xref: /linux/fs/btrfs/backref.c (revision b43ab901d671e3e3cad425ea5e9a3c74e266dcdd)
1 /*
2  * Copyright (C) 2011 STRATO.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include "ctree.h"
20 #include "disk-io.h"
21 #include "backref.h"
22 #include "ulist.h"
23 #include "transaction.h"
24 #include "delayed-ref.h"
25 
26 /*
27  * this structure records all encountered refs on the way up to the root
28  */
29 struct __prelim_ref {
30 	struct list_head list;
31 	u64 root_id;
32 	struct btrfs_key key;
33 	int level;
34 	int count;
35 	u64 parent;
36 	u64 wanted_disk_byte;
37 };
38 
39 static int __add_prelim_ref(struct list_head *head, u64 root_id,
40 			    struct btrfs_key *key, int level, u64 parent,
41 			    u64 wanted_disk_byte, int count)
42 {
43 	struct __prelim_ref *ref;
44 
45 	/* in case we're adding delayed refs, we're holding the refs spinlock */
46 	ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
47 	if (!ref)
48 		return -ENOMEM;
49 
50 	ref->root_id = root_id;
51 	if (key)
52 		ref->key = *key;
53 	else
54 		memset(&ref->key, 0, sizeof(ref->key));
55 
56 	ref->level = level;
57 	ref->count = count;
58 	ref->parent = parent;
59 	ref->wanted_disk_byte = wanted_disk_byte;
60 	list_add_tail(&ref->list, head);
61 
62 	return 0;
63 }
64 
65 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
66 				struct ulist *parents,
67 				struct extent_buffer *eb, int level,
68 				u64 wanted_objectid, u64 wanted_disk_byte)
69 {
70 	int ret;
71 	int slot;
72 	struct btrfs_file_extent_item *fi;
73 	struct btrfs_key key;
74 	u64 disk_byte;
75 
76 add_parent:
77 	ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
78 	if (ret < 0)
79 		return ret;
80 
81 	if (level != 0)
82 		return 0;
83 
84 	/*
85 	 * if the current leaf is full with EXTENT_DATA items, we must
86 	 * check the next one if that holds a reference as well.
87 	 * ref->count cannot be used to skip this check.
88 	 * repeat this until we don't find any additional EXTENT_DATA items.
89 	 */
90 	while (1) {
91 		ret = btrfs_next_leaf(root, path);
92 		if (ret < 0)
93 			return ret;
94 		if (ret)
95 			return 0;
96 
97 		eb = path->nodes[0];
98 		for (slot = 0; slot < btrfs_header_nritems(eb); ++slot) {
99 			btrfs_item_key_to_cpu(eb, &key, slot);
100 			if (key.objectid != wanted_objectid ||
101 			    key.type != BTRFS_EXTENT_DATA_KEY)
102 				return 0;
103 			fi = btrfs_item_ptr(eb, slot,
104 						struct btrfs_file_extent_item);
105 			disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
106 			if (disk_byte == wanted_disk_byte)
107 				goto add_parent;
108 		}
109 	}
110 
111 	return 0;
112 }
113 
114 /*
115  * resolve an indirect backref in the form (root_id, key, level)
116  * to a logical address
117  */
118 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
119 					struct __prelim_ref *ref,
120 					struct ulist *parents)
121 {
122 	struct btrfs_path *path;
123 	struct btrfs_root *root;
124 	struct btrfs_key root_key;
125 	struct btrfs_key key = {0};
126 	struct extent_buffer *eb;
127 	int ret = 0;
128 	int root_level;
129 	int level = ref->level;
130 
131 	path = btrfs_alloc_path();
132 	if (!path)
133 		return -ENOMEM;
134 
135 	root_key.objectid = ref->root_id;
136 	root_key.type = BTRFS_ROOT_ITEM_KEY;
137 	root_key.offset = (u64)-1;
138 	root = btrfs_read_fs_root_no_name(fs_info, &root_key);
139 	if (IS_ERR(root)) {
140 		ret = PTR_ERR(root);
141 		goto out;
142 	}
143 
144 	rcu_read_lock();
145 	root_level = btrfs_header_level(root->node);
146 	rcu_read_unlock();
147 
148 	if (root_level + 1 == level)
149 		goto out;
150 
151 	path->lowest_level = level;
152 	ret = btrfs_search_slot(NULL, root, &ref->key, path, 0, 0);
153 	pr_debug("search slot in root %llu (level %d, ref count %d) returned "
154 		 "%d for key (%llu %u %llu)\n",
155 		 (unsigned long long)ref->root_id, level, ref->count, ret,
156 		 (unsigned long long)ref->key.objectid, ref->key.type,
157 		 (unsigned long long)ref->key.offset);
158 	if (ret < 0)
159 		goto out;
160 
161 	eb = path->nodes[level];
162 	if (!eb) {
163 		WARN_ON(1);
164 		ret = 1;
165 		goto out;
166 	}
167 
168 	if (level == 0) {
169 		if (ret == 1 && path->slots[0] >= btrfs_header_nritems(eb)) {
170 			ret = btrfs_next_leaf(root, path);
171 			if (ret)
172 				goto out;
173 			eb = path->nodes[0];
174 		}
175 
176 		btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
177 	}
178 
179 	/* the last two parameters will only be used for level == 0 */
180 	ret = add_all_parents(root, path, parents, eb, level, key.objectid,
181 				ref->wanted_disk_byte);
182 out:
183 	btrfs_free_path(path);
184 	return ret;
185 }
186 
187 /*
188  * resolve all indirect backrefs from the list
189  */
190 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
191 				   struct list_head *head)
192 {
193 	int err;
194 	int ret = 0;
195 	struct __prelim_ref *ref;
196 	struct __prelim_ref *ref_safe;
197 	struct __prelim_ref *new_ref;
198 	struct ulist *parents;
199 	struct ulist_node *node;
200 
201 	parents = ulist_alloc(GFP_NOFS);
202 	if (!parents)
203 		return -ENOMEM;
204 
205 	/*
206 	 * _safe allows us to insert directly after the current item without
207 	 * iterating over the newly inserted items.
208 	 * we're also allowed to re-assign ref during iteration.
209 	 */
210 	list_for_each_entry_safe(ref, ref_safe, head, list) {
211 		if (ref->parent)	/* already direct */
212 			continue;
213 		if (ref->count == 0)
214 			continue;
215 		err = __resolve_indirect_ref(fs_info, ref, parents);
216 		if (err) {
217 			if (ret == 0)
218 				ret = err;
219 			continue;
220 		}
221 
222 		/* we put the first parent into the ref at hand */
223 		node = ulist_next(parents, NULL);
224 		ref->parent = node ? node->val : 0;
225 
226 		/* additional parents require new refs being added here */
227 		while ((node = ulist_next(parents, node))) {
228 			new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
229 			if (!new_ref) {
230 				ret = -ENOMEM;
231 				break;
232 			}
233 			memcpy(new_ref, ref, sizeof(*ref));
234 			new_ref->parent = node->val;
235 			list_add(&new_ref->list, &ref->list);
236 		}
237 		ulist_reinit(parents);
238 	}
239 
240 	ulist_free(parents);
241 	return ret;
242 }
243 
244 /*
245  * merge two lists of backrefs and adjust counts accordingly
246  *
247  * mode = 1: merge identical keys, if key is set
248  * mode = 2: merge identical parents
249  */
250 static int __merge_refs(struct list_head *head, int mode)
251 {
252 	struct list_head *pos1;
253 
254 	list_for_each(pos1, head) {
255 		struct list_head *n2;
256 		struct list_head *pos2;
257 		struct __prelim_ref *ref1;
258 
259 		ref1 = list_entry(pos1, struct __prelim_ref, list);
260 
261 		if (mode == 1 && ref1->key.type == 0)
262 			continue;
263 		for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
264 		     pos2 = n2, n2 = pos2->next) {
265 			struct __prelim_ref *ref2;
266 
267 			ref2 = list_entry(pos2, struct __prelim_ref, list);
268 
269 			if (mode == 1) {
270 				if (memcmp(&ref1->key, &ref2->key,
271 					   sizeof(ref1->key)) ||
272 				    ref1->level != ref2->level ||
273 				    ref1->root_id != ref2->root_id)
274 					continue;
275 				ref1->count += ref2->count;
276 			} else {
277 				if (ref1->parent != ref2->parent)
278 					continue;
279 				ref1->count += ref2->count;
280 			}
281 			list_del(&ref2->list);
282 			kfree(ref2);
283 		}
284 
285 	}
286 	return 0;
287 }
288 
289 /*
290  * add all currently queued delayed refs from this head whose seq nr is
291  * smaller or equal that seq to the list
292  */
293 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
294 			      struct btrfs_key *info_key,
295 			      struct list_head *prefs)
296 {
297 	struct btrfs_delayed_extent_op *extent_op = head->extent_op;
298 	struct rb_node *n = &head->node.rb_node;
299 	int sgn;
300 	int ret;
301 
302 	if (extent_op && extent_op->update_key)
303 		btrfs_disk_key_to_cpu(info_key, &extent_op->key);
304 
305 	while ((n = rb_prev(n))) {
306 		struct btrfs_delayed_ref_node *node;
307 		node = rb_entry(n, struct btrfs_delayed_ref_node,
308 				rb_node);
309 		if (node->bytenr != head->node.bytenr)
310 			break;
311 		WARN_ON(node->is_head);
312 
313 		if (node->seq > seq)
314 			continue;
315 
316 		switch (node->action) {
317 		case BTRFS_ADD_DELAYED_EXTENT:
318 		case BTRFS_UPDATE_DELAYED_HEAD:
319 			WARN_ON(1);
320 			continue;
321 		case BTRFS_ADD_DELAYED_REF:
322 			sgn = 1;
323 			break;
324 		case BTRFS_DROP_DELAYED_REF:
325 			sgn = -1;
326 			break;
327 		default:
328 			BUG_ON(1);
329 		}
330 		switch (node->type) {
331 		case BTRFS_TREE_BLOCK_REF_KEY: {
332 			struct btrfs_delayed_tree_ref *ref;
333 
334 			ref = btrfs_delayed_node_to_tree_ref(node);
335 			ret = __add_prelim_ref(prefs, ref->root, info_key,
336 					       ref->level + 1, 0, node->bytenr,
337 					       node->ref_mod * sgn);
338 			break;
339 		}
340 		case BTRFS_SHARED_BLOCK_REF_KEY: {
341 			struct btrfs_delayed_tree_ref *ref;
342 
343 			ref = btrfs_delayed_node_to_tree_ref(node);
344 			ret = __add_prelim_ref(prefs, ref->root, info_key,
345 					       ref->level + 1, ref->parent,
346 					       node->bytenr,
347 					       node->ref_mod * sgn);
348 			break;
349 		}
350 		case BTRFS_EXTENT_DATA_REF_KEY: {
351 			struct btrfs_delayed_data_ref *ref;
352 			struct btrfs_key key;
353 
354 			ref = btrfs_delayed_node_to_data_ref(node);
355 
356 			key.objectid = ref->objectid;
357 			key.type = BTRFS_EXTENT_DATA_KEY;
358 			key.offset = ref->offset;
359 			ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
360 					       node->bytenr,
361 					       node->ref_mod * sgn);
362 			break;
363 		}
364 		case BTRFS_SHARED_DATA_REF_KEY: {
365 			struct btrfs_delayed_data_ref *ref;
366 			struct btrfs_key key;
367 
368 			ref = btrfs_delayed_node_to_data_ref(node);
369 
370 			key.objectid = ref->objectid;
371 			key.type = BTRFS_EXTENT_DATA_KEY;
372 			key.offset = ref->offset;
373 			ret = __add_prelim_ref(prefs, ref->root, &key, 0,
374 					       ref->parent, node->bytenr,
375 					       node->ref_mod * sgn);
376 			break;
377 		}
378 		default:
379 			WARN_ON(1);
380 		}
381 		BUG_ON(ret);
382 	}
383 
384 	return 0;
385 }
386 
387 /*
388  * add all inline backrefs for bytenr to the list
389  */
390 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
391 			     struct btrfs_path *path, u64 bytenr,
392 			     struct btrfs_key *info_key, int *info_level,
393 			     struct list_head *prefs)
394 {
395 	int ret;
396 	int slot;
397 	struct extent_buffer *leaf;
398 	struct btrfs_key key;
399 	unsigned long ptr;
400 	unsigned long end;
401 	struct btrfs_extent_item *ei;
402 	u64 flags;
403 	u64 item_size;
404 
405 	/*
406 	 * enumerate all inline refs
407 	 */
408 	leaf = path->nodes[0];
409 	slot = path->slots[0] - 1;
410 
411 	item_size = btrfs_item_size_nr(leaf, slot);
412 	BUG_ON(item_size < sizeof(*ei));
413 
414 	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
415 	flags = btrfs_extent_flags(leaf, ei);
416 
417 	ptr = (unsigned long)(ei + 1);
418 	end = (unsigned long)ei + item_size;
419 
420 	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
421 		struct btrfs_tree_block_info *info;
422 		struct btrfs_disk_key disk_key;
423 
424 		info = (struct btrfs_tree_block_info *)ptr;
425 		*info_level = btrfs_tree_block_level(leaf, info);
426 		btrfs_tree_block_key(leaf, info, &disk_key);
427 		btrfs_disk_key_to_cpu(info_key, &disk_key);
428 		ptr += sizeof(struct btrfs_tree_block_info);
429 		BUG_ON(ptr > end);
430 	} else {
431 		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
432 	}
433 
434 	while (ptr < end) {
435 		struct btrfs_extent_inline_ref *iref;
436 		u64 offset;
437 		int type;
438 
439 		iref = (struct btrfs_extent_inline_ref *)ptr;
440 		type = btrfs_extent_inline_ref_type(leaf, iref);
441 		offset = btrfs_extent_inline_ref_offset(leaf, iref);
442 
443 		switch (type) {
444 		case BTRFS_SHARED_BLOCK_REF_KEY:
445 			ret = __add_prelim_ref(prefs, 0, info_key,
446 						*info_level + 1, offset,
447 						bytenr, 1);
448 			break;
449 		case BTRFS_SHARED_DATA_REF_KEY: {
450 			struct btrfs_shared_data_ref *sdref;
451 			int count;
452 
453 			sdref = (struct btrfs_shared_data_ref *)(iref + 1);
454 			count = btrfs_shared_data_ref_count(leaf, sdref);
455 			ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
456 					       bytenr, count);
457 			break;
458 		}
459 		case BTRFS_TREE_BLOCK_REF_KEY:
460 			ret = __add_prelim_ref(prefs, offset, info_key,
461 					       *info_level + 1, 0, bytenr, 1);
462 			break;
463 		case BTRFS_EXTENT_DATA_REF_KEY: {
464 			struct btrfs_extent_data_ref *dref;
465 			int count;
466 			u64 root;
467 
468 			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
469 			count = btrfs_extent_data_ref_count(leaf, dref);
470 			key.objectid = btrfs_extent_data_ref_objectid(leaf,
471 								      dref);
472 			key.type = BTRFS_EXTENT_DATA_KEY;
473 			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
474 			root = btrfs_extent_data_ref_root(leaf, dref);
475 			ret = __add_prelim_ref(prefs, root, &key, 0, 0, bytenr,
476 						count);
477 			break;
478 		}
479 		default:
480 			WARN_ON(1);
481 		}
482 		BUG_ON(ret);
483 		ptr += btrfs_extent_inline_ref_size(type);
484 	}
485 
486 	return 0;
487 }
488 
489 /*
490  * add all non-inline backrefs for bytenr to the list
491  */
492 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
493 			    struct btrfs_path *path, u64 bytenr,
494 			    struct btrfs_key *info_key, int info_level,
495 			    struct list_head *prefs)
496 {
497 	struct btrfs_root *extent_root = fs_info->extent_root;
498 	int ret;
499 	int slot;
500 	struct extent_buffer *leaf;
501 	struct btrfs_key key;
502 
503 	while (1) {
504 		ret = btrfs_next_item(extent_root, path);
505 		if (ret < 0)
506 			break;
507 		if (ret) {
508 			ret = 0;
509 			break;
510 		}
511 
512 		slot = path->slots[0];
513 		leaf = path->nodes[0];
514 		btrfs_item_key_to_cpu(leaf, &key, slot);
515 
516 		if (key.objectid != bytenr)
517 			break;
518 		if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
519 			continue;
520 		if (key.type > BTRFS_SHARED_DATA_REF_KEY)
521 			break;
522 
523 		switch (key.type) {
524 		case BTRFS_SHARED_BLOCK_REF_KEY:
525 			ret = __add_prelim_ref(prefs, 0, info_key,
526 						info_level + 1, key.offset,
527 						bytenr, 1);
528 			break;
529 		case BTRFS_SHARED_DATA_REF_KEY: {
530 			struct btrfs_shared_data_ref *sdref;
531 			int count;
532 
533 			sdref = btrfs_item_ptr(leaf, slot,
534 					      struct btrfs_shared_data_ref);
535 			count = btrfs_shared_data_ref_count(leaf, sdref);
536 			ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
537 						bytenr, count);
538 			break;
539 		}
540 		case BTRFS_TREE_BLOCK_REF_KEY:
541 			ret = __add_prelim_ref(prefs, key.offset, info_key,
542 						info_level + 1, 0, bytenr, 1);
543 			break;
544 		case BTRFS_EXTENT_DATA_REF_KEY: {
545 			struct btrfs_extent_data_ref *dref;
546 			int count;
547 			u64 root;
548 
549 			dref = btrfs_item_ptr(leaf, slot,
550 					      struct btrfs_extent_data_ref);
551 			count = btrfs_extent_data_ref_count(leaf, dref);
552 			key.objectid = btrfs_extent_data_ref_objectid(leaf,
553 								      dref);
554 			key.type = BTRFS_EXTENT_DATA_KEY;
555 			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
556 			root = btrfs_extent_data_ref_root(leaf, dref);
557 			ret = __add_prelim_ref(prefs, root, &key, 0, 0,
558 						bytenr, count);
559 			break;
560 		}
561 		default:
562 			WARN_ON(1);
563 		}
564 		BUG_ON(ret);
565 	}
566 
567 	return ret;
568 }
569 
570 /*
571  * this adds all existing backrefs (inline backrefs, backrefs and delayed
572  * refs) for the given bytenr to the refs list, merges duplicates and resolves
573  * indirect refs to their parent bytenr.
574  * When roots are found, they're added to the roots list
575  *
576  * FIXME some caching might speed things up
577  */
578 static int find_parent_nodes(struct btrfs_trans_handle *trans,
579 			     struct btrfs_fs_info *fs_info, u64 bytenr,
580 			     u64 seq, struct ulist *refs, struct ulist *roots)
581 {
582 	struct btrfs_key key;
583 	struct btrfs_path *path;
584 	struct btrfs_key info_key = { 0 };
585 	struct btrfs_delayed_ref_root *delayed_refs = NULL;
586 	struct btrfs_delayed_ref_head *head = NULL;
587 	int info_level = 0;
588 	int ret;
589 	struct list_head prefs_delayed;
590 	struct list_head prefs;
591 	struct __prelim_ref *ref;
592 
593 	INIT_LIST_HEAD(&prefs);
594 	INIT_LIST_HEAD(&prefs_delayed);
595 
596 	key.objectid = bytenr;
597 	key.type = BTRFS_EXTENT_ITEM_KEY;
598 	key.offset = (u64)-1;
599 
600 	path = btrfs_alloc_path();
601 	if (!path)
602 		return -ENOMEM;
603 
604 	/*
605 	 * grab both a lock on the path and a lock on the delayed ref head.
606 	 * We need both to get a consistent picture of how the refs look
607 	 * at a specified point in time
608 	 */
609 again:
610 	ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
611 	if (ret < 0)
612 		goto out;
613 	BUG_ON(ret == 0);
614 
615 	/*
616 	 * look if there are updates for this ref queued and lock the head
617 	 */
618 	delayed_refs = &trans->transaction->delayed_refs;
619 	spin_lock(&delayed_refs->lock);
620 	head = btrfs_find_delayed_ref_head(trans, bytenr);
621 	if (head) {
622 		if (!mutex_trylock(&head->mutex)) {
623 			atomic_inc(&head->node.refs);
624 			spin_unlock(&delayed_refs->lock);
625 
626 			btrfs_release_path(path);
627 
628 			/*
629 			 * Mutex was contended, block until it's
630 			 * released and try again
631 			 */
632 			mutex_lock(&head->mutex);
633 			mutex_unlock(&head->mutex);
634 			btrfs_put_delayed_ref(&head->node);
635 			goto again;
636 		}
637 		ret = __add_delayed_refs(head, seq, &info_key, &prefs_delayed);
638 		if (ret)
639 			goto out;
640 	}
641 	spin_unlock(&delayed_refs->lock);
642 
643 	if (path->slots[0]) {
644 		struct extent_buffer *leaf;
645 		int slot;
646 
647 		leaf = path->nodes[0];
648 		slot = path->slots[0] - 1;
649 		btrfs_item_key_to_cpu(leaf, &key, slot);
650 		if (key.objectid == bytenr &&
651 		    key.type == BTRFS_EXTENT_ITEM_KEY) {
652 			ret = __add_inline_refs(fs_info, path, bytenr,
653 						&info_key, &info_level, &prefs);
654 			if (ret)
655 				goto out;
656 			ret = __add_keyed_refs(fs_info, path, bytenr, &info_key,
657 					       info_level, &prefs);
658 			if (ret)
659 				goto out;
660 		}
661 	}
662 	btrfs_release_path(path);
663 
664 	/*
665 	 * when adding the delayed refs above, the info_key might not have
666 	 * been known yet. Go over the list and replace the missing keys
667 	 */
668 	list_for_each_entry(ref, &prefs_delayed, list) {
669 		if ((ref->key.offset | ref->key.type | ref->key.objectid) == 0)
670 			memcpy(&ref->key, &info_key, sizeof(ref->key));
671 	}
672 	list_splice_init(&prefs_delayed, &prefs);
673 
674 	ret = __merge_refs(&prefs, 1);
675 	if (ret)
676 		goto out;
677 
678 	ret = __resolve_indirect_refs(fs_info, &prefs);
679 	if (ret)
680 		goto out;
681 
682 	ret = __merge_refs(&prefs, 2);
683 	if (ret)
684 		goto out;
685 
686 	while (!list_empty(&prefs)) {
687 		ref = list_first_entry(&prefs, struct __prelim_ref, list);
688 		list_del(&ref->list);
689 		if (ref->count < 0)
690 			WARN_ON(1);
691 		if (ref->count && ref->root_id && ref->parent == 0) {
692 			/* no parent == root of tree */
693 			ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
694 			BUG_ON(ret < 0);
695 		}
696 		if (ref->count && ref->parent) {
697 			ret = ulist_add(refs, ref->parent, 0, GFP_NOFS);
698 			BUG_ON(ret < 0);
699 		}
700 		kfree(ref);
701 	}
702 
703 out:
704 	if (head)
705 		mutex_unlock(&head->mutex);
706 	btrfs_free_path(path);
707 	while (!list_empty(&prefs)) {
708 		ref = list_first_entry(&prefs, struct __prelim_ref, list);
709 		list_del(&ref->list);
710 		kfree(ref);
711 	}
712 	while (!list_empty(&prefs_delayed)) {
713 		ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
714 				       list);
715 		list_del(&ref->list);
716 		kfree(ref);
717 	}
718 
719 	return ret;
720 }
721 
722 /*
723  * Finds all leafs with a reference to the specified combination of bytenr and
724  * offset. key_list_head will point to a list of corresponding keys (caller must
725  * free each list element). The leafs will be stored in the leafs ulist, which
726  * must be freed with ulist_free.
727  *
728  * returns 0 on success, <0 on error
729  */
730 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
731 				struct btrfs_fs_info *fs_info, u64 bytenr,
732 				u64 num_bytes, u64 seq, struct ulist **leafs)
733 {
734 	struct ulist *tmp;
735 	int ret;
736 
737 	tmp = ulist_alloc(GFP_NOFS);
738 	if (!tmp)
739 		return -ENOMEM;
740 	*leafs = ulist_alloc(GFP_NOFS);
741 	if (!*leafs) {
742 		ulist_free(tmp);
743 		return -ENOMEM;
744 	}
745 
746 	ret = find_parent_nodes(trans, fs_info, bytenr, seq, *leafs, tmp);
747 	ulist_free(tmp);
748 
749 	if (ret < 0 && ret != -ENOENT) {
750 		ulist_free(*leafs);
751 		return ret;
752 	}
753 
754 	return 0;
755 }
756 
757 /*
758  * walk all backrefs for a given extent to find all roots that reference this
759  * extent. Walking a backref means finding all extents that reference this
760  * extent and in turn walk the backrefs of those, too. Naturally this is a
761  * recursive process, but here it is implemented in an iterative fashion: We
762  * find all referencing extents for the extent in question and put them on a
763  * list. In turn, we find all referencing extents for those, further appending
764  * to the list. The way we iterate the list allows adding more elements after
765  * the current while iterating. The process stops when we reach the end of the
766  * list. Found roots are added to the roots list.
767  *
768  * returns 0 on success, < 0 on error.
769  */
770 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
771 				struct btrfs_fs_info *fs_info, u64 bytenr,
772 				u64 num_bytes, u64 seq, struct ulist **roots)
773 {
774 	struct ulist *tmp;
775 	struct ulist_node *node = NULL;
776 	int ret;
777 
778 	tmp = ulist_alloc(GFP_NOFS);
779 	if (!tmp)
780 		return -ENOMEM;
781 	*roots = ulist_alloc(GFP_NOFS);
782 	if (!*roots) {
783 		ulist_free(tmp);
784 		return -ENOMEM;
785 	}
786 
787 	while (1) {
788 		ret = find_parent_nodes(trans, fs_info, bytenr, seq,
789 					tmp, *roots);
790 		if (ret < 0 && ret != -ENOENT) {
791 			ulist_free(tmp);
792 			ulist_free(*roots);
793 			return ret;
794 		}
795 		node = ulist_next(tmp, node);
796 		if (!node)
797 			break;
798 		bytenr = node->val;
799 	}
800 
801 	ulist_free(tmp);
802 	return 0;
803 }
804 
805 
806 static int __inode_info(u64 inum, u64 ioff, u8 key_type,
807 			struct btrfs_root *fs_root, struct btrfs_path *path,
808 			struct btrfs_key *found_key)
809 {
810 	int ret;
811 	struct btrfs_key key;
812 	struct extent_buffer *eb;
813 
814 	key.type = key_type;
815 	key.objectid = inum;
816 	key.offset = ioff;
817 
818 	ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
819 	if (ret < 0)
820 		return ret;
821 
822 	eb = path->nodes[0];
823 	if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
824 		ret = btrfs_next_leaf(fs_root, path);
825 		if (ret)
826 			return ret;
827 		eb = path->nodes[0];
828 	}
829 
830 	btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
831 	if (found_key->type != key.type || found_key->objectid != key.objectid)
832 		return 1;
833 
834 	return 0;
835 }
836 
837 /*
838  * this makes the path point to (inum INODE_ITEM ioff)
839  */
840 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
841 			struct btrfs_path *path)
842 {
843 	struct btrfs_key key;
844 	return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path,
845 				&key);
846 }
847 
848 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
849 				struct btrfs_path *path,
850 				struct btrfs_key *found_key)
851 {
852 	return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path,
853 				found_key);
854 }
855 
856 /*
857  * this iterates to turn a btrfs_inode_ref into a full filesystem path. elements
858  * of the path are separated by '/' and the path is guaranteed to be
859  * 0-terminated. the path is only given within the current file system.
860  * Therefore, it never starts with a '/'. the caller is responsible to provide
861  * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
862  * the start point of the resulting string is returned. this pointer is within
863  * dest, normally.
864  * in case the path buffer would overflow, the pointer is decremented further
865  * as if output was written to the buffer, though no more output is actually
866  * generated. that way, the caller can determine how much space would be
867  * required for the path to fit into the buffer. in that case, the returned
868  * value will be smaller than dest. callers must check this!
869  */
870 static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
871 				struct btrfs_inode_ref *iref,
872 				struct extent_buffer *eb_in, u64 parent,
873 				char *dest, u32 size)
874 {
875 	u32 len;
876 	int slot;
877 	u64 next_inum;
878 	int ret;
879 	s64 bytes_left = size - 1;
880 	struct extent_buffer *eb = eb_in;
881 	struct btrfs_key found_key;
882 
883 	if (bytes_left >= 0)
884 		dest[bytes_left] = '\0';
885 
886 	while (1) {
887 		len = btrfs_inode_ref_name_len(eb, iref);
888 		bytes_left -= len;
889 		if (bytes_left >= 0)
890 			read_extent_buffer(eb, dest + bytes_left,
891 						(unsigned long)(iref + 1), len);
892 		if (eb != eb_in)
893 			free_extent_buffer(eb);
894 		ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
895 		if (ret)
896 			break;
897 		next_inum = found_key.offset;
898 
899 		/* regular exit ahead */
900 		if (parent == next_inum)
901 			break;
902 
903 		slot = path->slots[0];
904 		eb = path->nodes[0];
905 		/* make sure we can use eb after releasing the path */
906 		if (eb != eb_in)
907 			atomic_inc(&eb->refs);
908 		btrfs_release_path(path);
909 
910 		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
911 		parent = next_inum;
912 		--bytes_left;
913 		if (bytes_left >= 0)
914 			dest[bytes_left] = '/';
915 	}
916 
917 	btrfs_release_path(path);
918 
919 	if (ret)
920 		return ERR_PTR(ret);
921 
922 	return dest + bytes_left;
923 }
924 
925 /*
926  * this makes the path point to (logical EXTENT_ITEM *)
927  * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
928  * tree blocks and <0 on error.
929  */
930 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
931 			struct btrfs_path *path, struct btrfs_key *found_key)
932 {
933 	int ret;
934 	u64 flags;
935 	u32 item_size;
936 	struct extent_buffer *eb;
937 	struct btrfs_extent_item *ei;
938 	struct btrfs_key key;
939 
940 	key.type = BTRFS_EXTENT_ITEM_KEY;
941 	key.objectid = logical;
942 	key.offset = (u64)-1;
943 
944 	ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
945 	if (ret < 0)
946 		return ret;
947 	ret = btrfs_previous_item(fs_info->extent_root, path,
948 					0, BTRFS_EXTENT_ITEM_KEY);
949 	if (ret < 0)
950 		return ret;
951 
952 	btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
953 	if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
954 	    found_key->objectid > logical ||
955 	    found_key->objectid + found_key->offset <= logical) {
956 		pr_debug("logical %llu is not within any extent\n",
957 			 (unsigned long long)logical);
958 		return -ENOENT;
959 	}
960 
961 	eb = path->nodes[0];
962 	item_size = btrfs_item_size_nr(eb, path->slots[0]);
963 	BUG_ON(item_size < sizeof(*ei));
964 
965 	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
966 	flags = btrfs_extent_flags(eb, ei);
967 
968 	pr_debug("logical %llu is at position %llu within the extent (%llu "
969 		 "EXTENT_ITEM %llu) flags %#llx size %u\n",
970 		 (unsigned long long)logical,
971 		 (unsigned long long)(logical - found_key->objectid),
972 		 (unsigned long long)found_key->objectid,
973 		 (unsigned long long)found_key->offset,
974 		 (unsigned long long)flags, item_size);
975 	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
976 		return BTRFS_EXTENT_FLAG_TREE_BLOCK;
977 	if (flags & BTRFS_EXTENT_FLAG_DATA)
978 		return BTRFS_EXTENT_FLAG_DATA;
979 
980 	return -EIO;
981 }
982 
983 /*
984  * helper function to iterate extent inline refs. ptr must point to a 0 value
985  * for the first call and may be modified. it is used to track state.
986  * if more refs exist, 0 is returned and the next call to
987  * __get_extent_inline_ref must pass the modified ptr parameter to get the
988  * next ref. after the last ref was processed, 1 is returned.
989  * returns <0 on error
990  */
991 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
992 				struct btrfs_extent_item *ei, u32 item_size,
993 				struct btrfs_extent_inline_ref **out_eiref,
994 				int *out_type)
995 {
996 	unsigned long end;
997 	u64 flags;
998 	struct btrfs_tree_block_info *info;
999 
1000 	if (!*ptr) {
1001 		/* first call */
1002 		flags = btrfs_extent_flags(eb, ei);
1003 		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1004 			info = (struct btrfs_tree_block_info *)(ei + 1);
1005 			*out_eiref =
1006 				(struct btrfs_extent_inline_ref *)(info + 1);
1007 		} else {
1008 			*out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1009 		}
1010 		*ptr = (unsigned long)*out_eiref;
1011 		if ((void *)*ptr >= (void *)ei + item_size)
1012 			return -ENOENT;
1013 	}
1014 
1015 	end = (unsigned long)ei + item_size;
1016 	*out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1017 	*out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1018 
1019 	*ptr += btrfs_extent_inline_ref_size(*out_type);
1020 	WARN_ON(*ptr > end);
1021 	if (*ptr == end)
1022 		return 1; /* last */
1023 
1024 	return 0;
1025 }
1026 
1027 /*
1028  * reads the tree block backref for an extent. tree level and root are returned
1029  * through out_level and out_root. ptr must point to a 0 value for the first
1030  * call and may be modified (see __get_extent_inline_ref comment).
1031  * returns 0 if data was provided, 1 if there was no more data to provide or
1032  * <0 on error.
1033  */
1034 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1035 				struct btrfs_extent_item *ei, u32 item_size,
1036 				u64 *out_root, u8 *out_level)
1037 {
1038 	int ret;
1039 	int type;
1040 	struct btrfs_tree_block_info *info;
1041 	struct btrfs_extent_inline_ref *eiref;
1042 
1043 	if (*ptr == (unsigned long)-1)
1044 		return 1;
1045 
1046 	while (1) {
1047 		ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1048 						&eiref, &type);
1049 		if (ret < 0)
1050 			return ret;
1051 
1052 		if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1053 		    type == BTRFS_SHARED_BLOCK_REF_KEY)
1054 			break;
1055 
1056 		if (ret == 1)
1057 			return 1;
1058 	}
1059 
1060 	/* we can treat both ref types equally here */
1061 	info = (struct btrfs_tree_block_info *)(ei + 1);
1062 	*out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1063 	*out_level = btrfs_tree_block_level(eb, info);
1064 
1065 	if (ret == 1)
1066 		*ptr = (unsigned long)-1;
1067 
1068 	return 0;
1069 }
1070 
1071 static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
1072 				struct btrfs_path *path, u64 logical,
1073 				u64 orig_extent_item_objectid,
1074 				u64 extent_item_pos, u64 root,
1075 				iterate_extent_inodes_t *iterate, void *ctx)
1076 {
1077 	u64 disk_byte;
1078 	struct btrfs_key key;
1079 	struct btrfs_file_extent_item *fi;
1080 	struct extent_buffer *eb;
1081 	int slot;
1082 	int nritems;
1083 	int ret = 0;
1084 	int extent_type;
1085 	u64 data_offset;
1086 	u64 data_len;
1087 
1088 	eb = read_tree_block(fs_info->tree_root, logical,
1089 				fs_info->tree_root->leafsize, 0);
1090 	if (!eb)
1091 		return -EIO;
1092 
1093 	/*
1094 	 * from the shared data ref, we only have the leaf but we need
1095 	 * the key. thus, we must look into all items and see that we
1096 	 * find one (some) with a reference to our extent item.
1097 	 */
1098 	nritems = btrfs_header_nritems(eb);
1099 	for (slot = 0; slot < nritems; ++slot) {
1100 		btrfs_item_key_to_cpu(eb, &key, slot);
1101 		if (key.type != BTRFS_EXTENT_DATA_KEY)
1102 			continue;
1103 		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
1104 		extent_type = btrfs_file_extent_type(eb, fi);
1105 		if (extent_type == BTRFS_FILE_EXTENT_INLINE)
1106 			continue;
1107 		/* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
1108 		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
1109 		if (disk_byte != orig_extent_item_objectid)
1110 			continue;
1111 
1112 		data_offset = btrfs_file_extent_offset(eb, fi);
1113 		data_len = btrfs_file_extent_num_bytes(eb, fi);
1114 
1115 		if (extent_item_pos < data_offset ||
1116 		    extent_item_pos >= data_offset + data_len)
1117 			continue;
1118 
1119 		pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1120 				"root %llu\n", orig_extent_item_objectid,
1121 				key.objectid, key.offset, root);
1122 		ret = iterate(key.objectid,
1123 				key.offset + (extent_item_pos - data_offset),
1124 				root, ctx);
1125 		if (ret) {
1126 			pr_debug("stopping iteration because ret=%d\n", ret);
1127 			break;
1128 		}
1129 	}
1130 
1131 	free_extent_buffer(eb);
1132 
1133 	return ret;
1134 }
1135 
1136 /*
1137  * calls iterate() for every inode that references the extent identified by
1138  * the given parameters.
1139  * when the iterator function returns a non-zero value, iteration stops.
1140  * path is guaranteed to be in released state when iterate() is called.
1141  */
1142 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1143 				struct btrfs_path *path,
1144 				u64 extent_item_objectid, u64 extent_item_pos,
1145 				iterate_extent_inodes_t *iterate, void *ctx)
1146 {
1147 	int ret;
1148 	struct list_head data_refs = LIST_HEAD_INIT(data_refs);
1149 	struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
1150 	struct btrfs_trans_handle *trans;
1151 	struct ulist *refs;
1152 	struct ulist *roots;
1153 	struct ulist_node *ref_node = NULL;
1154 	struct ulist_node *root_node = NULL;
1155 	struct seq_list seq_elem;
1156 	struct btrfs_delayed_ref_root *delayed_refs;
1157 
1158 	trans = btrfs_join_transaction(fs_info->extent_root);
1159 	if (IS_ERR(trans))
1160 		return PTR_ERR(trans);
1161 
1162 	pr_debug("resolving all inodes for extent %llu\n",
1163 			extent_item_objectid);
1164 
1165 	delayed_refs = &trans->transaction->delayed_refs;
1166 	spin_lock(&delayed_refs->lock);
1167 	btrfs_get_delayed_seq(delayed_refs, &seq_elem);
1168 	spin_unlock(&delayed_refs->lock);
1169 
1170 	ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1171 				   extent_item_pos, seq_elem.seq,
1172 				   &refs);
1173 
1174 	if (ret)
1175 		goto out;
1176 
1177 	while (!ret && (ref_node = ulist_next(refs, ref_node))) {
1178 		ret = btrfs_find_all_roots(trans, fs_info, ref_node->val, -1,
1179 						seq_elem.seq, &roots);
1180 		if (ret)
1181 			break;
1182 		while (!ret && (root_node = ulist_next(roots, root_node))) {
1183 			pr_debug("root %llu references leaf %llu\n",
1184 					root_node->val, ref_node->val);
1185 			ret = iterate_leaf_refs(fs_info, path, ref_node->val,
1186 						extent_item_objectid,
1187 						extent_item_pos, root_node->val,
1188 						iterate, ctx);
1189 		}
1190 	}
1191 
1192 	ulist_free(refs);
1193 	ulist_free(roots);
1194 out:
1195 	btrfs_put_delayed_seq(delayed_refs, &seq_elem);
1196 	btrfs_end_transaction(trans, fs_info->extent_root);
1197 	return ret;
1198 }
1199 
1200 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1201 				struct btrfs_path *path,
1202 				iterate_extent_inodes_t *iterate, void *ctx)
1203 {
1204 	int ret;
1205 	u64 extent_item_pos;
1206 	struct btrfs_key found_key;
1207 
1208 	ret = extent_from_logical(fs_info, logical, path,
1209 					&found_key);
1210 	btrfs_release_path(path);
1211 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1212 		ret = -EINVAL;
1213 	if (ret < 0)
1214 		return ret;
1215 
1216 	extent_item_pos = logical - found_key.objectid;
1217 	ret = iterate_extent_inodes(fs_info, path, found_key.objectid,
1218 					extent_item_pos, iterate, ctx);
1219 
1220 	return ret;
1221 }
1222 
1223 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1224 				struct btrfs_path *path,
1225 				iterate_irefs_t *iterate, void *ctx)
1226 {
1227 	int ret;
1228 	int slot;
1229 	u32 cur;
1230 	u32 len;
1231 	u32 name_len;
1232 	u64 parent = 0;
1233 	int found = 0;
1234 	struct extent_buffer *eb;
1235 	struct btrfs_item *item;
1236 	struct btrfs_inode_ref *iref;
1237 	struct btrfs_key found_key;
1238 
1239 	while (1) {
1240 		ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1241 					&found_key);
1242 		if (ret < 0)
1243 			break;
1244 		if (ret) {
1245 			ret = found ? 0 : -ENOENT;
1246 			break;
1247 		}
1248 		++found;
1249 
1250 		parent = found_key.offset;
1251 		slot = path->slots[0];
1252 		eb = path->nodes[0];
1253 		/* make sure we can use eb after releasing the path */
1254 		atomic_inc(&eb->refs);
1255 		btrfs_release_path(path);
1256 
1257 		item = btrfs_item_nr(eb, slot);
1258 		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1259 
1260 		for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1261 			name_len = btrfs_inode_ref_name_len(eb, iref);
1262 			/* path must be released before calling iterate()! */
1263 			pr_debug("following ref at offset %u for inode %llu in "
1264 				 "tree %llu\n", cur,
1265 				 (unsigned long long)found_key.objectid,
1266 				 (unsigned long long)fs_root->objectid);
1267 			ret = iterate(parent, iref, eb, ctx);
1268 			if (ret) {
1269 				free_extent_buffer(eb);
1270 				break;
1271 			}
1272 			len = sizeof(*iref) + name_len;
1273 			iref = (struct btrfs_inode_ref *)((char *)iref + len);
1274 		}
1275 		free_extent_buffer(eb);
1276 	}
1277 
1278 	btrfs_release_path(path);
1279 
1280 	return ret;
1281 }
1282 
1283 /*
1284  * returns 0 if the path could be dumped (probably truncated)
1285  * returns <0 in case of an error
1286  */
1287 static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
1288 				struct extent_buffer *eb, void *ctx)
1289 {
1290 	struct inode_fs_paths *ipath = ctx;
1291 	char *fspath;
1292 	char *fspath_min;
1293 	int i = ipath->fspath->elem_cnt;
1294 	const int s_ptr = sizeof(char *);
1295 	u32 bytes_left;
1296 
1297 	bytes_left = ipath->fspath->bytes_left > s_ptr ?
1298 					ipath->fspath->bytes_left - s_ptr : 0;
1299 
1300 	fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1301 	fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
1302 				inum, fspath_min, bytes_left);
1303 	if (IS_ERR(fspath))
1304 		return PTR_ERR(fspath);
1305 
1306 	if (fspath > fspath_min) {
1307 		pr_debug("path resolved: %s\n", fspath);
1308 		ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1309 		++ipath->fspath->elem_cnt;
1310 		ipath->fspath->bytes_left = fspath - fspath_min;
1311 	} else {
1312 		pr_debug("missed path, not enough space. missing bytes: %lu, "
1313 			 "constructed so far: %s\n",
1314 			 (unsigned long)(fspath_min - fspath), fspath_min);
1315 		++ipath->fspath->elem_missed;
1316 		ipath->fspath->bytes_missing += fspath_min - fspath;
1317 		ipath->fspath->bytes_left = 0;
1318 	}
1319 
1320 	return 0;
1321 }
1322 
1323 /*
1324  * this dumps all file system paths to the inode into the ipath struct, provided
1325  * is has been created large enough. each path is zero-terminated and accessed
1326  * from ipath->fspath->val[i].
1327  * when it returns, there are ipath->fspath->elem_cnt number of paths available
1328  * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1329  * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1330  * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1331  * have been needed to return all paths.
1332  */
1333 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1334 {
1335 	return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1336 				inode_to_path, ipath);
1337 }
1338 
1339 /*
1340  * allocates space to return multiple file system paths for an inode.
1341  * total_bytes to allocate are passed, note that space usable for actual path
1342  * information will be total_bytes - sizeof(struct inode_fs_paths).
1343  * the returned pointer must be freed with free_ipath() in the end.
1344  */
1345 struct btrfs_data_container *init_data_container(u32 total_bytes)
1346 {
1347 	struct btrfs_data_container *data;
1348 	size_t alloc_bytes;
1349 
1350 	alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1351 	data = kmalloc(alloc_bytes, GFP_NOFS);
1352 	if (!data)
1353 		return ERR_PTR(-ENOMEM);
1354 
1355 	if (total_bytes >= sizeof(*data)) {
1356 		data->bytes_left = total_bytes - sizeof(*data);
1357 		data->bytes_missing = 0;
1358 	} else {
1359 		data->bytes_missing = sizeof(*data) - total_bytes;
1360 		data->bytes_left = 0;
1361 	}
1362 
1363 	data->elem_cnt = 0;
1364 	data->elem_missed = 0;
1365 
1366 	return data;
1367 }
1368 
1369 /*
1370  * allocates space to return multiple file system paths for an inode.
1371  * total_bytes to allocate are passed, note that space usable for actual path
1372  * information will be total_bytes - sizeof(struct inode_fs_paths).
1373  * the returned pointer must be freed with free_ipath() in the end.
1374  */
1375 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1376 					struct btrfs_path *path)
1377 {
1378 	struct inode_fs_paths *ifp;
1379 	struct btrfs_data_container *fspath;
1380 
1381 	fspath = init_data_container(total_bytes);
1382 	if (IS_ERR(fspath))
1383 		return (void *)fspath;
1384 
1385 	ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1386 	if (!ifp) {
1387 		kfree(fspath);
1388 		return ERR_PTR(-ENOMEM);
1389 	}
1390 
1391 	ifp->btrfs_path = path;
1392 	ifp->fspath = fspath;
1393 	ifp->fs_root = fs_root;
1394 
1395 	return ifp;
1396 }
1397 
1398 void free_ipath(struct inode_fs_paths *ipath)
1399 {
1400 	kfree(ipath);
1401 }
1402