xref: /linux/fs/btrfs/ordered-data.c (revision 63307d015b91e626c97bb82e88054af3d0b74643)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #include <linux/slab.h>
7 #include <linux/blkdev.h>
8 #include <linux/writeback.h>
9 #include <linux/sched/mm.h>
10 #include "ctree.h"
11 #include "transaction.h"
12 #include "btrfs_inode.h"
13 #include "extent_io.h"
14 #include "disk-io.h"
15 #include "compression.h"
16 
17 static struct kmem_cache *btrfs_ordered_extent_cache;
18 
19 static u64 entry_end(struct btrfs_ordered_extent *entry)
20 {
21 	if (entry->file_offset + entry->len < entry->file_offset)
22 		return (u64)-1;
23 	return entry->file_offset + entry->len;
24 }
25 
26 /* returns NULL if the insertion worked, or it returns the node it did find
27  * in the tree
28  */
29 static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset,
30 				   struct rb_node *node)
31 {
32 	struct rb_node **p = &root->rb_node;
33 	struct rb_node *parent = NULL;
34 	struct btrfs_ordered_extent *entry;
35 
36 	while (*p) {
37 		parent = *p;
38 		entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node);
39 
40 		if (file_offset < entry->file_offset)
41 			p = &(*p)->rb_left;
42 		else if (file_offset >= entry_end(entry))
43 			p = &(*p)->rb_right;
44 		else
45 			return parent;
46 	}
47 
48 	rb_link_node(node, parent, p);
49 	rb_insert_color(node, root);
50 	return NULL;
51 }
52 
53 static void ordered_data_tree_panic(struct inode *inode, int errno,
54 					       u64 offset)
55 {
56 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
57 	btrfs_panic(fs_info, errno,
58 		    "Inconsistency in ordered tree at offset %llu", offset);
59 }
60 
61 /*
62  * look for a given offset in the tree, and if it can't be found return the
63  * first lesser offset
64  */
65 static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset,
66 				     struct rb_node **prev_ret)
67 {
68 	struct rb_node *n = root->rb_node;
69 	struct rb_node *prev = NULL;
70 	struct rb_node *test;
71 	struct btrfs_ordered_extent *entry;
72 	struct btrfs_ordered_extent *prev_entry = NULL;
73 
74 	while (n) {
75 		entry = rb_entry(n, struct btrfs_ordered_extent, rb_node);
76 		prev = n;
77 		prev_entry = entry;
78 
79 		if (file_offset < entry->file_offset)
80 			n = n->rb_left;
81 		else if (file_offset >= entry_end(entry))
82 			n = n->rb_right;
83 		else
84 			return n;
85 	}
86 	if (!prev_ret)
87 		return NULL;
88 
89 	while (prev && file_offset >= entry_end(prev_entry)) {
90 		test = rb_next(prev);
91 		if (!test)
92 			break;
93 		prev_entry = rb_entry(test, struct btrfs_ordered_extent,
94 				      rb_node);
95 		if (file_offset < entry_end(prev_entry))
96 			break;
97 
98 		prev = test;
99 	}
100 	if (prev)
101 		prev_entry = rb_entry(prev, struct btrfs_ordered_extent,
102 				      rb_node);
103 	while (prev && file_offset < entry_end(prev_entry)) {
104 		test = rb_prev(prev);
105 		if (!test)
106 			break;
107 		prev_entry = rb_entry(test, struct btrfs_ordered_extent,
108 				      rb_node);
109 		prev = test;
110 	}
111 	*prev_ret = prev;
112 	return NULL;
113 }
114 
115 /*
116  * helper to check if a given offset is inside a given entry
117  */
118 static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset)
119 {
120 	if (file_offset < entry->file_offset ||
121 	    entry->file_offset + entry->len <= file_offset)
122 		return 0;
123 	return 1;
124 }
125 
126 static int range_overlaps(struct btrfs_ordered_extent *entry, u64 file_offset,
127 			  u64 len)
128 {
129 	if (file_offset + len <= entry->file_offset ||
130 	    entry->file_offset + entry->len <= file_offset)
131 		return 0;
132 	return 1;
133 }
134 
135 /*
136  * look find the first ordered struct that has this offset, otherwise
137  * the first one less than this offset
138  */
139 static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
140 					  u64 file_offset)
141 {
142 	struct rb_root *root = &tree->tree;
143 	struct rb_node *prev = NULL;
144 	struct rb_node *ret;
145 	struct btrfs_ordered_extent *entry;
146 
147 	if (tree->last) {
148 		entry = rb_entry(tree->last, struct btrfs_ordered_extent,
149 				 rb_node);
150 		if (offset_in_entry(entry, file_offset))
151 			return tree->last;
152 	}
153 	ret = __tree_search(root, file_offset, &prev);
154 	if (!ret)
155 		ret = prev;
156 	if (ret)
157 		tree->last = ret;
158 	return ret;
159 }
160 
161 /* allocate and add a new ordered_extent into the per-inode tree.
162  * file_offset is the logical offset in the file
163  *
164  * start is the disk block number of an extent already reserved in the
165  * extent allocation tree
166  *
167  * len is the length of the extent
168  *
169  * The tree is given a single reference on the ordered extent that was
170  * inserted.
171  */
172 static int __btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
173 				      u64 start, u64 len, u64 disk_len,
174 				      int type, int dio, int compress_type)
175 {
176 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
177 	struct btrfs_root *root = BTRFS_I(inode)->root;
178 	struct btrfs_ordered_inode_tree *tree;
179 	struct rb_node *node;
180 	struct btrfs_ordered_extent *entry;
181 
182 	tree = &BTRFS_I(inode)->ordered_tree;
183 	entry = kmem_cache_zalloc(btrfs_ordered_extent_cache, GFP_NOFS);
184 	if (!entry)
185 		return -ENOMEM;
186 
187 	entry->file_offset = file_offset;
188 	entry->start = start;
189 	entry->len = len;
190 	entry->disk_len = disk_len;
191 	entry->bytes_left = len;
192 	entry->inode = igrab(inode);
193 	entry->compress_type = compress_type;
194 	entry->truncated_len = (u64)-1;
195 	if (type != BTRFS_ORDERED_IO_DONE && type != BTRFS_ORDERED_COMPLETE)
196 		set_bit(type, &entry->flags);
197 
198 	if (dio) {
199 		percpu_counter_add_batch(&fs_info->dio_bytes, len,
200 					 fs_info->delalloc_batch);
201 		set_bit(BTRFS_ORDERED_DIRECT, &entry->flags);
202 	}
203 
204 	/* one ref for the tree */
205 	refcount_set(&entry->refs, 1);
206 	init_waitqueue_head(&entry->wait);
207 	INIT_LIST_HEAD(&entry->list);
208 	INIT_LIST_HEAD(&entry->root_extent_list);
209 	INIT_LIST_HEAD(&entry->work_list);
210 	init_completion(&entry->completion);
211 	INIT_LIST_HEAD(&entry->log_list);
212 	INIT_LIST_HEAD(&entry->trans_list);
213 
214 	trace_btrfs_ordered_extent_add(inode, entry);
215 
216 	spin_lock_irq(&tree->lock);
217 	node = tree_insert(&tree->tree, file_offset,
218 			   &entry->rb_node);
219 	if (node)
220 		ordered_data_tree_panic(inode, -EEXIST, file_offset);
221 	spin_unlock_irq(&tree->lock);
222 
223 	spin_lock(&root->ordered_extent_lock);
224 	list_add_tail(&entry->root_extent_list,
225 		      &root->ordered_extents);
226 	root->nr_ordered_extents++;
227 	if (root->nr_ordered_extents == 1) {
228 		spin_lock(&fs_info->ordered_root_lock);
229 		BUG_ON(!list_empty(&root->ordered_root));
230 		list_add_tail(&root->ordered_root, &fs_info->ordered_roots);
231 		spin_unlock(&fs_info->ordered_root_lock);
232 	}
233 	spin_unlock(&root->ordered_extent_lock);
234 
235 	/*
236 	 * We don't need the count_max_extents here, we can assume that all of
237 	 * that work has been done at higher layers, so this is truly the
238 	 * smallest the extent is going to get.
239 	 */
240 	spin_lock(&BTRFS_I(inode)->lock);
241 	btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
242 	spin_unlock(&BTRFS_I(inode)->lock);
243 
244 	return 0;
245 }
246 
247 int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
248 			     u64 start, u64 len, u64 disk_len, int type)
249 {
250 	return __btrfs_add_ordered_extent(inode, file_offset, start, len,
251 					  disk_len, type, 0,
252 					  BTRFS_COMPRESS_NONE);
253 }
254 
255 int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset,
256 				 u64 start, u64 len, u64 disk_len, int type)
257 {
258 	return __btrfs_add_ordered_extent(inode, file_offset, start, len,
259 					  disk_len, type, 1,
260 					  BTRFS_COMPRESS_NONE);
261 }
262 
263 int btrfs_add_ordered_extent_compress(struct inode *inode, u64 file_offset,
264 				      u64 start, u64 len, u64 disk_len,
265 				      int type, int compress_type)
266 {
267 	return __btrfs_add_ordered_extent(inode, file_offset, start, len,
268 					  disk_len, type, 0,
269 					  compress_type);
270 }
271 
272 /*
273  * Add a struct btrfs_ordered_sum into the list of checksums to be inserted
274  * when an ordered extent is finished.  If the list covers more than one
275  * ordered extent, it is split across multiples.
276  */
277 void btrfs_add_ordered_sum(struct btrfs_ordered_extent *entry,
278 			   struct btrfs_ordered_sum *sum)
279 {
280 	struct btrfs_ordered_inode_tree *tree;
281 
282 	tree = &BTRFS_I(entry->inode)->ordered_tree;
283 	spin_lock_irq(&tree->lock);
284 	list_add_tail(&sum->list, &entry->list);
285 	spin_unlock_irq(&tree->lock);
286 }
287 
288 /*
289  * this is used to account for finished IO across a given range
290  * of the file.  The IO may span ordered extents.  If
291  * a given ordered_extent is completely done, 1 is returned, otherwise
292  * 0.
293  *
294  * test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
295  * to make sure this function only returns 1 once for a given ordered extent.
296  *
297  * file_offset is updated to one byte past the range that is recorded as
298  * complete.  This allows you to walk forward in the file.
299  */
300 int btrfs_dec_test_first_ordered_pending(struct inode *inode,
301 				   struct btrfs_ordered_extent **cached,
302 				   u64 *file_offset, u64 io_size, int uptodate)
303 {
304 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
305 	struct btrfs_ordered_inode_tree *tree;
306 	struct rb_node *node;
307 	struct btrfs_ordered_extent *entry = NULL;
308 	int ret;
309 	unsigned long flags;
310 	u64 dec_end;
311 	u64 dec_start;
312 	u64 to_dec;
313 
314 	tree = &BTRFS_I(inode)->ordered_tree;
315 	spin_lock_irqsave(&tree->lock, flags);
316 	node = tree_search(tree, *file_offset);
317 	if (!node) {
318 		ret = 1;
319 		goto out;
320 	}
321 
322 	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
323 	if (!offset_in_entry(entry, *file_offset)) {
324 		ret = 1;
325 		goto out;
326 	}
327 
328 	dec_start = max(*file_offset, entry->file_offset);
329 	dec_end = min(*file_offset + io_size, entry->file_offset +
330 		      entry->len);
331 	*file_offset = dec_end;
332 	if (dec_start > dec_end) {
333 		btrfs_crit(fs_info, "bad ordering dec_start %llu end %llu",
334 			   dec_start, dec_end);
335 	}
336 	to_dec = dec_end - dec_start;
337 	if (to_dec > entry->bytes_left) {
338 		btrfs_crit(fs_info,
339 			   "bad ordered accounting left %llu size %llu",
340 			   entry->bytes_left, to_dec);
341 	}
342 	entry->bytes_left -= to_dec;
343 	if (!uptodate)
344 		set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
345 
346 	if (entry->bytes_left == 0) {
347 		ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
348 		/* test_and_set_bit implies a barrier */
349 		cond_wake_up_nomb(&entry->wait);
350 	} else {
351 		ret = 1;
352 	}
353 out:
354 	if (!ret && cached && entry) {
355 		*cached = entry;
356 		refcount_inc(&entry->refs);
357 	}
358 	spin_unlock_irqrestore(&tree->lock, flags);
359 	return ret == 0;
360 }
361 
362 /*
363  * this is used to account for finished IO across a given range
364  * of the file.  The IO should not span ordered extents.  If
365  * a given ordered_extent is completely done, 1 is returned, otherwise
366  * 0.
367  *
368  * test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
369  * to make sure this function only returns 1 once for a given ordered extent.
370  */
371 int btrfs_dec_test_ordered_pending(struct inode *inode,
372 				   struct btrfs_ordered_extent **cached,
373 				   u64 file_offset, u64 io_size, int uptodate)
374 {
375 	struct btrfs_ordered_inode_tree *tree;
376 	struct rb_node *node;
377 	struct btrfs_ordered_extent *entry = NULL;
378 	unsigned long flags;
379 	int ret;
380 
381 	tree = &BTRFS_I(inode)->ordered_tree;
382 	spin_lock_irqsave(&tree->lock, flags);
383 	if (cached && *cached) {
384 		entry = *cached;
385 		goto have_entry;
386 	}
387 
388 	node = tree_search(tree, file_offset);
389 	if (!node) {
390 		ret = 1;
391 		goto out;
392 	}
393 
394 	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
395 have_entry:
396 	if (!offset_in_entry(entry, file_offset)) {
397 		ret = 1;
398 		goto out;
399 	}
400 
401 	if (io_size > entry->bytes_left) {
402 		btrfs_crit(BTRFS_I(inode)->root->fs_info,
403 			   "bad ordered accounting left %llu size %llu",
404 		       entry->bytes_left, io_size);
405 	}
406 	entry->bytes_left -= io_size;
407 	if (!uptodate)
408 		set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
409 
410 	if (entry->bytes_left == 0) {
411 		ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
412 		/* test_and_set_bit implies a barrier */
413 		cond_wake_up_nomb(&entry->wait);
414 	} else {
415 		ret = 1;
416 	}
417 out:
418 	if (!ret && cached && entry) {
419 		*cached = entry;
420 		refcount_inc(&entry->refs);
421 	}
422 	spin_unlock_irqrestore(&tree->lock, flags);
423 	return ret == 0;
424 }
425 
426 /*
427  * used to drop a reference on an ordered extent.  This will free
428  * the extent if the last reference is dropped
429  */
430 void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry)
431 {
432 	struct list_head *cur;
433 	struct btrfs_ordered_sum *sum;
434 
435 	trace_btrfs_ordered_extent_put(entry->inode, entry);
436 
437 	if (refcount_dec_and_test(&entry->refs)) {
438 		ASSERT(list_empty(&entry->log_list));
439 		ASSERT(list_empty(&entry->trans_list));
440 		ASSERT(list_empty(&entry->root_extent_list));
441 		ASSERT(RB_EMPTY_NODE(&entry->rb_node));
442 		if (entry->inode)
443 			btrfs_add_delayed_iput(entry->inode);
444 		while (!list_empty(&entry->list)) {
445 			cur = entry->list.next;
446 			sum = list_entry(cur, struct btrfs_ordered_sum, list);
447 			list_del(&sum->list);
448 			kvfree(sum);
449 		}
450 		kmem_cache_free(btrfs_ordered_extent_cache, entry);
451 	}
452 }
453 
454 /*
455  * remove an ordered extent from the tree.  No references are dropped
456  * and waiters are woken up.
457  */
458 void btrfs_remove_ordered_extent(struct inode *inode,
459 				 struct btrfs_ordered_extent *entry)
460 {
461 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
462 	struct btrfs_ordered_inode_tree *tree;
463 	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
464 	struct btrfs_root *root = btrfs_inode->root;
465 	struct rb_node *node;
466 
467 	/* This is paired with btrfs_add_ordered_extent. */
468 	spin_lock(&btrfs_inode->lock);
469 	btrfs_mod_outstanding_extents(btrfs_inode, -1);
470 	spin_unlock(&btrfs_inode->lock);
471 	if (root != fs_info->tree_root)
472 		btrfs_delalloc_release_metadata(btrfs_inode, entry->len, false);
473 
474 	if (test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
475 		percpu_counter_add_batch(&fs_info->dio_bytes, -entry->len,
476 					 fs_info->delalloc_batch);
477 
478 	tree = &btrfs_inode->ordered_tree;
479 	spin_lock_irq(&tree->lock);
480 	node = &entry->rb_node;
481 	rb_erase(node, &tree->tree);
482 	RB_CLEAR_NODE(node);
483 	if (tree->last == node)
484 		tree->last = NULL;
485 	set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
486 	spin_unlock_irq(&tree->lock);
487 
488 	spin_lock(&root->ordered_extent_lock);
489 	list_del_init(&entry->root_extent_list);
490 	root->nr_ordered_extents--;
491 
492 	trace_btrfs_ordered_extent_remove(inode, entry);
493 
494 	if (!root->nr_ordered_extents) {
495 		spin_lock(&fs_info->ordered_root_lock);
496 		BUG_ON(list_empty(&root->ordered_root));
497 		list_del_init(&root->ordered_root);
498 		spin_unlock(&fs_info->ordered_root_lock);
499 	}
500 	spin_unlock(&root->ordered_extent_lock);
501 	wake_up(&entry->wait);
502 }
503 
504 static void btrfs_run_ordered_extent_work(struct btrfs_work *work)
505 {
506 	struct btrfs_ordered_extent *ordered;
507 
508 	ordered = container_of(work, struct btrfs_ordered_extent, flush_work);
509 	btrfs_start_ordered_extent(ordered->inode, ordered, 1);
510 	complete(&ordered->completion);
511 }
512 
513 /*
514  * wait for all the ordered extents in a root.  This is done when balancing
515  * space between drives.
516  */
517 u64 btrfs_wait_ordered_extents(struct btrfs_root *root, u64 nr,
518 			       const u64 range_start, const u64 range_len)
519 {
520 	struct btrfs_fs_info *fs_info = root->fs_info;
521 	LIST_HEAD(splice);
522 	LIST_HEAD(skipped);
523 	LIST_HEAD(works);
524 	struct btrfs_ordered_extent *ordered, *next;
525 	u64 count = 0;
526 	const u64 range_end = range_start + range_len;
527 
528 	mutex_lock(&root->ordered_extent_mutex);
529 	spin_lock(&root->ordered_extent_lock);
530 	list_splice_init(&root->ordered_extents, &splice);
531 	while (!list_empty(&splice) && nr) {
532 		ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
533 					   root_extent_list);
534 
535 		if (range_end <= ordered->start ||
536 		    ordered->start + ordered->disk_len <= range_start) {
537 			list_move_tail(&ordered->root_extent_list, &skipped);
538 			cond_resched_lock(&root->ordered_extent_lock);
539 			continue;
540 		}
541 
542 		list_move_tail(&ordered->root_extent_list,
543 			       &root->ordered_extents);
544 		refcount_inc(&ordered->refs);
545 		spin_unlock(&root->ordered_extent_lock);
546 
547 		btrfs_init_work(&ordered->flush_work,
548 				btrfs_flush_delalloc_helper,
549 				btrfs_run_ordered_extent_work, NULL, NULL);
550 		list_add_tail(&ordered->work_list, &works);
551 		btrfs_queue_work(fs_info->flush_workers, &ordered->flush_work);
552 
553 		cond_resched();
554 		spin_lock(&root->ordered_extent_lock);
555 		if (nr != U64_MAX)
556 			nr--;
557 		count++;
558 	}
559 	list_splice_tail(&skipped, &root->ordered_extents);
560 	list_splice_tail(&splice, &root->ordered_extents);
561 	spin_unlock(&root->ordered_extent_lock);
562 
563 	list_for_each_entry_safe(ordered, next, &works, work_list) {
564 		list_del_init(&ordered->work_list);
565 		wait_for_completion(&ordered->completion);
566 		btrfs_put_ordered_extent(ordered);
567 		cond_resched();
568 	}
569 	mutex_unlock(&root->ordered_extent_mutex);
570 
571 	return count;
572 }
573 
574 u64 btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, u64 nr,
575 			     const u64 range_start, const u64 range_len)
576 {
577 	struct btrfs_root *root;
578 	struct list_head splice;
579 	u64 total_done = 0;
580 	u64 done;
581 
582 	INIT_LIST_HEAD(&splice);
583 
584 	mutex_lock(&fs_info->ordered_operations_mutex);
585 	spin_lock(&fs_info->ordered_root_lock);
586 	list_splice_init(&fs_info->ordered_roots, &splice);
587 	while (!list_empty(&splice) && nr) {
588 		root = list_first_entry(&splice, struct btrfs_root,
589 					ordered_root);
590 		root = btrfs_grab_fs_root(root);
591 		BUG_ON(!root);
592 		list_move_tail(&root->ordered_root,
593 			       &fs_info->ordered_roots);
594 		spin_unlock(&fs_info->ordered_root_lock);
595 
596 		done = btrfs_wait_ordered_extents(root, nr,
597 						  range_start, range_len);
598 		btrfs_put_fs_root(root);
599 		total_done += done;
600 
601 		spin_lock(&fs_info->ordered_root_lock);
602 		if (nr != U64_MAX) {
603 			nr -= done;
604 		}
605 	}
606 	list_splice_tail(&splice, &fs_info->ordered_roots);
607 	spin_unlock(&fs_info->ordered_root_lock);
608 	mutex_unlock(&fs_info->ordered_operations_mutex);
609 
610 	return total_done;
611 }
612 
613 /*
614  * Used to start IO or wait for a given ordered extent to finish.
615  *
616  * If wait is one, this effectively waits on page writeback for all the pages
617  * in the extent, and it waits on the io completion code to insert
618  * metadata into the btree corresponding to the extent
619  */
620 void btrfs_start_ordered_extent(struct inode *inode,
621 				       struct btrfs_ordered_extent *entry,
622 				       int wait)
623 {
624 	u64 start = entry->file_offset;
625 	u64 end = start + entry->len - 1;
626 
627 	trace_btrfs_ordered_extent_start(inode, entry);
628 
629 	/*
630 	 * pages in the range can be dirty, clean or writeback.  We
631 	 * start IO on any dirty ones so the wait doesn't stall waiting
632 	 * for the flusher thread to find them
633 	 */
634 	if (!test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
635 		filemap_fdatawrite_range(inode->i_mapping, start, end);
636 	if (wait) {
637 		wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE,
638 						 &entry->flags));
639 	}
640 }
641 
642 /*
643  * Used to wait on ordered extents across a large range of bytes.
644  */
645 int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
646 {
647 	int ret = 0;
648 	int ret_wb = 0;
649 	u64 end;
650 	u64 orig_end;
651 	struct btrfs_ordered_extent *ordered;
652 
653 	if (start + len < start) {
654 		orig_end = INT_LIMIT(loff_t);
655 	} else {
656 		orig_end = start + len - 1;
657 		if (orig_end > INT_LIMIT(loff_t))
658 			orig_end = INT_LIMIT(loff_t);
659 	}
660 
661 	/* start IO across the range first to instantiate any delalloc
662 	 * extents
663 	 */
664 	ret = btrfs_fdatawrite_range(inode, start, orig_end);
665 	if (ret)
666 		return ret;
667 
668 	/*
669 	 * If we have a writeback error don't return immediately. Wait first
670 	 * for any ordered extents that haven't completed yet. This is to make
671 	 * sure no one can dirty the same page ranges and call writepages()
672 	 * before the ordered extents complete - to avoid failures (-EEXIST)
673 	 * when adding the new ordered extents to the ordered tree.
674 	 */
675 	ret_wb = filemap_fdatawait_range(inode->i_mapping, start, orig_end);
676 
677 	end = orig_end;
678 	while (1) {
679 		ordered = btrfs_lookup_first_ordered_extent(inode, end);
680 		if (!ordered)
681 			break;
682 		if (ordered->file_offset > orig_end) {
683 			btrfs_put_ordered_extent(ordered);
684 			break;
685 		}
686 		if (ordered->file_offset + ordered->len <= start) {
687 			btrfs_put_ordered_extent(ordered);
688 			break;
689 		}
690 		btrfs_start_ordered_extent(inode, ordered, 1);
691 		end = ordered->file_offset;
692 		if (test_bit(BTRFS_ORDERED_IOERR, &ordered->flags))
693 			ret = -EIO;
694 		btrfs_put_ordered_extent(ordered);
695 		if (ret || end == 0 || end == start)
696 			break;
697 		end--;
698 	}
699 	return ret_wb ? ret_wb : ret;
700 }
701 
702 /*
703  * find an ordered extent corresponding to file_offset.  return NULL if
704  * nothing is found, otherwise take a reference on the extent and return it
705  */
706 struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
707 							 u64 file_offset)
708 {
709 	struct btrfs_ordered_inode_tree *tree;
710 	struct rb_node *node;
711 	struct btrfs_ordered_extent *entry = NULL;
712 
713 	tree = &BTRFS_I(inode)->ordered_tree;
714 	spin_lock_irq(&tree->lock);
715 	node = tree_search(tree, file_offset);
716 	if (!node)
717 		goto out;
718 
719 	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
720 	if (!offset_in_entry(entry, file_offset))
721 		entry = NULL;
722 	if (entry)
723 		refcount_inc(&entry->refs);
724 out:
725 	spin_unlock_irq(&tree->lock);
726 	return entry;
727 }
728 
729 /* Since the DIO code tries to lock a wide area we need to look for any ordered
730  * extents that exist in the range, rather than just the start of the range.
731  */
732 struct btrfs_ordered_extent *btrfs_lookup_ordered_range(
733 		struct btrfs_inode *inode, u64 file_offset, u64 len)
734 {
735 	struct btrfs_ordered_inode_tree *tree;
736 	struct rb_node *node;
737 	struct btrfs_ordered_extent *entry = NULL;
738 
739 	tree = &inode->ordered_tree;
740 	spin_lock_irq(&tree->lock);
741 	node = tree_search(tree, file_offset);
742 	if (!node) {
743 		node = tree_search(tree, file_offset + len);
744 		if (!node)
745 			goto out;
746 	}
747 
748 	while (1) {
749 		entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
750 		if (range_overlaps(entry, file_offset, len))
751 			break;
752 
753 		if (entry->file_offset >= file_offset + len) {
754 			entry = NULL;
755 			break;
756 		}
757 		entry = NULL;
758 		node = rb_next(node);
759 		if (!node)
760 			break;
761 	}
762 out:
763 	if (entry)
764 		refcount_inc(&entry->refs);
765 	spin_unlock_irq(&tree->lock);
766 	return entry;
767 }
768 
769 /*
770  * lookup and return any extent before 'file_offset'.  NULL is returned
771  * if none is found
772  */
773 struct btrfs_ordered_extent *
774 btrfs_lookup_first_ordered_extent(struct inode *inode, u64 file_offset)
775 {
776 	struct btrfs_ordered_inode_tree *tree;
777 	struct rb_node *node;
778 	struct btrfs_ordered_extent *entry = NULL;
779 
780 	tree = &BTRFS_I(inode)->ordered_tree;
781 	spin_lock_irq(&tree->lock);
782 	node = tree_search(tree, file_offset);
783 	if (!node)
784 		goto out;
785 
786 	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
787 	refcount_inc(&entry->refs);
788 out:
789 	spin_unlock_irq(&tree->lock);
790 	return entry;
791 }
792 
793 /*
794  * After an extent is done, call this to conditionally update the on disk
795  * i_size.  i_size is updated to cover any fully written part of the file.
796  */
797 int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
798 				struct btrfs_ordered_extent *ordered)
799 {
800 	struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
801 	u64 disk_i_size;
802 	u64 new_i_size;
803 	u64 i_size = i_size_read(inode);
804 	struct rb_node *node;
805 	struct rb_node *prev = NULL;
806 	struct btrfs_ordered_extent *test;
807 	int ret = 1;
808 	u64 orig_offset = offset;
809 
810 	spin_lock_irq(&tree->lock);
811 	if (ordered) {
812 		offset = entry_end(ordered);
813 		if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags))
814 			offset = min(offset,
815 				     ordered->file_offset +
816 				     ordered->truncated_len);
817 	} else {
818 		offset = ALIGN(offset, btrfs_inode_sectorsize(inode));
819 	}
820 	disk_i_size = BTRFS_I(inode)->disk_i_size;
821 
822 	/*
823 	 * truncate file.
824 	 * If ordered is not NULL, then this is called from endio and
825 	 * disk_i_size will be updated by either truncate itself or any
826 	 * in-flight IOs which are inside the disk_i_size.
827 	 *
828 	 * Because btrfs_setsize() may set i_size with disk_i_size if truncate
829 	 * fails somehow, we need to make sure we have a precise disk_i_size by
830 	 * updating it as usual.
831 	 *
832 	 */
833 	if (!ordered && disk_i_size > i_size) {
834 		BTRFS_I(inode)->disk_i_size = orig_offset;
835 		ret = 0;
836 		goto out;
837 	}
838 
839 	/*
840 	 * if the disk i_size is already at the inode->i_size, or
841 	 * this ordered extent is inside the disk i_size, we're done
842 	 */
843 	if (disk_i_size == i_size)
844 		goto out;
845 
846 	/*
847 	 * We still need to update disk_i_size if outstanding_isize is greater
848 	 * than disk_i_size.
849 	 */
850 	if (offset <= disk_i_size &&
851 	    (!ordered || ordered->outstanding_isize <= disk_i_size))
852 		goto out;
853 
854 	/*
855 	 * walk backward from this ordered extent to disk_i_size.
856 	 * if we find an ordered extent then we can't update disk i_size
857 	 * yet
858 	 */
859 	if (ordered) {
860 		node = rb_prev(&ordered->rb_node);
861 	} else {
862 		prev = tree_search(tree, offset);
863 		/*
864 		 * we insert file extents without involving ordered struct,
865 		 * so there should be no ordered struct cover this offset
866 		 */
867 		if (prev) {
868 			test = rb_entry(prev, struct btrfs_ordered_extent,
869 					rb_node);
870 			BUG_ON(offset_in_entry(test, offset));
871 		}
872 		node = prev;
873 	}
874 	for (; node; node = rb_prev(node)) {
875 		test = rb_entry(node, struct btrfs_ordered_extent, rb_node);
876 
877 		/* We treat this entry as if it doesn't exist */
878 		if (test_bit(BTRFS_ORDERED_UPDATED_ISIZE, &test->flags))
879 			continue;
880 
881 		if (entry_end(test) <= disk_i_size)
882 			break;
883 		if (test->file_offset >= i_size)
884 			break;
885 
886 		/*
887 		 * We don't update disk_i_size now, so record this undealt
888 		 * i_size. Or we will not know the real i_size.
889 		 */
890 		if (test->outstanding_isize < offset)
891 			test->outstanding_isize = offset;
892 		if (ordered &&
893 		    ordered->outstanding_isize > test->outstanding_isize)
894 			test->outstanding_isize = ordered->outstanding_isize;
895 		goto out;
896 	}
897 	new_i_size = min_t(u64, offset, i_size);
898 
899 	/*
900 	 * Some ordered extents may completed before the current one, and
901 	 * we hold the real i_size in ->outstanding_isize.
902 	 */
903 	if (ordered && ordered->outstanding_isize > new_i_size)
904 		new_i_size = min_t(u64, ordered->outstanding_isize, i_size);
905 	BTRFS_I(inode)->disk_i_size = new_i_size;
906 	ret = 0;
907 out:
908 	/*
909 	 * We need to do this because we can't remove ordered extents until
910 	 * after the i_disk_size has been updated and then the inode has been
911 	 * updated to reflect the change, so we need to tell anybody who finds
912 	 * this ordered extent that we've already done all the real work, we
913 	 * just haven't completed all the other work.
914 	 */
915 	if (ordered)
916 		set_bit(BTRFS_ORDERED_UPDATED_ISIZE, &ordered->flags);
917 	spin_unlock_irq(&tree->lock);
918 	return ret;
919 }
920 
921 /*
922  * search the ordered extents for one corresponding to 'offset' and
923  * try to find a checksum.  This is used because we allow pages to
924  * be reclaimed before their checksum is actually put into the btree
925  */
926 int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
927 			   u32 *sum, int len)
928 {
929 	struct btrfs_ordered_sum *ordered_sum;
930 	struct btrfs_ordered_extent *ordered;
931 	struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
932 	unsigned long num_sectors;
933 	unsigned long i;
934 	u32 sectorsize = btrfs_inode_sectorsize(inode);
935 	int index = 0;
936 
937 	ordered = btrfs_lookup_ordered_extent(inode, offset);
938 	if (!ordered)
939 		return 0;
940 
941 	spin_lock_irq(&tree->lock);
942 	list_for_each_entry_reverse(ordered_sum, &ordered->list, list) {
943 		if (disk_bytenr >= ordered_sum->bytenr &&
944 		    disk_bytenr < ordered_sum->bytenr + ordered_sum->len) {
945 			i = (disk_bytenr - ordered_sum->bytenr) >>
946 			    inode->i_sb->s_blocksize_bits;
947 			num_sectors = ordered_sum->len >>
948 				      inode->i_sb->s_blocksize_bits;
949 			num_sectors = min_t(int, len - index, num_sectors - i);
950 			memcpy(sum + index, ordered_sum->sums + i,
951 			       num_sectors);
952 
953 			index += (int)num_sectors;
954 			if (index == len)
955 				goto out;
956 			disk_bytenr += num_sectors * sectorsize;
957 		}
958 	}
959 out:
960 	spin_unlock_irq(&tree->lock);
961 	btrfs_put_ordered_extent(ordered);
962 	return index;
963 }
964 
965 int __init ordered_data_init(void)
966 {
967 	btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent",
968 				     sizeof(struct btrfs_ordered_extent), 0,
969 				     SLAB_MEM_SPREAD,
970 				     NULL);
971 	if (!btrfs_ordered_extent_cache)
972 		return -ENOMEM;
973 
974 	return 0;
975 }
976 
977 void __cold ordered_data_exit(void)
978 {
979 	kmem_cache_destroy(btrfs_ordered_extent_cache);
980 }
981