1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
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) |
6 | */ |
7 | #ifndef __MM_INTERNAL_H |
8 | #define __MM_INTERNAL_H |
9 | |
10 | #include <linux/fs.h> |
11 | #include <linux/mm.h> |
12 | #include <linux/pagemap.h> |
13 | #include <linux/rmap.h> |
14 | #include <linux/tracepoint-defs.h> |
15 | |
16 | struct folio_batch; |
17 | |
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|\ |
25 | __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\ |
26 | __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\ |
27 | __GFP_NOLOCKDEP) |
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 | |
38 | /* |
39 | * Different from WARN_ON_ONCE(), no warning will be issued |
40 | * when we specify __GFP_NOWARN. |
41 | */ |
42 | #define WARN_ON_ONCE_GFP(cond, gfp) ({ \ |
43 | static bool __section(".data.once") __warned; \ |
44 | int __ret_warn_once = !!(cond); \ |
45 | \ |
46 | if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && !__warned)) { \ |
47 | __warned = true; \ |
48 | WARN_ON(1); \ |
49 | } \ |
50 | unlikely(__ret_warn_once); \ |
51 | }) |
52 | |
53 | void page_writeback_init(void); |
54 | |
55 | /* |
56 | * If a 16GB hugetlb folio were mapped by PTEs of all of its 4kB pages, |
57 | * its nr_pages_mapped would be 0x400000: choose the COMPOUND_MAPPED bit |
58 | * above that range, instead of 2*(PMD_SIZE/PAGE_SIZE). Hugetlb currently |
59 | * leaves nr_pages_mapped at 0, but avoid surprise if it participates later. |
60 | */ |
61 | #define COMPOUND_MAPPED 0x800000 |
62 | #define FOLIO_PAGES_MAPPED (COMPOUND_MAPPED - 1) |
63 | |
64 | /* |
65 | * Flags passed to __show_mem() and show_free_areas() to suppress output in |
66 | * various contexts. |
67 | */ |
68 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ |
69 | |
70 | /* |
71 | * How many individual pages have an elevated _mapcount. Excludes |
72 | * the folio's entire_mapcount. |
73 | */ |
74 | static inline int folio_nr_pages_mapped(struct folio *folio) |
75 | { |
76 | return atomic_read(v: &folio->_nr_pages_mapped) & FOLIO_PAGES_MAPPED; |
77 | } |
78 | |
79 | static inline void *folio_raw_mapping(struct folio *folio) |
80 | { |
81 | unsigned long mapping = (unsigned long)folio->mapping; |
82 | |
83 | return (void *)(mapping & ~PAGE_MAPPING_FLAGS); |
84 | } |
85 | |
86 | void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, |
87 | int nr_throttled); |
88 | static inline void acct_reclaim_writeback(struct folio *folio) |
89 | { |
90 | pg_data_t *pgdat = folio_pgdat(folio); |
91 | int nr_throttled = atomic_read(v: &pgdat->nr_writeback_throttled); |
92 | |
93 | if (nr_throttled) |
94 | __acct_reclaim_writeback(pgdat, folio, nr_throttled); |
95 | } |
96 | |
97 | static inline void wake_throttle_isolated(pg_data_t *pgdat) |
98 | { |
99 | wait_queue_head_t *wqh; |
100 | |
101 | wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED]; |
102 | if (waitqueue_active(wq_head: wqh)) |
103 | wake_up(wqh); |
104 | } |
105 | |
106 | vm_fault_t do_swap_page(struct vm_fault *vmf); |
107 | void folio_rotate_reclaimable(struct folio *folio); |
108 | bool __folio_end_writeback(struct folio *folio); |
109 | void deactivate_file_folio(struct folio *folio); |
110 | void folio_activate(struct folio *folio); |
111 | |
112 | void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas, |
113 | struct vm_area_struct *start_vma, unsigned long floor, |
114 | unsigned long ceiling, bool mm_wr_locked); |
115 | void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); |
116 | |
117 | struct zap_details; |
118 | void unmap_page_range(struct mmu_gather *tlb, |
119 | struct vm_area_struct *vma, |
120 | unsigned long addr, unsigned long end, |
121 | struct zap_details *details); |
122 | |
123 | void page_cache_ra_order(struct readahead_control *, struct file_ra_state *, |
124 | unsigned int order); |
125 | void force_page_cache_ra(struct readahead_control *, unsigned long nr); |
126 | static inline void force_page_cache_readahead(struct address_space *mapping, |
127 | struct file *file, pgoff_t index, unsigned long nr_to_read) |
128 | { |
129 | DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index); |
130 | force_page_cache_ra(&ractl, nr: nr_to_read); |
131 | } |
132 | |
133 | unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start, |
134 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
135 | unsigned find_get_entries(struct address_space *mapping, pgoff_t *start, |
136 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
137 | void filemap_free_folio(struct address_space *mapping, struct folio *folio); |
138 | int truncate_inode_folio(struct address_space *mapping, struct folio *folio); |
139 | bool truncate_inode_partial_folio(struct folio *folio, loff_t start, |
140 | loff_t end); |
141 | long invalidate_inode_page(struct page *page); |
142 | unsigned long mapping_try_invalidate(struct address_space *mapping, |
143 | pgoff_t start, pgoff_t end, unsigned long *nr_failed); |
144 | |
145 | /** |
146 | * folio_evictable - Test whether a folio is evictable. |
147 | * @folio: The folio to test. |
148 | * |
149 | * Test whether @folio is evictable -- i.e., should be placed on |
150 | * active/inactive lists vs unevictable list. |
151 | * |
152 | * Reasons folio might not be evictable: |
153 | * 1. folio's mapping marked unevictable |
154 | * 2. One of the pages in the folio is part of an mlocked VMA |
155 | */ |
156 | static inline bool folio_evictable(struct folio *folio) |
157 | { |
158 | bool ret; |
159 | |
160 | /* Prevent address_space of inode and swap cache from being freed */ |
161 | rcu_read_lock(); |
162 | ret = !mapping_unevictable(mapping: folio_mapping(folio)) && |
163 | !folio_test_mlocked(folio); |
164 | rcu_read_unlock(); |
165 | return ret; |
166 | } |
167 | |
168 | /* |
169 | * Turn a non-refcounted page (->_refcount == 0) into refcounted with |
170 | * a count of one. |
171 | */ |
172 | static inline void set_page_refcounted(struct page *page) |
173 | { |
174 | VM_BUG_ON_PAGE(PageTail(page), page); |
175 | VM_BUG_ON_PAGE(page_ref_count(page), page); |
176 | set_page_count(page, v: 1); |
177 | } |
178 | |
179 | /* |
180 | * Return true if a folio needs ->release_folio() calling upon it. |
181 | */ |
182 | static inline bool folio_needs_release(struct folio *folio) |
183 | { |
184 | struct address_space *mapping = folio_mapping(folio); |
185 | |
186 | return folio_has_private(folio) || |
187 | (mapping && mapping_release_always(mapping)); |
188 | } |
189 | |
190 | extern unsigned long highest_memmap_pfn; |
191 | |
192 | /* |
193 | * Maximum number of reclaim retries without progress before the OOM |
194 | * killer is consider the only way forward. |
195 | */ |
196 | #define MAX_RECLAIM_RETRIES 16 |
197 | |
198 | /* |
199 | * in mm/vmscan.c: |
200 | */ |
201 | bool isolate_lru_page(struct page *page); |
202 | bool folio_isolate_lru(struct folio *folio); |
203 | void putback_lru_page(struct page *page); |
204 | void folio_putback_lru(struct folio *folio); |
205 | extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); |
206 | |
207 | /* |
208 | * in mm/rmap.c: |
209 | */ |
210 | pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); |
211 | |
212 | /* |
213 | * in mm/page_alloc.c |
214 | */ |
215 | #define K(x) ((x) << (PAGE_SHIFT-10)) |
216 | |
217 | extern char * const zone_names[MAX_NR_ZONES]; |
218 | |
219 | /* perform sanity checks on struct pages being allocated or freed */ |
220 | DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled); |
221 | |
222 | extern int min_free_kbytes; |
223 | |
224 | void setup_per_zone_wmarks(void); |
225 | void calculate_min_free_kbytes(void); |
226 | int __meminit init_per_zone_wmark_min(void); |
227 | void page_alloc_sysctl_init(void); |
228 | |
229 | /* |
230 | * Structure for holding the mostly immutable allocation parameters passed |
231 | * between functions involved in allocations, including the alloc_pages* |
232 | * family of functions. |
233 | * |
234 | * nodemask, migratetype and highest_zoneidx are initialized only once in |
235 | * __alloc_pages() and then never change. |
236 | * |
237 | * zonelist, preferred_zone and highest_zoneidx are set first in |
238 | * __alloc_pages() for the fast path, and might be later changed |
239 | * in __alloc_pages_slowpath(). All other functions pass the whole structure |
240 | * by a const pointer. |
241 | */ |
242 | struct alloc_context { |
243 | struct zonelist *zonelist; |
244 | nodemask_t *nodemask; |
245 | struct zoneref *preferred_zoneref; |
246 | int migratetype; |
247 | |
248 | /* |
249 | * highest_zoneidx represents highest usable zone index of |
250 | * the allocation request. Due to the nature of the zone, |
251 | * memory on lower zone than the highest_zoneidx will be |
252 | * protected by lowmem_reserve[highest_zoneidx]. |
253 | * |
254 | * highest_zoneidx is also used by reclaim/compaction to limit |
255 | * the target zone since higher zone than this index cannot be |
256 | * usable for this allocation request. |
257 | */ |
258 | enum zone_type highest_zoneidx; |
259 | bool spread_dirty_pages; |
260 | }; |
261 | |
262 | /* |
263 | * This function returns the order of a free page in the buddy system. In |
264 | * general, page_zone(page)->lock must be held by the caller to prevent the |
265 | * page from being allocated in parallel and returning garbage as the order. |
266 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the |
267 | * page cannot be allocated or merged in parallel. Alternatively, it must |
268 | * handle invalid values gracefully, and use buddy_order_unsafe() below. |
269 | */ |
270 | static inline unsigned int buddy_order(struct page *page) |
271 | { |
272 | /* PageBuddy() must be checked by the caller */ |
273 | return page_private(page); |
274 | } |
275 | |
276 | /* |
277 | * Like buddy_order(), but for callers who cannot afford to hold the zone lock. |
278 | * PageBuddy() should be checked first by the caller to minimize race window, |
279 | * and invalid values must be handled gracefully. |
280 | * |
281 | * READ_ONCE is used so that if the caller assigns the result into a local |
282 | * variable and e.g. tests it for valid range before using, the compiler cannot |
283 | * decide to remove the variable and inline the page_private(page) multiple |
284 | * times, potentially observing different values in the tests and the actual |
285 | * use of the result. |
286 | */ |
287 | #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) |
288 | |
289 | /* |
290 | * This function checks whether a page is free && is the buddy |
291 | * we can coalesce a page and its buddy if |
292 | * (a) the buddy is not in a hole (check before calling!) && |
293 | * (b) the buddy is in the buddy system && |
294 | * (c) a page and its buddy have the same order && |
295 | * (d) a page and its buddy are in the same zone. |
296 | * |
297 | * For recording whether a page is in the buddy system, we set PageBuddy. |
298 | * Setting, clearing, and testing PageBuddy is serialized by zone->lock. |
299 | * |
300 | * For recording page's order, we use page_private(page). |
301 | */ |
302 | static inline bool page_is_buddy(struct page *page, struct page *buddy, |
303 | unsigned int order) |
304 | { |
305 | if (!page_is_guard(page: buddy) && !PageBuddy(page: buddy)) |
306 | return false; |
307 | |
308 | if (buddy_order(page: buddy) != order) |
309 | return false; |
310 | |
311 | /* |
312 | * zone check is done late to avoid uselessly calculating |
313 | * zone/node ids for pages that could never merge. |
314 | */ |
315 | if (page_zone_id(page) != page_zone_id(page: buddy)) |
316 | return false; |
317 | |
318 | VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy); |
319 | |
320 | return true; |
321 | } |
322 | |
323 | /* |
324 | * Locate the struct page for both the matching buddy in our |
325 | * pair (buddy1) and the combined O(n+1) page they form (page). |
326 | * |
327 | * 1) Any buddy B1 will have an order O twin B2 which satisfies |
328 | * the following equation: |
329 | * B2 = B1 ^ (1 << O) |
330 | * For example, if the starting buddy (buddy2) is #8 its order |
331 | * 1 buddy is #10: |
332 | * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 |
333 | * |
334 | * 2) Any buddy B will have an order O+1 parent P which |
335 | * satisfies the following equation: |
336 | * P = B & ~(1 << O) |
337 | * |
338 | * Assumption: *_mem_map is contiguous at least up to MAX_ORDER |
339 | */ |
340 | static inline unsigned long |
341 | __find_buddy_pfn(unsigned long page_pfn, unsigned int order) |
342 | { |
343 | return page_pfn ^ (1 << order); |
344 | } |
345 | |
346 | /* |
347 | * Find the buddy of @page and validate it. |
348 | * @page: The input page |
349 | * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the |
350 | * function is used in the performance-critical __free_one_page(). |
351 | * @order: The order of the page |
352 | * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to |
353 | * page_to_pfn(). |
354 | * |
355 | * The found buddy can be a non PageBuddy, out of @page's zone, or its order is |
356 | * not the same as @page. The validation is necessary before use it. |
357 | * |
358 | * Return: the found buddy page or NULL if not found. |
359 | */ |
360 | static inline struct page *find_buddy_page_pfn(struct page *page, |
361 | unsigned long pfn, unsigned int order, unsigned long *buddy_pfn) |
362 | { |
363 | unsigned long __buddy_pfn = __find_buddy_pfn(page_pfn: pfn, order); |
364 | struct page *buddy; |
365 | |
366 | buddy = page + (__buddy_pfn - pfn); |
367 | if (buddy_pfn) |
368 | *buddy_pfn = __buddy_pfn; |
369 | |
370 | if (page_is_buddy(page, buddy, order)) |
371 | return buddy; |
372 | return NULL; |
373 | } |
374 | |
375 | extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, |
376 | unsigned long end_pfn, struct zone *zone); |
377 | |
378 | static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, |
379 | unsigned long end_pfn, struct zone *zone) |
380 | { |
381 | if (zone->contiguous) |
382 | return pfn_to_page(start_pfn); |
383 | |
384 | return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); |
385 | } |
386 | |
387 | void set_zone_contiguous(struct zone *zone); |
388 | |
389 | static inline void clear_zone_contiguous(struct zone *zone) |
390 | { |
391 | zone->contiguous = false; |
392 | } |
393 | |
394 | extern int __isolate_free_page(struct page *page, unsigned int order); |
395 | extern void __putback_isolated_page(struct page *page, unsigned int order, |
396 | int mt); |
397 | extern void memblock_free_pages(struct page *page, unsigned long pfn, |
398 | unsigned int order); |
399 | extern void __free_pages_core(struct page *page, unsigned int order); |
400 | |
401 | /* |
402 | * This will have no effect, other than possibly generating a warning, if the |
403 | * caller passes in a non-large folio. |
404 | */ |
405 | static inline void folio_set_order(struct folio *folio, unsigned int order) |
406 | { |
407 | if (WARN_ON_ONCE(!order || !folio_test_large(folio))) |
408 | return; |
409 | |
410 | folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order; |
411 | #ifdef CONFIG_64BIT |
412 | folio->_folio_nr_pages = 1U << order; |
413 | #endif |
414 | } |
415 | |
416 | void folio_undo_large_rmappable(struct folio *folio); |
417 | |
418 | static inline struct folio *page_rmappable_folio(struct page *page) |
419 | { |
420 | struct folio *folio = (struct folio *)page; |
421 | |
422 | if (folio && folio_order(folio) > 1) |
423 | folio_prep_large_rmappable(folio); |
424 | return folio; |
425 | } |
426 | |
427 | static inline void prep_compound_head(struct page *page, unsigned int order) |
428 | { |
429 | struct folio *folio = (struct folio *)page; |
430 | |
431 | folio_set_order(folio, order); |
432 | atomic_set(v: &folio->_entire_mapcount, i: -1); |
433 | atomic_set(v: &folio->_nr_pages_mapped, i: 0); |
434 | atomic_set(v: &folio->_pincount, i: 0); |
435 | } |
436 | |
437 | static inline void prep_compound_tail(struct page *head, int tail_idx) |
438 | { |
439 | struct page *p = head + tail_idx; |
440 | |
441 | p->mapping = TAIL_MAPPING; |
442 | set_compound_head(page: p, head); |
443 | set_page_private(page: p, private: 0); |
444 | } |
445 | |
446 | extern void prep_compound_page(struct page *page, unsigned int order); |
447 | |
448 | extern void post_alloc_hook(struct page *page, unsigned int order, |
449 | gfp_t gfp_flags); |
450 | extern int user_min_free_kbytes; |
451 | |
452 | extern void free_unref_page(struct page *page, unsigned int order); |
453 | extern void free_unref_page_list(struct list_head *list); |
454 | |
455 | extern void zone_pcp_reset(struct zone *zone); |
456 | extern void zone_pcp_disable(struct zone *zone); |
457 | extern void zone_pcp_enable(struct zone *zone); |
458 | extern void zone_pcp_init(struct zone *zone); |
459 | |
460 | extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, |
461 | phys_addr_t min_addr, |
462 | int nid, bool exact_nid); |
463 | |
464 | void memmap_init_range(unsigned long, int, unsigned long, unsigned long, |
465 | unsigned long, enum meminit_context, struct vmem_altmap *, int); |
466 | |
467 | |
468 | int split_free_page(struct page *free_page, |
469 | unsigned int order, unsigned long split_pfn_offset); |
470 | |
471 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
472 | |
473 | /* |
474 | * in mm/compaction.c |
475 | */ |
476 | /* |
477 | * compact_control is used to track pages being migrated and the free pages |
478 | * they are being migrated to during memory compaction. The free_pfn starts |
479 | * at the end of a zone and migrate_pfn begins at the start. Movable pages |
480 | * are moved to the end of a zone during a compaction run and the run |
481 | * completes when free_pfn <= migrate_pfn |
482 | */ |
483 | struct compact_control { |
484 | struct list_head freepages; /* List of free pages to migrate to */ |
485 | struct list_head migratepages; /* List of pages being migrated */ |
486 | unsigned int nr_freepages; /* Number of isolated free pages */ |
487 | unsigned int nr_migratepages; /* Number of pages to migrate */ |
488 | unsigned long free_pfn; /* isolate_freepages search base */ |
489 | /* |
490 | * Acts as an in/out parameter to page isolation for migration. |
491 | * isolate_migratepages uses it as a search base. |
492 | * isolate_migratepages_block will update the value to the next pfn |
493 | * after the last isolated one. |
494 | */ |
495 | unsigned long migrate_pfn; |
496 | unsigned long fast_start_pfn; /* a pfn to start linear scan from */ |
497 | struct zone *zone; |
498 | unsigned long total_migrate_scanned; |
499 | unsigned long total_free_scanned; |
500 | unsigned short fast_search_fail;/* failures to use free list searches */ |
501 | short search_order; /* order to start a fast search at */ |
502 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
503 | int order; /* order a direct compactor needs */ |
504 | int migratetype; /* migratetype of direct compactor */ |
505 | const unsigned int alloc_flags; /* alloc flags of a direct compactor */ |
506 | const int highest_zoneidx; /* zone index of a direct compactor */ |
507 | enum migrate_mode mode; /* Async or sync migration mode */ |
508 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
509 | bool no_set_skip_hint; /* Don't mark blocks for skipping */ |
510 | bool ignore_block_suitable; /* Scan blocks considered unsuitable */ |
511 | bool direct_compaction; /* False from kcompactd or /proc/... */ |
512 | bool proactive_compaction; /* kcompactd proactive compaction */ |
513 | bool whole_zone; /* Whole zone should/has been scanned */ |
514 | bool contended; /* Signal lock contention */ |
515 | bool finish_pageblock; /* Scan the remainder of a pageblock. Used |
516 | * when there are potentially transient |
517 | * isolation or migration failures to |
518 | * ensure forward progress. |
519 | */ |
520 | bool alloc_contig; /* alloc_contig_range allocation */ |
521 | }; |
522 | |
523 | /* |
524 | * Used in direct compaction when a page should be taken from the freelists |
525 | * immediately when one is created during the free path. |
526 | */ |
527 | struct capture_control { |
528 | struct compact_control *cc; |
529 | struct page *page; |
530 | }; |
531 | |
532 | unsigned long |
533 | isolate_freepages_range(struct compact_control *cc, |
534 | unsigned long start_pfn, unsigned long end_pfn); |
535 | int |
536 | isolate_migratepages_range(struct compact_control *cc, |
537 | unsigned long low_pfn, unsigned long end_pfn); |
538 | |
539 | int __alloc_contig_migrate_range(struct compact_control *cc, |
540 | unsigned long start, unsigned long end); |
541 | |
542 | /* Free whole pageblock and set its migration type to MIGRATE_CMA. */ |
543 | void init_cma_reserved_pageblock(struct page *page); |
544 | |
545 | #endif /* CONFIG_COMPACTION || CONFIG_CMA */ |
546 | |
547 | int find_suitable_fallback(struct free_area *area, unsigned int order, |
548 | int migratetype, bool only_stealable, bool *can_steal); |
549 | |
550 | static inline bool free_area_empty(struct free_area *area, int migratetype) |
551 | { |
552 | return list_empty(head: &area->free_list[migratetype]); |
553 | } |
554 | |
555 | /* |
556 | * These three helpers classifies VMAs for virtual memory accounting. |
557 | */ |
558 | |
559 | /* |
560 | * Executable code area - executable, not writable, not stack |
561 | */ |
562 | static inline bool is_exec_mapping(vm_flags_t flags) |
563 | { |
564 | return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; |
565 | } |
566 | |
567 | /* |
568 | * Stack area (including shadow stacks) |
569 | * |
570 | * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: |
571 | * do_mmap() forbids all other combinations. |
572 | */ |
573 | static inline bool is_stack_mapping(vm_flags_t flags) |
574 | { |
575 | return ((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK); |
576 | } |
577 | |
578 | /* |
579 | * Data area - private, writable, not stack |
580 | */ |
581 | static inline bool is_data_mapping(vm_flags_t flags) |
582 | { |
583 | return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; |
584 | } |
585 | |
586 | /* mm/util.c */ |
587 | struct anon_vma *folio_anon_vma(struct folio *folio); |
588 | |
589 | #ifdef CONFIG_MMU |
590 | void unmap_mapping_folio(struct folio *folio); |
591 | extern long populate_vma_page_range(struct vm_area_struct *vma, |
592 | unsigned long start, unsigned long end, int *locked); |
593 | extern long faultin_vma_page_range(struct vm_area_struct *vma, |
594 | unsigned long start, unsigned long end, |
595 | bool write, int *locked); |
596 | extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags, |
597 | unsigned long bytes); |
598 | |
599 | /* |
600 | * NOTE: This function can't tell whether the folio is "fully mapped" in the |
601 | * range. |
602 | * "fully mapped" means all the pages of folio is associated with the page |
603 | * table of range while this function just check whether the folio range is |
604 | * within the range [start, end). Function caller needs to do page table |
605 | * check if it cares about the page table association. |
606 | * |
607 | * Typical usage (like mlock or madvise) is: |
608 | * Caller knows at least 1 page of folio is associated with page table of VMA |
609 | * and the range [start, end) is intersect with the VMA range. Caller wants |
610 | * to know whether the folio is fully associated with the range. It calls |
611 | * this function to check whether the folio is in the range first. Then checks |
612 | * the page table to know whether the folio is fully mapped to the range. |
613 | */ |
614 | static inline bool |
615 | folio_within_range(struct folio *folio, struct vm_area_struct *vma, |
616 | unsigned long start, unsigned long end) |
617 | { |
618 | pgoff_t pgoff, addr; |
619 | unsigned long vma_pglen = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; |
620 | |
621 | VM_WARN_ON_FOLIO(folio_test_ksm(folio), folio); |
622 | if (start > end) |
623 | return false; |
624 | |
625 | if (start < vma->vm_start) |
626 | start = vma->vm_start; |
627 | |
628 | if (end > vma->vm_end) |
629 | end = vma->vm_end; |
630 | |
631 | pgoff = folio_pgoff(folio); |
632 | |
633 | /* if folio start address is not in vma range */ |
634 | if (!in_range(pgoff, vma->vm_pgoff, vma_pglen)) |
635 | return false; |
636 | |
637 | addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); |
638 | |
639 | return !(addr < start || end - addr < folio_size(folio)); |
640 | } |
641 | |
642 | static inline bool |
643 | folio_within_vma(struct folio *folio, struct vm_area_struct *vma) |
644 | { |
645 | return folio_within_range(folio, vma, start: vma->vm_start, end: vma->vm_end); |
646 | } |
647 | |
648 | /* |
649 | * mlock_vma_folio() and munlock_vma_folio(): |
650 | * should be called with vma's mmap_lock held for read or write, |
651 | * under page table lock for the pte/pmd being added or removed. |
652 | * |
653 | * mlock is usually called at the end of page_add_*_rmap(), munlock at |
654 | * the end of page_remove_rmap(); but new anon folios are managed by |
655 | * folio_add_lru_vma() calling mlock_new_folio(). |
656 | */ |
657 | void mlock_folio(struct folio *folio); |
658 | static inline void mlock_vma_folio(struct folio *folio, |
659 | struct vm_area_struct *vma) |
660 | { |
661 | /* |
662 | * The VM_SPECIAL check here serves two purposes. |
663 | * 1) VM_IO check prevents migration from double-counting during mlock. |
664 | * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED |
665 | * is never left set on a VM_SPECIAL vma, there is an interval while |
666 | * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may |
667 | * still be set while VM_SPECIAL bits are added: so ignore it then. |
668 | */ |
669 | if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED)) |
670 | mlock_folio(folio); |
671 | } |
672 | |
673 | void munlock_folio(struct folio *folio); |
674 | static inline void munlock_vma_folio(struct folio *folio, |
675 | struct vm_area_struct *vma) |
676 | { |
677 | /* |
678 | * munlock if the function is called. Ideally, we should only |
679 | * do munlock if any page of folio is unmapped from VMA and |
680 | * cause folio not fully mapped to VMA. |
681 | * |
682 | * But it's not easy to confirm that's the situation. So we |
683 | * always munlock the folio and page reclaim will correct it |
684 | * if it's wrong. |
685 | */ |
686 | if (unlikely(vma->vm_flags & VM_LOCKED)) |
687 | munlock_folio(folio); |
688 | } |
689 | |
690 | void mlock_new_folio(struct folio *folio); |
691 | bool need_mlock_drain(int cpu); |
692 | void mlock_drain_local(void); |
693 | void mlock_drain_remote(int cpu); |
694 | |
695 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
696 | |
697 | /* |
698 | * Return the start of user virtual address at the specific offset within |
699 | * a vma. |
700 | */ |
701 | static inline unsigned long |
702 | vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages, |
703 | struct vm_area_struct *vma) |
704 | { |
705 | unsigned long address; |
706 | |
707 | if (pgoff >= vma->vm_pgoff) { |
708 | address = vma->vm_start + |
709 | ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); |
710 | /* Check for address beyond vma (or wrapped through 0?) */ |
711 | if (address < vma->vm_start || address >= vma->vm_end) |
712 | address = -EFAULT; |
713 | } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) { |
714 | /* Test above avoids possibility of wrap to 0 on 32-bit */ |
715 | address = vma->vm_start; |
716 | } else { |
717 | address = -EFAULT; |
718 | } |
719 | return address; |
720 | } |
721 | |
722 | /* |
723 | * Return the start of user virtual address of a page within a vma. |
724 | * Returns -EFAULT if all of the page is outside the range of vma. |
725 | * If page is a compound head, the entire compound page is considered. |
726 | */ |
727 | static inline unsigned long |
728 | vma_address(struct page *page, struct vm_area_struct *vma) |
729 | { |
730 | VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ |
731 | return vma_pgoff_address(pgoff: page_to_pgoff(page), nr_pages: compound_nr(page), vma); |
732 | } |
733 | |
734 | /* |
735 | * Then at what user virtual address will none of the range be found in vma? |
736 | * Assumes that vma_address() already returned a good starting address. |
737 | */ |
738 | static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw) |
739 | { |
740 | struct vm_area_struct *vma = pvmw->vma; |
741 | pgoff_t pgoff; |
742 | unsigned long address; |
743 | |
744 | /* Common case, plus ->pgoff is invalid for KSM */ |
745 | if (pvmw->nr_pages == 1) |
746 | return pvmw->address + PAGE_SIZE; |
747 | |
748 | pgoff = pvmw->pgoff + pvmw->nr_pages; |
749 | address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); |
750 | /* Check for address beyond vma (or wrapped through 0?) */ |
751 | if (address < vma->vm_start || address > vma->vm_end) |
752 | address = vma->vm_end; |
753 | return address; |
754 | } |
755 | |
756 | static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf, |
757 | struct file *fpin) |
758 | { |
759 | int flags = vmf->flags; |
760 | |
761 | if (fpin) |
762 | return fpin; |
763 | |
764 | /* |
765 | * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or |
766 | * anything, so we only pin the file and drop the mmap_lock if only |
767 | * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. |
768 | */ |
769 | if (fault_flag_allow_retry_first(flags) && |
770 | !(flags & FAULT_FLAG_RETRY_NOWAIT)) { |
771 | fpin = get_file(f: vmf->vma->vm_file); |
772 | release_fault_lock(vmf); |
773 | } |
774 | return fpin; |
775 | } |
776 | #else /* !CONFIG_MMU */ |
777 | static inline void unmap_mapping_folio(struct folio *folio) { } |
778 | static inline void mlock_new_folio(struct folio *folio) { } |
779 | static inline bool need_mlock_drain(int cpu) { return false; } |
780 | static inline void mlock_drain_local(void) { } |
781 | static inline void mlock_drain_remote(int cpu) { } |
782 | static inline void vunmap_range_noflush(unsigned long start, unsigned long end) |
783 | { |
784 | } |
785 | #endif /* !CONFIG_MMU */ |
786 | |
787 | /* Memory initialisation debug and verification */ |
788 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
789 | DECLARE_STATIC_KEY_TRUE(deferred_pages); |
790 | |
791 | bool __init deferred_grow_zone(struct zone *zone, unsigned int order); |
792 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ |
793 | |
794 | enum mminit_level { |
795 | MMINIT_WARNING, |
796 | MMINIT_VERIFY, |
797 | MMINIT_TRACE |
798 | }; |
799 | |
800 | #ifdef CONFIG_DEBUG_MEMORY_INIT |
801 | |
802 | extern int mminit_loglevel; |
803 | |
804 | #define mminit_dprintk(level, prefix, fmt, arg...) \ |
805 | do { \ |
806 | if (level < mminit_loglevel) { \ |
807 | if (level <= MMINIT_WARNING) \ |
808 | pr_warn("mminit::" prefix " " fmt, ##arg); \ |
809 | else \ |
810 | printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ |
811 | } \ |
812 | } while (0) |
813 | |
814 | extern void mminit_verify_pageflags_layout(void); |
815 | extern void mminit_verify_zonelist(void); |
816 | #else |
817 | |
818 | static inline void mminit_dprintk(enum mminit_level level, |
819 | const char *prefix, const char *fmt, ...) |
820 | { |
821 | } |
822 | |
823 | static inline void mminit_verify_pageflags_layout(void) |
824 | { |
825 | } |
826 | |
827 | static inline void mminit_verify_zonelist(void) |
828 | { |
829 | } |
830 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
831 | |
832 | #define NODE_RECLAIM_NOSCAN -2 |
833 | #define NODE_RECLAIM_FULL -1 |
834 | #define NODE_RECLAIM_SOME 0 |
835 | #define NODE_RECLAIM_SUCCESS 1 |
836 | |
837 | #ifdef CONFIG_NUMA |
838 | extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); |
839 | extern int find_next_best_node(int node, nodemask_t *used_node_mask); |
840 | #else |
841 | static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, |
842 | unsigned int order) |
843 | { |
844 | return NODE_RECLAIM_NOSCAN; |
845 | } |
846 | static inline int find_next_best_node(int node, nodemask_t *used_node_mask) |
847 | { |
848 | return NUMA_NO_NODE; |
849 | } |
850 | #endif |
851 | |
852 | /* |
853 | * mm/memory-failure.c |
854 | */ |
855 | extern int hwpoison_filter(struct page *p); |
856 | |
857 | extern u32 hwpoison_filter_dev_major; |
858 | extern u32 hwpoison_filter_dev_minor; |
859 | extern u64 hwpoison_filter_flags_mask; |
860 | extern u64 hwpoison_filter_flags_value; |
861 | extern u64 hwpoison_filter_memcg; |
862 | extern u32 hwpoison_filter_enable; |
863 | |
864 | extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, |
865 | unsigned long, unsigned long, |
866 | unsigned long, unsigned long); |
867 | |
868 | extern void set_pageblock_order(void); |
869 | unsigned long reclaim_pages(struct list_head *folio_list); |
870 | unsigned int reclaim_clean_pages_from_list(struct zone *zone, |
871 | struct list_head *folio_list); |
872 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
873 | #define ALLOC_WMARK_MIN WMARK_MIN |
874 | #define ALLOC_WMARK_LOW WMARK_LOW |
875 | #define ALLOC_WMARK_HIGH WMARK_HIGH |
876 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ |
877 | |
878 | /* Mask to get the watermark bits */ |
879 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) |
880 | |
881 | /* |
882 | * Only MMU archs have async oom victim reclaim - aka oom_reaper so we |
883 | * cannot assume a reduced access to memory reserves is sufficient for |
884 | * !MMU |
885 | */ |
886 | #ifdef CONFIG_MMU |
887 | #define ALLOC_OOM 0x08 |
888 | #else |
889 | #define ALLOC_OOM ALLOC_NO_WATERMARKS |
890 | #endif |
891 | |
892 | #define ALLOC_NON_BLOCK 0x10 /* Caller cannot block. Allow access |
893 | * to 25% of the min watermark or |
894 | * 62.5% if __GFP_HIGH is set. |
895 | */ |
896 | #define ALLOC_MIN_RESERVE 0x20 /* __GFP_HIGH set. Allow access to 50% |
897 | * of the min watermark. |
898 | */ |
899 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ |
900 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ |
901 | #ifdef CONFIG_ZONE_DMA32 |
902 | #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */ |
903 | #else |
904 | #define ALLOC_NOFRAGMENT 0x0 |
905 | #endif |
906 | #define ALLOC_HIGHATOMIC 0x200 /* Allows access to MIGRATE_HIGHATOMIC */ |
907 | #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ |
908 | |
909 | /* Flags that allow allocations below the min watermark. */ |
910 | #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM) |
911 | |
912 | enum ttu_flags; |
913 | struct tlbflush_unmap_batch; |
914 | |
915 | |
916 | /* |
917 | * only for MM internal work items which do not depend on |
918 | * any allocations or locks which might depend on allocations |
919 | */ |
920 | extern struct workqueue_struct *mm_percpu_wq; |
921 | |
922 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
923 | void try_to_unmap_flush(void); |
924 | void try_to_unmap_flush_dirty(void); |
925 | void flush_tlb_batched_pending(struct mm_struct *mm); |
926 | #else |
927 | static inline void try_to_unmap_flush(void) |
928 | { |
929 | } |
930 | static inline void try_to_unmap_flush_dirty(void) |
931 | { |
932 | } |
933 | static inline void flush_tlb_batched_pending(struct mm_struct *mm) |
934 | { |
935 | } |
936 | #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ |
937 | |
938 | extern const struct trace_print_flags pageflag_names[]; |
939 | extern const struct trace_print_flags pagetype_names[]; |
940 | extern const struct trace_print_flags vmaflag_names[]; |
941 | extern const struct trace_print_flags gfpflag_names[]; |
942 | |
943 | static inline bool is_migrate_highatomic(enum migratetype migratetype) |
944 | { |
945 | return migratetype == MIGRATE_HIGHATOMIC; |
946 | } |
947 | |
948 | static inline bool is_migrate_highatomic_page(struct page *page) |
949 | { |
950 | return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC; |
951 | } |
952 | |
953 | void setup_zone_pageset(struct zone *zone); |
954 | |
955 | struct migration_target_control { |
956 | int nid; /* preferred node id */ |
957 | nodemask_t *nmask; |
958 | gfp_t gfp_mask; |
959 | }; |
960 | |
961 | /* |
962 | * mm/filemap.c |
963 | */ |
964 | size_t splice_folio_into_pipe(struct pipe_inode_info *pipe, |
965 | struct folio *folio, loff_t fpos, size_t size); |
966 | |
967 | /* |
968 | * mm/vmalloc.c |
969 | */ |
970 | #ifdef CONFIG_MMU |
971 | void __init vmalloc_init(void); |
972 | int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
973 | pgprot_t prot, struct page **pages, unsigned int page_shift); |
974 | #else |
975 | static inline void vmalloc_init(void) |
976 | { |
977 | } |
978 | |
979 | static inline |
980 | int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
981 | pgprot_t prot, struct page **pages, unsigned int page_shift) |
982 | { |
983 | return -EINVAL; |
984 | } |
985 | #endif |
986 | |
987 | int __must_check __vmap_pages_range_noflush(unsigned long addr, |
988 | unsigned long end, pgprot_t prot, |
989 | struct page **pages, unsigned int page_shift); |
990 | |
991 | void vunmap_range_noflush(unsigned long start, unsigned long end); |
992 | |
993 | void __vunmap_range_noflush(unsigned long start, unsigned long end); |
994 | |
995 | int numa_migrate_prep(struct folio *folio, struct vm_area_struct *vma, |
996 | unsigned long addr, int page_nid, int *flags); |
997 | |
998 | void free_zone_device_page(struct page *page); |
999 | int migrate_device_coherent_page(struct page *page); |
1000 | |
1001 | /* |
1002 | * mm/gup.c |
1003 | */ |
1004 | struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags); |
1005 | int __must_check try_grab_page(struct page *page, unsigned int flags); |
1006 | |
1007 | /* |
1008 | * mm/huge_memory.c |
1009 | */ |
1010 | struct page *follow_trans_huge_pmd(struct vm_area_struct *vma, |
1011 | unsigned long addr, pmd_t *pmd, |
1012 | unsigned int flags); |
1013 | |
1014 | /* |
1015 | * mm/mmap.c |
1016 | */ |
1017 | struct vm_area_struct *vma_merge_extend(struct vma_iterator *vmi, |
1018 | struct vm_area_struct *vma, |
1019 | unsigned long delta); |
1020 | |
1021 | enum { |
1022 | /* mark page accessed */ |
1023 | FOLL_TOUCH = 1 << 16, |
1024 | /* a retry, previous pass started an IO */ |
1025 | FOLL_TRIED = 1 << 17, |
1026 | /* we are working on non-current tsk/mm */ |
1027 | FOLL_REMOTE = 1 << 18, |
1028 | /* pages must be released via unpin_user_page */ |
1029 | FOLL_PIN = 1 << 19, |
1030 | /* gup_fast: prevent fall-back to slow gup */ |
1031 | FOLL_FAST_ONLY = 1 << 20, |
1032 | /* allow unlocking the mmap lock */ |
1033 | FOLL_UNLOCKABLE = 1 << 21, |
1034 | }; |
1035 | |
1036 | #define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \ |
1037 | FOLL_FAST_ONLY | FOLL_UNLOCKABLE) |
1038 | |
1039 | /* |
1040 | * Indicates for which pages that are write-protected in the page table, |
1041 | * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the |
1042 | * GUP pin will remain consistent with the pages mapped into the page tables |
1043 | * of the MM. |
1044 | * |
1045 | * Temporary unmapping of PageAnonExclusive() pages or clearing of |
1046 | * PageAnonExclusive() has to protect against concurrent GUP: |
1047 | * * Ordinary GUP: Using the PT lock |
1048 | * * GUP-fast and fork(): mm->write_protect_seq |
1049 | * * GUP-fast and KSM or temporary unmapping (swap, migration): see |
1050 | * page_try_share_anon_rmap() |
1051 | * |
1052 | * Must be called with the (sub)page that's actually referenced via the |
1053 | * page table entry, which might not necessarily be the head page for a |
1054 | * PTE-mapped THP. |
1055 | * |
1056 | * If the vma is NULL, we're coming from the GUP-fast path and might have |
1057 | * to fallback to the slow path just to lookup the vma. |
1058 | */ |
1059 | static inline bool gup_must_unshare(struct vm_area_struct *vma, |
1060 | unsigned int flags, struct page *page) |
1061 | { |
1062 | /* |
1063 | * FOLL_WRITE is implicitly handled correctly as the page table entry |
1064 | * has to be writable -- and if it references (part of) an anonymous |
1065 | * folio, that part is required to be marked exclusive. |
1066 | */ |
1067 | if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN) |
1068 | return false; |
1069 | /* |
1070 | * Note: PageAnon(page) is stable until the page is actually getting |
1071 | * freed. |
1072 | */ |
1073 | if (!PageAnon(page)) { |
1074 | /* |
1075 | * We only care about R/O long-term pining: R/O short-term |
1076 | * pinning does not have the semantics to observe successive |
1077 | * changes through the process page tables. |
1078 | */ |
1079 | if (!(flags & FOLL_LONGTERM)) |
1080 | return false; |
1081 | |
1082 | /* We really need the vma ... */ |
1083 | if (!vma) |
1084 | return true; |
1085 | |
1086 | /* |
1087 | * ... because we only care about writable private ("COW") |
1088 | * mappings where we have to break COW early. |
1089 | */ |
1090 | return is_cow_mapping(flags: vma->vm_flags); |
1091 | } |
1092 | |
1093 | /* Paired with a memory barrier in page_try_share_anon_rmap(). */ |
1094 | if (IS_ENABLED(CONFIG_HAVE_FAST_GUP)) |
1095 | smp_rmb(); |
1096 | |
1097 | /* |
1098 | * During GUP-fast we might not get called on the head page for a |
1099 | * hugetlb page that is mapped using cont-PTE, because GUP-fast does |
1100 | * not work with the abstracted hugetlb PTEs that always point at the |
1101 | * head page. For hugetlb, PageAnonExclusive only applies on the head |
1102 | * page (as it cannot be partially COW-shared), so lookup the head page. |
1103 | */ |
1104 | if (unlikely(!PageHead(page) && PageHuge(page))) |
1105 | page = compound_head(page); |
1106 | |
1107 | /* |
1108 | * Note that PageKsm() pages cannot be exclusive, and consequently, |
1109 | * cannot get pinned. |
1110 | */ |
1111 | return !PageAnonExclusive(page); |
1112 | } |
1113 | |
1114 | extern bool mirrored_kernelcore; |
1115 | extern bool memblock_has_mirror(void); |
1116 | |
1117 | static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma) |
1118 | { |
1119 | /* |
1120 | * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty |
1121 | * enablements, because when without soft-dirty being compiled in, |
1122 | * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY) |
1123 | * will be constantly true. |
1124 | */ |
1125 | if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY)) |
1126 | return false; |
1127 | |
1128 | /* |
1129 | * Soft-dirty is kind of special: its tracking is enabled when the |
1130 | * vma flags not set. |
1131 | */ |
1132 | return !(vma->vm_flags & VM_SOFTDIRTY); |
1133 | } |
1134 | |
1135 | static inline void vma_iter_config(struct vma_iterator *vmi, |
1136 | unsigned long index, unsigned long last) |
1137 | { |
1138 | MAS_BUG_ON(&vmi->mas, vmi->mas.node != MAS_START && |
1139 | (vmi->mas.index > index || vmi->mas.last < index)); |
1140 | __mas_set_range(mas: &vmi->mas, start: index, last: last - 1); |
1141 | } |
1142 | |
1143 | /* |
1144 | * VMA Iterator functions shared between nommu and mmap |
1145 | */ |
1146 | static inline int vma_iter_prealloc(struct vma_iterator *vmi, |
1147 | struct vm_area_struct *vma) |
1148 | { |
1149 | return mas_preallocate(mas: &vmi->mas, entry: vma, GFP_KERNEL); |
1150 | } |
1151 | |
1152 | static inline void vma_iter_clear(struct vma_iterator *vmi) |
1153 | { |
1154 | mas_store_prealloc(mas: &vmi->mas, NULL); |
1155 | } |
1156 | |
1157 | static inline int vma_iter_clear_gfp(struct vma_iterator *vmi, |
1158 | unsigned long start, unsigned long end, gfp_t gfp) |
1159 | { |
1160 | __mas_set_range(mas: &vmi->mas, start, last: end - 1); |
1161 | mas_store_gfp(mas: &vmi->mas, NULL, gfp); |
1162 | if (unlikely(mas_is_err(&vmi->mas))) |
1163 | return -ENOMEM; |
1164 | |
1165 | return 0; |
1166 | } |
1167 | |
1168 | static inline struct vm_area_struct *vma_iter_load(struct vma_iterator *vmi) |
1169 | { |
1170 | return mas_walk(mas: &vmi->mas); |
1171 | } |
1172 | |
1173 | /* Store a VMA with preallocated memory */ |
1174 | static inline void vma_iter_store(struct vma_iterator *vmi, |
1175 | struct vm_area_struct *vma) |
1176 | { |
1177 | |
1178 | #if defined(CONFIG_DEBUG_VM_MAPLE_TREE) |
1179 | if (MAS_WARN_ON(&vmi->mas, vmi->mas.node != MAS_START && |
1180 | vmi->mas.index > vma->vm_start)) { |
1181 | pr_warn("%lx > %lx\n store vma %lx-%lx\n into slot %lx-%lx\n" , |
1182 | vmi->mas.index, vma->vm_start, vma->vm_start, |
1183 | vma->vm_end, vmi->mas.index, vmi->mas.last); |
1184 | } |
1185 | if (MAS_WARN_ON(&vmi->mas, vmi->mas.node != MAS_START && |
1186 | vmi->mas.last < vma->vm_start)) { |
1187 | pr_warn("%lx < %lx\nstore vma %lx-%lx\ninto slot %lx-%lx\n" , |
1188 | vmi->mas.last, vma->vm_start, vma->vm_start, vma->vm_end, |
1189 | vmi->mas.index, vmi->mas.last); |
1190 | } |
1191 | #endif |
1192 | |
1193 | if (vmi->mas.node != MAS_START && |
1194 | ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start))) |
1195 | vma_iter_invalidate(vmi); |
1196 | |
1197 | __mas_set_range(mas: &vmi->mas, start: vma->vm_start, last: vma->vm_end - 1); |
1198 | mas_store_prealloc(mas: &vmi->mas, entry: vma); |
1199 | } |
1200 | |
1201 | static inline int vma_iter_store_gfp(struct vma_iterator *vmi, |
1202 | struct vm_area_struct *vma, gfp_t gfp) |
1203 | { |
1204 | if (vmi->mas.node != MAS_START && |
1205 | ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start))) |
1206 | vma_iter_invalidate(vmi); |
1207 | |
1208 | __mas_set_range(mas: &vmi->mas, start: vma->vm_start, last: vma->vm_end - 1); |
1209 | mas_store_gfp(mas: &vmi->mas, entry: vma, gfp); |
1210 | if (unlikely(mas_is_err(&vmi->mas))) |
1211 | return -ENOMEM; |
1212 | |
1213 | return 0; |
1214 | } |
1215 | |
1216 | /* |
1217 | * VMA lock generalization |
1218 | */ |
1219 | struct vma_prepare { |
1220 | struct vm_area_struct *vma; |
1221 | struct vm_area_struct *adj_next; |
1222 | struct file *file; |
1223 | struct address_space *mapping; |
1224 | struct anon_vma *anon_vma; |
1225 | struct vm_area_struct *insert; |
1226 | struct vm_area_struct *remove; |
1227 | struct vm_area_struct *remove2; |
1228 | }; |
1229 | |
1230 | void __meminit __init_single_page(struct page *page, unsigned long pfn, |
1231 | unsigned long zone, int nid); |
1232 | |
1233 | /* shrinker related functions */ |
1234 | unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, |
1235 | int priority); |
1236 | |
1237 | #ifdef CONFIG_SHRINKER_DEBUG |
1238 | static inline __printf(2, 0) int shrinker_debugfs_name_alloc( |
1239 | struct shrinker *shrinker, const char *fmt, va_list ap) |
1240 | { |
1241 | shrinker->name = kvasprintf_const(GFP_KERNEL, fmt, args: ap); |
1242 | |
1243 | return shrinker->name ? 0 : -ENOMEM; |
1244 | } |
1245 | |
1246 | static inline void shrinker_debugfs_name_free(struct shrinker *shrinker) |
1247 | { |
1248 | kfree_const(x: shrinker->name); |
1249 | shrinker->name = NULL; |
1250 | } |
1251 | |
1252 | extern int shrinker_debugfs_add(struct shrinker *shrinker); |
1253 | extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker, |
1254 | int *debugfs_id); |
1255 | extern void shrinker_debugfs_remove(struct dentry *debugfs_entry, |
1256 | int debugfs_id); |
1257 | #else /* CONFIG_SHRINKER_DEBUG */ |
1258 | static inline int shrinker_debugfs_add(struct shrinker *shrinker) |
1259 | { |
1260 | return 0; |
1261 | } |
1262 | static inline int shrinker_debugfs_name_alloc(struct shrinker *shrinker, |
1263 | const char *fmt, va_list ap) |
1264 | { |
1265 | return 0; |
1266 | } |
1267 | static inline void shrinker_debugfs_name_free(struct shrinker *shrinker) |
1268 | { |
1269 | } |
1270 | static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker, |
1271 | int *debugfs_id) |
1272 | { |
1273 | *debugfs_id = -1; |
1274 | return NULL; |
1275 | } |
1276 | static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry, |
1277 | int debugfs_id) |
1278 | { |
1279 | } |
1280 | #endif /* CONFIG_SHRINKER_DEBUG */ |
1281 | |
1282 | #endif /* __MM_INTERNAL_H */ |
1283 | |