xref: /linux/fs/btrfs/delayed-inode.c (revision c0e297dc61f8d4453e07afbea1fa8d0e67cd4a34)
1 /*
2  * Copyright (C) 2011 Fujitsu.  All rights reserved.
3  * Written by Miao Xie <miaox@cn.fujitsu.com>
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public
7  * License v2 as published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it will be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12  * General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public
15  * License along with this program; if not, write to the
16  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17  * Boston, MA 021110-1307, USA.
18  */
19 
20 #include <linux/slab.h>
21 #include "delayed-inode.h"
22 #include "disk-io.h"
23 #include "transaction.h"
24 #include "ctree.h"
25 
26 #define BTRFS_DELAYED_WRITEBACK		512
27 #define BTRFS_DELAYED_BACKGROUND	128
28 #define BTRFS_DELAYED_BATCH		16
29 
30 static struct kmem_cache *delayed_node_cache;
31 
32 int __init btrfs_delayed_inode_init(void)
33 {
34 	delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
35 					sizeof(struct btrfs_delayed_node),
36 					0,
37 					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
38 					NULL);
39 	if (!delayed_node_cache)
40 		return -ENOMEM;
41 	return 0;
42 }
43 
44 void btrfs_delayed_inode_exit(void)
45 {
46 	if (delayed_node_cache)
47 		kmem_cache_destroy(delayed_node_cache);
48 }
49 
50 static inline void btrfs_init_delayed_node(
51 				struct btrfs_delayed_node *delayed_node,
52 				struct btrfs_root *root, u64 inode_id)
53 {
54 	delayed_node->root = root;
55 	delayed_node->inode_id = inode_id;
56 	atomic_set(&delayed_node->refs, 0);
57 	delayed_node->count = 0;
58 	delayed_node->flags = 0;
59 	delayed_node->ins_root = RB_ROOT;
60 	delayed_node->del_root = RB_ROOT;
61 	mutex_init(&delayed_node->mutex);
62 	delayed_node->index_cnt = 0;
63 	INIT_LIST_HEAD(&delayed_node->n_list);
64 	INIT_LIST_HEAD(&delayed_node->p_list);
65 	delayed_node->bytes_reserved = 0;
66 	memset(&delayed_node->inode_item, 0, sizeof(delayed_node->inode_item));
67 }
68 
69 static inline int btrfs_is_continuous_delayed_item(
70 					struct btrfs_delayed_item *item1,
71 					struct btrfs_delayed_item *item2)
72 {
73 	if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
74 	    item1->key.objectid == item2->key.objectid &&
75 	    item1->key.type == item2->key.type &&
76 	    item1->key.offset + 1 == item2->key.offset)
77 		return 1;
78 	return 0;
79 }
80 
81 static inline struct btrfs_delayed_root *btrfs_get_delayed_root(
82 							struct btrfs_root *root)
83 {
84 	return root->fs_info->delayed_root;
85 }
86 
87 static struct btrfs_delayed_node *btrfs_get_delayed_node(struct inode *inode)
88 {
89 	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
90 	struct btrfs_root *root = btrfs_inode->root;
91 	u64 ino = btrfs_ino(inode);
92 	struct btrfs_delayed_node *node;
93 
94 	node = ACCESS_ONCE(btrfs_inode->delayed_node);
95 	if (node) {
96 		atomic_inc(&node->refs);
97 		return node;
98 	}
99 
100 	spin_lock(&root->inode_lock);
101 	node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
102 	if (node) {
103 		if (btrfs_inode->delayed_node) {
104 			atomic_inc(&node->refs);	/* can be accessed */
105 			BUG_ON(btrfs_inode->delayed_node != node);
106 			spin_unlock(&root->inode_lock);
107 			return node;
108 		}
109 		btrfs_inode->delayed_node = node;
110 		/* can be accessed and cached in the inode */
111 		atomic_add(2, &node->refs);
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 inode *inode)
123 {
124 	struct btrfs_delayed_node *node;
125 	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
126 	struct btrfs_root *root = btrfs_inode->root;
127 	u64 ino = btrfs_ino(inode);
128 	int ret;
129 
130 again:
131 	node = btrfs_get_delayed_node(inode);
132 	if (node)
133 		return node;
134 
135 	node = kmem_cache_alloc(delayed_node_cache, GFP_NOFS);
136 	if (!node)
137 		return ERR_PTR(-ENOMEM);
138 	btrfs_init_delayed_node(node, root, ino);
139 
140 	/* cached in the btrfs inode and can be accessed */
141 	atomic_add(2, &node->refs);
142 
143 	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
144 	if (ret) {
145 		kmem_cache_free(delayed_node_cache, node);
146 		return ERR_PTR(ret);
147 	}
148 
149 	spin_lock(&root->inode_lock);
150 	ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
151 	if (ret == -EEXIST) {
152 		spin_unlock(&root->inode_lock);
153 		kmem_cache_free(delayed_node_cache, node);
154 		radix_tree_preload_end();
155 		goto again;
156 	}
157 	btrfs_inode->delayed_node = node;
158 	spin_unlock(&root->inode_lock);
159 	radix_tree_preload_end();
160 
161 	return node;
162 }
163 
164 /*
165  * Call it when holding delayed_node->mutex
166  *
167  * If mod = 1, add this node into the prepared list.
168  */
169 static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
170 				     struct btrfs_delayed_node *node,
171 				     int mod)
172 {
173 	spin_lock(&root->lock);
174 	if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
175 		if (!list_empty(&node->p_list))
176 			list_move_tail(&node->p_list, &root->prepare_list);
177 		else if (mod)
178 			list_add_tail(&node->p_list, &root->prepare_list);
179 	} else {
180 		list_add_tail(&node->n_list, &root->node_list);
181 		list_add_tail(&node->p_list, &root->prepare_list);
182 		atomic_inc(&node->refs);	/* inserted into list */
183 		root->nodes++;
184 		set_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
185 	}
186 	spin_unlock(&root->lock);
187 }
188 
189 /* Call it when holding delayed_node->mutex */
190 static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
191 				       struct btrfs_delayed_node *node)
192 {
193 	spin_lock(&root->lock);
194 	if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
195 		root->nodes--;
196 		atomic_dec(&node->refs);	/* not in the list */
197 		list_del_init(&node->n_list);
198 		if (!list_empty(&node->p_list))
199 			list_del_init(&node->p_list);
200 		clear_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
201 	}
202 	spin_unlock(&root->lock);
203 }
204 
205 static struct btrfs_delayed_node *btrfs_first_delayed_node(
206 			struct btrfs_delayed_root *delayed_root)
207 {
208 	struct list_head *p;
209 	struct btrfs_delayed_node *node = NULL;
210 
211 	spin_lock(&delayed_root->lock);
212 	if (list_empty(&delayed_root->node_list))
213 		goto out;
214 
215 	p = delayed_root->node_list.next;
216 	node = list_entry(p, struct btrfs_delayed_node, n_list);
217 	atomic_inc(&node->refs);
218 out:
219 	spin_unlock(&delayed_root->lock);
220 
221 	return node;
222 }
223 
224 static struct btrfs_delayed_node *btrfs_next_delayed_node(
225 						struct btrfs_delayed_node *node)
226 {
227 	struct btrfs_delayed_root *delayed_root;
228 	struct list_head *p;
229 	struct btrfs_delayed_node *next = NULL;
230 
231 	delayed_root = node->root->fs_info->delayed_root;
232 	spin_lock(&delayed_root->lock);
233 	if (!test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
234 		/* not in the list */
235 		if (list_empty(&delayed_root->node_list))
236 			goto out;
237 		p = delayed_root->node_list.next;
238 	} else if (list_is_last(&node->n_list, &delayed_root->node_list))
239 		goto out;
240 	else
241 		p = node->n_list.next;
242 
243 	next = list_entry(p, struct btrfs_delayed_node, n_list);
244 	atomic_inc(&next->refs);
245 out:
246 	spin_unlock(&delayed_root->lock);
247 
248 	return next;
249 }
250 
251 static void __btrfs_release_delayed_node(
252 				struct btrfs_delayed_node *delayed_node,
253 				int mod)
254 {
255 	struct btrfs_delayed_root *delayed_root;
256 
257 	if (!delayed_node)
258 		return;
259 
260 	delayed_root = delayed_node->root->fs_info->delayed_root;
261 
262 	mutex_lock(&delayed_node->mutex);
263 	if (delayed_node->count)
264 		btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
265 	else
266 		btrfs_dequeue_delayed_node(delayed_root, delayed_node);
267 	mutex_unlock(&delayed_node->mutex);
268 
269 	if (atomic_dec_and_test(&delayed_node->refs)) {
270 		bool free = false;
271 		struct btrfs_root *root = delayed_node->root;
272 		spin_lock(&root->inode_lock);
273 		if (atomic_read(&delayed_node->refs) == 0) {
274 			radix_tree_delete(&root->delayed_nodes_tree,
275 					  delayed_node->inode_id);
276 			free = true;
277 		}
278 		spin_unlock(&root->inode_lock);
279 		if (free)
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 	atomic_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 		atomic_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 	struct btrfs_delayed_item *item;
395 
396 	item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
397 					   NULL, NULL);
398 	return item;
399 }
400 
401 static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
402 				    struct btrfs_delayed_item *ins,
403 				    int action)
404 {
405 	struct rb_node **p, *node;
406 	struct rb_node *parent_node = NULL;
407 	struct rb_root *root;
408 	struct btrfs_delayed_item *item;
409 	int cmp;
410 
411 	if (action == BTRFS_DELAYED_INSERTION_ITEM)
412 		root = &delayed_node->ins_root;
413 	else if (action == BTRFS_DELAYED_DELETION_ITEM)
414 		root = &delayed_node->del_root;
415 	else
416 		BUG();
417 	p = &root->rb_node;
418 	node = &ins->rb_node;
419 
420 	while (*p) {
421 		parent_node = *p;
422 		item = rb_entry(parent_node, struct btrfs_delayed_item,
423 				 rb_node);
424 
425 		cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
426 		if (cmp < 0)
427 			p = &(*p)->rb_right;
428 		else if (cmp > 0)
429 			p = &(*p)->rb_left;
430 		else
431 			return -EEXIST;
432 	}
433 
434 	rb_link_node(node, parent_node, p);
435 	rb_insert_color(node, root);
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 	if ((atomic_dec_return(&delayed_root->items) <
467 	    BTRFS_DELAYED_BACKGROUND || seq % BTRFS_DELAYED_BATCH == 0) &&
468 	    waitqueue_active(&delayed_root->wait))
469 		wake_up(&delayed_root->wait);
470 }
471 
472 static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
473 {
474 	struct rb_root *root;
475 	struct btrfs_delayed_root *delayed_root;
476 
477 	delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
478 
479 	BUG_ON(!delayed_root);
480 	BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
481 	       delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
482 
483 	if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
484 		root = &delayed_item->delayed_node->ins_root;
485 	else
486 		root = &delayed_item->delayed_node->del_root;
487 
488 	rb_erase(&delayed_item->rb_node, root);
489 	delayed_item->delayed_node->count--;
490 
491 	finish_one_item(delayed_root);
492 }
493 
494 static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
495 {
496 	if (item) {
497 		__btrfs_remove_delayed_item(item);
498 		if (atomic_dec_and_test(&item->refs))
499 			kfree(item);
500 	}
501 }
502 
503 static struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
504 					struct btrfs_delayed_node *delayed_node)
505 {
506 	struct rb_node *p;
507 	struct btrfs_delayed_item *item = NULL;
508 
509 	p = rb_first(&delayed_node->ins_root);
510 	if (p)
511 		item = rb_entry(p, struct btrfs_delayed_item, rb_node);
512 
513 	return item;
514 }
515 
516 static struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
517 					struct btrfs_delayed_node *delayed_node)
518 {
519 	struct rb_node *p;
520 	struct btrfs_delayed_item *item = NULL;
521 
522 	p = rb_first(&delayed_node->del_root);
523 	if (p)
524 		item = rb_entry(p, struct btrfs_delayed_item, rb_node);
525 
526 	return item;
527 }
528 
529 static struct btrfs_delayed_item *__btrfs_next_delayed_item(
530 						struct btrfs_delayed_item *item)
531 {
532 	struct rb_node *p;
533 	struct btrfs_delayed_item *next = NULL;
534 
535 	p = rb_next(&item->rb_node);
536 	if (p)
537 		next = rb_entry(p, struct btrfs_delayed_item, rb_node);
538 
539 	return next;
540 }
541 
542 static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
543 					       struct btrfs_root *root,
544 					       struct btrfs_delayed_item *item)
545 {
546 	struct btrfs_block_rsv *src_rsv;
547 	struct btrfs_block_rsv *dst_rsv;
548 	u64 num_bytes;
549 	int ret;
550 
551 	if (!trans->bytes_reserved)
552 		return 0;
553 
554 	src_rsv = trans->block_rsv;
555 	dst_rsv = &root->fs_info->delayed_block_rsv;
556 
557 	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
558 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
559 	if (!ret) {
560 		trace_btrfs_space_reservation(root->fs_info, "delayed_item",
561 					      item->key.objectid,
562 					      num_bytes, 1);
563 		item->bytes_reserved = num_bytes;
564 	}
565 
566 	return ret;
567 }
568 
569 static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
570 						struct btrfs_delayed_item *item)
571 {
572 	struct btrfs_block_rsv *rsv;
573 
574 	if (!item->bytes_reserved)
575 		return;
576 
577 	rsv = &root->fs_info->delayed_block_rsv;
578 	trace_btrfs_space_reservation(root->fs_info, "delayed_item",
579 				      item->key.objectid, item->bytes_reserved,
580 				      0);
581 	btrfs_block_rsv_release(root, rsv,
582 				item->bytes_reserved);
583 }
584 
585 static int btrfs_delayed_inode_reserve_metadata(
586 					struct btrfs_trans_handle *trans,
587 					struct btrfs_root *root,
588 					struct inode *inode,
589 					struct btrfs_delayed_node *node)
590 {
591 	struct btrfs_block_rsv *src_rsv;
592 	struct btrfs_block_rsv *dst_rsv;
593 	u64 num_bytes;
594 	int ret;
595 	bool release = false;
596 
597 	src_rsv = trans->block_rsv;
598 	dst_rsv = &root->fs_info->delayed_block_rsv;
599 
600 	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
601 
602 	/*
603 	 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
604 	 * which doesn't reserve space for speed.  This is a problem since we
605 	 * still need to reserve space for this update, so try to reserve the
606 	 * space.
607 	 *
608 	 * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
609 	 * we're accounted for.
610 	 */
611 	if (!src_rsv || (!trans->bytes_reserved &&
612 			 src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) {
613 		ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
614 					  BTRFS_RESERVE_NO_FLUSH);
615 		/*
616 		 * Since we're under a transaction reserve_metadata_bytes could
617 		 * try to commit the transaction which will make it return
618 		 * EAGAIN to make us stop the transaction we have, so return
619 		 * ENOSPC instead so that btrfs_dirty_inode knows what to do.
620 		 */
621 		if (ret == -EAGAIN)
622 			ret = -ENOSPC;
623 		if (!ret) {
624 			node->bytes_reserved = num_bytes;
625 			trace_btrfs_space_reservation(root->fs_info,
626 						      "delayed_inode",
627 						      btrfs_ino(inode),
628 						      num_bytes, 1);
629 		}
630 		return ret;
631 	} else if (src_rsv->type == BTRFS_BLOCK_RSV_DELALLOC) {
632 		spin_lock(&BTRFS_I(inode)->lock);
633 		if (test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
634 				       &BTRFS_I(inode)->runtime_flags)) {
635 			spin_unlock(&BTRFS_I(inode)->lock);
636 			release = true;
637 			goto migrate;
638 		}
639 		spin_unlock(&BTRFS_I(inode)->lock);
640 
641 		/* Ok we didn't have space pre-reserved.  This shouldn't happen
642 		 * too often but it can happen if we do delalloc to an existing
643 		 * inode which gets dirtied because of the time update, and then
644 		 * isn't touched again until after the transaction commits and
645 		 * then we try to write out the data.  First try to be nice and
646 		 * reserve something strictly for us.  If not be a pain and try
647 		 * to steal from the delalloc block rsv.
648 		 */
649 		ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
650 					  BTRFS_RESERVE_NO_FLUSH);
651 		if (!ret)
652 			goto out;
653 
654 		ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
655 		if (!WARN_ON(ret))
656 			goto out;
657 
658 		/*
659 		 * Ok this is a problem, let's just steal from the global rsv
660 		 * since this really shouldn't happen that often.
661 		 */
662 		ret = btrfs_block_rsv_migrate(&root->fs_info->global_block_rsv,
663 					      dst_rsv, num_bytes);
664 		goto out;
665 	}
666 
667 migrate:
668 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
669 
670 out:
671 	/*
672 	 * Migrate only takes a reservation, it doesn't touch the size of the
673 	 * block_rsv.  This is to simplify people who don't normally have things
674 	 * migrated from their block rsv.  If they go to release their
675 	 * reservation, that will decrease the size as well, so if migrate
676 	 * reduced size we'd end up with a negative size.  But for the
677 	 * delalloc_meta_reserved stuff we will only know to drop 1 reservation,
678 	 * but we could in fact do this reserve/migrate dance several times
679 	 * between the time we did the original reservation and we'd clean it
680 	 * up.  So to take care of this, release the space for the meta
681 	 * reservation here.  I think it may be time for a documentation page on
682 	 * how block rsvs. work.
683 	 */
684 	if (!ret) {
685 		trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
686 					      btrfs_ino(inode), num_bytes, 1);
687 		node->bytes_reserved = num_bytes;
688 	}
689 
690 	if (release) {
691 		trace_btrfs_space_reservation(root->fs_info, "delalloc",
692 					      btrfs_ino(inode), num_bytes, 0);
693 		btrfs_block_rsv_release(root, src_rsv, num_bytes);
694 	}
695 
696 	return ret;
697 }
698 
699 static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
700 						struct btrfs_delayed_node *node)
701 {
702 	struct btrfs_block_rsv *rsv;
703 
704 	if (!node->bytes_reserved)
705 		return;
706 
707 	rsv = &root->fs_info->delayed_block_rsv;
708 	trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
709 				      node->inode_id, node->bytes_reserved, 0);
710 	btrfs_block_rsv_release(root, rsv,
711 				node->bytes_reserved);
712 	node->bytes_reserved = 0;
713 }
714 
715 /*
716  * This helper will insert some continuous items into the same leaf according
717  * to the free space of the leaf.
718  */
719 static int btrfs_batch_insert_items(struct btrfs_root *root,
720 				    struct btrfs_path *path,
721 				    struct btrfs_delayed_item *item)
722 {
723 	struct btrfs_delayed_item *curr, *next;
724 	int free_space;
725 	int total_data_size = 0, total_size = 0;
726 	struct extent_buffer *leaf;
727 	char *data_ptr;
728 	struct btrfs_key *keys;
729 	u32 *data_size;
730 	struct list_head head;
731 	int slot;
732 	int nitems;
733 	int i;
734 	int ret = 0;
735 
736 	BUG_ON(!path->nodes[0]);
737 
738 	leaf = path->nodes[0];
739 	free_space = btrfs_leaf_free_space(root, leaf);
740 	INIT_LIST_HEAD(&head);
741 
742 	next = item;
743 	nitems = 0;
744 
745 	/*
746 	 * count the number of the continuous items that we can insert in batch
747 	 */
748 	while (total_size + next->data_len + sizeof(struct btrfs_item) <=
749 	       free_space) {
750 		total_data_size += next->data_len;
751 		total_size += next->data_len + sizeof(struct btrfs_item);
752 		list_add_tail(&next->tree_list, &head);
753 		nitems++;
754 
755 		curr = next;
756 		next = __btrfs_next_delayed_item(curr);
757 		if (!next)
758 			break;
759 
760 		if (!btrfs_is_continuous_delayed_item(curr, next))
761 			break;
762 	}
763 
764 	if (!nitems) {
765 		ret = 0;
766 		goto out;
767 	}
768 
769 	/*
770 	 * we need allocate some memory space, but it might cause the task
771 	 * to sleep, so we set all locked nodes in the path to blocking locks
772 	 * first.
773 	 */
774 	btrfs_set_path_blocking(path);
775 
776 	keys = kmalloc_array(nitems, sizeof(struct btrfs_key), GFP_NOFS);
777 	if (!keys) {
778 		ret = -ENOMEM;
779 		goto out;
780 	}
781 
782 	data_size = kmalloc_array(nitems, sizeof(u32), GFP_NOFS);
783 	if (!data_size) {
784 		ret = -ENOMEM;
785 		goto error;
786 	}
787 
788 	/* get keys of all the delayed items */
789 	i = 0;
790 	list_for_each_entry(next, &head, tree_list) {
791 		keys[i] = next->key;
792 		data_size[i] = next->data_len;
793 		i++;
794 	}
795 
796 	/* reset all the locked nodes in the patch to spinning locks. */
797 	btrfs_clear_path_blocking(path, NULL, 0);
798 
799 	/* insert the keys of the items */
800 	setup_items_for_insert(root, path, keys, data_size,
801 			       total_data_size, total_size, nitems);
802 
803 	/* insert the dir index items */
804 	slot = path->slots[0];
805 	list_for_each_entry_safe(curr, next, &head, tree_list) {
806 		data_ptr = btrfs_item_ptr(leaf, slot, char);
807 		write_extent_buffer(leaf, &curr->data,
808 				    (unsigned long)data_ptr,
809 				    curr->data_len);
810 		slot++;
811 
812 		btrfs_delayed_item_release_metadata(root, curr);
813 
814 		list_del(&curr->tree_list);
815 		btrfs_release_delayed_item(curr);
816 	}
817 
818 error:
819 	kfree(data_size);
820 	kfree(keys);
821 out:
822 	return ret;
823 }
824 
825 /*
826  * This helper can just do simple insertion that needn't extend item for new
827  * data, such as directory name index insertion, inode insertion.
828  */
829 static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
830 				     struct btrfs_root *root,
831 				     struct btrfs_path *path,
832 				     struct btrfs_delayed_item *delayed_item)
833 {
834 	struct extent_buffer *leaf;
835 	char *ptr;
836 	int ret;
837 
838 	ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
839 				      delayed_item->data_len);
840 	if (ret < 0 && ret != -EEXIST)
841 		return ret;
842 
843 	leaf = path->nodes[0];
844 
845 	ptr = btrfs_item_ptr(leaf, path->slots[0], char);
846 
847 	write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
848 			    delayed_item->data_len);
849 	btrfs_mark_buffer_dirty(leaf);
850 
851 	btrfs_delayed_item_release_metadata(root, delayed_item);
852 	return 0;
853 }
854 
855 /*
856  * we insert an item first, then if there are some continuous items, we try
857  * to insert those items into the same leaf.
858  */
859 static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
860 				      struct btrfs_path *path,
861 				      struct btrfs_root *root,
862 				      struct btrfs_delayed_node *node)
863 {
864 	struct btrfs_delayed_item *curr, *prev;
865 	int ret = 0;
866 
867 do_again:
868 	mutex_lock(&node->mutex);
869 	curr = __btrfs_first_delayed_insertion_item(node);
870 	if (!curr)
871 		goto insert_end;
872 
873 	ret = btrfs_insert_delayed_item(trans, root, path, curr);
874 	if (ret < 0) {
875 		btrfs_release_path(path);
876 		goto insert_end;
877 	}
878 
879 	prev = curr;
880 	curr = __btrfs_next_delayed_item(prev);
881 	if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
882 		/* insert the continuous items into the same leaf */
883 		path->slots[0]++;
884 		btrfs_batch_insert_items(root, path, curr);
885 	}
886 	btrfs_release_delayed_item(prev);
887 	btrfs_mark_buffer_dirty(path->nodes[0]);
888 
889 	btrfs_release_path(path);
890 	mutex_unlock(&node->mutex);
891 	goto do_again;
892 
893 insert_end:
894 	mutex_unlock(&node->mutex);
895 	return ret;
896 }
897 
898 static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
899 				    struct btrfs_root *root,
900 				    struct btrfs_path *path,
901 				    struct btrfs_delayed_item *item)
902 {
903 	struct btrfs_delayed_item *curr, *next;
904 	struct extent_buffer *leaf;
905 	struct btrfs_key key;
906 	struct list_head head;
907 	int nitems, i, last_item;
908 	int ret = 0;
909 
910 	BUG_ON(!path->nodes[0]);
911 
912 	leaf = path->nodes[0];
913 
914 	i = path->slots[0];
915 	last_item = btrfs_header_nritems(leaf) - 1;
916 	if (i > last_item)
917 		return -ENOENT;	/* FIXME: Is errno suitable? */
918 
919 	next = item;
920 	INIT_LIST_HEAD(&head);
921 	btrfs_item_key_to_cpu(leaf, &key, i);
922 	nitems = 0;
923 	/*
924 	 * count the number of the dir index items that we can delete in batch
925 	 */
926 	while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
927 		list_add_tail(&next->tree_list, &head);
928 		nitems++;
929 
930 		curr = next;
931 		next = __btrfs_next_delayed_item(curr);
932 		if (!next)
933 			break;
934 
935 		if (!btrfs_is_continuous_delayed_item(curr, next))
936 			break;
937 
938 		i++;
939 		if (i > last_item)
940 			break;
941 		btrfs_item_key_to_cpu(leaf, &key, i);
942 	}
943 
944 	if (!nitems)
945 		return 0;
946 
947 	ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
948 	if (ret)
949 		goto out;
950 
951 	list_for_each_entry_safe(curr, next, &head, tree_list) {
952 		btrfs_delayed_item_release_metadata(root, curr);
953 		list_del(&curr->tree_list);
954 		btrfs_release_delayed_item(curr);
955 	}
956 
957 out:
958 	return ret;
959 }
960 
961 static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
962 				      struct btrfs_path *path,
963 				      struct btrfs_root *root,
964 				      struct btrfs_delayed_node *node)
965 {
966 	struct btrfs_delayed_item *curr, *prev;
967 	int ret = 0;
968 
969 do_again:
970 	mutex_lock(&node->mutex);
971 	curr = __btrfs_first_delayed_deletion_item(node);
972 	if (!curr)
973 		goto delete_fail;
974 
975 	ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
976 	if (ret < 0)
977 		goto delete_fail;
978 	else if (ret > 0) {
979 		/*
980 		 * can't find the item which the node points to, so this node
981 		 * is invalid, just drop it.
982 		 */
983 		prev = curr;
984 		curr = __btrfs_next_delayed_item(prev);
985 		btrfs_release_delayed_item(prev);
986 		ret = 0;
987 		btrfs_release_path(path);
988 		if (curr) {
989 			mutex_unlock(&node->mutex);
990 			goto do_again;
991 		} else
992 			goto delete_fail;
993 	}
994 
995 	btrfs_batch_delete_items(trans, root, path, curr);
996 	btrfs_release_path(path);
997 	mutex_unlock(&node->mutex);
998 	goto do_again;
999 
1000 delete_fail:
1001 	btrfs_release_path(path);
1002 	mutex_unlock(&node->mutex);
1003 	return ret;
1004 }
1005 
1006 static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
1007 {
1008 	struct btrfs_delayed_root *delayed_root;
1009 
1010 	if (delayed_node &&
1011 	    test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1012 		BUG_ON(!delayed_node->root);
1013 		clear_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
1014 		delayed_node->count--;
1015 
1016 		delayed_root = delayed_node->root->fs_info->delayed_root;
1017 		finish_one_item(delayed_root);
1018 	}
1019 }
1020 
1021 static void btrfs_release_delayed_iref(struct btrfs_delayed_node *delayed_node)
1022 {
1023 	struct btrfs_delayed_root *delayed_root;
1024 
1025 	ASSERT(delayed_node->root);
1026 	clear_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
1027 	delayed_node->count--;
1028 
1029 	delayed_root = delayed_node->root->fs_info->delayed_root;
1030 	finish_one_item(delayed_root);
1031 }
1032 
1033 static int __btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1034 					struct btrfs_root *root,
1035 					struct btrfs_path *path,
1036 					struct btrfs_delayed_node *node)
1037 {
1038 	struct btrfs_key key;
1039 	struct btrfs_inode_item *inode_item;
1040 	struct extent_buffer *leaf;
1041 	int mod;
1042 	int ret;
1043 
1044 	key.objectid = node->inode_id;
1045 	key.type = BTRFS_INODE_ITEM_KEY;
1046 	key.offset = 0;
1047 
1048 	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
1049 		mod = -1;
1050 	else
1051 		mod = 1;
1052 
1053 	ret = btrfs_lookup_inode(trans, root, path, &key, mod);
1054 	if (ret > 0) {
1055 		btrfs_release_path(path);
1056 		return -ENOENT;
1057 	} else if (ret < 0) {
1058 		return ret;
1059 	}
1060 
1061 	leaf = path->nodes[0];
1062 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
1063 				    struct btrfs_inode_item);
1064 	write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
1065 			    sizeof(struct btrfs_inode_item));
1066 	btrfs_mark_buffer_dirty(leaf);
1067 
1068 	if (!test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
1069 		goto no_iref;
1070 
1071 	path->slots[0]++;
1072 	if (path->slots[0] >= btrfs_header_nritems(leaf))
1073 		goto search;
1074 again:
1075 	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1076 	if (key.objectid != node->inode_id)
1077 		goto out;
1078 
1079 	if (key.type != BTRFS_INODE_REF_KEY &&
1080 	    key.type != BTRFS_INODE_EXTREF_KEY)
1081 		goto out;
1082 
1083 	/*
1084 	 * Delayed iref deletion is for the inode who has only one link,
1085 	 * so there is only one iref. The case that several irefs are
1086 	 * in the same item doesn't exist.
1087 	 */
1088 	btrfs_del_item(trans, root, path);
1089 out:
1090 	btrfs_release_delayed_iref(node);
1091 no_iref:
1092 	btrfs_release_path(path);
1093 err_out:
1094 	btrfs_delayed_inode_release_metadata(root, node);
1095 	btrfs_release_delayed_inode(node);
1096 
1097 	return ret;
1098 
1099 search:
1100 	btrfs_release_path(path);
1101 
1102 	key.type = BTRFS_INODE_EXTREF_KEY;
1103 	key.offset = -1;
1104 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1105 	if (ret < 0)
1106 		goto err_out;
1107 	ASSERT(ret);
1108 
1109 	ret = 0;
1110 	leaf = path->nodes[0];
1111 	path->slots[0]--;
1112 	goto again;
1113 }
1114 
1115 static inline int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1116 					     struct btrfs_root *root,
1117 					     struct btrfs_path *path,
1118 					     struct btrfs_delayed_node *node)
1119 {
1120 	int ret;
1121 
1122 	mutex_lock(&node->mutex);
1123 	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &node->flags)) {
1124 		mutex_unlock(&node->mutex);
1125 		return 0;
1126 	}
1127 
1128 	ret = __btrfs_update_delayed_inode(trans, root, path, node);
1129 	mutex_unlock(&node->mutex);
1130 	return ret;
1131 }
1132 
1133 static inline int
1134 __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1135 				   struct btrfs_path *path,
1136 				   struct btrfs_delayed_node *node)
1137 {
1138 	int ret;
1139 
1140 	ret = btrfs_insert_delayed_items(trans, path, node->root, node);
1141 	if (ret)
1142 		return ret;
1143 
1144 	ret = btrfs_delete_delayed_items(trans, path, node->root, node);
1145 	if (ret)
1146 		return ret;
1147 
1148 	ret = btrfs_update_delayed_inode(trans, node->root, path, node);
1149 	return ret;
1150 }
1151 
1152 /*
1153  * Called when committing the transaction.
1154  * Returns 0 on success.
1155  * Returns < 0 on error and returns with an aborted transaction with any
1156  * outstanding delayed items cleaned up.
1157  */
1158 static int __btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1159 				     struct btrfs_root *root, int nr)
1160 {
1161 	struct btrfs_delayed_root *delayed_root;
1162 	struct btrfs_delayed_node *curr_node, *prev_node;
1163 	struct btrfs_path *path;
1164 	struct btrfs_block_rsv *block_rsv;
1165 	int ret = 0;
1166 	bool count = (nr > 0);
1167 
1168 	if (trans->aborted)
1169 		return -EIO;
1170 
1171 	path = btrfs_alloc_path();
1172 	if (!path)
1173 		return -ENOMEM;
1174 	path->leave_spinning = 1;
1175 
1176 	block_rsv = trans->block_rsv;
1177 	trans->block_rsv = &root->fs_info->delayed_block_rsv;
1178 
1179 	delayed_root = btrfs_get_delayed_root(root);
1180 
1181 	curr_node = btrfs_first_delayed_node(delayed_root);
1182 	while (curr_node && (!count || (count && nr--))) {
1183 		ret = __btrfs_commit_inode_delayed_items(trans, path,
1184 							 curr_node);
1185 		if (ret) {
1186 			btrfs_release_delayed_node(curr_node);
1187 			curr_node = NULL;
1188 			btrfs_abort_transaction(trans, root, ret);
1189 			break;
1190 		}
1191 
1192 		prev_node = curr_node;
1193 		curr_node = btrfs_next_delayed_node(curr_node);
1194 		btrfs_release_delayed_node(prev_node);
1195 	}
1196 
1197 	if (curr_node)
1198 		btrfs_release_delayed_node(curr_node);
1199 	btrfs_free_path(path);
1200 	trans->block_rsv = block_rsv;
1201 
1202 	return ret;
1203 }
1204 
1205 int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1206 			    struct btrfs_root *root)
1207 {
1208 	return __btrfs_run_delayed_items(trans, root, -1);
1209 }
1210 
1211 int btrfs_run_delayed_items_nr(struct btrfs_trans_handle *trans,
1212 			       struct btrfs_root *root, int nr)
1213 {
1214 	return __btrfs_run_delayed_items(trans, root, nr);
1215 }
1216 
1217 int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1218 				     struct inode *inode)
1219 {
1220 	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1221 	struct btrfs_path *path;
1222 	struct btrfs_block_rsv *block_rsv;
1223 	int ret;
1224 
1225 	if (!delayed_node)
1226 		return 0;
1227 
1228 	mutex_lock(&delayed_node->mutex);
1229 	if (!delayed_node->count) {
1230 		mutex_unlock(&delayed_node->mutex);
1231 		btrfs_release_delayed_node(delayed_node);
1232 		return 0;
1233 	}
1234 	mutex_unlock(&delayed_node->mutex);
1235 
1236 	path = btrfs_alloc_path();
1237 	if (!path) {
1238 		btrfs_release_delayed_node(delayed_node);
1239 		return -ENOMEM;
1240 	}
1241 	path->leave_spinning = 1;
1242 
1243 	block_rsv = trans->block_rsv;
1244 	trans->block_rsv = &delayed_node->root->fs_info->delayed_block_rsv;
1245 
1246 	ret = __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
1247 
1248 	btrfs_release_delayed_node(delayed_node);
1249 	btrfs_free_path(path);
1250 	trans->block_rsv = block_rsv;
1251 
1252 	return ret;
1253 }
1254 
1255 int btrfs_commit_inode_delayed_inode(struct inode *inode)
1256 {
1257 	struct btrfs_trans_handle *trans;
1258 	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1259 	struct btrfs_path *path;
1260 	struct btrfs_block_rsv *block_rsv;
1261 	int ret;
1262 
1263 	if (!delayed_node)
1264 		return 0;
1265 
1266 	mutex_lock(&delayed_node->mutex);
1267 	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1268 		mutex_unlock(&delayed_node->mutex);
1269 		btrfs_release_delayed_node(delayed_node);
1270 		return 0;
1271 	}
1272 	mutex_unlock(&delayed_node->mutex);
1273 
1274 	trans = btrfs_join_transaction(delayed_node->root);
1275 	if (IS_ERR(trans)) {
1276 		ret = PTR_ERR(trans);
1277 		goto out;
1278 	}
1279 
1280 	path = btrfs_alloc_path();
1281 	if (!path) {
1282 		ret = -ENOMEM;
1283 		goto trans_out;
1284 	}
1285 	path->leave_spinning = 1;
1286 
1287 	block_rsv = trans->block_rsv;
1288 	trans->block_rsv = &delayed_node->root->fs_info->delayed_block_rsv;
1289 
1290 	mutex_lock(&delayed_node->mutex);
1291 	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags))
1292 		ret = __btrfs_update_delayed_inode(trans, delayed_node->root,
1293 						   path, delayed_node);
1294 	else
1295 		ret = 0;
1296 	mutex_unlock(&delayed_node->mutex);
1297 
1298 	btrfs_free_path(path);
1299 	trans->block_rsv = block_rsv;
1300 trans_out:
1301 	btrfs_end_transaction(trans, delayed_node->root);
1302 	btrfs_btree_balance_dirty(delayed_node->root);
1303 out:
1304 	btrfs_release_delayed_node(delayed_node);
1305 
1306 	return ret;
1307 }
1308 
1309 void btrfs_remove_delayed_node(struct inode *inode)
1310 {
1311 	struct btrfs_delayed_node *delayed_node;
1312 
1313 	delayed_node = ACCESS_ONCE(BTRFS_I(inode)->delayed_node);
1314 	if (!delayed_node)
1315 		return;
1316 
1317 	BTRFS_I(inode)->delayed_node = NULL;
1318 	btrfs_release_delayed_node(delayed_node);
1319 }
1320 
1321 struct btrfs_async_delayed_work {
1322 	struct btrfs_delayed_root *delayed_root;
1323 	int nr;
1324 	struct btrfs_work work;
1325 };
1326 
1327 static void btrfs_async_run_delayed_root(struct btrfs_work *work)
1328 {
1329 	struct btrfs_async_delayed_work *async_work;
1330 	struct btrfs_delayed_root *delayed_root;
1331 	struct btrfs_trans_handle *trans;
1332 	struct btrfs_path *path;
1333 	struct btrfs_delayed_node *delayed_node = NULL;
1334 	struct btrfs_root *root;
1335 	struct btrfs_block_rsv *block_rsv;
1336 	int total_done = 0;
1337 
1338 	async_work = container_of(work, struct btrfs_async_delayed_work, work);
1339 	delayed_root = async_work->delayed_root;
1340 
1341 	path = btrfs_alloc_path();
1342 	if (!path)
1343 		goto out;
1344 
1345 again:
1346 	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND / 2)
1347 		goto free_path;
1348 
1349 	delayed_node = btrfs_first_prepared_delayed_node(delayed_root);
1350 	if (!delayed_node)
1351 		goto free_path;
1352 
1353 	path->leave_spinning = 1;
1354 	root = delayed_node->root;
1355 
1356 	trans = btrfs_join_transaction(root);
1357 	if (IS_ERR(trans))
1358 		goto release_path;
1359 
1360 	block_rsv = trans->block_rsv;
1361 	trans->block_rsv = &root->fs_info->delayed_block_rsv;
1362 
1363 	__btrfs_commit_inode_delayed_items(trans, path, delayed_node);
1364 
1365 	trans->block_rsv = block_rsv;
1366 	btrfs_end_transaction(trans, root);
1367 	btrfs_btree_balance_dirty_nodelay(root);
1368 
1369 release_path:
1370 	btrfs_release_path(path);
1371 	total_done++;
1372 
1373 	btrfs_release_prepared_delayed_node(delayed_node);
1374 	if (async_work->nr == 0 || total_done < async_work->nr)
1375 		goto again;
1376 
1377 free_path:
1378 	btrfs_free_path(path);
1379 out:
1380 	wake_up(&delayed_root->wait);
1381 	kfree(async_work);
1382 }
1383 
1384 
1385 static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
1386 				     struct btrfs_fs_info *fs_info, int nr)
1387 {
1388 	struct btrfs_async_delayed_work *async_work;
1389 
1390 	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1391 		return 0;
1392 
1393 	async_work = kmalloc(sizeof(*async_work), GFP_NOFS);
1394 	if (!async_work)
1395 		return -ENOMEM;
1396 
1397 	async_work->delayed_root = delayed_root;
1398 	btrfs_init_work(&async_work->work, btrfs_delayed_meta_helper,
1399 			btrfs_async_run_delayed_root, NULL, NULL);
1400 	async_work->nr = nr;
1401 
1402 	btrfs_queue_work(fs_info->delayed_workers, &async_work->work);
1403 	return 0;
1404 }
1405 
1406 void btrfs_assert_delayed_root_empty(struct btrfs_root *root)
1407 {
1408 	struct btrfs_delayed_root *delayed_root;
1409 	delayed_root = btrfs_get_delayed_root(root);
1410 	WARN_ON(btrfs_first_delayed_node(delayed_root));
1411 }
1412 
1413 static int could_end_wait(struct btrfs_delayed_root *delayed_root, int seq)
1414 {
1415 	int val = atomic_read(&delayed_root->items_seq);
1416 
1417 	if (val < seq || val >= seq + BTRFS_DELAYED_BATCH)
1418 		return 1;
1419 
1420 	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1421 		return 1;
1422 
1423 	return 0;
1424 }
1425 
1426 void btrfs_balance_delayed_items(struct btrfs_root *root)
1427 {
1428 	struct btrfs_delayed_root *delayed_root;
1429 	struct btrfs_fs_info *fs_info = root->fs_info;
1430 
1431 	delayed_root = btrfs_get_delayed_root(root);
1432 
1433 	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1434 		return;
1435 
1436 	if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
1437 		int seq;
1438 		int ret;
1439 
1440 		seq = atomic_read(&delayed_root->items_seq);
1441 
1442 		ret = btrfs_wq_run_delayed_node(delayed_root, fs_info, 0);
1443 		if (ret)
1444 			return;
1445 
1446 		wait_event_interruptible(delayed_root->wait,
1447 					 could_end_wait(delayed_root, seq));
1448 		return;
1449 	}
1450 
1451 	btrfs_wq_run_delayed_node(delayed_root, fs_info, BTRFS_DELAYED_BATCH);
1452 }
1453 
1454 /* Will return 0 or -ENOMEM */
1455 int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
1456 				   struct btrfs_root *root, const char *name,
1457 				   int name_len, struct inode *dir,
1458 				   struct btrfs_disk_key *disk_key, u8 type,
1459 				   u64 index)
1460 {
1461 	struct btrfs_delayed_node *delayed_node;
1462 	struct btrfs_delayed_item *delayed_item;
1463 	struct btrfs_dir_item *dir_item;
1464 	int ret;
1465 
1466 	delayed_node = btrfs_get_or_create_delayed_node(dir);
1467 	if (IS_ERR(delayed_node))
1468 		return PTR_ERR(delayed_node);
1469 
1470 	delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
1471 	if (!delayed_item) {
1472 		ret = -ENOMEM;
1473 		goto release_node;
1474 	}
1475 
1476 	delayed_item->key.objectid = btrfs_ino(dir);
1477 	delayed_item->key.type = BTRFS_DIR_INDEX_KEY;
1478 	delayed_item->key.offset = index;
1479 
1480 	dir_item = (struct btrfs_dir_item *)delayed_item->data;
1481 	dir_item->location = *disk_key;
1482 	btrfs_set_stack_dir_transid(dir_item, trans->transid);
1483 	btrfs_set_stack_dir_data_len(dir_item, 0);
1484 	btrfs_set_stack_dir_name_len(dir_item, name_len);
1485 	btrfs_set_stack_dir_type(dir_item, type);
1486 	memcpy((char *)(dir_item + 1), name, name_len);
1487 
1488 	ret = btrfs_delayed_item_reserve_metadata(trans, root, delayed_item);
1489 	/*
1490 	 * we have reserved enough space when we start a new transaction,
1491 	 * so reserving metadata failure is impossible
1492 	 */
1493 	BUG_ON(ret);
1494 
1495 
1496 	mutex_lock(&delayed_node->mutex);
1497 	ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
1498 	if (unlikely(ret)) {
1499 		btrfs_err(root->fs_info, "err add delayed dir index item(name: %.*s) "
1500 				"into the insertion tree of the delayed node"
1501 				"(root id: %llu, inode id: %llu, errno: %d)",
1502 				name_len, name, delayed_node->root->objectid,
1503 				delayed_node->inode_id, ret);
1504 		BUG();
1505 	}
1506 	mutex_unlock(&delayed_node->mutex);
1507 
1508 release_node:
1509 	btrfs_release_delayed_node(delayed_node);
1510 	return ret;
1511 }
1512 
1513 static int btrfs_delete_delayed_insertion_item(struct btrfs_root *root,
1514 					       struct btrfs_delayed_node *node,
1515 					       struct btrfs_key *key)
1516 {
1517 	struct btrfs_delayed_item *item;
1518 
1519 	mutex_lock(&node->mutex);
1520 	item = __btrfs_lookup_delayed_insertion_item(node, key);
1521 	if (!item) {
1522 		mutex_unlock(&node->mutex);
1523 		return 1;
1524 	}
1525 
1526 	btrfs_delayed_item_release_metadata(root, item);
1527 	btrfs_release_delayed_item(item);
1528 	mutex_unlock(&node->mutex);
1529 	return 0;
1530 }
1531 
1532 int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
1533 				   struct btrfs_root *root, struct inode *dir,
1534 				   u64 index)
1535 {
1536 	struct btrfs_delayed_node *node;
1537 	struct btrfs_delayed_item *item;
1538 	struct btrfs_key item_key;
1539 	int ret;
1540 
1541 	node = btrfs_get_or_create_delayed_node(dir);
1542 	if (IS_ERR(node))
1543 		return PTR_ERR(node);
1544 
1545 	item_key.objectid = btrfs_ino(dir);
1546 	item_key.type = BTRFS_DIR_INDEX_KEY;
1547 	item_key.offset = index;
1548 
1549 	ret = btrfs_delete_delayed_insertion_item(root, node, &item_key);
1550 	if (!ret)
1551 		goto end;
1552 
1553 	item = btrfs_alloc_delayed_item(0);
1554 	if (!item) {
1555 		ret = -ENOMEM;
1556 		goto end;
1557 	}
1558 
1559 	item->key = item_key;
1560 
1561 	ret = btrfs_delayed_item_reserve_metadata(trans, root, item);
1562 	/*
1563 	 * we have reserved enough space when we start a new transaction,
1564 	 * so reserving metadata failure is impossible.
1565 	 */
1566 	BUG_ON(ret);
1567 
1568 	mutex_lock(&node->mutex);
1569 	ret = __btrfs_add_delayed_deletion_item(node, item);
1570 	if (unlikely(ret)) {
1571 		btrfs_err(root->fs_info, "err add delayed dir index item(index: %llu) "
1572 				"into the deletion tree of the delayed node"
1573 				"(root id: %llu, inode id: %llu, errno: %d)",
1574 				index, node->root->objectid, node->inode_id,
1575 				ret);
1576 		BUG();
1577 	}
1578 	mutex_unlock(&node->mutex);
1579 end:
1580 	btrfs_release_delayed_node(node);
1581 	return ret;
1582 }
1583 
1584 int btrfs_inode_delayed_dir_index_count(struct inode *inode)
1585 {
1586 	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1587 
1588 	if (!delayed_node)
1589 		return -ENOENT;
1590 
1591 	/*
1592 	 * Since we have held i_mutex of this directory, it is impossible that
1593 	 * a new directory index is added into the delayed node and index_cnt
1594 	 * is updated now. So we needn't lock the delayed node.
1595 	 */
1596 	if (!delayed_node->index_cnt) {
1597 		btrfs_release_delayed_node(delayed_node);
1598 		return -EINVAL;
1599 	}
1600 
1601 	BTRFS_I(inode)->index_cnt = delayed_node->index_cnt;
1602 	btrfs_release_delayed_node(delayed_node);
1603 	return 0;
1604 }
1605 
1606 void btrfs_get_delayed_items(struct inode *inode, struct list_head *ins_list,
1607 			     struct list_head *del_list)
1608 {
1609 	struct btrfs_delayed_node *delayed_node;
1610 	struct btrfs_delayed_item *item;
1611 
1612 	delayed_node = btrfs_get_delayed_node(inode);
1613 	if (!delayed_node)
1614 		return;
1615 
1616 	mutex_lock(&delayed_node->mutex);
1617 	item = __btrfs_first_delayed_insertion_item(delayed_node);
1618 	while (item) {
1619 		atomic_inc(&item->refs);
1620 		list_add_tail(&item->readdir_list, ins_list);
1621 		item = __btrfs_next_delayed_item(item);
1622 	}
1623 
1624 	item = __btrfs_first_delayed_deletion_item(delayed_node);
1625 	while (item) {
1626 		atomic_inc(&item->refs);
1627 		list_add_tail(&item->readdir_list, del_list);
1628 		item = __btrfs_next_delayed_item(item);
1629 	}
1630 	mutex_unlock(&delayed_node->mutex);
1631 	/*
1632 	 * This delayed node is still cached in the btrfs inode, so refs
1633 	 * must be > 1 now, and we needn't check it is going to be freed
1634 	 * or not.
1635 	 *
1636 	 * Besides that, this function is used to read dir, we do not
1637 	 * insert/delete delayed items in this period. So we also needn't
1638 	 * requeue or dequeue this delayed node.
1639 	 */
1640 	atomic_dec(&delayed_node->refs);
1641 }
1642 
1643 void btrfs_put_delayed_items(struct list_head *ins_list,
1644 			     struct list_head *del_list)
1645 {
1646 	struct btrfs_delayed_item *curr, *next;
1647 
1648 	list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1649 		list_del(&curr->readdir_list);
1650 		if (atomic_dec_and_test(&curr->refs))
1651 			kfree(curr);
1652 	}
1653 
1654 	list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1655 		list_del(&curr->readdir_list);
1656 		if (atomic_dec_and_test(&curr->refs))
1657 			kfree(curr);
1658 	}
1659 }
1660 
1661 int btrfs_should_delete_dir_index(struct list_head *del_list,
1662 				  u64 index)
1663 {
1664 	struct btrfs_delayed_item *curr, *next;
1665 	int ret;
1666 
1667 	if (list_empty(del_list))
1668 		return 0;
1669 
1670 	list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1671 		if (curr->key.offset > index)
1672 			break;
1673 
1674 		list_del(&curr->readdir_list);
1675 		ret = (curr->key.offset == index);
1676 
1677 		if (atomic_dec_and_test(&curr->refs))
1678 			kfree(curr);
1679 
1680 		if (ret)
1681 			return 1;
1682 		else
1683 			continue;
1684 	}
1685 	return 0;
1686 }
1687 
1688 /*
1689  * btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
1690  *
1691  */
1692 int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
1693 				    struct list_head *ins_list)
1694 {
1695 	struct btrfs_dir_item *di;
1696 	struct btrfs_delayed_item *curr, *next;
1697 	struct btrfs_key location;
1698 	char *name;
1699 	int name_len;
1700 	int over = 0;
1701 	unsigned char d_type;
1702 
1703 	if (list_empty(ins_list))
1704 		return 0;
1705 
1706 	/*
1707 	 * Changing the data of the delayed item is impossible. So
1708 	 * we needn't lock them. And we have held i_mutex of the
1709 	 * directory, nobody can delete any directory indexes now.
1710 	 */
1711 	list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1712 		list_del(&curr->readdir_list);
1713 
1714 		if (curr->key.offset < ctx->pos) {
1715 			if (atomic_dec_and_test(&curr->refs))
1716 				kfree(curr);
1717 			continue;
1718 		}
1719 
1720 		ctx->pos = curr->key.offset;
1721 
1722 		di = (struct btrfs_dir_item *)curr->data;
1723 		name = (char *)(di + 1);
1724 		name_len = btrfs_stack_dir_name_len(di);
1725 
1726 		d_type = btrfs_filetype_table[di->type];
1727 		btrfs_disk_key_to_cpu(&location, &di->location);
1728 
1729 		over = !dir_emit(ctx, name, name_len,
1730 			       location.objectid, d_type);
1731 
1732 		if (atomic_dec_and_test(&curr->refs))
1733 			kfree(curr);
1734 
1735 		if (over)
1736 			return 1;
1737 	}
1738 	return 0;
1739 }
1740 
1741 static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
1742 				  struct btrfs_inode_item *inode_item,
1743 				  struct inode *inode)
1744 {
1745 	btrfs_set_stack_inode_uid(inode_item, i_uid_read(inode));
1746 	btrfs_set_stack_inode_gid(inode_item, i_gid_read(inode));
1747 	btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
1748 	btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
1749 	btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
1750 	btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
1751 	btrfs_set_stack_inode_generation(inode_item,
1752 					 BTRFS_I(inode)->generation);
1753 	btrfs_set_stack_inode_sequence(inode_item, inode->i_version);
1754 	btrfs_set_stack_inode_transid(inode_item, trans->transid);
1755 	btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
1756 	btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
1757 	btrfs_set_stack_inode_block_group(inode_item, 0);
1758 
1759 	btrfs_set_stack_timespec_sec(&inode_item->atime,
1760 				     inode->i_atime.tv_sec);
1761 	btrfs_set_stack_timespec_nsec(&inode_item->atime,
1762 				      inode->i_atime.tv_nsec);
1763 
1764 	btrfs_set_stack_timespec_sec(&inode_item->mtime,
1765 				     inode->i_mtime.tv_sec);
1766 	btrfs_set_stack_timespec_nsec(&inode_item->mtime,
1767 				      inode->i_mtime.tv_nsec);
1768 
1769 	btrfs_set_stack_timespec_sec(&inode_item->ctime,
1770 				     inode->i_ctime.tv_sec);
1771 	btrfs_set_stack_timespec_nsec(&inode_item->ctime,
1772 				      inode->i_ctime.tv_nsec);
1773 
1774 	btrfs_set_stack_timespec_sec(&inode_item->otime,
1775 				     BTRFS_I(inode)->i_otime.tv_sec);
1776 	btrfs_set_stack_timespec_nsec(&inode_item->otime,
1777 				     BTRFS_I(inode)->i_otime.tv_nsec);
1778 }
1779 
1780 int btrfs_fill_inode(struct inode *inode, u32 *rdev)
1781 {
1782 	struct btrfs_delayed_node *delayed_node;
1783 	struct btrfs_inode_item *inode_item;
1784 
1785 	delayed_node = btrfs_get_delayed_node(inode);
1786 	if (!delayed_node)
1787 		return -ENOENT;
1788 
1789 	mutex_lock(&delayed_node->mutex);
1790 	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1791 		mutex_unlock(&delayed_node->mutex);
1792 		btrfs_release_delayed_node(delayed_node);
1793 		return -ENOENT;
1794 	}
1795 
1796 	inode_item = &delayed_node->inode_item;
1797 
1798 	i_uid_write(inode, btrfs_stack_inode_uid(inode_item));
1799 	i_gid_write(inode, btrfs_stack_inode_gid(inode_item));
1800 	btrfs_i_size_write(inode, btrfs_stack_inode_size(inode_item));
1801 	inode->i_mode = btrfs_stack_inode_mode(inode_item);
1802 	set_nlink(inode, btrfs_stack_inode_nlink(inode_item));
1803 	inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
1804 	BTRFS_I(inode)->generation = btrfs_stack_inode_generation(inode_item);
1805         BTRFS_I(inode)->last_trans = btrfs_stack_inode_transid(inode_item);
1806 
1807 	inode->i_version = btrfs_stack_inode_sequence(inode_item);
1808 	inode->i_rdev = 0;
1809 	*rdev = btrfs_stack_inode_rdev(inode_item);
1810 	BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
1811 
1812 	inode->i_atime.tv_sec = btrfs_stack_timespec_sec(&inode_item->atime);
1813 	inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->atime);
1814 
1815 	inode->i_mtime.tv_sec = btrfs_stack_timespec_sec(&inode_item->mtime);
1816 	inode->i_mtime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->mtime);
1817 
1818 	inode->i_ctime.tv_sec = btrfs_stack_timespec_sec(&inode_item->ctime);
1819 	inode->i_ctime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->ctime);
1820 
1821 	BTRFS_I(inode)->i_otime.tv_sec =
1822 		btrfs_stack_timespec_sec(&inode_item->otime);
1823 	BTRFS_I(inode)->i_otime.tv_nsec =
1824 		btrfs_stack_timespec_nsec(&inode_item->otime);
1825 
1826 	inode->i_generation = BTRFS_I(inode)->generation;
1827 	BTRFS_I(inode)->index_cnt = (u64)-1;
1828 
1829 	mutex_unlock(&delayed_node->mutex);
1830 	btrfs_release_delayed_node(delayed_node);
1831 	return 0;
1832 }
1833 
1834 int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
1835 			       struct btrfs_root *root, struct inode *inode)
1836 {
1837 	struct btrfs_delayed_node *delayed_node;
1838 	int ret = 0;
1839 
1840 	delayed_node = btrfs_get_or_create_delayed_node(inode);
1841 	if (IS_ERR(delayed_node))
1842 		return PTR_ERR(delayed_node);
1843 
1844 	mutex_lock(&delayed_node->mutex);
1845 	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1846 		fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1847 		goto release_node;
1848 	}
1849 
1850 	ret = btrfs_delayed_inode_reserve_metadata(trans, root, inode,
1851 						   delayed_node);
1852 	if (ret)
1853 		goto release_node;
1854 
1855 	fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1856 	set_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
1857 	delayed_node->count++;
1858 	atomic_inc(&root->fs_info->delayed_root->items);
1859 release_node:
1860 	mutex_unlock(&delayed_node->mutex);
1861 	btrfs_release_delayed_node(delayed_node);
1862 	return ret;
1863 }
1864 
1865 int btrfs_delayed_delete_inode_ref(struct inode *inode)
1866 {
1867 	struct btrfs_delayed_node *delayed_node;
1868 
1869 	/*
1870 	 * we don't do delayed inode updates during log recovery because it
1871 	 * leads to enospc problems.  This means we also can't do
1872 	 * delayed inode refs
1873 	 */
1874 	if (BTRFS_I(inode)->root->fs_info->log_root_recovering)
1875 		return -EAGAIN;
1876 
1877 	delayed_node = btrfs_get_or_create_delayed_node(inode);
1878 	if (IS_ERR(delayed_node))
1879 		return PTR_ERR(delayed_node);
1880 
1881 	/*
1882 	 * We don't reserve space for inode ref deletion is because:
1883 	 * - We ONLY do async inode ref deletion for the inode who has only
1884 	 *   one link(i_nlink == 1), it means there is only one inode ref.
1885 	 *   And in most case, the inode ref and the inode item are in the
1886 	 *   same leaf, and we will deal with them at the same time.
1887 	 *   Since we are sure we will reserve the space for the inode item,
1888 	 *   it is unnecessary to reserve space for inode ref deletion.
1889 	 * - If the inode ref and the inode item are not in the same leaf,
1890 	 *   We also needn't worry about enospc problem, because we reserve
1891 	 *   much more space for the inode update than it needs.
1892 	 * - At the worst, we can steal some space from the global reservation.
1893 	 *   It is very rare.
1894 	 */
1895 	mutex_lock(&delayed_node->mutex);
1896 	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
1897 		goto release_node;
1898 
1899 	set_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
1900 	delayed_node->count++;
1901 	atomic_inc(&BTRFS_I(inode)->root->fs_info->delayed_root->items);
1902 release_node:
1903 	mutex_unlock(&delayed_node->mutex);
1904 	btrfs_release_delayed_node(delayed_node);
1905 	return 0;
1906 }
1907 
1908 static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
1909 {
1910 	struct btrfs_root *root = delayed_node->root;
1911 	struct btrfs_delayed_item *curr_item, *prev_item;
1912 
1913 	mutex_lock(&delayed_node->mutex);
1914 	curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
1915 	while (curr_item) {
1916 		btrfs_delayed_item_release_metadata(root, curr_item);
1917 		prev_item = curr_item;
1918 		curr_item = __btrfs_next_delayed_item(prev_item);
1919 		btrfs_release_delayed_item(prev_item);
1920 	}
1921 
1922 	curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
1923 	while (curr_item) {
1924 		btrfs_delayed_item_release_metadata(root, curr_item);
1925 		prev_item = curr_item;
1926 		curr_item = __btrfs_next_delayed_item(prev_item);
1927 		btrfs_release_delayed_item(prev_item);
1928 	}
1929 
1930 	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
1931 		btrfs_release_delayed_iref(delayed_node);
1932 
1933 	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1934 		btrfs_delayed_inode_release_metadata(root, delayed_node);
1935 		btrfs_release_delayed_inode(delayed_node);
1936 	}
1937 	mutex_unlock(&delayed_node->mutex);
1938 }
1939 
1940 void btrfs_kill_delayed_inode_items(struct inode *inode)
1941 {
1942 	struct btrfs_delayed_node *delayed_node;
1943 
1944 	delayed_node = btrfs_get_delayed_node(inode);
1945 	if (!delayed_node)
1946 		return;
1947 
1948 	__btrfs_kill_delayed_node(delayed_node);
1949 	btrfs_release_delayed_node(delayed_node);
1950 }
1951 
1952 void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
1953 {
1954 	u64 inode_id = 0;
1955 	struct btrfs_delayed_node *delayed_nodes[8];
1956 	int i, n;
1957 
1958 	while (1) {
1959 		spin_lock(&root->inode_lock);
1960 		n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
1961 					   (void **)delayed_nodes, inode_id,
1962 					   ARRAY_SIZE(delayed_nodes));
1963 		if (!n) {
1964 			spin_unlock(&root->inode_lock);
1965 			break;
1966 		}
1967 
1968 		inode_id = delayed_nodes[n - 1]->inode_id + 1;
1969 
1970 		for (i = 0; i < n; i++)
1971 			atomic_inc(&delayed_nodes[i]->refs);
1972 		spin_unlock(&root->inode_lock);
1973 
1974 		for (i = 0; i < n; i++) {
1975 			__btrfs_kill_delayed_node(delayed_nodes[i]);
1976 			btrfs_release_delayed_node(delayed_nodes[i]);
1977 		}
1978 	}
1979 }
1980 
1981 void btrfs_destroy_delayed_inodes(struct btrfs_root *root)
1982 {
1983 	struct btrfs_delayed_root *delayed_root;
1984 	struct btrfs_delayed_node *curr_node, *prev_node;
1985 
1986 	delayed_root = btrfs_get_delayed_root(root);
1987 
1988 	curr_node = btrfs_first_delayed_node(delayed_root);
1989 	while (curr_node) {
1990 		__btrfs_kill_delayed_node(curr_node);
1991 
1992 		prev_node = curr_node;
1993 		curr_node = btrfs_next_delayed_node(curr_node);
1994 		btrfs_release_delayed_node(prev_node);
1995 	}
1996 }
1997 
1998