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>
72 * finished 'unifying' the page and buffer cache and SMP-threaded the
73 * page-cache, 21.05.1999, Ingo Molnar <mingo@redhat.com>
75 * SMP-threaded pagemap-LRU 1999, Andrea Arcangeli <andrea@suse.de>
81 * ->i_mmap_rwsem (truncate_pagecache)
82 * ->private_lock (__free_pte->block_dirty_folio)
83 * ->swap_lock (exclusive_swap_page, others)
84 * ->i_pages lock
86 * ->i_rwsem
87 * ->invalidate_lock (acquired by fs in truncate path)
88 * ->i_mmap_rwsem (truncate->unmap_mapping_range)
90 * ->mmap_lock
91 * ->i_mmap_rwsem
92 * ->page_table_lock or pte_lock (various, mainly in memory.c)
93 * ->i_pages lock (arch-dependent flush_dcache_mmap_lock)
95 * ->mmap_lock
96 * ->invalidate_lock (filemap_fault)
97 * ->lock_page (filemap_fault, access_process_vm)
99 * ->i_rwsem (generic_perform_write)
100 * ->mmap_lock (fault_in_readable->do_page_fault)
102 * bdi->wb.list_lock
103 * sb_lock (fs/fs-writeback.c)
104 * ->i_pages lock (__sync_single_inode)
106 * ->i_mmap_rwsem
107 * ->anon_vma.lock (vma_merge)
109 * ->anon_vma.lock
110 * ->page_table_lock or pte_lock (anon_vma_prepare and various)
112 * ->page_table_lock or pte_lock
113 * ->swap_lock (try_to_unmap_one)
114 * ->private_lock (try_to_unmap_one)
115 * ->i_pages lock (try_to_unmap_one)
116 * ->lruvec->lru_lock (follow_page_mask->mark_page_accessed)
117 * ->lruvec->lru_lock (check_pte_range->folio_isolate_lru)
118 * ->private_lock (folio_remove_rmap_pte->set_page_dirty)
119 * ->i_pages lock (folio_remove_rmap_pte->set_page_dirty)
120 * bdi.wb->list_lock (folio_remove_rmap_pte->set_page_dirty)
121 * ->inode->i_lock (folio_remove_rmap_pte->set_page_dirty)
122 * ->memcg->move_lock (folio_remove_rmap_pte->folio_memcg_lock)
123 * bdi.wb->list_lock (zap_pte_range->set_page_dirty)
124 * ->inode->i_lock (zap_pte_range->set_page_dirty)
125 * ->private_lock (zap_pte_range->block_dirty_folio)
129 struct address_space *mapping) in mapping_set_update() argument
131 if (dax_mapping(mapping) || shmem_mapping(mapping)) in mapping_set_update()
137 static void page_cache_delete(struct address_space *mapping, in page_cache_delete() argument
140 XA_STATE(xas, &mapping->i_pages, folio->index); in page_cache_delete()
143 mapping_set_update(&xas, mapping); in page_cache_delete()
145 xas_set_order(&xas, folio->index, folio_order(folio)); in page_cache_delete()
153 folio->mapping = NULL; in page_cache_delete()
154 /* Leave page->index set: truncation lookup relies upon it */ in page_cache_delete()
155 mapping->nrpages -= nr; in page_cache_delete()
158 static void filemap_unaccount_folio(struct address_space *mapping, in filemap_unaccount_folio() argument
166 current->comm, folio_pfn(folio)); in filemap_unaccount_folio()
167 dump_page(&folio->page, "still mapped when deleted"); in filemap_unaccount_folio()
171 if (mapping_exiting(mapping) && !folio_test_large(folio)) { in filemap_unaccount_folio()
181 atomic_set(&folio->_mapcount, -1); in filemap_unaccount_folio()
193 __lruvec_stat_mod_folio(folio, NR_FILE_PAGES, -nr); in filemap_unaccount_folio()
195 __lruvec_stat_mod_folio(folio, NR_SHMEM, -nr); in filemap_unaccount_folio()
197 __lruvec_stat_mod_folio(folio, NR_SHMEM_THPS, -nr); in filemap_unaccount_folio()
199 __lruvec_stat_mod_folio(folio, NR_FILE_THPS, -nr); in filemap_unaccount_folio()
200 filemap_nr_thps_dec(mapping); in filemap_unaccount_folio()
206 * unwritten data - on ordinary filesystems. in filemap_unaccount_folio()
208 * But it's harmless on in-memory filesystems like tmpfs; and can in filemap_unaccount_folio()
218 mapping_can_writeback(mapping))) in filemap_unaccount_folio()
219 folio_account_cleaned(folio, inode_to_wb(mapping->host)); in filemap_unaccount_folio()
224 * sure the page is locked and that nobody else uses it - or that usage
229 struct address_space *mapping = folio->mapping; in __filemap_remove_folio() local
232 filemap_unaccount_folio(mapping, folio); in __filemap_remove_folio()
233 page_cache_delete(mapping, folio, shadow); in __filemap_remove_folio()
236 void filemap_free_folio(struct address_space *mapping, struct folio *folio) in filemap_free_folio() argument
241 free_folio = mapping->a_ops->free_folio; in filemap_free_folio()
251 * filemap_remove_folio - Remove folio from page cache.
260 struct address_space *mapping = folio->mapping; in filemap_remove_folio() local
263 spin_lock(&mapping->host->i_lock); in filemap_remove_folio()
264 xa_lock_irq(&mapping->i_pages); in filemap_remove_folio()
266 xa_unlock_irq(&mapping->i_pages); in filemap_remove_folio()
267 if (mapping_shrinkable(mapping)) in filemap_remove_folio()
268 inode_add_lru(mapping->host); in filemap_remove_folio()
269 spin_unlock(&mapping->host->i_lock); in filemap_remove_folio()
271 filemap_free_folio(mapping, folio); in filemap_remove_folio()
275 * page_cache_delete_batch - delete several folios from page cache
276 * @mapping: the mapping to which folios belong
279 * The function walks over mapping->i_pages and removes folios passed in
280 * @fbatch from the mapping. The function expects @fbatch to be sorted
282 * It tolerates holes in @fbatch (mapping entries at those indices are not
287 static void page_cache_delete_batch(struct address_space *mapping, in page_cache_delete_batch() argument
290 XA_STATE(xas, &mapping->i_pages, fbatch->folios[0]->index); in page_cache_delete_batch()
295 mapping_set_update(&xas, mapping); in page_cache_delete_batch()
310 if (folio != fbatch->folios[i]) { in page_cache_delete_batch()
311 VM_BUG_ON_FOLIO(folio->index > in page_cache_delete_batch()
312 fbatch->folios[i]->index, folio); in page_cache_delete_batch()
318 folio->mapping = NULL; in page_cache_delete_batch()
319 /* Leave folio->index set: truncation lookup relies on it */ in page_cache_delete_batch()
325 mapping->nrpages -= total_pages; in page_cache_delete_batch()
328 void delete_from_page_cache_batch(struct address_space *mapping, in delete_from_page_cache_batch() argument
336 spin_lock(&mapping->host->i_lock); in delete_from_page_cache_batch()
337 xa_lock_irq(&mapping->i_pages); in delete_from_page_cache_batch()
339 struct folio *folio = fbatch->folios[i]; in delete_from_page_cache_batch()
342 filemap_unaccount_folio(mapping, folio); in delete_from_page_cache_batch()
344 page_cache_delete_batch(mapping, fbatch); in delete_from_page_cache_batch()
345 xa_unlock_irq(&mapping->i_pages); in delete_from_page_cache_batch()
346 if (mapping_shrinkable(mapping)) in delete_from_page_cache_batch()
347 inode_add_lru(mapping->host); in delete_from_page_cache_batch()
348 spin_unlock(&mapping->host->i_lock); in delete_from_page_cache_batch()
351 filemap_free_folio(mapping, fbatch->folios[i]); in delete_from_page_cache_batch()
354 int filemap_check_errors(struct address_space *mapping) in filemap_check_errors() argument
358 if (test_bit(AS_ENOSPC, &mapping->flags) && in filemap_check_errors()
359 test_and_clear_bit(AS_ENOSPC, &mapping->flags)) in filemap_check_errors()
360 ret = -ENOSPC; in filemap_check_errors()
361 if (test_bit(AS_EIO, &mapping->flags) && in filemap_check_errors()
362 test_and_clear_bit(AS_EIO, &mapping->flags)) in filemap_check_errors()
363 ret = -EIO; in filemap_check_errors()
368 static int filemap_check_and_keep_errors(struct address_space *mapping) in filemap_check_and_keep_errors() argument
371 if (test_bit(AS_EIO, &mapping->flags)) in filemap_check_and_keep_errors()
372 return -EIO; in filemap_check_and_keep_errors()
373 if (test_bit(AS_ENOSPC, &mapping->flags)) in filemap_check_and_keep_errors()
374 return -ENOSPC; in filemap_check_and_keep_errors()
379 * filemap_fdatawrite_wbc - start writeback on mapping dirty pages in range
380 * @mapping: address space structure to write
383 * Call writepages on the mapping using the provided wbc to control the
388 int filemap_fdatawrite_wbc(struct address_space *mapping, in filemap_fdatawrite_wbc() argument
393 if (!mapping_can_writeback(mapping) || in filemap_fdatawrite_wbc()
394 !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) in filemap_fdatawrite_wbc()
397 wbc_attach_fdatawrite_inode(wbc, mapping->host); in filemap_fdatawrite_wbc()
398 ret = do_writepages(mapping, wbc); in filemap_fdatawrite_wbc()
405 * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range
406 * @mapping: address space structure to write
411 * Start writeback against all of a mapping's dirty pages that lie
414 * If sync_mode is WB_SYNC_ALL then this is a "data integrity" operation, as
421 int __filemap_fdatawrite_range(struct address_space *mapping, loff_t start, in __filemap_fdatawrite_range() argument
431 return filemap_fdatawrite_wbc(mapping, &wbc); in __filemap_fdatawrite_range()
434 static inline int __filemap_fdatawrite(struct address_space *mapping, in __filemap_fdatawrite() argument
437 return __filemap_fdatawrite_range(mapping, 0, LLONG_MAX, sync_mode); in __filemap_fdatawrite()
440 int filemap_fdatawrite(struct address_space *mapping) in filemap_fdatawrite() argument
442 return __filemap_fdatawrite(mapping, WB_SYNC_ALL); in filemap_fdatawrite()
446 int filemap_fdatawrite_range(struct address_space *mapping, loff_t start, in filemap_fdatawrite_range() argument
449 return __filemap_fdatawrite_range(mapping, start, end, WB_SYNC_ALL); in filemap_fdatawrite_range()
454 * filemap_flush - mostly a non-blocking flush
455 * @mapping: target address_space
457 * This is a mostly non-blocking flush. Not suitable for data-integrity
458 * purposes - I/O may not be started against all dirty pages.
462 int filemap_flush(struct address_space *mapping) in filemap_flush() argument
464 return __filemap_fdatawrite(mapping, WB_SYNC_NONE); in filemap_flush()
469 * filemap_range_has_page - check if a page exists in range.
470 * @mapping: address space within which to check
480 bool filemap_range_has_page(struct address_space *mapping, in filemap_range_has_page() argument
484 XA_STATE(xas, &mapping->i_pages, start_byte >> PAGE_SHIFT); in filemap_range_has_page()
511 static void __filemap_fdatawait_range(struct address_space *mapping, in __filemap_fdatawait_range() argument
524 nr_folios = filemap_get_folios_tag(mapping, &index, end, in __filemap_fdatawait_range()
541 * filemap_fdatawait_range - wait for writeback to complete
542 * @mapping: address space structure to wait for
546 * Walk the list of under-writeback pages of the given address space
556 int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte, in filemap_fdatawait_range() argument
559 __filemap_fdatawait_range(mapping, start_byte, end_byte); in filemap_fdatawait_range()
560 return filemap_check_errors(mapping); in filemap_fdatawait_range()
565 * filemap_fdatawait_range_keep_errors - wait for writeback to complete
566 * @mapping: address space structure to wait for
570 * Walk the list of under-writeback pages of the given address space in the
575 * call sites are system-wide / filesystem-wide data flushers: e.g. sync(2),
578 int filemap_fdatawait_range_keep_errors(struct address_space *mapping, in filemap_fdatawait_range_keep_errors() argument
581 __filemap_fdatawait_range(mapping, start_byte, end_byte); in filemap_fdatawait_range_keep_errors()
582 return filemap_check_and_keep_errors(mapping); in filemap_fdatawait_range_keep_errors()
587 * file_fdatawait_range - wait for writeback to complete
592 * Walk the list of under-writeback pages of the address space that file
594 * status of the address space vs. the file->f_wb_err cursor and return it.
600 * Return: error status of the address space vs. the file->f_wb_err cursor.
604 struct address_space *mapping = file->f_mapping; in file_fdatawait_range() local
606 __filemap_fdatawait_range(mapping, start_byte, end_byte); in file_fdatawait_range()
612 * filemap_fdatawait_keep_errors - wait for writeback without clearing errors
613 * @mapping: address space structure to wait for
615 * Walk the list of under-writeback pages of the given address space
620 * call sites are system-wide / filesystem-wide data flushers: e.g. sync(2),
625 int filemap_fdatawait_keep_errors(struct address_space *mapping) in filemap_fdatawait_keep_errors() argument
627 __filemap_fdatawait_range(mapping, 0, LLONG_MAX); in filemap_fdatawait_keep_errors()
628 return filemap_check_and_keep_errors(mapping); in filemap_fdatawait_keep_errors()
633 static bool mapping_needs_writeback(struct address_space *mapping) in mapping_needs_writeback() argument
635 return mapping->nrpages; in mapping_needs_writeback()
638 bool filemap_range_has_writeback(struct address_space *mapping, in filemap_range_has_writeback() argument
641 XA_STATE(xas, &mapping->i_pages, start_byte >> PAGE_SHIFT); in filemap_range_has_writeback()
664 * filemap_write_and_wait_range - write out & wait on a file range
665 * @mapping: the address_space for the pages
669 * Write out and wait upon file offsets lstart->lend, inclusive.
672 * that this function can be used to write to the very end-of-file (end = -1).
676 int filemap_write_and_wait_range(struct address_space *mapping, in filemap_write_and_wait_range() argument
684 if (mapping_needs_writeback(mapping)) { in filemap_write_and_wait_range()
685 err = __filemap_fdatawrite_range(mapping, lstart, lend, in filemap_write_and_wait_range()
689 * written partially (e.g. -ENOSPC), so we wait for it. in filemap_write_and_wait_range()
690 * But the -EIO is special case, it may indicate the worst in filemap_write_and_wait_range()
693 if (err != -EIO) in filemap_write_and_wait_range()
694 __filemap_fdatawait_range(mapping, lstart, lend); in filemap_write_and_wait_range()
696 err2 = filemap_check_errors(mapping); in filemap_write_and_wait_range()
703 void __filemap_set_wb_err(struct address_space *mapping, int err) in __filemap_set_wb_err() argument
705 errseq_t eseq = errseq_set(&mapping->wb_err, err); in __filemap_set_wb_err()
707 trace_filemap_set_wb_err(mapping, eseq); in __filemap_set_wb_err()
712 * file_check_and_advance_wb_err - report wb error (if any) that was previously
720 * Grab the wb_err from the mapping. If it matches what we have in the file,
723 * If it doesn't match, then take the mapping value, set the "seen" flag in
729 * While we handle mapping->wb_err with atomic operations, the f_wb_err
738 errseq_t old = READ_ONCE(file->f_wb_err); in file_check_and_advance_wb_err()
739 struct address_space *mapping = file->f_mapping; in file_check_and_advance_wb_err() local
742 if (errseq_check(&mapping->wb_err, old)) { in file_check_and_advance_wb_err()
744 spin_lock(&file->f_lock); in file_check_and_advance_wb_err()
745 old = file->f_wb_err; in file_check_and_advance_wb_err()
746 err = errseq_check_and_advance(&mapping->wb_err, in file_check_and_advance_wb_err()
747 &file->f_wb_err); in file_check_and_advance_wb_err()
749 spin_unlock(&file->f_lock); in file_check_and_advance_wb_err()
757 clear_bit(AS_EIO, &mapping->flags); in file_check_and_advance_wb_err()
758 clear_bit(AS_ENOSPC, &mapping->flags); in file_check_and_advance_wb_err()
764 * file_write_and_wait_range - write out & wait on a file range
769 * Write out and wait upon file offsets lstart->lend, inclusive.
772 * that this function can be used to write to the very end-of-file (end = -1).
774 * After writing out and waiting on the data, we check and advance the
782 struct address_space *mapping = file->f_mapping; in file_write_and_wait_range() local
787 if (mapping_needs_writeback(mapping)) { in file_write_and_wait_range()
788 err = __filemap_fdatawrite_range(mapping, lstart, lend, in file_write_and_wait_range()
791 if (err != -EIO) in file_write_and_wait_range()
792 __filemap_fdatawait_range(mapping, lstart, lend); in file_write_and_wait_range()
802 * replace_page_cache_folio - replace a pagecache folio with a new one
816 struct address_space *mapping = old->mapping; in replace_page_cache_folio() local
817 void (*free_folio)(struct folio *) = mapping->a_ops->free_folio; in replace_page_cache_folio()
818 pgoff_t offset = old->index; in replace_page_cache_folio()
819 XA_STATE(xas, &mapping->i_pages, offset); in replace_page_cache_folio()
823 VM_BUG_ON_FOLIO(new->mapping, new); in replace_page_cache_folio()
826 new->mapping = mapping; in replace_page_cache_folio()
827 new->index = offset; in replace_page_cache_folio()
834 old->mapping = NULL; in replace_page_cache_folio()
851 noinline int __filemap_add_folio(struct address_space *mapping, in __filemap_add_folio() argument
854 XA_STATE(xas, &mapping->i_pages, index); in __filemap_add_folio()
862 VM_BUG_ON_FOLIO(folio_order(folio) < mapping_min_folio_order(mapping), in __filemap_add_folio()
864 mapping_set_update(&xas, mapping); in __filemap_add_folio()
866 VM_BUG_ON_FOLIO(index & (folio_nr_pages(folio) - 1), folio); in __filemap_add_folio()
873 folio->mapping = mapping; in __filemap_add_folio()
874 folio->index = xas.xa_index; in __filemap_add_folio()
877 int order = -1, split_order = 0; in __filemap_add_folio()
884 xas_set_err(&xas, -EEXIST); in __filemap_add_folio()
891 if (order == -1) in __filemap_add_folio()
895 /* entry may have changed before we re-acquire the lock */ in __filemap_add_folio()
904 BUG_ON(shmem_mapping(mapping)); in __filemap_add_folio()
920 mapping->nrpages += nr; in __filemap_add_folio()
954 folio->mapping = NULL; in __filemap_add_folio()
955 /* Leave page->index set: truncation relies upon it */ in __filemap_add_folio()
961 int filemap_add_folio(struct address_space *mapping, struct folio *folio, in filemap_add_folio() argument
972 ret = __filemap_add_folio(mapping, folio, index, gfp, &shadow); in filemap_add_folio()
982 * data from the working set, only to cache data that will in filemap_add_folio()
1016 * filemap_invalidate_lock_two - lock invalidate_lock for two mappings
1018 * Lock exclusively invalidate_lock of any passed mapping that is not NULL.
1020 * @mapping1: the first mapping to lock
1021 * @mapping2: the second mapping to lock
1029 down_write(&mapping1->invalidate_lock); in filemap_invalidate_lock_two()
1031 down_write_nested(&mapping2->invalidate_lock, 1); in filemap_invalidate_lock_two()
1036 * filemap_invalidate_unlock_two - unlock invalidate_lock for two mappings
1038 * Unlock exclusive invalidate_lock of any passed mapping that is not NULL.
1040 * @mapping1: the first mapping to unlock
1041 * @mapping2: the second mapping to unlock
1047 up_write(&mapping1->invalidate_lock); in filemap_invalidate_unlock_two()
1049 up_write(&mapping2->invalidate_lock); in filemap_invalidate_unlock_two()
1083 * The page wait code treats the "wait->flags" somewhat unusually, because
1130 flags = wait->flags; in wake_page_function()
1132 if (test_bit(key->bit_nr, &key->folio->flags)) in wake_page_function()
1133 return -1; in wake_page_function()
1135 if (test_and_set_bit(key->bit_nr, &key->folio->flags)) in wake_page_function()
1136 return -1; in wake_page_function()
1142 * We are holding the wait-queue lock, but the waiter that in wake_page_function()
1147 * afterwards to avoid any races. This store-release pairs in wake_page_function()
1148 * with the load-acquire in folio_wait_bit_common(). in wake_page_function()
1150 smp_store_release(&wait->flags, flags | WQ_FLAG_WOKEN); in wake_page_function()
1151 wake_up_state(wait->private, mode); in wake_page_function()
1159 * After this list_del_init(&wait->entry) the wait entry in wake_page_function()
1160 * might be de-allocated and the process might even have in wake_page_function()
1163 list_del_init_careful(&wait->entry); in wake_page_function()
1177 spin_lock_irqsave(&q->lock, flags); in folio_wake_bit()
1192 spin_unlock_irqrestore(&q->lock, flags); in folio_wake_bit()
1217 if (wait->flags & WQ_FLAG_EXCLUSIVE) { in folio_trylock_flag()
1218 if (test_and_set_bit(bit_nr, &folio->flags)) in folio_trylock_flag()
1220 } else if (test_bit(bit_nr, &folio->flags)) in folio_trylock_flag()
1223 wait->flags |= WQ_FLAG_WOKEN | WQ_FLAG_DONE; in folio_trylock_flag()
1249 wait->func = wake_page_function; in folio_wait_bit_common()
1254 wait->flags = 0; in folio_wait_bit_common()
1256 wait->flags = WQ_FLAG_EXCLUSIVE; in folio_wait_bit_common()
1257 if (--unfairness < 0) in folio_wait_bit_common()
1258 wait->flags |= WQ_FLAG_CUSTOM; in folio_wait_bit_common()
1275 spin_lock_irq(&q->lock); in folio_wait_bit_common()
1279 spin_unlock_irq(&q->lock); in folio_wait_bit_common()
1295 * be very careful with the 'wait->flags', because in folio_wait_bit_common()
1304 flags = smp_load_acquire(&wait->flags); in folio_wait_bit_common()
1313 /* If we were non-exclusive, we're done */ in folio_wait_bit_common()
1330 wait->flags |= WQ_FLAG_DONE; in folio_wait_bit_common()
1336 * waiter from the wait-queues, but the folio waiters bit will remain in folio_wait_bit_common()
1348 * NOTE! The wait->flags weren't stable until we've done the in folio_wait_bit_common()
1357 * Also note that WQ_FLAG_WOKEN is sufficient for a non-exclusive in folio_wait_bit_common()
1361 return wait->flags & WQ_FLAG_DONE ? 0 : -EINTR; in folio_wait_bit_common()
1363 return wait->flags & WQ_FLAG_WOKEN ? 0 : -EINTR; in folio_wait_bit_common()
1368 * migration_entry_wait_on_locked - Wait for a migration entry to be removed
1402 wait->func = wake_page_function; in migration_entry_wait_on_locked()
1405 wait->flags = 0; in migration_entry_wait_on_locked()
1407 spin_lock_irq(&q->lock); in migration_entry_wait_on_locked()
1411 spin_unlock_irq(&q->lock); in migration_entry_wait_on_locked()
1426 flags = smp_load_acquire(&wait->flags); in migration_entry_wait_on_locked()
1459 * folio_put_wait_locked - Drop a reference and wait for it to be unlocked
1469 * Return: 0 if the folio was unlocked or -EINTR if interrupted by a signal.
1477 * folio_add_wait_queue - Add an arbitrary waiter to a folio's wait queue
1488 spin_lock_irqsave(&q->lock, flags); in folio_add_wait_queue()
1491 spin_unlock_irqrestore(&q->lock, flags); in folio_add_wait_queue()
1496 * folio_unlock - Unlock a locked folio.
1516 * folio_end_read - End read on a folio.
1546 * folio_end_private_2 - Clear PG_private_2 and wake any waiters.
1566 * folio_wait_private_2 - Wait for PG_private_2 to be cleared on a folio.
1579 * folio_wait_private_2_killable - Wait for PG_private_2 to be cleared on a folio.
1586 * - 0 if successful.
1587 * - -EINTR if a fatal signal was encountered.
1604 * folio_end_writeback - End writeback against a folio.
1642 * __folio_lock - Get a lock on the folio, assuming we need to sleep to get it.
1664 wait->folio = folio; in __folio_lock_async()
1665 wait->bit_nr = PG_locked; in __folio_lock_async()
1667 spin_lock_irq(&q->lock); in __folio_lock_async()
1668 __add_wait_queue_entry_tail(q, &wait->wait); in __folio_lock_async()
1678 __remove_wait_queue(q, &wait->wait); in __folio_lock_async()
1680 ret = -EIOCBQUEUED; in __folio_lock_async()
1681 spin_unlock_irq(&q->lock); in __folio_lock_async()
1687 * 0 - folio is locked.
1688 * non-zero - folio is not locked.
1689 * mmap_lock or per-VMA lock has been released (mmap_read_unlock() or
1694 * with the folio locked and the mmap_lock/per-VMA lock is left unperturbed.
1698 unsigned int flags = vmf->flags; in __folio_lock_or_retry()
1702 * CAUTION! In this case, mmap_lock/per-VMA lock is not in __folio_lock_or_retry()
1731 * page_cache_next_miss() - Find the next gap in the page cache.
1732 * @mapping: Mapping.
1736 * Search the range [index, min(index + max_scan - 1, ULONG_MAX)] for the
1746 * range specified (in which case 'return - index >= max_scan' will be true).
1747 * In the rare case of index wrap-around, 0 will be returned.
1749 pgoff_t page_cache_next_miss(struct address_space *mapping, in page_cache_next_miss() argument
1752 XA_STATE(xas, &mapping->i_pages, index); in page_cache_next_miss()
1754 while (max_scan--) { in page_cache_next_miss()
1767 * page_cache_prev_miss() - Find the previous gap in the page cache.
1768 * @mapping: Mapping.
1772 * Search the range [max(index - max_scan + 1, 0), index] for the
1782 * range specified (in which case 'index - return >= max_scan' will be true).
1783 * In the rare case of wrap-around, ULONG_MAX will be returned.
1785 pgoff_t page_cache_prev_miss(struct address_space *mapping, in page_cache_prev_miss() argument
1788 XA_STATE(xas, &mapping->i_pages, index); in page_cache_prev_miss()
1790 while (max_scan--) { in page_cache_prev_miss()
1815 * increased by a speculative page cache (or GUP-fast) lookup as it can
1823 * filemap_get_entry - Get a page cache entry.
1824 * @mapping: the address_space to search
1827 * Looks up the page cache entry at @mapping & @index. If it is a folio,
1834 void *filemap_get_entry(struct address_space *mapping, pgoff_t index) in filemap_get_entry() argument
1836 XA_STATE(xas, &mapping->i_pages, index); in filemap_get_entry()
1866 * __filemap_get_folio - Find and get a reference to a folio.
1867 * @mapping: The address_space to search.
1872 * Looks up the page cache entry at @mapping & @index.
1881 struct folio *__filemap_get_folio(struct address_space *mapping, pgoff_t index, in __filemap_get_folio() argument
1887 folio = filemap_get_entry(mapping, index); in __filemap_get_folio()
1897 return ERR_PTR(-EAGAIN); in __filemap_get_folio()
1904 if (unlikely(folio->mapping != mapping)) { in __filemap_get_folio()
1924 unsigned int min_order = mapping_min_folio_order(mapping); in __filemap_get_folio()
1927 index = mapping_align_index(mapping, index); in __filemap_get_folio()
1929 if ((fgp_flags & FGP_WRITE) && mapping_can_writeback(mapping)) in __filemap_get_folio()
1940 if (order > mapping_max_folio_order(mapping)) in __filemap_get_folio()
1941 order = mapping_max_folio_order(mapping); in __filemap_get_folio()
1943 if (index & ((1UL << order) - 1)) in __filemap_get_folio()
1949 err = -ENOMEM; in __filemap_get_folio()
1960 err = filemap_add_folio(mapping, folio, index, gfp); in __filemap_get_folio()
1965 } while (order-- > min_order); in __filemap_get_folio()
1967 if (err == -EEXIST) in __filemap_get_folio()
1980 return ERR_PTR(-ENOENT); in __filemap_get_folio()
2021 * find_get_entries - gang pagecache lookup
2022 * @mapping: The address_space to search
2029 * the mapping. The entries are placed in @fbatch. find_get_entries()
2033 * due to not-present entries or large folios.
2040 unsigned find_get_entries(struct address_space *mapping, pgoff_t *start, in find_get_entries() argument
2043 XA_STATE(xas, &mapping->i_pages, *start); in find_get_entries()
2048 indices[fbatch->nr] = xas.xa_index; in find_get_entries()
2055 int idx = folio_batch_count(fbatch) - 1; in find_get_entries()
2057 folio = fbatch->folios[idx]; in find_get_entries()
2061 nr = 1 << xa_get_order(&mapping->i_pages, indices[idx]); in find_get_entries()
2070 * find_lock_entries - Find a batch of pagecache entries.
2071 * @mapping: The address_space to search.
2077 * find_lock_entries() will return a batch of entries from @mapping.
2084 * due to not-present entries, large folios, folios which could not be
2089 unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start, in find_lock_entries() argument
2092 XA_STATE(xas, &mapping->i_pages, *start); in find_lock_entries()
2102 base = folio->index; in find_lock_entries()
2107 if (base + nr - 1 > end) in find_lock_entries()
2111 if (folio->mapping != mapping || in find_lock_entries()
2118 base = xas.xa_index & ~(nr - 1); in find_lock_entries()
2123 if (base + nr - 1 > end) in find_lock_entries()
2129 indices[fbatch->nr] = xas.xa_index; in find_lock_entries()
2144 * filemap_get_folios - Get a batch of folios
2145 * @mapping: The address_space to search
2150 * Search for and return a batch of folios in the mapping starting at
2157 unsigned filemap_get_folios(struct address_space *mapping, pgoff_t *start, in filemap_get_folios() argument
2160 return filemap_get_folios_tag(mapping, start, end, XA_PRESENT, fbatch); in filemap_get_folios()
2165 * filemap_get_folios_contig - Get a batch of contiguous folios
2166 * @mapping: The address_space to search
2179 unsigned filemap_get_folios_contig(struct address_space *mapping, in filemap_get_folios_contig() argument
2182 XA_STATE(xas, &mapping->i_pages, *start); in filemap_get_folios_contig()
2211 *start = folio->index + nr; in filemap_get_folios_contig()
2226 folio = fbatch->folios[nr - 1]; in filemap_get_folios_contig()
2236 * filemap_get_folios_tag - Get a batch of folios matching @tag
2237 * @mapping: The address_space to search
2254 unsigned filemap_get_folios_tag(struct address_space *mapping, pgoff_t *start, in filemap_get_folios_tag() argument
2257 XA_STATE(xas, &mapping->i_pages, *start); in filemap_get_folios_tag()
2271 *start = folio->index + nr; in filemap_get_folios_tag()
2278 * breaks the iteration when there is a page at index -1 but that is in filemap_get_folios_tag()
2281 if (end == (pgoff_t)-1) in filemap_get_folios_tag()
2282 *start = (pgoff_t)-1; in filemap_get_folios_tag()
2296 * ---R__________________________________________B__________
2309 ra->ra_pages /= 4; in shrink_readahead_size_eio()
2313 * filemap_get_read_batch - Get a batch of folios for read
2321 static void filemap_get_read_batch(struct address_space *mapping, in filemap_get_read_batch() argument
2324 XA_STATE(xas, &mapping->i_pages, index); in filemap_get_read_batch()
2347 xas_advance(&xas, folio_next_index(folio) - 1); in filemap_get_read_batch()
2379 shrink_readahead_size_eio(&file->f_ra); in filemap_read_folio()
2380 return -EIO; in filemap_read_folio()
2383 static bool filemap_range_uptodate(struct address_space *mapping, in filemap_range_uptodate() argument
2392 if (!mapping->a_ops->is_partially_uptodate) in filemap_range_uptodate()
2394 if (mapping->host->i_blkbits >= folio_shift(folio)) in filemap_range_uptodate()
2398 count -= folio_pos(folio) - pos; in filemap_range_uptodate()
2401 pos -= folio_pos(folio); in filemap_range_uptodate()
2404 return mapping->a_ops->is_partially_uptodate(folio, pos, count); in filemap_range_uptodate()
2408 struct address_space *mapping, size_t count, in filemap_update_page() argument
2413 if (iocb->ki_flags & IOCB_NOWAIT) { in filemap_update_page()
2414 if (!filemap_invalidate_trylock_shared(mapping)) in filemap_update_page()
2415 return -EAGAIN; in filemap_update_page()
2417 filemap_invalidate_lock_shared(mapping); in filemap_update_page()
2421 error = -EAGAIN; in filemap_update_page()
2422 if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_NOIO)) in filemap_update_page()
2424 if (!(iocb->ki_flags & IOCB_WAITQ)) { in filemap_update_page()
2425 filemap_invalidate_unlock_shared(mapping); in filemap_update_page()
2433 error = __folio_lock_async(folio, iocb->ki_waitq); in filemap_update_page()
2439 if (!folio->mapping) in filemap_update_page()
2443 if (filemap_range_uptodate(mapping, iocb->ki_pos, count, folio, in filemap_update_page()
2447 error = -EAGAIN; in filemap_update_page()
2448 if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) in filemap_update_page()
2451 error = filemap_read_folio(iocb->ki_filp, mapping->a_ops->read_folio, in filemap_update_page()
2457 filemap_invalidate_unlock_shared(mapping); in filemap_update_page()
2464 struct address_space *mapping, loff_t pos, in filemap_create_folio() argument
2469 unsigned int min_order = mapping_min_folio_order(mapping); in filemap_create_folio()
2472 folio = filemap_alloc_folio(mapping_gfp_mask(mapping), min_order); in filemap_create_folio()
2474 return -ENOMEM; in filemap_create_folio()
2485 * pages or ->readahead() that need to hold invalidate_lock in filemap_create_folio()
2486 * while mapping blocks for IO so let's hold the lock here as in filemap_create_folio()
2489 filemap_invalidate_lock_shared(mapping); in filemap_create_folio()
2491 error = filemap_add_folio(mapping, folio, index, in filemap_create_folio()
2492 mapping_gfp_constraint(mapping, GFP_KERNEL)); in filemap_create_folio()
2493 if (error == -EEXIST) in filemap_create_folio()
2498 error = filemap_read_folio(file, mapping->a_ops->read_folio, folio); in filemap_create_folio()
2502 filemap_invalidate_unlock_shared(mapping); in filemap_create_folio()
2506 filemap_invalidate_unlock_shared(mapping); in filemap_create_folio()
2512 struct address_space *mapping, struct folio *folio, in filemap_readahead() argument
2515 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, folio->index); in filemap_readahead()
2517 if (iocb->ki_flags & IOCB_NOIO) in filemap_readahead()
2518 return -EAGAIN; in filemap_readahead()
2519 page_cache_async_ra(&ractl, folio, last_index - folio->index); in filemap_readahead()
2526 struct file *filp = iocb->ki_filp; in filemap_get_pages()
2527 struct address_space *mapping = filp->f_mapping; in filemap_get_pages() local
2528 struct file_ra_state *ra = &filp->f_ra; in filemap_get_pages()
2529 pgoff_t index = iocb->ki_pos >> PAGE_SHIFT; in filemap_get_pages()
2536 last_index = DIV_ROUND_UP(iocb->ki_pos + count, PAGE_SIZE); in filemap_get_pages()
2539 return -EINTR; in filemap_get_pages()
2541 filemap_get_read_batch(mapping, index, last_index - 1, fbatch); in filemap_get_pages()
2543 if (iocb->ki_flags & IOCB_NOIO) in filemap_get_pages()
2544 return -EAGAIN; in filemap_get_pages()
2545 if (iocb->ki_flags & IOCB_NOWAIT) in filemap_get_pages()
2547 page_cache_sync_readahead(mapping, ra, filp, index, in filemap_get_pages()
2548 last_index - index); in filemap_get_pages()
2549 if (iocb->ki_flags & IOCB_NOWAIT) in filemap_get_pages()
2551 filemap_get_read_batch(mapping, index, last_index - 1, fbatch); in filemap_get_pages()
2554 if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_WAITQ)) in filemap_get_pages()
2555 return -EAGAIN; in filemap_get_pages()
2556 err = filemap_create_folio(filp, mapping, iocb->ki_pos, fbatch); in filemap_get_pages()
2562 folio = fbatch->folios[folio_batch_count(fbatch) - 1]; in filemap_get_pages()
2564 err = filemap_readahead(iocb, filp, mapping, folio, last_index); in filemap_get_pages()
2569 if ((iocb->ki_flags & IOCB_WAITQ) && in filemap_get_pages()
2571 iocb->ki_flags |= IOCB_NOWAIT; in filemap_get_pages()
2572 err = filemap_update_page(iocb, mapping, count, folio, in filemap_get_pages()
2578 trace_mm_filemap_get_pages(mapping, index, last_index - 1); in filemap_get_pages()
2583 if (likely(--fbatch->nr)) in filemap_get_pages()
2598 * filemap_read - Read data from the page cache.
2600 * @iter: Destination for the data.
2603 * Copies data from the page cache. If the data is not currently present,
2613 struct file *filp = iocb->ki_filp; in filemap_read()
2614 struct file_ra_state *ra = &filp->f_ra; in filemap_read()
2615 struct address_space *mapping = filp->f_mapping; in filemap_read() local
2616 struct inode *inode = mapping->host; in filemap_read()
2621 loff_t last_pos = ra->prev_pos; in filemap_read()
2623 if (unlikely(iocb->ki_pos >= inode->i_sb->s_maxbytes)) in filemap_read()
2628 iov_iter_truncate(iter, inode->i_sb->s_maxbytes - iocb->ki_pos); in filemap_read()
2635 * If we've already successfully copied some data, then we in filemap_read()
2636 * can no longer safely return -EIOCBQUEUED. Hence mark in filemap_read()
2639 if ((iocb->ki_flags & IOCB_WAITQ) && already_read) in filemap_read()
2640 iocb->ki_flags |= IOCB_NOWAIT; in filemap_read()
2642 if (unlikely(iocb->ki_pos >= i_size_read(inode))) in filemap_read()
2645 error = filemap_get_pages(iocb, iter->count, &fbatch, false); in filemap_read()
2653 * the correct value for "nr", which means the zero-filled in filemap_read()
2655 * another truncate extends the file - this is desired though). in filemap_read()
2658 if (unlikely(iocb->ki_pos >= isize)) in filemap_read()
2660 end_offset = min_t(loff_t, isize, iocb->ki_pos + iter->count); in filemap_read()
2663 * Once we start copying data, we don't want to be touching any in filemap_read()
2666 writably_mapped = mapping_writably_mapped(mapping); in filemap_read()
2672 if (!pos_same_folio(iocb->ki_pos, last_pos - 1, in filemap_read()
2679 size_t offset = iocb->ki_pos & (fsize - 1); in filemap_read()
2680 size_t bytes = min_t(loff_t, end_offset - iocb->ki_pos, in filemap_read()
2681 fsize - offset); in filemap_read()
2699 iocb->ki_pos += copied; in filemap_read()
2700 last_pos = iocb->ki_pos; in filemap_read()
2703 error = -EFAULT; in filemap_read()
2711 } while (iov_iter_count(iter) && iocb->ki_pos < isize && !error); in filemap_read()
2714 ra->prev_pos = last_pos; in filemap_read()
2721 struct address_space *mapping = iocb->ki_filp->f_mapping; in kiocb_write_and_wait() local
2722 loff_t pos = iocb->ki_pos; in kiocb_write_and_wait()
2723 loff_t end = pos + count - 1; in kiocb_write_and_wait()
2725 if (iocb->ki_flags & IOCB_NOWAIT) { in kiocb_write_and_wait()
2726 if (filemap_range_needs_writeback(mapping, pos, end)) in kiocb_write_and_wait()
2727 return -EAGAIN; in kiocb_write_and_wait()
2731 return filemap_write_and_wait_range(mapping, pos, end); in kiocb_write_and_wait()
2735 int filemap_invalidate_pages(struct address_space *mapping, in filemap_invalidate_pages() argument
2742 if (filemap_range_has_page(mapping, pos, end)) in filemap_invalidate_pages()
2743 return -EAGAIN; in filemap_invalidate_pages()
2745 ret = filemap_write_and_wait_range(mapping, pos, end); in filemap_invalidate_pages()
2752 * the new data. We invalidate clean cached page from the region we're in filemap_invalidate_pages()
2754 * without clobbering -EIOCBQUEUED from ->direct_IO(). in filemap_invalidate_pages()
2756 return invalidate_inode_pages2_range(mapping, pos >> PAGE_SHIFT, in filemap_invalidate_pages()
2762 struct address_space *mapping = iocb->ki_filp->f_mapping; in kiocb_invalidate_pages() local
2764 return filemap_invalidate_pages(mapping, iocb->ki_pos, in kiocb_invalidate_pages()
2765 iocb->ki_pos + count - 1, in kiocb_invalidate_pages()
2766 iocb->ki_flags & IOCB_NOWAIT); in kiocb_invalidate_pages()
2771 * generic_file_read_iter - generic filesystem read routine
2773 * @iter: destination for the data read
2778 * The IOCB_NOWAIT flag in iocb->ki_flags indicates that -EAGAIN shall
2779 * be returned when no data can be read without waiting for I/O requests
2782 * The IOCB_NOIO flag in iocb->ki_flags indicates that no new I/O
2783 * requests shall be made for the read or for readahead. When no data
2784 * can be read, -EAGAIN shall be returned. When readahead would be
2800 if (iocb->ki_flags & IOCB_DIRECT) { in generic_file_read_iter()
2801 struct file *file = iocb->ki_filp; in generic_file_read_iter()
2802 struct address_space *mapping = file->f_mapping; in generic_file_read_iter() local
2803 struct inode *inode = mapping->host; in generic_file_read_iter()
2810 retval = mapping->a_ops->direct_IO(iocb, iter); in generic_file_read_iter()
2812 iocb->ki_pos += retval; in generic_file_read_iter()
2813 count -= retval; in generic_file_read_iter()
2815 if (retval != -EIOCBQUEUED) in generic_file_read_iter()
2816 iov_iter_revert(iter, count - iov_iter_count(iter)); in generic_file_read_iter()
2829 if (iocb->ki_pos >= i_size_read(inode)) in generic_file_read_iter()
2847 size = min(size, folio_size(folio) - offset); in splice_folio_into_pipe()
2851 !pipe_full(pipe->head, pipe->tail, pipe->max_usage)) { in splice_folio_into_pipe()
2853 size_t part = min_t(size_t, PAGE_SIZE - offset, size - spliced); in splice_folio_into_pipe()
2862 pipe->head++; in splice_folio_into_pipe()
2872 * filemap_splice_read - Splice data from a file's pagecache into a pipe
2884 * will be updated if appropriate; 0 will be returned if there is no more data
2885 * to be read; -EAGAIN will be returned if the pipe had no space, and some
2887 * if the pipe has insufficient space, we reach the end of the data or we hit a
2901 if (unlikely(*ppos >= in->f_mapping->host->i_sb->s_maxbytes)) in filemap_splice_read()
2907 /* Work out how much data we can actually add into the pipe */ in filemap_splice_read()
2908 used = pipe_occupancy(pipe->head, pipe->tail); in filemap_splice_read()
2909 npages = max_t(ssize_t, pipe->max_usage - used, 0); in filemap_splice_read()
2917 if (*ppos >= i_size_read(in->f_mapping->host)) in filemap_splice_read()
2929 * the correct value for "nr", which means the zero-filled in filemap_splice_read()
2931 * another truncate extends the file - this is desired though). in filemap_splice_read()
2933 isize = i_size_read(in->f_mapping->host); in filemap_splice_read()
2939 * Once we start copying data, we don't want to be touching any in filemap_splice_read()
2942 writably_mapped = mapping_writably_mapped(in->f_mapping); in filemap_splice_read()
2960 n = min_t(loff_t, len, isize - *ppos); in filemap_splice_read()
2964 len -= n; in filemap_splice_read()
2967 in->f_ra.prev_pos = *ppos; in filemap_splice_read()
2968 if (pipe_full(pipe->head, pipe->tail, pipe->max_usage)) in filemap_splice_read()
2984 struct address_space *mapping, struct folio *folio, in folio_seek_hole_data() argument
2987 const struct address_space_operations *ops = mapping->a_ops; in folio_seek_hole_data()
2988 size_t offset, bsz = i_blocksize(mapping->host); in folio_seek_hole_data()
2992 if (!ops->is_partially_uptodate) in folio_seek_hole_data()
2998 if (unlikely(folio->mapping != mapping)) in folio_seek_hole_data()
3001 offset = offset_in_folio(folio, start) & ~(bsz - 1); in folio_seek_hole_data()
3004 if (ops->is_partially_uptodate(folio, offset, bsz) == in folio_seek_hole_data()
3007 start = (start + bsz) & ~(bsz - 1); in folio_seek_hole_data()
3024 * mapping_seek_hole_data - Seek for SEEK_DATA / SEEK_HOLE in the page cache.
3025 * @mapping: Address space to search.
3031 * contain data, your filesystem can use this function to implement
3033 * entirely memory-based such as tmpfs, and filesystems which support
3036 * Return: The requested offset on success, or -ENXIO if @whence specifies
3037 * SEEK_DATA and there is no data after @start. There is an implicit hole
3038 * after @end - 1, so SEEK_HOLE returns @end if all the bytes between @start
3039 * and @end contain data.
3041 loff_t mapping_seek_hole_data(struct address_space *mapping, loff_t start, in mapping_seek_hole_data() argument
3044 XA_STATE(xas, &mapping->i_pages, start >> PAGE_SHIFT); in mapping_seek_hole_data()
3045 pgoff_t max = (end - 1) >> PAGE_SHIFT; in mapping_seek_hole_data()
3050 return -ENXIO; in mapping_seek_hole_data()
3065 start = folio_seek_hole_data(&xas, mapping, folio, start, pos, in mapping_seek_hole_data()
3077 start = -ENXIO; in mapping_seek_hole_data()
3090 * lock_folio_maybe_drop_mmap - lock the page, possibly dropping the mmap_lock
3091 * @vmf - the vm_fault for this fault.
3092 * @folio - the folio to lock.
3093 * @fpin - the pointer to the file we may pin (or is already pinned).
3112 if (vmf->flags & FAULT_FLAG_RETRY_NOWAIT) in lock_folio_maybe_drop_mmap()
3116 if (vmf->flags & FAULT_FLAG_KILLABLE) { in lock_folio_maybe_drop_mmap()
3144 struct file *file = vmf->vma->vm_file; in do_sync_mmap_readahead()
3145 struct file_ra_state *ra = &file->f_ra; in do_sync_mmap_readahead()
3146 struct address_space *mapping = file->f_mapping; in do_sync_mmap_readahead() local
3147 DEFINE_READAHEAD(ractl, file, ra, mapping, vmf->pgoff); in do_sync_mmap_readahead()
3149 unsigned long vm_flags = vmf->vma->vm_flags; in do_sync_mmap_readahead()
3156 ractl._index &= ~((unsigned long)HPAGE_PMD_NR - 1); in do_sync_mmap_readahead()
3157 ra->size = HPAGE_PMD_NR; in do_sync_mmap_readahead()
3163 ra->size *= 2; in do_sync_mmap_readahead()
3164 ra->async_size = HPAGE_PMD_NR; in do_sync_mmap_readahead()
3170 /* If we don't want any read-ahead, don't bother */ in do_sync_mmap_readahead()
3173 if (!ra->ra_pages) in do_sync_mmap_readahead()
3178 page_cache_sync_ra(&ractl, ra->ra_pages); in do_sync_mmap_readahead()
3183 mmap_miss = READ_ONCE(ra->mmap_miss); in do_sync_mmap_readahead()
3185 WRITE_ONCE(ra->mmap_miss, ++mmap_miss); in do_sync_mmap_readahead()
3189 * stop bothering with read-ahead. It will only hurt. in do_sync_mmap_readahead()
3195 * mmap read-around in do_sync_mmap_readahead()
3198 ra->start = max_t(long, 0, vmf->pgoff - ra->ra_pages / 2); in do_sync_mmap_readahead()
3199 ra->size = ra->ra_pages; in do_sync_mmap_readahead()
3200 ra->async_size = ra->ra_pages / 4; in do_sync_mmap_readahead()
3201 ractl._index = ra->start; in do_sync_mmap_readahead()
3214 struct file *file = vmf->vma->vm_file; in do_async_mmap_readahead()
3215 struct file_ra_state *ra = &file->f_ra; in do_async_mmap_readahead()
3216 DEFINE_READAHEAD(ractl, file, ra, file->f_mapping, vmf->pgoff); in do_async_mmap_readahead()
3220 /* If we don't want any read-ahead, don't bother */ in do_async_mmap_readahead()
3221 if (vmf->vma->vm_flags & VM_RAND_READ || !ra->ra_pages) in do_async_mmap_readahead()
3224 mmap_miss = READ_ONCE(ra->mmap_miss); in do_async_mmap_readahead()
3226 WRITE_ONCE(ra->mmap_miss, --mmap_miss); in do_async_mmap_readahead()
3230 page_cache_async_ra(&ractl, folio, ra->ra_pages); in do_async_mmap_readahead()
3237 struct vm_area_struct *vma = vmf->vma; in filemap_fault_recheck_pte_none()
3251 * scenario while holding the PT lock, to not degrade non-mlocked in filemap_fault_recheck_pte_none()
3255 if (!(vma->vm_flags & VM_LOCKED)) in filemap_fault_recheck_pte_none()
3258 if (!(vmf->flags & FAULT_FLAG_ORIG_PTE_VALID)) in filemap_fault_recheck_pte_none()
3261 ptep = pte_offset_map_nolock(vma->vm_mm, vmf->pmd, vmf->address, in filemap_fault_recheck_pte_none()
3262 &vmf->ptl); in filemap_fault_recheck_pte_none()
3269 spin_lock(vmf->ptl); in filemap_fault_recheck_pte_none()
3272 spin_unlock(vmf->ptl); in filemap_fault_recheck_pte_none()
3279 * filemap_fault - read in file data for page fault handling
3283 * mapped memory region to read in file data during a page fault.
3289 * vma->vm_mm->mmap_lock must be held on entry.
3299 * Return: bitwise-OR of %VM_FAULT_ codes.
3304 struct file *file = vmf->vma->vm_file; in filemap_fault()
3306 struct address_space *mapping = file->f_mapping; in filemap_fault() local
3307 struct inode *inode = mapping->host; in filemap_fault()
3308 pgoff_t max_idx, index = vmf->pgoff; in filemap_fault()
3317 trace_mm_filemap_fault(mapping, index); in filemap_fault()
3322 folio = filemap_get_folio(mapping, index); in filemap_fault()
3328 if (!(vmf->flags & FAULT_FLAG_TRIED)) in filemap_fault()
3331 filemap_invalidate_lock_shared(mapping); in filemap_fault()
3341 count_memcg_event_mm(vmf->vma->vm_mm, PGMAJFAULT); in filemap_fault()
3350 filemap_invalidate_lock_shared(mapping); in filemap_fault()
3353 folio = __filemap_get_folio(mapping, index, in filemap_fault()
3355 vmf->gfp_mask); in filemap_fault()
3359 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3368 if (unlikely(folio->mapping != mapping)) { in filemap_fault()
3377 * that it's up-to-date. If not, it is going to be due to an error, in filemap_fault()
3403 * time to return to the upper layer and have it re-find the vma and in filemap_fault()
3411 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3424 vmf->page = folio_file_page(folio, index); in filemap_fault()
3429 * Umm, take care of errors if the page isn't up-to-date. in filemap_fault()
3430 * Try to re-read it _once_. We do this synchronously, in filemap_fault()
3435 error = filemap_read_folio(file, mapping->a_ops->read_folio, folio); in filemap_fault()
3442 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3449 * re-find the vma and come back and find our hopefully still populated in filemap_fault()
3455 filemap_invalidate_unlock_shared(mapping); in filemap_fault()
3465 struct mm_struct *mm = vmf->vma->vm_mm; in filemap_map_pmd()
3468 if (pmd_trans_huge(*vmf->pmd)) { in filemap_map_pmd()
3474 if (pmd_none(*vmf->pmd) && folio_test_pmd_mappable(folio)) { in filemap_map_pmd()
3484 if (pmd_none(*vmf->pmd) && vmf->prealloc_pte) in filemap_map_pmd()
3485 pmd_install(mm, vmf->pmd, &vmf->prealloc_pte); in filemap_map_pmd()
3491 struct address_space *mapping, pgoff_t end_pgoff) in next_uptodate_folio() argument
3514 if (folio->mapping != mapping) in next_uptodate_folio()
3518 max_idx = DIV_ROUND_UP(i_size_read(mapping->host), PAGE_SIZE); in next_uptodate_folio()
3519 if (xas->xa_index >= max_idx) in next_uptodate_folio()
3543 pte_t *old_ptep = vmf->pte; in filemap_map_folio_range()
3552 * In such situation, read-ahead is only a waste of IO. in filemap_map_folio_range()
3554 * we can stop read-ahead. in filemap_map_folio_range()
3562 * fault-around logic. in filemap_map_folio_range()
3564 if (!pte_none(ptep_get(&vmf->pte[count]))) in filemap_map_folio_range()
3574 if (in_range(vmf->address, addr, count * PAGE_SIZE)) in filemap_map_folio_range()
3580 vmf->pte += count; in filemap_map_folio_range()
3583 } while (--nr_pages > 0); in filemap_map_folio_range()
3589 if (in_range(vmf->address, addr, count * PAGE_SIZE)) in filemap_map_folio_range()
3593 vmf->pte = old_ptep; in filemap_map_folio_range()
3603 struct page *page = &folio->page; in filemap_map_order0_folio()
3615 * the fault-around logic. in filemap_map_order0_folio()
3617 if (!pte_none(ptep_get(vmf->pte))) in filemap_map_order0_folio()
3620 if (vmf->address == addr) in filemap_map_order0_folio()
3633 struct vm_area_struct *vma = vmf->vma; in filemap_map_pages()
3634 struct file *file = vma->vm_file; in filemap_map_pages()
3635 struct address_space *mapping = file->f_mapping; in filemap_map_pages() local
3638 XA_STATE(xas, &mapping->i_pages, start_pgoff); in filemap_map_pages()
3645 folio = next_uptodate_folio(&xas, mapping, end_pgoff); in filemap_map_pages()
3654 addr = vma->vm_start + ((start_pgoff - vma->vm_pgoff) << PAGE_SHIFT); in filemap_map_pages()
3655 vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd, addr, &vmf->ptl); in filemap_map_pages()
3656 if (!vmf->pte) { in filemap_map_pages()
3662 file_end = DIV_ROUND_UP(i_size_read(mapping->host), PAGE_SIZE) - 1; in filemap_map_pages()
3670 addr += (xas.xa_index - last_pgoff) << PAGE_SHIFT; in filemap_map_pages()
3671 vmf->pte += xas.xa_index - last_pgoff; in filemap_map_pages()
3673 end = folio_next_index(folio) - 1; in filemap_map_pages()
3674 nr_pages = min(end, end_pgoff) - xas.xa_index + 1; in filemap_map_pages()
3681 xas.xa_index - folio->index, addr, in filemap_map_pages()
3686 } while ((folio = next_uptodate_folio(&xas, mapping, end_pgoff)) != NULL); in filemap_map_pages()
3687 add_mm_counter(vma->vm_mm, folio_type, rss); in filemap_map_pages()
3688 pte_unmap_unlock(vmf->pte, vmf->ptl); in filemap_map_pages()
3689 trace_mm_filemap_map_pages(mapping, start_pgoff, end_pgoff); in filemap_map_pages()
3693 mmap_miss_saved = READ_ONCE(file->f_ra.mmap_miss); in filemap_map_pages()
3695 WRITE_ONCE(file->f_ra.mmap_miss, 0); in filemap_map_pages()
3697 WRITE_ONCE(file->f_ra.mmap_miss, mmap_miss_saved - mmap_miss); in filemap_map_pages()
3705 struct address_space *mapping = vmf->vma->vm_file->f_mapping; in filemap_page_mkwrite() local
3706 struct folio *folio = page_folio(vmf->page); in filemap_page_mkwrite()
3709 sb_start_pagefault(mapping->host->i_sb); in filemap_page_mkwrite()
3710 file_update_time(vmf->vma->vm_file); in filemap_page_mkwrite()
3712 if (folio->mapping != mapping) { in filemap_page_mkwrite()
3725 sb_end_pagefault(mapping->host->i_sb); in filemap_page_mkwrite()
3739 struct address_space *mapping = file->f_mapping; in generic_file_mmap() local
3741 if (!mapping->a_ops->read_folio) in generic_file_mmap()
3742 return -ENOEXEC; in generic_file_mmap()
3744 vma->vm_ops = &generic_file_vm_ops; in generic_file_mmap()
3749 * This is for filesystems which do not implement ->writepage.
3754 return -EINVAL; in generic_file_readonly_mmap()
3764 return -ENOSYS; in generic_file_mmap()
3768 return -ENOSYS; in generic_file_readonly_mmap()
3776 static struct folio *do_read_cache_folio(struct address_space *mapping, in do_read_cache_folio() argument
3783 filler = mapping->a_ops->read_folio; in do_read_cache_folio()
3785 folio = filemap_get_folio(mapping, index); in do_read_cache_folio()
3788 mapping_min_folio_order(mapping)); in do_read_cache_folio()
3790 return ERR_PTR(-ENOMEM); in do_read_cache_folio()
3791 index = mapping_align_index(mapping, index); in do_read_cache_folio()
3792 err = filemap_add_folio(mapping, folio, index, gfp); in do_read_cache_folio()
3795 if (err == -EEXIST) in do_read_cache_folio()
3811 /* Folio was truncated from mapping */ in do_read_cache_folio()
3812 if (!folio->mapping) { in do_read_cache_folio()
3839 * read_cache_folio - Read into page cache, fill it if needed.
3840 * @mapping: The address_space to read from.
3842 * @filler: Function to perform the read, or NULL to use aops->read_folio().
3851 * Context: May sleep. Expects mapping->invalidate_lock to be held.
3854 struct folio *read_cache_folio(struct address_space *mapping, pgoff_t index, in read_cache_folio() argument
3857 return do_read_cache_folio(mapping, index, filler, file, in read_cache_folio()
3858 mapping_gfp_mask(mapping)); in read_cache_folio()
3863 * mapping_read_folio_gfp - Read into page cache, using specified allocation flags.
3864 * @mapping: The address_space for the folio.
3868 * This is the same as "read_cache_folio(mapping, index, NULL, NULL)", but with
3872 * possible and so is EINTR. If ->read_folio returns another error,
3875 * The function expects mapping->invalidate_lock to be already held.
3879 struct folio *mapping_read_folio_gfp(struct address_space *mapping, in mapping_read_folio_gfp() argument
3882 return do_read_cache_folio(mapping, index, NULL, NULL, gfp); in mapping_read_folio_gfp()
3886 static struct page *do_read_cache_page(struct address_space *mapping, in do_read_cache_page() argument
3891 folio = do_read_cache_folio(mapping, index, filler, file, gfp); in do_read_cache_page()
3893 return &folio->page; in do_read_cache_page()
3897 struct page *read_cache_page(struct address_space *mapping, in read_cache_page() argument
3900 return do_read_cache_page(mapping, index, filler, file, in read_cache_page()
3901 mapping_gfp_mask(mapping)); in read_cache_page()
3906 * read_cache_page_gfp - read into page cache, using specified page allocation flags.
3907 * @mapping: the page's address_space
3911 * This is the same as "read_mapping_page(mapping, index, NULL)", but with
3914 * If the page does not get brought uptodate, return -EIO.
3916 * The function expects mapping->invalidate_lock to be already held.
3920 struct page *read_cache_page_gfp(struct address_space *mapping, in read_cache_page_gfp() argument
3924 return do_read_cache_page(mapping, index, NULL, NULL, gfp); in read_cache_page_gfp()
3937 errseq_set(&filp->f_mapping->wb_err, -EIO); in dio_warn_stale_pagecache()
3942 …pr_crit("Page cache invalidation failure on direct I/O. Possible data corruption due to collision… in dio_warn_stale_pagecache()
3943 pr_crit("File: %s PID: %d Comm: %.20s\n", path, current->pid, in dio_warn_stale_pagecache()
3944 current->comm); in dio_warn_stale_pagecache()
3950 struct address_space *mapping = iocb->ki_filp->f_mapping; in kiocb_invalidate_post_direct_write() local
3952 if (mapping->nrpages && in kiocb_invalidate_post_direct_write()
3953 invalidate_inode_pages2_range(mapping, in kiocb_invalidate_post_direct_write()
3954 iocb->ki_pos >> PAGE_SHIFT, in kiocb_invalidate_post_direct_write()
3955 (iocb->ki_pos + count - 1) >> PAGE_SHIFT)) in kiocb_invalidate_post_direct_write()
3956 dio_warn_stale_pagecache(iocb->ki_filp); in kiocb_invalidate_post_direct_write()
3962 struct address_space *mapping = iocb->ki_filp->f_mapping; in generic_file_direct_write() local
3972 if (written == -EBUSY) in generic_file_direct_write()
3977 written = mapping->a_ops->direct_IO(iocb, from); in generic_file_direct_write()
3981 * cached by non-direct readahead, or faulted in by get_user_pages() in generic_file_direct_write()
3994 * Skip invalidation for async writes or if mapping has no pages. in generic_file_direct_write()
3997 struct inode *inode = mapping->host; in generic_file_direct_write()
3998 loff_t pos = iocb->ki_pos; in generic_file_direct_write()
4002 write_len -= written; in generic_file_direct_write()
4003 if (pos > i_size_read(inode) && !S_ISBLK(inode->i_mode)) { in generic_file_direct_write()
4007 iocb->ki_pos = pos; in generic_file_direct_write()
4009 if (written != -EIOCBQUEUED) in generic_file_direct_write()
4010 iov_iter_revert(from, write_len - iov_iter_count(from)); in generic_file_direct_write()
4017 struct file *file = iocb->ki_filp; in generic_perform_write()
4018 loff_t pos = iocb->ki_pos; in generic_perform_write()
4019 struct address_space *mapping = file->f_mapping; in generic_perform_write() local
4020 const struct address_space_operations *a_ops = mapping->a_ops; in generic_perform_write()
4021 size_t chunk = mapping_max_folio_size(mapping); in generic_perform_write()
4034 offset = pos & (chunk - 1); in generic_perform_write()
4035 bytes = min(chunk - offset, bytes); in generic_perform_write()
4036 balance_dirty_pages_ratelimited(mapping); in generic_perform_write()
4042 * up-to-date. in generic_perform_write()
4045 status = -EFAULT; in generic_perform_write()
4050 status = -EINTR; in generic_perform_write()
4054 status = a_ops->write_begin(file, mapping, pos, bytes, in generic_perform_write()
4060 if (bytes > folio_size(folio) - offset) in generic_perform_write()
4061 bytes = folio_size(folio) - offset; in generic_perform_write()
4063 if (mapping_writably_mapped(mapping)) in generic_perform_write()
4069 status = a_ops->write_end(file, mapping, pos, bytes, copied, in generic_perform_write()
4072 iov_iter_revert(i, copied - max(status, 0L)); in generic_perform_write()
4080 * A short copy made ->write_end() reject the in generic_perform_write()
4099 iocb->ki_pos += written; in generic_perform_write()
4105 * __generic_file_write_iter - write data to a file
4107 * @from: iov_iter with data to write
4109 * This function does all the work needed for actually writing data to a
4117 * This function does *not* take care of syncing data in case of O_SYNC write.
4123 * * negative error code if no data has been written at all
4127 struct file *file = iocb->ki_filp; in __generic_file_write_iter()
4128 struct address_space *mapping = file->f_mapping; in __generic_file_write_iter() local
4129 struct inode *inode = mapping->host; in __generic_file_write_iter()
4140 if (iocb->ki_flags & IOCB_DIRECT) { in __generic_file_write_iter()
4147 * page-cache pages correctly). in __generic_file_write_iter()
4160 * generic_file_write_iter - write data to a file
4162 * @from: iov_iter with data to write
4168 * * negative error code if no data has been written at all of
4174 struct file *file = iocb->ki_filp; in generic_file_write_iter()
4175 struct inode *inode = file->f_mapping->host; in generic_file_write_iter()
4191 * filemap_release_folio() - Release fs-specific metadata on a folio.
4195 * The address_space is trying to release any data attached to a folio
4196 * (presumably at folio->private).
4209 struct address_space * const mapping = folio->mapping; in filemap_release_folio() local
4217 if (mapping && mapping->a_ops->release_folio) in filemap_release_folio()
4218 return mapping->a_ops->release_folio(folio, gfp); in filemap_release_folio()
4224 * filemap_invalidate_inode - Invalidate/forcibly write back a range of an inode's pagecache
4239 struct address_space *mapping = inode->i_mapping; in filemap_invalidate_inode() local
4242 pgoff_t nr = end == LLONG_MAX ? ULONG_MAX : last - first + 1; in filemap_invalidate_inode()
4244 if (!mapping || !mapping->nrpages || end < start) in filemap_invalidate_inode()
4248 filemap_invalidate_lock(mapping); in filemap_invalidate_inode()
4250 if (!mapping->nrpages) in filemap_invalidate_inode()
4253 unmap_mapping_pages(mapping, first, nr, false); in filemap_invalidate_inode()
4255 /* Write back the data if we're asked to. */ in filemap_invalidate_inode()
4264 filemap_fdatawrite_wbc(mapping, &wbc); in filemap_invalidate_inode()
4268 invalidate_inode_pages2_range(mapping, start / PAGE_SIZE, end / PAGE_SIZE); in filemap_invalidate_inode()
4271 filemap_invalidate_unlock(mapping); in filemap_invalidate_inode()
4273 return filemap_check_errors(mapping); in filemap_invalidate_inode()
4279 * filemap_cachestat() - compute the page cache statistics of a mapping
4280 * @mapping: The mapping to compute the statistics for.
4285 * This will query the page cache statistics of a mapping in the
4290 static void filemap_cachestat(struct address_space *mapping, in filemap_cachestat() argument
4293 XA_STATE(xas, &mapping->i_pages, first_index); in filemap_cachestat()
4313 * the rcu-protected xarray. in filemap_cachestat()
4322 folio_last_index = folio_first_index + nr_pages - 1; in filemap_cachestat()
4326 nr_pages -= first_index - folio_first_index; in filemap_cachestat()
4329 nr_pages -= folio_last_index - last_index; in filemap_cachestat()
4336 cs->nr_evicted += nr_pages; in filemap_cachestat()
4339 if (shmem_mapping(mapping)) { in filemap_cachestat()
4340 /* shmem file - in swap cache */ in filemap_cachestat()
4363 cs->nr_recently_evicted += nr_pages; in filemap_cachestat()
4369 cs->nr_cache += nr_pages; in filemap_cachestat()
4372 cs->nr_dirty += nr_pages; in filemap_cachestat()
4375 cs->nr_writeback += nr_pages; in filemap_cachestat()
4400 * `off` and `len` must be non-negative integers. If `len` > 0,
4414 * zero - success
4415 * -EFAULT - cstat or cstat_range points to an illegal address
4416 * -EINVAL - invalid flags
4417 * -EBADF - invalid file descriptor
4418 * -EOPNOTSUPP - file descriptor is of a hugetlbfs file
4425 struct address_space *mapping; in SYSCALL_DEFINE4() local
4431 return -EBADF; in SYSCALL_DEFINE4()
4436 return -EFAULT; in SYSCALL_DEFINE4()
4442 return -EOPNOTSUPP; in SYSCALL_DEFINE4()
4447 return -EINVAL; in SYSCALL_DEFINE4()
4452 csr.len == 0 ? ULONG_MAX : (csr.off + csr.len - 1) >> PAGE_SHIFT; in SYSCALL_DEFINE4()
4454 mapping = fd_file(f)->f_mapping; in SYSCALL_DEFINE4()
4455 filemap_cachestat(mapping, first_index, last_index, &cs); in SYSCALL_DEFINE4()
4459 return -EFAULT; in SYSCALL_DEFINE4()