Lines Matching +full:data +full:- +full:mapping
1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1994-1999 Linus Torvalds
30 #include <linux/error-injection.h>
33 #include <linux/backing-dev.h>
70 * finished 'unifying' the page and buffer cache and SMP-threaded the
71 * page-cache, 21.05.1999, Ingo Molnar <mingo@redhat.com>
73 * SMP-threaded pagemap-LRU 1999, Andrea Arcangeli <andrea@suse.de>
79 * ->i_mmap_rwsem (truncate_pagecache)
80 * ->private_lock (__free_pte->block_dirty_folio)
81 * ->swap_lock (exclusive_swap_page, others)
82 * ->i_pages lock
84 * ->i_rwsem
85 * ->invalidate_lock (acquired by fs in truncate path)
86 * ->i_mmap_rwsem (truncate->unmap_mapping_range)
88 * ->mmap_lock
89 * ->i_mmap_rwsem
90 * ->page_table_lock or pte_lock (various, mainly in memory.c)
91 * ->i_pages lock (arch-dependent flush_dcache_mmap_lock)
93 * ->mmap_lock
94 * ->invalidate_lock (filemap_fault)
95 * ->lock_page (filemap_fault, access_process_vm)
97 * ->i_rwsem (generic_perform_write)
98 * ->mmap_lock (fault_in_readable->do_page_fault)
100 * bdi->wb.list_lock
101 * sb_lock (fs/fs-writeback.c)
102 * ->i_pages lock (__sync_single_inode)
104 * ->i_mmap_rwsem
105 * ->anon_vma.lock (vma_merge)
107 * ->anon_vma.lock
108 * ->page_table_lock or pte_lock (anon_vma_prepare and various)
110 * ->page_table_lock or pte_lock
111 * ->swap_lock (try_to_unmap_one)
112 * ->private_lock (try_to_unmap_one)
113 * ->i_pages lock (try_to_unmap_one)
114 * ->lruvec->lru_lock (follow_page->mark_page_accessed)
115 * ->lruvec->lru_lock (check_pte_range->isolate_lru_page)
116 * ->private_lock (page_remove_rmap->set_page_dirty)
117 * ->i_pages lock (page_remove_rmap->set_page_dirty)
118 * bdi.wb->list_lock (page_remove_rmap->set_page_dirty)
119 * ->inode->i_lock (page_remove_rmap->set_page_dirty)
120 * ->memcg->move_lock (page_remove_rmap->folio_memcg_lock)
121 * bdi.wb->list_lock (zap_pte_range->set_page_dirty)
122 * ->inode->i_lock (zap_pte_range->set_page_dirty)
123 * ->private_lock (zap_pte_range->block_dirty_folio)
126 static void page_cache_delete(struct address_space *mapping, in page_cache_delete() argument
129 XA_STATE(xas, &mapping->i_pages, folio->index); in page_cache_delete()
132 mapping_set_update(&xas, mapping); in page_cache_delete()
136 xas_set_order(&xas, folio->index, folio_order(folio)); in page_cache_delete()
145 folio->mapping = NULL; in page_cache_delete()
146 /* Leave page->index set: truncation lookup relies upon it */ in page_cache_delete()
147 mapping->nrpages -= nr; in page_cache_delete()
150 static void filemap_unaccount_folio(struct address_space *mapping, in filemap_unaccount_folio() argument
158 current->comm, folio_pfn(folio)); in filemap_unaccount_folio()
159 dump_page(&folio->page, "still mapped when deleted"); in filemap_unaccount_folio()
163 if (mapping_exiting(mapping) && !folio_test_large(folio)) { in filemap_unaccount_folio()
164 int mapcount = page_mapcount(&folio->page); in filemap_unaccount_folio()
173 page_mapcount_reset(&folio->page); in filemap_unaccount_folio()
185 __lruvec_stat_mod_folio(folio, NR_FILE_PAGES, -nr); in filemap_unaccount_folio()
187 __lruvec_stat_mod_folio(folio, NR_SHMEM, -nr); in filemap_unaccount_folio()
189 __lruvec_stat_mod_folio(folio, NR_SHMEM_THPS, -nr); in filemap_unaccount_folio()
191 __lruvec_stat_mod_folio(folio, NR_FILE_THPS, -nr); in filemap_unaccount_folio()
192 filemap_nr_thps_dec(mapping); in filemap_unaccount_folio()
198 * unwritten data - on ordinary filesystems. in filemap_unaccount_folio()
200 * But it's harmless on in-memory filesystems like tmpfs; and can in filemap_unaccount_folio()
210 mapping_can_writeback(mapping))) in filemap_unaccount_folio()
211 folio_account_cleaned(folio, inode_to_wb(mapping->host)); in filemap_unaccount_folio()
216 * sure the page is locked and that nobody else uses it - or that usage
221 struct address_space *mapping = folio->mapping; in __filemap_remove_folio() local
224 filemap_unaccount_folio(mapping, folio); in __filemap_remove_folio()
225 page_cache_delete(mapping, folio, shadow); in __filemap_remove_folio()
228 void filemap_free_folio(struct address_space *mapping, struct folio *folio) in filemap_free_folio() argument
233 free_folio = mapping->a_ops->free_folio; in filemap_free_folio()
243 * filemap_remove_folio - Remove folio from page cache.
252 struct address_space *mapping = folio->mapping; in filemap_remove_folio() local
255 spin_lock(&mapping->host->i_lock); in filemap_remove_folio()
256 xa_lock_irq(&mapping->i_pages); in filemap_remove_folio()
258 xa_unlock_irq(&mapping->i_pages); in filemap_remove_folio()
259 if (mapping_shrinkable(mapping)) in filemap_remove_folio()
260 inode_add_lru(mapping->host); in filemap_remove_folio()
261 spin_unlock(&mapping->host->i_lock); in filemap_remove_folio()
263 filemap_free_folio(mapping, folio); in filemap_remove_folio()
267 * page_cache_delete_batch - delete several folios from page cache
268 * @mapping: the mapping to which folios belong
271 * The function walks over mapping->i_pages and removes folios passed in
272 * @fbatch from the mapping. The function expects @fbatch to be sorted
274 * It tolerates holes in @fbatch (mapping entries at those indices are not
279 static void page_cache_delete_batch(struct address_space *mapping, in page_cache_delete_batch() argument
282 XA_STATE(xas, &mapping->i_pages, fbatch->folios[0]->index); in page_cache_delete_batch()
287 mapping_set_update(&xas, mapping); in page_cache_delete_batch()
302 if (folio != fbatch->folios[i]) { in page_cache_delete_batch()
303 VM_BUG_ON_FOLIO(folio->index > in page_cache_delete_batch()
304 fbatch->folios[i]->index, folio); in page_cache_delete_batch()
310 folio->mapping = NULL; in page_cache_delete_batch()
311 /* Leave folio->index set: truncation lookup relies on it */ in page_cache_delete_batch()
317 mapping->nrpages -= total_pages; in page_cache_delete_batch()
320 void delete_from_page_cache_batch(struct address_space *mapping, in delete_from_page_cache_batch() argument
328 spin_lock(&mapping->host->i_lock); in delete_from_page_cache_batch()
329 xa_lock_irq(&mapping->i_pages); in delete_from_page_cache_batch()
331 struct folio *folio = fbatch->folios[i]; in delete_from_page_cache_batch()
334 filemap_unaccount_folio(mapping, folio); in delete_from_page_cache_batch()
336 page_cache_delete_batch(mapping, fbatch); in delete_from_page_cache_batch()
337 xa_unlock_irq(&mapping->i_pages); in delete_from_page_cache_batch()
338 if (mapping_shrinkable(mapping)) in delete_from_page_cache_batch()
339 inode_add_lru(mapping->host); in delete_from_page_cache_batch()
340 spin_unlock(&mapping->host->i_lock); in delete_from_page_cache_batch()
343 filemap_free_folio(mapping, fbatch->folios[i]); in delete_from_page_cache_batch()
346 int filemap_check_errors(struct address_space *mapping) in filemap_check_errors() argument
350 if (test_bit(AS_ENOSPC, &mapping->flags) && in filemap_check_errors()
351 test_and_clear_bit(AS_ENOSPC, &mapping->flags)) in filemap_check_errors()
352 ret = -ENOSPC; in filemap_check_errors()
353 if (test_bit(AS_EIO, &mapping->flags) && in filemap_check_errors()
354 test_and_clear_bit(AS_EIO, &mapping->flags)) in filemap_check_errors()
355 ret = -EIO; in filemap_check_errors()
360 static int filemap_check_and_keep_errors(struct address_space *mapping) in filemap_check_and_keep_errors() argument
363 if (test_bit(AS_EIO, &mapping->flags)) in filemap_check_and_keep_errors()
364 return -EIO; in filemap_check_and_keep_errors()
365 if (test_bit(AS_ENOSPC, &mapping->flags)) in filemap_check_and_keep_errors()
366 return -ENOSPC; in filemap_check_and_keep_errors()
371 * filemap_fdatawrite_wbc - start writeback on mapping dirty pages in range
372 * @mapping: address space structure to write
375 * Call writepages on the mapping using the provided wbc to control the
380 int filemap_fdatawrite_wbc(struct address_space *mapping, in filemap_fdatawrite_wbc() argument
385 if (!mapping_can_writeback(mapping) || in filemap_fdatawrite_wbc()
386 !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) in filemap_fdatawrite_wbc()
389 wbc_attach_fdatawrite_inode(wbc, mapping->host); in filemap_fdatawrite_wbc()
390 ret = do_writepages(mapping, wbc); in filemap_fdatawrite_wbc()
397 * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range
398 * @mapping: address space structure to write
403 * Start writeback against all of a mapping's dirty pages that lie
406 * If sync_mode is WB_SYNC_ALL then this is a "data integrity" operation, as
413 int __filemap_fdatawrite_range(struct address_space *mapping, loff_t start, in __filemap_fdatawrite_range() argument
423 return filemap_fdatawrite_wbc(mapping, &wbc); in __filemap_fdatawrite_range()
426 static inline int __filemap_fdatawrite(struct address_space *mapping, in __filemap_fdatawrite() argument
429 return __filemap_fdatawrite_range(mapping, 0, LLONG_MAX, sync_mode); in __filemap_fdatawrite()
432 int filemap_fdatawrite(struct address_space *mapping) in filemap_fdatawrite() argument
434 return __filemap_fdatawrite(mapping, WB_SYNC_ALL); in filemap_fdatawrite()
438 int filemap_fdatawrite_range(struct address_space *mapping, loff_t start, in filemap_fdatawrite_range() argument
441 return __filemap_fdatawrite_range(mapping, start, end, WB_SYNC_ALL); in filemap_fdatawrite_range()
446 * filemap_flush - mostly a non-blocking flush
447 * @mapping: target address_space
449 * This is a mostly non-blocking flush. Not suitable for data-integrity
450 * purposes - I/O may not be started against all dirty pages.
454 int filemap_flush(struct address_space *mapping) in filemap_flush() argument
456 return __filemap_fdatawrite(mapping, WB_SYNC_NONE); in filemap_flush()
461 * filemap_range_has_page - check if a page exists in range.
462 * @mapping: address space within which to check
472 bool filemap_range_has_page(struct address_space *mapping, in filemap_range_has_page() argument
476 XA_STATE(xas, &mapping->i_pages, start_byte >> PAGE_SHIFT); in filemap_range_has_page()
503 static void __filemap_fdatawait_range(struct address_space *mapping, in __filemap_fdatawait_range() argument
516 nr_folios = filemap_get_folios_tag(mapping, &index, end, in __filemap_fdatawait_range()
534 * filemap_fdatawait_range - wait for writeback to complete
535 * @mapping: address space structure to wait for
539 * Walk the list of under-writeback pages of the given address space
549 int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte, in filemap_fdatawait_range() argument
552 __filemap_fdatawait_range(mapping, start_byte, end_byte); in filemap_fdatawait_range()
553 return filemap_check_errors(mapping); in filemap_fdatawait_range()
558 * filemap_fdatawait_range_keep_errors - wait for writeback to complete
559 * @mapping: address space structure to wait for
563 * Walk the list of under-writeback pages of the given address space in the
568 * call sites are system-wide / filesystem-wide data flushers: e.g. sync(2),
571 int filemap_fdatawait_range_keep_errors(struct address_space *mapping, in filemap_fdatawait_range_keep_errors() argument
574 __filemap_fdatawait_range(mapping, start_byte, end_byte); in filemap_fdatawait_range_keep_errors()
575 return filemap_check_and_keep_errors(mapping); in filemap_fdatawait_range_keep_errors()
580 * file_fdatawait_range - wait for writeback to complete
585 * Walk the list of under-writeback pages of the address space that file
587 * status of the address space vs. the file->f_wb_err cursor and return it.
593 * Return: error status of the address space vs. the file->f_wb_err cursor.
597 struct address_space *mapping = file->f_mapping; in file_fdatawait_range() local
599 __filemap_fdatawait_range(mapping, start_byte, end_byte); in file_fdatawait_range()
605 * filemap_fdatawait_keep_errors - wait for writeback without clearing errors
606 * @mapping: address space structure to wait for
608 * Walk the list of under-writeback pages of the given address space
613 * call sites are system-wide / filesystem-wide data flushers: e.g. sync(2),
618 int filemap_fdatawait_keep_errors(struct address_space *mapping) in filemap_fdatawait_keep_errors() argument
620 __filemap_fdatawait_range(mapping, 0, LLONG_MAX); in filemap_fdatawait_keep_errors()
621 return filemap_check_and_keep_errors(mapping); in filemap_fdatawait_keep_errors()
626 static bool mapping_needs_writeback(struct address_space *mapping) in mapping_needs_writeback() argument
628 return mapping->nrpages; in mapping_needs_writeback()
631 bool filemap_range_has_writeback(struct address_space *mapping, in filemap_range_has_writeback() argument
634 XA_STATE(xas, &mapping->i_pages, start_byte >> PAGE_SHIFT); in filemap_range_has_writeback()
657 * filemap_write_and_wait_range - write out & wait on a file range
658 * @mapping: the address_space for the pages
662 * Write out and wait upon file offsets lstart->lend, inclusive.
665 * that this function can be used to write to the very end-of-file (end = -1).
669 int filemap_write_and_wait_range(struct address_space *mapping, in filemap_write_and_wait_range() argument
677 if (mapping_needs_writeback(mapping)) { in filemap_write_and_wait_range()
678 err = __filemap_fdatawrite_range(mapping, lstart, lend, in filemap_write_and_wait_range()
682 * written partially (e.g. -ENOSPC), so we wait for it. in filemap_write_and_wait_range()
683 * But the -EIO is special case, it may indicate the worst in filemap_write_and_wait_range()
686 if (err != -EIO) in filemap_write_and_wait_range()
687 __filemap_fdatawait_range(mapping, lstart, lend); in filemap_write_and_wait_range()
689 err2 = filemap_check_errors(mapping); in filemap_write_and_wait_range()
696 void __filemap_set_wb_err(struct address_space *mapping, int err) in __filemap_set_wb_err() argument
698 errseq_t eseq = errseq_set(&mapping->wb_err, err); in __filemap_set_wb_err()
700 trace_filemap_set_wb_err(mapping, eseq); in __filemap_set_wb_err()
705 * file_check_and_advance_wb_err - report wb error (if any) that was previously
713 * Grab the wb_err from the mapping. If it matches what we have in the file,
716 * If it doesn't match, then take the mapping value, set the "seen" flag in
722 * While we handle mapping->wb_err with atomic operations, the f_wb_err
731 errseq_t old = READ_ONCE(file->f_wb_err); in file_check_and_advance_wb_err()
732 struct address_space *mapping = file->f_mapping; in file_check_and_advance_wb_err() local
735 if (errseq_check(&mapping->wb_err, old)) { in file_check_and_advance_wb_err()
737 spin_lock(&file->f_lock); in file_check_and_advance_wb_err()
738 old = file->f_wb_err; in file_check_and_advance_wb_err()
739 err = errseq_check_and_advance(&mapping->wb_err, in file_check_and_advance_wb_err()
740 &file->f_wb_err); in file_check_and_advance_wb_err()
742 spin_unlock(&file->f_lock); in file_check_and_advance_wb_err()
750 clear_bit(AS_EIO, &mapping->flags); in file_check_and_advance_wb_err()
751 clear_bit(AS_ENOSPC, &mapping->flags); in file_check_and_advance_wb_err()
757 * file_write_and_wait_range - write out & wait on a file range
762 * Write out and wait upon file offsets lstart->lend, inclusive.
765 * that this function can be used to write to the very end-of-file (end = -1).
767 * After writing out and waiting on the data, we check and advance the
775 struct address_space *mapping = file->f_mapping; in file_write_and_wait_range() local
780 if (mapping_needs_writeback(mapping)) { in file_write_and_wait_range()
781 err = __filemap_fdatawrite_range(mapping, lstart, lend, in file_write_and_wait_range()
784 if (err != -EIO) in file_write_and_wait_range()
785 __filemap_fdatawait_range(mapping, lstart, lend); in file_write_and_wait_range()
795 * replace_page_cache_folio - replace a pagecache folio with a new one
809 struct address_space *mapping = old->mapping; in replace_page_cache_folio() local
810 void (*free_folio)(struct folio *) = mapping->a_ops->free_folio; in replace_page_cache_folio()
811 pgoff_t offset = old->index; in replace_page_cache_folio()
812 XA_STATE(xas, &mapping->i_pages, offset); in replace_page_cache_folio()
816 VM_BUG_ON_FOLIO(new->mapping, new); in replace_page_cache_folio()
819 new->mapping = mapping; in replace_page_cache_folio()
820 new->index = offset; in replace_page_cache_folio()
827 old->mapping = NULL; in replace_page_cache_folio()
844 noinline int __filemap_add_folio(struct address_space *mapping, in __filemap_add_folio() argument
847 XA_STATE(xas, &mapping->i_pages, index); in __filemap_add_folio()
854 mapping_set_update(&xas, mapping); in __filemap_add_folio()
858 VM_BUG_ON_FOLIO(index & (folio_nr_pages(folio) - 1), folio); in __filemap_add_folio()
868 folio->mapping = mapping; in __filemap_add_folio()
869 folio->index = xas.xa_index; in __filemap_add_folio()
882 xas_set_err(&xas, -EEXIST); in __filemap_add_folio()
894 BUG_ON(shmem_mapping(mapping)); in __filemap_add_folio()
904 mapping->nrpages += nr; in __filemap_add_folio()
925 folio->mapping = NULL; in __filemap_add_folio()
926 /* Leave page->index set: truncation relies upon it */ in __filemap_add_folio()
932 int filemap_add_folio(struct address_space *mapping, struct folio *folio, in filemap_add_folio() argument
939 ret = __filemap_add_folio(mapping, folio, index, gfp, &shadow); in filemap_add_folio()
948 * data from the working set, only to cache data that will in filemap_add_folio()
982 * filemap_invalidate_lock_two - lock invalidate_lock for two mappings
984 * Lock exclusively invalidate_lock of any passed mapping that is not NULL.
986 * @mapping1: the first mapping to lock
987 * @mapping2: the second mapping to lock
995 down_write(&mapping1->invalidate_lock); in filemap_invalidate_lock_two()
997 down_write_nested(&mapping2->invalidate_lock, 1); in filemap_invalidate_lock_two()
1002 * filemap_invalidate_unlock_two - unlock invalidate_lock for two mappings
1004 * Unlock exclusive invalidate_lock of any passed mapping that is not NULL.
1006 * @mapping1: the first mapping to unlock
1007 * @mapping2: the second mapping to unlock
1013 up_write(&mapping1->invalidate_lock); in filemap_invalidate_unlock_two()
1015 up_write(&mapping2->invalidate_lock); in filemap_invalidate_unlock_two()
1049 * The page wait code treats the "wait->flags" somewhat unusually, because
1096 flags = wait->flags; in wake_page_function()
1098 if (test_bit(key->bit_nr, &key->folio->flags)) in wake_page_function()
1099 return -1; in wake_page_function()
1101 if (test_and_set_bit(key->bit_nr, &key->folio->flags)) in wake_page_function()
1102 return -1; in wake_page_function()
1108 * We are holding the wait-queue lock, but the waiter that in wake_page_function()
1113 * afterwards to avoid any races. This store-release pairs in wake_page_function()
1114 * with the load-acquire in folio_wait_bit_common(). in wake_page_function()
1116 smp_store_release(&wait->flags, flags | WQ_FLAG_WOKEN); in wake_page_function()
1117 wake_up_state(wait->private, mode); in wake_page_function()
1125 * After this list_del_init(&wait->entry) the wait entry in wake_page_function()
1126 * might be de-allocated and the process might even have in wake_page_function()
1129 list_del_init_careful(&wait->entry); in wake_page_function()
1149 spin_lock_irqsave(&q->lock, flags); in folio_wake_bit()
1159 spin_unlock_irqrestore(&q->lock, flags); in folio_wake_bit()
1161 spin_lock_irqsave(&q->lock, flags); in folio_wake_bit()
1177 spin_unlock_irqrestore(&q->lock, flags); in folio_wake_bit()
1209 if (wait->flags & WQ_FLAG_EXCLUSIVE) { in folio_trylock_flag()
1210 if (test_and_set_bit(bit_nr, &folio->flags)) in folio_trylock_flag()
1212 } else if (test_bit(bit_nr, &folio->flags)) in folio_trylock_flag()
1215 wait->flags |= WQ_FLAG_WOKEN | WQ_FLAG_DONE; in folio_trylock_flag()
1241 wait->func = wake_page_function; in folio_wait_bit_common()
1246 wait->flags = 0; in folio_wait_bit_common()
1248 wait->flags = WQ_FLAG_EXCLUSIVE; in folio_wait_bit_common()
1249 if (--unfairness < 0) in folio_wait_bit_common()
1250 wait->flags |= WQ_FLAG_CUSTOM; in folio_wait_bit_common()
1267 spin_lock_irq(&q->lock); in folio_wait_bit_common()
1271 spin_unlock_irq(&q->lock); in folio_wait_bit_common()
1287 * be very careful with the 'wait->flags', because in folio_wait_bit_common()
1296 flags = smp_load_acquire(&wait->flags); in folio_wait_bit_common()
1305 /* If we were non-exclusive, we're done */ in folio_wait_bit_common()
1322 wait->flags |= WQ_FLAG_DONE; in folio_wait_bit_common()
1328 * waiter from the wait-queues, but the folio waiters bit will remain in folio_wait_bit_common()
1340 * NOTE! The wait->flags weren't stable until we've done the in folio_wait_bit_common()
1349 * Also note that WQ_FLAG_WOKEN is sufficient for a non-exclusive in folio_wait_bit_common()
1353 return wait->flags & WQ_FLAG_DONE ? 0 : -EINTR; in folio_wait_bit_common()
1355 return wait->flags & WQ_FLAG_WOKEN ? 0 : -EINTR; in folio_wait_bit_common()
1360 * migration_entry_wait_on_locked - Wait for a migration entry to be removed
1394 wait->func = wake_page_function; in migration_entry_wait_on_locked()
1397 wait->flags = 0; in migration_entry_wait_on_locked()
1399 spin_lock_irq(&q->lock); in migration_entry_wait_on_locked()
1403 spin_unlock_irq(&q->lock); in migration_entry_wait_on_locked()
1418 flags = smp_load_acquire(&wait->flags); in migration_entry_wait_on_locked()
1451 * folio_put_wait_locked - Drop a reference and wait for it to be unlocked
1461 * Return: 0 if the folio was unlocked or -EINTR if interrupted by a signal.
1469 * folio_add_wait_queue - Add an arbitrary waiter to a folio's wait queue
1480 spin_lock_irqsave(&q->lock, flags); in folio_add_wait_queue()
1483 spin_unlock_irqrestore(&q->lock, flags); in folio_add_wait_queue()
1511 * folio_unlock - Unlock a locked folio.
1531 * folio_end_private_2 - Clear PG_private_2 and wake any waiters.
1551 * folio_wait_private_2 - Wait for PG_private_2 to be cleared on a folio.
1564 * folio_wait_private_2_killable - Wait for PG_private_2 to be cleared on a folio.
1571 * - 0 if successful.
1572 * - -EINTR if a fatal signal was encountered.
1589 * folio_end_writeback - End writeback against a folio.
1624 * __folio_lock - Get a lock on the folio, assuming we need to sleep to get it.
1646 wait->folio = folio; in __folio_lock_async()
1647 wait->bit_nr = PG_locked; in __folio_lock_async()
1649 spin_lock_irq(&q->lock); in __folio_lock_async()
1650 __add_wait_queue_entry_tail(q, &wait->wait); in __folio_lock_async()
1660 __remove_wait_queue(q, &wait->wait); in __folio_lock_async()
1662 ret = -EIOCBQUEUED; in __folio_lock_async()
1663 spin_unlock_irq(&q->lock); in __folio_lock_async()
1669 * 0 - folio is locked.
1670 * non-zero - folio is not locked.
1671 * mmap_lock or per-VMA lock has been released (mmap_read_unlock() or
1676 * with the folio locked and the mmap_lock/per-VMA lock is left unperturbed.
1680 unsigned int flags = vmf->flags; in __folio_lock_or_retry()
1684 * CAUTION! In this case, mmap_lock/per-VMA lock is not in __folio_lock_or_retry()
1713 * page_cache_next_miss() - Find the next gap in the page cache.
1714 * @mapping: Mapping.
1718 * Search the range [index, min(index + max_scan - 1, ULONG_MAX)] for the
1728 * range specified (in which case 'return - index >= max_scan' will be true).
1729 * In the rare case of index wrap-around, 0 will be returned.
1731 pgoff_t page_cache_next_miss(struct address_space *mapping, in page_cache_next_miss() argument
1734 XA_STATE(xas, &mapping->i_pages, index); in page_cache_next_miss()
1736 while (max_scan--) { in page_cache_next_miss()
1749 * page_cache_prev_miss() - Find the previous gap in the page cache.
1750 * @mapping: Mapping.
1754 * Search the range [max(index - max_scan + 1, 0), index] for the
1764 * range specified (in which case 'index - return >= max_scan' will be true).
1765 * In the rare case of wrap-around, ULONG_MAX will be returned.
1767 pgoff_t page_cache_prev_miss(struct address_space *mapping, in page_cache_prev_miss() argument
1770 XA_STATE(xas, &mapping->i_pages, index); in page_cache_prev_miss()
1772 while (max_scan--) { in page_cache_prev_miss()
1805 * filemap_get_entry - Get a page cache entry.
1806 * @mapping: the address_space to search
1809 * Looks up the page cache entry at @mapping & @index. If it is a folio,
1816 void *filemap_get_entry(struct address_space *mapping, pgoff_t index) in filemap_get_entry() argument
1818 XA_STATE(xas, &mapping->i_pages, index); in filemap_get_entry()
1848 * __filemap_get_folio - Find and get a reference to a folio.
1849 * @mapping: The address_space to search.
1854 * Looks up the page cache entry at @mapping & @index.
1863 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index, in __filemap_get_folio() argument
1869 folio = filemap_get_entry(mapping, index); in __filemap_get_folio()
1879 return ERR_PTR(-EAGAIN); in __filemap_get_folio()
1886 if (unlikely(folio->mapping != mapping)) { in __filemap_get_folio()
1909 if ((fgp_flags & FGP_WRITE) && mapping_can_writeback(mapping)) in __filemap_get_folio()
1920 if (!mapping_large_folio_support(mapping)) in __filemap_get_folio()
1925 if (index & ((1UL << order) - 1)) in __filemap_get_folio()
1931 err = -ENOMEM; in __filemap_get_folio()
1944 err = filemap_add_folio(mapping, folio, index, gfp); in __filemap_get_folio()
1949 } while (order-- > 0); in __filemap_get_folio()
1951 if (err == -EEXIST) in __filemap_get_folio()
1964 return ERR_PTR(-ENOENT); in __filemap_get_folio()
2005 * find_get_entries - gang pagecache lookup
2006 * @mapping: The address_space to search
2013 * the mapping. The entries are placed in @fbatch. find_get_entries()
2017 * due to not-present entries or large folios.
2024 unsigned find_get_entries(struct address_space *mapping, pgoff_t *start, in find_get_entries() argument
2027 XA_STATE(xas, &mapping->i_pages, *start); in find_get_entries()
2032 indices[fbatch->nr] = xas.xa_index; in find_get_entries()
2040 int idx = folio_batch_count(fbatch) - 1; in find_get_entries()
2042 folio = fbatch->folios[idx]; in find_get_entries()
2051 * find_lock_entries - Find a batch of pagecache entries.
2052 * @mapping: The address_space to search.
2058 * find_lock_entries() will return a batch of entries from @mapping.
2065 * due to not-present entries, large folios, folios which could not be
2070 unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start, in find_lock_entries() argument
2073 XA_STATE(xas, &mapping->i_pages, *start); in find_lock_entries()
2079 if (folio->index < *start) in find_lock_entries()
2081 if (folio_next_index(folio) - 1 > end) in find_lock_entries()
2085 if (folio->mapping != mapping || in find_lock_entries()
2091 indices[fbatch->nr] = xas.xa_index; in find_lock_entries()
2104 int idx = folio_batch_count(fbatch) - 1; in find_lock_entries()
2106 folio = fbatch->folios[idx]; in find_lock_entries()
2115 * filemap_get_folios - Get a batch of folios
2116 * @mapping: The address_space to search
2121 * Search for and return a batch of folios in the mapping starting at
2135 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start, in filemap_get_folios() argument
2138 XA_STATE(xas, &mapping->i_pages, *start); in filemap_get_folios()
2151 *start = folio->index + nr; in filemap_get_folios()
2159 * breaks the iteration when there is a page at index -1 but that is in filemap_get_folios()
2162 if (end == (pgoff_t)-1) in filemap_get_folios()
2163 *start = (pgoff_t)-1; in filemap_get_folios()
2174 * filemap_get_folios_contig - Get a batch of contiguous folios
2175 * @mapping: The address_space to search
2188 unsigned filemap_get_folios_contig(struct address_space *mapping, in filemap_get_folios_contig() argument
2191 XA_STATE(xas, &mapping->i_pages, *start); in filemap_get_folios_contig()
2219 *start = folio->index + nr; in filemap_get_folios_contig()
2234 folio = fbatch->folios[nr - 1]; in filemap_get_folios_contig()
2236 *start = folio->index + 1; in filemap_get_folios_contig()
2247 * filemap_get_folios_tag - Get a batch of folios matching @tag
2248 * @mapping: The address_space to search
2259 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start, in filemap_get_folios_tag() argument
2262 XA_STATE(xas, &mapping->i_pages, *start); in filemap_get_folios_tag()
2279 *start = folio->index + nr; in filemap_get_folios_tag()
2286 * breaks the iteration when there is a page at index -1 but that is in filemap_get_folios_tag()
2289 if (end == (pgoff_t)-1) in filemap_get_folios_tag()
2290 *start = (pgoff_t)-1; in filemap_get_folios_tag()
2304 * ---R__________________________________________B__________
2317 ra->ra_pages /= 4; in shrink_readahead_size_eio()
2321 * filemap_get_read_batch - Get a batch of folios for read
2329 static void filemap_get_read_batch(struct address_space *mapping, in filemap_get_read_batch() argument
2332 XA_STATE(xas, &mapping->i_pages, index); in filemap_get_read_batch()
2355 xas_advance(&xas, folio_next_index(folio) - 1); in filemap_get_read_batch()
2394 shrink_readahead_size_eio(&file->f_ra); in filemap_read_folio()
2395 return -EIO; in filemap_read_folio()
2398 static bool filemap_range_uptodate(struct address_space *mapping, in filemap_range_uptodate() argument
2407 if (!mapping->a_ops->is_partially_uptodate) in filemap_range_uptodate()
2409 if (mapping->host->i_blkbits >= folio_shift(folio)) in filemap_range_uptodate()
2413 count -= folio_pos(folio) - pos; in filemap_range_uptodate()
2416 pos -= folio_pos(folio); in filemap_range_uptodate()
2419 return mapping->a_ops->is_partially_uptodate(folio, pos, count); in filemap_range_uptodate()
2423 struct address_space *mapping, size_t count, in filemap_update_page() argument
2428 if (iocb->ki_flags & IOCB_NOWAIT) { in filemap_update_page()
2429 if (!filemap_invalidate_trylock_shared(mapping)) in filemap_update_page()
2430 return -EAGAIN; in filemap_update_page()
2432 filemap_invalidate_lock_shared(mapping); in filemap_update_page()
2436 error = -EAGAIN; in filemap_update_page()
2437 if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_NOIO)) in filemap_update_page()
2439 if (!(iocb->ki_flags & IOCB_WAITQ)) { in filemap_update_page()
2440 filemap_invalidate_unlock_shared(mapping); in filemap_update_page()
2448 error = __folio_lock_async(folio, iocb->ki_waitq); in filemap_update_page()
2454 if (!folio->mapping) in filemap_update_page()
2458 if (filemap_range_uptodate(mapping, iocb->ki_pos, count, folio, in filemap_update_page()
2462 error = -EAGAIN; in filemap_update_page()
2463 if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) in filemap_update_page()
2466 error = filemap_read_folio(iocb->ki_filp, mapping->a_ops->read_folio, in filemap_update_page()
2472 filemap_invalidate_unlock_shared(mapping); in filemap_update_page()
2479 struct address_space *mapping, pgoff_t index, in filemap_create_folio() argument
2485 folio = filemap_alloc_folio(mapping_gfp_mask(mapping), 0); in filemap_create_folio()
2487 return -ENOMEM; in filemap_create_folio()
2498 * pages or ->readahead() that need to hold invalidate_lock in filemap_create_folio()
2499 * while mapping blocks for IO so let's hold the lock here as in filemap_create_folio()
2502 filemap_invalidate_lock_shared(mapping); in filemap_create_folio()
2503 error = filemap_add_folio(mapping, folio, index, in filemap_create_folio()
2504 mapping_gfp_constraint(mapping, GFP_KERNEL)); in filemap_create_folio()
2505 if (error == -EEXIST) in filemap_create_folio()
2510 error = filemap_read_folio(file, mapping->a_ops->read_folio, folio); in filemap_create_folio()
2514 filemap_invalidate_unlock_shared(mapping); in filemap_create_folio()
2518 filemap_invalidate_unlock_shared(mapping); in filemap_create_folio()
2524 struct address_space *mapping, struct folio *folio, in filemap_readahead() argument
2527 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, folio->index); in filemap_readahead()
2529 if (iocb->ki_flags & IOCB_NOIO) in filemap_readahead()
2530 return -EAGAIN; in filemap_readahead()
2531 page_cache_async_ra(&ractl, folio, last_index - folio->index); in filemap_readahead()
2538 struct file *filp = iocb->ki_filp; in filemap_get_pages()
2539 struct address_space *mapping = filp->f_mapping; in filemap_get_pages() local
2540 struct file_ra_state *ra = &filp->f_ra; in filemap_get_pages()
2541 pgoff_t index = iocb->ki_pos >> PAGE_SHIFT; in filemap_get_pages()
2547 last_index = DIV_ROUND_UP(iocb->ki_pos + count, PAGE_SIZE); in filemap_get_pages()
2550 return -EINTR; in filemap_get_pages()
2552 filemap_get_read_batch(mapping, index, last_index - 1, fbatch); in filemap_get_pages()
2554 if (iocb->ki_flags & IOCB_NOIO) in filemap_get_pages()
2555 return -EAGAIN; in filemap_get_pages()
2556 page_cache_sync_readahead(mapping, ra, filp, index, in filemap_get_pages()
2557 last_index - index); in filemap_get_pages()
2558 filemap_get_read_batch(mapping, index, last_index - 1, fbatch); in filemap_get_pages()
2561 if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_WAITQ)) in filemap_get_pages()
2562 return -EAGAIN; in filemap_get_pages()
2563 err = filemap_create_folio(filp, mapping, in filemap_get_pages()
2564 iocb->ki_pos >> PAGE_SHIFT, fbatch); in filemap_get_pages()
2570 folio = fbatch->folios[folio_batch_count(fbatch) - 1]; in filemap_get_pages()
2572 err = filemap_readahead(iocb, filp, mapping, folio, last_index); in filemap_get_pages()
2577 if ((iocb->ki_flags & IOCB_WAITQ) && in filemap_get_pages()
2579 iocb->ki_flags |= IOCB_NOWAIT; in filemap_get_pages()
2580 err = filemap_update_page(iocb, mapping, count, folio, in filemap_get_pages()
2590 if (likely(--fbatch->nr)) in filemap_get_pages()
2605 * filemap_read - Read data from the page cache.
2607 * @iter: Destination for the data.
2610 * Copies data from the page cache. If the data is not currently present,
2620 struct file *filp = iocb->ki_filp; in filemap_read()
2621 struct file_ra_state *ra = &filp->f_ra; in filemap_read()
2622 struct address_space *mapping = filp->f_mapping; in filemap_read() local
2623 struct inode *inode = mapping->host; in filemap_read()
2628 loff_t last_pos = ra->prev_pos; in filemap_read()
2630 if (unlikely(iocb->ki_pos >= inode->i_sb->s_maxbytes)) in filemap_read()
2635 iov_iter_truncate(iter, inode->i_sb->s_maxbytes); in filemap_read()
2642 * If we've already successfully copied some data, then we in filemap_read()
2643 * can no longer safely return -EIOCBQUEUED. Hence mark in filemap_read()
2646 if ((iocb->ki_flags & IOCB_WAITQ) && already_read) in filemap_read()
2647 iocb->ki_flags |= IOCB_NOWAIT; in filemap_read()
2649 if (unlikely(iocb->ki_pos >= i_size_read(inode))) in filemap_read()
2652 error = filemap_get_pages(iocb, iter->count, &fbatch, false); in filemap_read()
2660 * the correct value for "nr", which means the zero-filled in filemap_read()
2662 * another truncate extends the file - this is desired though). in filemap_read()
2665 if (unlikely(iocb->ki_pos >= isize)) in filemap_read()
2667 end_offset = min_t(loff_t, isize, iocb->ki_pos + iter->count); in filemap_read()
2670 * Once we start copying data, we don't want to be touching any in filemap_read()
2673 writably_mapped = mapping_writably_mapped(mapping); in filemap_read()
2679 if (!pos_same_folio(iocb->ki_pos, last_pos - 1, in filemap_read()
2686 size_t offset = iocb->ki_pos & (fsize - 1); in filemap_read()
2687 size_t bytes = min_t(loff_t, end_offset - iocb->ki_pos, in filemap_read()
2688 fsize - offset); in filemap_read()
2706 iocb->ki_pos += copied; in filemap_read()
2707 last_pos = iocb->ki_pos; in filemap_read()
2710 error = -EFAULT; in filemap_read()
2718 } while (iov_iter_count(iter) && iocb->ki_pos < isize && !error); in filemap_read()
2721 ra->prev_pos = last_pos; in filemap_read()
2728 struct address_space *mapping = iocb->ki_filp->f_mapping; in kiocb_write_and_wait() local
2729 loff_t pos = iocb->ki_pos; in kiocb_write_and_wait()
2730 loff_t end = pos + count - 1; in kiocb_write_and_wait()
2732 if (iocb->ki_flags & IOCB_NOWAIT) { in kiocb_write_and_wait()
2733 if (filemap_range_needs_writeback(mapping, pos, end)) in kiocb_write_and_wait()
2734 return -EAGAIN; in kiocb_write_and_wait()
2738 return filemap_write_and_wait_range(mapping, pos, end); in kiocb_write_and_wait()
2743 struct address_space *mapping = iocb->ki_filp->f_mapping; in kiocb_invalidate_pages() local
2744 loff_t pos = iocb->ki_pos; in kiocb_invalidate_pages()
2745 loff_t end = pos + count - 1; in kiocb_invalidate_pages()
2748 if (iocb->ki_flags & IOCB_NOWAIT) { in kiocb_invalidate_pages()
2750 if (filemap_range_has_page(mapping, pos, end)) in kiocb_invalidate_pages()
2751 return -EAGAIN; in kiocb_invalidate_pages()
2753 ret = filemap_write_and_wait_range(mapping, pos, end); in kiocb_invalidate_pages()
2760 * the new data. We invalidate clean cached page from the region we're in kiocb_invalidate_pages()
2762 * without clobbering -EIOCBQUEUED from ->direct_IO(). in kiocb_invalidate_pages()
2764 return invalidate_inode_pages2_range(mapping, pos >> PAGE_SHIFT, in kiocb_invalidate_pages()
2769 * generic_file_read_iter - generic filesystem read routine
2771 * @iter: destination for the data read
2776 * The IOCB_NOWAIT flag in iocb->ki_flags indicates that -EAGAIN shall
2777 * be returned when no data can be read without waiting for I/O requests
2780 * The IOCB_NOIO flag in iocb->ki_flags indicates that no new I/O
2781 * requests shall be made for the read or for readahead. When no data
2782 * can be read, -EAGAIN shall be returned. When readahead would be
2798 if (iocb->ki_flags & IOCB_DIRECT) { in generic_file_read_iter()
2799 struct file *file = iocb->ki_filp; in generic_file_read_iter()
2800 struct address_space *mapping = file->f_mapping; in generic_file_read_iter() local
2801 struct inode *inode = mapping->host; in generic_file_read_iter()
2808 retval = mapping->a_ops->direct_IO(iocb, iter); in generic_file_read_iter()
2810 iocb->ki_pos += retval; in generic_file_read_iter()
2811 count -= retval; in generic_file_read_iter()
2813 if (retval != -EIOCBQUEUED) in generic_file_read_iter()
2814 iov_iter_revert(iter, count - iov_iter_count(iter)); in generic_file_read_iter()
2827 if (iocb->ki_pos >= i_size_read(inode)) in generic_file_read_iter()
2845 size = min(size, folio_size(folio) - offset); in splice_folio_into_pipe()
2849 !pipe_full(pipe->head, pipe->tail, pipe->max_usage)) { in splice_folio_into_pipe()
2851 size_t part = min_t(size_t, PAGE_SIZE - offset, size - spliced); in splice_folio_into_pipe()
2860 pipe->head++; in splice_folio_into_pipe()
2870 * filemap_splice_read - Splice data from a file's pagecache into a pipe
2882 * will be updated if appropriate; 0 will be returned if there is no more data
2883 * to be read; -EAGAIN will be returned if the pipe had no space, and some
2885 * if the pipe has insufficient space, we reach the end of the data or we hit a
2899 if (unlikely(*ppos >= in->f_mapping->host->i_sb->s_maxbytes)) in filemap_splice_read()
2905 /* Work out how much data we can actually add into the pipe */ in filemap_splice_read()
2906 used = pipe_occupancy(pipe->head, pipe->tail); in filemap_splice_read()
2907 npages = max_t(ssize_t, pipe->max_usage - used, 0); in filemap_splice_read()
2915 if (*ppos >= i_size_read(in->f_mapping->host)) in filemap_splice_read()
2927 * the correct value for "nr", which means the zero-filled in filemap_splice_read()
2929 * another truncate extends the file - this is desired though). in filemap_splice_read()
2931 isize = i_size_read(in->f_mapping->host); in filemap_splice_read()
2937 * Once we start copying data, we don't want to be touching any in filemap_splice_read()
2940 writably_mapped = mapping_writably_mapped(in->f_mapping); in filemap_splice_read()
2958 n = min_t(loff_t, len, isize - *ppos); in filemap_splice_read()
2962 len -= n; in filemap_splice_read()
2965 in->f_ra.prev_pos = *ppos; in filemap_splice_read()
2966 if (pipe_full(pipe->head, pipe->tail, pipe->max_usage)) in filemap_splice_read()
2982 struct address_space *mapping, struct folio *folio, in folio_seek_hole_data() argument
2985 const struct address_space_operations *ops = mapping->a_ops; in folio_seek_hole_data()
2986 size_t offset, bsz = i_blocksize(mapping->host); in folio_seek_hole_data()
2990 if (!ops->is_partially_uptodate) in folio_seek_hole_data()
2996 if (unlikely(folio->mapping != mapping)) in folio_seek_hole_data()
2999 offset = offset_in_folio(folio, start) & ~(bsz - 1); in folio_seek_hole_data()
3002 if (ops->is_partially_uptodate(folio, offset, bsz) == in folio_seek_hole_data()
3005 start = (start + bsz) & ~(bsz - 1); in folio_seek_hole_data()
3017 return PAGE_SIZE << xa_get_order(xas->xa, xas->xa_index); in seek_folio_size()
3022 * mapping_seek_hole_data - Seek for SEEK_DATA / SEEK_HOLE in the page cache.
3023 * @mapping: Address space to search.
3029 * contain data, your filesystem can use this function to implement
3031 * entirely memory-based such as tmpfs, and filesystems which support
3034 * Return: The requested offset on success, or -ENXIO if @whence specifies
3035 * SEEK_DATA and there is no data after @start. There is an implicit hole
3036 * after @end - 1, so SEEK_HOLE returns @end if all the bytes between @start
3037 * and @end contain data.
3039 loff_t mapping_seek_hole_data(struct address_space *mapping, loff_t start, in mapping_seek_hole_data() argument
3042 XA_STATE(xas, &mapping->i_pages, start >> PAGE_SHIFT); in mapping_seek_hole_data()
3043 pgoff_t max = (end - 1) >> PAGE_SHIFT; in mapping_seek_hole_data()
3048 return -ENXIO; in mapping_seek_hole_data()
3063 start = folio_seek_hole_data(&xas, mapping, folio, start, pos, in mapping_seek_hole_data()
3075 start = -ENXIO; in mapping_seek_hole_data()
3088 * lock_folio_maybe_drop_mmap - lock the page, possibly dropping the mmap_lock
3089 * @vmf - the vm_fault for this fault.
3090 * @folio - the folio to lock.
3091 * @fpin - the pointer to the file we may pin (or is already pinned).
3110 if (vmf->flags & FAULT_FLAG_RETRY_NOWAIT) in lock_folio_maybe_drop_mmap()
3114 if (vmf->flags & FAULT_FLAG_KILLABLE) { in lock_folio_maybe_drop_mmap()
3123 mmap_read_unlock(vmf->vma->vm_mm); in lock_folio_maybe_drop_mmap()
3141 struct file *file = vmf->vma->vm_file; in do_sync_mmap_readahead()
3142 struct file_ra_state *ra = &file->f_ra; in do_sync_mmap_readahead()
3143 struct address_space *mapping = file->f_mapping; in do_sync_mmap_readahead() local
3144 DEFINE_READAHEAD(ractl, file, ra, mapping, vmf->pgoff); in do_sync_mmap_readahead()
3146 unsigned long vm_flags = vmf->vma->vm_flags; in do_sync_mmap_readahead()
3153 ractl._index &= ~((unsigned long)HPAGE_PMD_NR - 1); in do_sync_mmap_readahead()
3154 ra->size = HPAGE_PMD_NR; in do_sync_mmap_readahead()
3160 ra->size *= 2; in do_sync_mmap_readahead()
3161 ra->async_size = HPAGE_PMD_NR; in do_sync_mmap_readahead()
3167 /* If we don't want any read-ahead, don't bother */ in do_sync_mmap_readahead()
3170 if (!ra->ra_pages) in do_sync_mmap_readahead()
3175 page_cache_sync_ra(&ractl, ra->ra_pages); in do_sync_mmap_readahead()
3180 mmap_miss = READ_ONCE(ra->mmap_miss); in do_sync_mmap_readahead()
3182 WRITE_ONCE(ra->mmap_miss, ++mmap_miss); in do_sync_mmap_readahead()
3186 * stop bothering with read-ahead. It will only hurt. in do_sync_mmap_readahead()
3192 * mmap read-around in do_sync_mmap_readahead()
3195 ra->start = max_t(long, 0, vmf->pgoff - ra->ra_pages / 2); in do_sync_mmap_readahead()
3196 ra->size = ra->ra_pages; in do_sync_mmap_readahead()
3197 ra->async_size = ra->ra_pages / 4; in do_sync_mmap_readahead()
3198 ractl._index = ra->start; in do_sync_mmap_readahead()
3211 struct file *file = vmf->vma->vm_file; in do_async_mmap_readahead()
3212 struct file_ra_state *ra = &file->f_ra; in do_async_mmap_readahead()
3213 DEFINE_READAHEAD(ractl, file, ra, file->f_mapping, vmf->pgoff); in do_async_mmap_readahead()
3217 /* If we don't want any read-ahead, don't bother */ in do_async_mmap_readahead()
3218 if (vmf->vma->vm_flags & VM_RAND_READ || !ra->ra_pages) in do_async_mmap_readahead()
3221 mmap_miss = READ_ONCE(ra->mmap_miss); in do_async_mmap_readahead()
3223 WRITE_ONCE(ra->mmap_miss, --mmap_miss); in do_async_mmap_readahead()
3227 page_cache_async_ra(&ractl, folio, ra->ra_pages); in do_async_mmap_readahead()
3233 * filemap_fault - read in file data for page fault handling
3237 * mapped memory region to read in file data during a page fault.
3243 * vma->vm_mm->mmap_lock must be held on entry.
3253 * Return: bitwise-OR of %VM_FAULT_ codes.
3258 struct file *file = vmf->vma->vm_file; in filemap_fault()
3260 struct address_space *mapping = file->f_mapping; in filemap_fault() local
3261 struct inode *inode = mapping->host; in filemap_fault()
3262 pgoff_t max_idx, index = vmf->pgoff; in filemap_fault()
3274 folio = filemap_get_folio(mapping, index); in filemap_fault()
3280 if (!(vmf->flags & FAULT_FLAG_TRIED)) in filemap_fault()
3283 filemap_invalidate_lock_shared(mapping); in filemap_fault()
3289 count_memcg_event_mm(vmf->vma->vm_mm, PGMAJFAULT); in filemap_fault()
3298 filemap_invalidate_lock_shared(mapping); in filemap_fault()
3301 folio = __filemap_get_folio(mapping, index, in filemap_fault()
3303 vmf->gfp_mask); in filemap_fault()
3307 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3316 if (unlikely(folio->mapping != mapping)) { in filemap_fault()
3325 * that it's up-to-date. If not, it is going to be due to an error. in filemap_fault()
3344 * time to return to the upper layer and have it re-find the vma and in filemap_fault()
3352 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3365 vmf->page = folio_file_page(folio, index); in filemap_fault()
3370 * Umm, take care of errors if the page isn't up-to-date. in filemap_fault()
3371 * Try to re-read it _once_. We do this synchronously, in filemap_fault()
3376 error = filemap_read_folio(file, mapping->a_ops->read_folio, folio); in filemap_fault()
3383 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3390 * re-find the vma and come back and find our hopefully still populated in filemap_fault()
3396 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3406 struct mm_struct *mm = vmf->vma->vm_mm; in filemap_map_pmd()
3409 if (pmd_trans_huge(*vmf->pmd)) { in filemap_map_pmd()
3415 if (pmd_none(*vmf->pmd) && folio_test_pmd_mappable(folio)) { in filemap_map_pmd()
3425 if (pmd_none(*vmf->pmd)) in filemap_map_pmd()
3426 pmd_install(mm, vmf->pmd, &vmf->prealloc_pte); in filemap_map_pmd()
3432 struct address_space *mapping, pgoff_t end_pgoff) in next_uptodate_folio() argument
3455 if (folio->mapping != mapping) in next_uptodate_folio()
3459 max_idx = DIV_ROUND_UP(i_size_read(mapping->host), PAGE_SIZE); in next_uptodate_folio()
3460 if (xas->xa_index >= max_idx) in next_uptodate_folio()
3484 pte_t *old_ptep = vmf->pte; in filemap_map_folio_range()
3495 * fault-around logic. in filemap_map_folio_range()
3497 if (!pte_none(vmf->pte[count])) in filemap_map_folio_range()
3506 if (in_range(vmf->address, addr, count * PAGE_SIZE)) in filemap_map_folio_range()
3512 vmf->pte += count; in filemap_map_folio_range()
3515 } while (--nr_pages > 0); in filemap_map_folio_range()
3520 if (in_range(vmf->address, addr, count * PAGE_SIZE)) in filemap_map_folio_range()
3524 vmf->pte = old_ptep; in filemap_map_folio_range()
3534 struct page *page = &folio->page; in filemap_map_order0_folio()
3544 * the fault-around logic. in filemap_map_order0_folio()
3546 if (!pte_none(ptep_get(vmf->pte))) in filemap_map_order0_folio()
3549 if (vmf->address == addr) in filemap_map_order0_folio()
3561 struct vm_area_struct *vma = vmf->vma; in filemap_map_pages()
3562 struct file *file = vma->vm_file; in filemap_map_pages()
3563 struct address_space *mapping = file->f_mapping; in filemap_map_pages() local
3566 XA_STATE(xas, &mapping->i_pages, start_pgoff); in filemap_map_pages()
3572 folio = next_uptodate_folio(&xas, mapping, end_pgoff); in filemap_map_pages()
3581 addr = vma->vm_start + ((start_pgoff - vma->vm_pgoff) << PAGE_SHIFT); in filemap_map_pages()
3582 vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, addr, &vmf->ptl); in filemap_map_pages()
3583 if (!vmf->pte) { in filemap_map_pages()
3591 addr += (xas.xa_index - last_pgoff) << PAGE_SHIFT; in filemap_map_pages()
3592 vmf->pte += xas.xa_index - last_pgoff; in filemap_map_pages()
3594 end = folio->index + folio_nr_pages(folio) - 1; in filemap_map_pages()
3595 nr_pages = min(end, end_pgoff) - xas.xa_index + 1; in filemap_map_pages()
3602 xas.xa_index - folio->index, addr, in filemap_map_pages()
3607 } while ((folio = next_uptodate_folio(&xas, mapping, end_pgoff)) != NULL); in filemap_map_pages()
3608 pte_unmap_unlock(vmf->pte, vmf->ptl); in filemap_map_pages()
3612 mmap_miss_saved = READ_ONCE(file->f_ra.mmap_miss); in filemap_map_pages()
3614 WRITE_ONCE(file->f_ra.mmap_miss, 0); in filemap_map_pages()
3616 WRITE_ONCE(file->f_ra.mmap_miss, mmap_miss_saved - mmap_miss); in filemap_map_pages()
3624 struct address_space *mapping = vmf->vma->vm_file->f_mapping; in filemap_page_mkwrite() local
3625 struct folio *folio = page_folio(vmf->page); in filemap_page_mkwrite()
3628 sb_start_pagefault(mapping->host->i_sb); in filemap_page_mkwrite()
3629 file_update_time(vmf->vma->vm_file); in filemap_page_mkwrite()
3631 if (folio->mapping != mapping) { in filemap_page_mkwrite()
3644 sb_end_pagefault(mapping->host->i_sb); in filemap_page_mkwrite()
3658 struct address_space *mapping = file->f_mapping; in generic_file_mmap() local
3660 if (!mapping->a_ops->read_folio) in generic_file_mmap()
3661 return -ENOEXEC; in generic_file_mmap()
3663 vma->vm_ops = &generic_file_vm_ops; in generic_file_mmap()
3668 * This is for filesystems which do not implement ->writepage.
3672 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE)) in generic_file_readonly_mmap()
3673 return -EINVAL; in generic_file_readonly_mmap()
3683 return -ENOSYS; in generic_file_mmap()
3687 return -ENOSYS; in generic_file_readonly_mmap()
3695 static struct folio *do_read_cache_folio(struct address_space *mapping, in do_read_cache_folio() argument
3702 filler = mapping->a_ops->read_folio; in do_read_cache_folio()
3704 folio = filemap_get_folio(mapping, index); in do_read_cache_folio()
3708 return ERR_PTR(-ENOMEM); in do_read_cache_folio()
3709 err = filemap_add_folio(mapping, folio, index, gfp); in do_read_cache_folio()
3712 if (err == -EEXIST) in do_read_cache_folio()
3728 /* Folio was truncated from mapping */ in do_read_cache_folio()
3729 if (!folio->mapping) { in do_read_cache_folio()
3756 * read_cache_folio - Read into page cache, fill it if needed.
3757 * @mapping: The address_space to read from.
3759 * @filler: Function to perform the read, or NULL to use aops->read_folio().
3768 * Context: May sleep. Expects mapping->invalidate_lock to be held.
3771 struct folio *read_cache_folio(struct address_space *mapping, pgoff_t index, in read_cache_folio() argument
3774 return do_read_cache_folio(mapping, index, filler, file, in read_cache_folio()
3775 mapping_gfp_mask(mapping)); in read_cache_folio()
3780 * mapping_read_folio_gfp - Read into page cache, using specified allocation flags.
3781 * @mapping: The address_space for the folio.
3785 * This is the same as "read_cache_folio(mapping, index, NULL, NULL)", but with
3789 * possible and so is EINTR. If ->read_folio returns another error,
3792 * The function expects mapping->invalidate_lock to be already held.
3796 struct folio *mapping_read_folio_gfp(struct address_space *mapping, in mapping_read_folio_gfp() argument
3799 return do_read_cache_folio(mapping, index, NULL, NULL, gfp); in mapping_read_folio_gfp()
3803 static struct page *do_read_cache_page(struct address_space *mapping, in do_read_cache_page() argument
3808 folio = do_read_cache_folio(mapping, index, filler, file, gfp); in do_read_cache_page()
3810 return &folio->page; in do_read_cache_page()
3814 struct page *read_cache_page(struct address_space *mapping, in read_cache_page() argument
3817 return do_read_cache_page(mapping, index, filler, file, in read_cache_page()
3818 mapping_gfp_mask(mapping)); in read_cache_page()
3823 * read_cache_page_gfp - read into page cache, using specified page allocation flags.
3824 * @mapping: the page's address_space
3828 * This is the same as "read_mapping_page(mapping, index, NULL)", but with
3831 * If the page does not get brought uptodate, return -EIO.
3833 * The function expects mapping->invalidate_lock to be already held.
3837 struct page *read_cache_page_gfp(struct address_space *mapping, in read_cache_page_gfp() argument
3841 return do_read_cache_page(mapping, index, NULL, NULL, gfp); in read_cache_page_gfp()
3854 errseq_set(&filp->f_mapping->wb_err, -EIO); in dio_warn_stale_pagecache()
3859 …pr_crit("Page cache invalidation failure on direct I/O. Possible data corruption due to collision… in dio_warn_stale_pagecache()
3860 pr_crit("File: %s PID: %d Comm: %.20s\n", path, current->pid, in dio_warn_stale_pagecache()
3861 current->comm); in dio_warn_stale_pagecache()
3867 struct address_space *mapping = iocb->ki_filp->f_mapping; in kiocb_invalidate_post_direct_write() local
3869 if (mapping->nrpages && in kiocb_invalidate_post_direct_write()
3870 invalidate_inode_pages2_range(mapping, in kiocb_invalidate_post_direct_write()
3871 iocb->ki_pos >> PAGE_SHIFT, in kiocb_invalidate_post_direct_write()
3872 (iocb->ki_pos + count - 1) >> PAGE_SHIFT)) in kiocb_invalidate_post_direct_write()
3873 dio_warn_stale_pagecache(iocb->ki_filp); in kiocb_invalidate_post_direct_write()
3879 struct address_space *mapping = iocb->ki_filp->f_mapping; in generic_file_direct_write() local
3889 if (written == -EBUSY) in generic_file_direct_write()
3894 written = mapping->a_ops->direct_IO(iocb, from); in generic_file_direct_write()
3898 * cached by non-direct readahead, or faulted in by get_user_pages() in generic_file_direct_write()
3911 * Skip invalidation for async writes or if mapping has no pages. in generic_file_direct_write()
3914 struct inode *inode = mapping->host; in generic_file_direct_write()
3915 loff_t pos = iocb->ki_pos; in generic_file_direct_write()
3919 write_len -= written; in generic_file_direct_write()
3920 if (pos > i_size_read(inode) && !S_ISBLK(inode->i_mode)) { in generic_file_direct_write()
3924 iocb->ki_pos = pos; in generic_file_direct_write()
3926 if (written != -EIOCBQUEUED) in generic_file_direct_write()
3927 iov_iter_revert(from, write_len - iov_iter_count(from)); in generic_file_direct_write()
3934 struct file *file = iocb->ki_filp; in generic_perform_write()
3935 loff_t pos = iocb->ki_pos; in generic_perform_write()
3936 struct address_space *mapping = file->f_mapping; in generic_perform_write() local
3937 const struct address_space_operations *a_ops = mapping->a_ops; in generic_perform_write()
3948 offset = (pos & (PAGE_SIZE - 1)); in generic_perform_write()
3949 bytes = min_t(unsigned long, PAGE_SIZE - offset, in generic_perform_write()
3957 * up-to-date. in generic_perform_write()
3960 status = -EFAULT; in generic_perform_write()
3965 status = -EINTR; in generic_perform_write()
3969 status = a_ops->write_begin(file, mapping, pos, bytes, in generic_perform_write()
3974 if (mapping_writably_mapped(mapping)) in generic_perform_write()
3980 status = a_ops->write_end(file, mapping, pos, bytes, copied, in generic_perform_write()
3983 iov_iter_revert(i, copied - max(status, 0L)); in generic_perform_write()
3991 * A short copy made ->write_end() reject the in generic_perform_write()
4003 balance_dirty_pages_ratelimited(mapping); in generic_perform_write()
4008 iocb->ki_pos += written; in generic_perform_write()
4014 * __generic_file_write_iter - write data to a file
4016 * @from: iov_iter with data to write
4018 * This function does all the work needed for actually writing data to a
4026 * This function does *not* take care of syncing data in case of O_SYNC write.
4032 * * negative error code if no data has been written at all
4036 struct file *file = iocb->ki_filp; in __generic_file_write_iter()
4037 struct address_space *mapping = file->f_mapping; in __generic_file_write_iter() local
4038 struct inode *inode = mapping->host; in __generic_file_write_iter()
4049 if (iocb->ki_flags & IOCB_DIRECT) { in __generic_file_write_iter()
4056 * page-cache pages correctly). in __generic_file_write_iter()
4069 * generic_file_write_iter - write data to a file
4071 * @from: iov_iter with data to write
4077 * * negative error code if no data has been written at all of
4083 struct file *file = iocb->ki_filp; in generic_file_write_iter()
4084 struct inode *inode = file->f_mapping->host; in generic_file_write_iter()
4100 * filemap_release_folio() - Release fs-specific metadata on a folio.
4104 * The address_space is trying to release any data attached to a folio
4105 * (presumably at folio->private).
4118 struct address_space * const mapping = folio->mapping; in filemap_release_folio() local
4126 if (mapping && mapping->a_ops->release_folio) in filemap_release_folio()
4127 return mapping->a_ops->release_folio(folio, gfp); in filemap_release_folio()
4134 * filemap_cachestat() - compute the page cache statistics of a mapping
4135 * @mapping: The mapping to compute the statistics for.
4140 * This will query the page cache statistics of a mapping in the
4145 static void filemap_cachestat(struct address_space *mapping, in filemap_cachestat() argument
4148 XA_STATE(xas, &mapping->i_pages, first_index); in filemap_cachestat()
4167 folio_last_index = folio_first_index + nr_pages - 1; in filemap_cachestat()
4171 nr_pages -= first_index - folio_first_index; in filemap_cachestat()
4174 nr_pages -= folio_last_index - last_index; in filemap_cachestat()
4176 cs->nr_evicted += nr_pages; in filemap_cachestat()
4179 if (shmem_mapping(mapping)) { in filemap_cachestat()
4180 /* shmem file - in swap cache */ in filemap_cachestat()
4187 cs->nr_recently_evicted += nr_pages; in filemap_cachestat()
4194 folio_last_index = folio_first_index + nr_pages - 1; in filemap_cachestat()
4198 nr_pages -= first_index - folio_first_index; in filemap_cachestat()
4201 nr_pages -= folio_last_index - last_index; in filemap_cachestat()
4204 cs->nr_cache += nr_pages; in filemap_cachestat()
4207 cs->nr_dirty += nr_pages; in filemap_cachestat()
4210 cs->nr_writeback += nr_pages; in filemap_cachestat()
4235 * `off` and `len` must be non-negative integers. If `len` > 0,
4249 * zero - success
4250 * -EFAULT - cstat or cstat_range points to an illegal address
4251 * -EINVAL - invalid flags
4252 * -EBADF - invalid file descriptor
4253 * -EOPNOTSUPP - file descriptor is of a hugetlbfs file
4260 struct address_space *mapping; in SYSCALL_DEFINE4() local
4266 return -EBADF; in SYSCALL_DEFINE4()
4271 return -EFAULT; in SYSCALL_DEFINE4()
4277 return -EOPNOTSUPP; in SYSCALL_DEFINE4()
4282 return -EINVAL; in SYSCALL_DEFINE4()
4287 csr.len == 0 ? ULONG_MAX : (csr.off + csr.len - 1) >> PAGE_SHIFT; in SYSCALL_DEFINE4()
4289 mapping = f.file->f_mapping; in SYSCALL_DEFINE4()
4290 filemap_cachestat(mapping, first_index, last_index, &cs); in SYSCALL_DEFINE4()
4294 return -EFAULT; in SYSCALL_DEFINE4()