]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/pagemap.h
dmaengine: imx-sdma: fix size check for sdma script_number
[mirror_ubuntu-bionic-kernel.git] / include / linux / pagemap.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_PAGEMAP_H
3 #define _LINUX_PAGEMAP_H
4
5 /*
6 * Copyright 1995 Linus Torvalds
7 */
8 #include <linux/mm.h>
9 #include <linux/fs.h>
10 #include <linux/list.h>
11 #include <linux/highmem.h>
12 #include <linux/compiler.h>
13 #include <linux/uaccess.h>
14 #include <linux/gfp.h>
15 #include <linux/bitops.h>
16 #include <linux/hardirq.h> /* for in_interrupt() */
17 #include <linux/hugetlb_inline.h>
18
19 struct pagevec;
20
21 /*
22 * Bits in mapping->flags.
23 */
24 enum mapping_flags {
25 AS_EIO = 0, /* IO error on async write */
26 AS_ENOSPC = 1, /* ENOSPC on async write */
27 AS_MM_ALL_LOCKS = 2, /* under mm_take_all_locks() */
28 AS_UNEVICTABLE = 3, /* e.g., ramdisk, SHM_LOCK */
29 AS_EXITING = 4, /* final truncate in progress */
30 /* writeback related tags are not used */
31 AS_NO_WRITEBACK_TAGS = 5,
32 };
33
34 /**
35 * mapping_set_error - record a writeback error in the address_space
36 * @mapping - the mapping in which an error should be set
37 * @error - the error to set in the mapping
38 *
39 * When writeback fails in some way, we must record that error so that
40 * userspace can be informed when fsync and the like are called. We endeavor
41 * to report errors on any file that was open at the time of the error. Some
42 * internal callers also need to know when writeback errors have occurred.
43 *
44 * When a writeback error occurs, most filesystems will want to call
45 * mapping_set_error to record the error in the mapping so that it can be
46 * reported when the application calls fsync(2).
47 */
48 static inline void mapping_set_error(struct address_space *mapping, int error)
49 {
50 if (likely(!error))
51 return;
52
53 /* Record in wb_err for checkers using errseq_t based tracking */
54 filemap_set_wb_err(mapping, error);
55
56 /* Record it in flags for now, for legacy callers */
57 if (error == -ENOSPC)
58 set_bit(AS_ENOSPC, &mapping->flags);
59 else
60 set_bit(AS_EIO, &mapping->flags);
61 }
62
63 static inline void mapping_set_unevictable(struct address_space *mapping)
64 {
65 set_bit(AS_UNEVICTABLE, &mapping->flags);
66 }
67
68 static inline void mapping_clear_unevictable(struct address_space *mapping)
69 {
70 clear_bit(AS_UNEVICTABLE, &mapping->flags);
71 }
72
73 static inline int mapping_unevictable(struct address_space *mapping)
74 {
75 if (mapping)
76 return test_bit(AS_UNEVICTABLE, &mapping->flags);
77 return !!mapping;
78 }
79
80 static inline void mapping_set_exiting(struct address_space *mapping)
81 {
82 set_bit(AS_EXITING, &mapping->flags);
83 }
84
85 static inline int mapping_exiting(struct address_space *mapping)
86 {
87 return test_bit(AS_EXITING, &mapping->flags);
88 }
89
90 static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
91 {
92 set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
93 }
94
95 static inline int mapping_use_writeback_tags(struct address_space *mapping)
96 {
97 return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
98 }
99
100 static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
101 {
102 return mapping->gfp_mask;
103 }
104
105 /* Restricts the given gfp_mask to what the mapping allows. */
106 static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
107 gfp_t gfp_mask)
108 {
109 return mapping_gfp_mask(mapping) & gfp_mask;
110 }
111
112 /*
113 * This is non-atomic. Only to be used before the mapping is activated.
114 * Probably needs a barrier...
115 */
116 static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
117 {
118 m->gfp_mask = mask;
119 }
120
121 void release_pages(struct page **pages, int nr);
122
123 /*
124 * speculatively take a reference to a page.
125 * If the page is free (_refcount == 0), then _refcount is untouched, and 0
126 * is returned. Otherwise, _refcount is incremented by 1 and 1 is returned.
127 *
128 * This function must be called inside the same rcu_read_lock() section as has
129 * been used to lookup the page in the pagecache radix-tree (or page table):
130 * this allows allocators to use a synchronize_rcu() to stabilize _refcount.
131 *
132 * Unless an RCU grace period has passed, the count of all pages coming out
133 * of the allocator must be considered unstable. page_count may return higher
134 * than expected, and put_page must be able to do the right thing when the
135 * page has been finished with, no matter what it is subsequently allocated
136 * for (because put_page is what is used here to drop an invalid speculative
137 * reference).
138 *
139 * This is the interesting part of the lockless pagecache (and lockless
140 * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page)
141 * has the following pattern:
142 * 1. find page in radix tree
143 * 2. conditionally increment refcount
144 * 3. check the page is still in pagecache (if no, goto 1)
145 *
146 * Remove-side that cares about stability of _refcount (eg. reclaim) has the
147 * following (with tree_lock held for write):
148 * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg)
149 * B. remove page from pagecache
150 * C. free the page
151 *
152 * There are 2 critical interleavings that matter:
153 * - 2 runs before A: in this case, A sees elevated refcount and bails out
154 * - A runs before 2: in this case, 2 sees zero refcount and retries;
155 * subsequently, B will complete and 1 will find no page, causing the
156 * lookup to return NULL.
157 *
158 * It is possible that between 1 and 2, the page is removed then the exact same
159 * page is inserted into the same position in pagecache. That's OK: the
160 * old find_get_page using tree_lock could equally have run before or after
161 * such a re-insertion, depending on order that locks are granted.
162 *
163 * Lookups racing against pagecache insertion isn't a big problem: either 1
164 * will find the page or it will not. Likewise, the old find_get_page could run
165 * either before the insertion or afterwards, depending on timing.
166 */
167 static inline int page_cache_get_speculative(struct page *page)
168 {
169 #ifdef CONFIG_TINY_RCU
170 # ifdef CONFIG_PREEMPT_COUNT
171 VM_BUG_ON(!in_atomic() && !irqs_disabled());
172 # endif
173 /*
174 * Preempt must be disabled here - we rely on rcu_read_lock doing
175 * this for us.
176 *
177 * Pagecache won't be truncated from interrupt context, so if we have
178 * found a page in the radix tree here, we have pinned its refcount by
179 * disabling preempt, and hence no need for the "speculative get" that
180 * SMP requires.
181 */
182 VM_BUG_ON_PAGE(page_count(page) == 0, page);
183 page_ref_inc(page);
184
185 #else
186 if (unlikely(!get_page_unless_zero(page))) {
187 /*
188 * Either the page has been freed, or will be freed.
189 * In either case, retry here and the caller should
190 * do the right thing (see comments above).
191 */
192 return 0;
193 }
194 #endif
195 VM_BUG_ON_PAGE(PageTail(page), page);
196
197 return 1;
198 }
199
200 /*
201 * Same as above, but add instead of inc (could just be merged)
202 */
203 static inline int page_cache_add_speculative(struct page *page, int count)
204 {
205 VM_BUG_ON(in_interrupt());
206
207 #if !defined(CONFIG_SMP) && defined(CONFIG_TREE_RCU)
208 # ifdef CONFIG_PREEMPT_COUNT
209 VM_BUG_ON(!in_atomic() && !irqs_disabled());
210 # endif
211 VM_BUG_ON_PAGE(page_count(page) == 0, page);
212 page_ref_add(page, count);
213
214 #else
215 if (unlikely(!page_ref_add_unless(page, count, 0)))
216 return 0;
217 #endif
218 VM_BUG_ON_PAGE(PageCompound(page) && page != compound_head(page), page);
219
220 return 1;
221 }
222
223 #ifdef CONFIG_NUMA
224 extern struct page *__page_cache_alloc(gfp_t gfp);
225 #else
226 static inline struct page *__page_cache_alloc(gfp_t gfp)
227 {
228 return alloc_pages(gfp, 0);
229 }
230 #endif
231
232 static inline struct page *page_cache_alloc(struct address_space *x)
233 {
234 return __page_cache_alloc(mapping_gfp_mask(x));
235 }
236
237 static inline gfp_t readahead_gfp_mask(struct address_space *x)
238 {
239 return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
240 }
241
242 typedef int filler_t(void *, struct page *);
243
244 pgoff_t page_cache_next_hole(struct address_space *mapping,
245 pgoff_t index, unsigned long max_scan);
246 pgoff_t page_cache_prev_hole(struct address_space *mapping,
247 pgoff_t index, unsigned long max_scan);
248
249 #define FGP_ACCESSED 0x00000001
250 #define FGP_LOCK 0x00000002
251 #define FGP_CREAT 0x00000004
252 #define FGP_WRITE 0x00000008
253 #define FGP_NOFS 0x00000010
254 #define FGP_NOWAIT 0x00000020
255
256 struct page *pagecache_get_page(struct address_space *mapping, pgoff_t offset,
257 int fgp_flags, gfp_t cache_gfp_mask);
258
259 /**
260 * find_get_page - find and get a page reference
261 * @mapping: the address_space to search
262 * @offset: the page index
263 *
264 * Looks up the page cache slot at @mapping & @offset. If there is a
265 * page cache page, it is returned with an increased refcount.
266 *
267 * Otherwise, %NULL is returned.
268 */
269 static inline struct page *find_get_page(struct address_space *mapping,
270 pgoff_t offset)
271 {
272 return pagecache_get_page(mapping, offset, 0, 0);
273 }
274
275 static inline struct page *find_get_page_flags(struct address_space *mapping,
276 pgoff_t offset, int fgp_flags)
277 {
278 return pagecache_get_page(mapping, offset, fgp_flags, 0);
279 }
280
281 /**
282 * find_lock_page - locate, pin and lock a pagecache page
283 * @mapping: the address_space to search
284 * @offset: the page index
285 *
286 * Looks up the page cache slot at @mapping & @offset. If there is a
287 * page cache page, it is returned locked and with an increased
288 * refcount.
289 *
290 * Otherwise, %NULL is returned.
291 *
292 * find_lock_page() may sleep.
293 */
294 static inline struct page *find_lock_page(struct address_space *mapping,
295 pgoff_t offset)
296 {
297 return pagecache_get_page(mapping, offset, FGP_LOCK, 0);
298 }
299
300 /**
301 * find_or_create_page - locate or add a pagecache page
302 * @mapping: the page's address_space
303 * @index: the page's index into the mapping
304 * @gfp_mask: page allocation mode
305 *
306 * Looks up the page cache slot at @mapping & @offset. If there is a
307 * page cache page, it is returned locked and with an increased
308 * refcount.
309 *
310 * If the page is not present, a new page is allocated using @gfp_mask
311 * and added to the page cache and the VM's LRU list. The page is
312 * returned locked and with an increased refcount.
313 *
314 * On memory exhaustion, %NULL is returned.
315 *
316 * find_or_create_page() may sleep, even if @gfp_flags specifies an
317 * atomic allocation!
318 */
319 static inline struct page *find_or_create_page(struct address_space *mapping,
320 pgoff_t offset, gfp_t gfp_mask)
321 {
322 return pagecache_get_page(mapping, offset,
323 FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
324 gfp_mask);
325 }
326
327 /**
328 * grab_cache_page_nowait - returns locked page at given index in given cache
329 * @mapping: target address_space
330 * @index: the page index
331 *
332 * Same as grab_cache_page(), but do not wait if the page is unavailable.
333 * This is intended for speculative data generators, where the data can
334 * be regenerated if the page couldn't be grabbed. This routine should
335 * be safe to call while holding the lock for another page.
336 *
337 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
338 * and deadlock against the caller's locked page.
339 */
340 static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
341 pgoff_t index)
342 {
343 return pagecache_get_page(mapping, index,
344 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
345 mapping_gfp_mask(mapping));
346 }
347
348 struct page *find_get_entry(struct address_space *mapping, pgoff_t offset);
349 struct page *find_lock_entry(struct address_space *mapping, pgoff_t offset);
350 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
351 unsigned int nr_entries, struct page **entries,
352 pgoff_t *indices);
353 unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
354 pgoff_t end, unsigned int nr_pages,
355 struct page **pages);
356 static inline unsigned find_get_pages(struct address_space *mapping,
357 pgoff_t *start, unsigned int nr_pages,
358 struct page **pages)
359 {
360 return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
361 pages);
362 }
363 unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
364 unsigned int nr_pages, struct page **pages);
365 unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
366 pgoff_t end, int tag, unsigned int nr_pages,
367 struct page **pages);
368 static inline unsigned find_get_pages_tag(struct address_space *mapping,
369 pgoff_t *index, int tag, unsigned int nr_pages,
370 struct page **pages)
371 {
372 return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
373 nr_pages, pages);
374 }
375 unsigned find_get_entries_tag(struct address_space *mapping, pgoff_t start,
376 int tag, unsigned int nr_entries,
377 struct page **entries, pgoff_t *indices);
378
379 struct page *grab_cache_page_write_begin(struct address_space *mapping,
380 pgoff_t index, unsigned flags);
381
382 /*
383 * Returns locked page at given index in given cache, creating it if needed.
384 */
385 static inline struct page *grab_cache_page(struct address_space *mapping,
386 pgoff_t index)
387 {
388 return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
389 }
390
391 extern struct page * read_cache_page(struct address_space *mapping,
392 pgoff_t index, filler_t *filler, void *data);
393 extern struct page * read_cache_page_gfp(struct address_space *mapping,
394 pgoff_t index, gfp_t gfp_mask);
395 extern int read_cache_pages(struct address_space *mapping,
396 struct list_head *pages, filler_t *filler, void *data);
397
398 static inline struct page *read_mapping_page(struct address_space *mapping,
399 pgoff_t index, void *data)
400 {
401 filler_t *filler = (filler_t *)mapping->a_ops->readpage;
402 return read_cache_page(mapping, index, filler, data);
403 }
404
405 /*
406 * Get index of the page with in radix-tree
407 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
408 */
409 static inline pgoff_t page_to_index(struct page *page)
410 {
411 pgoff_t pgoff;
412
413 if (likely(!PageTransTail(page)))
414 return page->index;
415
416 /*
417 * We don't initialize ->index for tail pages: calculate based on
418 * head page
419 */
420 pgoff = compound_head(page)->index;
421 pgoff += page - compound_head(page);
422 return pgoff;
423 }
424
425 /*
426 * Get the offset in PAGE_SIZE.
427 * (TODO: hugepage should have ->index in PAGE_SIZE)
428 */
429 static inline pgoff_t page_to_pgoff(struct page *page)
430 {
431 if (unlikely(PageHeadHuge(page)))
432 return page->index << compound_order(page);
433
434 return page_to_index(page);
435 }
436
437 /*
438 * Return byte-offset into filesystem object for page.
439 */
440 static inline loff_t page_offset(struct page *page)
441 {
442 return ((loff_t)page->index) << PAGE_SHIFT;
443 }
444
445 static inline loff_t page_file_offset(struct page *page)
446 {
447 return ((loff_t)page_index(page)) << PAGE_SHIFT;
448 }
449
450 extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
451 unsigned long address);
452
453 static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
454 unsigned long address)
455 {
456 pgoff_t pgoff;
457 if (unlikely(is_vm_hugetlb_page(vma)))
458 return linear_hugepage_index(vma, address);
459 pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
460 pgoff += vma->vm_pgoff;
461 return pgoff;
462 }
463
464 extern void __lock_page(struct page *page);
465 extern int __lock_page_killable(struct page *page);
466 extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm,
467 unsigned int flags);
468 extern void unlock_page(struct page *page);
469
470 static inline int trylock_page(struct page *page)
471 {
472 page = compound_head(page);
473 return (likely(!test_and_set_bit_lock(PG_locked, &page->flags)));
474 }
475
476 /*
477 * lock_page may only be called if we have the page's inode pinned.
478 */
479 static inline void lock_page(struct page *page)
480 {
481 might_sleep();
482 if (!trylock_page(page))
483 __lock_page(page);
484 }
485
486 /*
487 * lock_page_killable is like lock_page but can be interrupted by fatal
488 * signals. It returns 0 if it locked the page and -EINTR if it was
489 * killed while waiting.
490 */
491 static inline int lock_page_killable(struct page *page)
492 {
493 might_sleep();
494 if (!trylock_page(page))
495 return __lock_page_killable(page);
496 return 0;
497 }
498
499 /*
500 * lock_page_or_retry - Lock the page, unless this would block and the
501 * caller indicated that it can handle a retry.
502 *
503 * Return value and mmap_sem implications depend on flags; see
504 * __lock_page_or_retry().
505 */
506 static inline int lock_page_or_retry(struct page *page, struct mm_struct *mm,
507 unsigned int flags)
508 {
509 might_sleep();
510 return trylock_page(page) || __lock_page_or_retry(page, mm, flags);
511 }
512
513 /*
514 * This is exported only for wait_on_page_locked/wait_on_page_writeback, etc.,
515 * and should not be used directly.
516 */
517 extern void wait_on_page_bit(struct page *page, int bit_nr);
518 extern int wait_on_page_bit_killable(struct page *page, int bit_nr);
519
520 /*
521 * Wait for a page to be unlocked.
522 *
523 * This must be called with the caller "holding" the page,
524 * ie with increased "page->count" so that the page won't
525 * go away during the wait..
526 */
527 static inline void wait_on_page_locked(struct page *page)
528 {
529 if (PageLocked(page))
530 wait_on_page_bit(compound_head(page), PG_locked);
531 }
532
533 static inline int wait_on_page_locked_killable(struct page *page)
534 {
535 if (!PageLocked(page))
536 return 0;
537 return wait_on_page_bit_killable(compound_head(page), PG_locked);
538 }
539
540 /*
541 * Wait for a page to complete writeback
542 */
543 static inline void wait_on_page_writeback(struct page *page)
544 {
545 if (PageWriteback(page))
546 wait_on_page_bit(page, PG_writeback);
547 }
548
549 extern void end_page_writeback(struct page *page);
550 void wait_for_stable_page(struct page *page);
551
552 void page_endio(struct page *page, bool is_write, int err);
553
554 /*
555 * Add an arbitrary waiter to a page's wait queue
556 */
557 extern void add_page_wait_queue(struct page *page, wait_queue_entry_t *waiter);
558
559 /*
560 * Fault everything in given userspace address range in.
561 */
562 static inline int fault_in_pages_writeable(char __user *uaddr, int size)
563 {
564 char __user *end = uaddr + size - 1;
565
566 if (unlikely(size == 0))
567 return 0;
568
569 if (unlikely(uaddr > end))
570 return -EFAULT;
571 /*
572 * Writing zeroes into userspace here is OK, because we know that if
573 * the zero gets there, we'll be overwriting it.
574 */
575 do {
576 if (unlikely(__put_user(0, uaddr) != 0))
577 return -EFAULT;
578 uaddr += PAGE_SIZE;
579 } while (uaddr <= end);
580
581 /* Check whether the range spilled into the next page. */
582 if (((unsigned long)uaddr & PAGE_MASK) ==
583 ((unsigned long)end & PAGE_MASK))
584 return __put_user(0, end);
585
586 return 0;
587 }
588
589 static inline int fault_in_pages_readable(const char __user *uaddr, int size)
590 {
591 volatile char c;
592 const char __user *end = uaddr + size - 1;
593
594 if (unlikely(size == 0))
595 return 0;
596
597 if (unlikely(uaddr > end))
598 return -EFAULT;
599
600 do {
601 if (unlikely(__get_user(c, uaddr) != 0))
602 return -EFAULT;
603 uaddr += PAGE_SIZE;
604 } while (uaddr <= end);
605
606 /* Check whether the range spilled into the next page. */
607 if (((unsigned long)uaddr & PAGE_MASK) ==
608 ((unsigned long)end & PAGE_MASK)) {
609 return __get_user(c, end);
610 }
611
612 (void)c;
613 return 0;
614 }
615
616 int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
617 pgoff_t index, gfp_t gfp_mask);
618 int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
619 pgoff_t index, gfp_t gfp_mask);
620 extern void delete_from_page_cache(struct page *page);
621 extern void __delete_from_page_cache(struct page *page, void *shadow);
622 int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
623 void delete_from_page_cache_batch(struct address_space *mapping,
624 struct pagevec *pvec);
625
626 /*
627 * Like add_to_page_cache_locked, but used to add newly allocated pages:
628 * the page is new, so we can just run __SetPageLocked() against it.
629 */
630 static inline int add_to_page_cache(struct page *page,
631 struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
632 {
633 int error;
634
635 __SetPageLocked(page);
636 error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
637 if (unlikely(error))
638 __ClearPageLocked(page);
639 return error;
640 }
641
642 static inline unsigned long dir_pages(struct inode *inode)
643 {
644 return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
645 PAGE_SHIFT;
646 }
647
648 #endif /* _LINUX_PAGEMAP_H */