xref: /linux/mm/internal.h (revision 9abd613a85af72fa560e49d9a0acc5b872840c72)
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 ENTIRELY_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 ENTIRELY_MAPPED		0x800000
62 #define FOLIO_PAGES_MAPPED	(ENTIRELY_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(&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 #ifdef CONFIG_MMU
87 
88 /* Flags for folio_pte_batch(). */
89 typedef int __bitwise fpb_t;
90 
91 /* Compare PTEs after pte_mkclean(), ignoring the dirty bit. */
92 #define FPB_IGNORE_DIRTY		((__force fpb_t)BIT(0))
93 
94 /* Compare PTEs after pte_clear_soft_dirty(), ignoring the soft-dirty bit. */
95 #define FPB_IGNORE_SOFT_DIRTY		((__force fpb_t)BIT(1))
96 
97 static inline pte_t __pte_batch_clear_ignored(pte_t pte, fpb_t flags)
98 {
99 	if (flags & FPB_IGNORE_DIRTY)
100 		pte = pte_mkclean(pte);
101 	if (likely(flags & FPB_IGNORE_SOFT_DIRTY))
102 		pte = pte_clear_soft_dirty(pte);
103 	return pte_wrprotect(pte_mkold(pte));
104 }
105 
106 /**
107  * folio_pte_batch - detect a PTE batch for a large folio
108  * @folio: The large folio to detect a PTE batch for.
109  * @addr: The user virtual address the first page is mapped at.
110  * @start_ptep: Page table pointer for the first entry.
111  * @pte: Page table entry for the first page.
112  * @max_nr: The maximum number of table entries to consider.
113  * @flags: Flags to modify the PTE batch semantics.
114  * @any_writable: Optional pointer to indicate whether any entry except the
115  *		  first one is writable.
116  *
117  * Detect a PTE batch: consecutive (present) PTEs that map consecutive
118  * pages of the same large folio.
119  *
120  * All PTEs inside a PTE batch have the same PTE bits set, excluding the PFN,
121  * the accessed bit, writable bit, dirty bit (with FPB_IGNORE_DIRTY) and
122  * soft-dirty bit (with FPB_IGNORE_SOFT_DIRTY).
123  *
124  * start_ptep must map any page of the folio. max_nr must be at least one and
125  * must be limited by the caller so scanning cannot exceed a single page table.
126  *
127  * Return: the number of table entries in the batch.
128  */
129 static inline int folio_pte_batch(struct folio *folio, unsigned long addr,
130 		pte_t *start_ptep, pte_t pte, int max_nr, fpb_t flags,
131 		bool *any_writable)
132 {
133 	unsigned long folio_end_pfn = folio_pfn(folio) + folio_nr_pages(folio);
134 	const pte_t *end_ptep = start_ptep + max_nr;
135 	pte_t expected_pte, *ptep;
136 	bool writable;
137 	int nr;
138 
139 	if (any_writable)
140 		*any_writable = false;
141 
142 	VM_WARN_ON_FOLIO(!pte_present(pte), folio);
143 	VM_WARN_ON_FOLIO(!folio_test_large(folio) || max_nr < 1, folio);
144 	VM_WARN_ON_FOLIO(page_folio(pfn_to_page(pte_pfn(pte))) != folio, folio);
145 
146 	nr = pte_batch_hint(start_ptep, pte);
147 	expected_pte = __pte_batch_clear_ignored(pte_advance_pfn(pte, nr), flags);
148 	ptep = start_ptep + nr;
149 
150 	while (ptep < end_ptep) {
151 		pte = ptep_get(ptep);
152 		if (any_writable)
153 			writable = !!pte_write(pte);
154 		pte = __pte_batch_clear_ignored(pte, flags);
155 
156 		if (!pte_same(pte, expected_pte))
157 			break;
158 
159 		/*
160 		 * Stop immediately once we reached the end of the folio. In
161 		 * corner cases the next PFN might fall into a different
162 		 * folio.
163 		 */
164 		if (pte_pfn(pte) >= folio_end_pfn)
165 			break;
166 
167 		if (any_writable)
168 			*any_writable |= writable;
169 
170 		nr = pte_batch_hint(ptep, pte);
171 		expected_pte = pte_advance_pfn(expected_pte, nr);
172 		ptep += nr;
173 	}
174 
175 	return min(ptep - start_ptep, max_nr);
176 }
177 #endif /* CONFIG_MMU */
178 
179 void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
180 						int nr_throttled);
181 static inline void acct_reclaim_writeback(struct folio *folio)
182 {
183 	pg_data_t *pgdat = folio_pgdat(folio);
184 	int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
185 
186 	if (nr_throttled)
187 		__acct_reclaim_writeback(pgdat, folio, nr_throttled);
188 }
189 
190 static inline void wake_throttle_isolated(pg_data_t *pgdat)
191 {
192 	wait_queue_head_t *wqh;
193 
194 	wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
195 	if (waitqueue_active(wqh))
196 		wake_up(wqh);
197 }
198 
199 vm_fault_t vmf_anon_prepare(struct vm_fault *vmf);
200 vm_fault_t do_swap_page(struct vm_fault *vmf);
201 void folio_rotate_reclaimable(struct folio *folio);
202 bool __folio_end_writeback(struct folio *folio);
203 void deactivate_file_folio(struct folio *folio);
204 void folio_activate(struct folio *folio);
205 
206 void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
207 		   struct vm_area_struct *start_vma, unsigned long floor,
208 		   unsigned long ceiling, bool mm_wr_locked);
209 void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
210 
211 struct zap_details;
212 void unmap_page_range(struct mmu_gather *tlb,
213 			     struct vm_area_struct *vma,
214 			     unsigned long addr, unsigned long end,
215 			     struct zap_details *details);
216 
217 void page_cache_ra_order(struct readahead_control *, struct file_ra_state *,
218 		unsigned int order);
219 void force_page_cache_ra(struct readahead_control *, unsigned long nr);
220 static inline void force_page_cache_readahead(struct address_space *mapping,
221 		struct file *file, pgoff_t index, unsigned long nr_to_read)
222 {
223 	DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
224 	force_page_cache_ra(&ractl, nr_to_read);
225 }
226 
227 unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start,
228 		pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
229 unsigned find_get_entries(struct address_space *mapping, pgoff_t *start,
230 		pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
231 void filemap_free_folio(struct address_space *mapping, struct folio *folio);
232 int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
233 bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
234 		loff_t end);
235 long mapping_evict_folio(struct address_space *mapping, struct folio *folio);
236 unsigned long mapping_try_invalidate(struct address_space *mapping,
237 		pgoff_t start, pgoff_t end, unsigned long *nr_failed);
238 
239 /**
240  * folio_evictable - Test whether a folio is evictable.
241  * @folio: The folio to test.
242  *
243  * Test whether @folio is evictable -- i.e., should be placed on
244  * active/inactive lists vs unevictable list.
245  *
246  * Reasons folio might not be evictable:
247  * 1. folio's mapping marked unevictable
248  * 2. One of the pages in the folio is part of an mlocked VMA
249  */
250 static inline bool folio_evictable(struct folio *folio)
251 {
252 	bool ret;
253 
254 	/* Prevent address_space of inode and swap cache from being freed */
255 	rcu_read_lock();
256 	ret = !mapping_unevictable(folio_mapping(folio)) &&
257 			!folio_test_mlocked(folio);
258 	rcu_read_unlock();
259 	return ret;
260 }
261 
262 /*
263  * Turn a non-refcounted page (->_refcount == 0) into refcounted with
264  * a count of one.
265  */
266 static inline void set_page_refcounted(struct page *page)
267 {
268 	VM_BUG_ON_PAGE(PageTail(page), page);
269 	VM_BUG_ON_PAGE(page_ref_count(page), page);
270 	set_page_count(page, 1);
271 }
272 
273 /*
274  * Return true if a folio needs ->release_folio() calling upon it.
275  */
276 static inline bool folio_needs_release(struct folio *folio)
277 {
278 	struct address_space *mapping = folio_mapping(folio);
279 
280 	return folio_has_private(folio) ||
281 		(mapping && mapping_release_always(mapping));
282 }
283 
284 extern unsigned long highest_memmap_pfn;
285 
286 /*
287  * Maximum number of reclaim retries without progress before the OOM
288  * killer is consider the only way forward.
289  */
290 #define MAX_RECLAIM_RETRIES 16
291 
292 /*
293  * in mm/vmscan.c:
294  */
295 bool isolate_lru_page(struct page *page);
296 bool folio_isolate_lru(struct folio *folio);
297 void putback_lru_page(struct page *page);
298 void folio_putback_lru(struct folio *folio);
299 extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
300 
301 /*
302  * in mm/rmap.c:
303  */
304 pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
305 
306 /*
307  * in mm/page_alloc.c
308  */
309 #define K(x) ((x) << (PAGE_SHIFT-10))
310 
311 extern char * const zone_names[MAX_NR_ZONES];
312 
313 /* perform sanity checks on struct pages being allocated or freed */
314 DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled);
315 
316 extern int min_free_kbytes;
317 
318 void setup_per_zone_wmarks(void);
319 void calculate_min_free_kbytes(void);
320 int __meminit init_per_zone_wmark_min(void);
321 void page_alloc_sysctl_init(void);
322 
323 /*
324  * Structure for holding the mostly immutable allocation parameters passed
325  * between functions involved in allocations, including the alloc_pages*
326  * family of functions.
327  *
328  * nodemask, migratetype and highest_zoneidx are initialized only once in
329  * __alloc_pages() and then never change.
330  *
331  * zonelist, preferred_zone and highest_zoneidx are set first in
332  * __alloc_pages() for the fast path, and might be later changed
333  * in __alloc_pages_slowpath(). All other functions pass the whole structure
334  * by a const pointer.
335  */
336 struct alloc_context {
337 	struct zonelist *zonelist;
338 	nodemask_t *nodemask;
339 	struct zoneref *preferred_zoneref;
340 	int migratetype;
341 
342 	/*
343 	 * highest_zoneidx represents highest usable zone index of
344 	 * the allocation request. Due to the nature of the zone,
345 	 * memory on lower zone than the highest_zoneidx will be
346 	 * protected by lowmem_reserve[highest_zoneidx].
347 	 *
348 	 * highest_zoneidx is also used by reclaim/compaction to limit
349 	 * the target zone since higher zone than this index cannot be
350 	 * usable for this allocation request.
351 	 */
352 	enum zone_type highest_zoneidx;
353 	bool spread_dirty_pages;
354 };
355 
356 /*
357  * This function returns the order of a free page in the buddy system. In
358  * general, page_zone(page)->lock must be held by the caller to prevent the
359  * page from being allocated in parallel and returning garbage as the order.
360  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
361  * page cannot be allocated or merged in parallel. Alternatively, it must
362  * handle invalid values gracefully, and use buddy_order_unsafe() below.
363  */
364 static inline unsigned int buddy_order(struct page *page)
365 {
366 	/* PageBuddy() must be checked by the caller */
367 	return page_private(page);
368 }
369 
370 /*
371  * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
372  * PageBuddy() should be checked first by the caller to minimize race window,
373  * and invalid values must be handled gracefully.
374  *
375  * READ_ONCE is used so that if the caller assigns the result into a local
376  * variable and e.g. tests it for valid range before using, the compiler cannot
377  * decide to remove the variable and inline the page_private(page) multiple
378  * times, potentially observing different values in the tests and the actual
379  * use of the result.
380  */
381 #define buddy_order_unsafe(page)	READ_ONCE(page_private(page))
382 
383 /*
384  * This function checks whether a page is free && is the buddy
385  * we can coalesce a page and its buddy if
386  * (a) the buddy is not in a hole (check before calling!) &&
387  * (b) the buddy is in the buddy system &&
388  * (c) a page and its buddy have the same order &&
389  * (d) a page and its buddy are in the same zone.
390  *
391  * For recording whether a page is in the buddy system, we set PageBuddy.
392  * Setting, clearing, and testing PageBuddy is serialized by zone->lock.
393  *
394  * For recording page's order, we use page_private(page).
395  */
396 static inline bool page_is_buddy(struct page *page, struct page *buddy,
397 				 unsigned int order)
398 {
399 	if (!page_is_guard(buddy) && !PageBuddy(buddy))
400 		return false;
401 
402 	if (buddy_order(buddy) != order)
403 		return false;
404 
405 	/*
406 	 * zone check is done late to avoid uselessly calculating
407 	 * zone/node ids for pages that could never merge.
408 	 */
409 	if (page_zone_id(page) != page_zone_id(buddy))
410 		return false;
411 
412 	VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
413 
414 	return true;
415 }
416 
417 /*
418  * Locate the struct page for both the matching buddy in our
419  * pair (buddy1) and the combined O(n+1) page they form (page).
420  *
421  * 1) Any buddy B1 will have an order O twin B2 which satisfies
422  * the following equation:
423  *     B2 = B1 ^ (1 << O)
424  * For example, if the starting buddy (buddy2) is #8 its order
425  * 1 buddy is #10:
426  *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
427  *
428  * 2) Any buddy B will have an order O+1 parent P which
429  * satisfies the following equation:
430  *     P = B & ~(1 << O)
431  *
432  * Assumption: *_mem_map is contiguous at least up to MAX_PAGE_ORDER
433  */
434 static inline unsigned long
435 __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
436 {
437 	return page_pfn ^ (1 << order);
438 }
439 
440 /*
441  * Find the buddy of @page and validate it.
442  * @page: The input page
443  * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the
444  *       function is used in the performance-critical __free_one_page().
445  * @order: The order of the page
446  * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to
447  *             page_to_pfn().
448  *
449  * The found buddy can be a non PageBuddy, out of @page's zone, or its order is
450  * not the same as @page. The validation is necessary before use it.
451  *
452  * Return: the found buddy page or NULL if not found.
453  */
454 static inline struct page *find_buddy_page_pfn(struct page *page,
455 			unsigned long pfn, unsigned int order, unsigned long *buddy_pfn)
456 {
457 	unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order);
458 	struct page *buddy;
459 
460 	buddy = page + (__buddy_pfn - pfn);
461 	if (buddy_pfn)
462 		*buddy_pfn = __buddy_pfn;
463 
464 	if (page_is_buddy(page, buddy, order))
465 		return buddy;
466 	return NULL;
467 }
468 
469 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
470 				unsigned long end_pfn, struct zone *zone);
471 
472 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
473 				unsigned long end_pfn, struct zone *zone)
474 {
475 	if (zone->contiguous)
476 		return pfn_to_page(start_pfn);
477 
478 	return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
479 }
480 
481 void set_zone_contiguous(struct zone *zone);
482 
483 static inline void clear_zone_contiguous(struct zone *zone)
484 {
485 	zone->contiguous = false;
486 }
487 
488 extern int __isolate_free_page(struct page *page, unsigned int order);
489 extern void __putback_isolated_page(struct page *page, unsigned int order,
490 				    int mt);
491 extern void memblock_free_pages(struct page *page, unsigned long pfn,
492 					unsigned int order);
493 extern void __free_pages_core(struct page *page, unsigned int order);
494 
495 /*
496  * This will have no effect, other than possibly generating a warning, if the
497  * caller passes in a non-large folio.
498  */
499 static inline void folio_set_order(struct folio *folio, unsigned int order)
500 {
501 	if (WARN_ON_ONCE(!order || !folio_test_large(folio)))
502 		return;
503 
504 	folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order;
505 #ifdef CONFIG_64BIT
506 	folio->_folio_nr_pages = 1U << order;
507 #endif
508 }
509 
510 void folio_undo_large_rmappable(struct folio *folio);
511 
512 static inline struct folio *page_rmappable_folio(struct page *page)
513 {
514 	struct folio *folio = (struct folio *)page;
515 
516 	folio_prep_large_rmappable(folio);
517 	return folio;
518 }
519 
520 static inline void prep_compound_head(struct page *page, unsigned int order)
521 {
522 	struct folio *folio = (struct folio *)page;
523 
524 	folio_set_order(folio, order);
525 	atomic_set(&folio->_entire_mapcount, -1);
526 	atomic_set(&folio->_nr_pages_mapped, 0);
527 	atomic_set(&folio->_pincount, 0);
528 }
529 
530 static inline void prep_compound_tail(struct page *head, int tail_idx)
531 {
532 	struct page *p = head + tail_idx;
533 
534 	p->mapping = TAIL_MAPPING;
535 	set_compound_head(p, head);
536 	set_page_private(p, 0);
537 }
538 
539 extern void prep_compound_page(struct page *page, unsigned int order);
540 
541 extern void post_alloc_hook(struct page *page, unsigned int order,
542 					gfp_t gfp_flags);
543 extern bool free_pages_prepare(struct page *page, unsigned int order);
544 
545 extern int user_min_free_kbytes;
546 
547 void free_unref_page(struct page *page, unsigned int order);
548 void free_unref_folios(struct folio_batch *fbatch);
549 
550 extern void zone_pcp_reset(struct zone *zone);
551 extern void zone_pcp_disable(struct zone *zone);
552 extern void zone_pcp_enable(struct zone *zone);
553 extern void zone_pcp_init(struct zone *zone);
554 
555 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
556 			  phys_addr_t min_addr,
557 			  int nid, bool exact_nid);
558 
559 void memmap_init_range(unsigned long, int, unsigned long, unsigned long,
560 		unsigned long, enum meminit_context, struct vmem_altmap *, int);
561 
562 
563 int split_free_page(struct page *free_page,
564 			unsigned int order, unsigned long split_pfn_offset);
565 
566 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
567 
568 /*
569  * in mm/compaction.c
570  */
571 /*
572  * compact_control is used to track pages being migrated and the free pages
573  * they are being migrated to during memory compaction. The free_pfn starts
574  * at the end of a zone and migrate_pfn begins at the start. Movable pages
575  * are moved to the end of a zone during a compaction run and the run
576  * completes when free_pfn <= migrate_pfn
577  */
578 struct compact_control {
579 	struct list_head freepages[NR_PAGE_ORDERS];	/* List of free pages to migrate to */
580 	struct list_head migratepages;	/* List of pages being migrated */
581 	unsigned int nr_freepages;	/* Number of isolated free pages */
582 	unsigned int nr_migratepages;	/* Number of pages to migrate */
583 	unsigned long free_pfn;		/* isolate_freepages search base */
584 	/*
585 	 * Acts as an in/out parameter to page isolation for migration.
586 	 * isolate_migratepages uses it as a search base.
587 	 * isolate_migratepages_block will update the value to the next pfn
588 	 * after the last isolated one.
589 	 */
590 	unsigned long migrate_pfn;
591 	unsigned long fast_start_pfn;	/* a pfn to start linear scan from */
592 	struct zone *zone;
593 	unsigned long total_migrate_scanned;
594 	unsigned long total_free_scanned;
595 	unsigned short fast_search_fail;/* failures to use free list searches */
596 	short search_order;		/* order to start a fast search at */
597 	const gfp_t gfp_mask;		/* gfp mask of a direct compactor */
598 	int order;			/* order a direct compactor needs */
599 	int migratetype;		/* migratetype of direct compactor */
600 	const unsigned int alloc_flags;	/* alloc flags of a direct compactor */
601 	const int highest_zoneidx;	/* zone index of a direct compactor */
602 	enum migrate_mode mode;		/* Async or sync migration mode */
603 	bool ignore_skip_hint;		/* Scan blocks even if marked skip */
604 	bool no_set_skip_hint;		/* Don't mark blocks for skipping */
605 	bool ignore_block_suitable;	/* Scan blocks considered unsuitable */
606 	bool direct_compaction;		/* False from kcompactd or /proc/... */
607 	bool proactive_compaction;	/* kcompactd proactive compaction */
608 	bool whole_zone;		/* Whole zone should/has been scanned */
609 	bool contended;			/* Signal lock contention */
610 	bool finish_pageblock;		/* Scan the remainder of a pageblock. Used
611 					 * when there are potentially transient
612 					 * isolation or migration failures to
613 					 * ensure forward progress.
614 					 */
615 	bool alloc_contig;		/* alloc_contig_range allocation */
616 };
617 
618 /*
619  * Used in direct compaction when a page should be taken from the freelists
620  * immediately when one is created during the free path.
621  */
622 struct capture_control {
623 	struct compact_control *cc;
624 	struct page *page;
625 };
626 
627 unsigned long
628 isolate_freepages_range(struct compact_control *cc,
629 			unsigned long start_pfn, unsigned long end_pfn);
630 int
631 isolate_migratepages_range(struct compact_control *cc,
632 			   unsigned long low_pfn, unsigned long end_pfn);
633 
634 int __alloc_contig_migrate_range(struct compact_control *cc,
635 					unsigned long start, unsigned long end,
636 					int migratetype);
637 
638 /* Free whole pageblock and set its migration type to MIGRATE_CMA. */
639 void init_cma_reserved_pageblock(struct page *page);
640 
641 #endif /* CONFIG_COMPACTION || CONFIG_CMA */
642 
643 int find_suitable_fallback(struct free_area *area, unsigned int order,
644 			int migratetype, bool only_stealable, bool *can_steal);
645 
646 static inline bool free_area_empty(struct free_area *area, int migratetype)
647 {
648 	return list_empty(&area->free_list[migratetype]);
649 }
650 
651 /*
652  * These three helpers classifies VMAs for virtual memory accounting.
653  */
654 
655 /*
656  * Executable code area - executable, not writable, not stack
657  */
658 static inline bool is_exec_mapping(vm_flags_t flags)
659 {
660 	return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
661 }
662 
663 /*
664  * Stack area (including shadow stacks)
665  *
666  * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
667  * do_mmap() forbids all other combinations.
668  */
669 static inline bool is_stack_mapping(vm_flags_t flags)
670 {
671 	return ((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK);
672 }
673 
674 /*
675  * Data area - private, writable, not stack
676  */
677 static inline bool is_data_mapping(vm_flags_t flags)
678 {
679 	return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
680 }
681 
682 /* mm/util.c */
683 struct anon_vma *folio_anon_vma(struct folio *folio);
684 
685 #ifdef CONFIG_MMU
686 void unmap_mapping_folio(struct folio *folio);
687 extern long populate_vma_page_range(struct vm_area_struct *vma,
688 		unsigned long start, unsigned long end, int *locked);
689 extern long faultin_vma_page_range(struct vm_area_struct *vma,
690 				   unsigned long start, unsigned long end,
691 				   bool write, int *locked);
692 extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags,
693 			       unsigned long bytes);
694 
695 /*
696  * NOTE: This function can't tell whether the folio is "fully mapped" in the
697  * range.
698  * "fully mapped" means all the pages of folio is associated with the page
699  * table of range while this function just check whether the folio range is
700  * within the range [start, end). Function caller needs to do page table
701  * check if it cares about the page table association.
702  *
703  * Typical usage (like mlock or madvise) is:
704  * Caller knows at least 1 page of folio is associated with page table of VMA
705  * and the range [start, end) is intersect with the VMA range. Caller wants
706  * to know whether the folio is fully associated with the range. It calls
707  * this function to check whether the folio is in the range first. Then checks
708  * the page table to know whether the folio is fully mapped to the range.
709  */
710 static inline bool
711 folio_within_range(struct folio *folio, struct vm_area_struct *vma,
712 		unsigned long start, unsigned long end)
713 {
714 	pgoff_t pgoff, addr;
715 	unsigned long vma_pglen = vma_pages(vma);
716 
717 	VM_WARN_ON_FOLIO(folio_test_ksm(folio), folio);
718 	if (start > end)
719 		return false;
720 
721 	if (start < vma->vm_start)
722 		start = vma->vm_start;
723 
724 	if (end > vma->vm_end)
725 		end = vma->vm_end;
726 
727 	pgoff = folio_pgoff(folio);
728 
729 	/* if folio start address is not in vma range */
730 	if (!in_range(pgoff, vma->vm_pgoff, vma_pglen))
731 		return false;
732 
733 	addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
734 
735 	return !(addr < start || end - addr < folio_size(folio));
736 }
737 
738 static inline bool
739 folio_within_vma(struct folio *folio, struct vm_area_struct *vma)
740 {
741 	return folio_within_range(folio, vma, vma->vm_start, vma->vm_end);
742 }
743 
744 /*
745  * mlock_vma_folio() and munlock_vma_folio():
746  * should be called with vma's mmap_lock held for read or write,
747  * under page table lock for the pte/pmd being added or removed.
748  *
749  * mlock is usually called at the end of folio_add_*_rmap_*(), munlock at
750  * the end of folio_remove_rmap_*(); but new anon folios are managed by
751  * folio_add_lru_vma() calling mlock_new_folio().
752  */
753 void mlock_folio(struct folio *folio);
754 static inline void mlock_vma_folio(struct folio *folio,
755 				struct vm_area_struct *vma)
756 {
757 	/*
758 	 * The VM_SPECIAL check here serves two purposes.
759 	 * 1) VM_IO check prevents migration from double-counting during mlock.
760 	 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
761 	 *    is never left set on a VM_SPECIAL vma, there is an interval while
762 	 *    file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
763 	 *    still be set while VM_SPECIAL bits are added: so ignore it then.
764 	 */
765 	if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED))
766 		mlock_folio(folio);
767 }
768 
769 void munlock_folio(struct folio *folio);
770 static inline void munlock_vma_folio(struct folio *folio,
771 					struct vm_area_struct *vma)
772 {
773 	/*
774 	 * munlock if the function is called. Ideally, we should only
775 	 * do munlock if any page of folio is unmapped from VMA and
776 	 * cause folio not fully mapped to VMA.
777 	 *
778 	 * But it's not easy to confirm that's the situation. So we
779 	 * always munlock the folio and page reclaim will correct it
780 	 * if it's wrong.
781 	 */
782 	if (unlikely(vma->vm_flags & VM_LOCKED))
783 		munlock_folio(folio);
784 }
785 
786 void mlock_new_folio(struct folio *folio);
787 bool need_mlock_drain(int cpu);
788 void mlock_drain_local(void);
789 void mlock_drain_remote(int cpu);
790 
791 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
792 
793 /*
794  * Return the start of user virtual address at the specific offset within
795  * a vma.
796  */
797 static inline unsigned long
798 vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages,
799 		  struct vm_area_struct *vma)
800 {
801 	unsigned long address;
802 
803 	if (pgoff >= vma->vm_pgoff) {
804 		address = vma->vm_start +
805 			((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
806 		/* Check for address beyond vma (or wrapped through 0?) */
807 		if (address < vma->vm_start || address >= vma->vm_end)
808 			address = -EFAULT;
809 	} else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) {
810 		/* Test above avoids possibility of wrap to 0 on 32-bit */
811 		address = vma->vm_start;
812 	} else {
813 		address = -EFAULT;
814 	}
815 	return address;
816 }
817 
818 /*
819  * Return the start of user virtual address of a page within a vma.
820  * Returns -EFAULT if all of the page is outside the range of vma.
821  * If page is a compound head, the entire compound page is considered.
822  */
823 static inline unsigned long
824 vma_address(struct page *page, struct vm_area_struct *vma)
825 {
826 	VM_BUG_ON_PAGE(PageKsm(page), page);	/* KSM page->index unusable */
827 	return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma);
828 }
829 
830 /*
831  * Then at what user virtual address will none of the range be found in vma?
832  * Assumes that vma_address() already returned a good starting address.
833  */
834 static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw)
835 {
836 	struct vm_area_struct *vma = pvmw->vma;
837 	pgoff_t pgoff;
838 	unsigned long address;
839 
840 	/* Common case, plus ->pgoff is invalid for KSM */
841 	if (pvmw->nr_pages == 1)
842 		return pvmw->address + PAGE_SIZE;
843 
844 	pgoff = pvmw->pgoff + pvmw->nr_pages;
845 	address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
846 	/* Check for address beyond vma (or wrapped through 0?) */
847 	if (address < vma->vm_start || address > vma->vm_end)
848 		address = vma->vm_end;
849 	return address;
850 }
851 
852 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
853 						    struct file *fpin)
854 {
855 	int flags = vmf->flags;
856 
857 	if (fpin)
858 		return fpin;
859 
860 	/*
861 	 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
862 	 * anything, so we only pin the file and drop the mmap_lock if only
863 	 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
864 	 */
865 	if (fault_flag_allow_retry_first(flags) &&
866 	    !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
867 		fpin = get_file(vmf->vma->vm_file);
868 		release_fault_lock(vmf);
869 	}
870 	return fpin;
871 }
872 #else /* !CONFIG_MMU */
873 static inline void unmap_mapping_folio(struct folio *folio) { }
874 static inline void mlock_new_folio(struct folio *folio) { }
875 static inline bool need_mlock_drain(int cpu) { return false; }
876 static inline void mlock_drain_local(void) { }
877 static inline void mlock_drain_remote(int cpu) { }
878 static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
879 {
880 }
881 #endif /* !CONFIG_MMU */
882 
883 /* Memory initialisation debug and verification */
884 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
885 DECLARE_STATIC_KEY_TRUE(deferred_pages);
886 
887 bool __init deferred_grow_zone(struct zone *zone, unsigned int order);
888 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
889 
890 enum mminit_level {
891 	MMINIT_WARNING,
892 	MMINIT_VERIFY,
893 	MMINIT_TRACE
894 };
895 
896 #ifdef CONFIG_DEBUG_MEMORY_INIT
897 
898 extern int mminit_loglevel;
899 
900 #define mminit_dprintk(level, prefix, fmt, arg...) \
901 do { \
902 	if (level < mminit_loglevel) { \
903 		if (level <= MMINIT_WARNING) \
904 			pr_warn("mminit::" prefix " " fmt, ##arg);	\
905 		else \
906 			printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
907 	} \
908 } while (0)
909 
910 extern void mminit_verify_pageflags_layout(void);
911 extern void mminit_verify_zonelist(void);
912 #else
913 
914 static inline void mminit_dprintk(enum mminit_level level,
915 				const char *prefix, const char *fmt, ...)
916 {
917 }
918 
919 static inline void mminit_verify_pageflags_layout(void)
920 {
921 }
922 
923 static inline void mminit_verify_zonelist(void)
924 {
925 }
926 #endif /* CONFIG_DEBUG_MEMORY_INIT */
927 
928 #define NODE_RECLAIM_NOSCAN	-2
929 #define NODE_RECLAIM_FULL	-1
930 #define NODE_RECLAIM_SOME	0
931 #define NODE_RECLAIM_SUCCESS	1
932 
933 #ifdef CONFIG_NUMA
934 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
935 extern int find_next_best_node(int node, nodemask_t *used_node_mask);
936 #else
937 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
938 				unsigned int order)
939 {
940 	return NODE_RECLAIM_NOSCAN;
941 }
942 static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
943 {
944 	return NUMA_NO_NODE;
945 }
946 #endif
947 
948 /*
949  * mm/memory-failure.c
950  */
951 extern int hwpoison_filter(struct page *p);
952 
953 extern u32 hwpoison_filter_dev_major;
954 extern u32 hwpoison_filter_dev_minor;
955 extern u64 hwpoison_filter_flags_mask;
956 extern u64 hwpoison_filter_flags_value;
957 extern u64 hwpoison_filter_memcg;
958 extern u32 hwpoison_filter_enable;
959 
960 extern unsigned long  __must_check vm_mmap_pgoff(struct file *, unsigned long,
961         unsigned long, unsigned long,
962         unsigned long, unsigned long);
963 
964 extern void set_pageblock_order(void);
965 unsigned long reclaim_pages(struct list_head *folio_list, bool ignore_references);
966 unsigned int reclaim_clean_pages_from_list(struct zone *zone,
967 					    struct list_head *folio_list);
968 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
969 #define ALLOC_WMARK_MIN		WMARK_MIN
970 #define ALLOC_WMARK_LOW		WMARK_LOW
971 #define ALLOC_WMARK_HIGH	WMARK_HIGH
972 #define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
973 
974 /* Mask to get the watermark bits */
975 #define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
976 
977 /*
978  * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
979  * cannot assume a reduced access to memory reserves is sufficient for
980  * !MMU
981  */
982 #ifdef CONFIG_MMU
983 #define ALLOC_OOM		0x08
984 #else
985 #define ALLOC_OOM		ALLOC_NO_WATERMARKS
986 #endif
987 
988 #define ALLOC_NON_BLOCK		 0x10 /* Caller cannot block. Allow access
989 				       * to 25% of the min watermark or
990 				       * 62.5% if __GFP_HIGH is set.
991 				       */
992 #define ALLOC_MIN_RESERVE	 0x20 /* __GFP_HIGH set. Allow access to 50%
993 				       * of the min watermark.
994 				       */
995 #define ALLOC_CPUSET		 0x40 /* check for correct cpuset */
996 #define ALLOC_CMA		 0x80 /* allow allocations from CMA areas */
997 #ifdef CONFIG_ZONE_DMA32
998 #define ALLOC_NOFRAGMENT	0x100 /* avoid mixing pageblock types */
999 #else
1000 #define ALLOC_NOFRAGMENT	  0x0
1001 #endif
1002 #define ALLOC_HIGHATOMIC	0x200 /* Allows access to MIGRATE_HIGHATOMIC */
1003 #define ALLOC_KSWAPD		0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
1004 
1005 /* Flags that allow allocations below the min watermark. */
1006 #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM)
1007 
1008 enum ttu_flags;
1009 struct tlbflush_unmap_batch;
1010 
1011 
1012 /*
1013  * only for MM internal work items which do not depend on
1014  * any allocations or locks which might depend on allocations
1015  */
1016 extern struct workqueue_struct *mm_percpu_wq;
1017 
1018 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1019 void try_to_unmap_flush(void);
1020 void try_to_unmap_flush_dirty(void);
1021 void flush_tlb_batched_pending(struct mm_struct *mm);
1022 #else
1023 static inline void try_to_unmap_flush(void)
1024 {
1025 }
1026 static inline void try_to_unmap_flush_dirty(void)
1027 {
1028 }
1029 static inline void flush_tlb_batched_pending(struct mm_struct *mm)
1030 {
1031 }
1032 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
1033 
1034 extern const struct trace_print_flags pageflag_names[];
1035 extern const struct trace_print_flags pagetype_names[];
1036 extern const struct trace_print_flags vmaflag_names[];
1037 extern const struct trace_print_flags gfpflag_names[];
1038 
1039 static inline bool is_migrate_highatomic(enum migratetype migratetype)
1040 {
1041 	return migratetype == MIGRATE_HIGHATOMIC;
1042 }
1043 
1044 static inline bool is_migrate_highatomic_page(struct page *page)
1045 {
1046 	return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
1047 }
1048 
1049 void setup_zone_pageset(struct zone *zone);
1050 
1051 struct migration_target_control {
1052 	int nid;		/* preferred node id */
1053 	nodemask_t *nmask;
1054 	gfp_t gfp_mask;
1055 };
1056 
1057 /*
1058  * mm/filemap.c
1059  */
1060 size_t splice_folio_into_pipe(struct pipe_inode_info *pipe,
1061 			      struct folio *folio, loff_t fpos, size_t size);
1062 
1063 /*
1064  * mm/vmalloc.c
1065  */
1066 #ifdef CONFIG_MMU
1067 void __init vmalloc_init(void);
1068 int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
1069                 pgprot_t prot, struct page **pages, unsigned int page_shift);
1070 #else
1071 static inline void vmalloc_init(void)
1072 {
1073 }
1074 
1075 static inline
1076 int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
1077                 pgprot_t prot, struct page **pages, unsigned int page_shift)
1078 {
1079 	return -EINVAL;
1080 }
1081 #endif
1082 
1083 int __must_check __vmap_pages_range_noflush(unsigned long addr,
1084 			       unsigned long end, pgprot_t prot,
1085 			       struct page **pages, unsigned int page_shift);
1086 
1087 void vunmap_range_noflush(unsigned long start, unsigned long end);
1088 
1089 void __vunmap_range_noflush(unsigned long start, unsigned long end);
1090 
1091 int numa_migrate_prep(struct folio *folio, struct vm_area_struct *vma,
1092 		      unsigned long addr, int page_nid, int *flags);
1093 
1094 void free_zone_device_page(struct page *page);
1095 int migrate_device_coherent_page(struct page *page);
1096 
1097 /*
1098  * mm/gup.c
1099  */
1100 struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags);
1101 int __must_check try_grab_page(struct page *page, unsigned int flags);
1102 
1103 /*
1104  * mm/huge_memory.c
1105  */
1106 struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
1107 				   unsigned long addr, pmd_t *pmd,
1108 				   unsigned int flags);
1109 
1110 /*
1111  * mm/mmap.c
1112  */
1113 struct vm_area_struct *vma_merge_extend(struct vma_iterator *vmi,
1114 					struct vm_area_struct *vma,
1115 					unsigned long delta);
1116 
1117 enum {
1118 	/* mark page accessed */
1119 	FOLL_TOUCH = 1 << 16,
1120 	/* a retry, previous pass started an IO */
1121 	FOLL_TRIED = 1 << 17,
1122 	/* we are working on non-current tsk/mm */
1123 	FOLL_REMOTE = 1 << 18,
1124 	/* pages must be released via unpin_user_page */
1125 	FOLL_PIN = 1 << 19,
1126 	/* gup_fast: prevent fall-back to slow gup */
1127 	FOLL_FAST_ONLY = 1 << 20,
1128 	/* allow unlocking the mmap lock */
1129 	FOLL_UNLOCKABLE = 1 << 21,
1130 };
1131 
1132 #define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \
1133 			    FOLL_FAST_ONLY | FOLL_UNLOCKABLE)
1134 
1135 /*
1136  * Indicates for which pages that are write-protected in the page table,
1137  * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
1138  * GUP pin will remain consistent with the pages mapped into the page tables
1139  * of the MM.
1140  *
1141  * Temporary unmapping of PageAnonExclusive() pages or clearing of
1142  * PageAnonExclusive() has to protect against concurrent GUP:
1143  * * Ordinary GUP: Using the PT lock
1144  * * GUP-fast and fork(): mm->write_protect_seq
1145  * * GUP-fast and KSM or temporary unmapping (swap, migration): see
1146  *    folio_try_share_anon_rmap_*()
1147  *
1148  * Must be called with the (sub)page that's actually referenced via the
1149  * page table entry, which might not necessarily be the head page for a
1150  * PTE-mapped THP.
1151  *
1152  * If the vma is NULL, we're coming from the GUP-fast path and might have
1153  * to fallback to the slow path just to lookup the vma.
1154  */
1155 static inline bool gup_must_unshare(struct vm_area_struct *vma,
1156 				    unsigned int flags, struct page *page)
1157 {
1158 	/*
1159 	 * FOLL_WRITE is implicitly handled correctly as the page table entry
1160 	 * has to be writable -- and if it references (part of) an anonymous
1161 	 * folio, that part is required to be marked exclusive.
1162 	 */
1163 	if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
1164 		return false;
1165 	/*
1166 	 * Note: PageAnon(page) is stable until the page is actually getting
1167 	 * freed.
1168 	 */
1169 	if (!PageAnon(page)) {
1170 		/*
1171 		 * We only care about R/O long-term pining: R/O short-term
1172 		 * pinning does not have the semantics to observe successive
1173 		 * changes through the process page tables.
1174 		 */
1175 		if (!(flags & FOLL_LONGTERM))
1176 			return false;
1177 
1178 		/* We really need the vma ... */
1179 		if (!vma)
1180 			return true;
1181 
1182 		/*
1183 		 * ... because we only care about writable private ("COW")
1184 		 * mappings where we have to break COW early.
1185 		 */
1186 		return is_cow_mapping(vma->vm_flags);
1187 	}
1188 
1189 	/* Paired with a memory barrier in folio_try_share_anon_rmap_*(). */
1190 	if (IS_ENABLED(CONFIG_HAVE_FAST_GUP))
1191 		smp_rmb();
1192 
1193 	/*
1194 	 * During GUP-fast we might not get called on the head page for a
1195 	 * hugetlb page that is mapped using cont-PTE, because GUP-fast does
1196 	 * not work with the abstracted hugetlb PTEs that always point at the
1197 	 * head page. For hugetlb, PageAnonExclusive only applies on the head
1198 	 * page (as it cannot be partially COW-shared), so lookup the head page.
1199 	 */
1200 	if (unlikely(!PageHead(page) && PageHuge(page)))
1201 		page = compound_head(page);
1202 
1203 	/*
1204 	 * Note that PageKsm() pages cannot be exclusive, and consequently,
1205 	 * cannot get pinned.
1206 	 */
1207 	return !PageAnonExclusive(page);
1208 }
1209 
1210 extern bool mirrored_kernelcore;
1211 extern bool memblock_has_mirror(void);
1212 
1213 static __always_inline void vma_set_range(struct vm_area_struct *vma,
1214 					  unsigned long start, unsigned long end,
1215 					  pgoff_t pgoff)
1216 {
1217 	vma->vm_start = start;
1218 	vma->vm_end = end;
1219 	vma->vm_pgoff = pgoff;
1220 }
1221 
1222 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma)
1223 {
1224 	/*
1225 	 * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty
1226 	 * enablements, because when without soft-dirty being compiled in,
1227 	 * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY)
1228 	 * will be constantly true.
1229 	 */
1230 	if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY))
1231 		return false;
1232 
1233 	/*
1234 	 * Soft-dirty is kind of special: its tracking is enabled when the
1235 	 * vma flags not set.
1236 	 */
1237 	return !(vma->vm_flags & VM_SOFTDIRTY);
1238 }
1239 
1240 static inline void vma_iter_config(struct vma_iterator *vmi,
1241 		unsigned long index, unsigned long last)
1242 {
1243 	__mas_set_range(&vmi->mas, index, last - 1);
1244 }
1245 
1246 /*
1247  * VMA Iterator functions shared between nommu and mmap
1248  */
1249 static inline int vma_iter_prealloc(struct vma_iterator *vmi,
1250 		struct vm_area_struct *vma)
1251 {
1252 	return mas_preallocate(&vmi->mas, vma, GFP_KERNEL);
1253 }
1254 
1255 static inline void vma_iter_clear(struct vma_iterator *vmi)
1256 {
1257 	mas_store_prealloc(&vmi->mas, NULL);
1258 }
1259 
1260 static inline struct vm_area_struct *vma_iter_load(struct vma_iterator *vmi)
1261 {
1262 	return mas_walk(&vmi->mas);
1263 }
1264 
1265 /* Store a VMA with preallocated memory */
1266 static inline void vma_iter_store(struct vma_iterator *vmi,
1267 				  struct vm_area_struct *vma)
1268 {
1269 
1270 #if defined(CONFIG_DEBUG_VM_MAPLE_TREE)
1271 	if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start &&
1272 			vmi->mas.index > vma->vm_start)) {
1273 		pr_warn("%lx > %lx\n store vma %lx-%lx\n into slot %lx-%lx\n",
1274 			vmi->mas.index, vma->vm_start, vma->vm_start,
1275 			vma->vm_end, vmi->mas.index, vmi->mas.last);
1276 	}
1277 	if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start &&
1278 			vmi->mas.last <  vma->vm_start)) {
1279 		pr_warn("%lx < %lx\nstore vma %lx-%lx\ninto slot %lx-%lx\n",
1280 		       vmi->mas.last, vma->vm_start, vma->vm_start, vma->vm_end,
1281 		       vmi->mas.index, vmi->mas.last);
1282 	}
1283 #endif
1284 
1285 	if (vmi->mas.status != ma_start &&
1286 	    ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start)))
1287 		vma_iter_invalidate(vmi);
1288 
1289 	__mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1);
1290 	mas_store_prealloc(&vmi->mas, vma);
1291 }
1292 
1293 static inline int vma_iter_store_gfp(struct vma_iterator *vmi,
1294 			struct vm_area_struct *vma, gfp_t gfp)
1295 {
1296 	if (vmi->mas.status != ma_start &&
1297 	    ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start)))
1298 		vma_iter_invalidate(vmi);
1299 
1300 	__mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1);
1301 	mas_store_gfp(&vmi->mas, vma, gfp);
1302 	if (unlikely(mas_is_err(&vmi->mas)))
1303 		return -ENOMEM;
1304 
1305 	return 0;
1306 }
1307 
1308 /*
1309  * VMA lock generalization
1310  */
1311 struct vma_prepare {
1312 	struct vm_area_struct *vma;
1313 	struct vm_area_struct *adj_next;
1314 	struct file *file;
1315 	struct address_space *mapping;
1316 	struct anon_vma *anon_vma;
1317 	struct vm_area_struct *insert;
1318 	struct vm_area_struct *remove;
1319 	struct vm_area_struct *remove2;
1320 };
1321 
1322 void __meminit __init_single_page(struct page *page, unsigned long pfn,
1323 				unsigned long zone, int nid);
1324 
1325 /* shrinker related functions */
1326 unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg,
1327 			  int priority);
1328 
1329 #ifdef CONFIG_SHRINKER_DEBUG
1330 static inline __printf(2, 0) int shrinker_debugfs_name_alloc(
1331 			struct shrinker *shrinker, const char *fmt, va_list ap)
1332 {
1333 	shrinker->name = kvasprintf_const(GFP_KERNEL, fmt, ap);
1334 
1335 	return shrinker->name ? 0 : -ENOMEM;
1336 }
1337 
1338 static inline void shrinker_debugfs_name_free(struct shrinker *shrinker)
1339 {
1340 	kfree_const(shrinker->name);
1341 	shrinker->name = NULL;
1342 }
1343 
1344 extern int shrinker_debugfs_add(struct shrinker *shrinker);
1345 extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
1346 					      int *debugfs_id);
1347 extern void shrinker_debugfs_remove(struct dentry *debugfs_entry,
1348 				    int debugfs_id);
1349 #else /* CONFIG_SHRINKER_DEBUG */
1350 static inline int shrinker_debugfs_add(struct shrinker *shrinker)
1351 {
1352 	return 0;
1353 }
1354 static inline int shrinker_debugfs_name_alloc(struct shrinker *shrinker,
1355 					      const char *fmt, va_list ap)
1356 {
1357 	return 0;
1358 }
1359 static inline void shrinker_debugfs_name_free(struct shrinker *shrinker)
1360 {
1361 }
1362 static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
1363 						     int *debugfs_id)
1364 {
1365 	*debugfs_id = -1;
1366 	return NULL;
1367 }
1368 static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry,
1369 					   int debugfs_id)
1370 {
1371 }
1372 #endif /* CONFIG_SHRINKER_DEBUG */
1373 
1374 /* Only track the nodes of mappings with shadow entries */
1375 void workingset_update_node(struct xa_node *node);
1376 extern struct list_lru shadow_nodes;
1377 
1378 #endif	/* __MM_INTERNAL_H */
1379