6462a HD |
149 | extern unsigned long highest_memmap_pfn; |
150 | ||
c73322d0 JW |
151 | /* |
152 | * Maximum number of reclaim retries without progress before the OOM | |
153 | * killer is consider the only way forward. | |
154 | */ | |
155 | #define MAX_RECLAIM_RETRIES 16 | |
156 | ||
be4893d9 VB |
157 | /* |
158 | * in mm/early_ioremap.c | |
159 | */ | |
160 | pgprot_t __init early_memremap_pgprot_adjust(resource_size_t phys_addr, | |
161 | unsigned long size, pgprot_t prot); | |
162 | ||
894bc310 LS |
163 | /* |
164 | * in mm/vmscan.c: | |
165 | */ | |
d1d8a3b4 MWO |
166 | int isolate_lru_page(struct page *page); |
167 | int folio_isolate_lru(struct folio *folio); | |
ca6d60f3 MWO |
168 | void putback_lru_page(struct page *page); |
169 | void folio_putback_lru(struct folio *folio); | |
c3f4a9a2 | 170 | extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); |
62695a84 | 171 | |
6219049a BL |
172 | /* |
173 | * in mm/rmap.c: | |
174 | */ | |
175 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); | |
176 | ||
894bc310 LS |
177 | /* |
178 | * in mm/page_alloc.c | |
179 | */ | |
3c605096 | 180 | |
1a6d53a1 VB |
181 | /* |
182 | * Structure for holding the mostly immutable allocation parameters passed | |
183 | * between functions involved in allocations, including the alloc_pages* | |
184 | * family of functions. | |
185 | * | |
97a225e6 | 186 | * nodemask, migratetype and highest_zoneidx are initialized only once in |
84172f4b | 187 | * __alloc_pages() and then never change. |
1a6d53a1 | 188 | * |
97a225e6 | 189 | * zonelist, preferred_zone and highest_zoneidx are set first in |
84172f4b | 190 | * __alloc_pages() for the fast path, and might be later changed |
68956ccb | 191 | * in __alloc_pages_slowpath(). All other functions pass the whole structure |
1a6d53a1 VB |
192 | * by a const pointer. |
193 | */ | |
194 | struct alloc_context { | |
195 | struct zonelist *zonelist; | |
196 | nodemask_t *nodemask; | |
c33d6c06 | 197 | struct zoneref *preferred_zoneref; |
1a6d53a1 | 198 | int migratetype; |
97a225e6 JK |
199 | |
200 | /* | |
201 | * highest_zoneidx represents highest usable zone index of | |
202 | * the allocation request. Due to the nature of the zone, | |
203 | * memory on lower zone than the highest_zoneidx will be | |
204 | * protected by lowmem_reserve[highest_zoneidx]. | |
205 | * | |
206 | * highest_zoneidx is also used by reclaim/compaction to limit | |
207 | * the target zone since higher zone than this index cannot be | |
208 | * usable for this allocation request. | |
209 | */ | |
210 | enum zone_type highest_zoneidx; | |
c9ab0c4f | 211 | bool spread_dirty_pages; |
1a6d53a1 VB |
212 | }; |
213 | ||
3c605096 JK |
214 | /* |
215 | * Locate the struct page for both the matching buddy in our | |
216 | * pair (buddy1) and the combined O(n+1) page they form (page). | |
217 | * | |
218 | * 1) Any buddy B1 will have an order O twin B2 which satisfies | |
219 | * the following equation: | |
220 | * B2 = B1 ^ (1 << O) | |
221 | * For example, if the starting buddy (buddy2) is #8 its order | |
222 | * 1 buddy is #10: | |
223 | * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 | |
224 | * | |
225 | * 2) Any buddy B will have an order O+1 parent P which | |
226 | * satisfies the following equation: | |
227 | * P = B & ~(1 << O) | |
228 | * | |
229 | * Assumption: *_mem_map is contiguous at least up to MAX_ORDER | |
230 | */ | |
231 | static inline unsigned long | |
76741e77 | 232 | __find_buddy_pfn(unsigned long page_pfn, unsigned int order) |
3c605096 | 233 | { |
76741e77 | 234 | return page_pfn ^ (1 << order); |
3c605096 JK |
235 | } |
236 | ||
7cf91a98 JK |
237 | extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, |
238 | unsigned long end_pfn, struct zone *zone); | |
239 | ||
240 | static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, | |
241 | unsigned long end_pfn, struct zone *zone) | |
242 | { | |
243 | if (zone->contiguous) | |
244 | return pfn_to_page(start_pfn); | |
245 | ||
246 | return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); | |
247 | } | |
248 | ||
3c605096 | 249 | extern int __isolate_free_page(struct page *page, unsigned int order); |
624f58d8 AD |
250 | extern void __putback_isolated_page(struct page *page, unsigned int order, |
251 | int mt); | |
7c2ee349 | 252 | extern void memblock_free_pages(struct page *page, unsigned long pfn, |
d70ddd7a | 253 | unsigned int order); |
a9cd410a | 254 | extern void __free_pages_core(struct page *page, unsigned int order); |
d00181b9 | 255 | extern void prep_compound_page(struct page *page, unsigned int order); |
46f24fd8 JK |
256 | extern void post_alloc_hook(struct page *page, unsigned int order, |
257 | gfp_t gfp_flags); | |
42aa83cb | 258 | extern int user_min_free_kbytes; |
20a0307c | 259 | |
44042b44 | 260 | extern void free_unref_page(struct page *page, unsigned int order); |
0966aeb4 MWO |
261 | extern void free_unref_page_list(struct list_head *list); |
262 | ||
04f8cfea | 263 | extern void zone_pcp_update(struct zone *zone, int cpu_online); |
68265390 | 264 | extern void zone_pcp_reset(struct zone *zone); |
ec6e8c7e VB |
265 | extern void zone_pcp_disable(struct zone *zone); |
266 | extern void zone_pcp_enable(struct zone *zone); | |
68265390 | 267 | |
c803b3c8 MR |
268 | extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, |
269 | phys_addr_t min_addr, | |
270 | int nid, bool exact_nid); | |
271 | ||
ff9543fd MN |
272 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
273 | ||
274 | /* | |
275 | * in mm/compaction.c | |
276 | */ | |
277 | /* | |
278 | * compact_control is used to track pages being migrated and the free pages | |
279 | * they are being migrated to during memory compaction. The free_pfn starts | |
280 | * at the end of a zone and migrate_pfn begins at the start. Movable pages | |
281 | * are moved to the end of a zone during a compaction run and the run | |
282 | * completes when free_pfn <= migrate_pfn | |
283 | */ | |
284 | struct compact_control { | |
285 | struct list_head freepages; /* List of free pages to migrate to */ | |
286 | struct list_head migratepages; /* List of pages being migrated */ | |
c5fbd937 MG |
287 | unsigned int nr_freepages; /* Number of isolated free pages */ |
288 | unsigned int nr_migratepages; /* Number of pages to migrate */ | |
ff9543fd | 289 | unsigned long free_pfn; /* isolate_freepages search base */ |
c2ad7a1f OS |
290 | /* |
291 | * Acts as an in/out parameter to page isolation for migration. | |
292 | * isolate_migratepages uses it as a search base. | |
293 | * isolate_migratepages_block will update the value to the next pfn | |
294 | * after the last isolated one. | |
295 | */ | |
296 | unsigned long migrate_pfn; | |
70b44595 | 297 | unsigned long fast_start_pfn; /* a pfn to start linear scan from */ |
c5943b9c MG |
298 | struct zone *zone; |
299 | unsigned long total_migrate_scanned; | |
300 | unsigned long total_free_scanned; | |
dbe2d4e4 MG |
301 | unsigned short fast_search_fail;/* failures to use free list searches */ |
302 | short search_order; /* order to start a fast search at */ | |
f25ba6dc VB |
303 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
304 | int order; /* order a direct compactor needs */ | |
d39773a0 | 305 | int migratetype; /* migratetype of direct compactor */ |
f25ba6dc | 306 | const unsigned int alloc_flags; /* alloc flags of a direct compactor */ |
97a225e6 | 307 | const int highest_zoneidx; /* zone index of a direct compactor */ |
e0b9daeb | 308 | enum migrate_mode mode; /* Async or sync migration mode */ |
bb13ffeb | 309 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
2583d671 | 310 | bool no_set_skip_hint; /* Don't mark blocks for skipping */ |
9f7e3387 | 311 | bool ignore_block_suitable; /* Scan blocks considered unsuitable */ |
accf6242 | 312 | bool direct_compaction; /* False from kcompactd or /proc/... */ |
facdaa91 | 313 | bool proactive_compaction; /* kcompactd proactive compaction */ |
06ed2998 | 314 | bool whole_zone; /* Whole zone should/has been scanned */ |
c3486f53 | 315 | bool contended; /* Signal lock or sched contention */ |
804d3121 | 316 | bool rescan; /* Rescanning the same pageblock */ |
b06eda09 | 317 | bool alloc_contig; /* alloc_contig_range allocation */ |
ff9543fd MN |
318 | }; |
319 | ||
5e1f0f09 MG |
320 | /* |
321 | * Used in direct compaction when a page should be taken from the freelists | |
322 | * immediately when one is created during the free path. | |
323 | */ | |
324 | struct capture_control { | |
325 | struct compact_control *cc; | |
326 | struct page *page; | |
327 | }; | |
328 | ||
ff9543fd | 329 | unsigned long |
bb13ffeb MG |
330 | isolate_freepages_range(struct compact_control *cc, |
331 | unsigned long start_pfn, unsigned long end_pfn); | |
c2ad7a1f | 332 | int |
edc2ca61 VB |
333 | isolate_migratepages_range(struct compact_control *cc, |
334 | unsigned long low_pfn, unsigned long end_pfn); | |
ffd8f251 | 335 | #endif |
2149cdae JK |
336 | int find_suitable_fallback(struct free_area *area, unsigned int order, |
337 | int migratetype, bool only_stealable, bool *can_steal); | |
ff9543fd | 338 | |
48f13bf3 | 339 | /* |
6c14466c MG |
340 | * This function returns the order of a free page in the buddy system. In |
341 | * general, page_zone(page)->lock must be held by the caller to prevent the | |
342 | * page from being allocated in parallel and returning garbage as the order. | |
343 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the | |
99c0fd5e | 344 | * page cannot be allocated or merged in parallel. Alternatively, it must |
ab130f91 | 345 | * handle invalid values gracefully, and use buddy_order_unsafe() below. |
48f13bf3 | 346 | */ |
ab130f91 | 347 | static inline unsigned int buddy_order(struct page *page) |
48f13bf3 | 348 | { |
572438f9 | 349 | /* PageBuddy() must be checked by the caller */ |
48f13bf3 MG |
350 | return page_private(page); |
351 | } | |
b5a0e011 | 352 | |
99c0fd5e | 353 | /* |
ab130f91 | 354 | * Like buddy_order(), but for callers who cannot afford to hold the zone lock. |
99c0fd5e VB |
355 | * PageBuddy() should be checked first by the caller to minimize race window, |
356 | * and invalid values must be handled gracefully. | |
357 | * | |
4db0c3c2 | 358 | * READ_ONCE is used so that if the caller assigns the result into a local |
99c0fd5e VB |
359 | * variable and e.g. tests it for valid range before using, the compiler cannot |
360 | * decide to remove the variable and inline the page_private(page) multiple | |
361 | * times, potentially observing different values in the tests and the actual | |
362 | * use of the result. | |
363 | */ | |
ab130f91 | 364 | #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) |
99c0fd5e | 365 | |
30bdbb78 KK |
366 | /* |
367 | * These three helpers classifies VMAs for virtual memory accounting. | |
368 | */ | |
369 | ||
370 | /* | |
371 | * Executable code area - executable, not writable, not stack | |
372 | */ | |
d977d56c KK |
373 | static inline bool is_exec_mapping(vm_flags_t flags) |
374 | { | |
30bdbb78 | 375 | return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; |
d977d56c KK |
376 | } |
377 | ||
30bdbb78 | 378 | /* |
f0953a1b | 379 | * Stack area - automatically grows in one direction |
30bdbb78 KK |
380 | * |
381 | * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: | |
382 | * do_mmap() forbids all other combinations. | |
383 | */ | |
d977d56c KK |
384 | static inline bool is_stack_mapping(vm_flags_t flags) |
385 | { | |
30bdbb78 | 386 | return (flags & VM_STACK) == VM_STACK; |
d977d56c KK |
387 | } |
388 | ||
30bdbb78 KK |
389 | /* |
390 | * Data area - private, writable, not stack | |
391 | */ | |
d977d56c KK |
392 | static inline bool is_data_mapping(vm_flags_t flags) |
393 | { | |
30bdbb78 | 394 | return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; |
d977d56c KK |
395 | } |
396 | ||
6038def0 NK |
397 | /* mm/util.c */ |
398 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, | |
aba6dfb7 | 399 | struct vm_area_struct *prev); |
1b9fc5b2 | 400 | void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma); |
e05b3453 | 401 | struct anon_vma *folio_anon_vma(struct folio *folio); |
6038def0 | 402 | |
af8e3354 | 403 | #ifdef CONFIG_MMU |
3506659e | 404 | void unmap_mapping_folio(struct folio *folio); |
fc05f566 | 405 | extern long populate_vma_page_range(struct vm_area_struct *vma, |
a78f1ccd | 406 | unsigned long start, unsigned long end, int *locked); |
4ca9b385 DH |
407 | extern long faultin_vma_page_range(struct vm_area_struct *vma, |
408 | unsigned long start, unsigned long end, | |
409 | bool write, int *locked); | |
6aeb2542 MR |
410 | extern int mlock_future_check(struct mm_struct *mm, unsigned long flags, |
411 | unsigned long len); | |
b291f000 | 412 | /* |
cea86fe2 HD |
413 | * mlock_vma_page() and munlock_vma_page(): |
414 | * should be called with vma's mmap_lock held for read or write, | |
415 | * under page table lock for the pte/pmd being added or removed. | |
b291f000 | 416 | * |
cea86fe2 HD |
417 | * mlock is usually called at the end of page_add_*_rmap(), |
418 | * munlock at the end of page_remove_rmap(); but new anon | |
2fbb0c10 HD |
419 | * pages are managed by lru_cache_add_inactive_or_unevictable() |
420 | * calling mlock_new_page(). | |
cea86fe2 HD |
421 | * |
422 | * @compound is used to include pmd mappings of THPs, but filter out | |
423 | * pte mappings of THPs, which cannot be consistently counted: a pte | |
424 | * mapping of the THP head cannot be distinguished by the page alone. | |
b291f000 | 425 | */ |
dcc5d337 MWO |
426 | void mlock_folio(struct folio *folio); |
427 | static inline void mlock_vma_folio(struct folio *folio, | |
cea86fe2 HD |
428 | struct vm_area_struct *vma, bool compound) |
429 | { | |
c8263bd6 HD |
430 | /* |
431 | * The VM_SPECIAL check here serves two purposes. | |
432 | * 1) VM_IO check prevents migration from double-counting during mlock. | |
433 | * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED | |
434 | * is never left set on a VM_SPECIAL vma, there is an interval while | |
435 | * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may | |
436 | * still be set while VM_SPECIAL bits are added: so ignore it then. | |
437 | */ | |
438 | if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED) && | |
dcc5d337 MWO |
439 | (compound || !folio_test_large(folio))) |
440 | mlock_folio(folio); | |
441 | } | |
442 | ||
443 | static inline void mlock_vma_page(struct page *page, | |
444 | struct vm_area_struct *vma, bool compound) | |
445 | { | |
446 | mlock_vma_folio(page_folio(page), vma, compound); | |
cea86fe2 | 447 | } |
dcc5d337 | 448 | |
cea86fe2 HD |
449 | void munlock_page(struct page *page); |
450 | static inline void munlock_vma_page(struct page *page, | |
451 | struct vm_area_struct *vma, bool compound) | |
452 | { | |
453 | if (unlikely(vma->vm_flags & VM_LOCKED) && | |
454 | (compound || !PageTransCompound(page))) | |
455 | munlock_page(page); | |
456 | } | |
2fbb0c10 HD |
457 | void mlock_new_page(struct page *page); |
458 | bool need_mlock_page_drain(int cpu); | |
adb11e78 SAS |
459 | void mlock_page_drain_local(void); |
460 | void mlock_page_drain_remote(int cpu); | |
b291f000 | 461 | |
f55e1014 | 462 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
b32967ff | 463 | |
e9b61f19 | 464 | /* |
494334e4 HD |
465 | * At what user virtual address is page expected in vma? |
466 | * Returns -EFAULT if all of the page is outside the range of vma. | |
467 | * If page is a compound head, the entire compound page is considered. | |
e9b61f19 KS |
468 | */ |
469 | static inline unsigned long | |
494334e4 | 470 | vma_address(struct page *page, struct vm_area_struct *vma) |
e9b61f19 | 471 | { |
494334e4 HD |
472 | pgoff_t pgoff; |
473 | unsigned long address; | |
474 | ||
475 | VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ | |
476 | pgoff = page_to_pgoff(page); | |
477 | if (pgoff >= vma->vm_pgoff) { | |
478 | address = vma->vm_start + | |
479 | ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); | |
480 | /* Check for address beyond vma (or wrapped through 0?) */ | |
481 | if (address < vma->vm_start || address >= vma->vm_end) | |
482 | address = -EFAULT; | |
483 | } else if (PageHead(page) && | |
484 | pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) { | |
485 | /* Test above avoids possibility of wrap to 0 on 32-bit */ | |
486 | address = vma->vm_start; | |
487 | } else { | |
488 | address = -EFAULT; | |
489 | } | |
490 | return address; | |
e9b61f19 KS |
491 | } |
492 | ||
494334e4 | 493 | /* |
2aff7a47 | 494 | * Then at what user virtual address will none of the range be found in vma? |
494334e4 | 495 | * Assumes that vma_address() already returned a good starting address. |
494334e4 | 496 | */ |
2aff7a47 | 497 | static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw) |
e9b61f19 | 498 | { |
2aff7a47 | 499 | struct vm_area_struct *vma = pvmw->vma; |
494334e4 HD |
500 | pgoff_t pgoff; |
501 | unsigned long address; | |
502 | ||
2aff7a47 MWO |
503 | /* Common case, plus ->pgoff is invalid for KSM */ |
504 | if (pvmw->nr_pages == 1) | |
505 | return pvmw->address + PAGE_SIZE; | |
506 | ||
507 | pgoff = pvmw->pgoff + pvmw->nr_pages; | |
494334e4 HD |
508 | address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); |
509 | /* Check for address beyond vma (or wrapped through 0?) */ | |
510 | if (address < vma->vm_start || address > vma->vm_end) | |
511 | address = vma->vm_end; | |
512 | return address; | |
e9b61f19 KS |
513 | } |
514 | ||
89b15332 JW |
515 | static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf, |
516 | struct file *fpin) | |
517 | { | |
518 | int flags = vmf->flags; | |
519 | ||
520 | if (fpin) | |
521 | return fpin; | |
522 | ||
523 | /* | |
524 | * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or | |
c1e8d7c6 | 525 | * anything, so we only pin the file and drop the mmap_lock if only |
4064b982 | 526 | * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. |
89b15332 | 527 | */ |
4064b982 PX |
528 | if (fault_flag_allow_retry_first(flags) && |
529 | !(flags & FAULT_FLAG_RETRY_NOWAIT)) { | |
89b15332 | 530 | fpin = get_file(vmf->vma->vm_file); |
d8ed45c5 | 531 | mmap_read_unlock(vmf->vma->vm_mm); |
89b15332 JW |
532 | } |
533 | return fpin; | |
534 | } | |
af8e3354 | 535 | #else /* !CONFIG_MMU */ |
3506659e | 536 | static inline void unmap_mapping_folio(struct folio *folio) { } |
cea86fe2 HD |
537 | static inline void mlock_vma_page(struct page *page, |
538 | struct vm_area_struct *vma, bool compound) { } | |
539 | static inline void munlock_vma_page(struct page *page, | |
540 | struct vm_area_struct *vma, bool compound) { } | |
2fbb0c10 HD |
541 | static inline void mlock_new_page(struct page *page) { } |
542 | static inline bool need_mlock_page_drain(int cpu) { return false; } | |
adb11e78 SAS |
543 | static inline void mlock_page_drain_local(void) { } |
544 | static inline void mlock_page_drain_remote(int cpu) { } | |
4ad0ae8c NP |
545 | static inline void vunmap_range_noflush(unsigned long start, unsigned long end) |
546 | { | |
547 | } | |
af8e3354 | 548 | #endif /* !CONFIG_MMU */ |
894bc310 | 549 | |
69d177c2 AW |
550 | /* |
551 | * Return the mem_map entry representing the 'offset' subpage within | |
552 | * the maximally aligned gigantic page 'base'. Handle any discontiguity | |
553 | * in the mem_map at MAX_ORDER_NR_PAGES boundaries. | |
554 | */ | |
555 | static inline struct page *mem_map_offset(struct page *base, int offset) | |
556 | { | |
557 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) | |
bc7f84c0 | 558 | return nth_page(base, offset); |
69d177c2 AW |
559 | return base + offset; |
560 | } | |
561 | ||
562 | /* | |
25985edc | 563 | * Iterator over all subpages within the maximally aligned gigantic |
69d177c2 AW |
564 | * page 'base'. Handle any discontiguity in the mem_map. |
565 | */ | |
566 | static inline struct page *mem_map_next(struct page *iter, | |
567 | struct page *base, int offset) | |
568 | { | |
569 | if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { | |
570 | unsigned long pfn = page_to_pfn(base) + offset; | |
571 | if (!pfn_valid(pfn)) | |
572 | return NULL; | |
573 | return pfn_to_page(pfn); | |
574 | } | |
575 | return iter + 1; | |
576 | } | |
577 | ||
6b74ab97 MG |
578 | /* Memory initialisation debug and verification */ |
579 | enum mminit_level { | |
580 | MMINIT_WARNING, | |
581 | MMINIT_VERIFY, | |
582 | MMINIT_TRACE | |
583 | }; | |
584 | ||
585 | #ifdef CONFIG_DEBUG_MEMORY_INIT | |
586 | ||
587 | extern int mminit_loglevel; | |
588 | ||
589 | #define mminit_dprintk(level, prefix, fmt, arg...) \ | |
590 | do { \ | |
591 | if (level < mminit_loglevel) { \ | |
fc5199d1 | 592 | if (level <= MMINIT_WARNING) \ |
1170532b | 593 | pr_warn("mminit::" prefix " " fmt, ##arg); \ |
fc5199d1 RV |
594 | else \ |
595 | printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ | |
6b74ab97 MG |
596 | } \ |
597 | } while (0) | |
598 | ||
708614e6 | 599 | extern void mminit_verify_pageflags_layout(void); |
68ad8df4 | 600 | extern void mminit_verify_zonelist(void); |
6b74ab97 MG |
601 | #else |
602 | ||
603 | static inline void mminit_dprintk(enum mminit_level level, | |
604 | const char *prefix, const char *fmt, ...) | |
605 | { | |
606 | } | |
607 | ||
708614e6 MG |
608 | static inline void mminit_verify_pageflags_layout(void) |
609 | { | |
610 | } | |
611 | ||
68ad8df4 MG |
612 | static inline void mminit_verify_zonelist(void) |
613 | { | |
614 | } | |
6b74ab97 | 615 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
2dbb51c4 | 616 | |
a5f5f91d MG |
617 | #define NODE_RECLAIM_NOSCAN -2 |
618 | #define NODE_RECLAIM_FULL -1 | |
619 | #define NODE_RECLAIM_SOME 0 | |
620 | #define NODE_RECLAIM_SUCCESS 1 | |
7c116f2b | 621 | |
8b09549c WY |
622 | #ifdef CONFIG_NUMA |
623 | extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); | |
79c28a41 | 624 | extern int find_next_best_node(int node, nodemask_t *used_node_mask); |
8b09549c WY |
625 | #else |
626 | static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, | |
627 | unsigned int order) | |
628 | { | |
629 | return NODE_RECLAIM_NOSCAN; | |
630 | } | |
79c28a41 DH |
631 | static inline int find_next_best_node(int node, nodemask_t *used_node_mask) |
632 | { | |
633 | return NUMA_NO_NODE; | |
634 | } | |
8b09549c WY |
635 | #endif |
636 | ||
31d3d348 WF |
637 | extern int hwpoison_filter(struct page *p); |
638 | ||
7c116f2b WF |
639 | extern u32 hwpoison_filter_dev_major; |
640 | extern u32 hwpoison_filter_dev_minor; | |
478c5ffc WF |
641 | extern u64 hwpoison_filter_flags_mask; |
642 | extern u64 hwpoison_filter_flags_value; | |
4fd466eb | 643 | extern u64 hwpoison_filter_memcg; |
1bfe5feb | 644 | extern u32 hwpoison_filter_enable; |
eb36c587 | 645 | |
dc0ef0df | 646 | extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, |
eb36c587 | 647 | unsigned long, unsigned long, |
9fbeb5ab | 648 | unsigned long, unsigned long); |
ca57df79 XQ |
649 | |
650 | extern void set_pageblock_order(void); | |
730ec8c0 | 651 | unsigned int reclaim_clean_pages_from_list(struct zone *zone, |
02c6de8d | 652 | struct list_head *page_list); |
d95ea5d1 BZ |
653 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
654 | #define ALLOC_WMARK_MIN WMARK_MIN | |
655 | #define ALLOC_WMARK_LOW WMARK_LOW | |
656 | #define ALLOC_WMARK_HIGH WMARK_HIGH | |
657 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ | |
658 | ||
659 | /* Mask to get the watermark bits */ | |
660 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) | |
661 | ||
cd04ae1e MH |
662 | /* |
663 | * Only MMU archs have async oom victim reclaim - aka oom_reaper so we | |
664 | * cannot assume a reduced access to memory reserves is sufficient for | |
665 | * !MMU | |
666 | */ | |
667 | #ifdef CONFIG_MMU | |
668 | #define ALLOC_OOM 0x08 | |
669 | #else | |
670 | #define ALLOC_OOM ALLOC_NO_WATERMARKS | |
671 | #endif | |
672 | ||
6bb15450 MG |
673 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ |
674 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ | |
675 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ | |
676 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ | |
677 | #ifdef CONFIG_ZONE_DMA32 | |
678 | #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */ | |
679 | #else | |
680 | #define ALLOC_NOFRAGMENT 0x0 | |
681 | #endif | |
736838e9 | 682 | #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ |
d95ea5d1 | 683 | |
72b252ae MG |
684 | enum ttu_flags; |
685 | struct tlbflush_unmap_batch; | |
686 | ||
ce612879 MH |
687 | |
688 | /* | |
689 | * only for MM internal work items which do not depend on | |
690 | * any allocations or locks which might depend on allocations | |
691 | */ | |
692 | extern struct workqueue_struct *mm_percpu_wq; | |
693 | ||
72b252ae MG |
694 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
695 | void try_to_unmap_flush(void); | |
d950c947 | 696 | void try_to_unmap_flush_dirty(void); |
3ea27719 | 697 | void flush_tlb_batched_pending(struct mm_struct *mm); |
72b252ae MG |
698 | #else |
699 | static inline void try_to_unmap_flush(void) | |
700 | { | |
701 | } | |
d950c947 MG |
702 | static inline void try_to_unmap_flush_dirty(void) |
703 | { | |
704 | } | |
3ea27719 MG |
705 | static inline void flush_tlb_batched_pending(struct mm_struct *mm) |
706 | { | |
707 | } | |
72b252ae | 708 | #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ |
edf14cdb VB |
709 | |
710 | extern const struct trace_print_flags pageflag_names[]; | |
711 | extern const struct trace_print_flags vmaflag_names[]; | |
712 | extern const struct trace_print_flags gfpflag_names[]; | |
713 | ||
a6ffdc07 XQ |
714 | static inline bool is_migrate_highatomic(enum migratetype migratetype) |
715 | { | |
716 | return migratetype == MIGRATE_HIGHATOMIC; | |
717 | } | |
718 | ||
719 | static inline bool is_migrate_highatomic_page(struct page *page) | |
720 | { | |
721 | return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC; | |
722 | } | |
723 | ||
72675e13 | 724 | void setup_zone_pageset(struct zone *zone); |
19fc7bed JK |
725 | |
726 | struct migration_target_control { | |
727 | int nid; /* preferred node id */ | |
728 | nodemask_t *nmask; | |
729 | gfp_t gfp_mask; | |
730 | }; | |
731 | ||
b67177ec NP |
732 | /* |
733 | * mm/vmalloc.c | |
734 | */ | |
4ad0ae8c | 735 | #ifdef CONFIG_MMU |
b67177ec NP |
736 | int vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
737 | pgprot_t prot, struct page **pages, unsigned int page_shift); | |
4ad0ae8c NP |
738 | #else |
739 | static inline | |
740 | int vmap_pages_range_noflush(unsigned long addr, unsigned long end, | |
741 | pgprot_t prot, struct page **pages, unsigned int page_shift) | |
742 | { | |
743 | return -EINVAL; | |
744 | } | |
745 | #endif | |
746 | ||
747 | void vunmap_range_noflush(unsigned long start, unsigned long end); | |
b67177ec | 748 | |
f4c0d836 YS |
749 | int numa_migrate_prep(struct page *page, struct vm_area_struct *vma, |
750 | unsigned long addr, int page_nid, int *flags); | |
751 | ||
27674ef6 CH |
752 | void free_zone_device_page(struct page *page); |
753 | ||
ece1ed7b MWO |
754 | /* |
755 | * mm/gup.c | |
756 | */ | |
757 | struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags); | |
758 | ||
09f49dca MH |
759 | DECLARE_PER_CPU(struct per_cpu_nodestat, boot_nodestats); |
760 | ||
db971418 | 761 | #endif /* __MM_INTERNAL_H */ |