1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/slab.h>
7 #include <linux/blkdev.h>
8 #include <linux/writeback.h>
9 #include <linux/sched/mm.h>
12 #include "transaction.h"
13 #include "btrfs_inode.h"
14 #include "extent_io.h"
16 #include "compression.h"
17 #include "delalloc-space.h"
20 static struct kmem_cache *btrfs_ordered_extent_cache;
22 static u64 entry_end(struct btrfs_ordered_extent *entry)
24 if (entry->file_offset + entry->num_bytes < entry->file_offset)
26 return entry->file_offset + entry->num_bytes;
29 /* returns NULL if the insertion worked, or it returns the node it did find
32 static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset,
35 struct rb_node **p = &root->rb_node;
36 struct rb_node *parent = NULL;
37 struct btrfs_ordered_extent *entry;
41 entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node);
43 if (file_offset < entry->file_offset)
45 else if (file_offset >= entry_end(entry))
51 rb_link_node(node, parent, p);
52 rb_insert_color(node, root);
57 * look for a given offset in the tree, and if it can't be found return the
60 static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset,
61 struct rb_node **prev_ret)
63 struct rb_node *n = root->rb_node;
64 struct rb_node *prev = NULL;
66 struct btrfs_ordered_extent *entry;
67 struct btrfs_ordered_extent *prev_entry = NULL;
70 entry = rb_entry(n, struct btrfs_ordered_extent, rb_node);
74 if (file_offset < entry->file_offset)
76 else if (file_offset >= entry_end(entry))
84 while (prev && file_offset >= entry_end(prev_entry)) {
88 prev_entry = rb_entry(test, struct btrfs_ordered_extent,
90 if (file_offset < entry_end(prev_entry))
96 prev_entry = rb_entry(prev, struct btrfs_ordered_extent,
98 while (prev && file_offset < entry_end(prev_entry)) {
102 prev_entry = rb_entry(test, struct btrfs_ordered_extent,
111 * helper to check if a given offset is inside a given entry
113 static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset)
115 if (file_offset < entry->file_offset ||
116 entry->file_offset + entry->num_bytes <= file_offset)
121 static int range_overlaps(struct btrfs_ordered_extent *entry, u64 file_offset,
124 if (file_offset + len <= entry->file_offset ||
125 entry->file_offset + entry->num_bytes <= file_offset)
131 * look find the first ordered struct that has this offset, otherwise
132 * the first one less than this offset
134 static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
137 struct rb_root *root = &tree->tree;
138 struct rb_node *prev = NULL;
140 struct btrfs_ordered_extent *entry;
143 entry = rb_entry(tree->last, struct btrfs_ordered_extent,
145 if (offset_in_entry(entry, file_offset))
148 ret = __tree_search(root, file_offset, &prev);
157 * Allocate and add a new ordered_extent into the per-inode tree.
159 * The tree is given a single reference on the ordered extent that was
162 static int __btrfs_add_ordered_extent(struct btrfs_inode *inode, u64 file_offset,
163 u64 disk_bytenr, u64 num_bytes,
164 u64 disk_num_bytes, int type, int dio,
167 struct btrfs_root *root = inode->root;
168 struct btrfs_fs_info *fs_info = root->fs_info;
169 struct btrfs_ordered_inode_tree *tree = &inode->ordered_tree;
170 struct rb_node *node;
171 struct btrfs_ordered_extent *entry;
174 if (type == BTRFS_ORDERED_NOCOW || type == BTRFS_ORDERED_PREALLOC) {
175 /* For nocow write, we can release the qgroup rsv right now */
176 ret = btrfs_qgroup_free_data(inode, NULL, file_offset, num_bytes);
182 * The ordered extent has reserved qgroup space, release now
183 * and pass the reserved number for qgroup_record to free.
185 ret = btrfs_qgroup_release_data(inode, file_offset, num_bytes);
189 entry = kmem_cache_zalloc(btrfs_ordered_extent_cache, GFP_NOFS);
193 entry->file_offset = file_offset;
194 entry->disk_bytenr = disk_bytenr;
195 entry->num_bytes = num_bytes;
196 entry->disk_num_bytes = disk_num_bytes;
197 entry->bytes_left = num_bytes;
198 entry->inode = igrab(&inode->vfs_inode);
199 entry->compress_type = compress_type;
200 entry->truncated_len = (u64)-1;
201 entry->qgroup_rsv = ret;
202 entry->physical = (u64)-1;
204 entry->partno = (u8)-1;
206 ASSERT(type == BTRFS_ORDERED_REGULAR ||
207 type == BTRFS_ORDERED_NOCOW ||
208 type == BTRFS_ORDERED_PREALLOC ||
209 type == BTRFS_ORDERED_COMPRESSED);
210 set_bit(type, &entry->flags);
212 percpu_counter_add_batch(&fs_info->ordered_bytes, num_bytes,
213 fs_info->delalloc_batch);
216 set_bit(BTRFS_ORDERED_DIRECT, &entry->flags);
218 /* one ref for the tree */
219 refcount_set(&entry->refs, 1);
220 init_waitqueue_head(&entry->wait);
221 INIT_LIST_HEAD(&entry->list);
222 INIT_LIST_HEAD(&entry->log_list);
223 INIT_LIST_HEAD(&entry->root_extent_list);
224 INIT_LIST_HEAD(&entry->work_list);
225 init_completion(&entry->completion);
227 trace_btrfs_ordered_extent_add(inode, entry);
229 spin_lock_irq(&tree->lock);
230 node = tree_insert(&tree->tree, file_offset,
233 btrfs_panic(fs_info, -EEXIST,
234 "inconsistency in ordered tree at offset %llu",
236 spin_unlock_irq(&tree->lock);
238 spin_lock(&root->ordered_extent_lock);
239 list_add_tail(&entry->root_extent_list,
240 &root->ordered_extents);
241 root->nr_ordered_extents++;
242 if (root->nr_ordered_extents == 1) {
243 spin_lock(&fs_info->ordered_root_lock);
244 BUG_ON(!list_empty(&root->ordered_root));
245 list_add_tail(&root->ordered_root, &fs_info->ordered_roots);
246 spin_unlock(&fs_info->ordered_root_lock);
248 spin_unlock(&root->ordered_extent_lock);
251 * We don't need the count_max_extents here, we can assume that all of
252 * that work has been done at higher layers, so this is truly the
253 * smallest the extent is going to get.
255 spin_lock(&inode->lock);
256 btrfs_mod_outstanding_extents(inode, 1);
257 spin_unlock(&inode->lock);
262 int btrfs_add_ordered_extent(struct btrfs_inode *inode, u64 file_offset,
263 u64 disk_bytenr, u64 num_bytes, u64 disk_num_bytes,
266 ASSERT(type == BTRFS_ORDERED_REGULAR ||
267 type == BTRFS_ORDERED_NOCOW ||
268 type == BTRFS_ORDERED_PREALLOC);
269 return __btrfs_add_ordered_extent(inode, file_offset, disk_bytenr,
270 num_bytes, disk_num_bytes, type, 0,
271 BTRFS_COMPRESS_NONE);
274 int btrfs_add_ordered_extent_dio(struct btrfs_inode *inode, u64 file_offset,
275 u64 disk_bytenr, u64 num_bytes,
276 u64 disk_num_bytes, int type)
278 ASSERT(type == BTRFS_ORDERED_REGULAR ||
279 type == BTRFS_ORDERED_NOCOW ||
280 type == BTRFS_ORDERED_PREALLOC);
281 return __btrfs_add_ordered_extent(inode, file_offset, disk_bytenr,
282 num_bytes, disk_num_bytes, type, 1,
283 BTRFS_COMPRESS_NONE);
286 int btrfs_add_ordered_extent_compress(struct btrfs_inode *inode, u64 file_offset,
287 u64 disk_bytenr, u64 num_bytes,
288 u64 disk_num_bytes, int compress_type)
290 ASSERT(compress_type != BTRFS_COMPRESS_NONE);
291 return __btrfs_add_ordered_extent(inode, file_offset, disk_bytenr,
292 num_bytes, disk_num_bytes,
293 BTRFS_ORDERED_COMPRESSED, 0,
298 * Add a struct btrfs_ordered_sum into the list of checksums to be inserted
299 * when an ordered extent is finished. If the list covers more than one
300 * ordered extent, it is split across multiples.
302 void btrfs_add_ordered_sum(struct btrfs_ordered_extent *entry,
303 struct btrfs_ordered_sum *sum)
305 struct btrfs_ordered_inode_tree *tree;
307 tree = &BTRFS_I(entry->inode)->ordered_tree;
308 spin_lock_irq(&tree->lock);
309 list_add_tail(&sum->list, &entry->list);
310 spin_unlock_irq(&tree->lock);
314 * Finish IO for one ordered extent across a given range. The range can
315 * contain several ordered extents.
317 * @found_ret: Return the finished ordered extent
318 * @file_offset: File offset for the finished IO
319 * Will also be updated to one byte past the range that is
320 * recordered as finished. This allows caller to walk forward.
321 * @io_size: Length of the finish IO range
322 * @uptodate: If the IO finished without problem
324 * Return true if any ordered extent is finished in the range, and update
325 * @found_ret and @file_offset.
326 * Return false otherwise.
328 * NOTE: Although The range can cross multiple ordered extents, only one
329 * ordered extent will be updated during one call. The caller is responsible to
330 * iterate all ordered extents in the range.
332 bool btrfs_dec_test_first_ordered_pending(struct btrfs_inode *inode,
333 struct btrfs_ordered_extent **finished_ret,
334 u64 *file_offset, u64 io_size, int uptodate)
336 struct btrfs_fs_info *fs_info = inode->root->fs_info;
337 struct btrfs_ordered_inode_tree *tree = &inode->ordered_tree;
338 struct rb_node *node;
339 struct btrfs_ordered_extent *entry = NULL;
340 bool finished = false;
346 spin_lock_irqsave(&tree->lock, flags);
347 node = tree_search(tree, *file_offset);
351 entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
352 if (!offset_in_entry(entry, *file_offset))
355 dec_start = max(*file_offset, entry->file_offset);
356 dec_end = min(*file_offset + io_size,
357 entry->file_offset + entry->num_bytes);
358 *file_offset = dec_end;
359 if (dec_start > dec_end) {
360 btrfs_crit(fs_info, "bad ordering dec_start %llu end %llu",
363 to_dec = dec_end - dec_start;
364 if (to_dec > entry->bytes_left) {
366 "bad ordered accounting left %llu size %llu",
367 entry->bytes_left, to_dec);
369 entry->bytes_left -= to_dec;
371 set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
373 if (entry->bytes_left == 0) {
375 * Ensure only one caller can set the flag and finished_ret
378 finished = !test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
379 /* test_and_set_bit implies a barrier */
380 cond_wake_up_nomb(&entry->wait);
383 if (finished && finished_ret && entry) {
384 *finished_ret = entry;
385 refcount_inc(&entry->refs);
387 spin_unlock_irqrestore(&tree->lock, flags);
392 * Finish IO for one ordered extent across a given range. The range can only
393 * contain one ordered extent.
395 * @cached: The cached ordered extent. If not NULL, we can skip the tree
396 * search and use the ordered extent directly.
397 * Will be also used to store the finished ordered extent.
398 * @file_offset: File offset for the finished IO
399 * @io_size: Length of the finish IO range
400 * @uptodate: If the IO finishes without problem
402 * Return true if the ordered extent is finished in the range, and update
404 * Return false otherwise.
406 * NOTE: The range can NOT cross multiple ordered extents.
407 * Thus caller should ensure the range doesn't cross ordered extents.
409 bool btrfs_dec_test_ordered_pending(struct btrfs_inode *inode,
410 struct btrfs_ordered_extent **cached,
411 u64 file_offset, u64 io_size, int uptodate)
413 struct btrfs_ordered_inode_tree *tree = &inode->ordered_tree;
414 struct rb_node *node;
415 struct btrfs_ordered_extent *entry = NULL;
417 bool finished = false;
419 spin_lock_irqsave(&tree->lock, flags);
420 if (cached && *cached) {
425 node = tree_search(tree, file_offset);
429 entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
431 if (!offset_in_entry(entry, file_offset))
434 if (io_size > entry->bytes_left)
435 btrfs_crit(inode->root->fs_info,
436 "bad ordered accounting left %llu size %llu",
437 entry->bytes_left, io_size);
439 entry->bytes_left -= io_size;
441 set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
443 if (entry->bytes_left == 0) {
445 * Ensure only one caller can set the flag and finished_ret
448 finished = !test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
449 /* test_and_set_bit implies a barrier */
450 cond_wake_up_nomb(&entry->wait);
453 if (finished && cached && entry) {
455 refcount_inc(&entry->refs);
457 spin_unlock_irqrestore(&tree->lock, flags);
462 * used to drop a reference on an ordered extent. This will free
463 * the extent if the last reference is dropped
465 void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry)
467 struct list_head *cur;
468 struct btrfs_ordered_sum *sum;
470 trace_btrfs_ordered_extent_put(BTRFS_I(entry->inode), entry);
472 if (refcount_dec_and_test(&entry->refs)) {
473 ASSERT(list_empty(&entry->root_extent_list));
474 ASSERT(list_empty(&entry->log_list));
475 ASSERT(RB_EMPTY_NODE(&entry->rb_node));
477 btrfs_add_delayed_iput(entry->inode);
478 while (!list_empty(&entry->list)) {
479 cur = entry->list.next;
480 sum = list_entry(cur, struct btrfs_ordered_sum, list);
481 list_del(&sum->list);
484 kmem_cache_free(btrfs_ordered_extent_cache, entry);
489 * remove an ordered extent from the tree. No references are dropped
490 * and waiters are woken up.
492 void btrfs_remove_ordered_extent(struct btrfs_inode *btrfs_inode,
493 struct btrfs_ordered_extent *entry)
495 struct btrfs_ordered_inode_tree *tree;
496 struct btrfs_root *root = btrfs_inode->root;
497 struct btrfs_fs_info *fs_info = root->fs_info;
498 struct rb_node *node;
501 /* This is paired with btrfs_add_ordered_extent. */
502 spin_lock(&btrfs_inode->lock);
503 btrfs_mod_outstanding_extents(btrfs_inode, -1);
504 spin_unlock(&btrfs_inode->lock);
505 if (root != fs_info->tree_root)
506 btrfs_delalloc_release_metadata(btrfs_inode, entry->num_bytes,
509 percpu_counter_add_batch(&fs_info->ordered_bytes, -entry->num_bytes,
510 fs_info->delalloc_batch);
512 tree = &btrfs_inode->ordered_tree;
513 spin_lock_irq(&tree->lock);
514 node = &entry->rb_node;
515 rb_erase(node, &tree->tree);
517 if (tree->last == node)
519 set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
520 pending = test_and_clear_bit(BTRFS_ORDERED_PENDING, &entry->flags);
521 spin_unlock_irq(&tree->lock);
524 * The current running transaction is waiting on us, we need to let it
525 * know that we're complete and wake it up.
528 struct btrfs_transaction *trans;
531 * The checks for trans are just a formality, it should be set,
532 * but if it isn't we don't want to deref/assert under the spin
533 * lock, so be nice and check if trans is set, but ASSERT() so
534 * if it isn't set a developer will notice.
536 spin_lock(&fs_info->trans_lock);
537 trans = fs_info->running_transaction;
539 refcount_inc(&trans->use_count);
540 spin_unlock(&fs_info->trans_lock);
544 if (atomic_dec_and_test(&trans->pending_ordered))
545 wake_up(&trans->pending_wait);
546 btrfs_put_transaction(trans);
550 spin_lock(&root->ordered_extent_lock);
551 list_del_init(&entry->root_extent_list);
552 root->nr_ordered_extents--;
554 trace_btrfs_ordered_extent_remove(btrfs_inode, entry);
556 if (!root->nr_ordered_extents) {
557 spin_lock(&fs_info->ordered_root_lock);
558 BUG_ON(list_empty(&root->ordered_root));
559 list_del_init(&root->ordered_root);
560 spin_unlock(&fs_info->ordered_root_lock);
562 spin_unlock(&root->ordered_extent_lock);
563 wake_up(&entry->wait);
566 static void btrfs_run_ordered_extent_work(struct btrfs_work *work)
568 struct btrfs_ordered_extent *ordered;
570 ordered = container_of(work, struct btrfs_ordered_extent, flush_work);
571 btrfs_start_ordered_extent(ordered, 1);
572 complete(&ordered->completion);
576 * wait for all the ordered extents in a root. This is done when balancing
577 * space between drives.
579 u64 btrfs_wait_ordered_extents(struct btrfs_root *root, u64 nr,
580 const u64 range_start, const u64 range_len)
582 struct btrfs_fs_info *fs_info = root->fs_info;
586 struct btrfs_ordered_extent *ordered, *next;
588 const u64 range_end = range_start + range_len;
590 mutex_lock(&root->ordered_extent_mutex);
591 spin_lock(&root->ordered_extent_lock);
592 list_splice_init(&root->ordered_extents, &splice);
593 while (!list_empty(&splice) && nr) {
594 ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
597 if (range_end <= ordered->disk_bytenr ||
598 ordered->disk_bytenr + ordered->disk_num_bytes <= range_start) {
599 list_move_tail(&ordered->root_extent_list, &skipped);
600 cond_resched_lock(&root->ordered_extent_lock);
604 list_move_tail(&ordered->root_extent_list,
605 &root->ordered_extents);
606 refcount_inc(&ordered->refs);
607 spin_unlock(&root->ordered_extent_lock);
609 btrfs_init_work(&ordered->flush_work,
610 btrfs_run_ordered_extent_work, NULL, NULL);
611 list_add_tail(&ordered->work_list, &works);
612 btrfs_queue_work(fs_info->flush_workers, &ordered->flush_work);
615 spin_lock(&root->ordered_extent_lock);
620 list_splice_tail(&skipped, &root->ordered_extents);
621 list_splice_tail(&splice, &root->ordered_extents);
622 spin_unlock(&root->ordered_extent_lock);
624 list_for_each_entry_safe(ordered, next, &works, work_list) {
625 list_del_init(&ordered->work_list);
626 wait_for_completion(&ordered->completion);
627 btrfs_put_ordered_extent(ordered);
630 mutex_unlock(&root->ordered_extent_mutex);
635 void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, u64 nr,
636 const u64 range_start, const u64 range_len)
638 struct btrfs_root *root;
639 struct list_head splice;
642 INIT_LIST_HEAD(&splice);
644 mutex_lock(&fs_info->ordered_operations_mutex);
645 spin_lock(&fs_info->ordered_root_lock);
646 list_splice_init(&fs_info->ordered_roots, &splice);
647 while (!list_empty(&splice) && nr) {
648 root = list_first_entry(&splice, struct btrfs_root,
650 root = btrfs_grab_root(root);
652 list_move_tail(&root->ordered_root,
653 &fs_info->ordered_roots);
654 spin_unlock(&fs_info->ordered_root_lock);
656 done = btrfs_wait_ordered_extents(root, nr,
657 range_start, range_len);
658 btrfs_put_root(root);
660 spin_lock(&fs_info->ordered_root_lock);
665 list_splice_tail(&splice, &fs_info->ordered_roots);
666 spin_unlock(&fs_info->ordered_root_lock);
667 mutex_unlock(&fs_info->ordered_operations_mutex);
671 * Used to start IO or wait for a given ordered extent to finish.
673 * If wait is one, this effectively waits on page writeback for all the pages
674 * in the extent, and it waits on the io completion code to insert
675 * metadata into the btree corresponding to the extent
677 void btrfs_start_ordered_extent(struct btrfs_ordered_extent *entry, int wait)
679 u64 start = entry->file_offset;
680 u64 end = start + entry->num_bytes - 1;
681 struct btrfs_inode *inode = BTRFS_I(entry->inode);
683 trace_btrfs_ordered_extent_start(inode, entry);
686 * pages in the range can be dirty, clean or writeback. We
687 * start IO on any dirty ones so the wait doesn't stall waiting
688 * for the flusher thread to find them
690 if (!test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
691 filemap_fdatawrite_range(inode->vfs_inode.i_mapping, start, end);
693 wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE,
699 * Used to wait on ordered extents across a large range of bytes.
701 int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
707 struct btrfs_ordered_extent *ordered;
709 if (start + len < start) {
710 orig_end = INT_LIMIT(loff_t);
712 orig_end = start + len - 1;
713 if (orig_end > INT_LIMIT(loff_t))
714 orig_end = INT_LIMIT(loff_t);
717 /* start IO across the range first to instantiate any delalloc
720 ret = btrfs_fdatawrite_range(inode, start, orig_end);
725 * If we have a writeback error don't return immediately. Wait first
726 * for any ordered extents that haven't completed yet. This is to make
727 * sure no one can dirty the same page ranges and call writepages()
728 * before the ordered extents complete - to avoid failures (-EEXIST)
729 * when adding the new ordered extents to the ordered tree.
731 ret_wb = filemap_fdatawait_range(inode->i_mapping, start, orig_end);
735 ordered = btrfs_lookup_first_ordered_extent(BTRFS_I(inode), end);
738 if (ordered->file_offset > orig_end) {
739 btrfs_put_ordered_extent(ordered);
742 if (ordered->file_offset + ordered->num_bytes <= start) {
743 btrfs_put_ordered_extent(ordered);
746 btrfs_start_ordered_extent(ordered, 1);
747 end = ordered->file_offset;
749 * If the ordered extent had an error save the error but don't
750 * exit without waiting first for all other ordered extents in
751 * the range to complete.
753 if (test_bit(BTRFS_ORDERED_IOERR, &ordered->flags))
755 btrfs_put_ordered_extent(ordered);
756 if (end == 0 || end == start)
760 return ret_wb ? ret_wb : ret;
764 * find an ordered extent corresponding to file_offset. return NULL if
765 * nothing is found, otherwise take a reference on the extent and return it
767 struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct btrfs_inode *inode,
770 struct btrfs_ordered_inode_tree *tree;
771 struct rb_node *node;
772 struct btrfs_ordered_extent *entry = NULL;
775 tree = &inode->ordered_tree;
776 spin_lock_irqsave(&tree->lock, flags);
777 node = tree_search(tree, file_offset);
781 entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
782 if (!offset_in_entry(entry, file_offset))
785 refcount_inc(&entry->refs);
787 spin_unlock_irqrestore(&tree->lock, flags);
791 /* Since the DIO code tries to lock a wide area we need to look for any ordered
792 * extents that exist in the range, rather than just the start of the range.
794 struct btrfs_ordered_extent *btrfs_lookup_ordered_range(
795 struct btrfs_inode *inode, u64 file_offset, u64 len)
797 struct btrfs_ordered_inode_tree *tree;
798 struct rb_node *node;
799 struct btrfs_ordered_extent *entry = NULL;
801 tree = &inode->ordered_tree;
802 spin_lock_irq(&tree->lock);
803 node = tree_search(tree, file_offset);
805 node = tree_search(tree, file_offset + len);
811 entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
812 if (range_overlaps(entry, file_offset, len))
815 if (entry->file_offset >= file_offset + len) {
820 node = rb_next(node);
826 refcount_inc(&entry->refs);
827 spin_unlock_irq(&tree->lock);
832 * Adds all ordered extents to the given list. The list ends up sorted by the
833 * file_offset of the ordered extents.
835 void btrfs_get_ordered_extents_for_logging(struct btrfs_inode *inode,
836 struct list_head *list)
838 struct btrfs_ordered_inode_tree *tree = &inode->ordered_tree;
841 ASSERT(inode_is_locked(&inode->vfs_inode));
843 spin_lock_irq(&tree->lock);
844 for (n = rb_first(&tree->tree); n; n = rb_next(n)) {
845 struct btrfs_ordered_extent *ordered;
847 ordered = rb_entry(n, struct btrfs_ordered_extent, rb_node);
849 if (test_bit(BTRFS_ORDERED_LOGGED, &ordered->flags))
852 ASSERT(list_empty(&ordered->log_list));
853 list_add_tail(&ordered->log_list, list);
854 refcount_inc(&ordered->refs);
856 spin_unlock_irq(&tree->lock);
860 * lookup and return any extent before 'file_offset'. NULL is returned
863 struct btrfs_ordered_extent *
864 btrfs_lookup_first_ordered_extent(struct btrfs_inode *inode, u64 file_offset)
866 struct btrfs_ordered_inode_tree *tree;
867 struct rb_node *node;
868 struct btrfs_ordered_extent *entry = NULL;
870 tree = &inode->ordered_tree;
871 spin_lock_irq(&tree->lock);
872 node = tree_search(tree, file_offset);
876 entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
877 refcount_inc(&entry->refs);
879 spin_unlock_irq(&tree->lock);
884 * btrfs_flush_ordered_range - Lock the passed range and ensures all pending
885 * ordered extents in it are run to completion.
887 * @inode: Inode whose ordered tree is to be searched
888 * @start: Beginning of range to flush
889 * @end: Last byte of range to lock
890 * @cached_state: If passed, will return the extent state responsible for the
891 * locked range. It's the caller's responsibility to free the cached state.
893 * This function always returns with the given range locked, ensuring after it's
894 * called no order extent can be pending.
896 void btrfs_lock_and_flush_ordered_range(struct btrfs_inode *inode, u64 start,
898 struct extent_state **cached_state)
900 struct btrfs_ordered_extent *ordered;
901 struct extent_state *cache = NULL;
902 struct extent_state **cachedp = &cache;
905 cachedp = cached_state;
908 lock_extent_bits(&inode->io_tree, start, end, cachedp);
909 ordered = btrfs_lookup_ordered_range(inode, start,
913 * If no external cached_state has been passed then
914 * decrement the extra ref taken for cachedp since we
915 * aren't exposing it outside of this function
918 refcount_dec(&cache->refs);
921 unlock_extent_cached(&inode->io_tree, start, end, cachedp);
922 btrfs_start_ordered_extent(ordered, 1);
923 btrfs_put_ordered_extent(ordered);
927 static int clone_ordered_extent(struct btrfs_ordered_extent *ordered, u64 pos,
930 struct inode *inode = ordered->inode;
931 u64 file_offset = ordered->file_offset + pos;
932 u64 disk_bytenr = ordered->disk_bytenr + pos;
934 u64 disk_num_bytes = len;
936 unsigned long flags_masked = ordered->flags & ~(1 << BTRFS_ORDERED_DIRECT);
937 int compress_type = ordered->compress_type;
938 unsigned long weight;
941 weight = hweight_long(flags_masked);
942 WARN_ON_ONCE(weight > 1);
946 type = __ffs(flags_masked);
948 if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered->flags)) {
950 ret = btrfs_add_ordered_extent_compress(BTRFS_I(inode),
951 file_offset, disk_bytenr, num_bytes,
952 disk_num_bytes, compress_type);
953 } else if (test_bit(BTRFS_ORDERED_DIRECT, &ordered->flags)) {
954 ret = btrfs_add_ordered_extent_dio(BTRFS_I(inode), file_offset,
955 disk_bytenr, num_bytes, disk_num_bytes, type);
957 ret = btrfs_add_ordered_extent(BTRFS_I(inode), file_offset,
958 disk_bytenr, num_bytes, disk_num_bytes, type);
964 int btrfs_split_ordered_extent(struct btrfs_ordered_extent *ordered, u64 pre,
967 struct inode *inode = ordered->inode;
968 struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
969 struct rb_node *node;
970 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
973 spin_lock_irq(&tree->lock);
974 /* Remove from tree once */
975 node = &ordered->rb_node;
976 rb_erase(node, &tree->tree);
978 if (tree->last == node)
981 ordered->file_offset += pre;
982 ordered->disk_bytenr += pre;
983 ordered->num_bytes -= (pre + post);
984 ordered->disk_num_bytes -= (pre + post);
985 ordered->bytes_left -= (pre + post);
987 /* Re-insert the node */
988 node = tree_insert(&tree->tree, ordered->file_offset, &ordered->rb_node);
990 btrfs_panic(fs_info, -EEXIST,
991 "zoned: inconsistency in ordered tree at offset %llu",
992 ordered->file_offset);
994 spin_unlock_irq(&tree->lock);
997 ret = clone_ordered_extent(ordered, 0, pre);
999 ret = clone_ordered_extent(ordered, pre + ordered->disk_num_bytes,
1005 int __init ordered_data_init(void)
1007 btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent",
1008 sizeof(struct btrfs_ordered_extent), 0,
1011 if (!btrfs_ordered_extent_cache)
1017 void __cold ordered_data_exit(void)
1019 kmem_cache_destroy(btrfs_ordered_extent_cache);