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
 19struct pagevec;
 20
 21/*
 22 * Bits in mapping->flags.
 
 23 */
 24enum 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 */
 48static 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 superblock */
 57	if (mapping->host)
 58		errseq_set(&mapping->host->i_sb->s_wb_err, error);
 59
 60	/* Record it in flags for now, for legacy callers */
 61	if (error == -ENOSPC)
 62		set_bit(AS_ENOSPC, &mapping->flags);
 63	else
 64		set_bit(AS_EIO, &mapping->flags);
 65}
 66
 67static inline void mapping_set_unevictable(struct address_space *mapping)
 68{
 69	set_bit(AS_UNEVICTABLE, &mapping->flags);
 70}
 71
 72static inline void mapping_clear_unevictable(struct address_space *mapping)
 73{
 74	clear_bit(AS_UNEVICTABLE, &mapping->flags);
 75}
 76
 77static inline bool mapping_unevictable(struct address_space *mapping)
 78{
 79	return mapping && test_bit(AS_UNEVICTABLE, &mapping->flags);
 80}
 81
 82static inline void mapping_set_exiting(struct address_space *mapping)
 83{
 84	set_bit(AS_EXITING, &mapping->flags);
 85}
 86
 87static inline int mapping_exiting(struct address_space *mapping)
 88{
 89	return test_bit(AS_EXITING, &mapping->flags);
 90}
 91
 92static inline void mapping_set_no_writeback_tags(struct address_space *mapping)
 93{
 94	set_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
 95}
 96
 97static inline int mapping_use_writeback_tags(struct address_space *mapping)
 98{
 99	return !test_bit(AS_NO_WRITEBACK_TAGS, &mapping->flags);
100}
101
102static inline gfp_t mapping_gfp_mask(struct address_space * mapping)
103{
104	return mapping->gfp_mask;
105}
106
107/* Restricts the given gfp_mask to what the mapping allows. */
108static inline gfp_t mapping_gfp_constraint(struct address_space *mapping,
109		gfp_t gfp_mask)
110{
111	return mapping_gfp_mask(mapping) & gfp_mask;
112}
113
114/*
115 * This is non-atomic.  Only to be used before the mapping is activated.
116 * Probably needs a barrier...
117 */
118static inline void mapping_set_gfp_mask(struct address_space *m, gfp_t mask)
119{
120	m->gfp_mask = mask;
 
121}
122
123void release_pages(struct page **pages, int nr);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
124
125/*
126 * speculatively take a reference to a page.
127 * If the page is free (_refcount == 0), then _refcount is untouched, and 0
128 * is returned. Otherwise, _refcount is incremented by 1 and 1 is returned.
129 *
130 * This function must be called inside the same rcu_read_lock() section as has
131 * been used to lookup the page in the pagecache radix-tree (or page table):
132 * this allows allocators to use a synchronize_rcu() to stabilize _refcount.
133 *
134 * Unless an RCU grace period has passed, the count of all pages coming out
135 * of the allocator must be considered unstable. page_count may return higher
136 * than expected, and put_page must be able to do the right thing when the
137 * page has been finished with, no matter what it is subsequently allocated
138 * for (because put_page is what is used here to drop an invalid speculative
139 * reference).
140 *
141 * This is the interesting part of the lockless pagecache (and lockless
142 * get_user_pages) locking protocol, where the lookup-side (eg. find_get_page)
143 * has the following pattern:
144 * 1. find page in radix tree
145 * 2. conditionally increment refcount
146 * 3. check the page is still in pagecache (if no, goto 1)
147 *
148 * Remove-side that cares about stability of _refcount (eg. reclaim) has the
149 * following (with the i_pages lock held):
150 * A. atomically check refcount is correct and set it to 0 (atomic_cmpxchg)
151 * B. remove page from pagecache
152 * C. free the page
153 *
154 * There are 2 critical interleavings that matter:
155 * - 2 runs before A: in this case, A sees elevated refcount and bails out
156 * - A runs before 2: in this case, 2 sees zero refcount and retries;
157 *   subsequently, B will complete and 1 will find no page, causing the
158 *   lookup to return NULL.
159 *
160 * It is possible that between 1 and 2, the page is removed then the exact same
161 * page is inserted into the same position in pagecache. That's OK: the
162 * old find_get_page using a lock could equally have run before or after
163 * such a re-insertion, depending on order that locks are granted.
164 *
165 * Lookups racing against pagecache insertion isn't a big problem: either 1
166 * will find the page or it will not. Likewise, the old find_get_page could run
167 * either before the insertion or afterwards, depending on timing.
168 */
169static inline int __page_cache_add_speculative(struct page *page, int count)
170{
171#ifdef CONFIG_TINY_RCU
 
 
172# ifdef CONFIG_PREEMPT_COUNT
173	VM_BUG_ON(!in_atomic() && !irqs_disabled());
174# endif
175	/*
176	 * Preempt must be disabled here - we rely on rcu_read_lock doing
177	 * this for us.
178	 *
179	 * Pagecache won't be truncated from interrupt context, so if we have
180	 * found a page in the radix tree here, we have pinned its refcount by
181	 * disabling preempt, and hence no need for the "speculative get" that
182	 * SMP requires.
183	 */
184	VM_BUG_ON_PAGE(page_count(page) == 0, page);
185	page_ref_add(page, count);
186
187#else
188	if (unlikely(!page_ref_add_unless(page, count, 0))) {
189		/*
190		 * Either the page has been freed, or will be freed.
191		 * In either case, retry here and the caller should
192		 * do the right thing (see comments above).
193		 */
194		return 0;
195	}
196#endif
197	VM_BUG_ON_PAGE(PageTail(page), page);
198
199	return 1;
200}
201
202static inline int page_cache_get_speculative(struct page *page)
203{
204	return __page_cache_add_speculative(page, 1);
205}
206
207static inline int page_cache_add_speculative(struct page *page, int count)
208{
209	return __page_cache_add_speculative(page, count);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
210}
211
212/**
213 * attach_page_private - Attach private data to a page.
214 * @page: Page to attach data to.
215 * @data: Data to attach to page.
216 *
217 * Attaching private data to a page increments the page's reference count.
218 * The data must be detached before the page will be freed.
219 */
220static inline void attach_page_private(struct page *page, void *data)
221{
222	get_page(page);
223	set_page_private(page, (unsigned long)data);
224	SetPagePrivate(page);
225}
226
227/**
228 * detach_page_private - Detach private data from a page.
229 * @page: Page to detach data from.
230 *
231 * Removes the data that was previously attached to the page and decrements
232 * the refcount on the page.
233 *
234 * Return: Data that was attached to the page.
235 */
236static inline void *detach_page_private(struct page *page)
237{
238	void *data = (void *)page_private(page);
 
239
240	if (!PagePrivate(page))
241		return NULL;
242	ClearPagePrivate(page);
243	set_page_private(page, 0);
244	put_page(page);
245
246	return data;
247}
248
249#ifdef CONFIG_NUMA
250extern struct page *__page_cache_alloc(gfp_t gfp);
251#else
252static inline struct page *__page_cache_alloc(gfp_t gfp)
253{
254	return alloc_pages(gfp, 0);
255}
256#endif
257
258static inline struct page *page_cache_alloc(struct address_space *x)
259{
260	return __page_cache_alloc(mapping_gfp_mask(x));
261}
262
263static inline gfp_t readahead_gfp_mask(struct address_space *x)
264{
265	return mapping_gfp_mask(x) | __GFP_NORETRY | __GFP_NOWARN;
266}
267
268typedef int filler_t(void *, struct page *);
269
270pgoff_t page_cache_next_miss(struct address_space *mapping,
271			     pgoff_t index, unsigned long max_scan);
272pgoff_t page_cache_prev_miss(struct address_space *mapping,
273			     pgoff_t index, unsigned long max_scan);
274
275#define FGP_ACCESSED		0x00000001
276#define FGP_LOCK		0x00000002
277#define FGP_CREAT		0x00000004
278#define FGP_WRITE		0x00000008
279#define FGP_NOFS		0x00000010
280#define FGP_NOWAIT		0x00000020
281#define FGP_FOR_MMAP		0x00000040
282
283struct page *pagecache_get_page(struct address_space *mapping, pgoff_t offset,
284		int fgp_flags, gfp_t cache_gfp_mask);
285
286/**
287 * find_get_page - find and get a page reference
288 * @mapping: the address_space to search
289 * @offset: the page index
290 *
291 * Looks up the page cache slot at @mapping & @offset.  If there is a
292 * page cache page, it is returned with an increased refcount.
293 *
294 * Otherwise, %NULL is returned.
295 */
296static inline struct page *find_get_page(struct address_space *mapping,
297					pgoff_t offset)
298{
299	return pagecache_get_page(mapping, offset, 0, 0);
300}
301
302static inline struct page *find_get_page_flags(struct address_space *mapping,
303					pgoff_t offset, int fgp_flags)
304{
305	return pagecache_get_page(mapping, offset, fgp_flags, 0);
 
306}
307
308/**
309 * find_lock_page - locate, pin and lock a pagecache page
310 * @mapping: the address_space to search
311 * @offset: the page index
312 *
313 * Looks up the page cache slot at @mapping & @offset.  If there is a
314 * page cache page, it is returned locked and with an increased
315 * refcount.
316 *
317 * Otherwise, %NULL is returned.
318 *
319 * find_lock_page() may sleep.
320 */
321static inline struct page *find_lock_page(struct address_space *mapping,
322					pgoff_t offset)
323{
324	return pagecache_get_page(mapping, offset, FGP_LOCK, 0);
325}
326
327/**
328 * find_or_create_page - locate or add a pagecache page
329 * @mapping: the page's address_space
330 * @index: the page's index into the mapping
331 * @gfp_mask: page allocation mode
332 *
333 * Looks up the page cache slot at @mapping & @offset.  If there is a
334 * page cache page, it is returned locked and with an increased
335 * refcount.
336 *
337 * If the page is not present, a new page is allocated using @gfp_mask
338 * and added to the page cache and the VM's LRU list.  The page is
339 * returned locked and with an increased refcount.
340 *
341 * On memory exhaustion, %NULL is returned.
342 *
343 * find_or_create_page() may sleep, even if @gfp_flags specifies an
344 * atomic allocation!
345 */
346static inline struct page *find_or_create_page(struct address_space *mapping,
347					pgoff_t index, gfp_t gfp_mask)
348{
349	return pagecache_get_page(mapping, index,
350					FGP_LOCK|FGP_ACCESSED|FGP_CREAT,
351					gfp_mask);
352}
353
354/**
355 * grab_cache_page_nowait - returns locked page at given index in given cache
356 * @mapping: target address_space
357 * @index: the page index
358 *
359 * Same as grab_cache_page(), but do not wait if the page is unavailable.
360 * This is intended for speculative data generators, where the data can
361 * be regenerated if the page couldn't be grabbed.  This routine should
362 * be safe to call while holding the lock for another page.
363 *
364 * Clear __GFP_FS when allocating the page to avoid recursion into the fs
365 * and deadlock against the caller's locked page.
366 */
367static inline struct page *grab_cache_page_nowait(struct address_space *mapping,
368				pgoff_t index)
369{
370	return pagecache_get_page(mapping, index,
371			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
372			mapping_gfp_mask(mapping));
373}
374
375/*
376 * Given the page we found in the page cache, return the page corresponding
377 * to this index in the file
378 */
379static inline struct page *find_subpage(struct page *head, pgoff_t index)
380{
381	/* HugeTLBfs wants the head page regardless */
382	if (PageHuge(head))
383		return head;
384
385	return head + (index & (thp_nr_pages(head) - 1));
386}
387
388struct page *find_get_entry(struct address_space *mapping, pgoff_t offset);
389struct page *find_lock_entry(struct address_space *mapping, pgoff_t offset);
390unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
391			  unsigned int nr_entries, struct page **entries,
392			  pgoff_t *indices);
393unsigned find_get_pages_range(struct address_space *mapping, pgoff_t *start,
394			pgoff_t end, unsigned int nr_pages,
395			struct page **pages);
396static inline unsigned find_get_pages(struct address_space *mapping,
397			pgoff_t *start, unsigned int nr_pages,
398			struct page **pages)
399{
400	return find_get_pages_range(mapping, start, (pgoff_t)-1, nr_pages,
401				    pages);
402}
403unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t start,
404			       unsigned int nr_pages, struct page **pages);
405unsigned find_get_pages_range_tag(struct address_space *mapping, pgoff_t *index,
406			pgoff_t end, xa_mark_t tag, unsigned int nr_pages,
407			struct page **pages);
408static inline unsigned find_get_pages_tag(struct address_space *mapping,
409			pgoff_t *index, xa_mark_t tag, unsigned int nr_pages,
410			struct page **pages)
411{
412	return find_get_pages_range_tag(mapping, index, (pgoff_t)-1, tag,
413					nr_pages, pages);
414}
415
416struct page *grab_cache_page_write_begin(struct address_space *mapping,
417			pgoff_t index, unsigned flags);
418
419/*
420 * Returns locked page at given index in given cache, creating it if needed.
421 */
422static inline struct page *grab_cache_page(struct address_space *mapping,
423								pgoff_t index)
424{
425	return find_or_create_page(mapping, index, mapping_gfp_mask(mapping));
426}
427
 
 
 
 
428extern struct page * read_cache_page(struct address_space *mapping,
429				pgoff_t index, filler_t *filler, void *data);
430extern struct page * read_cache_page_gfp(struct address_space *mapping,
431				pgoff_t index, gfp_t gfp_mask);
432extern int read_cache_pages(struct address_space *mapping,
433		struct list_head *pages, filler_t *filler, void *data);
434
435static inline struct page *read_mapping_page(struct address_space *mapping,
 
436				pgoff_t index, void *data)
437{
438	return read_cache_page(mapping, index, NULL, data);
439}
440
441/*
442 * Get index of the page with in radix-tree
443 * (TODO: remove once hugetlb pages will have ->index in PAGE_SIZE)
444 */
445static inline pgoff_t page_to_index(struct page *page)
446{
447	pgoff_t pgoff;
448
449	if (likely(!PageTransTail(page)))
450		return page->index;
451
452	/*
453	 *  We don't initialize ->index for tail pages: calculate based on
454	 *  head page
455	 */
456	pgoff = compound_head(page)->index;
457	pgoff += page - compound_head(page);
458	return pgoff;
459}
460
461/*
462 * Get the offset in PAGE_SIZE.
463 * (TODO: hugepage should have ->index in PAGE_SIZE)
464 */
465static inline pgoff_t page_to_pgoff(struct page *page)
466{
467	if (unlikely(PageHeadHuge(page)))
468		return page->index << compound_order(page);
469
470	return page_to_index(page);
471}
472
473/*
474 * Return byte-offset into filesystem object for page.
475 */
476static inline loff_t page_offset(struct page *page)
477{
478	return ((loff_t)page->index) << PAGE_SHIFT;
479}
480
481static inline loff_t page_file_offset(struct page *page)
482{
483	return ((loff_t)page_index(page)) << PAGE_SHIFT;
484}
485
486extern pgoff_t linear_hugepage_index(struct vm_area_struct *vma,
487				     unsigned long address);
488
489static inline pgoff_t linear_page_index(struct vm_area_struct *vma,
490					unsigned long address)
491{
492	pgoff_t pgoff;
493	if (unlikely(is_vm_hugetlb_page(vma)))
494		return linear_hugepage_index(vma, address);
495	pgoff = (address - vma->vm_start) >> PAGE_SHIFT;
496	pgoff += vma->vm_pgoff;
497	return pgoff;
498}
499
500/* This has the same layout as wait_bit_key - see fs/cachefiles/rdwr.c */
501struct wait_page_key {
502	struct page *page;
503	int bit_nr;
504	int page_match;
505};
506
507struct wait_page_queue {
508	struct page *page;
509	int bit_nr;
510	wait_queue_entry_t wait;
511};
512
513static inline bool wake_page_match(struct wait_page_queue *wait_page,
514				  struct wait_page_key *key)
515{
516	if (wait_page->page != key->page)
517	       return false;
518	key->page_match = 1;
519
520	if (wait_page->bit_nr != key->bit_nr)
521		return false;
522
523	return true;
524}
525
526extern void __lock_page(struct page *page);
527extern int __lock_page_killable(struct page *page);
528extern int __lock_page_async(struct page *page, struct wait_page_queue *wait);
529extern int __lock_page_or_retry(struct page *page, struct mm_struct *mm,
530				unsigned int flags);
531extern void unlock_page(struct page *page);
532
533/*
534 * Return true if the page was successfully locked
535 */
 
 
 
 
 
 
 
536static inline int trylock_page(struct page *page)
537{
538	page = compound_head(page);
539	return (likely(!test_and_set_bit_lock(PG_locked, &page->flags)));
540}
541
542/*
543 * lock_page may only be called if we have the page's inode pinned.
544 */
545static inline void lock_page(struct page *page)
546{
547	might_sleep();
548	if (!trylock_page(page))
549		__lock_page(page);
550}
551
552/*
553 * lock_page_killable is like lock_page but can be interrupted by fatal
554 * signals.  It returns 0 if it locked the page and -EINTR if it was
555 * killed while waiting.
556 */
557static inline int lock_page_killable(struct page *page)
558{
559	might_sleep();
560	if (!trylock_page(page))
561		return __lock_page_killable(page);
562	return 0;
563}
564
565/*
566 * lock_page_async - Lock the page, unless this would block. If the page
567 * is already locked, then queue a callback when the page becomes unlocked.
568 * This callback can then retry the operation.
569 *
570 * Returns 0 if the page is locked successfully, or -EIOCBQUEUED if the page
571 * was already locked and the callback defined in 'wait' was queued.
572 */
573static inline int lock_page_async(struct page *page,
574				  struct wait_page_queue *wait)
575{
576	if (!trylock_page(page))
577		return __lock_page_async(page, wait);
578	return 0;
579}
580
581/*
582 * lock_page_or_retry - Lock the page, unless this would block and the
583 * caller indicated that it can handle a retry.
584 *
585 * Return value and mmap_lock implications depend on flags; see
586 * __lock_page_or_retry().
587 */
588static inline int lock_page_or_retry(struct page *page, struct mm_struct *mm,
589				     unsigned int flags)
590{
591	might_sleep();
592	return trylock_page(page) || __lock_page_or_retry(page, mm, flags);
593}
594
595/*
596 * This is exported only for wait_on_page_locked/wait_on_page_writeback, etc.,
597 * and should not be used directly.
598 */
599extern void wait_on_page_bit(struct page *page, int bit_nr);
 
600extern int wait_on_page_bit_killable(struct page *page, int bit_nr);
601
 
 
 
 
 
 
 
602/* 
603 * Wait for a page to be unlocked.
604 *
605 * This must be called with the caller "holding" the page,
606 * ie with increased "page->count" so that the page won't
607 * go away during the wait..
608 */
609static inline void wait_on_page_locked(struct page *page)
610{
611	if (PageLocked(page))
612		wait_on_page_bit(compound_head(page), PG_locked);
613}
614
615static inline int wait_on_page_locked_killable(struct page *page)
 
 
 
616{
617	if (!PageLocked(page))
618		return 0;
619	return wait_on_page_bit_killable(compound_head(page), PG_locked);
620}
621
622extern void put_and_wait_on_page_locked(struct page *page);
623
624void wait_on_page_writeback(struct page *page);
625extern void end_page_writeback(struct page *page);
626void wait_for_stable_page(struct page *page);
627
628void page_endio(struct page *page, bool is_write, int err);
629
630/*
631 * Add an arbitrary waiter to a page's wait queue
632 */
633extern void add_page_wait_queue(struct page *page, wait_queue_entry_t *waiter);
634
635/*
636 * Fault everything in given userspace address range in.
 
 
 
637 */
638static inline int fault_in_pages_writeable(char __user *uaddr, int size)
639{
640	char __user *end = uaddr + size - 1;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
641
642	if (unlikely(size == 0))
643		return 0;
644
645	if (unlikely(uaddr > end))
646		return -EFAULT;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
647	/*
648	 * Writing zeroes into userspace here is OK, because we know that if
649	 * the zero gets there, we'll be overwriting it.
650	 */
651	do {
652		if (unlikely(__put_user(0, uaddr) != 0))
653			return -EFAULT;
 
654		uaddr += PAGE_SIZE;
655	} while (uaddr <= end);
656
657	/* Check whether the range spilled into the next page. */
658	if (((unsigned long)uaddr & PAGE_MASK) ==
659			((unsigned long)end & PAGE_MASK))
660		return __put_user(0, end);
661
662	return 0;
663}
664
665static inline int fault_in_pages_readable(const char __user *uaddr, int size)
 
666{
667	volatile char c;
 
668	const char __user *end = uaddr + size - 1;
669
670	if (unlikely(size == 0))
671		return 0;
672
673	if (unlikely(uaddr > end))
674		return -EFAULT;
675
676	do {
677		if (unlikely(__get_user(c, uaddr) != 0))
678			return -EFAULT;
 
679		uaddr += PAGE_SIZE;
680	} while (uaddr <= end);
681
682	/* Check whether the range spilled into the next page. */
683	if (((unsigned long)uaddr & PAGE_MASK) ==
684			((unsigned long)end & PAGE_MASK)) {
685		return __get_user(c, end);
 
686	}
687
688	(void)c;
689	return 0;
690}
691
692int add_to_page_cache_locked(struct page *page, struct address_space *mapping,
693				pgoff_t index, gfp_t gfp_mask);
694int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
695				pgoff_t index, gfp_t gfp_mask);
696extern void delete_from_page_cache(struct page *page);
697extern void __delete_from_page_cache(struct page *page, void *shadow);
698int replace_page_cache_page(struct page *old, struct page *new, gfp_t gfp_mask);
699void delete_from_page_cache_batch(struct address_space *mapping,
700				  struct pagevec *pvec);
701
702#define VM_READAHEAD_PAGES	(SZ_128K / PAGE_SIZE)
703
704void page_cache_sync_readahead(struct address_space *, struct file_ra_state *,
705		struct file *, pgoff_t index, unsigned long req_count);
706void page_cache_async_readahead(struct address_space *, struct file_ra_state *,
707		struct file *, struct page *, pgoff_t index,
708		unsigned long req_count);
709void page_cache_readahead_unbounded(struct address_space *, struct file *,
710		pgoff_t index, unsigned long nr_to_read,
711		unsigned long lookahead_count);
712
713/*
714 * Like add_to_page_cache_locked, but used to add newly allocated pages:
715 * the page is new, so we can just run __SetPageLocked() against it.
716 */
717static inline int add_to_page_cache(struct page *page,
718		struct address_space *mapping, pgoff_t offset, gfp_t gfp_mask)
719{
720	int error;
721
722	__SetPageLocked(page);
723	error = add_to_page_cache_locked(page, mapping, offset, gfp_mask);
724	if (unlikely(error))
725		__ClearPageLocked(page);
726	return error;
727}
728
729/**
730 * struct readahead_control - Describes a readahead request.
731 *
732 * A readahead request is for consecutive pages.  Filesystems which
733 * implement the ->readahead method should call readahead_page() or
734 * readahead_page_batch() in a loop and attempt to start I/O against
735 * each page in the request.
736 *
737 * Most of the fields in this struct are private and should be accessed
738 * by the functions below.
739 *
740 * @file: The file, used primarily by network filesystems for authentication.
741 *	  May be NULL if invoked internally by the filesystem.
742 * @mapping: Readahead this filesystem object.
743 */
744struct readahead_control {
745	struct file *file;
746	struct address_space *mapping;
747/* private: use the readahead_* accessors instead */
748	pgoff_t _index;
749	unsigned int _nr_pages;
750	unsigned int _batch_count;
751};
752
753/**
754 * readahead_page - Get the next page to read.
755 * @rac: The current readahead request.
756 *
757 * Context: The page is locked and has an elevated refcount.  The caller
758 * should decreases the refcount once the page has been submitted for I/O
759 * and unlock the page once all I/O to that page has completed.
760 * Return: A pointer to the next page, or %NULL if we are done.
761 */
762static inline struct page *readahead_page(struct readahead_control *rac)
763{
764	struct page *page;
765
766	BUG_ON(rac->_batch_count > rac->_nr_pages);
767	rac->_nr_pages -= rac->_batch_count;
768	rac->_index += rac->_batch_count;
769
770	if (!rac->_nr_pages) {
771		rac->_batch_count = 0;
772		return NULL;
773	}
774
775	page = xa_load(&rac->mapping->i_pages, rac->_index);
776	VM_BUG_ON_PAGE(!PageLocked(page), page);
777	rac->_batch_count = thp_nr_pages(page);
778
779	return page;
780}
781
782static inline unsigned int __readahead_batch(struct readahead_control *rac,
783		struct page **array, unsigned int array_sz)
784{
785	unsigned int i = 0;
786	XA_STATE(xas, &rac->mapping->i_pages, 0);
787	struct page *page;
788
789	BUG_ON(rac->_batch_count > rac->_nr_pages);
790	rac->_nr_pages -= rac->_batch_count;
791	rac->_index += rac->_batch_count;
792	rac->_batch_count = 0;
793
794	xas_set(&xas, rac->_index);
795	rcu_read_lock();
796	xas_for_each(&xas, page, rac->_index + rac->_nr_pages - 1) {
797		VM_BUG_ON_PAGE(!PageLocked(page), page);
798		VM_BUG_ON_PAGE(PageTail(page), page);
799		array[i++] = page;
800		rac->_batch_count += thp_nr_pages(page);
801
802		/*
803		 * The page cache isn't using multi-index entries yet,
804		 * so the xas cursor needs to be manually moved to the
805		 * next index.  This can be removed once the page cache
806		 * is converted.
807		 */
808		if (PageHead(page))
809			xas_set(&xas, rac->_index + rac->_batch_count);
810
811		if (i == array_sz)
812			break;
813	}
814	rcu_read_unlock();
815
816	return i;
817}
818
819/**
820 * readahead_page_batch - Get a batch of pages to read.
821 * @rac: The current readahead request.
822 * @array: An array of pointers to struct page.
823 *
824 * Context: The pages are locked and have an elevated refcount.  The caller
825 * should decreases the refcount once the page has been submitted for I/O
826 * and unlock the page once all I/O to that page has completed.
827 * Return: The number of pages placed in the array.  0 indicates the request
828 * is complete.
829 */
830#define readahead_page_batch(rac, array)				\
831	__readahead_batch(rac, array, ARRAY_SIZE(array))
832
833/**
834 * readahead_pos - The byte offset into the file of this readahead request.
835 * @rac: The readahead request.
836 */
837static inline loff_t readahead_pos(struct readahead_control *rac)
838{
839	return (loff_t)rac->_index * PAGE_SIZE;
840}
841
842/**
843 * readahead_length - The number of bytes in this readahead request.
844 * @rac: The readahead request.
845 */
846static inline loff_t readahead_length(struct readahead_control *rac)
847{
848	return (loff_t)rac->_nr_pages * PAGE_SIZE;
849}
850
851/**
852 * readahead_index - The index of the first page in this readahead request.
853 * @rac: The readahead request.
854 */
855static inline pgoff_t readahead_index(struct readahead_control *rac)
856{
857	return rac->_index;
858}
859
860/**
861 * readahead_count - The number of pages in this readahead request.
862 * @rac: The readahead request.
863 */
864static inline unsigned int readahead_count(struct readahead_control *rac)
865{
866	return rac->_nr_pages;
867}
868
869static inline unsigned long dir_pages(struct inode *inode)
870{
871	return (unsigned long)(inode->i_size + PAGE_SIZE - 1) >>
872			       PAGE_SHIFT;
873}
874
875/**
876 * page_mkwrite_check_truncate - check if page was truncated
877 * @page: the page to check
878 * @inode: the inode to check the page against
879 *
880 * Returns the number of bytes in the page up to EOF,
881 * or -EFAULT if the page was truncated.
882 */
883static inline int page_mkwrite_check_truncate(struct page *page,
884					      struct inode *inode)
885{
886	loff_t size = i_size_read(inode);
887	pgoff_t index = size >> PAGE_SHIFT;
888	int offset = offset_in_page(size);
889
890	if (page->mapping != inode->i_mapping)
891		return -EFAULT;
892
893	/* page is wholly inside EOF */
894	if (page->index < index)
895		return PAGE_SIZE;
896	/* page is wholly past EOF */
897	if (page->index > index || !offset)
898		return -EFAULT;
899	/* page is partially inside EOF */
900	return offset;
901}
902
903#endif /* _LINUX_PAGEMAP_H */