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 */
folio_nr_pages_mapped(struct folio * folio)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
folio_raw_mapping(struct folio * folio)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 /*
87 * This is a file-backed mapping, and is about to be memory mapped - invoke its
88 * mmap hook and safely handle error conditions. On error, VMA hooks will be
89 * mutated.
90 *
91 * @file: File which backs the mapping.
92 * @vma: VMA which we are mapping.
93 *
94 * Returns: 0 if success, error otherwise.
95 */
mmap_file(struct file * file,struct vm_area_struct * vma)96 static inline int mmap_file(struct file *file, struct vm_area_struct *vma)
97 {
98 int err = call_mmap(file, vma);
99
100 if (likely(!err))
101 return 0;
102
103 /*
104 * OK, we tried to call the file hook for mmap(), but an error
105 * arose. The mapping is in an inconsistent state and we most not invoke
106 * any further hooks on it.
107 */
108 vma->vm_ops = &vma_dummy_vm_ops;
109
110 return err;
111 }
112
113 /*
114 * If the VMA has a close hook then close it, and since closing it might leave
115 * it in an inconsistent state which makes the use of any hooks suspect, clear
116 * them down by installing dummy empty hooks.
117 */
vma_close(struct vm_area_struct * vma)118 static inline void vma_close(struct vm_area_struct *vma)
119 {
120 if (vma->vm_ops && vma->vm_ops->close) {
121 vma->vm_ops->close(vma);
122
123 /*
124 * The mapping is in an inconsistent state, and no further hooks
125 * may be invoked upon it.
126 */
127 vma->vm_ops = &vma_dummy_vm_ops;
128 }
129 }
130
131 void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
132 int nr_throttled);
acct_reclaim_writeback(struct folio * folio)133 static inline void acct_reclaim_writeback(struct folio *folio)
134 {
135 pg_data_t *pgdat = folio_pgdat(folio);
136 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
137
138 if (nr_throttled)
139 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
140 }
141
wake_throttle_isolated(pg_data_t * pgdat)142 static inline void wake_throttle_isolated(pg_data_t *pgdat)
143 {
144 wait_queue_head_t *wqh;
145
146 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
147 if (waitqueue_active(wqh))
148 wake_up(wqh);
149 }
150
151 vm_fault_t do_swap_page(struct vm_fault *vmf);
152 void folio_rotate_reclaimable(struct folio *folio);
153 bool __folio_end_writeback(struct folio *folio);
154 void deactivate_file_folio(struct folio *folio);
155 void folio_activate(struct folio *folio);
156
157 void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
158 struct vm_area_struct *start_vma, unsigned long floor,
159 unsigned long ceiling, bool mm_wr_locked);
160 void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
161
162 struct zap_details;
163 void unmap_page_range(struct mmu_gather *tlb,
164 struct vm_area_struct *vma,
165 unsigned long addr, unsigned long end,
166 struct zap_details *details);
167
168 void page_cache_ra_order(struct readahead_control *, struct file_ra_state *,
169 unsigned int order);
170 void force_page_cache_ra(struct readahead_control *, unsigned long nr);
force_page_cache_readahead(struct address_space * mapping,struct file * file,pgoff_t index,unsigned long nr_to_read)171 static inline void force_page_cache_readahead(struct address_space *mapping,
172 struct file *file, pgoff_t index, unsigned long nr_to_read)
173 {
174 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
175 force_page_cache_ra(&ractl, nr_to_read);
176 }
177
178 unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start,
179 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
180 unsigned find_get_entries(struct address_space *mapping, pgoff_t *start,
181 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
182 void filemap_free_folio(struct address_space *mapping, struct folio *folio);
183 int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
184 bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
185 loff_t end);
186 long invalidate_inode_page(struct page *page);
187 unsigned long mapping_try_invalidate(struct address_space *mapping,
188 pgoff_t start, pgoff_t end, unsigned long *nr_failed);
189
190 /**
191 * folio_evictable - Test whether a folio is evictable.
192 * @folio: The folio to test.
193 *
194 * Test whether @folio is evictable -- i.e., should be placed on
195 * active/inactive lists vs unevictable list.
196 *
197 * Reasons folio might not be evictable:
198 * 1. folio's mapping marked unevictable
199 * 2. One of the pages in the folio is part of an mlocked VMA
200 */
folio_evictable(struct folio * folio)201 static inline bool folio_evictable(struct folio *folio)
202 {
203 bool ret;
204
205 /* Prevent address_space of inode and swap cache from being freed */
206 rcu_read_lock();
207 ret = !mapping_unevictable(folio_mapping(folio)) &&
208 !folio_test_mlocked(folio);
209 rcu_read_unlock();
210 return ret;
211 }
212
213 /*
214 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
215 * a count of one.
216 */
set_page_refcounted(struct page * page)217 static inline void set_page_refcounted(struct page *page)
218 {
219 VM_BUG_ON_PAGE(PageTail(page), page);
220 VM_BUG_ON_PAGE(page_ref_count(page), page);
221 set_page_count(page, 1);
222 }
223
224 /*
225 * Return true if a folio needs ->release_folio() calling upon it.
226 */
folio_needs_release(struct folio * folio)227 static inline bool folio_needs_release(struct folio *folio)
228 {
229 struct address_space *mapping = folio_mapping(folio);
230
231 return folio_has_private(folio) ||
232 (mapping && mapping_release_always(mapping));
233 }
234
235 extern unsigned long highest_memmap_pfn;
236
237 /*
238 * Maximum number of reclaim retries without progress before the OOM
239 * killer is consider the only way forward.
240 */
241 #define MAX_RECLAIM_RETRIES 16
242
243 /*
244 * in mm/vmscan.c:
245 */
246 bool isolate_lru_page(struct page *page);
247 bool folio_isolate_lru(struct folio *folio);
248 void putback_lru_page(struct page *page);
249 void folio_putback_lru(struct folio *folio);
250 extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
251
252 /*
253 * in mm/rmap.c:
254 */
255 pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
256
257 /*
258 * in mm/page_alloc.c
259 */
260 #define K(x) ((x) << (PAGE_SHIFT-10))
261
262 extern char * const zone_names[MAX_NR_ZONES];
263
264 /* perform sanity checks on struct pages being allocated or freed */
265 DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled);
266
267 extern int min_free_kbytes;
268
269 void setup_per_zone_wmarks(void);
270 void calculate_min_free_kbytes(void);
271 int __meminit init_per_zone_wmark_min(void);
272 void page_alloc_sysctl_init(void);
273
274 /*
275 * Structure for holding the mostly immutable allocation parameters passed
276 * between functions involved in allocations, including the alloc_pages*
277 * family of functions.
278 *
279 * nodemask, migratetype and highest_zoneidx are initialized only once in
280 * __alloc_pages() and then never change.
281 *
282 * zonelist, preferred_zone and highest_zoneidx are set first in
283 * __alloc_pages() for the fast path, and might be later changed
284 * in __alloc_pages_slowpath(). All other functions pass the whole structure
285 * by a const pointer.
286 */
287 struct alloc_context {
288 struct zonelist *zonelist;
289 nodemask_t *nodemask;
290 struct zoneref *preferred_zoneref;
291 int migratetype;
292
293 /*
294 * highest_zoneidx represents highest usable zone index of
295 * the allocation request. Due to the nature of the zone,
296 * memory on lower zone than the highest_zoneidx will be
297 * protected by lowmem_reserve[highest_zoneidx].
298 *
299 * highest_zoneidx is also used by reclaim/compaction to limit
300 * the target zone since higher zone than this index cannot be
301 * usable for this allocation request.
302 */
303 enum zone_type highest_zoneidx;
304 bool spread_dirty_pages;
305 };
306
307 /*
308 * This function returns the order of a free page in the buddy system. In
309 * general, page_zone(page)->lock must be held by the caller to prevent the
310 * page from being allocated in parallel and returning garbage as the order.
311 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
312 * page cannot be allocated or merged in parallel. Alternatively, it must
313 * handle invalid values gracefully, and use buddy_order_unsafe() below.
314 */
buddy_order(struct page * page)315 static inline unsigned int buddy_order(struct page *page)
316 {
317 /* PageBuddy() must be checked by the caller */
318 return page_private(page);
319 }
320
321 /*
322 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
323 * PageBuddy() should be checked first by the caller to minimize race window,
324 * and invalid values must be handled gracefully.
325 *
326 * READ_ONCE is used so that if the caller assigns the result into a local
327 * variable and e.g. tests it for valid range before using, the compiler cannot
328 * decide to remove the variable and inline the page_private(page) multiple
329 * times, potentially observing different values in the tests and the actual
330 * use of the result.
331 */
332 #define buddy_order_unsafe(page) READ_ONCE(page_private(page))
333
334 /*
335 * This function checks whether a page is free && is the buddy
336 * we can coalesce a page and its buddy if
337 * (a) the buddy is not in a hole (check before calling!) &&
338 * (b) the buddy is in the buddy system &&
339 * (c) a page and its buddy have the same order &&
340 * (d) a page and its buddy are in the same zone.
341 *
342 * For recording whether a page is in the buddy system, we set PageBuddy.
343 * Setting, clearing, and testing PageBuddy is serialized by zone->lock.
344 *
345 * For recording page's order, we use page_private(page).
346 */
page_is_buddy(struct page * page,struct page * buddy,unsigned int order)347 static inline bool page_is_buddy(struct page *page, struct page *buddy,
348 unsigned int order)
349 {
350 if (!page_is_guard(buddy) && !PageBuddy(buddy))
351 return false;
352
353 if (buddy_order(buddy) != order)
354 return false;
355
356 /*
357 * zone check is done late to avoid uselessly calculating
358 * zone/node ids for pages that could never merge.
359 */
360 if (page_zone_id(page) != page_zone_id(buddy))
361 return false;
362
363 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
364
365 return true;
366 }
367
368 /*
369 * Locate the struct page for both the matching buddy in our
370 * pair (buddy1) and the combined O(n+1) page they form (page).
371 *
372 * 1) Any buddy B1 will have an order O twin B2 which satisfies
373 * the following equation:
374 * B2 = B1 ^ (1 << O)
375 * For example, if the starting buddy (buddy2) is #8 its order
376 * 1 buddy is #10:
377 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
378 *
379 * 2) Any buddy B will have an order O+1 parent P which
380 * satisfies the following equation:
381 * P = B & ~(1 << O)
382 *
383 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
384 */
385 static inline unsigned long
__find_buddy_pfn(unsigned long page_pfn,unsigned int order)386 __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
387 {
388 return page_pfn ^ (1 << order);
389 }
390
391 /*
392 * Find the buddy of @page and validate it.
393 * @page: The input page
394 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the
395 * function is used in the performance-critical __free_one_page().
396 * @order: The order of the page
397 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to
398 * page_to_pfn().
399 *
400 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is
401 * not the same as @page. The validation is necessary before use it.
402 *
403 * Return: the found buddy page or NULL if not found.
404 */
find_buddy_page_pfn(struct page * page,unsigned long pfn,unsigned int order,unsigned long * buddy_pfn)405 static inline struct page *find_buddy_page_pfn(struct page *page,
406 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn)
407 {
408 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order);
409 struct page *buddy;
410
411 buddy = page + (__buddy_pfn - pfn);
412 if (buddy_pfn)
413 *buddy_pfn = __buddy_pfn;
414
415 if (page_is_buddy(page, buddy, order))
416 return buddy;
417 return NULL;
418 }
419
420 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
421 unsigned long end_pfn, struct zone *zone);
422
pageblock_pfn_to_page(unsigned long start_pfn,unsigned long end_pfn,struct zone * zone)423 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
424 unsigned long end_pfn, struct zone *zone)
425 {
426 if (zone->contiguous)
427 return pfn_to_page(start_pfn);
428
429 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
430 }
431
432 void set_zone_contiguous(struct zone *zone);
433
clear_zone_contiguous(struct zone * zone)434 static inline void clear_zone_contiguous(struct zone *zone)
435 {
436 zone->contiguous = false;
437 }
438
439 extern int __isolate_free_page(struct page *page, unsigned int order);
440 extern void __putback_isolated_page(struct page *page, unsigned int order,
441 int mt);
442 extern void memblock_free_pages(struct page *page, unsigned long pfn,
443 unsigned int order);
444 extern void __free_pages_core(struct page *page, unsigned int order);
445
446 /*
447 * This will have no effect, other than possibly generating a warning, if the
448 * caller passes in a non-large folio.
449 */
folio_set_order(struct folio * folio,unsigned int order)450 static inline void folio_set_order(struct folio *folio, unsigned int order)
451 {
452 if (WARN_ON_ONCE(!order || !folio_test_large(folio)))
453 return;
454
455 folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order;
456 #ifdef CONFIG_64BIT
457 folio->_folio_nr_pages = 1U << order;
458 #endif
459 }
460
461 bool __folio_unqueue_deferred_split(struct folio *folio);
folio_unqueue_deferred_split(struct folio * folio)462 static inline bool folio_unqueue_deferred_split(struct folio *folio)
463 {
464 if (folio_order(folio) <= 1 || !folio_test_large_rmappable(folio))
465 return false;
466
467 /*
468 * At this point, there is no one trying to add the folio to
469 * deferred_list. If folio is not in deferred_list, it's safe
470 * to check without acquiring the split_queue_lock.
471 */
472 if (data_race(list_empty(&folio->_deferred_list)))
473 return false;
474
475 return __folio_unqueue_deferred_split(folio);
476 }
477
page_rmappable_folio(struct page * page)478 static inline struct folio *page_rmappable_folio(struct page *page)
479 {
480 struct folio *folio = (struct folio *)page;
481
482 folio_prep_large_rmappable(folio);
483 return folio;
484 }
485
prep_compound_head(struct page * page,unsigned int order)486 static inline void prep_compound_head(struct page *page, unsigned int order)
487 {
488 struct folio *folio = (struct folio *)page;
489
490 folio_set_order(folio, order);
491 atomic_set(&folio->_entire_mapcount, -1);
492 atomic_set(&folio->_nr_pages_mapped, 0);
493 atomic_set(&folio->_pincount, 0);
494 if (order > 1)
495 INIT_LIST_HEAD(&folio->_deferred_list);
496 }
497
prep_compound_tail(struct page * head,int tail_idx)498 static inline void prep_compound_tail(struct page *head, int tail_idx)
499 {
500 struct page *p = head + tail_idx;
501
502 p->mapping = TAIL_MAPPING;
503 set_compound_head(p, head);
504 set_page_private(p, 0);
505 }
506
507 extern void prep_compound_page(struct page *page, unsigned int order);
508
509 extern void post_alloc_hook(struct page *page, unsigned int order,
510 gfp_t gfp_flags);
511 extern int user_min_free_kbytes;
512
513 extern void free_unref_page(struct page *page, unsigned int order);
514 extern void free_unref_page_list(struct list_head *list);
515
516 extern void zone_pcp_reset(struct zone *zone);
517 extern void zone_pcp_disable(struct zone *zone);
518 extern void zone_pcp_enable(struct zone *zone);
519 extern void zone_pcp_init(struct zone *zone);
520
521 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
522 phys_addr_t min_addr,
523 int nid, bool exact_nid);
524
525 void memmap_init_range(unsigned long, int, unsigned long, unsigned long,
526 unsigned long, enum meminit_context, struct vmem_altmap *, int);
527
528
529 int split_free_page(struct page *free_page,
530 unsigned int order, unsigned long split_pfn_offset);
531
532 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
533
534 /*
535 * in mm/compaction.c
536 */
537 /*
538 * compact_control is used to track pages being migrated and the free pages
539 * they are being migrated to during memory compaction. The free_pfn starts
540 * at the end of a zone and migrate_pfn begins at the start. Movable pages
541 * are moved to the end of a zone during a compaction run and the run
542 * completes when free_pfn <= migrate_pfn
543 */
544 struct compact_control {
545 struct list_head freepages; /* List of free pages to migrate to */
546 struct list_head migratepages; /* List of pages being migrated */
547 unsigned int nr_freepages; /* Number of isolated free pages */
548 unsigned int nr_migratepages; /* Number of pages to migrate */
549 unsigned long free_pfn; /* isolate_freepages search base */
550 /*
551 * Acts as an in/out parameter to page isolation for migration.
552 * isolate_migratepages uses it as a search base.
553 * isolate_migratepages_block will update the value to the next pfn
554 * after the last isolated one.
555 */
556 unsigned long migrate_pfn;
557 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
558 struct zone *zone;
559 unsigned long total_migrate_scanned;
560 unsigned long total_free_scanned;
561 unsigned short fast_search_fail;/* failures to use free list searches */
562 short search_order; /* order to start a fast search at */
563 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
564 int order; /* order a direct compactor needs */
565 int migratetype; /* migratetype of direct compactor */
566 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
567 const int highest_zoneidx; /* zone index of a direct compactor */
568 enum migrate_mode mode; /* Async or sync migration mode */
569 bool ignore_skip_hint; /* Scan blocks even if marked skip */
570 bool no_set_skip_hint; /* Don't mark blocks for skipping */
571 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
572 bool direct_compaction; /* False from kcompactd or /proc/... */
573 bool proactive_compaction; /* kcompactd proactive compaction */
574 bool whole_zone; /* Whole zone should/has been scanned */
575 bool contended; /* Signal lock contention */
576 bool finish_pageblock; /* Scan the remainder of a pageblock. Used
577 * when there are potentially transient
578 * isolation or migration failures to
579 * ensure forward progress.
580 */
581 bool alloc_contig; /* alloc_contig_range allocation */
582 };
583
584 /*
585 * Used in direct compaction when a page should be taken from the freelists
586 * immediately when one is created during the free path.
587 */
588 struct capture_control {
589 struct compact_control *cc;
590 struct page *page;
591 };
592
593 unsigned long
594 isolate_freepages_range(struct compact_control *cc,
595 unsigned long start_pfn, unsigned long end_pfn);
596 int
597 isolate_migratepages_range(struct compact_control *cc,
598 unsigned long low_pfn, unsigned long end_pfn);
599
600 int __alloc_contig_migrate_range(struct compact_control *cc,
601 unsigned long start, unsigned long end);
602
603 /* Free whole pageblock and set its migration type to MIGRATE_CMA. */
604 void init_cma_reserved_pageblock(struct page *page);
605
606 #endif /* CONFIG_COMPACTION || CONFIG_CMA */
607
608 int find_suitable_fallback(struct free_area *area, unsigned int order,
609 int migratetype, bool only_stealable, bool *can_steal);
610
free_area_empty(struct free_area * area,int migratetype)611 static inline bool free_area_empty(struct free_area *area, int migratetype)
612 {
613 return list_empty(&area->free_list[migratetype]);
614 }
615
616 /*
617 * These three helpers classifies VMAs for virtual memory accounting.
618 */
619
620 /*
621 * Executable code area - executable, not writable, not stack
622 */
is_exec_mapping(vm_flags_t flags)623 static inline bool is_exec_mapping(vm_flags_t flags)
624 {
625 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
626 }
627
628 /*
629 * Stack area (including shadow stacks)
630 *
631 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
632 * do_mmap() forbids all other combinations.
633 */
is_stack_mapping(vm_flags_t flags)634 static inline bool is_stack_mapping(vm_flags_t flags)
635 {
636 return ((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK);
637 }
638
639 /*
640 * Data area - private, writable, not stack
641 */
is_data_mapping(vm_flags_t flags)642 static inline bool is_data_mapping(vm_flags_t flags)
643 {
644 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
645 }
646
647 /* mm/util.c */
648 struct anon_vma *folio_anon_vma(struct folio *folio);
649
650 #ifdef CONFIG_MMU
651 void unmap_mapping_folio(struct folio *folio);
652 extern long populate_vma_page_range(struct vm_area_struct *vma,
653 unsigned long start, unsigned long end, int *locked);
654 extern long faultin_page_range(struct mm_struct *mm, unsigned long start,
655 unsigned long end, bool write, int *locked);
656 extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags,
657 unsigned long bytes);
658 /*
659 * mlock_vma_folio() and munlock_vma_folio():
660 * should be called with vma's mmap_lock held for read or write,
661 * under page table lock for the pte/pmd being added or removed.
662 *
663 * mlock is usually called at the end of page_add_*_rmap(), munlock at
664 * the end of page_remove_rmap(); but new anon folios are managed by
665 * folio_add_lru_vma() calling mlock_new_folio().
666 *
667 * @compound is used to include pmd mappings of THPs, but filter out
668 * pte mappings of THPs, which cannot be consistently counted: a pte
669 * mapping of the THP head cannot be distinguished by the page alone.
670 */
671 void mlock_folio(struct folio *folio);
mlock_vma_folio(struct folio * folio,struct vm_area_struct * vma,bool compound)672 static inline void mlock_vma_folio(struct folio *folio,
673 struct vm_area_struct *vma, bool compound)
674 {
675 /*
676 * The VM_SPECIAL check here serves two purposes.
677 * 1) VM_IO check prevents migration from double-counting during mlock.
678 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
679 * is never left set on a VM_SPECIAL vma, there is an interval while
680 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
681 * still be set while VM_SPECIAL bits are added: so ignore it then.
682 */
683 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED) &&
684 (compound || !folio_test_large(folio)))
685 mlock_folio(folio);
686 }
687
688 void munlock_folio(struct folio *folio);
munlock_vma_folio(struct folio * folio,struct vm_area_struct * vma,bool compound)689 static inline void munlock_vma_folio(struct folio *folio,
690 struct vm_area_struct *vma, bool compound)
691 {
692 if (unlikely(vma->vm_flags & VM_LOCKED) &&
693 (compound || !folio_test_large(folio)))
694 munlock_folio(folio);
695 }
696
697 void mlock_new_folio(struct folio *folio);
698 bool need_mlock_drain(int cpu);
699 void mlock_drain_local(void);
700 void mlock_drain_remote(int cpu);
701
702 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
703
704 /*
705 * Return the start of user virtual address at the specific offset within
706 * a vma.
707 */
708 static inline unsigned long
vma_pgoff_address(pgoff_t pgoff,unsigned long nr_pages,struct vm_area_struct * vma)709 vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages,
710 struct vm_area_struct *vma)
711 {
712 unsigned long address;
713
714 if (pgoff >= vma->vm_pgoff) {
715 address = vma->vm_start +
716 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
717 /* Check for address beyond vma (or wrapped through 0?) */
718 if (address < vma->vm_start || address >= vma->vm_end)
719 address = -EFAULT;
720 } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) {
721 /* Test above avoids possibility of wrap to 0 on 32-bit */
722 address = vma->vm_start;
723 } else {
724 address = -EFAULT;
725 }
726 return address;
727 }
728
729 /*
730 * Return the start of user virtual address of a page within a vma.
731 * Returns -EFAULT if all of the page is outside the range of vma.
732 * If page is a compound head, the entire compound page is considered.
733 */
734 static inline unsigned long
vma_address(struct page * page,struct vm_area_struct * vma)735 vma_address(struct page *page, struct vm_area_struct *vma)
736 {
737 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
738 return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma);
739 }
740
741 /*
742 * Then at what user virtual address will none of the range be found in vma?
743 * Assumes that vma_address() already returned a good starting address.
744 */
vma_address_end(struct page_vma_mapped_walk * pvmw)745 static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw)
746 {
747 struct vm_area_struct *vma = pvmw->vma;
748 pgoff_t pgoff;
749 unsigned long address;
750
751 /* Common case, plus ->pgoff is invalid for KSM */
752 if (pvmw->nr_pages == 1)
753 return pvmw->address + PAGE_SIZE;
754
755 pgoff = pvmw->pgoff + pvmw->nr_pages;
756 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
757 /* Check for address beyond vma (or wrapped through 0?) */
758 if (address < vma->vm_start || address > vma->vm_end)
759 address = vma->vm_end;
760 return address;
761 }
762
maybe_unlock_mmap_for_io(struct vm_fault * vmf,struct file * fpin)763 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
764 struct file *fpin)
765 {
766 int flags = vmf->flags;
767
768 if (fpin)
769 return fpin;
770
771 /*
772 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
773 * anything, so we only pin the file and drop the mmap_lock if only
774 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
775 */
776 if (fault_flag_allow_retry_first(flags) &&
777 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
778 fpin = get_file(vmf->vma->vm_file);
779 release_fault_lock(vmf);
780 }
781 return fpin;
782 }
783 #else /* !CONFIG_MMU */
unmap_mapping_folio(struct folio * folio)784 static inline void unmap_mapping_folio(struct folio *folio) { }
mlock_new_folio(struct folio * folio)785 static inline void mlock_new_folio(struct folio *folio) { }
need_mlock_drain(int cpu)786 static inline bool need_mlock_drain(int cpu) { return false; }
mlock_drain_local(void)787 static inline void mlock_drain_local(void) { }
mlock_drain_remote(int cpu)788 static inline void mlock_drain_remote(int cpu) { }
vunmap_range_noflush(unsigned long start,unsigned long end)789 static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
790 {
791 }
792 #endif /* !CONFIG_MMU */
793
794 /* Memory initialisation debug and verification */
795 #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
796 DECLARE_STATIC_KEY_TRUE(deferred_pages);
797
798 bool __init deferred_grow_zone(struct zone *zone, unsigned int order);
799 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
800
801 enum mminit_level {
802 MMINIT_WARNING,
803 MMINIT_VERIFY,
804 MMINIT_TRACE
805 };
806
807 #ifdef CONFIG_DEBUG_MEMORY_INIT
808
809 extern int mminit_loglevel;
810
811 #define mminit_dprintk(level, prefix, fmt, arg...) \
812 do { \
813 if (level < mminit_loglevel) { \
814 if (level <= MMINIT_WARNING) \
815 pr_warn("mminit::" prefix " " fmt, ##arg); \
816 else \
817 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
818 } \
819 } while (0)
820
821 extern void mminit_verify_pageflags_layout(void);
822 extern void mminit_verify_zonelist(void);
823 #else
824
mminit_dprintk(enum mminit_level level,const char * prefix,const char * fmt,...)825 static inline void mminit_dprintk(enum mminit_level level,
826 const char *prefix, const char *fmt, ...)
827 {
828 }
829
mminit_verify_pageflags_layout(void)830 static inline void mminit_verify_pageflags_layout(void)
831 {
832 }
833
mminit_verify_zonelist(void)834 static inline void mminit_verify_zonelist(void)
835 {
836 }
837 #endif /* CONFIG_DEBUG_MEMORY_INIT */
838
839 #define NODE_RECLAIM_NOSCAN -2
840 #define NODE_RECLAIM_FULL -1
841 #define NODE_RECLAIM_SOME 0
842 #define NODE_RECLAIM_SUCCESS 1
843
844 #ifdef CONFIG_NUMA
845 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
846 extern int find_next_best_node(int node, nodemask_t *used_node_mask);
847 #else
node_reclaim(struct pglist_data * pgdat,gfp_t mask,unsigned int order)848 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
849 unsigned int order)
850 {
851 return NODE_RECLAIM_NOSCAN;
852 }
find_next_best_node(int node,nodemask_t * used_node_mask)853 static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
854 {
855 return NUMA_NO_NODE;
856 }
857 #endif
858
859 /*
860 * mm/memory-failure.c
861 */
862 extern int hwpoison_filter(struct page *p);
863
864 extern u32 hwpoison_filter_dev_major;
865 extern u32 hwpoison_filter_dev_minor;
866 extern u64 hwpoison_filter_flags_mask;
867 extern u64 hwpoison_filter_flags_value;
868 extern u64 hwpoison_filter_memcg;
869 extern u32 hwpoison_filter_enable;
870
871 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
872 unsigned long, unsigned long,
873 unsigned long, unsigned long);
874
875 extern void set_pageblock_order(void);
876 unsigned long reclaim_pages(struct list_head *folio_list);
877 unsigned int reclaim_clean_pages_from_list(struct zone *zone,
878 struct list_head *folio_list);
879 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
880 #define ALLOC_WMARK_MIN WMARK_MIN
881 #define ALLOC_WMARK_LOW WMARK_LOW
882 #define ALLOC_WMARK_HIGH WMARK_HIGH
883 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
884
885 /* Mask to get the watermark bits */
886 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
887
888 /*
889 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
890 * cannot assume a reduced access to memory reserves is sufficient for
891 * !MMU
892 */
893 #ifdef CONFIG_MMU
894 #define ALLOC_OOM 0x08
895 #else
896 #define ALLOC_OOM ALLOC_NO_WATERMARKS
897 #endif
898
899 #define ALLOC_NON_BLOCK 0x10 /* Caller cannot block. Allow access
900 * to 25% of the min watermark or
901 * 62.5% if __GFP_HIGH is set.
902 */
903 #define ALLOC_MIN_RESERVE 0x20 /* __GFP_HIGH set. Allow access to 50%
904 * of the min watermark.
905 */
906 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
907 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
908 #ifdef CONFIG_ZONE_DMA32
909 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
910 #else
911 #define ALLOC_NOFRAGMENT 0x0
912 #endif
913 #define ALLOC_HIGHATOMIC 0x200 /* Allows access to MIGRATE_HIGHATOMIC */
914 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
915
916 /* Flags that allow allocations below the min watermark. */
917 #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM)
918
919 enum ttu_flags;
920 struct tlbflush_unmap_batch;
921
922
923 /*
924 * only for MM internal work items which do not depend on
925 * any allocations or locks which might depend on allocations
926 */
927 extern struct workqueue_struct *mm_percpu_wq;
928
929 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
930 void try_to_unmap_flush(void);
931 void try_to_unmap_flush_dirty(void);
932 void flush_tlb_batched_pending(struct mm_struct *mm);
933 #else
try_to_unmap_flush(void)934 static inline void try_to_unmap_flush(void)
935 {
936 }
try_to_unmap_flush_dirty(void)937 static inline void try_to_unmap_flush_dirty(void)
938 {
939 }
flush_tlb_batched_pending(struct mm_struct * mm)940 static inline void flush_tlb_batched_pending(struct mm_struct *mm)
941 {
942 }
943 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
944
945 extern const struct trace_print_flags pageflag_names[];
946 extern const struct trace_print_flags pagetype_names[];
947 extern const struct trace_print_flags vmaflag_names[];
948 extern const struct trace_print_flags gfpflag_names[];
949
is_migrate_highatomic(enum migratetype migratetype)950 static inline bool is_migrate_highatomic(enum migratetype migratetype)
951 {
952 return migratetype == MIGRATE_HIGHATOMIC;
953 }
954
is_migrate_highatomic_page(struct page * page)955 static inline bool is_migrate_highatomic_page(struct page *page)
956 {
957 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
958 }
959
960 void setup_zone_pageset(struct zone *zone);
961
962 struct migration_target_control {
963 int nid; /* preferred node id */
964 nodemask_t *nmask;
965 gfp_t gfp_mask;
966 };
967
968 /*
969 * mm/filemap.c
970 */
971 size_t splice_folio_into_pipe(struct pipe_inode_info *pipe,
972 struct folio *folio, loff_t fpos, size_t size);
973
974 /*
975 * mm/vmalloc.c
976 */
977 #ifdef CONFIG_MMU
978 void __init vmalloc_init(void);
979 int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
980 pgprot_t prot, struct page **pages, unsigned int page_shift);
981 #else
vmalloc_init(void)982 static inline void vmalloc_init(void)
983 {
984 }
985
986 static inline
vmap_pages_range_noflush(unsigned long addr,unsigned long end,pgprot_t prot,struct page ** pages,unsigned int page_shift)987 int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
988 pgprot_t prot, struct page **pages, unsigned int page_shift)
989 {
990 return -EINVAL;
991 }
992 #endif
993
994 int __must_check __vmap_pages_range_noflush(unsigned long addr,
995 unsigned long end, pgprot_t prot,
996 struct page **pages, unsigned int page_shift);
997
998 void vunmap_range_noflush(unsigned long start, unsigned long end);
999
1000 void __vunmap_range_noflush(unsigned long start, unsigned long end);
1001
1002 int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
1003 unsigned long addr, int page_nid, int *flags);
1004
1005 void free_zone_device_page(struct page *page);
1006 int migrate_device_coherent_page(struct page *page);
1007
1008 /*
1009 * mm/gup.c
1010 */
1011 int __must_check try_grab_folio(struct folio *folio, int refs,
1012 unsigned int flags);
1013
1014 /*
1015 * mm/huge_memory.c
1016 */
1017 struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
1018 unsigned long addr, pmd_t *pmd,
1019 unsigned int flags);
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 /* VMA lookup+checks compatible with MADV_POPULATE_(READ|WRITE) */
1035 FOLL_MADV_POPULATE = 1 << 22,
1036 };
1037
1038 #define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \
1039 FOLL_FAST_ONLY | FOLL_UNLOCKABLE | \
1040 FOLL_MADV_POPULATE)
1041
1042 /*
1043 * Indicates for which pages that are write-protected in the page table,
1044 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
1045 * GUP pin will remain consistent with the pages mapped into the page tables
1046 * of the MM.
1047 *
1048 * Temporary unmapping of PageAnonExclusive() pages or clearing of
1049 * PageAnonExclusive() has to protect against concurrent GUP:
1050 * * Ordinary GUP: Using the PT lock
1051 * * GUP-fast and fork(): mm->write_protect_seq
1052 * * GUP-fast and KSM or temporary unmapping (swap, migration): see
1053 * page_try_share_anon_rmap()
1054 *
1055 * Must be called with the (sub)page that's actually referenced via the
1056 * page table entry, which might not necessarily be the head page for a
1057 * PTE-mapped THP.
1058 *
1059 * If the vma is NULL, we're coming from the GUP-fast path and might have
1060 * to fallback to the slow path just to lookup the vma.
1061 */
gup_must_unshare(struct vm_area_struct * vma,unsigned int flags,struct page * page)1062 static inline bool gup_must_unshare(struct vm_area_struct *vma,
1063 unsigned int flags, struct page *page)
1064 {
1065 /*
1066 * FOLL_WRITE is implicitly handled correctly as the page table entry
1067 * has to be writable -- and if it references (part of) an anonymous
1068 * folio, that part is required to be marked exclusive.
1069 */
1070 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
1071 return false;
1072 /*
1073 * Note: PageAnon(page) is stable until the page is actually getting
1074 * freed.
1075 */
1076 if (!PageAnon(page)) {
1077 /*
1078 * We only care about R/O long-term pining: R/O short-term
1079 * pinning does not have the semantics to observe successive
1080 * changes through the process page tables.
1081 */
1082 if (!(flags & FOLL_LONGTERM))
1083 return false;
1084
1085 /* We really need the vma ... */
1086 if (!vma)
1087 return true;
1088
1089 /*
1090 * ... because we only care about writable private ("COW")
1091 * mappings where we have to break COW early.
1092 */
1093 return is_cow_mapping(vma->vm_flags);
1094 }
1095
1096 /* Paired with a memory barrier in page_try_share_anon_rmap(). */
1097 if (IS_ENABLED(CONFIG_HAVE_FAST_GUP))
1098 smp_rmb();
1099
1100 /*
1101 * During GUP-fast we might not get called on the head page for a
1102 * hugetlb page that is mapped using cont-PTE, because GUP-fast does
1103 * not work with the abstracted hugetlb PTEs that always point at the
1104 * head page. For hugetlb, PageAnonExclusive only applies on the head
1105 * page (as it cannot be partially COW-shared), so lookup the head page.
1106 */
1107 if (unlikely(!PageHead(page) && PageHuge(page)))
1108 page = compound_head(page);
1109
1110 /*
1111 * Note that PageKsm() pages cannot be exclusive, and consequently,
1112 * cannot get pinned.
1113 */
1114 return !PageAnonExclusive(page);
1115 }
1116
1117 extern bool mirrored_kernelcore;
1118 extern bool memblock_has_mirror(void);
1119
vma_soft_dirty_enabled(struct vm_area_struct * vma)1120 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma)
1121 {
1122 /*
1123 * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty
1124 * enablements, because when without soft-dirty being compiled in,
1125 * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY)
1126 * will be constantly true.
1127 */
1128 if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY))
1129 return false;
1130
1131 /*
1132 * Soft-dirty is kind of special: its tracking is enabled when the
1133 * vma flags not set.
1134 */
1135 return !(vma->vm_flags & VM_SOFTDIRTY);
1136 }
1137
vma_iter_config(struct vma_iterator * vmi,unsigned long index,unsigned long last)1138 static inline void vma_iter_config(struct vma_iterator *vmi,
1139 unsigned long index, unsigned long last)
1140 {
1141 MAS_BUG_ON(&vmi->mas, vmi->mas.node != MAS_START &&
1142 (vmi->mas.index > index || vmi->mas.last < index));
1143 __mas_set_range(&vmi->mas, index, last - 1);
1144 }
1145
1146 /*
1147 * VMA Iterator functions shared between nommu and mmap
1148 */
vma_iter_prealloc(struct vma_iterator * vmi,struct vm_area_struct * vma)1149 static inline int vma_iter_prealloc(struct vma_iterator *vmi,
1150 struct vm_area_struct *vma)
1151 {
1152 return mas_preallocate(&vmi->mas, vma, GFP_KERNEL);
1153 }
1154
vma_iter_clear(struct vma_iterator * vmi)1155 static inline void vma_iter_clear(struct vma_iterator *vmi)
1156 {
1157 mas_store_prealloc(&vmi->mas, NULL);
1158 }
1159
vma_iter_clear_gfp(struct vma_iterator * vmi,unsigned long start,unsigned long end,gfp_t gfp)1160 static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
1161 unsigned long start, unsigned long end, gfp_t gfp)
1162 {
1163 __mas_set_range(&vmi->mas, start, end - 1);
1164 mas_store_gfp(&vmi->mas, NULL, gfp);
1165 if (unlikely(mas_is_err(&vmi->mas)))
1166 return -ENOMEM;
1167
1168 return 0;
1169 }
1170
vma_iter_load(struct vma_iterator * vmi)1171 static inline struct vm_area_struct *vma_iter_load(struct vma_iterator *vmi)
1172 {
1173 return mas_walk(&vmi->mas);
1174 }
1175
1176 /* Store a VMA with preallocated memory */
vma_iter_store(struct vma_iterator * vmi,struct vm_area_struct * vma)1177 static inline void vma_iter_store(struct vma_iterator *vmi,
1178 struct vm_area_struct *vma)
1179 {
1180
1181 #if defined(CONFIG_DEBUG_VM_MAPLE_TREE)
1182 if (MAS_WARN_ON(&vmi->mas, vmi->mas.node != MAS_START &&
1183 vmi->mas.index > vma->vm_start)) {
1184 pr_warn("%lx > %lx\n store vma %lx-%lx\n into slot %lx-%lx\n",
1185 vmi->mas.index, vma->vm_start, vma->vm_start,
1186 vma->vm_end, vmi->mas.index, vmi->mas.last);
1187 }
1188 if (MAS_WARN_ON(&vmi->mas, vmi->mas.node != MAS_START &&
1189 vmi->mas.last < vma->vm_start)) {
1190 pr_warn("%lx < %lx\nstore vma %lx-%lx\ninto slot %lx-%lx\n",
1191 vmi->mas.last, vma->vm_start, vma->vm_start, vma->vm_end,
1192 vmi->mas.index, vmi->mas.last);
1193 }
1194 #endif
1195
1196 if (vmi->mas.node != MAS_START &&
1197 ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start)))
1198 vma_iter_invalidate(vmi);
1199
1200 __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1);
1201 mas_store_prealloc(&vmi->mas, vma);
1202 }
1203
vma_iter_store_gfp(struct vma_iterator * vmi,struct vm_area_struct * vma,gfp_t gfp)1204 static inline int vma_iter_store_gfp(struct vma_iterator *vmi,
1205 struct vm_area_struct *vma, gfp_t gfp)
1206 {
1207 if (vmi->mas.node != MAS_START &&
1208 ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start)))
1209 vma_iter_invalidate(vmi);
1210
1211 __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1);
1212 mas_store_gfp(&vmi->mas, vma, gfp);
1213 if (unlikely(mas_is_err(&vmi->mas)))
1214 return -ENOMEM;
1215
1216 return 0;
1217 }
1218
1219 /*
1220 * VMA lock generalization
1221 */
1222 struct vma_prepare {
1223 struct vm_area_struct *vma;
1224 struct vm_area_struct *adj_next;
1225 struct file *file;
1226 struct address_space *mapping;
1227 struct anon_vma *anon_vma;
1228 struct vm_area_struct *insert;
1229 struct vm_area_struct *remove;
1230 struct vm_area_struct *remove2;
1231 };
1232 #endif /* __MM_INTERNAL_H */
1233