xref: /openbmc/linux/fs/btrfs/delayed-inode.c (revision 8795a739)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2011 Fujitsu.  All rights reserved.
4  * Written by Miao Xie <miaox@cn.fujitsu.com>
5  */
6 
7 #include <linux/slab.h>
8 #include <linux/iversion.h>
9 #include "misc.h"
10 #include "delayed-inode.h"
11 #include "disk-io.h"
12 #include "transaction.h"
13 #include "ctree.h"
14 #include "qgroup.h"
15 
16 #define BTRFS_DELAYED_WRITEBACK		512
17 #define BTRFS_DELAYED_BACKGROUND	128
18 #define BTRFS_DELAYED_BATCH		16
19 
20 static struct kmem_cache *delayed_node_cache;
21 
22 int __init btrfs_delayed_inode_init(void)
23 {
24 	delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
25 					sizeof(struct btrfs_delayed_node),
26 					0,
27 					SLAB_MEM_SPREAD,
28 					NULL);
29 	if (!delayed_node_cache)
30 		return -ENOMEM;
31 	return 0;
32 }
33 
34 void __cold btrfs_delayed_inode_exit(void)
35 {
36 	kmem_cache_destroy(delayed_node_cache);
37 }
38 
39 static inline void btrfs_init_delayed_node(
40 				struct btrfs_delayed_node *delayed_node,
41 				struct btrfs_root *root, u64 inode_id)
42 {
43 	delayed_node->root = root;
44 	delayed_node->inode_id = inode_id;
45 	refcount_set(&delayed_node->refs, 0);
46 	delayed_node->ins_root = RB_ROOT_CACHED;
47 	delayed_node->del_root = RB_ROOT_CACHED;
48 	mutex_init(&delayed_node->mutex);
49 	INIT_LIST_HEAD(&delayed_node->n_list);
50 	INIT_LIST_HEAD(&delayed_node->p_list);
51 }
52 
53 static inline int btrfs_is_continuous_delayed_item(
54 					struct btrfs_delayed_item *item1,
55 					struct btrfs_delayed_item *item2)
56 {
57 	if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
58 	    item1->key.objectid == item2->key.objectid &&
59 	    item1->key.type == item2->key.type &&
60 	    item1->key.offset + 1 == item2->key.offset)
61 		return 1;
62 	return 0;
63 }
64 
65 static struct btrfs_delayed_node *btrfs_get_delayed_node(
66 		struct btrfs_inode *btrfs_inode)
67 {
68 	struct btrfs_root *root = btrfs_inode->root;
69 	u64 ino = btrfs_ino(btrfs_inode);
70 	struct btrfs_delayed_node *node;
71 
72 	node = READ_ONCE(btrfs_inode->delayed_node);
73 	if (node) {
74 		refcount_inc(&node->refs);
75 		return node;
76 	}
77 
78 	spin_lock(&root->inode_lock);
79 	node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
80 
81 	if (node) {
82 		if (btrfs_inode->delayed_node) {
83 			refcount_inc(&node->refs);	/* can be accessed */
84 			BUG_ON(btrfs_inode->delayed_node != node);
85 			spin_unlock(&root->inode_lock);
86 			return node;
87 		}
88 
89 		/*
90 		 * It's possible that we're racing into the middle of removing
91 		 * this node from the radix tree.  In this case, the refcount
92 		 * was zero and it should never go back to one.  Just return
93 		 * NULL like it was never in the radix at all; our release
94 		 * function is in the process of removing it.
95 		 *
96 		 * Some implementations of refcount_inc refuse to bump the
97 		 * refcount once it has hit zero.  If we don't do this dance
98 		 * here, refcount_inc() may decide to just WARN_ONCE() instead
99 		 * of actually bumping the refcount.
100 		 *
101 		 * If this node is properly in the radix, we want to bump the
102 		 * refcount twice, once for the inode and once for this get
103 		 * operation.
104 		 */
105 		if (refcount_inc_not_zero(&node->refs)) {
106 			refcount_inc(&node->refs);
107 			btrfs_inode->delayed_node = node;
108 		} else {
109 			node = NULL;
110 		}
111 
112 		spin_unlock(&root->inode_lock);
113 		return node;
114 	}
115 	spin_unlock(&root->inode_lock);
116 
117 	return NULL;
118 }
119 
120 /* Will return either the node or PTR_ERR(-ENOMEM) */
121 static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
122 		struct btrfs_inode *btrfs_inode)
123 {
124 	struct btrfs_delayed_node *node;
125 	struct btrfs_root *root = btrfs_inode->root;
126 	u64 ino = btrfs_ino(btrfs_inode);
127 	int ret;
128 
129 again:
130 	node = btrfs_get_delayed_node(btrfs_inode);
131 	if (node)
132 		return node;
133 
134 	node = kmem_cache_zalloc(delayed_node_cache, GFP_NOFS);
135 	if (!node)
136 		return ERR_PTR(-ENOMEM);
137 	btrfs_init_delayed_node(node, root, ino);
138 
139 	/* cached in the btrfs inode and can be accessed */
140 	refcount_set(&node->refs, 2);
141 
142 	ret = radix_tree_preload(GFP_NOFS);
143 	if (ret) {
144 		kmem_cache_free(delayed_node_cache, node);
145 		return ERR_PTR(ret);
146 	}
147 
148 	spin_lock(&root->inode_lock);
149 	ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
150 	if (ret == -EEXIST) {
151 		spin_unlock(&root->inode_lock);
152 		kmem_cache_free(delayed_node_cache, node);
153 		radix_tree_preload_end();
154 		goto again;
155 	}
156 	btrfs_inode->delayed_node = node;
157 	spin_unlock(&root->inode_lock);
158 	radix_tree_preload_end();
159 
160 	return node;
161 }
162 
163 /*
164  * Call it when holding delayed_node->mutex
165  *
166  * If mod = 1, add this node into the prepared list.
167  */
168 static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
169 				     struct btrfs_delayed_node *node,
170 				     int mod)
171 {
172 	spin_lock(&root->lock);
173 	if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
174 		if (!list_empty(&node->p_list))
175 			list_move_tail(&node->p_list, &root->prepare_list);
176 		else if (mod)
177 			list_add_tail(&node->p_list, &root->prepare_list);
178 	} else {
179 		list_add_tail(&node->n_list, &root->node_list);
180 		list_add_tail(&node->p_list, &root->prepare_list);
181 		refcount_inc(&node->refs);	/* inserted into list */
182 		root->nodes++;
183 		set_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
184 	}
185 	spin_unlock(&root->lock);
186 }
187 
188 /* Call it when holding delayed_node->mutex */
189 static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
190 				       struct btrfs_delayed_node *node)
191 {
192 	spin_lock(&root->lock);
193 	if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
194 		root->nodes--;
195 		refcount_dec(&node->refs);	/* not in the list */
196 		list_del_init(&node->n_list);
197 		if (!list_empty(&node->p_list))
198 			list_del_init(&node->p_list);
199 		clear_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
200 	}
201 	spin_unlock(&root->lock);
202 }
203 
204 static struct btrfs_delayed_node *btrfs_first_delayed_node(
205 			struct btrfs_delayed_root *delayed_root)
206 {
207 	struct list_head *p;
208 	struct btrfs_delayed_node *node = NULL;
209 
210 	spin_lock(&delayed_root->lock);
211 	if (list_empty(&delayed_root->node_list))
212 		goto out;
213 
214 	p = delayed_root->node_list.next;
215 	node = list_entry(p, struct btrfs_delayed_node, n_list);
216 	refcount_inc(&node->refs);
217 out:
218 	spin_unlock(&delayed_root->lock);
219 
220 	return node;
221 }
222 
223 static struct btrfs_delayed_node *btrfs_next_delayed_node(
224 						struct btrfs_delayed_node *node)
225 {
226 	struct btrfs_delayed_root *delayed_root;
227 	struct list_head *p;
228 	struct btrfs_delayed_node *next = NULL;
229 
230 	delayed_root = node->root->fs_info->delayed_root;
231 	spin_lock(&delayed_root->lock);
232 	if (!test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
233 		/* not in the list */
234 		if (list_empty(&delayed_root->node_list))
235 			goto out;
236 		p = delayed_root->node_list.next;
237 	} else if (list_is_last(&node->n_list, &delayed_root->node_list))
238 		goto out;
239 	else
240 		p = node->n_list.next;
241 
242 	next = list_entry(p, struct btrfs_delayed_node, n_list);
243 	refcount_inc(&next->refs);
244 out:
245 	spin_unlock(&delayed_root->lock);
246 
247 	return next;
248 }
249 
250 static void __btrfs_release_delayed_node(
251 				struct btrfs_delayed_node *delayed_node,
252 				int mod)
253 {
254 	struct btrfs_delayed_root *delayed_root;
255 
256 	if (!delayed_node)
257 		return;
258 
259 	delayed_root = delayed_node->root->fs_info->delayed_root;
260 
261 	mutex_lock(&delayed_node->mutex);
262 	if (delayed_node->count)
263 		btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
264 	else
265 		btrfs_dequeue_delayed_node(delayed_root, delayed_node);
266 	mutex_unlock(&delayed_node->mutex);
267 
268 	if (refcount_dec_and_test(&delayed_node->refs)) {
269 		struct btrfs_root *root = delayed_node->root;
270 
271 		spin_lock(&root->inode_lock);
272 		/*
273 		 * Once our refcount goes to zero, nobody is allowed to bump it
274 		 * back up.  We can delete it now.
275 		 */
276 		ASSERT(refcount_read(&delayed_node->refs) == 0);
277 		radix_tree_delete(&root->delayed_nodes_tree,
278 				  delayed_node->inode_id);
279 		spin_unlock(&root->inode_lock);
280 		kmem_cache_free(delayed_node_cache, delayed_node);
281 	}
282 }
283 
284 static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
285 {
286 	__btrfs_release_delayed_node(node, 0);
287 }
288 
289 static struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
290 					struct btrfs_delayed_root *delayed_root)
291 {
292 	struct list_head *p;
293 	struct btrfs_delayed_node *node = NULL;
294 
295 	spin_lock(&delayed_root->lock);
296 	if (list_empty(&delayed_root->prepare_list))
297 		goto out;
298 
299 	p = delayed_root->prepare_list.next;
300 	list_del_init(p);
301 	node = list_entry(p, struct btrfs_delayed_node, p_list);
302 	refcount_inc(&node->refs);
303 out:
304 	spin_unlock(&delayed_root->lock);
305 
306 	return node;
307 }
308 
309 static inline void btrfs_release_prepared_delayed_node(
310 					struct btrfs_delayed_node *node)
311 {
312 	__btrfs_release_delayed_node(node, 1);
313 }
314 
315 static struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
316 {
317 	struct btrfs_delayed_item *item;
318 	item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
319 	if (item) {
320 		item->data_len = data_len;
321 		item->ins_or_del = 0;
322 		item->bytes_reserved = 0;
323 		item->delayed_node = NULL;
324 		refcount_set(&item->refs, 1);
325 	}
326 	return item;
327 }
328 
329 /*
330  * __btrfs_lookup_delayed_item - look up the delayed item by key
331  * @delayed_node: pointer to the delayed node
332  * @key:	  the key to look up
333  * @prev:	  used to store the prev item if the right item isn't found
334  * @next:	  used to store the next item if the right item isn't found
335  *
336  * Note: if we don't find the right item, we will return the prev item and
337  * the next item.
338  */
339 static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
340 				struct rb_root *root,
341 				struct btrfs_key *key,
342 				struct btrfs_delayed_item **prev,
343 				struct btrfs_delayed_item **next)
344 {
345 	struct rb_node *node, *prev_node = NULL;
346 	struct btrfs_delayed_item *delayed_item = NULL;
347 	int ret = 0;
348 
349 	node = root->rb_node;
350 
351 	while (node) {
352 		delayed_item = rb_entry(node, struct btrfs_delayed_item,
353 					rb_node);
354 		prev_node = node;
355 		ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
356 		if (ret < 0)
357 			node = node->rb_right;
358 		else if (ret > 0)
359 			node = node->rb_left;
360 		else
361 			return delayed_item;
362 	}
363 
364 	if (prev) {
365 		if (!prev_node)
366 			*prev = NULL;
367 		else if (ret < 0)
368 			*prev = delayed_item;
369 		else if ((node = rb_prev(prev_node)) != NULL) {
370 			*prev = rb_entry(node, struct btrfs_delayed_item,
371 					 rb_node);
372 		} else
373 			*prev = NULL;
374 	}
375 
376 	if (next) {
377 		if (!prev_node)
378 			*next = NULL;
379 		else if (ret > 0)
380 			*next = delayed_item;
381 		else if ((node = rb_next(prev_node)) != NULL) {
382 			*next = rb_entry(node, struct btrfs_delayed_item,
383 					 rb_node);
384 		} else
385 			*next = NULL;
386 	}
387 	return NULL;
388 }
389 
390 static struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
391 					struct btrfs_delayed_node *delayed_node,
392 					struct btrfs_key *key)
393 {
394 	return __btrfs_lookup_delayed_item(&delayed_node->ins_root.rb_root, key,
395 					   NULL, NULL);
396 }
397 
398 static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
399 				    struct btrfs_delayed_item *ins,
400 				    int action)
401 {
402 	struct rb_node **p, *node;
403 	struct rb_node *parent_node = NULL;
404 	struct rb_root_cached *root;
405 	struct btrfs_delayed_item *item;
406 	int cmp;
407 	bool leftmost = true;
408 
409 	if (action == BTRFS_DELAYED_INSERTION_ITEM)
410 		root = &delayed_node->ins_root;
411 	else if (action == BTRFS_DELAYED_DELETION_ITEM)
412 		root = &delayed_node->del_root;
413 	else
414 		BUG();
415 	p = &root->rb_root.rb_node;
416 	node = &ins->rb_node;
417 
418 	while (*p) {
419 		parent_node = *p;
420 		item = rb_entry(parent_node, struct btrfs_delayed_item,
421 				 rb_node);
422 
423 		cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
424 		if (cmp < 0) {
425 			p = &(*p)->rb_right;
426 			leftmost = false;
427 		} else if (cmp > 0) {
428 			p = &(*p)->rb_left;
429 		} else {
430 			return -EEXIST;
431 		}
432 	}
433 
434 	rb_link_node(node, parent_node, p);
435 	rb_insert_color_cached(node, root, leftmost);
436 	ins->delayed_node = delayed_node;
437 	ins->ins_or_del = action;
438 
439 	if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
440 	    action == BTRFS_DELAYED_INSERTION_ITEM &&
441 	    ins->key.offset >= delayed_node->index_cnt)
442 			delayed_node->index_cnt = ins->key.offset + 1;
443 
444 	delayed_node->count++;
445 	atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
446 	return 0;
447 }
448 
449 static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
450 					      struct btrfs_delayed_item *item)
451 {
452 	return __btrfs_add_delayed_item(node, item,
453 					BTRFS_DELAYED_INSERTION_ITEM);
454 }
455 
456 static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
457 					     struct btrfs_delayed_item *item)
458 {
459 	return __btrfs_add_delayed_item(node, item,
460 					BTRFS_DELAYED_DELETION_ITEM);
461 }
462 
463 static void finish_one_item(struct btrfs_delayed_root *delayed_root)
464 {
465 	int seq = atomic_inc_return(&delayed_root->items_seq);
466 
467 	/* atomic_dec_return implies a barrier */
468 	if ((atomic_dec_return(&delayed_root->items) <
469 	    BTRFS_DELAYED_BACKGROUND || seq % BTRFS_DELAYED_BATCH == 0))
470 		cond_wake_up_nomb(&delayed_root->wait);
471 }
472 
473 static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
474 {
475 	struct rb_root_cached *root;
476 	struct btrfs_delayed_root *delayed_root;
477 
478 	/* Not associated with any delayed_node */
479 	if (!delayed_item->delayed_node)
480 		return;
481 	delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
482 
483 	BUG_ON(!delayed_root);
484 	BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
485 	       delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
486 
487 	if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
488 		root = &delayed_item->delayed_node->ins_root;
489 	else
490 		root = &delayed_item->delayed_node->del_root;
491 
492 	rb_erase_cached(&delayed_item->rb_node, root);
493 	delayed_item->delayed_node->count--;
494 
495 	finish_one_item(delayed_root);
496 }
497 
498 static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
499 {
500 	if (item) {
501 		__btrfs_remove_delayed_item(item);
502 		if (refcount_dec_and_test(&item->refs))
503 			kfree(item);
504 	}
505 }
506 
507 static struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
508 					struct btrfs_delayed_node *delayed_node)
509 {
510 	struct rb_node *p;
511 	struct btrfs_delayed_item *item = NULL;
512 
513 	p = rb_first_cached(&delayed_node->ins_root);
514 	if (p)
515 		item = rb_entry(p, struct btrfs_delayed_item, rb_node);
516 
517 	return item;
518 }
519 
520 static struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
521 					struct btrfs_delayed_node *delayed_node)
522 {
523 	struct rb_node *p;
524 	struct btrfs_delayed_item *item = NULL;
525 
526 	p = rb_first_cached(&delayed_node->del_root);
527 	if (p)
528 		item = rb_entry(p, struct btrfs_delayed_item, rb_node);
529 
530 	return item;
531 }
532 
533 static struct btrfs_delayed_item *__btrfs_next_delayed_item(
534 						struct btrfs_delayed_item *item)
535 {
536 	struct rb_node *p;
537 	struct btrfs_delayed_item *next = NULL;
538 
539 	p = rb_next(&item->rb_node);
540 	if (p)
541 		next = rb_entry(p, struct btrfs_delayed_item, rb_node);
542 
543 	return next;
544 }
545 
546 static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
547 					       struct btrfs_root *root,
548 					       struct btrfs_delayed_item *item)
549 {
550 	struct btrfs_block_rsv *src_rsv;
551 	struct btrfs_block_rsv *dst_rsv;
552 	struct btrfs_fs_info *fs_info = root->fs_info;
553 	u64 num_bytes;
554 	int ret;
555 
556 	if (!trans->bytes_reserved)
557 		return 0;
558 
559 	src_rsv = trans->block_rsv;
560 	dst_rsv = &fs_info->delayed_block_rsv;
561 
562 	num_bytes = btrfs_calc_insert_metadata_size(fs_info, 1);
563 
564 	/*
565 	 * Here we migrate space rsv from transaction rsv, since have already
566 	 * reserved space when starting a transaction.  So no need to reserve
567 	 * qgroup space here.
568 	 */
569 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes, true);
570 	if (!ret) {
571 		trace_btrfs_space_reservation(fs_info, "delayed_item",
572 					      item->key.objectid,
573 					      num_bytes, 1);
574 		item->bytes_reserved = num_bytes;
575 	}
576 
577 	return ret;
578 }
579 
580 static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
581 						struct btrfs_delayed_item *item)
582 {
583 	struct btrfs_block_rsv *rsv;
584 	struct btrfs_fs_info *fs_info = root->fs_info;
585 
586 	if (!item->bytes_reserved)
587 		return;
588 
589 	rsv = &fs_info->delayed_block_rsv;
590 	/*
591 	 * Check btrfs_delayed_item_reserve_metadata() to see why we don't need
592 	 * to release/reserve qgroup space.
593 	 */
594 	trace_btrfs_space_reservation(fs_info, "delayed_item",
595 				      item->key.objectid, item->bytes_reserved,
596 				      0);
597 	btrfs_block_rsv_release(fs_info, rsv,
598 				item->bytes_reserved);
599 }
600 
601 static int btrfs_delayed_inode_reserve_metadata(
602 					struct btrfs_trans_handle *trans,
603 					struct btrfs_root *root,
604 					struct btrfs_inode *inode,
605 					struct btrfs_delayed_node *node)
606 {
607 	struct btrfs_fs_info *fs_info = root->fs_info;
608 	struct btrfs_block_rsv *src_rsv;
609 	struct btrfs_block_rsv *dst_rsv;
610 	u64 num_bytes;
611 	int ret;
612 
613 	src_rsv = trans->block_rsv;
614 	dst_rsv = &fs_info->delayed_block_rsv;
615 
616 	num_bytes = btrfs_calc_metadata_size(fs_info, 1);
617 
618 	/*
619 	 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
620 	 * which doesn't reserve space for speed.  This is a problem since we
621 	 * still need to reserve space for this update, so try to reserve the
622 	 * space.
623 	 *
624 	 * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
625 	 * we always reserve enough to update the inode item.
626 	 */
627 	if (!src_rsv || (!trans->bytes_reserved &&
628 			 src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) {
629 		ret = btrfs_qgroup_reserve_meta_prealloc(root,
630 				fs_info->nodesize, true);
631 		if (ret < 0)
632 			return ret;
633 		ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
634 					  BTRFS_RESERVE_NO_FLUSH);
635 		/*
636 		 * Since we're under a transaction reserve_metadata_bytes could
637 		 * try to commit the transaction which will make it return
638 		 * EAGAIN to make us stop the transaction we have, so return
639 		 * ENOSPC instead so that btrfs_dirty_inode knows what to do.
640 		 */
641 		if (ret == -EAGAIN) {
642 			ret = -ENOSPC;
643 			btrfs_qgroup_free_meta_prealloc(root, num_bytes);
644 		}
645 		if (!ret) {
646 			node->bytes_reserved = num_bytes;
647 			trace_btrfs_space_reservation(fs_info,
648 						      "delayed_inode",
649 						      btrfs_ino(inode),
650 						      num_bytes, 1);
651 		} else {
652 			btrfs_qgroup_free_meta_prealloc(root, fs_info->nodesize);
653 		}
654 		return ret;
655 	}
656 
657 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes, true);
658 	if (!ret) {
659 		trace_btrfs_space_reservation(fs_info, "delayed_inode",
660 					      btrfs_ino(inode), num_bytes, 1);
661 		node->bytes_reserved = num_bytes;
662 	}
663 
664 	return ret;
665 }
666 
667 static void btrfs_delayed_inode_release_metadata(struct btrfs_fs_info *fs_info,
668 						struct btrfs_delayed_node *node,
669 						bool qgroup_free)
670 {
671 	struct btrfs_block_rsv *rsv;
672 
673 	if (!node->bytes_reserved)
674 		return;
675 
676 	rsv = &fs_info->delayed_block_rsv;
677 	trace_btrfs_space_reservation(fs_info, "delayed_inode",
678 				      node->inode_id, node->bytes_reserved, 0);
679 	btrfs_block_rsv_release(fs_info, rsv,
680 				node->bytes_reserved);
681 	if (qgroup_free)
682 		btrfs_qgroup_free_meta_prealloc(node->root,
683 				node->bytes_reserved);
684 	else
685 		btrfs_qgroup_convert_reserved_meta(node->root,
686 				node->bytes_reserved);
687 	node->bytes_reserved = 0;
688 }
689 
690 /*
691  * This helper will insert some continuous items into the same leaf according
692  * to the free space of the leaf.
693  */
694 static int btrfs_batch_insert_items(struct btrfs_root *root,
695 				    struct btrfs_path *path,
696 				    struct btrfs_delayed_item *item)
697 {
698 	struct btrfs_delayed_item *curr, *next;
699 	int free_space;
700 	int total_data_size = 0, total_size = 0;
701 	struct extent_buffer *leaf;
702 	char *data_ptr;
703 	struct btrfs_key *keys;
704 	u32 *data_size;
705 	struct list_head head;
706 	int slot;
707 	int nitems;
708 	int i;
709 	int ret = 0;
710 
711 	BUG_ON(!path->nodes[0]);
712 
713 	leaf = path->nodes[0];
714 	free_space = btrfs_leaf_free_space(leaf);
715 	INIT_LIST_HEAD(&head);
716 
717 	next = item;
718 	nitems = 0;
719 
720 	/*
721 	 * count the number of the continuous items that we can insert in batch
722 	 */
723 	while (total_size + next->data_len + sizeof(struct btrfs_item) <=
724 	       free_space) {
725 		total_data_size += next->data_len;
726 		total_size += next->data_len + sizeof(struct btrfs_item);
727 		list_add_tail(&next->tree_list, &head);
728 		nitems++;
729 
730 		curr = next;
731 		next = __btrfs_next_delayed_item(curr);
732 		if (!next)
733 			break;
734 
735 		if (!btrfs_is_continuous_delayed_item(curr, next))
736 			break;
737 	}
738 
739 	if (!nitems) {
740 		ret = 0;
741 		goto out;
742 	}
743 
744 	/*
745 	 * we need allocate some memory space, but it might cause the task
746 	 * to sleep, so we set all locked nodes in the path to blocking locks
747 	 * first.
748 	 */
749 	btrfs_set_path_blocking(path);
750 
751 	keys = kmalloc_array(nitems, sizeof(struct btrfs_key), GFP_NOFS);
752 	if (!keys) {
753 		ret = -ENOMEM;
754 		goto out;
755 	}
756 
757 	data_size = kmalloc_array(nitems, sizeof(u32), GFP_NOFS);
758 	if (!data_size) {
759 		ret = -ENOMEM;
760 		goto error;
761 	}
762 
763 	/* get keys of all the delayed items */
764 	i = 0;
765 	list_for_each_entry(next, &head, tree_list) {
766 		keys[i] = next->key;
767 		data_size[i] = next->data_len;
768 		i++;
769 	}
770 
771 	/* insert the keys of the items */
772 	setup_items_for_insert(root, path, keys, data_size,
773 			       total_data_size, total_size, nitems);
774 
775 	/* insert the dir index items */
776 	slot = path->slots[0];
777 	list_for_each_entry_safe(curr, next, &head, tree_list) {
778 		data_ptr = btrfs_item_ptr(leaf, slot, char);
779 		write_extent_buffer(leaf, &curr->data,
780 				    (unsigned long)data_ptr,
781 				    curr->data_len);
782 		slot++;
783 
784 		btrfs_delayed_item_release_metadata(root, curr);
785 
786 		list_del(&curr->tree_list);
787 		btrfs_release_delayed_item(curr);
788 	}
789 
790 error:
791 	kfree(data_size);
792 	kfree(keys);
793 out:
794 	return ret;
795 }
796 
797 /*
798  * This helper can just do simple insertion that needn't extend item for new
799  * data, such as directory name index insertion, inode insertion.
800  */
801 static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
802 				     struct btrfs_root *root,
803 				     struct btrfs_path *path,
804 				     struct btrfs_delayed_item *delayed_item)
805 {
806 	struct extent_buffer *leaf;
807 	char *ptr;
808 	int ret;
809 
810 	ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
811 				      delayed_item->data_len);
812 	if (ret < 0 && ret != -EEXIST)
813 		return ret;
814 
815 	leaf = path->nodes[0];
816 
817 	ptr = btrfs_item_ptr(leaf, path->slots[0], char);
818 
819 	write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
820 			    delayed_item->data_len);
821 	btrfs_mark_buffer_dirty(leaf);
822 
823 	btrfs_delayed_item_release_metadata(root, delayed_item);
824 	return 0;
825 }
826 
827 /*
828  * we insert an item first, then if there are some continuous items, we try
829  * to insert those items into the same leaf.
830  */
831 static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
832 				      struct btrfs_path *path,
833 				      struct btrfs_root *root,
834 				      struct btrfs_delayed_node *node)
835 {
836 	struct btrfs_delayed_item *curr, *prev;
837 	int ret = 0;
838 
839 do_again:
840 	mutex_lock(&node->mutex);
841 	curr = __btrfs_first_delayed_insertion_item(node);
842 	if (!curr)
843 		goto insert_end;
844 
845 	ret = btrfs_insert_delayed_item(trans, root, path, curr);
846 	if (ret < 0) {
847 		btrfs_release_path(path);
848 		goto insert_end;
849 	}
850 
851 	prev = curr;
852 	curr = __btrfs_next_delayed_item(prev);
853 	if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
854 		/* insert the continuous items into the same leaf */
855 		path->slots[0]++;
856 		btrfs_batch_insert_items(root, path, curr);
857 	}
858 	btrfs_release_delayed_item(prev);
859 	btrfs_mark_buffer_dirty(path->nodes[0]);
860 
861 	btrfs_release_path(path);
862 	mutex_unlock(&node->mutex);
863 	goto do_again;
864 
865 insert_end:
866 	mutex_unlock(&node->mutex);
867 	return ret;
868 }
869 
870 static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
871 				    struct btrfs_root *root,
872 				    struct btrfs_path *path,
873 				    struct btrfs_delayed_item *item)
874 {
875 	struct btrfs_delayed_item *curr, *next;
876 	struct extent_buffer *leaf;
877 	struct btrfs_key key;
878 	struct list_head head;
879 	int nitems, i, last_item;
880 	int ret = 0;
881 
882 	BUG_ON(!path->nodes[0]);
883 
884 	leaf = path->nodes[0];
885 
886 	i = path->slots[0];
887 	last_item = btrfs_header_nritems(leaf) - 1;
888 	if (i > last_item)
889 		return -ENOENT;	/* FIXME: Is errno suitable? */
890 
891 	next = item;
892 	INIT_LIST_HEAD(&head);
893 	btrfs_item_key_to_cpu(leaf, &key, i);
894 	nitems = 0;
895 	/*
896 	 * count the number of the dir index items that we can delete in batch
897 	 */
898 	while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
899 		list_add_tail(&next->tree_list, &head);
900 		nitems++;
901 
902 		curr = next;
903 		next = __btrfs_next_delayed_item(curr);
904 		if (!next)
905 			break;
906 
907 		if (!btrfs_is_continuous_delayed_item(curr, next))
908 			break;
909 
910 		i++;
911 		if (i > last_item)
912 			break;
913 		btrfs_item_key_to_cpu(leaf, &key, i);
914 	}
915 
916 	if (!nitems)
917 		return 0;
918 
919 	ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
920 	if (ret)
921 		goto out;
922 
923 	list_for_each_entry_safe(curr, next, &head, tree_list) {
924 		btrfs_delayed_item_release_metadata(root, curr);
925 		list_del(&curr->tree_list);
926 		btrfs_release_delayed_item(curr);
927 	}
928 
929 out:
930 	return ret;
931 }
932 
933 static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
934 				      struct btrfs_path *path,
935 				      struct btrfs_root *root,
936 				      struct btrfs_delayed_node *node)
937 {
938 	struct btrfs_delayed_item *curr, *prev;
939 	int ret = 0;
940 
941 do_again:
942 	mutex_lock(&node->mutex);
943 	curr = __btrfs_first_delayed_deletion_item(node);
944 	if (!curr)
945 		goto delete_fail;
946 
947 	ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
948 	if (ret < 0)
949 		goto delete_fail;
950 	else if (ret > 0) {
951 		/*
952 		 * can't find the item which the node points to, so this node
953 		 * is invalid, just drop it.
954 		 */
955 		prev = curr;
956 		curr = __btrfs_next_delayed_item(prev);
957 		btrfs_release_delayed_item(prev);
958 		ret = 0;
959 		btrfs_release_path(path);
960 		if (curr) {
961 			mutex_unlock(&node->mutex);
962 			goto do_again;
963 		} else
964 			goto delete_fail;
965 	}
966 
967 	btrfs_batch_delete_items(trans, root, path, curr);
968 	btrfs_release_path(path);
969 	mutex_unlock(&node->mutex);
970 	goto do_again;
971 
972 delete_fail:
973 	btrfs_release_path(path);
974 	mutex_unlock(&node->mutex);
975 	return ret;
976 }
977 
978 static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
979 {
980 	struct btrfs_delayed_root *delayed_root;
981 
982 	if (delayed_node &&
983 	    test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
984 		BUG_ON(!delayed_node->root);
985 		clear_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
986 		delayed_node->count--;
987 
988 		delayed_root = delayed_node->root->fs_info->delayed_root;
989 		finish_one_item(delayed_root);
990 	}
991 }
992 
993 static void btrfs_release_delayed_iref(struct btrfs_delayed_node *delayed_node)
994 {
995 	struct btrfs_delayed_root *delayed_root;
996 
997 	ASSERT(delayed_node->root);
998 	clear_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
999 	delayed_node->count--;
1000 
1001 	delayed_root = delayed_node->root->fs_info->delayed_root;
1002 	finish_one_item(delayed_root);
1003 }
1004 
1005 static int __btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1006 					struct btrfs_root *root,
1007 					struct btrfs_path *path,
1008 					struct btrfs_delayed_node *node)
1009 {
1010 	struct btrfs_fs_info *fs_info = root->fs_info;
1011 	struct btrfs_key key;
1012 	struct btrfs_inode_item *inode_item;
1013 	struct extent_buffer *leaf;
1014 	int mod;
1015 	int ret;
1016 
1017 	key.objectid = node->inode_id;
1018 	key.type = BTRFS_INODE_ITEM_KEY;
1019 	key.offset = 0;
1020 
1021 	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
1022 		mod = -1;
1023 	else
1024 		mod = 1;
1025 
1026 	ret = btrfs_lookup_inode(trans, root, path, &key, mod);
1027 	if (ret > 0) {
1028 		btrfs_release_path(path);
1029 		return -ENOENT;
1030 	} else if (ret < 0) {
1031 		return ret;
1032 	}
1033 
1034 	leaf = path->nodes[0];
1035 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
1036 				    struct btrfs_inode_item);
1037 	write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
1038 			    sizeof(struct btrfs_inode_item));
1039 	btrfs_mark_buffer_dirty(leaf);
1040 
1041 	if (!test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
1042 		goto no_iref;
1043 
1044 	path->slots[0]++;
1045 	if (path->slots[0] >= btrfs_header_nritems(leaf))
1046 		goto search;
1047 again:
1048 	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1049 	if (key.objectid != node->inode_id)
1050 		goto out;
1051 
1052 	if (key.type != BTRFS_INODE_REF_KEY &&
1053 	    key.type != BTRFS_INODE_EXTREF_KEY)
1054 		goto out;
1055 
1056 	/*
1057 	 * Delayed iref deletion is for the inode who has only one link,
1058 	 * so there is only one iref. The case that several irefs are
1059 	 * in the same item doesn't exist.
1060 	 */
1061 	btrfs_del_item(trans, root, path);
1062 out:
1063 	btrfs_release_delayed_iref(node);
1064 no_iref:
1065 	btrfs_release_path(path);
1066 err_out:
1067 	btrfs_delayed_inode_release_metadata(fs_info, node, (ret < 0));
1068 	btrfs_release_delayed_inode(node);
1069 
1070 	return ret;
1071 
1072 search:
1073 	btrfs_release_path(path);
1074 
1075 	key.type = BTRFS_INODE_EXTREF_KEY;
1076 	key.offset = -1;
1077 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1078 	if (ret < 0)
1079 		goto err_out;
1080 	ASSERT(ret);
1081 
1082 	ret = 0;
1083 	leaf = path->nodes[0];
1084 	path->slots[0]--;
1085 	goto again;
1086 }
1087 
1088 static inline int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1089 					     struct btrfs_root *root,
1090 					     struct btrfs_path *path,
1091 					     struct btrfs_delayed_node *node)
1092 {
1093 	int ret;
1094 
1095 	mutex_lock(&node->mutex);
1096 	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &node->flags)) {
1097 		mutex_unlock(&node->mutex);
1098 		return 0;
1099 	}
1100 
1101 	ret = __btrfs_update_delayed_inode(trans, root, path, node);
1102 	mutex_unlock(&node->mutex);
1103 	return ret;
1104 }
1105 
1106 static inline int
1107 __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1108 				   struct btrfs_path *path,
1109 				   struct btrfs_delayed_node *node)
1110 {
1111 	int ret;
1112 
1113 	ret = btrfs_insert_delayed_items(trans, path, node->root, node);
1114 	if (ret)
1115 		return ret;
1116 
1117 	ret = btrfs_delete_delayed_items(trans, path, node->root, node);
1118 	if (ret)
1119 		return ret;
1120 
1121 	ret = btrfs_update_delayed_inode(trans, node->root, path, node);
1122 	return ret;
1123 }
1124 
1125 /*
1126  * Called when committing the transaction.
1127  * Returns 0 on success.
1128  * Returns < 0 on error and returns with an aborted transaction with any
1129  * outstanding delayed items cleaned up.
1130  */
1131 static int __btrfs_run_delayed_items(struct btrfs_trans_handle *trans, int nr)
1132 {
1133 	struct btrfs_fs_info *fs_info = trans->fs_info;
1134 	struct btrfs_delayed_root *delayed_root;
1135 	struct btrfs_delayed_node *curr_node, *prev_node;
1136 	struct btrfs_path *path;
1137 	struct btrfs_block_rsv *block_rsv;
1138 	int ret = 0;
1139 	bool count = (nr > 0);
1140 
1141 	if (trans->aborted)
1142 		return -EIO;
1143 
1144 	path = btrfs_alloc_path();
1145 	if (!path)
1146 		return -ENOMEM;
1147 	path->leave_spinning = 1;
1148 
1149 	block_rsv = trans->block_rsv;
1150 	trans->block_rsv = &fs_info->delayed_block_rsv;
1151 
1152 	delayed_root = fs_info->delayed_root;
1153 
1154 	curr_node = btrfs_first_delayed_node(delayed_root);
1155 	while (curr_node && (!count || (count && nr--))) {
1156 		ret = __btrfs_commit_inode_delayed_items(trans, path,
1157 							 curr_node);
1158 		if (ret) {
1159 			btrfs_release_delayed_node(curr_node);
1160 			curr_node = NULL;
1161 			btrfs_abort_transaction(trans, ret);
1162 			break;
1163 		}
1164 
1165 		prev_node = curr_node;
1166 		curr_node = btrfs_next_delayed_node(curr_node);
1167 		btrfs_release_delayed_node(prev_node);
1168 	}
1169 
1170 	if (curr_node)
1171 		btrfs_release_delayed_node(curr_node);
1172 	btrfs_free_path(path);
1173 	trans->block_rsv = block_rsv;
1174 
1175 	return ret;
1176 }
1177 
1178 int btrfs_run_delayed_items(struct btrfs_trans_handle *trans)
1179 {
1180 	return __btrfs_run_delayed_items(trans, -1);
1181 }
1182 
1183 int btrfs_run_delayed_items_nr(struct btrfs_trans_handle *trans, int nr)
1184 {
1185 	return __btrfs_run_delayed_items(trans, nr);
1186 }
1187 
1188 int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1189 				     struct btrfs_inode *inode)
1190 {
1191 	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1192 	struct btrfs_path *path;
1193 	struct btrfs_block_rsv *block_rsv;
1194 	int ret;
1195 
1196 	if (!delayed_node)
1197 		return 0;
1198 
1199 	mutex_lock(&delayed_node->mutex);
1200 	if (!delayed_node->count) {
1201 		mutex_unlock(&delayed_node->mutex);
1202 		btrfs_release_delayed_node(delayed_node);
1203 		return 0;
1204 	}
1205 	mutex_unlock(&delayed_node->mutex);
1206 
1207 	path = btrfs_alloc_path();
1208 	if (!path) {
1209 		btrfs_release_delayed_node(delayed_node);
1210 		return -ENOMEM;
1211 	}
1212 	path->leave_spinning = 1;
1213 
1214 	block_rsv = trans->block_rsv;
1215 	trans->block_rsv = &delayed_node->root->fs_info->delayed_block_rsv;
1216 
1217 	ret = __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
1218 
1219 	btrfs_release_delayed_node(delayed_node);
1220 	btrfs_free_path(path);
1221 	trans->block_rsv = block_rsv;
1222 
1223 	return ret;
1224 }
1225 
1226 int btrfs_commit_inode_delayed_inode(struct btrfs_inode *inode)
1227 {
1228 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
1229 	struct btrfs_trans_handle *trans;
1230 	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1231 	struct btrfs_path *path;
1232 	struct btrfs_block_rsv *block_rsv;
1233 	int ret;
1234 
1235 	if (!delayed_node)
1236 		return 0;
1237 
1238 	mutex_lock(&delayed_node->mutex);
1239 	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1240 		mutex_unlock(&delayed_node->mutex);
1241 		btrfs_release_delayed_node(delayed_node);
1242 		return 0;
1243 	}
1244 	mutex_unlock(&delayed_node->mutex);
1245 
1246 	trans = btrfs_join_transaction(delayed_node->root);
1247 	if (IS_ERR(trans)) {
1248 		ret = PTR_ERR(trans);
1249 		goto out;
1250 	}
1251 
1252 	path = btrfs_alloc_path();
1253 	if (!path) {
1254 		ret = -ENOMEM;
1255 		goto trans_out;
1256 	}
1257 	path->leave_spinning = 1;
1258 
1259 	block_rsv = trans->block_rsv;
1260 	trans->block_rsv = &fs_info->delayed_block_rsv;
1261 
1262 	mutex_lock(&delayed_node->mutex);
1263 	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags))
1264 		ret = __btrfs_update_delayed_inode(trans, delayed_node->root,
1265 						   path, delayed_node);
1266 	else
1267 		ret = 0;
1268 	mutex_unlock(&delayed_node->mutex);
1269 
1270 	btrfs_free_path(path);
1271 	trans->block_rsv = block_rsv;
1272 trans_out:
1273 	btrfs_end_transaction(trans);
1274 	btrfs_btree_balance_dirty(fs_info);
1275 out:
1276 	btrfs_release_delayed_node(delayed_node);
1277 
1278 	return ret;
1279 }
1280 
1281 void btrfs_remove_delayed_node(struct btrfs_inode *inode)
1282 {
1283 	struct btrfs_delayed_node *delayed_node;
1284 
1285 	delayed_node = READ_ONCE(inode->delayed_node);
1286 	if (!delayed_node)
1287 		return;
1288 
1289 	inode->delayed_node = NULL;
1290 	btrfs_release_delayed_node(delayed_node);
1291 }
1292 
1293 struct btrfs_async_delayed_work {
1294 	struct btrfs_delayed_root *delayed_root;
1295 	int nr;
1296 	struct btrfs_work work;
1297 };
1298 
1299 static void btrfs_async_run_delayed_root(struct btrfs_work *work)
1300 {
1301 	struct btrfs_async_delayed_work *async_work;
1302 	struct btrfs_delayed_root *delayed_root;
1303 	struct btrfs_trans_handle *trans;
1304 	struct btrfs_path *path;
1305 	struct btrfs_delayed_node *delayed_node = NULL;
1306 	struct btrfs_root *root;
1307 	struct btrfs_block_rsv *block_rsv;
1308 	int total_done = 0;
1309 
1310 	async_work = container_of(work, struct btrfs_async_delayed_work, work);
1311 	delayed_root = async_work->delayed_root;
1312 
1313 	path = btrfs_alloc_path();
1314 	if (!path)
1315 		goto out;
1316 
1317 	do {
1318 		if (atomic_read(&delayed_root->items) <
1319 		    BTRFS_DELAYED_BACKGROUND / 2)
1320 			break;
1321 
1322 		delayed_node = btrfs_first_prepared_delayed_node(delayed_root);
1323 		if (!delayed_node)
1324 			break;
1325 
1326 		path->leave_spinning = 1;
1327 		root = delayed_node->root;
1328 
1329 		trans = btrfs_join_transaction(root);
1330 		if (IS_ERR(trans)) {
1331 			btrfs_release_path(path);
1332 			btrfs_release_prepared_delayed_node(delayed_node);
1333 			total_done++;
1334 			continue;
1335 		}
1336 
1337 		block_rsv = trans->block_rsv;
1338 		trans->block_rsv = &root->fs_info->delayed_block_rsv;
1339 
1340 		__btrfs_commit_inode_delayed_items(trans, path, delayed_node);
1341 
1342 		trans->block_rsv = block_rsv;
1343 		btrfs_end_transaction(trans);
1344 		btrfs_btree_balance_dirty_nodelay(root->fs_info);
1345 
1346 		btrfs_release_path(path);
1347 		btrfs_release_prepared_delayed_node(delayed_node);
1348 		total_done++;
1349 
1350 	} while ((async_work->nr == 0 && total_done < BTRFS_DELAYED_WRITEBACK)
1351 		 || total_done < async_work->nr);
1352 
1353 	btrfs_free_path(path);
1354 out:
1355 	wake_up(&delayed_root->wait);
1356 	kfree(async_work);
1357 }
1358 
1359 
1360 static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
1361 				     struct btrfs_fs_info *fs_info, int nr)
1362 {
1363 	struct btrfs_async_delayed_work *async_work;
1364 
1365 	async_work = kmalloc(sizeof(*async_work), GFP_NOFS);
1366 	if (!async_work)
1367 		return -ENOMEM;
1368 
1369 	async_work->delayed_root = delayed_root;
1370 	btrfs_init_work(&async_work->work, btrfs_delayed_meta_helper,
1371 			btrfs_async_run_delayed_root, NULL, NULL);
1372 	async_work->nr = nr;
1373 
1374 	btrfs_queue_work(fs_info->delayed_workers, &async_work->work);
1375 	return 0;
1376 }
1377 
1378 void btrfs_assert_delayed_root_empty(struct btrfs_fs_info *fs_info)
1379 {
1380 	WARN_ON(btrfs_first_delayed_node(fs_info->delayed_root));
1381 }
1382 
1383 static int could_end_wait(struct btrfs_delayed_root *delayed_root, int seq)
1384 {
1385 	int val = atomic_read(&delayed_root->items_seq);
1386 
1387 	if (val < seq || val >= seq + BTRFS_DELAYED_BATCH)
1388 		return 1;
1389 
1390 	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1391 		return 1;
1392 
1393 	return 0;
1394 }
1395 
1396 void btrfs_balance_delayed_items(struct btrfs_fs_info *fs_info)
1397 {
1398 	struct btrfs_delayed_root *delayed_root = fs_info->delayed_root;
1399 
1400 	if ((atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND) ||
1401 		btrfs_workqueue_normal_congested(fs_info->delayed_workers))
1402 		return;
1403 
1404 	if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
1405 		int seq;
1406 		int ret;
1407 
1408 		seq = atomic_read(&delayed_root->items_seq);
1409 
1410 		ret = btrfs_wq_run_delayed_node(delayed_root, fs_info, 0);
1411 		if (ret)
1412 			return;
1413 
1414 		wait_event_interruptible(delayed_root->wait,
1415 					 could_end_wait(delayed_root, seq));
1416 		return;
1417 	}
1418 
1419 	btrfs_wq_run_delayed_node(delayed_root, fs_info, BTRFS_DELAYED_BATCH);
1420 }
1421 
1422 /* Will return 0 or -ENOMEM */
1423 int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
1424 				   const char *name, int name_len,
1425 				   struct btrfs_inode *dir,
1426 				   struct btrfs_disk_key *disk_key, u8 type,
1427 				   u64 index)
1428 {
1429 	struct btrfs_delayed_node *delayed_node;
1430 	struct btrfs_delayed_item *delayed_item;
1431 	struct btrfs_dir_item *dir_item;
1432 	int ret;
1433 
1434 	delayed_node = btrfs_get_or_create_delayed_node(dir);
1435 	if (IS_ERR(delayed_node))
1436 		return PTR_ERR(delayed_node);
1437 
1438 	delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
1439 	if (!delayed_item) {
1440 		ret = -ENOMEM;
1441 		goto release_node;
1442 	}
1443 
1444 	delayed_item->key.objectid = btrfs_ino(dir);
1445 	delayed_item->key.type = BTRFS_DIR_INDEX_KEY;
1446 	delayed_item->key.offset = index;
1447 
1448 	dir_item = (struct btrfs_dir_item *)delayed_item->data;
1449 	dir_item->location = *disk_key;
1450 	btrfs_set_stack_dir_transid(dir_item, trans->transid);
1451 	btrfs_set_stack_dir_data_len(dir_item, 0);
1452 	btrfs_set_stack_dir_name_len(dir_item, name_len);
1453 	btrfs_set_stack_dir_type(dir_item, type);
1454 	memcpy((char *)(dir_item + 1), name, name_len);
1455 
1456 	ret = btrfs_delayed_item_reserve_metadata(trans, dir->root, delayed_item);
1457 	/*
1458 	 * we have reserved enough space when we start a new transaction,
1459 	 * so reserving metadata failure is impossible
1460 	 */
1461 	BUG_ON(ret);
1462 
1463 	mutex_lock(&delayed_node->mutex);
1464 	ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
1465 	if (unlikely(ret)) {
1466 		btrfs_err(trans->fs_info,
1467 			  "err add delayed dir index item(name: %.*s) into the insertion tree of the delayed node(root id: %llu, inode id: %llu, errno: %d)",
1468 			  name_len, name, delayed_node->root->root_key.objectid,
1469 			  delayed_node->inode_id, ret);
1470 		BUG();
1471 	}
1472 	mutex_unlock(&delayed_node->mutex);
1473 
1474 release_node:
1475 	btrfs_release_delayed_node(delayed_node);
1476 	return ret;
1477 }
1478 
1479 static int btrfs_delete_delayed_insertion_item(struct btrfs_fs_info *fs_info,
1480 					       struct btrfs_delayed_node *node,
1481 					       struct btrfs_key *key)
1482 {
1483 	struct btrfs_delayed_item *item;
1484 
1485 	mutex_lock(&node->mutex);
1486 	item = __btrfs_lookup_delayed_insertion_item(node, key);
1487 	if (!item) {
1488 		mutex_unlock(&node->mutex);
1489 		return 1;
1490 	}
1491 
1492 	btrfs_delayed_item_release_metadata(node->root, item);
1493 	btrfs_release_delayed_item(item);
1494 	mutex_unlock(&node->mutex);
1495 	return 0;
1496 }
1497 
1498 int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
1499 				   struct btrfs_inode *dir, u64 index)
1500 {
1501 	struct btrfs_delayed_node *node;
1502 	struct btrfs_delayed_item *item;
1503 	struct btrfs_key item_key;
1504 	int ret;
1505 
1506 	node = btrfs_get_or_create_delayed_node(dir);
1507 	if (IS_ERR(node))
1508 		return PTR_ERR(node);
1509 
1510 	item_key.objectid = btrfs_ino(dir);
1511 	item_key.type = BTRFS_DIR_INDEX_KEY;
1512 	item_key.offset = index;
1513 
1514 	ret = btrfs_delete_delayed_insertion_item(trans->fs_info, node,
1515 						  &item_key);
1516 	if (!ret)
1517 		goto end;
1518 
1519 	item = btrfs_alloc_delayed_item(0);
1520 	if (!item) {
1521 		ret = -ENOMEM;
1522 		goto end;
1523 	}
1524 
1525 	item->key = item_key;
1526 
1527 	ret = btrfs_delayed_item_reserve_metadata(trans, dir->root, item);
1528 	/*
1529 	 * we have reserved enough space when we start a new transaction,
1530 	 * so reserving metadata failure is impossible.
1531 	 */
1532 	if (ret < 0) {
1533 		btrfs_err(trans->fs_info,
1534 "metadata reservation failed for delayed dir item deltiona, should have been reserved");
1535 		btrfs_release_delayed_item(item);
1536 		goto end;
1537 	}
1538 
1539 	mutex_lock(&node->mutex);
1540 	ret = __btrfs_add_delayed_deletion_item(node, item);
1541 	if (unlikely(ret)) {
1542 		btrfs_err(trans->fs_info,
1543 			  "err add delayed dir index item(index: %llu) into the deletion tree of the delayed node(root id: %llu, inode id: %llu, errno: %d)",
1544 			  index, node->root->root_key.objectid,
1545 			  node->inode_id, ret);
1546 		btrfs_delayed_item_release_metadata(dir->root, item);
1547 		btrfs_release_delayed_item(item);
1548 	}
1549 	mutex_unlock(&node->mutex);
1550 end:
1551 	btrfs_release_delayed_node(node);
1552 	return ret;
1553 }
1554 
1555 int btrfs_inode_delayed_dir_index_count(struct btrfs_inode *inode)
1556 {
1557 	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1558 
1559 	if (!delayed_node)
1560 		return -ENOENT;
1561 
1562 	/*
1563 	 * Since we have held i_mutex of this directory, it is impossible that
1564 	 * a new directory index is added into the delayed node and index_cnt
1565 	 * is updated now. So we needn't lock the delayed node.
1566 	 */
1567 	if (!delayed_node->index_cnt) {
1568 		btrfs_release_delayed_node(delayed_node);
1569 		return -EINVAL;
1570 	}
1571 
1572 	inode->index_cnt = delayed_node->index_cnt;
1573 	btrfs_release_delayed_node(delayed_node);
1574 	return 0;
1575 }
1576 
1577 bool btrfs_readdir_get_delayed_items(struct inode *inode,
1578 				     struct list_head *ins_list,
1579 				     struct list_head *del_list)
1580 {
1581 	struct btrfs_delayed_node *delayed_node;
1582 	struct btrfs_delayed_item *item;
1583 
1584 	delayed_node = btrfs_get_delayed_node(BTRFS_I(inode));
1585 	if (!delayed_node)
1586 		return false;
1587 
1588 	/*
1589 	 * We can only do one readdir with delayed items at a time because of
1590 	 * item->readdir_list.
1591 	 */
1592 	inode_unlock_shared(inode);
1593 	inode_lock(inode);
1594 
1595 	mutex_lock(&delayed_node->mutex);
1596 	item = __btrfs_first_delayed_insertion_item(delayed_node);
1597 	while (item) {
1598 		refcount_inc(&item->refs);
1599 		list_add_tail(&item->readdir_list, ins_list);
1600 		item = __btrfs_next_delayed_item(item);
1601 	}
1602 
1603 	item = __btrfs_first_delayed_deletion_item(delayed_node);
1604 	while (item) {
1605 		refcount_inc(&item->refs);
1606 		list_add_tail(&item->readdir_list, del_list);
1607 		item = __btrfs_next_delayed_item(item);
1608 	}
1609 	mutex_unlock(&delayed_node->mutex);
1610 	/*
1611 	 * This delayed node is still cached in the btrfs inode, so refs
1612 	 * must be > 1 now, and we needn't check it is going to be freed
1613 	 * or not.
1614 	 *
1615 	 * Besides that, this function is used to read dir, we do not
1616 	 * insert/delete delayed items in this period. So we also needn't
1617 	 * requeue or dequeue this delayed node.
1618 	 */
1619 	refcount_dec(&delayed_node->refs);
1620 
1621 	return true;
1622 }
1623 
1624 void btrfs_readdir_put_delayed_items(struct inode *inode,
1625 				     struct list_head *ins_list,
1626 				     struct list_head *del_list)
1627 {
1628 	struct btrfs_delayed_item *curr, *next;
1629 
1630 	list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1631 		list_del(&curr->readdir_list);
1632 		if (refcount_dec_and_test(&curr->refs))
1633 			kfree(curr);
1634 	}
1635 
1636 	list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1637 		list_del(&curr->readdir_list);
1638 		if (refcount_dec_and_test(&curr->refs))
1639 			kfree(curr);
1640 	}
1641 
1642 	/*
1643 	 * The VFS is going to do up_read(), so we need to downgrade back to a
1644 	 * read lock.
1645 	 */
1646 	downgrade_write(&inode->i_rwsem);
1647 }
1648 
1649 int btrfs_should_delete_dir_index(struct list_head *del_list,
1650 				  u64 index)
1651 {
1652 	struct btrfs_delayed_item *curr;
1653 	int ret = 0;
1654 
1655 	list_for_each_entry(curr, del_list, readdir_list) {
1656 		if (curr->key.offset > index)
1657 			break;
1658 		if (curr->key.offset == index) {
1659 			ret = 1;
1660 			break;
1661 		}
1662 	}
1663 	return ret;
1664 }
1665 
1666 /*
1667  * btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
1668  *
1669  */
1670 int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
1671 				    struct list_head *ins_list)
1672 {
1673 	struct btrfs_dir_item *di;
1674 	struct btrfs_delayed_item *curr, *next;
1675 	struct btrfs_key location;
1676 	char *name;
1677 	int name_len;
1678 	int over = 0;
1679 	unsigned char d_type;
1680 
1681 	if (list_empty(ins_list))
1682 		return 0;
1683 
1684 	/*
1685 	 * Changing the data of the delayed item is impossible. So
1686 	 * we needn't lock them. And we have held i_mutex of the
1687 	 * directory, nobody can delete any directory indexes now.
1688 	 */
1689 	list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1690 		list_del(&curr->readdir_list);
1691 
1692 		if (curr->key.offset < ctx->pos) {
1693 			if (refcount_dec_and_test(&curr->refs))
1694 				kfree(curr);
1695 			continue;
1696 		}
1697 
1698 		ctx->pos = curr->key.offset;
1699 
1700 		di = (struct btrfs_dir_item *)curr->data;
1701 		name = (char *)(di + 1);
1702 		name_len = btrfs_stack_dir_name_len(di);
1703 
1704 		d_type = fs_ftype_to_dtype(di->type);
1705 		btrfs_disk_key_to_cpu(&location, &di->location);
1706 
1707 		over = !dir_emit(ctx, name, name_len,
1708 			       location.objectid, d_type);
1709 
1710 		if (refcount_dec_and_test(&curr->refs))
1711 			kfree(curr);
1712 
1713 		if (over)
1714 			return 1;
1715 		ctx->pos++;
1716 	}
1717 	return 0;
1718 }
1719 
1720 static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
1721 				  struct btrfs_inode_item *inode_item,
1722 				  struct inode *inode)
1723 {
1724 	btrfs_set_stack_inode_uid(inode_item, i_uid_read(inode));
1725 	btrfs_set_stack_inode_gid(inode_item, i_gid_read(inode));
1726 	btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
1727 	btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
1728 	btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
1729 	btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
1730 	btrfs_set_stack_inode_generation(inode_item,
1731 					 BTRFS_I(inode)->generation);
1732 	btrfs_set_stack_inode_sequence(inode_item,
1733 				       inode_peek_iversion(inode));
1734 	btrfs_set_stack_inode_transid(inode_item, trans->transid);
1735 	btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
1736 	btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
1737 	btrfs_set_stack_inode_block_group(inode_item, 0);
1738 
1739 	btrfs_set_stack_timespec_sec(&inode_item->atime,
1740 				     inode->i_atime.tv_sec);
1741 	btrfs_set_stack_timespec_nsec(&inode_item->atime,
1742 				      inode->i_atime.tv_nsec);
1743 
1744 	btrfs_set_stack_timespec_sec(&inode_item->mtime,
1745 				     inode->i_mtime.tv_sec);
1746 	btrfs_set_stack_timespec_nsec(&inode_item->mtime,
1747 				      inode->i_mtime.tv_nsec);
1748 
1749 	btrfs_set_stack_timespec_sec(&inode_item->ctime,
1750 				     inode->i_ctime.tv_sec);
1751 	btrfs_set_stack_timespec_nsec(&inode_item->ctime,
1752 				      inode->i_ctime.tv_nsec);
1753 
1754 	btrfs_set_stack_timespec_sec(&inode_item->otime,
1755 				     BTRFS_I(inode)->i_otime.tv_sec);
1756 	btrfs_set_stack_timespec_nsec(&inode_item->otime,
1757 				     BTRFS_I(inode)->i_otime.tv_nsec);
1758 }
1759 
1760 int btrfs_fill_inode(struct inode *inode, u32 *rdev)
1761 {
1762 	struct btrfs_delayed_node *delayed_node;
1763 	struct btrfs_inode_item *inode_item;
1764 
1765 	delayed_node = btrfs_get_delayed_node(BTRFS_I(inode));
1766 	if (!delayed_node)
1767 		return -ENOENT;
1768 
1769 	mutex_lock(&delayed_node->mutex);
1770 	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1771 		mutex_unlock(&delayed_node->mutex);
1772 		btrfs_release_delayed_node(delayed_node);
1773 		return -ENOENT;
1774 	}
1775 
1776 	inode_item = &delayed_node->inode_item;
1777 
1778 	i_uid_write(inode, btrfs_stack_inode_uid(inode_item));
1779 	i_gid_write(inode, btrfs_stack_inode_gid(inode_item));
1780 	btrfs_i_size_write(BTRFS_I(inode), btrfs_stack_inode_size(inode_item));
1781 	inode->i_mode = btrfs_stack_inode_mode(inode_item);
1782 	set_nlink(inode, btrfs_stack_inode_nlink(inode_item));
1783 	inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
1784 	BTRFS_I(inode)->generation = btrfs_stack_inode_generation(inode_item);
1785         BTRFS_I(inode)->last_trans = btrfs_stack_inode_transid(inode_item);
1786 
1787 	inode_set_iversion_queried(inode,
1788 				   btrfs_stack_inode_sequence(inode_item));
1789 	inode->i_rdev = 0;
1790 	*rdev = btrfs_stack_inode_rdev(inode_item);
1791 	BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
1792 
1793 	inode->i_atime.tv_sec = btrfs_stack_timespec_sec(&inode_item->atime);
1794 	inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->atime);
1795 
1796 	inode->i_mtime.tv_sec = btrfs_stack_timespec_sec(&inode_item->mtime);
1797 	inode->i_mtime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->mtime);
1798 
1799 	inode->i_ctime.tv_sec = btrfs_stack_timespec_sec(&inode_item->ctime);
1800 	inode->i_ctime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->ctime);
1801 
1802 	BTRFS_I(inode)->i_otime.tv_sec =
1803 		btrfs_stack_timespec_sec(&inode_item->otime);
1804 	BTRFS_I(inode)->i_otime.tv_nsec =
1805 		btrfs_stack_timespec_nsec(&inode_item->otime);
1806 
1807 	inode->i_generation = BTRFS_I(inode)->generation;
1808 	BTRFS_I(inode)->index_cnt = (u64)-1;
1809 
1810 	mutex_unlock(&delayed_node->mutex);
1811 	btrfs_release_delayed_node(delayed_node);
1812 	return 0;
1813 }
1814 
1815 int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
1816 			       struct btrfs_root *root, struct inode *inode)
1817 {
1818 	struct btrfs_delayed_node *delayed_node;
1819 	int ret = 0;
1820 
1821 	delayed_node = btrfs_get_or_create_delayed_node(BTRFS_I(inode));
1822 	if (IS_ERR(delayed_node))
1823 		return PTR_ERR(delayed_node);
1824 
1825 	mutex_lock(&delayed_node->mutex);
1826 	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1827 		fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1828 		goto release_node;
1829 	}
1830 
1831 	ret = btrfs_delayed_inode_reserve_metadata(trans, root, BTRFS_I(inode),
1832 						   delayed_node);
1833 	if (ret)
1834 		goto release_node;
1835 
1836 	fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1837 	set_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
1838 	delayed_node->count++;
1839 	atomic_inc(&root->fs_info->delayed_root->items);
1840 release_node:
1841 	mutex_unlock(&delayed_node->mutex);
1842 	btrfs_release_delayed_node(delayed_node);
1843 	return ret;
1844 }
1845 
1846 int btrfs_delayed_delete_inode_ref(struct btrfs_inode *inode)
1847 {
1848 	struct btrfs_fs_info *fs_info = inode->root->fs_info;
1849 	struct btrfs_delayed_node *delayed_node;
1850 
1851 	/*
1852 	 * we don't do delayed inode updates during log recovery because it
1853 	 * leads to enospc problems.  This means we also can't do
1854 	 * delayed inode refs
1855 	 */
1856 	if (test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags))
1857 		return -EAGAIN;
1858 
1859 	delayed_node = btrfs_get_or_create_delayed_node(inode);
1860 	if (IS_ERR(delayed_node))
1861 		return PTR_ERR(delayed_node);
1862 
1863 	/*
1864 	 * We don't reserve space for inode ref deletion is because:
1865 	 * - We ONLY do async inode ref deletion for the inode who has only
1866 	 *   one link(i_nlink == 1), it means there is only one inode ref.
1867 	 *   And in most case, the inode ref and the inode item are in the
1868 	 *   same leaf, and we will deal with them at the same time.
1869 	 *   Since we are sure we will reserve the space for the inode item,
1870 	 *   it is unnecessary to reserve space for inode ref deletion.
1871 	 * - If the inode ref and the inode item are not in the same leaf,
1872 	 *   We also needn't worry about enospc problem, because we reserve
1873 	 *   much more space for the inode update than it needs.
1874 	 * - At the worst, we can steal some space from the global reservation.
1875 	 *   It is very rare.
1876 	 */
1877 	mutex_lock(&delayed_node->mutex);
1878 	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
1879 		goto release_node;
1880 
1881 	set_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
1882 	delayed_node->count++;
1883 	atomic_inc(&fs_info->delayed_root->items);
1884 release_node:
1885 	mutex_unlock(&delayed_node->mutex);
1886 	btrfs_release_delayed_node(delayed_node);
1887 	return 0;
1888 }
1889 
1890 static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
1891 {
1892 	struct btrfs_root *root = delayed_node->root;
1893 	struct btrfs_fs_info *fs_info = root->fs_info;
1894 	struct btrfs_delayed_item *curr_item, *prev_item;
1895 
1896 	mutex_lock(&delayed_node->mutex);
1897 	curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
1898 	while (curr_item) {
1899 		btrfs_delayed_item_release_metadata(root, curr_item);
1900 		prev_item = curr_item;
1901 		curr_item = __btrfs_next_delayed_item(prev_item);
1902 		btrfs_release_delayed_item(prev_item);
1903 	}
1904 
1905 	curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
1906 	while (curr_item) {
1907 		btrfs_delayed_item_release_metadata(root, curr_item);
1908 		prev_item = curr_item;
1909 		curr_item = __btrfs_next_delayed_item(prev_item);
1910 		btrfs_release_delayed_item(prev_item);
1911 	}
1912 
1913 	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
1914 		btrfs_release_delayed_iref(delayed_node);
1915 
1916 	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1917 		btrfs_delayed_inode_release_metadata(fs_info, delayed_node, false);
1918 		btrfs_release_delayed_inode(delayed_node);
1919 	}
1920 	mutex_unlock(&delayed_node->mutex);
1921 }
1922 
1923 void btrfs_kill_delayed_inode_items(struct btrfs_inode *inode)
1924 {
1925 	struct btrfs_delayed_node *delayed_node;
1926 
1927 	delayed_node = btrfs_get_delayed_node(inode);
1928 	if (!delayed_node)
1929 		return;
1930 
1931 	__btrfs_kill_delayed_node(delayed_node);
1932 	btrfs_release_delayed_node(delayed_node);
1933 }
1934 
1935 void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
1936 {
1937 	u64 inode_id = 0;
1938 	struct btrfs_delayed_node *delayed_nodes[8];
1939 	int i, n;
1940 
1941 	while (1) {
1942 		spin_lock(&root->inode_lock);
1943 		n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
1944 					   (void **)delayed_nodes, inode_id,
1945 					   ARRAY_SIZE(delayed_nodes));
1946 		if (!n) {
1947 			spin_unlock(&root->inode_lock);
1948 			break;
1949 		}
1950 
1951 		inode_id = delayed_nodes[n - 1]->inode_id + 1;
1952 
1953 		for (i = 0; i < n; i++)
1954 			refcount_inc(&delayed_nodes[i]->refs);
1955 		spin_unlock(&root->inode_lock);
1956 
1957 		for (i = 0; i < n; i++) {
1958 			__btrfs_kill_delayed_node(delayed_nodes[i]);
1959 			btrfs_release_delayed_node(delayed_nodes[i]);
1960 		}
1961 	}
1962 }
1963 
1964 void btrfs_destroy_delayed_inodes(struct btrfs_fs_info *fs_info)
1965 {
1966 	struct btrfs_delayed_node *curr_node, *prev_node;
1967 
1968 	curr_node = btrfs_first_delayed_node(fs_info->delayed_root);
1969 	while (curr_node) {
1970 		__btrfs_kill_delayed_node(curr_node);
1971 
1972 		prev_node = curr_node;
1973 		curr_node = btrfs_next_delayed_node(curr_node);
1974 		btrfs_release_delayed_node(prev_node);
1975 	}
1976 }
1977 
1978