1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2008 Oracle. All rights reserved.
6 #include <linux/sched.h>
7 #include <linux/slab.h>
8 #include <linux/blkdev.h>
9 #include <linux/list_sort.h>
10 #include <linux/iversion.h>
16 #include "print-tree.h"
18 #include "compression.h"
20 #include "block-group.h"
21 #include "space-info.h"
23 #include "inode-item.h"
25 /* magic values for the inode_only field in btrfs_log_inode:
27 * LOG_INODE_ALL means to log everything
28 * LOG_INODE_EXISTS means to log just enough to recreate the inode
39 * directory trouble cases
41 * 1) on rename or unlink, if the inode being unlinked isn't in the fsync
42 * log, we must force a full commit before doing an fsync of the directory
43 * where the unlink was done.
44 * ---> record transid of last unlink/rename per directory
48 * rename foo/some_dir foo2/some_dir
50 * fsync foo/some_dir/some_file
52 * The fsync above will unlink the original some_dir without recording
53 * it in its new location (foo2). After a crash, some_dir will be gone
54 * unless the fsync of some_file forces a full commit
56 * 2) we must log any new names for any file or dir that is in the fsync
57 * log. ---> check inode while renaming/linking.
59 * 2a) we must log any new names for any file or dir during rename
60 * when the directory they are being removed from was logged.
61 * ---> check inode and old parent dir during rename
63 * 2a is actually the more important variant. With the extra logging
64 * a crash might unlink the old name without recreating the new one
66 * 3) after a crash, we must go through any directories with a link count
67 * of zero and redo the rm -rf
74 * The directory f1 was fully removed from the FS, but fsync was never
75 * called on f1, only its parent dir. After a crash the rm -rf must
76 * be replayed. This must be able to recurse down the entire
77 * directory tree. The inode link count fixup code takes care of the
82 * stages for the tree walking. The first
83 * stage (0) is to only pin down the blocks we find
84 * the second stage (1) is to make sure that all the inodes
85 * we find in the log are created in the subvolume.
87 * The last stage is to deal with directories and links and extents
88 * and all the other fun semantics
92 LOG_WALK_REPLAY_INODES,
93 LOG_WALK_REPLAY_DIR_INDEX,
97 static int btrfs_log_inode(struct btrfs_trans_handle *trans,
98 struct btrfs_inode *inode,
100 struct btrfs_log_ctx *ctx);
101 static int link_to_fixup_dir(struct btrfs_trans_handle *trans,
102 struct btrfs_root *root,
103 struct btrfs_path *path, u64 objectid);
104 static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
105 struct btrfs_root *root,
106 struct btrfs_root *log,
107 struct btrfs_path *path,
108 u64 dirid, int del_all);
109 static void wait_log_commit(struct btrfs_root *root, int transid);
112 * tree logging is a special write ahead log used to make sure that
113 * fsyncs and O_SYNCs can happen without doing full tree commits.
115 * Full tree commits are expensive because they require commonly
116 * modified blocks to be recowed, creating many dirty pages in the
117 * extent tree an 4x-6x higher write load than ext3.
119 * Instead of doing a tree commit on every fsync, we use the
120 * key ranges and transaction ids to find items for a given file or directory
121 * that have changed in this transaction. Those items are copied into
122 * a special tree (one per subvolume root), that tree is written to disk
123 * and then the fsync is considered complete.
125 * After a crash, items are copied out of the log-tree back into the
126 * subvolume tree. Any file data extents found are recorded in the extent
127 * allocation tree, and the log-tree freed.
129 * The log tree is read three times, once to pin down all the extents it is
130 * using in ram and once, once to create all the inodes logged in the tree
131 * and once to do all the other items.
135 * start a sub transaction and setup the log tree
136 * this increments the log tree writer count to make the people
137 * syncing the tree wait for us to finish
139 static int start_log_trans(struct btrfs_trans_handle *trans,
140 struct btrfs_root *root,
141 struct btrfs_log_ctx *ctx)
143 struct btrfs_fs_info *fs_info = root->fs_info;
144 struct btrfs_root *tree_root = fs_info->tree_root;
145 const bool zoned = btrfs_is_zoned(fs_info);
147 bool created = false;
150 * First check if the log root tree was already created. If not, create
151 * it before locking the root's log_mutex, just to keep lockdep happy.
153 if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &tree_root->state)) {
154 mutex_lock(&tree_root->log_mutex);
155 if (!fs_info->log_root_tree) {
156 ret = btrfs_init_log_root_tree(trans, fs_info);
158 set_bit(BTRFS_ROOT_HAS_LOG_TREE, &tree_root->state);
162 mutex_unlock(&tree_root->log_mutex);
167 mutex_lock(&root->log_mutex);
170 if (root->log_root) {
171 int index = (root->log_transid + 1) % 2;
173 if (btrfs_need_log_full_commit(trans)) {
174 ret = BTRFS_LOG_FORCE_COMMIT;
178 if (zoned && atomic_read(&root->log_commit[index])) {
179 wait_log_commit(root, root->log_transid - 1);
183 if (!root->log_start_pid) {
184 clear_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
185 root->log_start_pid = current->pid;
186 } else if (root->log_start_pid != current->pid) {
187 set_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
191 * This means fs_info->log_root_tree was already created
192 * for some other FS trees. Do the full commit not to mix
193 * nodes from multiple log transactions to do sequential
196 if (zoned && !created) {
197 ret = BTRFS_LOG_FORCE_COMMIT;
201 ret = btrfs_add_log_tree(trans, root);
205 set_bit(BTRFS_ROOT_HAS_LOG_TREE, &root->state);
206 clear_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
207 root->log_start_pid = current->pid;
210 atomic_inc(&root->log_writers);
211 if (!ctx->logging_new_name) {
212 int index = root->log_transid % 2;
213 list_add_tail(&ctx->list, &root->log_ctxs[index]);
214 ctx->log_transid = root->log_transid;
218 mutex_unlock(&root->log_mutex);
223 * returns 0 if there was a log transaction running and we were able
224 * to join, or returns -ENOENT if there were not transactions
227 static int join_running_log_trans(struct btrfs_root *root)
229 const bool zoned = btrfs_is_zoned(root->fs_info);
232 if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &root->state))
235 mutex_lock(&root->log_mutex);
237 if (root->log_root) {
238 int index = (root->log_transid + 1) % 2;
241 if (zoned && atomic_read(&root->log_commit[index])) {
242 wait_log_commit(root, root->log_transid - 1);
245 atomic_inc(&root->log_writers);
247 mutex_unlock(&root->log_mutex);
252 * This either makes the current running log transaction wait
253 * until you call btrfs_end_log_trans() or it makes any future
254 * log transactions wait until you call btrfs_end_log_trans()
256 void btrfs_pin_log_trans(struct btrfs_root *root)
258 atomic_inc(&root->log_writers);
262 * indicate we're done making changes to the log tree
263 * and wake up anyone waiting to do a sync
265 void btrfs_end_log_trans(struct btrfs_root *root)
267 if (atomic_dec_and_test(&root->log_writers)) {
268 /* atomic_dec_and_test implies a barrier */
269 cond_wake_up_nomb(&root->log_writer_wait);
273 static void btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
275 filemap_fdatawait_range(buf->pages[0]->mapping,
276 buf->start, buf->start + buf->len - 1);
280 * the walk control struct is used to pass state down the chain when
281 * processing the log tree. The stage field tells us which part
282 * of the log tree processing we are currently doing. The others
283 * are state fields used for that specific part
285 struct walk_control {
286 /* should we free the extent on disk when done? This is used
287 * at transaction commit time while freeing a log tree
291 /* pin only walk, we record which extents on disk belong to the
296 /* what stage of the replay code we're currently in */
300 * Ignore any items from the inode currently being processed. Needs
301 * to be set every time we find a BTRFS_INODE_ITEM_KEY and we are in
302 * the LOG_WALK_REPLAY_INODES stage.
304 bool ignore_cur_inode;
306 /* the root we are currently replaying */
307 struct btrfs_root *replay_dest;
309 /* the trans handle for the current replay */
310 struct btrfs_trans_handle *trans;
312 /* the function that gets used to process blocks we find in the
313 * tree. Note the extent_buffer might not be up to date when it is
314 * passed in, and it must be checked or read if you need the data
317 int (*process_func)(struct btrfs_root *log, struct extent_buffer *eb,
318 struct walk_control *wc, u64 gen, int level);
322 * process_func used to pin down extents, write them or wait on them
324 static int process_one_buffer(struct btrfs_root *log,
325 struct extent_buffer *eb,
326 struct walk_control *wc, u64 gen, int level)
328 struct btrfs_fs_info *fs_info = log->fs_info;
332 * If this fs is mixed then we need to be able to process the leaves to
333 * pin down any logged extents, so we have to read the block.
335 if (btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
336 ret = btrfs_read_extent_buffer(eb, gen, level, NULL);
342 ret = btrfs_pin_extent_for_log_replay(wc->trans, eb->start,
347 if (btrfs_buffer_uptodate(eb, gen, 0) &&
348 btrfs_header_level(eb) == 0)
349 ret = btrfs_exclude_logged_extents(eb);
354 static int do_overwrite_item(struct btrfs_trans_handle *trans,
355 struct btrfs_root *root,
356 struct btrfs_path *path,
357 struct extent_buffer *eb, int slot,
358 struct btrfs_key *key)
362 u64 saved_i_size = 0;
363 int save_old_i_size = 0;
364 unsigned long src_ptr;
365 unsigned long dst_ptr;
366 int overwrite_root = 0;
367 bool inode_item = key->type == BTRFS_INODE_ITEM_KEY;
369 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
372 item_size = btrfs_item_size(eb, slot);
373 src_ptr = btrfs_item_ptr_offset(eb, slot);
375 /* Our caller must have done a search for the key for us. */
376 ASSERT(path->nodes[0] != NULL);
379 * And the slot must point to the exact key or the slot where the key
380 * should be at (the first item with a key greater than 'key')
382 if (path->slots[0] < btrfs_header_nritems(path->nodes[0])) {
383 struct btrfs_key found_key;
385 btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
386 ret = btrfs_comp_cpu_keys(&found_key, key);
395 u32 dst_size = btrfs_item_size(path->nodes[0],
397 if (dst_size != item_size)
400 if (item_size == 0) {
401 btrfs_release_path(path);
404 dst_copy = kmalloc(item_size, GFP_NOFS);
405 src_copy = kmalloc(item_size, GFP_NOFS);
406 if (!dst_copy || !src_copy) {
407 btrfs_release_path(path);
413 read_extent_buffer(eb, src_copy, src_ptr, item_size);
415 dst_ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
416 read_extent_buffer(path->nodes[0], dst_copy, dst_ptr,
418 ret = memcmp(dst_copy, src_copy, item_size);
423 * they have the same contents, just return, this saves
424 * us from cowing blocks in the destination tree and doing
425 * extra writes that may not have been done by a previous
429 btrfs_release_path(path);
434 * We need to load the old nbytes into the inode so when we
435 * replay the extents we've logged we get the right nbytes.
438 struct btrfs_inode_item *item;
442 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
443 struct btrfs_inode_item);
444 nbytes = btrfs_inode_nbytes(path->nodes[0], item);
445 item = btrfs_item_ptr(eb, slot,
446 struct btrfs_inode_item);
447 btrfs_set_inode_nbytes(eb, item, nbytes);
450 * If this is a directory we need to reset the i_size to
451 * 0 so that we can set it up properly when replaying
452 * the rest of the items in this log.
454 mode = btrfs_inode_mode(eb, item);
456 btrfs_set_inode_size(eb, item, 0);
458 } else if (inode_item) {
459 struct btrfs_inode_item *item;
463 * New inode, set nbytes to 0 so that the nbytes comes out
464 * properly when we replay the extents.
466 item = btrfs_item_ptr(eb, slot, struct btrfs_inode_item);
467 btrfs_set_inode_nbytes(eb, item, 0);
470 * If this is a directory we need to reset the i_size to 0 so
471 * that we can set it up properly when replaying the rest of
472 * the items in this log.
474 mode = btrfs_inode_mode(eb, item);
476 btrfs_set_inode_size(eb, item, 0);
479 btrfs_release_path(path);
480 /* try to insert the key into the destination tree */
481 path->skip_release_on_error = 1;
482 ret = btrfs_insert_empty_item(trans, root, path,
484 path->skip_release_on_error = 0;
486 /* make sure any existing item is the correct size */
487 if (ret == -EEXIST || ret == -EOVERFLOW) {
489 found_size = btrfs_item_size(path->nodes[0],
491 if (found_size > item_size)
492 btrfs_truncate_item(path, item_size, 1);
493 else if (found_size < item_size)
494 btrfs_extend_item(path, item_size - found_size);
498 dst_ptr = btrfs_item_ptr_offset(path->nodes[0],
501 /* don't overwrite an existing inode if the generation number
502 * was logged as zero. This is done when the tree logging code
503 * is just logging an inode to make sure it exists after recovery.
505 * Also, don't overwrite i_size on directories during replay.
506 * log replay inserts and removes directory items based on the
507 * state of the tree found in the subvolume, and i_size is modified
510 if (key->type == BTRFS_INODE_ITEM_KEY && ret == -EEXIST) {
511 struct btrfs_inode_item *src_item;
512 struct btrfs_inode_item *dst_item;
514 src_item = (struct btrfs_inode_item *)src_ptr;
515 dst_item = (struct btrfs_inode_item *)dst_ptr;
517 if (btrfs_inode_generation(eb, src_item) == 0) {
518 struct extent_buffer *dst_eb = path->nodes[0];
519 const u64 ino_size = btrfs_inode_size(eb, src_item);
522 * For regular files an ino_size == 0 is used only when
523 * logging that an inode exists, as part of a directory
524 * fsync, and the inode wasn't fsynced before. In this
525 * case don't set the size of the inode in the fs/subvol
526 * tree, otherwise we would be throwing valid data away.
528 if (S_ISREG(btrfs_inode_mode(eb, src_item)) &&
529 S_ISREG(btrfs_inode_mode(dst_eb, dst_item)) &&
531 btrfs_set_inode_size(dst_eb, dst_item, ino_size);
535 if (overwrite_root &&
536 S_ISDIR(btrfs_inode_mode(eb, src_item)) &&
537 S_ISDIR(btrfs_inode_mode(path->nodes[0], dst_item))) {
539 saved_i_size = btrfs_inode_size(path->nodes[0],
544 copy_extent_buffer(path->nodes[0], eb, dst_ptr,
547 if (save_old_i_size) {
548 struct btrfs_inode_item *dst_item;
549 dst_item = (struct btrfs_inode_item *)dst_ptr;
550 btrfs_set_inode_size(path->nodes[0], dst_item, saved_i_size);
553 /* make sure the generation is filled in */
554 if (key->type == BTRFS_INODE_ITEM_KEY) {
555 struct btrfs_inode_item *dst_item;
556 dst_item = (struct btrfs_inode_item *)dst_ptr;
557 if (btrfs_inode_generation(path->nodes[0], dst_item) == 0) {
558 btrfs_set_inode_generation(path->nodes[0], dst_item,
563 btrfs_mark_buffer_dirty(path->nodes[0]);
564 btrfs_release_path(path);
569 * Item overwrite used by replay and tree logging. eb, slot and key all refer
570 * to the src data we are copying out.
572 * root is the tree we are copying into, and path is a scratch
573 * path for use in this function (it should be released on entry and
574 * will be released on exit).
576 * If the key is already in the destination tree the existing item is
577 * overwritten. If the existing item isn't big enough, it is extended.
578 * If it is too large, it is truncated.
580 * If the key isn't in the destination yet, a new item is inserted.
582 static int overwrite_item(struct btrfs_trans_handle *trans,
583 struct btrfs_root *root,
584 struct btrfs_path *path,
585 struct extent_buffer *eb, int slot,
586 struct btrfs_key *key)
590 /* Look for the key in the destination tree. */
591 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
595 return do_overwrite_item(trans, root, path, eb, slot, key);
599 * simple helper to read an inode off the disk from a given root
600 * This can only be called for subvolume roots and not for the log
602 static noinline struct inode *read_one_inode(struct btrfs_root *root,
607 inode = btrfs_iget(root->fs_info->sb, objectid, root);
613 /* replays a single extent in 'eb' at 'slot' with 'key' into the
614 * subvolume 'root'. path is released on entry and should be released
617 * extents in the log tree have not been allocated out of the extent
618 * tree yet. So, this completes the allocation, taking a reference
619 * as required if the extent already exists or creating a new extent
620 * if it isn't in the extent allocation tree yet.
622 * The extent is inserted into the file, dropping any existing extents
623 * from the file that overlap the new one.
625 static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
626 struct btrfs_root *root,
627 struct btrfs_path *path,
628 struct extent_buffer *eb, int slot,
629 struct btrfs_key *key)
631 struct btrfs_drop_extents_args drop_args = { 0 };
632 struct btrfs_fs_info *fs_info = root->fs_info;
635 u64 start = key->offset;
637 struct btrfs_file_extent_item *item;
638 struct inode *inode = NULL;
642 item = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
643 found_type = btrfs_file_extent_type(eb, item);
645 if (found_type == BTRFS_FILE_EXTENT_REG ||
646 found_type == BTRFS_FILE_EXTENT_PREALLOC) {
647 nbytes = btrfs_file_extent_num_bytes(eb, item);
648 extent_end = start + nbytes;
651 * We don't add to the inodes nbytes if we are prealloc or a
654 if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
656 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
657 size = btrfs_file_extent_ram_bytes(eb, item);
658 nbytes = btrfs_file_extent_ram_bytes(eb, item);
659 extent_end = ALIGN(start + size,
660 fs_info->sectorsize);
666 inode = read_one_inode(root, key->objectid);
673 * first check to see if we already have this extent in the
674 * file. This must be done before the btrfs_drop_extents run
675 * so we don't try to drop this extent.
677 ret = btrfs_lookup_file_extent(trans, root, path,
678 btrfs_ino(BTRFS_I(inode)), start, 0);
681 (found_type == BTRFS_FILE_EXTENT_REG ||
682 found_type == BTRFS_FILE_EXTENT_PREALLOC)) {
683 struct btrfs_file_extent_item cmp1;
684 struct btrfs_file_extent_item cmp2;
685 struct btrfs_file_extent_item *existing;
686 struct extent_buffer *leaf;
688 leaf = path->nodes[0];
689 existing = btrfs_item_ptr(leaf, path->slots[0],
690 struct btrfs_file_extent_item);
692 read_extent_buffer(eb, &cmp1, (unsigned long)item,
694 read_extent_buffer(leaf, &cmp2, (unsigned long)existing,
698 * we already have a pointer to this exact extent,
699 * we don't have to do anything
701 if (memcmp(&cmp1, &cmp2, sizeof(cmp1)) == 0) {
702 btrfs_release_path(path);
706 btrfs_release_path(path);
708 /* drop any overlapping extents */
709 drop_args.start = start;
710 drop_args.end = extent_end;
711 drop_args.drop_cache = true;
712 ret = btrfs_drop_extents(trans, root, BTRFS_I(inode), &drop_args);
716 if (found_type == BTRFS_FILE_EXTENT_REG ||
717 found_type == BTRFS_FILE_EXTENT_PREALLOC) {
719 unsigned long dest_offset;
720 struct btrfs_key ins;
722 if (btrfs_file_extent_disk_bytenr(eb, item) == 0 &&
723 btrfs_fs_incompat(fs_info, NO_HOLES))
726 ret = btrfs_insert_empty_item(trans, root, path, key,
730 dest_offset = btrfs_item_ptr_offset(path->nodes[0],
732 copy_extent_buffer(path->nodes[0], eb, dest_offset,
733 (unsigned long)item, sizeof(*item));
735 ins.objectid = btrfs_file_extent_disk_bytenr(eb, item);
736 ins.offset = btrfs_file_extent_disk_num_bytes(eb, item);
737 ins.type = BTRFS_EXTENT_ITEM_KEY;
738 offset = key->offset - btrfs_file_extent_offset(eb, item);
741 * Manually record dirty extent, as here we did a shallow
742 * file extent item copy and skip normal backref update,
743 * but modifying extent tree all by ourselves.
744 * So need to manually record dirty extent for qgroup,
745 * as the owner of the file extent changed from log tree
746 * (doesn't affect qgroup) to fs/file tree(affects qgroup)
748 ret = btrfs_qgroup_trace_extent(trans,
749 btrfs_file_extent_disk_bytenr(eb, item),
750 btrfs_file_extent_disk_num_bytes(eb, item),
755 if (ins.objectid > 0) {
756 struct btrfs_ref ref = { 0 };
759 LIST_HEAD(ordered_sums);
762 * is this extent already allocated in the extent
763 * allocation tree? If so, just add a reference
765 ret = btrfs_lookup_data_extent(fs_info, ins.objectid,
769 } else if (ret == 0) {
770 btrfs_init_generic_ref(&ref,
771 BTRFS_ADD_DELAYED_REF,
772 ins.objectid, ins.offset, 0);
773 btrfs_init_data_ref(&ref,
774 root->root_key.objectid,
775 key->objectid, offset, 0, false);
776 ret = btrfs_inc_extent_ref(trans, &ref);
781 * insert the extent pointer in the extent
784 ret = btrfs_alloc_logged_file_extent(trans,
785 root->root_key.objectid,
786 key->objectid, offset, &ins);
790 btrfs_release_path(path);
792 if (btrfs_file_extent_compression(eb, item)) {
793 csum_start = ins.objectid;
794 csum_end = csum_start + ins.offset;
796 csum_start = ins.objectid +
797 btrfs_file_extent_offset(eb, item);
798 csum_end = csum_start +
799 btrfs_file_extent_num_bytes(eb, item);
802 ret = btrfs_lookup_csums_range(root->log_root,
803 csum_start, csum_end - 1,
808 * Now delete all existing cums in the csum root that
809 * cover our range. We do this because we can have an
810 * extent that is completely referenced by one file
811 * extent item and partially referenced by another
812 * file extent item (like after using the clone or
813 * extent_same ioctls). In this case if we end up doing
814 * the replay of the one that partially references the
815 * extent first, and we do not do the csum deletion
816 * below, we can get 2 csum items in the csum tree that
817 * overlap each other. For example, imagine our log has
818 * the two following file extent items:
820 * key (257 EXTENT_DATA 409600)
821 * extent data disk byte 12845056 nr 102400
822 * extent data offset 20480 nr 20480 ram 102400
824 * key (257 EXTENT_DATA 819200)
825 * extent data disk byte 12845056 nr 102400
826 * extent data offset 0 nr 102400 ram 102400
828 * Where the second one fully references the 100K extent
829 * that starts at disk byte 12845056, and the log tree
830 * has a single csum item that covers the entire range
833 * key (EXTENT_CSUM EXTENT_CSUM 12845056) itemsize 100
835 * After the first file extent item is replayed, the
836 * csum tree gets the following csum item:
838 * key (EXTENT_CSUM EXTENT_CSUM 12865536) itemsize 20
840 * Which covers the 20K sub-range starting at offset 20K
841 * of our extent. Now when we replay the second file
842 * extent item, if we do not delete existing csum items
843 * that cover any of its blocks, we end up getting two
844 * csum items in our csum tree that overlap each other:
846 * key (EXTENT_CSUM EXTENT_CSUM 12845056) itemsize 100
847 * key (EXTENT_CSUM EXTENT_CSUM 12865536) itemsize 20
849 * Which is a problem, because after this anyone trying
850 * to lookup up for the checksum of any block of our
851 * extent starting at an offset of 40K or higher, will
852 * end up looking at the second csum item only, which
853 * does not contain the checksum for any block starting
854 * at offset 40K or higher of our extent.
856 while (!list_empty(&ordered_sums)) {
857 struct btrfs_ordered_sum *sums;
858 struct btrfs_root *csum_root;
860 sums = list_entry(ordered_sums.next,
861 struct btrfs_ordered_sum,
863 csum_root = btrfs_csum_root(fs_info,
866 ret = btrfs_del_csums(trans, csum_root,
870 ret = btrfs_csum_file_blocks(trans,
873 list_del(&sums->list);
879 btrfs_release_path(path);
881 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
882 /* inline extents are easy, we just overwrite them */
883 ret = overwrite_item(trans, root, path, eb, slot, key);
888 ret = btrfs_inode_set_file_extent_range(BTRFS_I(inode), start,
894 btrfs_update_inode_bytes(BTRFS_I(inode), nbytes, drop_args.bytes_found);
895 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
901 static int unlink_inode_for_log_replay(struct btrfs_trans_handle *trans,
902 struct btrfs_inode *dir,
903 struct btrfs_inode *inode,
909 ret = btrfs_unlink_inode(trans, dir, inode, name, name_len);
913 * Whenever we need to check if a name exists or not, we check the
914 * fs/subvolume tree. So after an unlink we must run delayed items, so
915 * that future checks for a name during log replay see that the name
916 * does not exists anymore.
918 return btrfs_run_delayed_items(trans);
922 * when cleaning up conflicts between the directory names in the
923 * subvolume, directory names in the log and directory names in the
924 * inode back references, we may have to unlink inodes from directories.
926 * This is a helper function to do the unlink of a specific directory
929 static noinline int drop_one_dir_item(struct btrfs_trans_handle *trans,
930 struct btrfs_path *path,
931 struct btrfs_inode *dir,
932 struct btrfs_dir_item *di)
934 struct btrfs_root *root = dir->root;
938 struct extent_buffer *leaf;
939 struct btrfs_key location;
942 leaf = path->nodes[0];
944 btrfs_dir_item_key_to_cpu(leaf, di, &location);
945 name_len = btrfs_dir_name_len(leaf, di);
946 name = kmalloc(name_len, GFP_NOFS);
950 read_extent_buffer(leaf, name, (unsigned long)(di + 1), name_len);
951 btrfs_release_path(path);
953 inode = read_one_inode(root, location.objectid);
959 ret = link_to_fixup_dir(trans, root, path, location.objectid);
963 ret = unlink_inode_for_log_replay(trans, dir, BTRFS_I(inode), name,
972 * See if a given name and sequence number found in an inode back reference are
973 * already in a directory and correctly point to this inode.
975 * Returns: < 0 on error, 0 if the directory entry does not exists and 1 if it
978 static noinline int inode_in_dir(struct btrfs_root *root,
979 struct btrfs_path *path,
980 u64 dirid, u64 objectid, u64 index,
981 const char *name, int name_len)
983 struct btrfs_dir_item *di;
984 struct btrfs_key location;
987 di = btrfs_lookup_dir_index_item(NULL, root, path, dirid,
988 index, name, name_len, 0);
993 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
994 if (location.objectid != objectid)
1000 btrfs_release_path(path);
1001 di = btrfs_lookup_dir_item(NULL, root, path, dirid, name, name_len, 0);
1006 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1007 if (location.objectid == objectid)
1011 btrfs_release_path(path);
1016 * helper function to check a log tree for a named back reference in
1017 * an inode. This is used to decide if a back reference that is
1018 * found in the subvolume conflicts with what we find in the log.
1020 * inode backreferences may have multiple refs in a single item,
1021 * during replay we process one reference at a time, and we don't
1022 * want to delete valid links to a file from the subvolume if that
1023 * link is also in the log.
1025 static noinline int backref_in_log(struct btrfs_root *log,
1026 struct btrfs_key *key,
1028 const char *name, int namelen)
1030 struct btrfs_path *path;
1033 path = btrfs_alloc_path();
1037 ret = btrfs_search_slot(NULL, log, key, path, 0, 0);
1040 } else if (ret == 1) {
1045 if (key->type == BTRFS_INODE_EXTREF_KEY)
1046 ret = !!btrfs_find_name_in_ext_backref(path->nodes[0],
1051 ret = !!btrfs_find_name_in_backref(path->nodes[0],
1055 btrfs_free_path(path);
1059 static inline int __add_inode_ref(struct btrfs_trans_handle *trans,
1060 struct btrfs_root *root,
1061 struct btrfs_path *path,
1062 struct btrfs_root *log_root,
1063 struct btrfs_inode *dir,
1064 struct btrfs_inode *inode,
1065 u64 inode_objectid, u64 parent_objectid,
1066 u64 ref_index, char *name, int namelen,
1071 int victim_name_len;
1072 struct extent_buffer *leaf;
1073 struct btrfs_dir_item *di;
1074 struct btrfs_key search_key;
1075 struct btrfs_inode_extref *extref;
1078 /* Search old style refs */
1079 search_key.objectid = inode_objectid;
1080 search_key.type = BTRFS_INODE_REF_KEY;
1081 search_key.offset = parent_objectid;
1082 ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
1084 struct btrfs_inode_ref *victim_ref;
1086 unsigned long ptr_end;
1088 leaf = path->nodes[0];
1090 /* are we trying to overwrite a back ref for the root directory
1091 * if so, just jump out, we're done
1093 if (search_key.objectid == search_key.offset)
1096 /* check all the names in this back reference to see
1097 * if they are in the log. if so, we allow them to stay
1098 * otherwise they must be unlinked as a conflict
1100 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1101 ptr_end = ptr + btrfs_item_size(leaf, path->slots[0]);
1102 while (ptr < ptr_end) {
1103 victim_ref = (struct btrfs_inode_ref *)ptr;
1104 victim_name_len = btrfs_inode_ref_name_len(leaf,
1106 victim_name = kmalloc(victim_name_len, GFP_NOFS);
1110 read_extent_buffer(leaf, victim_name,
1111 (unsigned long)(victim_ref + 1),
1114 ret = backref_in_log(log_root, &search_key,
1115 parent_objectid, victim_name,
1121 inc_nlink(&inode->vfs_inode);
1122 btrfs_release_path(path);
1124 ret = unlink_inode_for_log_replay(trans, dir, inode,
1125 victim_name, victim_name_len);
1134 ptr = (unsigned long)(victim_ref + 1) + victim_name_len;
1138 * NOTE: we have searched root tree and checked the
1139 * corresponding ref, it does not need to check again.
1143 btrfs_release_path(path);
1145 /* Same search but for extended refs */
1146 extref = btrfs_lookup_inode_extref(NULL, root, path, name, namelen,
1147 inode_objectid, parent_objectid, 0,
1149 if (IS_ERR(extref)) {
1150 return PTR_ERR(extref);
1151 } else if (extref) {
1155 struct inode *victim_parent;
1157 leaf = path->nodes[0];
1159 item_size = btrfs_item_size(leaf, path->slots[0]);
1160 base = btrfs_item_ptr_offset(leaf, path->slots[0]);
1162 while (cur_offset < item_size) {
1163 extref = (struct btrfs_inode_extref *)(base + cur_offset);
1165 victim_name_len = btrfs_inode_extref_name_len(leaf, extref);
1167 if (btrfs_inode_extref_parent(leaf, extref) != parent_objectid)
1170 victim_name = kmalloc(victim_name_len, GFP_NOFS);
1173 read_extent_buffer(leaf, victim_name, (unsigned long)&extref->name,
1176 search_key.objectid = inode_objectid;
1177 search_key.type = BTRFS_INODE_EXTREF_KEY;
1178 search_key.offset = btrfs_extref_hash(parent_objectid,
1181 ret = backref_in_log(log_root, &search_key,
1182 parent_objectid, victim_name,
1189 victim_parent = read_one_inode(root,
1191 if (victim_parent) {
1192 inc_nlink(&inode->vfs_inode);
1193 btrfs_release_path(path);
1195 ret = unlink_inode_for_log_replay(trans,
1196 BTRFS_I(victim_parent),
1201 iput(victim_parent);
1210 cur_offset += victim_name_len + sizeof(*extref);
1214 btrfs_release_path(path);
1216 /* look for a conflicting sequence number */
1217 di = btrfs_lookup_dir_index_item(trans, root, path, btrfs_ino(dir),
1218 ref_index, name, namelen, 0);
1222 ret = drop_one_dir_item(trans, path, dir, di);
1226 btrfs_release_path(path);
1228 /* look for a conflicting name */
1229 di = btrfs_lookup_dir_item(trans, root, path, btrfs_ino(dir),
1234 ret = drop_one_dir_item(trans, path, dir, di);
1238 btrfs_release_path(path);
1243 static int extref_get_fields(struct extent_buffer *eb, unsigned long ref_ptr,
1244 u32 *namelen, char **name, u64 *index,
1245 u64 *parent_objectid)
1247 struct btrfs_inode_extref *extref;
1249 extref = (struct btrfs_inode_extref *)ref_ptr;
1251 *namelen = btrfs_inode_extref_name_len(eb, extref);
1252 *name = kmalloc(*namelen, GFP_NOFS);
1256 read_extent_buffer(eb, *name, (unsigned long)&extref->name,
1260 *index = btrfs_inode_extref_index(eb, extref);
1261 if (parent_objectid)
1262 *parent_objectid = btrfs_inode_extref_parent(eb, extref);
1267 static int ref_get_fields(struct extent_buffer *eb, unsigned long ref_ptr,
1268 u32 *namelen, char **name, u64 *index)
1270 struct btrfs_inode_ref *ref;
1272 ref = (struct btrfs_inode_ref *)ref_ptr;
1274 *namelen = btrfs_inode_ref_name_len(eb, ref);
1275 *name = kmalloc(*namelen, GFP_NOFS);
1279 read_extent_buffer(eb, *name, (unsigned long)(ref + 1), *namelen);
1282 *index = btrfs_inode_ref_index(eb, ref);
1288 * Take an inode reference item from the log tree and iterate all names from the
1289 * inode reference item in the subvolume tree with the same key (if it exists).
1290 * For any name that is not in the inode reference item from the log tree, do a
1291 * proper unlink of that name (that is, remove its entry from the inode
1292 * reference item and both dir index keys).
1294 static int unlink_old_inode_refs(struct btrfs_trans_handle *trans,
1295 struct btrfs_root *root,
1296 struct btrfs_path *path,
1297 struct btrfs_inode *inode,
1298 struct extent_buffer *log_eb,
1300 struct btrfs_key *key)
1303 unsigned long ref_ptr;
1304 unsigned long ref_end;
1305 struct extent_buffer *eb;
1308 btrfs_release_path(path);
1309 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
1317 eb = path->nodes[0];
1318 ref_ptr = btrfs_item_ptr_offset(eb, path->slots[0]);
1319 ref_end = ref_ptr + btrfs_item_size(eb, path->slots[0]);
1320 while (ref_ptr < ref_end) {
1325 if (key->type == BTRFS_INODE_EXTREF_KEY) {
1326 ret = extref_get_fields(eb, ref_ptr, &namelen, &name,
1329 parent_id = key->offset;
1330 ret = ref_get_fields(eb, ref_ptr, &namelen, &name,
1336 if (key->type == BTRFS_INODE_EXTREF_KEY)
1337 ret = !!btrfs_find_name_in_ext_backref(log_eb, log_slot,
1341 ret = !!btrfs_find_name_in_backref(log_eb, log_slot,
1347 btrfs_release_path(path);
1348 dir = read_one_inode(root, parent_id);
1354 ret = unlink_inode_for_log_replay(trans, BTRFS_I(dir),
1355 inode, name, namelen);
1365 if (key->type == BTRFS_INODE_EXTREF_KEY)
1366 ref_ptr += sizeof(struct btrfs_inode_extref);
1368 ref_ptr += sizeof(struct btrfs_inode_ref);
1372 btrfs_release_path(path);
1376 static int btrfs_inode_ref_exists(struct inode *inode, struct inode *dir,
1377 const u8 ref_type, const char *name,
1380 struct btrfs_key key;
1381 struct btrfs_path *path;
1382 const u64 parent_id = btrfs_ino(BTRFS_I(dir));
1385 path = btrfs_alloc_path();
1389 key.objectid = btrfs_ino(BTRFS_I(inode));
1390 key.type = ref_type;
1391 if (key.type == BTRFS_INODE_REF_KEY)
1392 key.offset = parent_id;
1394 key.offset = btrfs_extref_hash(parent_id, name, namelen);
1396 ret = btrfs_search_slot(NULL, BTRFS_I(inode)->root, &key, path, 0, 0);
1403 if (key.type == BTRFS_INODE_EXTREF_KEY)
1404 ret = !!btrfs_find_name_in_ext_backref(path->nodes[0],
1405 path->slots[0], parent_id, name, namelen);
1407 ret = !!btrfs_find_name_in_backref(path->nodes[0], path->slots[0],
1411 btrfs_free_path(path);
1415 static int add_link(struct btrfs_trans_handle *trans,
1416 struct inode *dir, struct inode *inode, const char *name,
1417 int namelen, u64 ref_index)
1419 struct btrfs_root *root = BTRFS_I(dir)->root;
1420 struct btrfs_dir_item *dir_item;
1421 struct btrfs_key key;
1422 struct btrfs_path *path;
1423 struct inode *other_inode = NULL;
1426 path = btrfs_alloc_path();
1430 dir_item = btrfs_lookup_dir_item(NULL, root, path,
1431 btrfs_ino(BTRFS_I(dir)),
1434 btrfs_release_path(path);
1436 } else if (IS_ERR(dir_item)) {
1437 ret = PTR_ERR(dir_item);
1442 * Our inode's dentry collides with the dentry of another inode which is
1443 * in the log but not yet processed since it has a higher inode number.
1444 * So delete that other dentry.
1446 btrfs_dir_item_key_to_cpu(path->nodes[0], dir_item, &key);
1447 btrfs_release_path(path);
1448 other_inode = read_one_inode(root, key.objectid);
1453 ret = unlink_inode_for_log_replay(trans, BTRFS_I(dir), BTRFS_I(other_inode),
1458 * If we dropped the link count to 0, bump it so that later the iput()
1459 * on the inode will not free it. We will fixup the link count later.
1461 if (other_inode->i_nlink == 0)
1462 set_nlink(other_inode, 1);
1464 ret = btrfs_add_link(trans, BTRFS_I(dir), BTRFS_I(inode),
1465 name, namelen, 0, ref_index);
1468 btrfs_free_path(path);
1474 * replay one inode back reference item found in the log tree.
1475 * eb, slot and key refer to the buffer and key found in the log tree.
1476 * root is the destination we are replaying into, and path is for temp
1477 * use by this function. (it should be released on return).
1479 static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
1480 struct btrfs_root *root,
1481 struct btrfs_root *log,
1482 struct btrfs_path *path,
1483 struct extent_buffer *eb, int slot,
1484 struct btrfs_key *key)
1486 struct inode *dir = NULL;
1487 struct inode *inode = NULL;
1488 unsigned long ref_ptr;
1489 unsigned long ref_end;
1493 int search_done = 0;
1494 int log_ref_ver = 0;
1495 u64 parent_objectid;
1498 int ref_struct_size;
1500 ref_ptr = btrfs_item_ptr_offset(eb, slot);
1501 ref_end = ref_ptr + btrfs_item_size(eb, slot);
1503 if (key->type == BTRFS_INODE_EXTREF_KEY) {
1504 struct btrfs_inode_extref *r;
1506 ref_struct_size = sizeof(struct btrfs_inode_extref);
1508 r = (struct btrfs_inode_extref *)ref_ptr;
1509 parent_objectid = btrfs_inode_extref_parent(eb, r);
1511 ref_struct_size = sizeof(struct btrfs_inode_ref);
1512 parent_objectid = key->offset;
1514 inode_objectid = key->objectid;
1517 * it is possible that we didn't log all the parent directories
1518 * for a given inode. If we don't find the dir, just don't
1519 * copy the back ref in. The link count fixup code will take
1522 dir = read_one_inode(root, parent_objectid);
1528 inode = read_one_inode(root, inode_objectid);
1534 while (ref_ptr < ref_end) {
1536 ret = extref_get_fields(eb, ref_ptr, &namelen, &name,
1537 &ref_index, &parent_objectid);
1539 * parent object can change from one array
1543 dir = read_one_inode(root, parent_objectid);
1549 ret = ref_get_fields(eb, ref_ptr, &namelen, &name,
1555 ret = inode_in_dir(root, path, btrfs_ino(BTRFS_I(dir)),
1556 btrfs_ino(BTRFS_I(inode)), ref_index,
1560 } else if (ret == 0) {
1562 * look for a conflicting back reference in the
1563 * metadata. if we find one we have to unlink that name
1564 * of the file before we add our new link. Later on, we
1565 * overwrite any existing back reference, and we don't
1566 * want to create dangling pointers in the directory.
1570 ret = __add_inode_ref(trans, root, path, log,
1575 ref_index, name, namelen,
1585 * If a reference item already exists for this inode
1586 * with the same parent and name, but different index,
1587 * drop it and the corresponding directory index entries
1588 * from the parent before adding the new reference item
1589 * and dir index entries, otherwise we would fail with
1590 * -EEXIST returned from btrfs_add_link() below.
1592 ret = btrfs_inode_ref_exists(inode, dir, key->type,
1595 ret = unlink_inode_for_log_replay(trans,
1600 * If we dropped the link count to 0, bump it so
1601 * that later the iput() on the inode will not
1602 * free it. We will fixup the link count later.
1604 if (!ret && inode->i_nlink == 0)
1605 set_nlink(inode, 1);
1610 /* insert our name */
1611 ret = add_link(trans, dir, inode, name, namelen,
1616 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
1620 /* Else, ret == 1, we already have a perfect match, we're done. */
1622 ref_ptr = (unsigned long)(ref_ptr + ref_struct_size) + namelen;
1632 * Before we overwrite the inode reference item in the subvolume tree
1633 * with the item from the log tree, we must unlink all names from the
1634 * parent directory that are in the subvolume's tree inode reference
1635 * item, otherwise we end up with an inconsistent subvolume tree where
1636 * dir index entries exist for a name but there is no inode reference
1637 * item with the same name.
1639 ret = unlink_old_inode_refs(trans, root, path, BTRFS_I(inode), eb, slot,
1644 /* finally write the back reference in the inode */
1645 ret = overwrite_item(trans, root, path, eb, slot, key);
1647 btrfs_release_path(path);
1654 static int count_inode_extrefs(struct btrfs_root *root,
1655 struct btrfs_inode *inode, struct btrfs_path *path)
1659 unsigned int nlink = 0;
1662 u64 inode_objectid = btrfs_ino(inode);
1665 struct btrfs_inode_extref *extref;
1666 struct extent_buffer *leaf;
1669 ret = btrfs_find_one_extref(root, inode_objectid, offset, path,
1674 leaf = path->nodes[0];
1675 item_size = btrfs_item_size(leaf, path->slots[0]);
1676 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1679 while (cur_offset < item_size) {
1680 extref = (struct btrfs_inode_extref *) (ptr + cur_offset);
1681 name_len = btrfs_inode_extref_name_len(leaf, extref);
1685 cur_offset += name_len + sizeof(*extref);
1689 btrfs_release_path(path);
1691 btrfs_release_path(path);
1693 if (ret < 0 && ret != -ENOENT)
1698 static int count_inode_refs(struct btrfs_root *root,
1699 struct btrfs_inode *inode, struct btrfs_path *path)
1702 struct btrfs_key key;
1703 unsigned int nlink = 0;
1705 unsigned long ptr_end;
1707 u64 ino = btrfs_ino(inode);
1710 key.type = BTRFS_INODE_REF_KEY;
1711 key.offset = (u64)-1;
1714 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1718 if (path->slots[0] == 0)
1723 btrfs_item_key_to_cpu(path->nodes[0], &key,
1725 if (key.objectid != ino ||
1726 key.type != BTRFS_INODE_REF_KEY)
1728 ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
1729 ptr_end = ptr + btrfs_item_size(path->nodes[0],
1731 while (ptr < ptr_end) {
1732 struct btrfs_inode_ref *ref;
1734 ref = (struct btrfs_inode_ref *)ptr;
1735 name_len = btrfs_inode_ref_name_len(path->nodes[0],
1737 ptr = (unsigned long)(ref + 1) + name_len;
1741 if (key.offset == 0)
1743 if (path->slots[0] > 0) {
1748 btrfs_release_path(path);
1750 btrfs_release_path(path);
1756 * There are a few corners where the link count of the file can't
1757 * be properly maintained during replay. So, instead of adding
1758 * lots of complexity to the log code, we just scan the backrefs
1759 * for any file that has been through replay.
1761 * The scan will update the link count on the inode to reflect the
1762 * number of back refs found. If it goes down to zero, the iput
1763 * will free the inode.
1765 static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
1766 struct btrfs_root *root,
1767 struct inode *inode)
1769 struct btrfs_path *path;
1772 u64 ino = btrfs_ino(BTRFS_I(inode));
1774 path = btrfs_alloc_path();
1778 ret = count_inode_refs(root, BTRFS_I(inode), path);
1784 ret = count_inode_extrefs(root, BTRFS_I(inode), path);
1792 if (nlink != inode->i_nlink) {
1793 set_nlink(inode, nlink);
1794 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
1798 BTRFS_I(inode)->index_cnt = (u64)-1;
1800 if (inode->i_nlink == 0) {
1801 if (S_ISDIR(inode->i_mode)) {
1802 ret = replay_dir_deletes(trans, root, NULL, path,
1807 ret = btrfs_insert_orphan_item(trans, root, ino);
1813 btrfs_free_path(path);
1817 static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
1818 struct btrfs_root *root,
1819 struct btrfs_path *path)
1822 struct btrfs_key key;
1823 struct inode *inode;
1825 key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
1826 key.type = BTRFS_ORPHAN_ITEM_KEY;
1827 key.offset = (u64)-1;
1829 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1835 if (path->slots[0] == 0)
1840 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1841 if (key.objectid != BTRFS_TREE_LOG_FIXUP_OBJECTID ||
1842 key.type != BTRFS_ORPHAN_ITEM_KEY)
1845 ret = btrfs_del_item(trans, root, path);
1849 btrfs_release_path(path);
1850 inode = read_one_inode(root, key.offset);
1856 ret = fixup_inode_link_count(trans, root, inode);
1862 * fixup on a directory may create new entries,
1863 * make sure we always look for the highset possible
1866 key.offset = (u64)-1;
1868 btrfs_release_path(path);
1874 * record a given inode in the fixup dir so we can check its link
1875 * count when replay is done. The link count is incremented here
1876 * so the inode won't go away until we check it
1878 static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans,
1879 struct btrfs_root *root,
1880 struct btrfs_path *path,
1883 struct btrfs_key key;
1885 struct inode *inode;
1887 inode = read_one_inode(root, objectid);
1891 key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
1892 key.type = BTRFS_ORPHAN_ITEM_KEY;
1893 key.offset = objectid;
1895 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1897 btrfs_release_path(path);
1899 if (!inode->i_nlink)
1900 set_nlink(inode, 1);
1903 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
1904 } else if (ret == -EEXIST) {
1913 * when replaying the log for a directory, we only insert names
1914 * for inodes that actually exist. This means an fsync on a directory
1915 * does not implicitly fsync all the new files in it
1917 static noinline int insert_one_name(struct btrfs_trans_handle *trans,
1918 struct btrfs_root *root,
1919 u64 dirid, u64 index,
1920 char *name, int name_len,
1921 struct btrfs_key *location)
1923 struct inode *inode;
1927 inode = read_one_inode(root, location->objectid);
1931 dir = read_one_inode(root, dirid);
1937 ret = btrfs_add_link(trans, BTRFS_I(dir), BTRFS_I(inode), name,
1938 name_len, 1, index);
1940 /* FIXME, put inode into FIXUP list */
1947 static int delete_conflicting_dir_entry(struct btrfs_trans_handle *trans,
1948 struct btrfs_inode *dir,
1949 struct btrfs_path *path,
1950 struct btrfs_dir_item *dst_di,
1951 const struct btrfs_key *log_key,
1955 struct btrfs_key found_key;
1957 btrfs_dir_item_key_to_cpu(path->nodes[0], dst_di, &found_key);
1958 /* The existing dentry points to the same inode, don't delete it. */
1959 if (found_key.objectid == log_key->objectid &&
1960 found_key.type == log_key->type &&
1961 found_key.offset == log_key->offset &&
1962 btrfs_dir_type(path->nodes[0], dst_di) == log_type)
1966 * Don't drop the conflicting directory entry if the inode for the new
1967 * entry doesn't exist.
1972 return drop_one_dir_item(trans, path, dir, dst_di);
1976 * take a single entry in a log directory item and replay it into
1979 * if a conflicting item exists in the subdirectory already,
1980 * the inode it points to is unlinked and put into the link count
1983 * If a name from the log points to a file or directory that does
1984 * not exist in the FS, it is skipped. fsyncs on directories
1985 * do not force down inodes inside that directory, just changes to the
1986 * names or unlinks in a directory.
1988 * Returns < 0 on error, 0 if the name wasn't replayed (dentry points to a
1989 * non-existing inode) and 1 if the name was replayed.
1991 static noinline int replay_one_name(struct btrfs_trans_handle *trans,
1992 struct btrfs_root *root,
1993 struct btrfs_path *path,
1994 struct extent_buffer *eb,
1995 struct btrfs_dir_item *di,
1996 struct btrfs_key *key)
2000 struct btrfs_dir_item *dir_dst_di;
2001 struct btrfs_dir_item *index_dst_di;
2002 bool dir_dst_matches = false;
2003 bool index_dst_matches = false;
2004 struct btrfs_key log_key;
2005 struct btrfs_key search_key;
2010 bool update_size = true;
2011 bool name_added = false;
2013 dir = read_one_inode(root, key->objectid);
2017 name_len = btrfs_dir_name_len(eb, di);
2018 name = kmalloc(name_len, GFP_NOFS);
2024 log_type = btrfs_dir_type(eb, di);
2025 read_extent_buffer(eb, name, (unsigned long)(di + 1),
2028 btrfs_dir_item_key_to_cpu(eb, di, &log_key);
2029 ret = btrfs_lookup_inode(trans, root, path, &log_key, 0);
2030 btrfs_release_path(path);
2033 exists = (ret == 0);
2036 dir_dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid,
2038 if (IS_ERR(dir_dst_di)) {
2039 ret = PTR_ERR(dir_dst_di);
2041 } else if (dir_dst_di) {
2042 ret = delete_conflicting_dir_entry(trans, BTRFS_I(dir), path,
2043 dir_dst_di, &log_key, log_type,
2047 dir_dst_matches = (ret == 1);
2050 btrfs_release_path(path);
2052 index_dst_di = btrfs_lookup_dir_index_item(trans, root, path,
2053 key->objectid, key->offset,
2055 if (IS_ERR(index_dst_di)) {
2056 ret = PTR_ERR(index_dst_di);
2058 } else if (index_dst_di) {
2059 ret = delete_conflicting_dir_entry(trans, BTRFS_I(dir), path,
2060 index_dst_di, &log_key,
2064 index_dst_matches = (ret == 1);
2067 btrfs_release_path(path);
2069 if (dir_dst_matches && index_dst_matches) {
2071 update_size = false;
2076 * Check if the inode reference exists in the log for the given name,
2077 * inode and parent inode
2079 search_key.objectid = log_key.objectid;
2080 search_key.type = BTRFS_INODE_REF_KEY;
2081 search_key.offset = key->objectid;
2082 ret = backref_in_log(root->log_root, &search_key, 0, name, name_len);
2086 /* The dentry will be added later. */
2088 update_size = false;
2092 search_key.objectid = log_key.objectid;
2093 search_key.type = BTRFS_INODE_EXTREF_KEY;
2094 search_key.offset = key->objectid;
2095 ret = backref_in_log(root->log_root, &search_key, key->objectid, name,
2100 /* The dentry will be added later. */
2102 update_size = false;
2105 btrfs_release_path(path);
2106 ret = insert_one_name(trans, root, key->objectid, key->offset,
2107 name, name_len, &log_key);
2108 if (ret && ret != -ENOENT && ret != -EEXIST)
2112 update_size = false;
2116 if (!ret && update_size) {
2117 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + name_len * 2);
2118 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
2122 if (!ret && name_added)
2127 /* Replay one dir item from a BTRFS_DIR_INDEX_KEY key. */
2128 static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans,
2129 struct btrfs_root *root,
2130 struct btrfs_path *path,
2131 struct extent_buffer *eb, int slot,
2132 struct btrfs_key *key)
2135 struct btrfs_dir_item *di;
2137 /* We only log dir index keys, which only contain a single dir item. */
2138 ASSERT(key->type == BTRFS_DIR_INDEX_KEY);
2140 di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
2141 ret = replay_one_name(trans, root, path, eb, di, key);
2146 * If this entry refers to a non-directory (directories can not have a
2147 * link count > 1) and it was added in the transaction that was not
2148 * committed, make sure we fixup the link count of the inode the entry
2149 * points to. Otherwise something like the following would result in a
2150 * directory pointing to an inode with a wrong link that does not account
2151 * for this dir entry:
2158 * ln testdir/bar testdir/bar_link
2159 * ln testdir/foo testdir/foo_link
2160 * xfs_io -c "fsync" testdir/bar
2164 * mount fs, log replay happens
2166 * File foo would remain with a link count of 1 when it has two entries
2167 * pointing to it in the directory testdir. This would make it impossible
2168 * to ever delete the parent directory has it would result in stale
2169 * dentries that can never be deleted.
2171 if (ret == 1 && btrfs_dir_type(eb, di) != BTRFS_FT_DIR) {
2172 struct btrfs_path *fixup_path;
2173 struct btrfs_key di_key;
2175 fixup_path = btrfs_alloc_path();
2179 btrfs_dir_item_key_to_cpu(eb, di, &di_key);
2180 ret = link_to_fixup_dir(trans, root, fixup_path, di_key.objectid);
2181 btrfs_free_path(fixup_path);
2188 * directory replay has two parts. There are the standard directory
2189 * items in the log copied from the subvolume, and range items
2190 * created in the log while the subvolume was logged.
2192 * The range items tell us which parts of the key space the log
2193 * is authoritative for. During replay, if a key in the subvolume
2194 * directory is in a logged range item, but not actually in the log
2195 * that means it was deleted from the directory before the fsync
2196 * and should be removed.
2198 static noinline int find_dir_range(struct btrfs_root *root,
2199 struct btrfs_path *path,
2201 u64 *start_ret, u64 *end_ret)
2203 struct btrfs_key key;
2205 struct btrfs_dir_log_item *item;
2209 if (*start_ret == (u64)-1)
2212 key.objectid = dirid;
2213 key.type = BTRFS_DIR_LOG_INDEX_KEY;
2214 key.offset = *start_ret;
2216 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2220 if (path->slots[0] == 0)
2225 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
2227 if (key.type != BTRFS_DIR_LOG_INDEX_KEY || key.objectid != dirid) {
2231 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2232 struct btrfs_dir_log_item);
2233 found_end = btrfs_dir_log_end(path->nodes[0], item);
2235 if (*start_ret >= key.offset && *start_ret <= found_end) {
2237 *start_ret = key.offset;
2238 *end_ret = found_end;
2243 /* check the next slot in the tree to see if it is a valid item */
2244 nritems = btrfs_header_nritems(path->nodes[0]);
2246 if (path->slots[0] >= nritems) {
2247 ret = btrfs_next_leaf(root, path);
2252 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
2254 if (key.type != BTRFS_DIR_LOG_INDEX_KEY || key.objectid != dirid) {
2258 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2259 struct btrfs_dir_log_item);
2260 found_end = btrfs_dir_log_end(path->nodes[0], item);
2261 *start_ret = key.offset;
2262 *end_ret = found_end;
2265 btrfs_release_path(path);
2270 * this looks for a given directory item in the log. If the directory
2271 * item is not in the log, the item is removed and the inode it points
2274 static noinline int check_item_in_log(struct btrfs_trans_handle *trans,
2275 struct btrfs_root *log,
2276 struct btrfs_path *path,
2277 struct btrfs_path *log_path,
2279 struct btrfs_key *dir_key)
2281 struct btrfs_root *root = BTRFS_I(dir)->root;
2283 struct extent_buffer *eb;
2285 struct btrfs_dir_item *di;
2288 struct inode *inode = NULL;
2289 struct btrfs_key location;
2292 * Currently we only log dir index keys. Even if we replay a log created
2293 * by an older kernel that logged both dir index and dir item keys, all
2294 * we need to do is process the dir index keys, we (and our caller) can
2295 * safely ignore dir item keys (key type BTRFS_DIR_ITEM_KEY).
2297 ASSERT(dir_key->type == BTRFS_DIR_INDEX_KEY);
2299 eb = path->nodes[0];
2300 slot = path->slots[0];
2301 di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
2302 name_len = btrfs_dir_name_len(eb, di);
2303 name = kmalloc(name_len, GFP_NOFS);
2309 read_extent_buffer(eb, name, (unsigned long)(di + 1), name_len);
2312 struct btrfs_dir_item *log_di;
2314 log_di = btrfs_lookup_dir_index_item(trans, log, log_path,
2318 if (IS_ERR(log_di)) {
2319 ret = PTR_ERR(log_di);
2321 } else if (log_di) {
2322 /* The dentry exists in the log, we have nothing to do. */
2328 btrfs_dir_item_key_to_cpu(eb, di, &location);
2329 btrfs_release_path(path);
2330 btrfs_release_path(log_path);
2331 inode = read_one_inode(root, location.objectid);
2337 ret = link_to_fixup_dir(trans, root, path, location.objectid);
2342 ret = unlink_inode_for_log_replay(trans, BTRFS_I(dir), BTRFS_I(inode),
2345 * Unlike dir item keys, dir index keys can only have one name (entry) in
2346 * them, as there are no key collisions since each key has a unique offset
2347 * (an index number), so we're done.
2350 btrfs_release_path(path);
2351 btrfs_release_path(log_path);
2357 static int replay_xattr_deletes(struct btrfs_trans_handle *trans,
2358 struct btrfs_root *root,
2359 struct btrfs_root *log,
2360 struct btrfs_path *path,
2363 struct btrfs_key search_key;
2364 struct btrfs_path *log_path;
2369 log_path = btrfs_alloc_path();
2373 search_key.objectid = ino;
2374 search_key.type = BTRFS_XATTR_ITEM_KEY;
2375 search_key.offset = 0;
2377 ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
2381 nritems = btrfs_header_nritems(path->nodes[0]);
2382 for (i = path->slots[0]; i < nritems; i++) {
2383 struct btrfs_key key;
2384 struct btrfs_dir_item *di;
2385 struct btrfs_dir_item *log_di;
2389 btrfs_item_key_to_cpu(path->nodes[0], &key, i);
2390 if (key.objectid != ino || key.type != BTRFS_XATTR_ITEM_KEY) {
2395 di = btrfs_item_ptr(path->nodes[0], i, struct btrfs_dir_item);
2396 total_size = btrfs_item_size(path->nodes[0], i);
2398 while (cur < total_size) {
2399 u16 name_len = btrfs_dir_name_len(path->nodes[0], di);
2400 u16 data_len = btrfs_dir_data_len(path->nodes[0], di);
2401 u32 this_len = sizeof(*di) + name_len + data_len;
2404 name = kmalloc(name_len, GFP_NOFS);
2409 read_extent_buffer(path->nodes[0], name,
2410 (unsigned long)(di + 1), name_len);
2412 log_di = btrfs_lookup_xattr(NULL, log, log_path, ino,
2414 btrfs_release_path(log_path);
2416 /* Doesn't exist in log tree, so delete it. */
2417 btrfs_release_path(path);
2418 di = btrfs_lookup_xattr(trans, root, path, ino,
2419 name, name_len, -1);
2426 ret = btrfs_delete_one_dir_name(trans, root,
2430 btrfs_release_path(path);
2435 if (IS_ERR(log_di)) {
2436 ret = PTR_ERR(log_di);
2440 di = (struct btrfs_dir_item *)((char *)di + this_len);
2443 ret = btrfs_next_leaf(root, path);
2449 btrfs_free_path(log_path);
2450 btrfs_release_path(path);
2456 * deletion replay happens before we copy any new directory items
2457 * out of the log or out of backreferences from inodes. It
2458 * scans the log to find ranges of keys that log is authoritative for,
2459 * and then scans the directory to find items in those ranges that are
2460 * not present in the log.
2462 * Anything we don't find in the log is unlinked and removed from the
2465 static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
2466 struct btrfs_root *root,
2467 struct btrfs_root *log,
2468 struct btrfs_path *path,
2469 u64 dirid, int del_all)
2474 struct btrfs_key dir_key;
2475 struct btrfs_key found_key;
2476 struct btrfs_path *log_path;
2479 dir_key.objectid = dirid;
2480 dir_key.type = BTRFS_DIR_INDEX_KEY;
2481 log_path = btrfs_alloc_path();
2485 dir = read_one_inode(root, dirid);
2486 /* it isn't an error if the inode isn't there, that can happen
2487 * because we replay the deletes before we copy in the inode item
2491 btrfs_free_path(log_path);
2499 range_end = (u64)-1;
2501 ret = find_dir_range(log, path, dirid,
2502 &range_start, &range_end);
2509 dir_key.offset = range_start;
2512 ret = btrfs_search_slot(NULL, root, &dir_key, path,
2517 nritems = btrfs_header_nritems(path->nodes[0]);
2518 if (path->slots[0] >= nritems) {
2519 ret = btrfs_next_leaf(root, path);
2525 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2527 if (found_key.objectid != dirid ||
2528 found_key.type != dir_key.type) {
2533 if (found_key.offset > range_end)
2536 ret = check_item_in_log(trans, log, path,
2541 if (found_key.offset == (u64)-1)
2543 dir_key.offset = found_key.offset + 1;
2545 btrfs_release_path(path);
2546 if (range_end == (u64)-1)
2548 range_start = range_end + 1;
2552 btrfs_release_path(path);
2553 btrfs_free_path(log_path);
2559 * the process_func used to replay items from the log tree. This
2560 * gets called in two different stages. The first stage just looks
2561 * for inodes and makes sure they are all copied into the subvolume.
2563 * The second stage copies all the other item types from the log into
2564 * the subvolume. The two stage approach is slower, but gets rid of
2565 * lots of complexity around inodes referencing other inodes that exist
2566 * only in the log (references come from either directory items or inode
2569 static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
2570 struct walk_control *wc, u64 gen, int level)
2573 struct btrfs_path *path;
2574 struct btrfs_root *root = wc->replay_dest;
2575 struct btrfs_key key;
2579 ret = btrfs_read_extent_buffer(eb, gen, level, NULL);
2583 level = btrfs_header_level(eb);
2588 path = btrfs_alloc_path();
2592 nritems = btrfs_header_nritems(eb);
2593 for (i = 0; i < nritems; i++) {
2594 btrfs_item_key_to_cpu(eb, &key, i);
2596 /* inode keys are done during the first stage */
2597 if (key.type == BTRFS_INODE_ITEM_KEY &&
2598 wc->stage == LOG_WALK_REPLAY_INODES) {
2599 struct btrfs_inode_item *inode_item;
2602 inode_item = btrfs_item_ptr(eb, i,
2603 struct btrfs_inode_item);
2605 * If we have a tmpfile (O_TMPFILE) that got fsync'ed
2606 * and never got linked before the fsync, skip it, as
2607 * replaying it is pointless since it would be deleted
2608 * later. We skip logging tmpfiles, but it's always
2609 * possible we are replaying a log created with a kernel
2610 * that used to log tmpfiles.
2612 if (btrfs_inode_nlink(eb, inode_item) == 0) {
2613 wc->ignore_cur_inode = true;
2616 wc->ignore_cur_inode = false;
2618 ret = replay_xattr_deletes(wc->trans, root, log,
2619 path, key.objectid);
2622 mode = btrfs_inode_mode(eb, inode_item);
2623 if (S_ISDIR(mode)) {
2624 ret = replay_dir_deletes(wc->trans,
2625 root, log, path, key.objectid, 0);
2629 ret = overwrite_item(wc->trans, root, path,
2635 * Before replaying extents, truncate the inode to its
2636 * size. We need to do it now and not after log replay
2637 * because before an fsync we can have prealloc extents
2638 * added beyond the inode's i_size. If we did it after,
2639 * through orphan cleanup for example, we would drop
2640 * those prealloc extents just after replaying them.
2642 if (S_ISREG(mode)) {
2643 struct btrfs_drop_extents_args drop_args = { 0 };
2644 struct inode *inode;
2647 inode = read_one_inode(root, key.objectid);
2652 from = ALIGN(i_size_read(inode),
2653 root->fs_info->sectorsize);
2654 drop_args.start = from;
2655 drop_args.end = (u64)-1;
2656 drop_args.drop_cache = true;
2657 ret = btrfs_drop_extents(wc->trans, root,
2661 inode_sub_bytes(inode,
2662 drop_args.bytes_found);
2663 /* Update the inode's nbytes. */
2664 ret = btrfs_update_inode(wc->trans,
2665 root, BTRFS_I(inode));
2672 ret = link_to_fixup_dir(wc->trans, root,
2673 path, key.objectid);
2678 if (wc->ignore_cur_inode)
2681 if (key.type == BTRFS_DIR_INDEX_KEY &&
2682 wc->stage == LOG_WALK_REPLAY_DIR_INDEX) {
2683 ret = replay_one_dir_item(wc->trans, root, path,
2689 if (wc->stage < LOG_WALK_REPLAY_ALL)
2692 /* these keys are simply copied */
2693 if (key.type == BTRFS_XATTR_ITEM_KEY) {
2694 ret = overwrite_item(wc->trans, root, path,
2698 } else if (key.type == BTRFS_INODE_REF_KEY ||
2699 key.type == BTRFS_INODE_EXTREF_KEY) {
2700 ret = add_inode_ref(wc->trans, root, log, path,
2702 if (ret && ret != -ENOENT)
2705 } else if (key.type == BTRFS_EXTENT_DATA_KEY) {
2706 ret = replay_one_extent(wc->trans, root, path,
2712 * We don't log BTRFS_DIR_ITEM_KEY keys anymore, only the
2713 * BTRFS_DIR_INDEX_KEY items which we use to derive the
2714 * BTRFS_DIR_ITEM_KEY items. If we are replaying a log from an
2715 * older kernel with such keys, ignore them.
2718 btrfs_free_path(path);
2723 * Correctly adjust the reserved bytes occupied by a log tree extent buffer
2725 static void unaccount_log_buffer(struct btrfs_fs_info *fs_info, u64 start)
2727 struct btrfs_block_group *cache;
2729 cache = btrfs_lookup_block_group(fs_info, start);
2731 btrfs_err(fs_info, "unable to find block group for %llu", start);
2735 spin_lock(&cache->space_info->lock);
2736 spin_lock(&cache->lock);
2737 cache->reserved -= fs_info->nodesize;
2738 cache->space_info->bytes_reserved -= fs_info->nodesize;
2739 spin_unlock(&cache->lock);
2740 spin_unlock(&cache->space_info->lock);
2742 btrfs_put_block_group(cache);
2745 static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
2746 struct btrfs_root *root,
2747 struct btrfs_path *path, int *level,
2748 struct walk_control *wc)
2750 struct btrfs_fs_info *fs_info = root->fs_info;
2753 struct extent_buffer *next;
2754 struct extent_buffer *cur;
2758 while (*level > 0) {
2759 struct btrfs_key first_key;
2761 cur = path->nodes[*level];
2763 WARN_ON(btrfs_header_level(cur) != *level);
2765 if (path->slots[*level] >=
2766 btrfs_header_nritems(cur))
2769 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
2770 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
2771 btrfs_node_key_to_cpu(cur, &first_key, path->slots[*level]);
2772 blocksize = fs_info->nodesize;
2774 next = btrfs_find_create_tree_block(fs_info, bytenr,
2775 btrfs_header_owner(cur),
2778 return PTR_ERR(next);
2781 ret = wc->process_func(root, next, wc, ptr_gen,
2784 free_extent_buffer(next);
2788 path->slots[*level]++;
2790 ret = btrfs_read_extent_buffer(next, ptr_gen,
2791 *level - 1, &first_key);
2793 free_extent_buffer(next);
2798 btrfs_tree_lock(next);
2799 btrfs_clean_tree_block(next);
2800 btrfs_wait_tree_block_writeback(next);
2801 btrfs_tree_unlock(next);
2802 ret = btrfs_pin_reserved_extent(trans,
2805 free_extent_buffer(next);
2808 btrfs_redirty_list_add(
2809 trans->transaction, next);
2811 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags))
2812 clear_extent_buffer_dirty(next);
2813 unaccount_log_buffer(fs_info, bytenr);
2816 free_extent_buffer(next);
2819 ret = btrfs_read_extent_buffer(next, ptr_gen, *level - 1, &first_key);
2821 free_extent_buffer(next);
2825 if (path->nodes[*level-1])
2826 free_extent_buffer(path->nodes[*level-1]);
2827 path->nodes[*level-1] = next;
2828 *level = btrfs_header_level(next);
2829 path->slots[*level] = 0;
2832 path->slots[*level] = btrfs_header_nritems(path->nodes[*level]);
2838 static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
2839 struct btrfs_root *root,
2840 struct btrfs_path *path, int *level,
2841 struct walk_control *wc)
2843 struct btrfs_fs_info *fs_info = root->fs_info;
2848 for (i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
2849 slot = path->slots[i];
2850 if (slot + 1 < btrfs_header_nritems(path->nodes[i])) {
2853 WARN_ON(*level == 0);
2856 ret = wc->process_func(root, path->nodes[*level], wc,
2857 btrfs_header_generation(path->nodes[*level]),
2863 struct extent_buffer *next;
2865 next = path->nodes[*level];
2868 btrfs_tree_lock(next);
2869 btrfs_clean_tree_block(next);
2870 btrfs_wait_tree_block_writeback(next);
2871 btrfs_tree_unlock(next);
2872 ret = btrfs_pin_reserved_extent(trans,
2873 path->nodes[*level]->start,
2874 path->nodes[*level]->len);
2877 btrfs_redirty_list_add(trans->transaction,
2880 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags))
2881 clear_extent_buffer_dirty(next);
2883 unaccount_log_buffer(fs_info,
2884 path->nodes[*level]->start);
2887 free_extent_buffer(path->nodes[*level]);
2888 path->nodes[*level] = NULL;
2896 * drop the reference count on the tree rooted at 'snap'. This traverses
2897 * the tree freeing any blocks that have a ref count of zero after being
2900 static int walk_log_tree(struct btrfs_trans_handle *trans,
2901 struct btrfs_root *log, struct walk_control *wc)
2903 struct btrfs_fs_info *fs_info = log->fs_info;
2907 struct btrfs_path *path;
2910 path = btrfs_alloc_path();
2914 level = btrfs_header_level(log->node);
2916 path->nodes[level] = log->node;
2917 atomic_inc(&log->node->refs);
2918 path->slots[level] = 0;
2921 wret = walk_down_log_tree(trans, log, path, &level, wc);
2929 wret = walk_up_log_tree(trans, log, path, &level, wc);
2938 /* was the root node processed? if not, catch it here */
2939 if (path->nodes[orig_level]) {
2940 ret = wc->process_func(log, path->nodes[orig_level], wc,
2941 btrfs_header_generation(path->nodes[orig_level]),
2946 struct extent_buffer *next;
2948 next = path->nodes[orig_level];
2951 btrfs_tree_lock(next);
2952 btrfs_clean_tree_block(next);
2953 btrfs_wait_tree_block_writeback(next);
2954 btrfs_tree_unlock(next);
2955 ret = btrfs_pin_reserved_extent(trans,
2956 next->start, next->len);
2959 btrfs_redirty_list_add(trans->transaction, next);
2961 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &next->bflags))
2962 clear_extent_buffer_dirty(next);
2963 unaccount_log_buffer(fs_info, next->start);
2969 btrfs_free_path(path);
2974 * helper function to update the item for a given subvolumes log root
2975 * in the tree of log roots
2977 static int update_log_root(struct btrfs_trans_handle *trans,
2978 struct btrfs_root *log,
2979 struct btrfs_root_item *root_item)
2981 struct btrfs_fs_info *fs_info = log->fs_info;
2984 if (log->log_transid == 1) {
2985 /* insert root item on the first sync */
2986 ret = btrfs_insert_root(trans, fs_info->log_root_tree,
2987 &log->root_key, root_item);
2989 ret = btrfs_update_root(trans, fs_info->log_root_tree,
2990 &log->root_key, root_item);
2995 static void wait_log_commit(struct btrfs_root *root, int transid)
2998 int index = transid % 2;
3001 * we only allow two pending log transactions at a time,
3002 * so we know that if ours is more than 2 older than the
3003 * current transaction, we're done
3006 prepare_to_wait(&root->log_commit_wait[index],
3007 &wait, TASK_UNINTERRUPTIBLE);
3009 if (!(root->log_transid_committed < transid &&
3010 atomic_read(&root->log_commit[index])))
3013 mutex_unlock(&root->log_mutex);
3015 mutex_lock(&root->log_mutex);
3017 finish_wait(&root->log_commit_wait[index], &wait);
3020 static void wait_for_writer(struct btrfs_root *root)
3025 prepare_to_wait(&root->log_writer_wait, &wait,
3026 TASK_UNINTERRUPTIBLE);
3027 if (!atomic_read(&root->log_writers))
3030 mutex_unlock(&root->log_mutex);
3032 mutex_lock(&root->log_mutex);
3034 finish_wait(&root->log_writer_wait, &wait);
3037 static inline void btrfs_remove_log_ctx(struct btrfs_root *root,
3038 struct btrfs_log_ctx *ctx)
3040 mutex_lock(&root->log_mutex);
3041 list_del_init(&ctx->list);
3042 mutex_unlock(&root->log_mutex);
3046 * Invoked in log mutex context, or be sure there is no other task which
3047 * can access the list.
3049 static inline void btrfs_remove_all_log_ctxs(struct btrfs_root *root,
3050 int index, int error)
3052 struct btrfs_log_ctx *ctx;
3053 struct btrfs_log_ctx *safe;
3055 list_for_each_entry_safe(ctx, safe, &root->log_ctxs[index], list) {
3056 list_del_init(&ctx->list);
3057 ctx->log_ret = error;
3062 * btrfs_sync_log does sends a given tree log down to the disk and
3063 * updates the super blocks to record it. When this call is done,
3064 * you know that any inodes previously logged are safely on disk only
3067 * Any other return value means you need to call btrfs_commit_transaction.
3068 * Some of the edge cases for fsyncing directories that have had unlinks
3069 * or renames done in the past mean that sometimes the only safe
3070 * fsync is to commit the whole FS. When btrfs_sync_log returns -EAGAIN,
3071 * that has happened.
3073 int btrfs_sync_log(struct btrfs_trans_handle *trans,
3074 struct btrfs_root *root, struct btrfs_log_ctx *ctx)
3080 struct btrfs_fs_info *fs_info = root->fs_info;
3081 struct btrfs_root *log = root->log_root;
3082 struct btrfs_root *log_root_tree = fs_info->log_root_tree;
3083 struct btrfs_root_item new_root_item;
3084 int log_transid = 0;
3085 struct btrfs_log_ctx root_log_ctx;
3086 struct blk_plug plug;
3090 mutex_lock(&root->log_mutex);
3091 log_transid = ctx->log_transid;
3092 if (root->log_transid_committed >= log_transid) {
3093 mutex_unlock(&root->log_mutex);
3094 return ctx->log_ret;
3097 index1 = log_transid % 2;
3098 if (atomic_read(&root->log_commit[index1])) {
3099 wait_log_commit(root, log_transid);
3100 mutex_unlock(&root->log_mutex);
3101 return ctx->log_ret;
3103 ASSERT(log_transid == root->log_transid);
3104 atomic_set(&root->log_commit[index1], 1);
3106 /* wait for previous tree log sync to complete */
3107 if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
3108 wait_log_commit(root, log_transid - 1);
3111 int batch = atomic_read(&root->log_batch);
3112 /* when we're on an ssd, just kick the log commit out */
3113 if (!btrfs_test_opt(fs_info, SSD) &&
3114 test_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state)) {
3115 mutex_unlock(&root->log_mutex);
3116 schedule_timeout_uninterruptible(1);
3117 mutex_lock(&root->log_mutex);
3119 wait_for_writer(root);
3120 if (batch == atomic_read(&root->log_batch))
3124 /* bail out if we need to do a full commit */
3125 if (btrfs_need_log_full_commit(trans)) {
3126 ret = BTRFS_LOG_FORCE_COMMIT;
3127 mutex_unlock(&root->log_mutex);
3131 if (log_transid % 2 == 0)
3132 mark = EXTENT_DIRTY;
3136 /* we start IO on all the marked extents here, but we don't actually
3137 * wait for them until later.
3139 blk_start_plug(&plug);
3140 ret = btrfs_write_marked_extents(fs_info, &log->dirty_log_pages, mark);
3142 * -EAGAIN happens when someone, e.g., a concurrent transaction
3143 * commit, writes a dirty extent in this tree-log commit. This
3144 * concurrent write will create a hole writing out the extents,
3145 * and we cannot proceed on a zoned filesystem, requiring
3146 * sequential writing. While we can bail out to a full commit
3147 * here, but we can continue hoping the concurrent writing fills
3150 if (ret == -EAGAIN && btrfs_is_zoned(fs_info))
3153 blk_finish_plug(&plug);
3154 btrfs_abort_transaction(trans, ret);
3155 btrfs_set_log_full_commit(trans);
3156 mutex_unlock(&root->log_mutex);
3161 * We _must_ update under the root->log_mutex in order to make sure we
3162 * have a consistent view of the log root we are trying to commit at
3165 * We _must_ copy this into a local copy, because we are not holding the
3166 * log_root_tree->log_mutex yet. This is important because when we
3167 * commit the log_root_tree we must have a consistent view of the
3168 * log_root_tree when we update the super block to point at the
3169 * log_root_tree bytenr. If we update the log_root_tree here we'll race
3170 * with the commit and possibly point at the new block which we may not
3173 btrfs_set_root_node(&log->root_item, log->node);
3174 memcpy(&new_root_item, &log->root_item, sizeof(new_root_item));
3176 root->log_transid++;
3177 log->log_transid = root->log_transid;
3178 root->log_start_pid = 0;
3180 * IO has been started, blocks of the log tree have WRITTEN flag set
3181 * in their headers. new modifications of the log will be written to
3182 * new positions. so it's safe to allow log writers to go in.
3184 mutex_unlock(&root->log_mutex);
3186 if (btrfs_is_zoned(fs_info)) {
3187 mutex_lock(&fs_info->tree_root->log_mutex);
3188 if (!log_root_tree->node) {
3189 ret = btrfs_alloc_log_tree_node(trans, log_root_tree);
3191 mutex_unlock(&fs_info->tree_root->log_mutex);
3192 blk_finish_plug(&plug);
3196 mutex_unlock(&fs_info->tree_root->log_mutex);
3199 btrfs_init_log_ctx(&root_log_ctx, NULL);
3201 mutex_lock(&log_root_tree->log_mutex);
3203 index2 = log_root_tree->log_transid % 2;
3204 list_add_tail(&root_log_ctx.list, &log_root_tree->log_ctxs[index2]);
3205 root_log_ctx.log_transid = log_root_tree->log_transid;
3208 * Now we are safe to update the log_root_tree because we're under the
3209 * log_mutex, and we're a current writer so we're holding the commit
3210 * open until we drop the log_mutex.
3212 ret = update_log_root(trans, log, &new_root_item);
3214 if (!list_empty(&root_log_ctx.list))
3215 list_del_init(&root_log_ctx.list);
3217 blk_finish_plug(&plug);
3218 btrfs_set_log_full_commit(trans);
3220 if (ret != -ENOSPC) {
3221 btrfs_abort_transaction(trans, ret);
3222 mutex_unlock(&log_root_tree->log_mutex);
3225 btrfs_wait_tree_log_extents(log, mark);
3226 mutex_unlock(&log_root_tree->log_mutex);
3227 ret = BTRFS_LOG_FORCE_COMMIT;
3231 if (log_root_tree->log_transid_committed >= root_log_ctx.log_transid) {
3232 blk_finish_plug(&plug);
3233 list_del_init(&root_log_ctx.list);
3234 mutex_unlock(&log_root_tree->log_mutex);
3235 ret = root_log_ctx.log_ret;
3239 index2 = root_log_ctx.log_transid % 2;
3240 if (atomic_read(&log_root_tree->log_commit[index2])) {
3241 blk_finish_plug(&plug);
3242 ret = btrfs_wait_tree_log_extents(log, mark);
3243 wait_log_commit(log_root_tree,
3244 root_log_ctx.log_transid);
3245 mutex_unlock(&log_root_tree->log_mutex);
3247 ret = root_log_ctx.log_ret;
3250 ASSERT(root_log_ctx.log_transid == log_root_tree->log_transid);
3251 atomic_set(&log_root_tree->log_commit[index2], 1);
3253 if (atomic_read(&log_root_tree->log_commit[(index2 + 1) % 2])) {
3254 wait_log_commit(log_root_tree,
3255 root_log_ctx.log_transid - 1);
3259 * now that we've moved on to the tree of log tree roots,
3260 * check the full commit flag again
3262 if (btrfs_need_log_full_commit(trans)) {
3263 blk_finish_plug(&plug);
3264 btrfs_wait_tree_log_extents(log, mark);
3265 mutex_unlock(&log_root_tree->log_mutex);
3266 ret = BTRFS_LOG_FORCE_COMMIT;
3267 goto out_wake_log_root;
3270 ret = btrfs_write_marked_extents(fs_info,
3271 &log_root_tree->dirty_log_pages,
3272 EXTENT_DIRTY | EXTENT_NEW);
3273 blk_finish_plug(&plug);
3275 * As described above, -EAGAIN indicates a hole in the extents. We
3276 * cannot wait for these write outs since the waiting cause a
3277 * deadlock. Bail out to the full commit instead.
3279 if (ret == -EAGAIN && btrfs_is_zoned(fs_info)) {
3280 btrfs_set_log_full_commit(trans);
3281 btrfs_wait_tree_log_extents(log, mark);
3282 mutex_unlock(&log_root_tree->log_mutex);
3283 goto out_wake_log_root;
3285 btrfs_set_log_full_commit(trans);
3286 btrfs_abort_transaction(trans, ret);
3287 mutex_unlock(&log_root_tree->log_mutex);
3288 goto out_wake_log_root;
3290 ret = btrfs_wait_tree_log_extents(log, mark);
3292 ret = btrfs_wait_tree_log_extents(log_root_tree,
3293 EXTENT_NEW | EXTENT_DIRTY);
3295 btrfs_set_log_full_commit(trans);
3296 mutex_unlock(&log_root_tree->log_mutex);
3297 goto out_wake_log_root;
3300 log_root_start = log_root_tree->node->start;
3301 log_root_level = btrfs_header_level(log_root_tree->node);
3302 log_root_tree->log_transid++;
3303 mutex_unlock(&log_root_tree->log_mutex);
3306 * Here we are guaranteed that nobody is going to write the superblock
3307 * for the current transaction before us and that neither we do write
3308 * our superblock before the previous transaction finishes its commit
3309 * and writes its superblock, because:
3311 * 1) We are holding a handle on the current transaction, so no body
3312 * can commit it until we release the handle;
3314 * 2) Before writing our superblock we acquire the tree_log_mutex, so
3315 * if the previous transaction is still committing, and hasn't yet
3316 * written its superblock, we wait for it to do it, because a
3317 * transaction commit acquires the tree_log_mutex when the commit
3318 * begins and releases it only after writing its superblock.
3320 mutex_lock(&fs_info->tree_log_mutex);
3323 * The previous transaction writeout phase could have failed, and thus
3324 * marked the fs in an error state. We must not commit here, as we
3325 * could have updated our generation in the super_for_commit and
3326 * writing the super here would result in transid mismatches. If there
3327 * is an error here just bail.
3329 if (BTRFS_FS_ERROR(fs_info)) {
3331 btrfs_set_log_full_commit(trans);
3332 btrfs_abort_transaction(trans, ret);
3333 mutex_unlock(&fs_info->tree_log_mutex);
3334 goto out_wake_log_root;
3337 btrfs_set_super_log_root(fs_info->super_for_commit, log_root_start);
3338 btrfs_set_super_log_root_level(fs_info->super_for_commit, log_root_level);
3339 ret = write_all_supers(fs_info, 1);
3340 mutex_unlock(&fs_info->tree_log_mutex);
3342 btrfs_set_log_full_commit(trans);
3343 btrfs_abort_transaction(trans, ret);
3344 goto out_wake_log_root;
3348 * We know there can only be one task here, since we have not yet set
3349 * root->log_commit[index1] to 0 and any task attempting to sync the
3350 * log must wait for the previous log transaction to commit if it's
3351 * still in progress or wait for the current log transaction commit if
3352 * someone else already started it. We use <= and not < because the
3353 * first log transaction has an ID of 0.
3355 ASSERT(root->last_log_commit <= log_transid);
3356 root->last_log_commit = log_transid;
3359 mutex_lock(&log_root_tree->log_mutex);
3360 btrfs_remove_all_log_ctxs(log_root_tree, index2, ret);
3362 log_root_tree->log_transid_committed++;
3363 atomic_set(&log_root_tree->log_commit[index2], 0);
3364 mutex_unlock(&log_root_tree->log_mutex);
3367 * The barrier before waitqueue_active (in cond_wake_up) is needed so
3368 * all the updates above are seen by the woken threads. It might not be
3369 * necessary, but proving that seems to be hard.
3371 cond_wake_up(&log_root_tree->log_commit_wait[index2]);
3373 mutex_lock(&root->log_mutex);
3374 btrfs_remove_all_log_ctxs(root, index1, ret);
3375 root->log_transid_committed++;
3376 atomic_set(&root->log_commit[index1], 0);
3377 mutex_unlock(&root->log_mutex);
3380 * The barrier before waitqueue_active (in cond_wake_up) is needed so
3381 * all the updates above are seen by the woken threads. It might not be
3382 * necessary, but proving that seems to be hard.
3384 cond_wake_up(&root->log_commit_wait[index1]);
3388 static void free_log_tree(struct btrfs_trans_handle *trans,
3389 struct btrfs_root *log)
3392 struct walk_control wc = {
3394 .process_func = process_one_buffer
3398 ret = walk_log_tree(trans, log, &wc);
3401 * We weren't able to traverse the entire log tree, the
3402 * typical scenario is getting an -EIO when reading an
3403 * extent buffer of the tree, due to a previous writeback
3406 set_bit(BTRFS_FS_STATE_LOG_CLEANUP_ERROR,
3407 &log->fs_info->fs_state);
3410 * Some extent buffers of the log tree may still be dirty
3411 * and not yet written back to storage, because we may
3412 * have updates to a log tree without syncing a log tree,
3413 * such as during rename and link operations. So flush
3414 * them out and wait for their writeback to complete, so
3415 * that we properly cleanup their state and pages.
3417 btrfs_write_marked_extents(log->fs_info,
3418 &log->dirty_log_pages,
3419 EXTENT_DIRTY | EXTENT_NEW);
3420 btrfs_wait_tree_log_extents(log,
3421 EXTENT_DIRTY | EXTENT_NEW);
3424 btrfs_abort_transaction(trans, ret);
3426 btrfs_handle_fs_error(log->fs_info, ret, NULL);
3430 clear_extent_bits(&log->dirty_log_pages, 0, (u64)-1,
3431 EXTENT_DIRTY | EXTENT_NEW | EXTENT_NEED_WAIT);
3432 extent_io_tree_release(&log->log_csum_range);
3434 btrfs_put_root(log);
3438 * free all the extents used by the tree log. This should be called
3439 * at commit time of the full transaction
3441 int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
3443 if (root->log_root) {
3444 free_log_tree(trans, root->log_root);
3445 root->log_root = NULL;
3446 clear_bit(BTRFS_ROOT_HAS_LOG_TREE, &root->state);
3451 int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
3452 struct btrfs_fs_info *fs_info)
3454 if (fs_info->log_root_tree) {
3455 free_log_tree(trans, fs_info->log_root_tree);
3456 fs_info->log_root_tree = NULL;
3457 clear_bit(BTRFS_ROOT_HAS_LOG_TREE, &fs_info->tree_root->state);
3463 * Check if an inode was logged in the current transaction. This correctly deals
3464 * with the case where the inode was logged but has a logged_trans of 0, which
3465 * happens if the inode is evicted and loaded again, as logged_trans is an in
3466 * memory only field (not persisted).
3468 * Returns 1 if the inode was logged before in the transaction, 0 if it was not,
3471 static int inode_logged(struct btrfs_trans_handle *trans,
3472 struct btrfs_inode *inode,
3473 struct btrfs_path *path_in)
3475 struct btrfs_path *path = path_in;
3476 struct btrfs_key key;
3479 if (inode->logged_trans == trans->transid)
3483 * If logged_trans is not 0, then we know the inode logged was not logged
3484 * in this transaction, so we can return false right away.
3486 if (inode->logged_trans > 0)
3490 * If no log tree was created for this root in this transaction, then
3491 * the inode can not have been logged in this transaction. In that case
3492 * set logged_trans to anything greater than 0 and less than the current
3493 * transaction's ID, to avoid the search below in a future call in case
3494 * a log tree gets created after this.
3496 if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &inode->root->state)) {
3497 inode->logged_trans = trans->transid - 1;
3502 * We have a log tree and the inode's logged_trans is 0. We can't tell
3503 * for sure if the inode was logged before in this transaction by looking
3504 * only at logged_trans. We could be pessimistic and assume it was, but
3505 * that can lead to unnecessarily logging an inode during rename and link
3506 * operations, and then further updating the log in followup rename and
3507 * link operations, specially if it's a directory, which adds latency
3508 * visible to applications doing a series of rename or link operations.
3510 * A logged_trans of 0 here can mean several things:
3512 * 1) The inode was never logged since the filesystem was mounted, and may
3513 * or may have not been evicted and loaded again;
3515 * 2) The inode was logged in a previous transaction, then evicted and
3516 * then loaded again;
3518 * 3) The inode was logged in the current transaction, then evicted and
3519 * then loaded again.
3521 * For cases 1) and 2) we don't want to return true, but we need to detect
3522 * case 3) and return true. So we do a search in the log root for the inode
3525 key.objectid = btrfs_ino(inode);
3526 key.type = BTRFS_INODE_ITEM_KEY;
3530 path = btrfs_alloc_path();
3535 ret = btrfs_search_slot(NULL, inode->root->log_root, &key, path, 0, 0);
3538 btrfs_release_path(path);
3540 btrfs_free_path(path);
3543 * Logging an inode always results in logging its inode item. So if we
3544 * did not find the item we know the inode was not logged for sure.
3548 } else if (ret > 0) {
3550 * Set logged_trans to a value greater than 0 and less then the
3551 * current transaction to avoid doing the search in future calls.
3553 inode->logged_trans = trans->transid - 1;
3558 * The inode was previously logged and then evicted, set logged_trans to
3559 * the current transacion's ID, to avoid future tree searches as long as
3560 * the inode is not evicted again.
3562 inode->logged_trans = trans->transid;
3565 * If it's a directory, then we must set last_dir_index_offset to the
3566 * maximum possible value, so that the next attempt to log the inode does
3567 * not skip checking if dir index keys found in modified subvolume tree
3568 * leaves have been logged before, otherwise it would result in attempts
3569 * to insert duplicate dir index keys in the log tree. This must be done
3570 * because last_dir_index_offset is an in-memory only field, not persisted
3571 * in the inode item or any other on-disk structure, so its value is lost
3572 * once the inode is evicted.
3574 if (S_ISDIR(inode->vfs_inode.i_mode))
3575 inode->last_dir_index_offset = (u64)-1;
3581 * Delete a directory entry from the log if it exists.
3583 * Returns < 0 on error
3584 * 1 if the entry does not exists
3585 * 0 if the entry existed and was successfully deleted
3587 static int del_logged_dentry(struct btrfs_trans_handle *trans,
3588 struct btrfs_root *log,
3589 struct btrfs_path *path,
3591 const char *name, int name_len,
3594 struct btrfs_dir_item *di;
3597 * We only log dir index items of a directory, so we don't need to look
3598 * for dir item keys.
3600 di = btrfs_lookup_dir_index_item(trans, log, path, dir_ino,
3601 index, name, name_len, -1);
3608 * We do not need to update the size field of the directory's
3609 * inode item because on log replay we update the field to reflect
3610 * all existing entries in the directory (see overwrite_item()).
3612 return btrfs_delete_one_dir_name(trans, log, path, di);
3616 * If both a file and directory are logged, and unlinks or renames are
3617 * mixed in, we have a few interesting corners:
3619 * create file X in dir Y
3620 * link file X to X.link in dir Y
3622 * unlink file X but leave X.link
3625 * After a crash we would expect only X.link to exist. But file X
3626 * didn't get fsync'd again so the log has back refs for X and X.link.
3628 * We solve this by removing directory entries and inode backrefs from the
3629 * log when a file that was logged in the current transaction is
3630 * unlinked. Any later fsync will include the updated log entries, and
3631 * we'll be able to reconstruct the proper directory items from backrefs.
3633 * This optimizations allows us to avoid relogging the entire inode
3634 * or the entire directory.
3636 void btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
3637 struct btrfs_root *root,
3638 const char *name, int name_len,
3639 struct btrfs_inode *dir, u64 index)
3641 struct btrfs_path *path;
3644 ret = inode_logged(trans, dir, NULL);
3648 btrfs_set_log_full_commit(trans);
3652 ret = join_running_log_trans(root);
3656 mutex_lock(&dir->log_mutex);
3658 path = btrfs_alloc_path();
3664 ret = del_logged_dentry(trans, root->log_root, path, btrfs_ino(dir),
3665 name, name_len, index);
3666 btrfs_free_path(path);
3668 mutex_unlock(&dir->log_mutex);
3670 btrfs_set_log_full_commit(trans);
3671 btrfs_end_log_trans(root);
3674 /* see comments for btrfs_del_dir_entries_in_log */
3675 void btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
3676 struct btrfs_root *root,
3677 const char *name, int name_len,
3678 struct btrfs_inode *inode, u64 dirid)
3680 struct btrfs_root *log;
3684 ret = inode_logged(trans, inode, NULL);
3688 btrfs_set_log_full_commit(trans);
3692 ret = join_running_log_trans(root);
3695 log = root->log_root;
3696 mutex_lock(&inode->log_mutex);
3698 ret = btrfs_del_inode_ref(trans, log, name, name_len, btrfs_ino(inode),
3700 mutex_unlock(&inode->log_mutex);
3701 if (ret < 0 && ret != -ENOENT)
3702 btrfs_set_log_full_commit(trans);
3703 btrfs_end_log_trans(root);
3707 * creates a range item in the log for 'dirid'. first_offset and
3708 * last_offset tell us which parts of the key space the log should
3709 * be considered authoritative for.
3711 static noinline int insert_dir_log_key(struct btrfs_trans_handle *trans,
3712 struct btrfs_root *log,
3713 struct btrfs_path *path,
3715 u64 first_offset, u64 last_offset)
3718 struct btrfs_key key;
3719 struct btrfs_dir_log_item *item;
3721 key.objectid = dirid;
3722 key.offset = first_offset;
3723 key.type = BTRFS_DIR_LOG_INDEX_KEY;
3724 ret = btrfs_insert_empty_item(trans, log, path, &key, sizeof(*item));
3726 * -EEXIST is fine and can happen sporadically when we are logging a
3727 * directory and have concurrent insertions in the subvolume's tree for
3728 * items from other inodes and that result in pushing off some dir items
3729 * from one leaf to another in order to accommodate for the new items.
3730 * This results in logging the same dir index range key.
3732 if (ret && ret != -EEXIST)
3735 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
3736 struct btrfs_dir_log_item);
3737 if (ret == -EEXIST) {
3738 const u64 curr_end = btrfs_dir_log_end(path->nodes[0], item);
3741 * btrfs_del_dir_entries_in_log() might have been called during
3742 * an unlink between the initial insertion of this key and the
3743 * current update, or we might be logging a single entry deletion
3744 * during a rename, so set the new last_offset to the max value.
3746 last_offset = max(last_offset, curr_end);
3748 btrfs_set_dir_log_end(path->nodes[0], item, last_offset);
3749 btrfs_mark_buffer_dirty(path->nodes[0]);
3750 btrfs_release_path(path);
3754 static int flush_dir_items_batch(struct btrfs_trans_handle *trans,
3755 struct btrfs_root *log,
3756 struct extent_buffer *src,
3757 struct btrfs_path *dst_path,
3761 char *ins_data = NULL;
3762 struct btrfs_item_batch batch;
3763 struct extent_buffer *dst;
3764 unsigned long src_offset;
3765 unsigned long dst_offset;
3766 struct btrfs_key key;
3775 btrfs_item_key_to_cpu(src, &key, start_slot);
3776 item_size = btrfs_item_size(src, start_slot);
3778 batch.data_sizes = &item_size;
3779 batch.total_data_size = item_size;
3781 struct btrfs_key *ins_keys;
3784 ins_data = kmalloc(count * sizeof(u32) +
3785 count * sizeof(struct btrfs_key), GFP_NOFS);
3789 ins_sizes = (u32 *)ins_data;
3790 ins_keys = (struct btrfs_key *)(ins_data + count * sizeof(u32));
3791 batch.keys = ins_keys;
3792 batch.data_sizes = ins_sizes;
3793 batch.total_data_size = 0;
3795 for (i = 0; i < count; i++) {
3796 const int slot = start_slot + i;
3798 btrfs_item_key_to_cpu(src, &ins_keys[i], slot);
3799 ins_sizes[i] = btrfs_item_size(src, slot);
3800 batch.total_data_size += ins_sizes[i];
3804 ret = btrfs_insert_empty_items(trans, log, dst_path, &batch);
3808 dst = dst_path->nodes[0];
3810 * Copy all the items in bulk, in a single copy operation. Item data is
3811 * organized such that it's placed at the end of a leaf and from right
3812 * to left. For example, the data for the second item ends at an offset
3813 * that matches the offset where the data for the first item starts, the
3814 * data for the third item ends at an offset that matches the offset
3815 * where the data of the second items starts, and so on.
3816 * Therefore our source and destination start offsets for copy match the
3817 * offsets of the last items (highest slots).
3819 dst_offset = btrfs_item_ptr_offset(dst, dst_path->slots[0] + count - 1);
3820 src_offset = btrfs_item_ptr_offset(src, start_slot + count - 1);
3821 copy_extent_buffer(dst, src, dst_offset, src_offset, batch.total_data_size);
3822 btrfs_release_path(dst_path);
3829 static int process_dir_items_leaf(struct btrfs_trans_handle *trans,
3830 struct btrfs_inode *inode,
3831 struct btrfs_path *path,
3832 struct btrfs_path *dst_path,
3833 struct btrfs_log_ctx *ctx,
3834 u64 *last_old_dentry_offset)
3836 struct btrfs_root *log = inode->root->log_root;
3837 struct extent_buffer *src = path->nodes[0];
3838 const int nritems = btrfs_header_nritems(src);
3839 const u64 ino = btrfs_ino(inode);
3840 bool last_found = false;
3841 int batch_start = 0;
3845 for (i = path->slots[0]; i < nritems; i++) {
3846 struct btrfs_dir_item *di;
3847 struct btrfs_key key;
3850 btrfs_item_key_to_cpu(src, &key, i);
3852 if (key.objectid != ino || key.type != BTRFS_DIR_INDEX_KEY) {
3857 di = btrfs_item_ptr(src, i, struct btrfs_dir_item);
3858 ctx->last_dir_item_offset = key.offset;
3861 * Skip ranges of items that consist only of dir item keys created
3862 * in past transactions. However if we find a gap, we must log a
3863 * dir index range item for that gap, so that index keys in that
3864 * gap are deleted during log replay.
3866 if (btrfs_dir_transid(src, di) < trans->transid) {
3867 if (key.offset > *last_old_dentry_offset + 1) {
3868 ret = insert_dir_log_key(trans, log, dst_path,
3869 ino, *last_old_dentry_offset + 1,
3875 *last_old_dentry_offset = key.offset;
3879 * We must make sure that when we log a directory entry, the
3880 * corresponding inode, after log replay, has a matching link
3881 * count. For example:
3887 * xfs_io -c "fsync" mydir
3889 * <mount fs and log replay>
3891 * Would result in a fsync log that when replayed, our file inode
3892 * would have a link count of 1, but we get two directory entries
3893 * pointing to the same inode. After removing one of the names,
3894 * it would not be possible to remove the other name, which
3895 * resulted always in stale file handle errors, and would not be
3896 * possible to rmdir the parent directory, since its i_size could
3897 * never be decremented to the value BTRFS_EMPTY_DIR_SIZE,
3898 * resulting in -ENOTEMPTY errors.
3900 if (!ctx->log_new_dentries) {
3901 struct btrfs_key di_key;
3903 btrfs_dir_item_key_to_cpu(src, di, &di_key);
3904 if (di_key.type != BTRFS_ROOT_ITEM_KEY)
3905 ctx->log_new_dentries = true;
3908 if (!ctx->logged_before)
3912 * If we were logged before and have logged dir items, we can skip
3913 * checking if any item with a key offset larger than the last one
3914 * we logged is in the log tree, saving time and avoiding adding
3915 * contention on the log tree. We can only rely on the value of
3916 * last_dir_index_offset when we know for sure that the inode was
3917 * previously logged in the current transaction.
3919 if (key.offset > inode->last_dir_index_offset)
3922 * Check if the key was already logged before. If not we can add
3923 * it to a batch for bulk insertion.
3925 ret = btrfs_search_slot(NULL, log, &key, dst_path, 0, 0);
3928 } else if (ret > 0) {
3929 btrfs_release_path(dst_path);
3934 * Item exists in the log. Overwrite the item in the log if it
3935 * has different content or do nothing if it has exactly the same
3936 * content. And then flush the current batch if any - do it after
3937 * overwriting the current item, or we would deadlock otherwise,
3938 * since we are holding a path for the existing item.
3940 ret = do_overwrite_item(trans, log, dst_path, src, i, &key);
3944 if (batch_size > 0) {
3945 ret = flush_dir_items_batch(trans, log, src, dst_path,
3946 batch_start, batch_size);
3953 if (batch_size == 0)
3958 if (batch_size > 0) {
3961 ret = flush_dir_items_batch(trans, log, src, dst_path,
3962 batch_start, batch_size);
3967 return last_found ? 1 : 0;
3971 * log all the items included in the current transaction for a given
3972 * directory. This also creates the range items in the log tree required
3973 * to replay anything deleted before the fsync
3975 static noinline int log_dir_items(struct btrfs_trans_handle *trans,
3976 struct btrfs_inode *inode,
3977 struct btrfs_path *path,
3978 struct btrfs_path *dst_path,
3979 struct btrfs_log_ctx *ctx,
3980 u64 min_offset, u64 *last_offset_ret)
3982 struct btrfs_key min_key;
3983 struct btrfs_root *root = inode->root;
3984 struct btrfs_root *log = root->log_root;
3987 u64 last_old_dentry_offset = min_offset - 1;
3988 u64 last_offset = (u64)-1;
3989 u64 ino = btrfs_ino(inode);
3991 min_key.objectid = ino;
3992 min_key.type = BTRFS_DIR_INDEX_KEY;
3993 min_key.offset = min_offset;
3995 ret = btrfs_search_forward(root, &min_key, path, trans->transid);
3998 * we didn't find anything from this transaction, see if there
3999 * is anything at all
4001 if (ret != 0 || min_key.objectid != ino ||
4002 min_key.type != BTRFS_DIR_INDEX_KEY) {
4003 min_key.objectid = ino;
4004 min_key.type = BTRFS_DIR_INDEX_KEY;
4005 min_key.offset = (u64)-1;
4006 btrfs_release_path(path);
4007 ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
4009 btrfs_release_path(path);
4012 ret = btrfs_previous_item(root, path, ino, BTRFS_DIR_INDEX_KEY);
4014 /* if ret == 0 there are items for this type,
4015 * create a range to tell us the last key of this type.
4016 * otherwise, there are no items in this directory after
4017 * *min_offset, and we create a range to indicate that.
4020 struct btrfs_key tmp;
4022 btrfs_item_key_to_cpu(path->nodes[0], &tmp,
4024 if (tmp.type == BTRFS_DIR_INDEX_KEY)
4025 last_old_dentry_offset = tmp.offset;
4030 /* go backward to find any previous key */
4031 ret = btrfs_previous_item(root, path, ino, BTRFS_DIR_INDEX_KEY);
4033 struct btrfs_key tmp;
4035 btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
4037 * The dir index key before the first one we found that needs to
4038 * be logged might be in a previous leaf, and there might be a
4039 * gap between these keys, meaning that we had deletions that
4040 * happened. So the key range item we log (key type
4041 * BTRFS_DIR_LOG_INDEX_KEY) must cover a range that starts at the
4042 * previous key's offset plus 1, so that those deletes are replayed.
4044 if (tmp.type == BTRFS_DIR_INDEX_KEY)
4045 last_old_dentry_offset = tmp.offset;
4047 btrfs_release_path(path);
4050 * Find the first key from this transaction again. See the note for
4051 * log_new_dir_dentries, if we're logging a directory recursively we
4052 * won't be holding its i_mutex, which means we can modify the directory
4053 * while we're logging it. If we remove an entry between our first
4054 * search and this search we'll not find the key again and can just
4058 ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
4063 * we have a block from this transaction, log every item in it
4064 * from our directory
4067 ret = process_dir_items_leaf(trans, inode, path, dst_path, ctx,
4068 &last_old_dentry_offset);
4074 path->slots[0] = btrfs_header_nritems(path->nodes[0]);
4077 * look ahead to the next item and see if it is also
4078 * from this directory and from this transaction
4080 ret = btrfs_next_leaf(root, path);
4083 last_offset = (u64)-1;
4088 btrfs_item_key_to_cpu(path->nodes[0], &min_key, path->slots[0]);
4089 if (min_key.objectid != ino || min_key.type != BTRFS_DIR_INDEX_KEY) {
4090 last_offset = (u64)-1;
4093 if (btrfs_header_generation(path->nodes[0]) != trans->transid) {
4095 * The next leaf was not changed in the current transaction
4096 * and has at least one dir index key.
4097 * We check for the next key because there might have been
4098 * one or more deletions between the last key we logged and
4099 * that next key. So the key range item we log (key type
4100 * BTRFS_DIR_LOG_INDEX_KEY) must end at the next key's
4101 * offset minus 1, so that those deletes are replayed.
4103 last_offset = min_key.offset - 1;
4106 if (need_resched()) {
4107 btrfs_release_path(path);
4113 btrfs_release_path(path);
4114 btrfs_release_path(dst_path);
4117 *last_offset_ret = last_offset;
4119 * In case the leaf was changed in the current transaction but
4120 * all its dir items are from a past transaction, the last item
4121 * in the leaf is a dir item and there's no gap between that last
4122 * dir item and the first one on the next leaf (which did not
4123 * change in the current transaction), then we don't need to log
4124 * a range, last_old_dentry_offset is == to last_offset.
4126 ASSERT(last_old_dentry_offset <= last_offset);
4127 if (last_old_dentry_offset < last_offset) {
4128 ret = insert_dir_log_key(trans, log, path, ino,
4129 last_old_dentry_offset + 1,
4139 * logging directories is very similar to logging inodes, We find all the items
4140 * from the current transaction and write them to the log.
4142 * The recovery code scans the directory in the subvolume, and if it finds a
4143 * key in the range logged that is not present in the log tree, then it means
4144 * that dir entry was unlinked during the transaction.
4146 * In order for that scan to work, we must include one key smaller than
4147 * the smallest logged by this transaction and one key larger than the largest
4148 * key logged by this transaction.
4150 static noinline int log_directory_changes(struct btrfs_trans_handle *trans,
4151 struct btrfs_inode *inode,
4152 struct btrfs_path *path,
4153 struct btrfs_path *dst_path,
4154 struct btrfs_log_ctx *ctx)
4160 min_key = BTRFS_DIR_START_INDEX;
4162 ctx->last_dir_item_offset = inode->last_dir_index_offset;
4165 ret = log_dir_items(trans, inode, path, dst_path,
4166 ctx, min_key, &max_key);
4169 if (max_key == (u64)-1)
4171 min_key = max_key + 1;
4174 inode->last_dir_index_offset = ctx->last_dir_item_offset;
4180 * a helper function to drop items from the log before we relog an
4181 * inode. max_key_type indicates the highest item type to remove.
4182 * This cannot be run for file data extents because it does not
4183 * free the extents they point to.
4185 static int drop_inode_items(struct btrfs_trans_handle *trans,
4186 struct btrfs_root *log,
4187 struct btrfs_path *path,
4188 struct btrfs_inode *inode,
4192 struct btrfs_key key;
4193 struct btrfs_key found_key;
4196 key.objectid = btrfs_ino(inode);
4197 key.type = max_key_type;
4198 key.offset = (u64)-1;
4201 ret = btrfs_search_slot(trans, log, &key, path, -1, 1);
4202 BUG_ON(ret == 0); /* Logic error */
4206 if (path->slots[0] == 0)
4210 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
4213 if (found_key.objectid != key.objectid)
4216 found_key.offset = 0;
4218 ret = btrfs_bin_search(path->nodes[0], &found_key, &start_slot);
4222 ret = btrfs_del_items(trans, log, path, start_slot,
4223 path->slots[0] - start_slot + 1);
4225 * If start slot isn't 0 then we don't need to re-search, we've
4226 * found the last guy with the objectid in this tree.
4228 if (ret || start_slot != 0)
4230 btrfs_release_path(path);
4232 btrfs_release_path(path);
4238 static int truncate_inode_items(struct btrfs_trans_handle *trans,
4239 struct btrfs_root *log_root,
4240 struct btrfs_inode *inode,
4241 u64 new_size, u32 min_type)
4243 struct btrfs_truncate_control control = {
4244 .new_size = new_size,
4245 .ino = btrfs_ino(inode),
4246 .min_type = min_type,
4247 .skip_ref_updates = true,
4250 return btrfs_truncate_inode_items(trans, log_root, &control);
4253 static void fill_inode_item(struct btrfs_trans_handle *trans,
4254 struct extent_buffer *leaf,
4255 struct btrfs_inode_item *item,
4256 struct inode *inode, int log_inode_only,
4259 struct btrfs_map_token token;
4262 btrfs_init_map_token(&token, leaf);
4264 if (log_inode_only) {
4265 /* set the generation to zero so the recover code
4266 * can tell the difference between an logging
4267 * just to say 'this inode exists' and a logging
4268 * to say 'update this inode with these values'
4270 btrfs_set_token_inode_generation(&token, item, 0);
4271 btrfs_set_token_inode_size(&token, item, logged_isize);
4273 btrfs_set_token_inode_generation(&token, item,
4274 BTRFS_I(inode)->generation);
4275 btrfs_set_token_inode_size(&token, item, inode->i_size);
4278 btrfs_set_token_inode_uid(&token, item, i_uid_read(inode));
4279 btrfs_set_token_inode_gid(&token, item, i_gid_read(inode));
4280 btrfs_set_token_inode_mode(&token, item, inode->i_mode);
4281 btrfs_set_token_inode_nlink(&token, item, inode->i_nlink);
4283 btrfs_set_token_timespec_sec(&token, &item->atime,
4284 inode->i_atime.tv_sec);
4285 btrfs_set_token_timespec_nsec(&token, &item->atime,
4286 inode->i_atime.tv_nsec);
4288 btrfs_set_token_timespec_sec(&token, &item->mtime,
4289 inode->i_mtime.tv_sec);
4290 btrfs_set_token_timespec_nsec(&token, &item->mtime,
4291 inode->i_mtime.tv_nsec);
4293 btrfs_set_token_timespec_sec(&token, &item->ctime,
4294 inode->i_ctime.tv_sec);
4295 btrfs_set_token_timespec_nsec(&token, &item->ctime,
4296 inode->i_ctime.tv_nsec);
4299 * We do not need to set the nbytes field, in fact during a fast fsync
4300 * its value may not even be correct, since a fast fsync does not wait
4301 * for ordered extent completion, which is where we update nbytes, it
4302 * only waits for writeback to complete. During log replay as we find
4303 * file extent items and replay them, we adjust the nbytes field of the
4304 * inode item in subvolume tree as needed (see overwrite_item()).
4307 btrfs_set_token_inode_sequence(&token, item, inode_peek_iversion(inode));
4308 btrfs_set_token_inode_transid(&token, item, trans->transid);
4309 btrfs_set_token_inode_rdev(&token, item, inode->i_rdev);
4310 flags = btrfs_inode_combine_flags(BTRFS_I(inode)->flags,
4311 BTRFS_I(inode)->ro_flags);
4312 btrfs_set_token_inode_flags(&token, item, flags);
4313 btrfs_set_token_inode_block_group(&token, item, 0);
4316 static int log_inode_item(struct btrfs_trans_handle *trans,
4317 struct btrfs_root *log, struct btrfs_path *path,
4318 struct btrfs_inode *inode, bool inode_item_dropped)
4320 struct btrfs_inode_item *inode_item;
4324 * If we are doing a fast fsync and the inode was logged before in the
4325 * current transaction, then we know the inode was previously logged and
4326 * it exists in the log tree. For performance reasons, in this case use
4327 * btrfs_search_slot() directly with ins_len set to 0 so that we never
4328 * attempt a write lock on the leaf's parent, which adds unnecessary lock
4329 * contention in case there are concurrent fsyncs for other inodes of the
4330 * same subvolume. Using btrfs_insert_empty_item() when the inode item
4331 * already exists can also result in unnecessarily splitting a leaf.
4333 if (!inode_item_dropped && inode->logged_trans == trans->transid) {
4334 ret = btrfs_search_slot(trans, log, &inode->location, path, 0, 1);
4340 * This means it is the first fsync in the current transaction,
4341 * so the inode item is not in the log and we need to insert it.
4342 * We can never get -EEXIST because we are only called for a fast
4343 * fsync and in case an inode eviction happens after the inode was
4344 * logged before in the current transaction, when we load again
4345 * the inode, we set BTRFS_INODE_NEEDS_FULL_SYNC on its runtime
4346 * flags and set ->logged_trans to 0.
4348 ret = btrfs_insert_empty_item(trans, log, path, &inode->location,
4349 sizeof(*inode_item));
4350 ASSERT(ret != -EEXIST);
4354 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
4355 struct btrfs_inode_item);
4356 fill_inode_item(trans, path->nodes[0], inode_item, &inode->vfs_inode,
4358 btrfs_release_path(path);
4362 static int log_csums(struct btrfs_trans_handle *trans,
4363 struct btrfs_inode *inode,
4364 struct btrfs_root *log_root,
4365 struct btrfs_ordered_sum *sums)
4367 const u64 lock_end = sums->bytenr + sums->len - 1;
4368 struct extent_state *cached_state = NULL;
4372 * If this inode was not used for reflink operations in the current
4373 * transaction with new extents, then do the fast path, no need to
4374 * worry about logging checksum items with overlapping ranges.
4376 if (inode->last_reflink_trans < trans->transid)
4377 return btrfs_csum_file_blocks(trans, log_root, sums);
4380 * Serialize logging for checksums. This is to avoid racing with the
4381 * same checksum being logged by another task that is logging another
4382 * file which happens to refer to the same extent as well. Such races
4383 * can leave checksum items in the log with overlapping ranges.
4385 ret = lock_extent_bits(&log_root->log_csum_range, sums->bytenr,
4386 lock_end, &cached_state);
4390 * Due to extent cloning, we might have logged a csum item that covers a
4391 * subrange of a cloned extent, and later we can end up logging a csum
4392 * item for a larger subrange of the same extent or the entire range.
4393 * This would leave csum items in the log tree that cover the same range
4394 * and break the searches for checksums in the log tree, resulting in
4395 * some checksums missing in the fs/subvolume tree. So just delete (or
4396 * trim and adjust) any existing csum items in the log for this range.
4398 ret = btrfs_del_csums(trans, log_root, sums->bytenr, sums->len);
4400 ret = btrfs_csum_file_blocks(trans, log_root, sums);
4402 unlock_extent_cached(&log_root->log_csum_range, sums->bytenr, lock_end,
4408 static noinline int copy_items(struct btrfs_trans_handle *trans,
4409 struct btrfs_inode *inode,
4410 struct btrfs_path *dst_path,
4411 struct btrfs_path *src_path,
4412 int start_slot, int nr, int inode_only,
4415 struct btrfs_root *log = inode->root->log_root;
4416 struct btrfs_file_extent_item *extent;
4417 struct extent_buffer *src = src_path->nodes[0];
4419 struct btrfs_key *ins_keys;
4421 struct btrfs_item_batch batch;
4425 const bool skip_csum = (inode->flags & BTRFS_INODE_NODATASUM);
4426 const u64 i_size = i_size_read(&inode->vfs_inode);
4428 ins_data = kmalloc(nr * sizeof(struct btrfs_key) +
4429 nr * sizeof(u32), GFP_NOFS);
4433 ins_sizes = (u32 *)ins_data;
4434 ins_keys = (struct btrfs_key *)(ins_data + nr * sizeof(u32));
4435 batch.keys = ins_keys;
4436 batch.data_sizes = ins_sizes;
4437 batch.total_data_size = 0;
4441 for (i = 0; i < nr; i++) {
4442 const int src_slot = start_slot + i;
4443 struct btrfs_root *csum_root;
4444 struct btrfs_ordered_sum *sums;
4445 struct btrfs_ordered_sum *sums_next;
4446 LIST_HEAD(ordered_sums);
4450 u64 extent_num_bytes;
4453 btrfs_item_key_to_cpu(src, &ins_keys[dst_index], src_slot);
4455 if (ins_keys[dst_index].type != BTRFS_EXTENT_DATA_KEY)
4458 extent = btrfs_item_ptr(src, src_slot,
4459 struct btrfs_file_extent_item);
4461 is_old_extent = (btrfs_file_extent_generation(src, extent) <
4465 * Don't copy extents from past generations. That would make us
4466 * log a lot more metadata for common cases like doing only a
4467 * few random writes into a file and then fsync it for the first
4468 * time or after the full sync flag is set on the inode. We can
4469 * get leaves full of extent items, most of which are from past
4470 * generations, so we can skip them - as long as the inode has
4471 * not been the target of a reflink operation in this transaction,
4472 * as in that case it might have had file extent items with old
4473 * generations copied into it. We also must always log prealloc
4474 * extents that start at or beyond eof, otherwise we would lose
4475 * them on log replay.
4477 if (is_old_extent &&
4478 ins_keys[dst_index].offset < i_size &&
4479 inode->last_reflink_trans < trans->transid)
4485 /* Only regular extents have checksums. */
4486 if (btrfs_file_extent_type(src, extent) != BTRFS_FILE_EXTENT_REG)
4490 * If it's an extent created in a past transaction, then its
4491 * checksums are already accessible from the committed csum tree,
4492 * no need to log them.
4497 disk_bytenr = btrfs_file_extent_disk_bytenr(src, extent);
4498 /* If it's an explicit hole, there are no checksums. */
4499 if (disk_bytenr == 0)
4502 disk_num_bytes = btrfs_file_extent_disk_num_bytes(src, extent);
4504 if (btrfs_file_extent_compression(src, extent)) {
4506 extent_num_bytes = disk_num_bytes;
4508 extent_offset = btrfs_file_extent_offset(src, extent);
4509 extent_num_bytes = btrfs_file_extent_num_bytes(src, extent);
4512 csum_root = btrfs_csum_root(trans->fs_info, disk_bytenr);
4513 disk_bytenr += extent_offset;
4514 ret = btrfs_lookup_csums_range(csum_root, disk_bytenr,
4515 disk_bytenr + extent_num_bytes - 1,
4520 list_for_each_entry_safe(sums, sums_next, &ordered_sums, list) {
4522 ret = log_csums(trans, inode, log, sums);
4523 list_del(&sums->list);
4530 ins_sizes[dst_index] = btrfs_item_size(src, src_slot);
4531 batch.total_data_size += ins_sizes[dst_index];
4537 * We have a leaf full of old extent items that don't need to be logged,
4538 * so we don't need to do anything.
4543 ret = btrfs_insert_empty_items(trans, log, dst_path, &batch);
4548 for (i = 0; i < nr; i++) {
4549 const int src_slot = start_slot + i;
4550 const int dst_slot = dst_path->slots[0] + dst_index;
4551 struct btrfs_key key;
4552 unsigned long src_offset;
4553 unsigned long dst_offset;
4556 * We're done, all the remaining items in the source leaf
4557 * correspond to old file extent items.
4559 if (dst_index >= batch.nr)
4562 btrfs_item_key_to_cpu(src, &key, src_slot);
4564 if (key.type != BTRFS_EXTENT_DATA_KEY)
4567 extent = btrfs_item_ptr(src, src_slot,
4568 struct btrfs_file_extent_item);
4570 /* See the comment in the previous loop, same logic. */
4571 if (btrfs_file_extent_generation(src, extent) < trans->transid &&
4572 key.offset < i_size &&
4573 inode->last_reflink_trans < trans->transid)
4577 dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0], dst_slot);
4578 src_offset = btrfs_item_ptr_offset(src, src_slot);
4580 if (key.type == BTRFS_INODE_ITEM_KEY) {
4581 struct btrfs_inode_item *inode_item;
4583 inode_item = btrfs_item_ptr(dst_path->nodes[0], dst_slot,
4584 struct btrfs_inode_item);
4585 fill_inode_item(trans, dst_path->nodes[0], inode_item,
4587 inode_only == LOG_INODE_EXISTS,
4590 copy_extent_buffer(dst_path->nodes[0], src, dst_offset,
4591 src_offset, ins_sizes[dst_index]);
4597 btrfs_mark_buffer_dirty(dst_path->nodes[0]);
4598 btrfs_release_path(dst_path);
4605 static int extent_cmp(void *priv, const struct list_head *a,
4606 const struct list_head *b)
4608 const struct extent_map *em1, *em2;
4610 em1 = list_entry(a, struct extent_map, list);
4611 em2 = list_entry(b, struct extent_map, list);
4613 if (em1->start < em2->start)
4615 else if (em1->start > em2->start)
4620 static int log_extent_csums(struct btrfs_trans_handle *trans,
4621 struct btrfs_inode *inode,
4622 struct btrfs_root *log_root,
4623 const struct extent_map *em,
4624 struct btrfs_log_ctx *ctx)
4626 struct btrfs_ordered_extent *ordered;
4627 struct btrfs_root *csum_root;
4630 u64 mod_start = em->mod_start;
4631 u64 mod_len = em->mod_len;
4632 LIST_HEAD(ordered_sums);
4635 if (inode->flags & BTRFS_INODE_NODATASUM ||
4636 test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
4637 em->block_start == EXTENT_MAP_HOLE)
4640 list_for_each_entry(ordered, &ctx->ordered_extents, log_list) {
4641 const u64 ordered_end = ordered->file_offset + ordered->num_bytes;
4642 const u64 mod_end = mod_start + mod_len;
4643 struct btrfs_ordered_sum *sums;
4648 if (ordered_end <= mod_start)
4650 if (mod_end <= ordered->file_offset)
4654 * We are going to copy all the csums on this ordered extent, so
4655 * go ahead and adjust mod_start and mod_len in case this ordered
4656 * extent has already been logged.
4658 if (ordered->file_offset > mod_start) {
4659 if (ordered_end >= mod_end)
4660 mod_len = ordered->file_offset - mod_start;
4662 * If we have this case
4664 * |--------- logged extent ---------|
4665 * |----- ordered extent ----|
4667 * Just don't mess with mod_start and mod_len, we'll
4668 * just end up logging more csums than we need and it
4672 if (ordered_end < mod_end) {
4673 mod_len = mod_end - ordered_end;
4674 mod_start = ordered_end;
4681 * To keep us from looping for the above case of an ordered
4682 * extent that falls inside of the logged extent.
4684 if (test_and_set_bit(BTRFS_ORDERED_LOGGED_CSUM, &ordered->flags))
4687 list_for_each_entry(sums, &ordered->list, list) {
4688 ret = log_csums(trans, inode, log_root, sums);
4694 /* We're done, found all csums in the ordered extents. */
4698 /* If we're compressed we have to save the entire range of csums. */
4699 if (em->compress_type) {
4701 csum_len = max(em->block_len, em->orig_block_len);
4703 csum_offset = mod_start - em->start;
4707 /* block start is already adjusted for the file extent offset. */
4708 csum_root = btrfs_csum_root(trans->fs_info, em->block_start);
4709 ret = btrfs_lookup_csums_range(csum_root,
4710 em->block_start + csum_offset,
4711 em->block_start + csum_offset +
4712 csum_len - 1, &ordered_sums, 0);
4716 while (!list_empty(&ordered_sums)) {
4717 struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next,
4718 struct btrfs_ordered_sum,
4721 ret = log_csums(trans, inode, log_root, sums);
4722 list_del(&sums->list);
4729 static int log_one_extent(struct btrfs_trans_handle *trans,
4730 struct btrfs_inode *inode,
4731 const struct extent_map *em,
4732 struct btrfs_path *path,
4733 struct btrfs_log_ctx *ctx)
4735 struct btrfs_drop_extents_args drop_args = { 0 };
4736 struct btrfs_root *log = inode->root->log_root;
4737 struct btrfs_file_extent_item fi = { 0 };
4738 struct extent_buffer *leaf;
4739 struct btrfs_key key;
4740 u64 extent_offset = em->start - em->orig_start;
4744 btrfs_set_stack_file_extent_generation(&fi, trans->transid);
4745 if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
4746 btrfs_set_stack_file_extent_type(&fi, BTRFS_FILE_EXTENT_PREALLOC);
4748 btrfs_set_stack_file_extent_type(&fi, BTRFS_FILE_EXTENT_REG);
4750 block_len = max(em->block_len, em->orig_block_len);
4751 if (em->compress_type != BTRFS_COMPRESS_NONE) {
4752 btrfs_set_stack_file_extent_disk_bytenr(&fi, em->block_start);
4753 btrfs_set_stack_file_extent_disk_num_bytes(&fi, block_len);
4754 } else if (em->block_start < EXTENT_MAP_LAST_BYTE) {
4755 btrfs_set_stack_file_extent_disk_bytenr(&fi, em->block_start -
4757 btrfs_set_stack_file_extent_disk_num_bytes(&fi, block_len);
4760 btrfs_set_stack_file_extent_offset(&fi, extent_offset);
4761 btrfs_set_stack_file_extent_num_bytes(&fi, em->len);
4762 btrfs_set_stack_file_extent_ram_bytes(&fi, em->ram_bytes);
4763 btrfs_set_stack_file_extent_compression(&fi, em->compress_type);
4765 ret = log_extent_csums(trans, inode, log, em, ctx);
4770 * If this is the first time we are logging the inode in the current
4771 * transaction, we can avoid btrfs_drop_extents(), which is expensive
4772 * because it does a deletion search, which always acquires write locks
4773 * for extent buffers at levels 2, 1 and 0. This not only wastes time
4774 * but also adds significant contention in a log tree, since log trees
4775 * are small, with a root at level 2 or 3 at most, due to their short
4778 if (ctx->logged_before) {
4779 drop_args.path = path;
4780 drop_args.start = em->start;
4781 drop_args.end = em->start + em->len;
4782 drop_args.replace_extent = true;
4783 drop_args.extent_item_size = sizeof(fi);
4784 ret = btrfs_drop_extents(trans, log, inode, &drop_args);
4789 if (!drop_args.extent_inserted) {
4790 key.objectid = btrfs_ino(inode);
4791 key.type = BTRFS_EXTENT_DATA_KEY;
4792 key.offset = em->start;
4794 ret = btrfs_insert_empty_item(trans, log, path, &key,
4799 leaf = path->nodes[0];
4800 write_extent_buffer(leaf, &fi,
4801 btrfs_item_ptr_offset(leaf, path->slots[0]),
4803 btrfs_mark_buffer_dirty(leaf);
4805 btrfs_release_path(path);
4811 * Log all prealloc extents beyond the inode's i_size to make sure we do not
4812 * lose them after doing a full/fast fsync and replaying the log. We scan the
4813 * subvolume's root instead of iterating the inode's extent map tree because
4814 * otherwise we can log incorrect extent items based on extent map conversion.
4815 * That can happen due to the fact that extent maps are merged when they
4816 * are not in the extent map tree's list of modified extents.
4818 static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans,
4819 struct btrfs_inode *inode,
4820 struct btrfs_path *path)
4822 struct btrfs_root *root = inode->root;
4823 struct btrfs_key key;
4824 const u64 i_size = i_size_read(&inode->vfs_inode);
4825 const u64 ino = btrfs_ino(inode);
4826 struct btrfs_path *dst_path = NULL;
4827 bool dropped_extents = false;
4828 u64 truncate_offset = i_size;
4829 struct extent_buffer *leaf;
4835 if (!(inode->flags & BTRFS_INODE_PREALLOC))
4839 key.type = BTRFS_EXTENT_DATA_KEY;
4840 key.offset = i_size;
4841 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
4846 * We must check if there is a prealloc extent that starts before the
4847 * i_size and crosses the i_size boundary. This is to ensure later we
4848 * truncate down to the end of that extent and not to the i_size, as
4849 * otherwise we end up losing part of the prealloc extent after a log
4850 * replay and with an implicit hole if there is another prealloc extent
4851 * that starts at an offset beyond i_size.
4853 ret = btrfs_previous_item(root, path, ino, BTRFS_EXTENT_DATA_KEY);
4858 struct btrfs_file_extent_item *ei;
4860 leaf = path->nodes[0];
4861 slot = path->slots[0];
4862 ei = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
4864 if (btrfs_file_extent_type(leaf, ei) ==
4865 BTRFS_FILE_EXTENT_PREALLOC) {
4868 btrfs_item_key_to_cpu(leaf, &key, slot);
4869 extent_end = key.offset +
4870 btrfs_file_extent_num_bytes(leaf, ei);
4872 if (extent_end > i_size)
4873 truncate_offset = extent_end;
4880 leaf = path->nodes[0];
4881 slot = path->slots[0];
4883 if (slot >= btrfs_header_nritems(leaf)) {
4885 ret = copy_items(trans, inode, dst_path, path,
4886 start_slot, ins_nr, 1, 0);
4891 ret = btrfs_next_leaf(root, path);
4901 btrfs_item_key_to_cpu(leaf, &key, slot);
4902 if (key.objectid > ino)
4904 if (WARN_ON_ONCE(key.objectid < ino) ||
4905 key.type < BTRFS_EXTENT_DATA_KEY ||
4906 key.offset < i_size) {
4910 if (!dropped_extents) {
4912 * Avoid logging extent items logged in past fsync calls
4913 * and leading to duplicate keys in the log tree.
4915 ret = truncate_inode_items(trans, root->log_root, inode,
4917 BTRFS_EXTENT_DATA_KEY);
4920 dropped_extents = true;
4927 dst_path = btrfs_alloc_path();
4935 ret = copy_items(trans, inode, dst_path, path,
4936 start_slot, ins_nr, 1, 0);
4938 btrfs_release_path(path);
4939 btrfs_free_path(dst_path);
4943 static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
4944 struct btrfs_inode *inode,
4945 struct btrfs_path *path,
4946 struct btrfs_log_ctx *ctx)
4948 struct btrfs_ordered_extent *ordered;
4949 struct btrfs_ordered_extent *tmp;
4950 struct extent_map *em, *n;
4951 struct list_head extents;
4952 struct extent_map_tree *tree = &inode->extent_tree;
4956 INIT_LIST_HEAD(&extents);
4958 write_lock(&tree->lock);
4960 list_for_each_entry_safe(em, n, &tree->modified_extents, list) {
4961 list_del_init(&em->list);
4963 * Just an arbitrary number, this can be really CPU intensive
4964 * once we start getting a lot of extents, and really once we
4965 * have a bunch of extents we just want to commit since it will
4968 if (++num > 32768) {
4969 list_del_init(&tree->modified_extents);
4974 if (em->generation < trans->transid)
4977 /* We log prealloc extents beyond eof later. */
4978 if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) &&
4979 em->start >= i_size_read(&inode->vfs_inode))
4982 /* Need a ref to keep it from getting evicted from cache */
4983 refcount_inc(&em->refs);
4984 set_bit(EXTENT_FLAG_LOGGING, &em->flags);
4985 list_add_tail(&em->list, &extents);
4989 list_sort(NULL, &extents, extent_cmp);
4991 while (!list_empty(&extents)) {
4992 em = list_entry(extents.next, struct extent_map, list);
4994 list_del_init(&em->list);
4997 * If we had an error we just need to delete everybody from our
5001 clear_em_logging(tree, em);
5002 free_extent_map(em);
5006 write_unlock(&tree->lock);
5008 ret = log_one_extent(trans, inode, em, path, ctx);
5009 write_lock(&tree->lock);
5010 clear_em_logging(tree, em);
5011 free_extent_map(em);
5013 WARN_ON(!list_empty(&extents));
5014 write_unlock(&tree->lock);
5017 ret = btrfs_log_prealloc_extents(trans, inode, path);
5022 * We have logged all extents successfully, now make sure the commit of
5023 * the current transaction waits for the ordered extents to complete
5024 * before it commits and wipes out the log trees, otherwise we would
5025 * lose data if an ordered extents completes after the transaction
5026 * commits and a power failure happens after the transaction commit.
5028 list_for_each_entry_safe(ordered, tmp, &ctx->ordered_extents, log_list) {
5029 list_del_init(&ordered->log_list);
5030 set_bit(BTRFS_ORDERED_LOGGED, &ordered->flags);
5032 if (!test_bit(BTRFS_ORDERED_COMPLETE, &ordered->flags)) {
5033 spin_lock_irq(&inode->ordered_tree.lock);
5034 if (!test_bit(BTRFS_ORDERED_COMPLETE, &ordered->flags)) {
5035 set_bit(BTRFS_ORDERED_PENDING, &ordered->flags);
5036 atomic_inc(&trans->transaction->pending_ordered);
5038 spin_unlock_irq(&inode->ordered_tree.lock);
5040 btrfs_put_ordered_extent(ordered);
5046 static int logged_inode_size(struct btrfs_root *log, struct btrfs_inode *inode,
5047 struct btrfs_path *path, u64 *size_ret)
5049 struct btrfs_key key;
5052 key.objectid = btrfs_ino(inode);
5053 key.type = BTRFS_INODE_ITEM_KEY;
5056 ret = btrfs_search_slot(NULL, log, &key, path, 0, 0);
5059 } else if (ret > 0) {
5062 struct btrfs_inode_item *item;
5064 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
5065 struct btrfs_inode_item);
5066 *size_ret = btrfs_inode_size(path->nodes[0], item);
5068 * If the in-memory inode's i_size is smaller then the inode
5069 * size stored in the btree, return the inode's i_size, so
5070 * that we get a correct inode size after replaying the log
5071 * when before a power failure we had a shrinking truncate
5072 * followed by addition of a new name (rename / new hard link).
5073 * Otherwise return the inode size from the btree, to avoid
5074 * data loss when replaying a log due to previously doing a
5075 * write that expands the inode's size and logging a new name
5076 * immediately after.
5078 if (*size_ret > inode->vfs_inode.i_size)
5079 *size_ret = inode->vfs_inode.i_size;
5082 btrfs_release_path(path);
5087 * At the moment we always log all xattrs. This is to figure out at log replay
5088 * time which xattrs must have their deletion replayed. If a xattr is missing
5089 * in the log tree and exists in the fs/subvol tree, we delete it. This is
5090 * because if a xattr is deleted, the inode is fsynced and a power failure
5091 * happens, causing the log to be replayed the next time the fs is mounted,
5092 * we want the xattr to not exist anymore (same behaviour as other filesystems
5093 * with a journal, ext3/4, xfs, f2fs, etc).
5095 static int btrfs_log_all_xattrs(struct btrfs_trans_handle *trans,
5096 struct btrfs_inode *inode,
5097 struct btrfs_path *path,
5098 struct btrfs_path *dst_path)
5100 struct btrfs_root *root = inode->root;
5102 struct btrfs_key key;
5103 const u64 ino = btrfs_ino(inode);
5106 bool found_xattrs = false;
5108 if (test_bit(BTRFS_INODE_NO_XATTRS, &inode->runtime_flags))
5112 key.type = BTRFS_XATTR_ITEM_KEY;
5115 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5120 int slot = path->slots[0];
5121 struct extent_buffer *leaf = path->nodes[0];
5122 int nritems = btrfs_header_nritems(leaf);
5124 if (slot >= nritems) {
5126 ret = copy_items(trans, inode, dst_path, path,
5127 start_slot, ins_nr, 1, 0);
5132 ret = btrfs_next_leaf(root, path);
5140 btrfs_item_key_to_cpu(leaf, &key, slot);
5141 if (key.objectid != ino || key.type != BTRFS_XATTR_ITEM_KEY)
5148 found_xattrs = true;
5152 ret = copy_items(trans, inode, dst_path, path,
5153 start_slot, ins_nr, 1, 0);
5159 set_bit(BTRFS_INODE_NO_XATTRS, &inode->runtime_flags);
5165 * When using the NO_HOLES feature if we punched a hole that causes the
5166 * deletion of entire leafs or all the extent items of the first leaf (the one
5167 * that contains the inode item and references) we may end up not processing
5168 * any extents, because there are no leafs with a generation matching the
5169 * current transaction that have extent items for our inode. So we need to find
5170 * if any holes exist and then log them. We also need to log holes after any
5171 * truncate operation that changes the inode's size.
5173 static int btrfs_log_holes(struct btrfs_trans_handle *trans,
5174 struct btrfs_inode *inode,
5175 struct btrfs_path *path)
5177 struct btrfs_root *root = inode->root;
5178 struct btrfs_fs_info *fs_info = root->fs_info;
5179 struct btrfs_key key;
5180 const u64 ino = btrfs_ino(inode);
5181 const u64 i_size = i_size_read(&inode->vfs_inode);
5182 u64 prev_extent_end = 0;
5185 if (!btrfs_fs_incompat(fs_info, NO_HOLES) || i_size == 0)
5189 key.type = BTRFS_EXTENT_DATA_KEY;
5192 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5197 struct extent_buffer *leaf = path->nodes[0];
5199 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
5200 ret = btrfs_next_leaf(root, path);
5207 leaf = path->nodes[0];
5210 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
5211 if (key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY)
5214 /* We have a hole, log it. */
5215 if (prev_extent_end < key.offset) {
5216 const u64 hole_len = key.offset - prev_extent_end;
5219 * Release the path to avoid deadlocks with other code
5220 * paths that search the root while holding locks on
5221 * leafs from the log root.
5223 btrfs_release_path(path);
5224 ret = btrfs_insert_file_extent(trans, root->log_root,
5225 ino, prev_extent_end, 0,
5226 0, hole_len, 0, hole_len,
5232 * Search for the same key again in the root. Since it's
5233 * an extent item and we are holding the inode lock, the
5234 * key must still exist. If it doesn't just emit warning
5235 * and return an error to fall back to a transaction
5238 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5241 if (WARN_ON(ret > 0))
5243 leaf = path->nodes[0];
5246 prev_extent_end = btrfs_file_extent_end(path);
5251 if (prev_extent_end < i_size) {
5254 btrfs_release_path(path);
5255 hole_len = ALIGN(i_size - prev_extent_end, fs_info->sectorsize);
5256 ret = btrfs_insert_file_extent(trans, root->log_root,
5257 ino, prev_extent_end, 0, 0,
5258 hole_len, 0, hole_len,
5268 * When we are logging a new inode X, check if it doesn't have a reference that
5269 * matches the reference from some other inode Y created in a past transaction
5270 * and that was renamed in the current transaction. If we don't do this, then at
5271 * log replay time we can lose inode Y (and all its files if it's a directory):
5274 * echo "hello world" > /mnt/x/foobar
5277 * mkdir /mnt/x # or touch /mnt/x
5278 * xfs_io -c fsync /mnt/x
5280 * mount fs, trigger log replay
5282 * After the log replay procedure, we would lose the first directory and all its
5283 * files (file foobar).
5284 * For the case where inode Y is not a directory we simply end up losing it:
5286 * echo "123" > /mnt/foo
5288 * mv /mnt/foo /mnt/bar
5289 * echo "abc" > /mnt/foo
5290 * xfs_io -c fsync /mnt/foo
5293 * We also need this for cases where a snapshot entry is replaced by some other
5294 * entry (file or directory) otherwise we end up with an unreplayable log due to
5295 * attempts to delete the snapshot entry (entry of type BTRFS_ROOT_ITEM_KEY) as
5296 * if it were a regular entry:
5299 * btrfs subvolume snapshot /mnt /mnt/x/snap
5300 * btrfs subvolume delete /mnt/x/snap
5303 * fsync /mnt/x or fsync some new file inside it
5306 * The snapshot delete, rmdir of x, mkdir of a new x and the fsync all happen in
5307 * the same transaction.
5309 static int btrfs_check_ref_name_override(struct extent_buffer *eb,
5311 const struct btrfs_key *key,
5312 struct btrfs_inode *inode,
5313 u64 *other_ino, u64 *other_parent)
5316 struct btrfs_path *search_path;
5319 u32 item_size = btrfs_item_size(eb, slot);
5321 unsigned long ptr = btrfs_item_ptr_offset(eb, slot);
5323 search_path = btrfs_alloc_path();
5326 search_path->search_commit_root = 1;
5327 search_path->skip_locking = 1;
5329 while (cur_offset < item_size) {
5333 unsigned long name_ptr;
5334 struct btrfs_dir_item *di;
5336 if (key->type == BTRFS_INODE_REF_KEY) {
5337 struct btrfs_inode_ref *iref;
5339 iref = (struct btrfs_inode_ref *)(ptr + cur_offset);
5340 parent = key->offset;
5341 this_name_len = btrfs_inode_ref_name_len(eb, iref);
5342 name_ptr = (unsigned long)(iref + 1);
5343 this_len = sizeof(*iref) + this_name_len;
5345 struct btrfs_inode_extref *extref;
5347 extref = (struct btrfs_inode_extref *)(ptr +
5349 parent = btrfs_inode_extref_parent(eb, extref);
5350 this_name_len = btrfs_inode_extref_name_len(eb, extref);
5351 name_ptr = (unsigned long)&extref->name;
5352 this_len = sizeof(*extref) + this_name_len;
5355 if (this_name_len > name_len) {
5358 new_name = krealloc(name, this_name_len, GFP_NOFS);
5363 name_len = this_name_len;
5367 read_extent_buffer(eb, name, name_ptr, this_name_len);
5368 di = btrfs_lookup_dir_item(NULL, inode->root, search_path,
5369 parent, name, this_name_len, 0);
5370 if (di && !IS_ERR(di)) {
5371 struct btrfs_key di_key;
5373 btrfs_dir_item_key_to_cpu(search_path->nodes[0],
5375 if (di_key.type == BTRFS_INODE_ITEM_KEY) {
5376 if (di_key.objectid != key->objectid) {
5378 *other_ino = di_key.objectid;
5379 *other_parent = parent;
5387 } else if (IS_ERR(di)) {
5391 btrfs_release_path(search_path);
5393 cur_offset += this_len;
5397 btrfs_free_path(search_path);
5402 struct btrfs_ino_list {
5405 struct list_head list;
5408 static int log_conflicting_inodes(struct btrfs_trans_handle *trans,
5409 struct btrfs_root *root,
5410 struct btrfs_path *path,
5411 struct btrfs_log_ctx *ctx,
5412 u64 ino, u64 parent)
5414 struct btrfs_ino_list *ino_elem;
5415 LIST_HEAD(inode_list);
5418 ino_elem = kmalloc(sizeof(*ino_elem), GFP_NOFS);
5421 ino_elem->ino = ino;
5422 ino_elem->parent = parent;
5423 list_add_tail(&ino_elem->list, &inode_list);
5425 while (!list_empty(&inode_list)) {
5426 struct btrfs_fs_info *fs_info = root->fs_info;
5427 struct btrfs_key key;
5428 struct inode *inode;
5430 ino_elem = list_first_entry(&inode_list, struct btrfs_ino_list,
5432 ino = ino_elem->ino;
5433 parent = ino_elem->parent;
5434 list_del(&ino_elem->list);
5439 btrfs_release_path(path);
5441 inode = btrfs_iget(fs_info->sb, ino, root);
5443 * If the other inode that had a conflicting dir entry was
5444 * deleted in the current transaction, we need to log its parent
5447 if (IS_ERR(inode)) {
5448 ret = PTR_ERR(inode);
5449 if (ret == -ENOENT) {
5450 inode = btrfs_iget(fs_info->sb, parent, root);
5451 if (IS_ERR(inode)) {
5452 ret = PTR_ERR(inode);
5454 ret = btrfs_log_inode(trans,
5456 LOG_OTHER_INODE_ALL,
5458 btrfs_add_delayed_iput(inode);
5464 * If the inode was already logged skip it - otherwise we can
5465 * hit an infinite loop. Example:
5467 * From the commit root (previous transaction) we have the
5470 * inode 257 a directory
5471 * inode 258 with references "zz" and "zz_link" on inode 257
5472 * inode 259 with reference "a" on inode 257
5474 * And in the current (uncommitted) transaction we have:
5476 * inode 257 a directory, unchanged
5477 * inode 258 with references "a" and "a2" on inode 257
5478 * inode 259 with reference "zz_link" on inode 257
5479 * inode 261 with reference "zz" on inode 257
5481 * When logging inode 261 the following infinite loop could
5482 * happen if we don't skip already logged inodes:
5484 * - we detect inode 258 as a conflicting inode, with inode 261
5485 * on reference "zz", and log it;
5487 * - we detect inode 259 as a conflicting inode, with inode 258
5488 * on reference "a", and log it;
5490 * - we detect inode 258 as a conflicting inode, with inode 259
5491 * on reference "zz_link", and log it - again! After this we
5492 * repeat the above steps forever.
5494 spin_lock(&BTRFS_I(inode)->lock);
5496 * Check the inode's logged_trans only instead of
5497 * btrfs_inode_in_log(). This is because the last_log_commit of
5498 * the inode is not updated when we only log that it exists (see
5499 * btrfs_log_inode()).
5501 if (BTRFS_I(inode)->logged_trans == trans->transid) {
5502 spin_unlock(&BTRFS_I(inode)->lock);
5503 btrfs_add_delayed_iput(inode);
5506 spin_unlock(&BTRFS_I(inode)->lock);
5508 * We are safe logging the other inode without acquiring its
5509 * lock as long as we log with the LOG_INODE_EXISTS mode. We
5510 * are safe against concurrent renames of the other inode as
5511 * well because during a rename we pin the log and update the
5512 * log with the new name before we unpin it.
5514 ret = btrfs_log_inode(trans, BTRFS_I(inode), LOG_OTHER_INODE, ctx);
5516 btrfs_add_delayed_iput(inode);
5521 key.type = BTRFS_INODE_REF_KEY;
5523 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5525 btrfs_add_delayed_iput(inode);
5530 struct extent_buffer *leaf = path->nodes[0];
5531 int slot = path->slots[0];
5533 u64 other_parent = 0;
5535 if (slot >= btrfs_header_nritems(leaf)) {
5536 ret = btrfs_next_leaf(root, path);
5539 } else if (ret > 0) {
5546 btrfs_item_key_to_cpu(leaf, &key, slot);
5547 if (key.objectid != ino ||
5548 (key.type != BTRFS_INODE_REF_KEY &&
5549 key.type != BTRFS_INODE_EXTREF_KEY)) {
5554 ret = btrfs_check_ref_name_override(leaf, slot, &key,
5555 BTRFS_I(inode), &other_ino,
5560 ino_elem = kmalloc(sizeof(*ino_elem), GFP_NOFS);
5565 ino_elem->ino = other_ino;
5566 ino_elem->parent = other_parent;
5567 list_add_tail(&ino_elem->list, &inode_list);
5572 btrfs_add_delayed_iput(inode);
5578 static int copy_inode_items_to_log(struct btrfs_trans_handle *trans,
5579 struct btrfs_inode *inode,
5580 struct btrfs_key *min_key,
5581 const struct btrfs_key *max_key,
5582 struct btrfs_path *path,
5583 struct btrfs_path *dst_path,
5584 const u64 logged_isize,
5585 const bool recursive_logging,
5586 const int inode_only,
5587 struct btrfs_log_ctx *ctx,
5588 bool *need_log_inode_item)
5590 const u64 i_size = i_size_read(&inode->vfs_inode);
5591 struct btrfs_root *root = inode->root;
5592 int ins_start_slot = 0;
5597 ret = btrfs_search_forward(root, min_key, path, trans->transid);
5605 /* Note, ins_nr might be > 0 here, cleanup outside the loop */
5606 if (min_key->objectid != max_key->objectid)
5608 if (min_key->type > max_key->type)
5611 if (min_key->type == BTRFS_INODE_ITEM_KEY) {
5612 *need_log_inode_item = false;
5613 } else if (min_key->type == BTRFS_EXTENT_DATA_KEY &&
5614 min_key->offset >= i_size) {
5616 * Extents at and beyond eof are logged with
5617 * btrfs_log_prealloc_extents().
5618 * Only regular files have BTRFS_EXTENT_DATA_KEY keys,
5619 * and no keys greater than that, so bail out.
5622 } else if ((min_key->type == BTRFS_INODE_REF_KEY ||
5623 min_key->type == BTRFS_INODE_EXTREF_KEY) &&
5624 inode->generation == trans->transid &&
5625 !recursive_logging) {
5627 u64 other_parent = 0;
5629 ret = btrfs_check_ref_name_override(path->nodes[0],
5630 path->slots[0], min_key, inode,
5631 &other_ino, &other_parent);
5634 } else if (ret > 0 &&
5635 other_ino != btrfs_ino(BTRFS_I(ctx->inode))) {
5640 ins_start_slot = path->slots[0];
5642 ret = copy_items(trans, inode, dst_path, path,
5643 ins_start_slot, ins_nr,
5644 inode_only, logged_isize);
5649 ret = log_conflicting_inodes(trans, root, path,
5650 ctx, other_ino, other_parent);
5653 btrfs_release_path(path);
5656 } else if (min_key->type == BTRFS_XATTR_ITEM_KEY) {
5657 /* Skip xattrs, logged later with btrfs_log_all_xattrs() */
5660 ret = copy_items(trans, inode, dst_path, path,
5662 ins_nr, inode_only, logged_isize);
5669 if (ins_nr && ins_start_slot + ins_nr == path->slots[0]) {
5672 } else if (!ins_nr) {
5673 ins_start_slot = path->slots[0];
5678 ret = copy_items(trans, inode, dst_path, path, ins_start_slot,
5679 ins_nr, inode_only, logged_isize);
5683 ins_start_slot = path->slots[0];
5686 if (path->slots[0] < btrfs_header_nritems(path->nodes[0])) {
5687 btrfs_item_key_to_cpu(path->nodes[0], min_key,
5692 ret = copy_items(trans, inode, dst_path, path,
5693 ins_start_slot, ins_nr, inode_only,
5699 btrfs_release_path(path);
5701 if (min_key->offset < (u64)-1) {
5703 } else if (min_key->type < max_key->type) {
5705 min_key->offset = 0;
5711 * We may process many leaves full of items for our inode, so
5712 * avoid monopolizing a cpu for too long by rescheduling while
5713 * not holding locks on any tree.
5718 ret = copy_items(trans, inode, dst_path, path, ins_start_slot,
5719 ins_nr, inode_only, logged_isize);
5724 if (inode_only == LOG_INODE_ALL && S_ISREG(inode->vfs_inode.i_mode)) {
5726 * Release the path because otherwise we might attempt to double
5727 * lock the same leaf with btrfs_log_prealloc_extents() below.
5729 btrfs_release_path(path);
5730 ret = btrfs_log_prealloc_extents(trans, inode, dst_path);
5736 /* log a single inode in the tree log.
5737 * At least one parent directory for this inode must exist in the tree
5738 * or be logged already.
5740 * Any items from this inode changed by the current transaction are copied
5741 * to the log tree. An extra reference is taken on any extents in this
5742 * file, allowing us to avoid a whole pile of corner cases around logging
5743 * blocks that have been removed from the tree.
5745 * See LOG_INODE_ALL and related defines for a description of what inode_only
5748 * This handles both files and directories.
5750 static int btrfs_log_inode(struct btrfs_trans_handle *trans,
5751 struct btrfs_inode *inode,
5753 struct btrfs_log_ctx *ctx)
5755 struct btrfs_path *path;
5756 struct btrfs_path *dst_path;
5757 struct btrfs_key min_key;
5758 struct btrfs_key max_key;
5759 struct btrfs_root *log = inode->root->log_root;
5761 bool fast_search = false;
5762 u64 ino = btrfs_ino(inode);
5763 struct extent_map_tree *em_tree = &inode->extent_tree;
5764 u64 logged_isize = 0;
5765 bool need_log_inode_item = true;
5766 bool xattrs_logged = false;
5767 bool recursive_logging = false;
5768 bool inode_item_dropped = true;
5769 const bool orig_logged_before = ctx->logged_before;
5771 path = btrfs_alloc_path();
5774 dst_path = btrfs_alloc_path();
5776 btrfs_free_path(path);
5780 min_key.objectid = ino;
5781 min_key.type = BTRFS_INODE_ITEM_KEY;
5784 max_key.objectid = ino;
5787 /* today the code can only do partial logging of directories */
5788 if (S_ISDIR(inode->vfs_inode.i_mode) ||
5789 (!test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
5790 &inode->runtime_flags) &&
5791 inode_only >= LOG_INODE_EXISTS))
5792 max_key.type = BTRFS_XATTR_ITEM_KEY;
5794 max_key.type = (u8)-1;
5795 max_key.offset = (u64)-1;
5798 * Only run delayed items if we are a directory. We want to make sure
5799 * all directory indexes hit the fs/subvolume tree so we can find them
5800 * and figure out which index ranges have to be logged.
5802 if (S_ISDIR(inode->vfs_inode.i_mode)) {
5803 ret = btrfs_commit_inode_delayed_items(trans, inode);
5808 if (inode_only == LOG_OTHER_INODE || inode_only == LOG_OTHER_INODE_ALL) {
5809 recursive_logging = true;
5810 if (inode_only == LOG_OTHER_INODE)
5811 inode_only = LOG_INODE_EXISTS;
5813 inode_only = LOG_INODE_ALL;
5814 mutex_lock_nested(&inode->log_mutex, SINGLE_DEPTH_NESTING);
5816 mutex_lock(&inode->log_mutex);
5820 * For symlinks, we must always log their content, which is stored in an
5821 * inline extent, otherwise we could end up with an empty symlink after
5822 * log replay, which is invalid on linux (symlink(2) returns -ENOENT if
5823 * one attempts to create an empty symlink).
5824 * We don't need to worry about flushing delalloc, because when we create
5825 * the inline extent when the symlink is created (we never have delalloc
5828 if (S_ISLNK(inode->vfs_inode.i_mode))
5829 inode_only = LOG_INODE_ALL;
5832 * Before logging the inode item, cache the value returned by
5833 * inode_logged(), because after that we have the need to figure out if
5834 * the inode was previously logged in this transaction.
5836 ret = inode_logged(trans, inode, path);
5839 ctx->logged_before = (ret == 1);
5843 * This is for cases where logging a directory could result in losing a
5844 * a file after replaying the log. For example, if we move a file from a
5845 * directory A to a directory B, then fsync directory A, we have no way
5846 * to known the file was moved from A to B, so logging just A would
5847 * result in losing the file after a log replay.
5849 if (S_ISDIR(inode->vfs_inode.i_mode) &&
5850 inode_only == LOG_INODE_ALL &&
5851 inode->last_unlink_trans >= trans->transid) {
5852 btrfs_set_log_full_commit(trans);
5853 ret = BTRFS_LOG_FORCE_COMMIT;
5858 * a brute force approach to making sure we get the most uptodate
5859 * copies of everything.
5861 if (S_ISDIR(inode->vfs_inode.i_mode)) {
5862 int max_key_type = BTRFS_DIR_LOG_INDEX_KEY;
5864 clear_bit(BTRFS_INODE_COPY_EVERYTHING, &inode->runtime_flags);
5865 if (inode_only == LOG_INODE_EXISTS)
5866 max_key_type = BTRFS_XATTR_ITEM_KEY;
5867 if (ctx->logged_before)
5868 ret = drop_inode_items(trans, log, path, inode,
5871 if (inode_only == LOG_INODE_EXISTS && ctx->logged_before) {
5873 * Make sure the new inode item we write to the log has
5874 * the same isize as the current one (if it exists).
5875 * This is necessary to prevent data loss after log
5876 * replay, and also to prevent doing a wrong expanding
5877 * truncate - for e.g. create file, write 4K into offset
5878 * 0, fsync, write 4K into offset 4096, add hard link,
5879 * fsync some other file (to sync log), power fail - if
5880 * we use the inode's current i_size, after log replay
5881 * we get a 8Kb file, with the last 4Kb extent as a hole
5882 * (zeroes), as if an expanding truncate happened,
5883 * instead of getting a file of 4Kb only.
5885 ret = logged_inode_size(log, inode, path, &logged_isize);
5889 if (test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
5890 &inode->runtime_flags)) {
5891 if (inode_only == LOG_INODE_EXISTS) {
5892 max_key.type = BTRFS_XATTR_ITEM_KEY;
5893 if (ctx->logged_before)
5894 ret = drop_inode_items(trans, log, path,
5895 inode, max_key.type);
5897 clear_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
5898 &inode->runtime_flags);
5899 clear_bit(BTRFS_INODE_COPY_EVERYTHING,
5900 &inode->runtime_flags);
5901 if (ctx->logged_before)
5902 ret = truncate_inode_items(trans, log,
5905 } else if (test_and_clear_bit(BTRFS_INODE_COPY_EVERYTHING,
5906 &inode->runtime_flags) ||
5907 inode_only == LOG_INODE_EXISTS) {
5908 if (inode_only == LOG_INODE_ALL)
5910 max_key.type = BTRFS_XATTR_ITEM_KEY;
5911 if (ctx->logged_before)
5912 ret = drop_inode_items(trans, log, path, inode,
5915 if (inode_only == LOG_INODE_ALL)
5917 inode_item_dropped = false;
5925 ret = copy_inode_items_to_log(trans, inode, &min_key, &max_key,
5926 path, dst_path, logged_isize,
5927 recursive_logging, inode_only, ctx,
5928 &need_log_inode_item);
5932 btrfs_release_path(path);
5933 btrfs_release_path(dst_path);
5934 ret = btrfs_log_all_xattrs(trans, inode, path, dst_path);
5937 xattrs_logged = true;
5938 if (max_key.type >= BTRFS_EXTENT_DATA_KEY && !fast_search) {
5939 btrfs_release_path(path);
5940 btrfs_release_path(dst_path);
5941 ret = btrfs_log_holes(trans, inode, path);
5946 btrfs_release_path(path);
5947 btrfs_release_path(dst_path);
5948 if (need_log_inode_item) {
5949 ret = log_inode_item(trans, log, dst_path, inode, inode_item_dropped);
5953 * If we are doing a fast fsync and the inode was logged before
5954 * in this transaction, we don't need to log the xattrs because
5955 * they were logged before. If xattrs were added, changed or
5956 * deleted since the last time we logged the inode, then we have
5957 * already logged them because the inode had the runtime flag
5958 * BTRFS_INODE_COPY_EVERYTHING set.
5960 if (!xattrs_logged && inode->logged_trans < trans->transid) {
5961 ret = btrfs_log_all_xattrs(trans, inode, path, dst_path);
5964 btrfs_release_path(path);
5968 ret = btrfs_log_changed_extents(trans, inode, dst_path, ctx);
5971 } else if (inode_only == LOG_INODE_ALL) {
5972 struct extent_map *em, *n;
5974 write_lock(&em_tree->lock);
5975 list_for_each_entry_safe(em, n, &em_tree->modified_extents, list)
5976 list_del_init(&em->list);
5977 write_unlock(&em_tree->lock);
5980 if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->vfs_inode.i_mode)) {
5981 ret = log_directory_changes(trans, inode, path, dst_path, ctx);
5986 spin_lock(&inode->lock);
5987 inode->logged_trans = trans->transid;
5989 * Don't update last_log_commit if we logged that an inode exists.
5990 * We do this for three reasons:
5992 * 1) We might have had buffered writes to this inode that were
5993 * flushed and had their ordered extents completed in this
5994 * transaction, but we did not previously log the inode with
5995 * LOG_INODE_ALL. Later the inode was evicted and after that
5996 * it was loaded again and this LOG_INODE_EXISTS log operation
5997 * happened. We must make sure that if an explicit fsync against
5998 * the inode is performed later, it logs the new extents, an
5999 * updated inode item, etc, and syncs the log. The same logic
6000 * applies to direct IO writes instead of buffered writes.
6002 * 2) When we log the inode with LOG_INODE_EXISTS, its inode item
6003 * is logged with an i_size of 0 or whatever value was logged
6004 * before. If later the i_size of the inode is increased by a
6005 * truncate operation, the log is synced through an fsync of
6006 * some other inode and then finally an explicit fsync against
6007 * this inode is made, we must make sure this fsync logs the
6008 * inode with the new i_size, the hole between old i_size and
6009 * the new i_size, and syncs the log.
6011 * 3) If we are logging that an ancestor inode exists as part of
6012 * logging a new name from a link or rename operation, don't update
6013 * its last_log_commit - otherwise if an explicit fsync is made
6014 * against an ancestor, the fsync considers the inode in the log
6015 * and doesn't sync the log, resulting in the ancestor missing after
6016 * a power failure unless the log was synced as part of an fsync
6017 * against any other unrelated inode.
6019 if (inode_only != LOG_INODE_EXISTS)
6020 inode->last_log_commit = inode->last_sub_trans;
6021 spin_unlock(&inode->lock);
6024 * Reset the last_reflink_trans so that the next fsync does not need to
6025 * go through the slower path when logging extents and their checksums.
6027 if (inode_only == LOG_INODE_ALL)
6028 inode->last_reflink_trans = 0;
6031 mutex_unlock(&inode->log_mutex);
6033 btrfs_free_path(path);
6034 btrfs_free_path(dst_path);
6036 if (recursive_logging)
6037 ctx->logged_before = orig_logged_before;
6043 * Check if we need to log an inode. This is used in contexts where while
6044 * logging an inode we need to log another inode (either that it exists or in
6045 * full mode). This is used instead of btrfs_inode_in_log() because the later
6046 * requires the inode to be in the log and have the log transaction committed,
6047 * while here we do not care if the log transaction was already committed - our
6048 * caller will commit the log later - and we want to avoid logging an inode
6049 * multiple times when multiple tasks have joined the same log transaction.
6051 static bool need_log_inode(struct btrfs_trans_handle *trans,
6052 struct btrfs_inode *inode)
6055 * If a directory was not modified, no dentries added or removed, we can
6056 * and should avoid logging it.
6058 if (S_ISDIR(inode->vfs_inode.i_mode) && inode->last_trans < trans->transid)
6062 * If this inode does not have new/updated/deleted xattrs since the last
6063 * time it was logged and is flagged as logged in the current transaction,
6064 * we can skip logging it. As for new/deleted names, those are updated in
6065 * the log by link/unlink/rename operations.
6066 * In case the inode was logged and then evicted and reloaded, its
6067 * logged_trans will be 0, in which case we have to fully log it since
6068 * logged_trans is a transient field, not persisted.
6070 if (inode->logged_trans == trans->transid &&
6071 !test_bit(BTRFS_INODE_COPY_EVERYTHING, &inode->runtime_flags))
6077 struct btrfs_dir_list {
6079 struct list_head list;
6083 * Log the inodes of the new dentries of a directory. See log_dir_items() for
6084 * details about the why it is needed.
6085 * This is a recursive operation - if an existing dentry corresponds to a
6086 * directory, that directory's new entries are logged too (same behaviour as
6087 * ext3/4, xfs, f2fs, reiserfs, nilfs2). Note that when logging the inodes
6088 * the dentries point to we do not lock their i_mutex, otherwise lockdep
6089 * complains about the following circular lock dependency / possible deadlock:
6093 * lock(&type->i_mutex_dir_key#3/2);
6094 * lock(sb_internal#2);
6095 * lock(&type->i_mutex_dir_key#3/2);
6096 * lock(&sb->s_type->i_mutex_key#14);
6098 * Where sb_internal is the lock (a counter that works as a lock) acquired by
6099 * sb_start_intwrite() in btrfs_start_transaction().
6100 * Not locking i_mutex of the inodes is still safe because:
6102 * 1) For regular files we log with a mode of LOG_INODE_EXISTS. It's possible
6103 * that while logging the inode new references (names) are added or removed
6104 * from the inode, leaving the logged inode item with a link count that does
6105 * not match the number of logged inode reference items. This is fine because
6106 * at log replay time we compute the real number of links and correct the
6107 * link count in the inode item (see replay_one_buffer() and
6108 * link_to_fixup_dir());
6110 * 2) For directories we log with a mode of LOG_INODE_ALL. It's possible that
6111 * while logging the inode's items new index items (key type
6112 * BTRFS_DIR_INDEX_KEY) are added to fs/subvol tree and the logged inode item
6113 * has a size that doesn't match the sum of the lengths of all the logged
6114 * names - this is ok, not a problem, because at log replay time we set the
6115 * directory's i_size to the correct value (see replay_one_name() and
6116 * do_overwrite_item()).
6118 static int log_new_dir_dentries(struct btrfs_trans_handle *trans,
6119 struct btrfs_root *root,
6120 struct btrfs_inode *start_inode,
6121 struct btrfs_log_ctx *ctx)
6123 struct btrfs_fs_info *fs_info = root->fs_info;
6124 struct btrfs_path *path;
6125 LIST_HEAD(dir_list);
6126 struct btrfs_dir_list *dir_elem;
6130 * If we are logging a new name, as part of a link or rename operation,
6131 * don't bother logging new dentries, as we just want to log the names
6132 * of an inode and that any new parents exist.
6134 if (ctx->logging_new_name)
6137 path = btrfs_alloc_path();
6141 dir_elem = kmalloc(sizeof(*dir_elem), GFP_NOFS);
6143 btrfs_free_path(path);
6146 dir_elem->ino = btrfs_ino(start_inode);
6147 list_add_tail(&dir_elem->list, &dir_list);
6149 while (!list_empty(&dir_list)) {
6150 struct extent_buffer *leaf;
6151 struct btrfs_key min_key;
6155 dir_elem = list_first_entry(&dir_list, struct btrfs_dir_list,
6158 goto next_dir_inode;
6160 min_key.objectid = dir_elem->ino;
6161 min_key.type = BTRFS_DIR_INDEX_KEY;
6164 btrfs_release_path(path);
6165 ret = btrfs_search_forward(root, &min_key, path, trans->transid);
6167 goto next_dir_inode;
6168 } else if (ret > 0) {
6170 goto next_dir_inode;
6173 leaf = path->nodes[0];
6174 nritems = btrfs_header_nritems(leaf);
6175 for (i = path->slots[0]; i < nritems; i++) {
6176 struct btrfs_dir_item *di;
6177 struct btrfs_key di_key;
6178 struct inode *di_inode;
6179 struct btrfs_dir_list *new_dir_elem;
6180 int log_mode = LOG_INODE_EXISTS;
6183 btrfs_item_key_to_cpu(leaf, &min_key, i);
6184 if (min_key.objectid != dir_elem->ino ||
6185 min_key.type != BTRFS_DIR_INDEX_KEY)
6186 goto next_dir_inode;
6188 di = btrfs_item_ptr(leaf, i, struct btrfs_dir_item);
6189 type = btrfs_dir_type(leaf, di);
6190 if (btrfs_dir_transid(leaf, di) < trans->transid)
6192 btrfs_dir_item_key_to_cpu(leaf, di, &di_key);
6193 if (di_key.type == BTRFS_ROOT_ITEM_KEY)
6196 btrfs_release_path(path);
6197 di_inode = btrfs_iget(fs_info->sb, di_key.objectid, root);
6198 if (IS_ERR(di_inode)) {
6199 ret = PTR_ERR(di_inode);
6200 goto next_dir_inode;
6203 if (!need_log_inode(trans, BTRFS_I(di_inode))) {
6204 btrfs_add_delayed_iput(di_inode);
6208 ctx->log_new_dentries = false;
6209 if (type == BTRFS_FT_DIR)
6210 log_mode = LOG_INODE_ALL;
6211 ret = btrfs_log_inode(trans, BTRFS_I(di_inode),
6213 btrfs_add_delayed_iput(di_inode);
6215 goto next_dir_inode;
6216 if (ctx->log_new_dentries) {
6217 new_dir_elem = kmalloc(sizeof(*new_dir_elem),
6219 if (!new_dir_elem) {
6221 goto next_dir_inode;
6223 new_dir_elem->ino = di_key.objectid;
6224 list_add_tail(&new_dir_elem->list, &dir_list);
6228 if (min_key.offset < (u64)-1) {
6233 list_del(&dir_elem->list);
6237 btrfs_free_path(path);
6241 static int btrfs_log_all_parents(struct btrfs_trans_handle *trans,
6242 struct btrfs_inode *inode,
6243 struct btrfs_log_ctx *ctx)
6245 struct btrfs_fs_info *fs_info = trans->fs_info;
6247 struct btrfs_path *path;
6248 struct btrfs_key key;
6249 struct btrfs_root *root = inode->root;
6250 const u64 ino = btrfs_ino(inode);
6252 path = btrfs_alloc_path();
6255 path->skip_locking = 1;
6256 path->search_commit_root = 1;
6259 key.type = BTRFS_INODE_REF_KEY;
6261 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
6266 struct extent_buffer *leaf = path->nodes[0];
6267 int slot = path->slots[0];
6272 if (slot >= btrfs_header_nritems(leaf)) {
6273 ret = btrfs_next_leaf(root, path);
6281 btrfs_item_key_to_cpu(leaf, &key, slot);
6282 /* BTRFS_INODE_EXTREF_KEY is BTRFS_INODE_REF_KEY + 1 */
6283 if (key.objectid != ino || key.type > BTRFS_INODE_EXTREF_KEY)
6286 item_size = btrfs_item_size(leaf, slot);
6287 ptr = btrfs_item_ptr_offset(leaf, slot);
6288 while (cur_offset < item_size) {
6289 struct btrfs_key inode_key;
6290 struct inode *dir_inode;
6292 inode_key.type = BTRFS_INODE_ITEM_KEY;
6293 inode_key.offset = 0;
6295 if (key.type == BTRFS_INODE_EXTREF_KEY) {
6296 struct btrfs_inode_extref *extref;
6298 extref = (struct btrfs_inode_extref *)
6300 inode_key.objectid = btrfs_inode_extref_parent(
6302 cur_offset += sizeof(*extref);
6303 cur_offset += btrfs_inode_extref_name_len(leaf,
6306 inode_key.objectid = key.offset;
6307 cur_offset = item_size;
6310 dir_inode = btrfs_iget(fs_info->sb, inode_key.objectid,
6313 * If the parent inode was deleted, return an error to
6314 * fallback to a transaction commit. This is to prevent
6315 * getting an inode that was moved from one parent A to
6316 * a parent B, got its former parent A deleted and then
6317 * it got fsync'ed, from existing at both parents after
6318 * a log replay (and the old parent still existing).
6325 * mv /mnt/B/bar /mnt/A/bar
6326 * mv -T /mnt/A /mnt/B
6330 * If we ignore the old parent B which got deleted,
6331 * after a log replay we would have file bar linked
6332 * at both parents and the old parent B would still
6335 if (IS_ERR(dir_inode)) {
6336 ret = PTR_ERR(dir_inode);
6340 if (!need_log_inode(trans, BTRFS_I(dir_inode))) {
6341 btrfs_add_delayed_iput(dir_inode);
6345 ctx->log_new_dentries = false;
6346 ret = btrfs_log_inode(trans, BTRFS_I(dir_inode),
6347 LOG_INODE_ALL, ctx);
6348 if (!ret && ctx->log_new_dentries)
6349 ret = log_new_dir_dentries(trans, root,
6350 BTRFS_I(dir_inode), ctx);
6351 btrfs_add_delayed_iput(dir_inode);
6359 btrfs_free_path(path);
6363 static int log_new_ancestors(struct btrfs_trans_handle *trans,
6364 struct btrfs_root *root,
6365 struct btrfs_path *path,
6366 struct btrfs_log_ctx *ctx)
6368 struct btrfs_key found_key;
6370 btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
6373 struct btrfs_fs_info *fs_info = root->fs_info;
6374 struct extent_buffer *leaf = path->nodes[0];
6375 int slot = path->slots[0];
6376 struct btrfs_key search_key;
6377 struct inode *inode;
6381 btrfs_release_path(path);
6383 ino = found_key.offset;
6385 search_key.objectid = found_key.offset;
6386 search_key.type = BTRFS_INODE_ITEM_KEY;
6387 search_key.offset = 0;
6388 inode = btrfs_iget(fs_info->sb, ino, root);
6390 return PTR_ERR(inode);
6392 if (BTRFS_I(inode)->generation >= trans->transid &&
6393 need_log_inode(trans, BTRFS_I(inode)))
6394 ret = btrfs_log_inode(trans, BTRFS_I(inode),
6395 LOG_INODE_EXISTS, ctx);
6396 btrfs_add_delayed_iput(inode);
6400 if (search_key.objectid == BTRFS_FIRST_FREE_OBJECTID)
6403 search_key.type = BTRFS_INODE_REF_KEY;
6404 ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
6408 leaf = path->nodes[0];
6409 slot = path->slots[0];
6410 if (slot >= btrfs_header_nritems(leaf)) {
6411 ret = btrfs_next_leaf(root, path);
6416 leaf = path->nodes[0];
6417 slot = path->slots[0];
6420 btrfs_item_key_to_cpu(leaf, &found_key, slot);
6421 if (found_key.objectid != search_key.objectid ||
6422 found_key.type != BTRFS_INODE_REF_KEY)
6428 static int log_new_ancestors_fast(struct btrfs_trans_handle *trans,
6429 struct btrfs_inode *inode,
6430 struct dentry *parent,
6431 struct btrfs_log_ctx *ctx)
6433 struct btrfs_root *root = inode->root;
6434 struct dentry *old_parent = NULL;
6435 struct super_block *sb = inode->vfs_inode.i_sb;
6439 if (!parent || d_really_is_negative(parent) ||
6443 inode = BTRFS_I(d_inode(parent));
6444 if (root != inode->root)
6447 if (inode->generation >= trans->transid &&
6448 need_log_inode(trans, inode)) {
6449 ret = btrfs_log_inode(trans, inode,
6450 LOG_INODE_EXISTS, ctx);
6454 if (IS_ROOT(parent))
6457 parent = dget_parent(parent);
6459 old_parent = parent;
6466 static int log_all_new_ancestors(struct btrfs_trans_handle *trans,
6467 struct btrfs_inode *inode,
6468 struct dentry *parent,
6469 struct btrfs_log_ctx *ctx)
6471 struct btrfs_root *root = inode->root;
6472 const u64 ino = btrfs_ino(inode);
6473 struct btrfs_path *path;
6474 struct btrfs_key search_key;
6478 * For a single hard link case, go through a fast path that does not
6479 * need to iterate the fs/subvolume tree.
6481 if (inode->vfs_inode.i_nlink < 2)
6482 return log_new_ancestors_fast(trans, inode, parent, ctx);
6484 path = btrfs_alloc_path();
6488 search_key.objectid = ino;
6489 search_key.type = BTRFS_INODE_REF_KEY;
6490 search_key.offset = 0;
6492 ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
6499 struct extent_buffer *leaf = path->nodes[0];
6500 int slot = path->slots[0];
6501 struct btrfs_key found_key;
6503 if (slot >= btrfs_header_nritems(leaf)) {
6504 ret = btrfs_next_leaf(root, path);
6512 btrfs_item_key_to_cpu(leaf, &found_key, slot);
6513 if (found_key.objectid != ino ||
6514 found_key.type > BTRFS_INODE_EXTREF_KEY)
6518 * Don't deal with extended references because they are rare
6519 * cases and too complex to deal with (we would need to keep
6520 * track of which subitem we are processing for each item in
6521 * this loop, etc). So just return some error to fallback to
6522 * a transaction commit.
6524 if (found_key.type == BTRFS_INODE_EXTREF_KEY) {
6530 * Logging ancestors needs to do more searches on the fs/subvol
6531 * tree, so it releases the path as needed to avoid deadlocks.
6532 * Keep track of the last inode ref key and resume from that key
6533 * after logging all new ancestors for the current hard link.
6535 memcpy(&search_key, &found_key, sizeof(search_key));
6537 ret = log_new_ancestors(trans, root, path, ctx);
6540 btrfs_release_path(path);
6545 btrfs_free_path(path);
6550 * helper function around btrfs_log_inode to make sure newly created
6551 * parent directories also end up in the log. A minimal inode and backref
6552 * only logging is done of any parent directories that are older than
6553 * the last committed transaction
6555 static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
6556 struct btrfs_inode *inode,
6557 struct dentry *parent,
6559 struct btrfs_log_ctx *ctx)
6561 struct btrfs_root *root = inode->root;
6562 struct btrfs_fs_info *fs_info = root->fs_info;
6564 bool log_dentries = false;
6566 if (btrfs_test_opt(fs_info, NOTREELOG)) {
6567 ret = BTRFS_LOG_FORCE_COMMIT;
6571 if (btrfs_root_refs(&root->root_item) == 0) {
6572 ret = BTRFS_LOG_FORCE_COMMIT;
6577 * Skip already logged inodes or inodes corresponding to tmpfiles
6578 * (since logging them is pointless, a link count of 0 means they
6579 * will never be accessible).
6581 if ((btrfs_inode_in_log(inode, trans->transid) &&
6582 list_empty(&ctx->ordered_extents)) ||
6583 inode->vfs_inode.i_nlink == 0) {
6584 ret = BTRFS_NO_LOG_SYNC;
6588 ret = start_log_trans(trans, root, ctx);
6592 ret = btrfs_log_inode(trans, inode, inode_only, ctx);
6597 * for regular files, if its inode is already on disk, we don't
6598 * have to worry about the parents at all. This is because
6599 * we can use the last_unlink_trans field to record renames
6600 * and other fun in this file.
6602 if (S_ISREG(inode->vfs_inode.i_mode) &&
6603 inode->generation < trans->transid &&
6604 inode->last_unlink_trans < trans->transid) {
6609 if (S_ISDIR(inode->vfs_inode.i_mode) && ctx->log_new_dentries)
6610 log_dentries = true;
6613 * On unlink we must make sure all our current and old parent directory
6614 * inodes are fully logged. This is to prevent leaving dangling
6615 * directory index entries in directories that were our parents but are
6616 * not anymore. Not doing this results in old parent directory being
6617 * impossible to delete after log replay (rmdir will always fail with
6618 * error -ENOTEMPTY).
6624 * ln testdir/foo testdir/bar
6626 * unlink testdir/bar
6627 * xfs_io -c fsync testdir/foo
6629 * mount fs, triggers log replay
6631 * If we don't log the parent directory (testdir), after log replay the
6632 * directory still has an entry pointing to the file inode using the bar
6633 * name, but a matching BTRFS_INODE_[REF|EXTREF]_KEY does not exist and
6634 * the file inode has a link count of 1.
6640 * ln foo testdir/foo2
6641 * ln foo testdir/foo3
6643 * unlink testdir/foo3
6644 * xfs_io -c fsync foo
6646 * mount fs, triggers log replay
6648 * Similar as the first example, after log replay the parent directory
6649 * testdir still has an entry pointing to the inode file with name foo3
6650 * but the file inode does not have a matching BTRFS_INODE_REF_KEY item
6651 * and has a link count of 2.
6653 if (inode->last_unlink_trans >= trans->transid) {
6654 ret = btrfs_log_all_parents(trans, inode, ctx);
6659 ret = log_all_new_ancestors(trans, inode, parent, ctx);
6664 ret = log_new_dir_dentries(trans, root, inode, ctx);
6669 btrfs_set_log_full_commit(trans);
6670 ret = BTRFS_LOG_FORCE_COMMIT;
6674 btrfs_remove_log_ctx(root, ctx);
6675 btrfs_end_log_trans(root);
6681 * it is not safe to log dentry if the chunk root has added new
6682 * chunks. This returns 0 if the dentry was logged, and 1 otherwise.
6683 * If this returns 1, you must commit the transaction to safely get your
6686 int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
6687 struct dentry *dentry,
6688 struct btrfs_log_ctx *ctx)
6690 struct dentry *parent = dget_parent(dentry);
6693 ret = btrfs_log_inode_parent(trans, BTRFS_I(d_inode(dentry)), parent,
6694 LOG_INODE_ALL, ctx);
6701 * should be called during mount to recover any replay any log trees
6704 int btrfs_recover_log_trees(struct btrfs_root *log_root_tree)
6707 struct btrfs_path *path;
6708 struct btrfs_trans_handle *trans;
6709 struct btrfs_key key;
6710 struct btrfs_key found_key;
6711 struct btrfs_root *log;
6712 struct btrfs_fs_info *fs_info = log_root_tree->fs_info;
6713 struct walk_control wc = {
6714 .process_func = process_one_buffer,
6715 .stage = LOG_WALK_PIN_ONLY,
6718 path = btrfs_alloc_path();
6722 set_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags);
6724 trans = btrfs_start_transaction(fs_info->tree_root, 0);
6725 if (IS_ERR(trans)) {
6726 ret = PTR_ERR(trans);
6733 ret = walk_log_tree(trans, log_root_tree, &wc);
6735 btrfs_abort_transaction(trans, ret);
6740 key.objectid = BTRFS_TREE_LOG_OBJECTID;
6741 key.offset = (u64)-1;
6742 key.type = BTRFS_ROOT_ITEM_KEY;
6745 ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0);
6748 btrfs_abort_transaction(trans, ret);
6752 if (path->slots[0] == 0)
6756 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
6758 btrfs_release_path(path);
6759 if (found_key.objectid != BTRFS_TREE_LOG_OBJECTID)
6762 log = btrfs_read_tree_root(log_root_tree, &found_key);
6765 btrfs_abort_transaction(trans, ret);
6769 wc.replay_dest = btrfs_get_fs_root(fs_info, found_key.offset,
6771 if (IS_ERR(wc.replay_dest)) {
6772 ret = PTR_ERR(wc.replay_dest);
6775 * We didn't find the subvol, likely because it was
6776 * deleted. This is ok, simply skip this log and go to
6779 * We need to exclude the root because we can't have
6780 * other log replays overwriting this log as we'll read
6781 * it back in a few more times. This will keep our
6782 * block from being modified, and we'll just bail for
6783 * each subsequent pass.
6786 ret = btrfs_pin_extent_for_log_replay(trans,
6789 btrfs_put_root(log);
6793 btrfs_abort_transaction(trans, ret);
6797 wc.replay_dest->log_root = log;
6798 ret = btrfs_record_root_in_trans(trans, wc.replay_dest);
6800 /* The loop needs to continue due to the root refs */
6801 btrfs_abort_transaction(trans, ret);
6803 ret = walk_log_tree(trans, log, &wc);
6805 if (!ret && wc.stage == LOG_WALK_REPLAY_ALL) {
6806 ret = fixup_inode_link_counts(trans, wc.replay_dest,
6809 btrfs_abort_transaction(trans, ret);
6812 if (!ret && wc.stage == LOG_WALK_REPLAY_ALL) {
6813 struct btrfs_root *root = wc.replay_dest;
6815 btrfs_release_path(path);
6818 * We have just replayed everything, and the highest
6819 * objectid of fs roots probably has changed in case
6820 * some inode_item's got replayed.
6822 * root->objectid_mutex is not acquired as log replay
6823 * could only happen during mount.
6825 ret = btrfs_init_root_free_objectid(root);
6827 btrfs_abort_transaction(trans, ret);
6830 wc.replay_dest->log_root = NULL;
6831 btrfs_put_root(wc.replay_dest);
6832 btrfs_put_root(log);
6837 if (found_key.offset == 0)
6839 key.offset = found_key.offset - 1;
6841 btrfs_release_path(path);
6843 /* step one is to pin it all, step two is to replay just inodes */
6846 wc.process_func = replay_one_buffer;
6847 wc.stage = LOG_WALK_REPLAY_INODES;
6850 /* step three is to replay everything */
6851 if (wc.stage < LOG_WALK_REPLAY_ALL) {
6856 btrfs_free_path(path);
6858 /* step 4: commit the transaction, which also unpins the blocks */
6859 ret = btrfs_commit_transaction(trans);
6863 log_root_tree->log_root = NULL;
6864 clear_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags);
6865 btrfs_put_root(log_root_tree);
6870 btrfs_end_transaction(wc.trans);
6871 clear_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags);
6872 btrfs_free_path(path);
6877 * there are some corner cases where we want to force a full
6878 * commit instead of allowing a directory to be logged.
6880 * They revolve around files there were unlinked from the directory, and
6881 * this function updates the parent directory so that a full commit is
6882 * properly done if it is fsync'd later after the unlinks are done.
6884 * Must be called before the unlink operations (updates to the subvolume tree,
6885 * inodes, etc) are done.
6887 void btrfs_record_unlink_dir(struct btrfs_trans_handle *trans,
6888 struct btrfs_inode *dir, struct btrfs_inode *inode,
6892 * when we're logging a file, if it hasn't been renamed
6893 * or unlinked, and its inode is fully committed on disk,
6894 * we don't have to worry about walking up the directory chain
6895 * to log its parents.
6897 * So, we use the last_unlink_trans field to put this transid
6898 * into the file. When the file is logged we check it and
6899 * don't log the parents if the file is fully on disk.
6901 mutex_lock(&inode->log_mutex);
6902 inode->last_unlink_trans = trans->transid;
6903 mutex_unlock(&inode->log_mutex);
6906 * if this directory was already logged any new
6907 * names for this file/dir will get recorded
6909 if (dir->logged_trans == trans->transid)
6913 * if the inode we're about to unlink was logged,
6914 * the log will be properly updated for any new names
6916 if (inode->logged_trans == trans->transid)
6920 * when renaming files across directories, if the directory
6921 * there we're unlinking from gets fsync'd later on, there's
6922 * no way to find the destination directory later and fsync it
6923 * properly. So, we have to be conservative and force commits
6924 * so the new name gets discovered.
6929 /* we can safely do the unlink without any special recording */
6933 mutex_lock(&dir->log_mutex);
6934 dir->last_unlink_trans = trans->transid;
6935 mutex_unlock(&dir->log_mutex);
6939 * Make sure that if someone attempts to fsync the parent directory of a deleted
6940 * snapshot, it ends up triggering a transaction commit. This is to guarantee
6941 * that after replaying the log tree of the parent directory's root we will not
6942 * see the snapshot anymore and at log replay time we will not see any log tree
6943 * corresponding to the deleted snapshot's root, which could lead to replaying
6944 * it after replaying the log tree of the parent directory (which would replay
6945 * the snapshot delete operation).
6947 * Must be called before the actual snapshot destroy operation (updates to the
6948 * parent root and tree of tree roots trees, etc) are done.
6950 void btrfs_record_snapshot_destroy(struct btrfs_trans_handle *trans,
6951 struct btrfs_inode *dir)
6953 mutex_lock(&dir->log_mutex);
6954 dir->last_unlink_trans = trans->transid;
6955 mutex_unlock(&dir->log_mutex);
6959 * Update the log after adding a new name for an inode.
6961 * @trans: Transaction handle.
6962 * @old_dentry: The dentry associated with the old name and the old
6964 * @old_dir: The inode of the previous parent directory for the case
6965 * of a rename. For a link operation, it must be NULL.
6966 * @old_dir_index: The index number associated with the old name, meaningful
6967 * only for rename operations (when @old_dir is not NULL).
6968 * Ignored for link operations.
6969 * @parent: The dentry associated with the directory under which the
6970 * new name is located.
6972 * Call this after adding a new name for an inode, as a result of a link or
6973 * rename operation, and it will properly update the log to reflect the new name.
6975 void btrfs_log_new_name(struct btrfs_trans_handle *trans,
6976 struct dentry *old_dentry, struct btrfs_inode *old_dir,
6977 u64 old_dir_index, struct dentry *parent)
6979 struct btrfs_inode *inode = BTRFS_I(d_inode(old_dentry));
6980 struct btrfs_root *root = inode->root;
6981 struct btrfs_log_ctx ctx;
6982 bool log_pinned = false;
6986 * this will force the logging code to walk the dentry chain
6989 if (!S_ISDIR(inode->vfs_inode.i_mode))
6990 inode->last_unlink_trans = trans->transid;
6993 * if this inode hasn't been logged and directory we're renaming it
6994 * from hasn't been logged, we don't need to log it
6996 ret = inode_logged(trans, inode, NULL);
6999 } else if (ret == 0) {
7003 * If the inode was not logged and we are doing a rename (old_dir is not
7004 * NULL), check if old_dir was logged - if it was not we can return and
7007 ret = inode_logged(trans, old_dir, NULL);
7016 * If we are doing a rename (old_dir is not NULL) from a directory that
7017 * was previously logged, make sure that on log replay we get the old
7018 * dir entry deleted. This is needed because we will also log the new
7019 * name of the renamed inode, so we need to make sure that after log
7020 * replay we don't end up with both the new and old dir entries existing.
7022 if (old_dir && old_dir->logged_trans == trans->transid) {
7023 struct btrfs_root *log = old_dir->root->log_root;
7024 struct btrfs_path *path;
7026 ASSERT(old_dir_index >= BTRFS_DIR_START_INDEX);
7029 * We have two inodes to update in the log, the old directory and
7030 * the inode that got renamed, so we must pin the log to prevent
7031 * anyone from syncing the log until we have updated both inodes
7034 ret = join_running_log_trans(root);
7036 * At least one of the inodes was logged before, so this should
7037 * not fail, but if it does, it's not serious, just bail out and
7038 * mark the log for a full commit.
7040 if (WARN_ON_ONCE(ret < 0))
7044 path = btrfs_alloc_path();
7051 * Other concurrent task might be logging the old directory,
7052 * as it can be triggered when logging other inode that had or
7053 * still has a dentry in the old directory. We lock the old
7054 * directory's log_mutex to ensure the deletion of the old
7055 * name is persisted, because during directory logging we
7056 * delete all BTRFS_DIR_LOG_INDEX_KEY keys and the deletion of
7057 * the old name's dir index item is in the delayed items, so
7058 * it could be missed by an in progress directory logging.
7060 mutex_lock(&old_dir->log_mutex);
7061 ret = del_logged_dentry(trans, log, path, btrfs_ino(old_dir),
7062 old_dentry->d_name.name,
7063 old_dentry->d_name.len, old_dir_index);
7066 * The dentry does not exist in the log, so record its
7069 btrfs_release_path(path);
7070 ret = insert_dir_log_key(trans, log, path,
7072 old_dir_index, old_dir_index);
7074 mutex_unlock(&old_dir->log_mutex);
7076 btrfs_free_path(path);
7081 btrfs_init_log_ctx(&ctx, &inode->vfs_inode);
7082 ctx.logging_new_name = true;
7084 * We don't care about the return value. If we fail to log the new name
7085 * then we know the next attempt to sync the log will fallback to a full
7086 * transaction commit (due to a call to btrfs_set_log_full_commit()), so
7087 * we don't need to worry about getting a log committed that has an
7088 * inconsistent state after a rename operation.
7090 btrfs_log_inode_parent(trans, inode, parent, LOG_INODE_EXISTS, &ctx);
7093 * If an error happened mark the log for a full commit because it's not
7094 * consistent and up to date or we couldn't find out if one of the
7095 * inodes was logged before in this transaction. Do it before unpinning
7096 * the log, to avoid any races with someone else trying to commit it.
7099 btrfs_set_log_full_commit(trans);
7101 btrfs_end_log_trans(root);