xref: /openbmc/linux/fs/btrfs/backref.c (revision 538f72cdf03cad1c21c551ea542c8ce7d9fa2d81)
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 <linux/vmalloc.h>
20 #include "ctree.h"
21 #include "disk-io.h"
22 #include "backref.h"
23 #include "ulist.h"
24 #include "transaction.h"
25 #include "delayed-ref.h"
26 #include "locking.h"
27 
28 struct extent_inode_elem {
29 	u64 inum;
30 	u64 offset;
31 	struct extent_inode_elem *next;
32 };
33 
34 static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb,
35 				struct btrfs_file_extent_item *fi,
36 				u64 extent_item_pos,
37 				struct extent_inode_elem **eie)
38 {
39 	u64 offset = 0;
40 	struct extent_inode_elem *e;
41 
42 	if (!btrfs_file_extent_compression(eb, fi) &&
43 	    !btrfs_file_extent_encryption(eb, fi) &&
44 	    !btrfs_file_extent_other_encoding(eb, fi)) {
45 		u64 data_offset;
46 		u64 data_len;
47 
48 		data_offset = btrfs_file_extent_offset(eb, fi);
49 		data_len = btrfs_file_extent_num_bytes(eb, fi);
50 
51 		if (extent_item_pos < data_offset ||
52 		    extent_item_pos >= data_offset + data_len)
53 			return 1;
54 		offset = extent_item_pos - data_offset;
55 	}
56 
57 	e = kmalloc(sizeof(*e), GFP_NOFS);
58 	if (!e)
59 		return -ENOMEM;
60 
61 	e->next = *eie;
62 	e->inum = key->objectid;
63 	e->offset = key->offset + offset;
64 	*eie = e;
65 
66 	return 0;
67 }
68 
69 static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
70 				u64 extent_item_pos,
71 				struct extent_inode_elem **eie)
72 {
73 	u64 disk_byte;
74 	struct btrfs_key key;
75 	struct btrfs_file_extent_item *fi;
76 	int slot;
77 	int nritems;
78 	int extent_type;
79 	int ret;
80 
81 	/*
82 	 * from the shared data ref, we only have the leaf but we need
83 	 * the key. thus, we must look into all items and see that we
84 	 * find one (some) with a reference to our extent item.
85 	 */
86 	nritems = btrfs_header_nritems(eb);
87 	for (slot = 0; slot < nritems; ++slot) {
88 		btrfs_item_key_to_cpu(eb, &key, slot);
89 		if (key.type != BTRFS_EXTENT_DATA_KEY)
90 			continue;
91 		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
92 		extent_type = btrfs_file_extent_type(eb, fi);
93 		if (extent_type == BTRFS_FILE_EXTENT_INLINE)
94 			continue;
95 		/* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
96 		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
97 		if (disk_byte != wanted_disk_byte)
98 			continue;
99 
100 		ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
101 		if (ret < 0)
102 			return ret;
103 	}
104 
105 	return 0;
106 }
107 
108 /*
109  * this structure records all encountered refs on the way up to the root
110  */
111 struct __prelim_ref {
112 	struct list_head list;
113 	u64 root_id;
114 	struct btrfs_key key_for_search;
115 	int level;
116 	int count;
117 	struct extent_inode_elem *inode_list;
118 	u64 parent;
119 	u64 wanted_disk_byte;
120 };
121 
122 static struct kmem_cache *btrfs_prelim_ref_cache;
123 
124 int __init btrfs_prelim_ref_init(void)
125 {
126 	btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
127 					sizeof(struct __prelim_ref),
128 					0,
129 					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
130 					NULL);
131 	if (!btrfs_prelim_ref_cache)
132 		return -ENOMEM;
133 	return 0;
134 }
135 
136 void btrfs_prelim_ref_exit(void)
137 {
138 	if (btrfs_prelim_ref_cache)
139 		kmem_cache_destroy(btrfs_prelim_ref_cache);
140 }
141 
142 /*
143  * the rules for all callers of this function are:
144  * - obtaining the parent is the goal
145  * - if you add a key, you must know that it is a correct key
146  * - if you cannot add the parent or a correct key, then we will look into the
147  *   block later to set a correct key
148  *
149  * delayed refs
150  * ============
151  *        backref type | shared | indirect | shared | indirect
152  * information         |   tree |     tree |   data |     data
153  * --------------------+--------+----------+--------+----------
154  *      parent logical |    y   |     -    |    -   |     -
155  *      key to resolve |    -   |     y    |    y   |     y
156  *  tree block logical |    -   |     -    |    -   |     -
157  *  root for resolving |    y   |     y    |    y   |     y
158  *
159  * - column 1:       we've the parent -> done
160  * - column 2, 3, 4: we use the key to find the parent
161  *
162  * on disk refs (inline or keyed)
163  * ==============================
164  *        backref type | shared | indirect | shared | indirect
165  * information         |   tree |     tree |   data |     data
166  * --------------------+--------+----------+--------+----------
167  *      parent logical |    y   |     -    |    y   |     -
168  *      key to resolve |    -   |     -    |    -   |     y
169  *  tree block logical |    y   |     y    |    y   |     y
170  *  root for resolving |    -   |     y    |    y   |     y
171  *
172  * - column 1, 3: we've the parent -> done
173  * - column 2:    we take the first key from the block to find the parent
174  *                (see __add_missing_keys)
175  * - column 4:    we use the key to find the parent
176  *
177  * additional information that's available but not required to find the parent
178  * block might help in merging entries to gain some speed.
179  */
180 
181 static int __add_prelim_ref(struct list_head *head, u64 root_id,
182 			    struct btrfs_key *key, int level,
183 			    u64 parent, u64 wanted_disk_byte, int count,
184 			    gfp_t gfp_mask)
185 {
186 	struct __prelim_ref *ref;
187 
188 	if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
189 		return 0;
190 
191 	ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
192 	if (!ref)
193 		return -ENOMEM;
194 
195 	ref->root_id = root_id;
196 	if (key)
197 		ref->key_for_search = *key;
198 	else
199 		memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
200 
201 	ref->inode_list = NULL;
202 	ref->level = level;
203 	ref->count = count;
204 	ref->parent = parent;
205 	ref->wanted_disk_byte = wanted_disk_byte;
206 	list_add_tail(&ref->list, head);
207 
208 	return 0;
209 }
210 
211 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
212 			   struct ulist *parents, struct __prelim_ref *ref,
213 			   int level, u64 time_seq, const u64 *extent_item_pos)
214 {
215 	int ret = 0;
216 	int slot;
217 	struct extent_buffer *eb;
218 	struct btrfs_key key;
219 	struct btrfs_key *key_for_search = &ref->key_for_search;
220 	struct btrfs_file_extent_item *fi;
221 	struct extent_inode_elem *eie = NULL, *old = NULL;
222 	u64 disk_byte;
223 	u64 wanted_disk_byte = ref->wanted_disk_byte;
224 	u64 count = 0;
225 
226 	if (level != 0) {
227 		eb = path->nodes[level];
228 		ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
229 		if (ret < 0)
230 			return ret;
231 		return 0;
232 	}
233 
234 	/*
235 	 * We normally enter this function with the path already pointing to
236 	 * the first item to check. But sometimes, we may enter it with
237 	 * slot==nritems. In that case, go to the next leaf before we continue.
238 	 */
239 	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0]))
240 		ret = btrfs_next_old_leaf(root, path, time_seq);
241 
242 	while (!ret && count < ref->count) {
243 		eb = path->nodes[0];
244 		slot = path->slots[0];
245 
246 		btrfs_item_key_to_cpu(eb, &key, slot);
247 
248 		if (key.objectid != key_for_search->objectid ||
249 		    key.type != BTRFS_EXTENT_DATA_KEY)
250 			break;
251 
252 		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
253 		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
254 
255 		if (disk_byte == wanted_disk_byte) {
256 			eie = NULL;
257 			old = NULL;
258 			count++;
259 			if (extent_item_pos) {
260 				ret = check_extent_in_eb(&key, eb, fi,
261 						*extent_item_pos,
262 						&eie);
263 				if (ret < 0)
264 					break;
265 			}
266 			if (ret > 0)
267 				goto next;
268 			ret = ulist_add_merge(parents, eb->start,
269 					      (uintptr_t)eie,
270 					      (u64 *)&old, GFP_NOFS);
271 			if (ret < 0)
272 				break;
273 			if (!ret && extent_item_pos) {
274 				while (old->next)
275 					old = old->next;
276 				old->next = eie;
277 			}
278 		}
279 next:
280 		ret = btrfs_next_old_item(root, path, time_seq);
281 	}
282 
283 	if (ret > 0)
284 		ret = 0;
285 	return ret;
286 }
287 
288 /*
289  * resolve an indirect backref in the form (root_id, key, level)
290  * to a logical address
291  */
292 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
293 				  struct btrfs_path *path, u64 time_seq,
294 				  struct __prelim_ref *ref,
295 				  struct ulist *parents,
296 				  const u64 *extent_item_pos)
297 {
298 	struct btrfs_root *root;
299 	struct btrfs_key root_key;
300 	struct extent_buffer *eb;
301 	int ret = 0;
302 	int root_level;
303 	int level = ref->level;
304 	int index;
305 
306 	root_key.objectid = ref->root_id;
307 	root_key.type = BTRFS_ROOT_ITEM_KEY;
308 	root_key.offset = (u64)-1;
309 
310 	index = srcu_read_lock(&fs_info->subvol_srcu);
311 
312 	root = btrfs_read_fs_root_no_name(fs_info, &root_key);
313 	if (IS_ERR(root)) {
314 		srcu_read_unlock(&fs_info->subvol_srcu, index);
315 		ret = PTR_ERR(root);
316 		goto out;
317 	}
318 
319 	root_level = btrfs_old_root_level(root, time_seq);
320 
321 	if (root_level + 1 == level) {
322 		srcu_read_unlock(&fs_info->subvol_srcu, index);
323 		goto out;
324 	}
325 
326 	path->lowest_level = level;
327 	ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
328 
329 	/* root node has been locked, we can release @subvol_srcu safely here */
330 	srcu_read_unlock(&fs_info->subvol_srcu, index);
331 
332 	pr_debug("search slot in root %llu (level %d, ref count %d) returned "
333 		 "%d for key (%llu %u %llu)\n",
334 		 ref->root_id, level, ref->count, ret,
335 		 ref->key_for_search.objectid, ref->key_for_search.type,
336 		 ref->key_for_search.offset);
337 	if (ret < 0)
338 		goto out;
339 
340 	eb = path->nodes[level];
341 	while (!eb) {
342 		if (WARN_ON(!level)) {
343 			ret = 1;
344 			goto out;
345 		}
346 		level--;
347 		eb = path->nodes[level];
348 	}
349 
350 	ret = add_all_parents(root, path, parents, ref, level, time_seq,
351 			      extent_item_pos);
352 out:
353 	path->lowest_level = 0;
354 	btrfs_release_path(path);
355 	return ret;
356 }
357 
358 /*
359  * resolve all indirect backrefs from the list
360  */
361 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
362 				   struct btrfs_path *path, u64 time_seq,
363 				   struct list_head *head,
364 				   const u64 *extent_item_pos)
365 {
366 	int err;
367 	int ret = 0;
368 	struct __prelim_ref *ref;
369 	struct __prelim_ref *ref_safe;
370 	struct __prelim_ref *new_ref;
371 	struct ulist *parents;
372 	struct ulist_node *node;
373 	struct ulist_iterator uiter;
374 
375 	parents = ulist_alloc(GFP_NOFS);
376 	if (!parents)
377 		return -ENOMEM;
378 
379 	/*
380 	 * _safe allows us to insert directly after the current item without
381 	 * iterating over the newly inserted items.
382 	 * we're also allowed to re-assign ref during iteration.
383 	 */
384 	list_for_each_entry_safe(ref, ref_safe, head, list) {
385 		if (ref->parent)	/* already direct */
386 			continue;
387 		if (ref->count == 0)
388 			continue;
389 		err = __resolve_indirect_ref(fs_info, path, time_seq, ref,
390 					     parents, extent_item_pos);
391 		if (err == -ENOMEM)
392 			goto out;
393 		if (err)
394 			continue;
395 
396 		/* we put the first parent into the ref at hand */
397 		ULIST_ITER_INIT(&uiter);
398 		node = ulist_next(parents, &uiter);
399 		ref->parent = node ? node->val : 0;
400 		ref->inode_list = node ?
401 			(struct extent_inode_elem *)(uintptr_t)node->aux : NULL;
402 
403 		/* additional parents require new refs being added here */
404 		while ((node = ulist_next(parents, &uiter))) {
405 			new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
406 						   GFP_NOFS);
407 			if (!new_ref) {
408 				ret = -ENOMEM;
409 				goto out;
410 			}
411 			memcpy(new_ref, ref, sizeof(*ref));
412 			new_ref->parent = node->val;
413 			new_ref->inode_list = (struct extent_inode_elem *)
414 							(uintptr_t)node->aux;
415 			list_add(&new_ref->list, &ref->list);
416 		}
417 		ulist_reinit(parents);
418 	}
419 out:
420 	ulist_free(parents);
421 	return ret;
422 }
423 
424 static inline int ref_for_same_block(struct __prelim_ref *ref1,
425 				     struct __prelim_ref *ref2)
426 {
427 	if (ref1->level != ref2->level)
428 		return 0;
429 	if (ref1->root_id != ref2->root_id)
430 		return 0;
431 	if (ref1->key_for_search.type != ref2->key_for_search.type)
432 		return 0;
433 	if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
434 		return 0;
435 	if (ref1->key_for_search.offset != ref2->key_for_search.offset)
436 		return 0;
437 	if (ref1->parent != ref2->parent)
438 		return 0;
439 
440 	return 1;
441 }
442 
443 /*
444  * read tree blocks and add keys where required.
445  */
446 static int __add_missing_keys(struct btrfs_fs_info *fs_info,
447 			      struct list_head *head)
448 {
449 	struct list_head *pos;
450 	struct extent_buffer *eb;
451 
452 	list_for_each(pos, head) {
453 		struct __prelim_ref *ref;
454 		ref = list_entry(pos, struct __prelim_ref, list);
455 
456 		if (ref->parent)
457 			continue;
458 		if (ref->key_for_search.type)
459 			continue;
460 		BUG_ON(!ref->wanted_disk_byte);
461 		eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
462 				     fs_info->tree_root->leafsize, 0);
463 		if (!eb || !extent_buffer_uptodate(eb)) {
464 			free_extent_buffer(eb);
465 			return -EIO;
466 		}
467 		btrfs_tree_read_lock(eb);
468 		if (btrfs_header_level(eb) == 0)
469 			btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
470 		else
471 			btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
472 		btrfs_tree_read_unlock(eb);
473 		free_extent_buffer(eb);
474 	}
475 	return 0;
476 }
477 
478 /*
479  * merge two lists of backrefs and adjust counts accordingly
480  *
481  * mode = 1: merge identical keys, if key is set
482  *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
483  *           additionally, we could even add a key range for the blocks we
484  *           looked into to merge even more (-> replace unresolved refs by those
485  *           having a parent).
486  * mode = 2: merge identical parents
487  */
488 static void __merge_refs(struct list_head *head, int mode)
489 {
490 	struct list_head *pos1;
491 
492 	list_for_each(pos1, head) {
493 		struct list_head *n2;
494 		struct list_head *pos2;
495 		struct __prelim_ref *ref1;
496 
497 		ref1 = list_entry(pos1, struct __prelim_ref, list);
498 
499 		for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
500 		     pos2 = n2, n2 = pos2->next) {
501 			struct __prelim_ref *ref2;
502 			struct __prelim_ref *xchg;
503 			struct extent_inode_elem *eie;
504 
505 			ref2 = list_entry(pos2, struct __prelim_ref, list);
506 
507 			if (mode == 1) {
508 				if (!ref_for_same_block(ref1, ref2))
509 					continue;
510 				if (!ref1->parent && ref2->parent) {
511 					xchg = ref1;
512 					ref1 = ref2;
513 					ref2 = xchg;
514 				}
515 			} else {
516 				if (ref1->parent != ref2->parent)
517 					continue;
518 			}
519 
520 			eie = ref1->inode_list;
521 			while (eie && eie->next)
522 				eie = eie->next;
523 			if (eie)
524 				eie->next = ref2->inode_list;
525 			else
526 				ref1->inode_list = ref2->inode_list;
527 			ref1->count += ref2->count;
528 
529 			list_del(&ref2->list);
530 			kmem_cache_free(btrfs_prelim_ref_cache, ref2);
531 		}
532 
533 	}
534 }
535 
536 /*
537  * add all currently queued delayed refs from this head whose seq nr is
538  * smaller or equal that seq to the list
539  */
540 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
541 			      struct list_head *prefs)
542 {
543 	struct btrfs_delayed_extent_op *extent_op = head->extent_op;
544 	struct rb_node *n = &head->node.rb_node;
545 	struct btrfs_key key;
546 	struct btrfs_key op_key = {0};
547 	int sgn;
548 	int ret = 0;
549 
550 	if (extent_op && extent_op->update_key)
551 		btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
552 
553 	spin_lock(&head->lock);
554 	n = rb_first(&head->ref_root);
555 	while (n) {
556 		struct btrfs_delayed_ref_node *node;
557 		node = rb_entry(n, struct btrfs_delayed_ref_node,
558 				rb_node);
559 		n = rb_next(n);
560 		if (node->seq > seq)
561 			continue;
562 
563 		switch (node->action) {
564 		case BTRFS_ADD_DELAYED_EXTENT:
565 		case BTRFS_UPDATE_DELAYED_HEAD:
566 			WARN_ON(1);
567 			continue;
568 		case BTRFS_ADD_DELAYED_REF:
569 			sgn = 1;
570 			break;
571 		case BTRFS_DROP_DELAYED_REF:
572 			sgn = -1;
573 			break;
574 		default:
575 			BUG_ON(1);
576 		}
577 		switch (node->type) {
578 		case BTRFS_TREE_BLOCK_REF_KEY: {
579 			struct btrfs_delayed_tree_ref *ref;
580 
581 			ref = btrfs_delayed_node_to_tree_ref(node);
582 			ret = __add_prelim_ref(prefs, ref->root, &op_key,
583 					       ref->level + 1, 0, node->bytenr,
584 					       node->ref_mod * sgn, GFP_ATOMIC);
585 			break;
586 		}
587 		case BTRFS_SHARED_BLOCK_REF_KEY: {
588 			struct btrfs_delayed_tree_ref *ref;
589 
590 			ref = btrfs_delayed_node_to_tree_ref(node);
591 			ret = __add_prelim_ref(prefs, ref->root, NULL,
592 					       ref->level + 1, ref->parent,
593 					       node->bytenr,
594 					       node->ref_mod * sgn, GFP_ATOMIC);
595 			break;
596 		}
597 		case BTRFS_EXTENT_DATA_REF_KEY: {
598 			struct btrfs_delayed_data_ref *ref;
599 			ref = btrfs_delayed_node_to_data_ref(node);
600 
601 			key.objectid = ref->objectid;
602 			key.type = BTRFS_EXTENT_DATA_KEY;
603 			key.offset = ref->offset;
604 			ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
605 					       node->bytenr,
606 					       node->ref_mod * sgn, GFP_ATOMIC);
607 			break;
608 		}
609 		case BTRFS_SHARED_DATA_REF_KEY: {
610 			struct btrfs_delayed_data_ref *ref;
611 
612 			ref = btrfs_delayed_node_to_data_ref(node);
613 
614 			key.objectid = ref->objectid;
615 			key.type = BTRFS_EXTENT_DATA_KEY;
616 			key.offset = ref->offset;
617 			ret = __add_prelim_ref(prefs, ref->root, &key, 0,
618 					       ref->parent, node->bytenr,
619 					       node->ref_mod * sgn, GFP_ATOMIC);
620 			break;
621 		}
622 		default:
623 			WARN_ON(1);
624 		}
625 		if (ret)
626 			break;
627 	}
628 	spin_unlock(&head->lock);
629 	return ret;
630 }
631 
632 /*
633  * add all inline backrefs for bytenr to the list
634  */
635 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
636 			     struct btrfs_path *path, u64 bytenr,
637 			     int *info_level, struct list_head *prefs)
638 {
639 	int ret = 0;
640 	int slot;
641 	struct extent_buffer *leaf;
642 	struct btrfs_key key;
643 	struct btrfs_key found_key;
644 	unsigned long ptr;
645 	unsigned long end;
646 	struct btrfs_extent_item *ei;
647 	u64 flags;
648 	u64 item_size;
649 
650 	/*
651 	 * enumerate all inline refs
652 	 */
653 	leaf = path->nodes[0];
654 	slot = path->slots[0];
655 
656 	item_size = btrfs_item_size_nr(leaf, slot);
657 	BUG_ON(item_size < sizeof(*ei));
658 
659 	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
660 	flags = btrfs_extent_flags(leaf, ei);
661 	btrfs_item_key_to_cpu(leaf, &found_key, slot);
662 
663 	ptr = (unsigned long)(ei + 1);
664 	end = (unsigned long)ei + item_size;
665 
666 	if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
667 	    flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
668 		struct btrfs_tree_block_info *info;
669 
670 		info = (struct btrfs_tree_block_info *)ptr;
671 		*info_level = btrfs_tree_block_level(leaf, info);
672 		ptr += sizeof(struct btrfs_tree_block_info);
673 		BUG_ON(ptr > end);
674 	} else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
675 		*info_level = found_key.offset;
676 	} else {
677 		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
678 	}
679 
680 	while (ptr < end) {
681 		struct btrfs_extent_inline_ref *iref;
682 		u64 offset;
683 		int type;
684 
685 		iref = (struct btrfs_extent_inline_ref *)ptr;
686 		type = btrfs_extent_inline_ref_type(leaf, iref);
687 		offset = btrfs_extent_inline_ref_offset(leaf, iref);
688 
689 		switch (type) {
690 		case BTRFS_SHARED_BLOCK_REF_KEY:
691 			ret = __add_prelim_ref(prefs, 0, NULL,
692 						*info_level + 1, offset,
693 						bytenr, 1, GFP_NOFS);
694 			break;
695 		case BTRFS_SHARED_DATA_REF_KEY: {
696 			struct btrfs_shared_data_ref *sdref;
697 			int count;
698 
699 			sdref = (struct btrfs_shared_data_ref *)(iref + 1);
700 			count = btrfs_shared_data_ref_count(leaf, sdref);
701 			ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
702 					       bytenr, count, GFP_NOFS);
703 			break;
704 		}
705 		case BTRFS_TREE_BLOCK_REF_KEY:
706 			ret = __add_prelim_ref(prefs, offset, NULL,
707 					       *info_level + 1, 0,
708 					       bytenr, 1, GFP_NOFS);
709 			break;
710 		case BTRFS_EXTENT_DATA_REF_KEY: {
711 			struct btrfs_extent_data_ref *dref;
712 			int count;
713 			u64 root;
714 
715 			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
716 			count = btrfs_extent_data_ref_count(leaf, dref);
717 			key.objectid = btrfs_extent_data_ref_objectid(leaf,
718 								      dref);
719 			key.type = BTRFS_EXTENT_DATA_KEY;
720 			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
721 			root = btrfs_extent_data_ref_root(leaf, dref);
722 			ret = __add_prelim_ref(prefs, root, &key, 0, 0,
723 					       bytenr, count, GFP_NOFS);
724 			break;
725 		}
726 		default:
727 			WARN_ON(1);
728 		}
729 		if (ret)
730 			return ret;
731 		ptr += btrfs_extent_inline_ref_size(type);
732 	}
733 
734 	return 0;
735 }
736 
737 /*
738  * add all non-inline backrefs for bytenr to the list
739  */
740 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
741 			    struct btrfs_path *path, u64 bytenr,
742 			    int info_level, struct list_head *prefs)
743 {
744 	struct btrfs_root *extent_root = fs_info->extent_root;
745 	int ret;
746 	int slot;
747 	struct extent_buffer *leaf;
748 	struct btrfs_key key;
749 
750 	while (1) {
751 		ret = btrfs_next_item(extent_root, path);
752 		if (ret < 0)
753 			break;
754 		if (ret) {
755 			ret = 0;
756 			break;
757 		}
758 
759 		slot = path->slots[0];
760 		leaf = path->nodes[0];
761 		btrfs_item_key_to_cpu(leaf, &key, slot);
762 
763 		if (key.objectid != bytenr)
764 			break;
765 		if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
766 			continue;
767 		if (key.type > BTRFS_SHARED_DATA_REF_KEY)
768 			break;
769 
770 		switch (key.type) {
771 		case BTRFS_SHARED_BLOCK_REF_KEY:
772 			ret = __add_prelim_ref(prefs, 0, NULL,
773 						info_level + 1, key.offset,
774 						bytenr, 1, GFP_NOFS);
775 			break;
776 		case BTRFS_SHARED_DATA_REF_KEY: {
777 			struct btrfs_shared_data_ref *sdref;
778 			int count;
779 
780 			sdref = btrfs_item_ptr(leaf, slot,
781 					      struct btrfs_shared_data_ref);
782 			count = btrfs_shared_data_ref_count(leaf, sdref);
783 			ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
784 						bytenr, count, GFP_NOFS);
785 			break;
786 		}
787 		case BTRFS_TREE_BLOCK_REF_KEY:
788 			ret = __add_prelim_ref(prefs, key.offset, NULL,
789 					       info_level + 1, 0,
790 					       bytenr, 1, GFP_NOFS);
791 			break;
792 		case BTRFS_EXTENT_DATA_REF_KEY: {
793 			struct btrfs_extent_data_ref *dref;
794 			int count;
795 			u64 root;
796 
797 			dref = btrfs_item_ptr(leaf, slot,
798 					      struct btrfs_extent_data_ref);
799 			count = btrfs_extent_data_ref_count(leaf, dref);
800 			key.objectid = btrfs_extent_data_ref_objectid(leaf,
801 								      dref);
802 			key.type = BTRFS_EXTENT_DATA_KEY;
803 			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
804 			root = btrfs_extent_data_ref_root(leaf, dref);
805 			ret = __add_prelim_ref(prefs, root, &key, 0, 0,
806 					       bytenr, count, GFP_NOFS);
807 			break;
808 		}
809 		default:
810 			WARN_ON(1);
811 		}
812 		if (ret)
813 			return ret;
814 
815 	}
816 
817 	return ret;
818 }
819 
820 /*
821  * this adds all existing backrefs (inline backrefs, backrefs and delayed
822  * refs) for the given bytenr to the refs list, merges duplicates and resolves
823  * indirect refs to their parent bytenr.
824  * When roots are found, they're added to the roots list
825  *
826  * FIXME some caching might speed things up
827  */
828 static int find_parent_nodes(struct btrfs_trans_handle *trans,
829 			     struct btrfs_fs_info *fs_info, u64 bytenr,
830 			     u64 time_seq, struct ulist *refs,
831 			     struct ulist *roots, const u64 *extent_item_pos)
832 {
833 	struct btrfs_key key;
834 	struct btrfs_path *path;
835 	struct btrfs_delayed_ref_root *delayed_refs = NULL;
836 	struct btrfs_delayed_ref_head *head;
837 	int info_level = 0;
838 	int ret;
839 	struct list_head prefs_delayed;
840 	struct list_head prefs;
841 	struct __prelim_ref *ref;
842 
843 	INIT_LIST_HEAD(&prefs);
844 	INIT_LIST_HEAD(&prefs_delayed);
845 
846 	key.objectid = bytenr;
847 	key.offset = (u64)-1;
848 	if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
849 		key.type = BTRFS_METADATA_ITEM_KEY;
850 	else
851 		key.type = BTRFS_EXTENT_ITEM_KEY;
852 
853 	path = btrfs_alloc_path();
854 	if (!path)
855 		return -ENOMEM;
856 	if (!trans)
857 		path->search_commit_root = 1;
858 
859 	/*
860 	 * grab both a lock on the path and a lock on the delayed ref head.
861 	 * We need both to get a consistent picture of how the refs look
862 	 * at a specified point in time
863 	 */
864 again:
865 	head = NULL;
866 
867 	ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
868 	if (ret < 0)
869 		goto out;
870 	BUG_ON(ret == 0);
871 
872 	if (trans) {
873 		/*
874 		 * look if there are updates for this ref queued and lock the
875 		 * head
876 		 */
877 		delayed_refs = &trans->transaction->delayed_refs;
878 		spin_lock(&delayed_refs->lock);
879 		head = btrfs_find_delayed_ref_head(trans, bytenr);
880 		if (head) {
881 			if (!mutex_trylock(&head->mutex)) {
882 				atomic_inc(&head->node.refs);
883 				spin_unlock(&delayed_refs->lock);
884 
885 				btrfs_release_path(path);
886 
887 				/*
888 				 * Mutex was contended, block until it's
889 				 * released and try again
890 				 */
891 				mutex_lock(&head->mutex);
892 				mutex_unlock(&head->mutex);
893 				btrfs_put_delayed_ref(&head->node);
894 				goto again;
895 			}
896 			spin_unlock(&delayed_refs->lock);
897 			ret = __add_delayed_refs(head, time_seq,
898 						 &prefs_delayed);
899 			mutex_unlock(&head->mutex);
900 			if (ret)
901 				goto out;
902 		} else {
903 			spin_unlock(&delayed_refs->lock);
904 		}
905 	}
906 
907 	if (path->slots[0]) {
908 		struct extent_buffer *leaf;
909 		int slot;
910 
911 		path->slots[0]--;
912 		leaf = path->nodes[0];
913 		slot = path->slots[0];
914 		btrfs_item_key_to_cpu(leaf, &key, slot);
915 		if (key.objectid == bytenr &&
916 		    (key.type == BTRFS_EXTENT_ITEM_KEY ||
917 		     key.type == BTRFS_METADATA_ITEM_KEY)) {
918 			ret = __add_inline_refs(fs_info, path, bytenr,
919 						&info_level, &prefs);
920 			if (ret)
921 				goto out;
922 			ret = __add_keyed_refs(fs_info, path, bytenr,
923 					       info_level, &prefs);
924 			if (ret)
925 				goto out;
926 		}
927 	}
928 	btrfs_release_path(path);
929 
930 	list_splice_init(&prefs_delayed, &prefs);
931 
932 	ret = __add_missing_keys(fs_info, &prefs);
933 	if (ret)
934 		goto out;
935 
936 	__merge_refs(&prefs, 1);
937 
938 	ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs,
939 				      extent_item_pos);
940 	if (ret)
941 		goto out;
942 
943 	__merge_refs(&prefs, 2);
944 
945 	while (!list_empty(&prefs)) {
946 		ref = list_first_entry(&prefs, struct __prelim_ref, list);
947 		WARN_ON(ref->count < 0);
948 		if (ref->count && ref->root_id && ref->parent == 0) {
949 			/* no parent == root of tree */
950 			ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
951 			if (ret < 0)
952 				goto out;
953 		}
954 		if (ref->count && ref->parent) {
955 			struct extent_inode_elem *eie = NULL;
956 			if (extent_item_pos && !ref->inode_list) {
957 				u32 bsz;
958 				struct extent_buffer *eb;
959 				bsz = btrfs_level_size(fs_info->extent_root,
960 							info_level);
961 				eb = read_tree_block(fs_info->extent_root,
962 							   ref->parent, bsz, 0);
963 				if (!eb || !extent_buffer_uptodate(eb)) {
964 					free_extent_buffer(eb);
965 					ret = -EIO;
966 					goto out;
967 				}
968 				ret = find_extent_in_eb(eb, bytenr,
969 							*extent_item_pos, &eie);
970 				free_extent_buffer(eb);
971 				if (ret < 0)
972 					goto out;
973 				ref->inode_list = eie;
974 			}
975 			ret = ulist_add_merge(refs, ref->parent,
976 					      (uintptr_t)ref->inode_list,
977 					      (u64 *)&eie, GFP_NOFS);
978 			if (ret < 0)
979 				goto out;
980 			if (!ret && extent_item_pos) {
981 				/*
982 				 * we've recorded that parent, so we must extend
983 				 * its inode list here
984 				 */
985 				BUG_ON(!eie);
986 				while (eie->next)
987 					eie = eie->next;
988 				eie->next = ref->inode_list;
989 			}
990 		}
991 		list_del(&ref->list);
992 		kmem_cache_free(btrfs_prelim_ref_cache, ref);
993 	}
994 
995 out:
996 	btrfs_free_path(path);
997 	while (!list_empty(&prefs)) {
998 		ref = list_first_entry(&prefs, struct __prelim_ref, list);
999 		list_del(&ref->list);
1000 		kmem_cache_free(btrfs_prelim_ref_cache, ref);
1001 	}
1002 	while (!list_empty(&prefs_delayed)) {
1003 		ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
1004 				       list);
1005 		list_del(&ref->list);
1006 		kmem_cache_free(btrfs_prelim_ref_cache, ref);
1007 	}
1008 
1009 	return ret;
1010 }
1011 
1012 static void free_leaf_list(struct ulist *blocks)
1013 {
1014 	struct ulist_node *node = NULL;
1015 	struct extent_inode_elem *eie;
1016 	struct extent_inode_elem *eie_next;
1017 	struct ulist_iterator uiter;
1018 
1019 	ULIST_ITER_INIT(&uiter);
1020 	while ((node = ulist_next(blocks, &uiter))) {
1021 		if (!node->aux)
1022 			continue;
1023 		eie = (struct extent_inode_elem *)(uintptr_t)node->aux;
1024 		for (; eie; eie = eie_next) {
1025 			eie_next = eie->next;
1026 			kfree(eie);
1027 		}
1028 		node->aux = 0;
1029 	}
1030 
1031 	ulist_free(blocks);
1032 }
1033 
1034 /*
1035  * Finds all leafs with a reference to the specified combination of bytenr and
1036  * offset. key_list_head will point to a list of corresponding keys (caller must
1037  * free each list element). The leafs will be stored in the leafs ulist, which
1038  * must be freed with ulist_free.
1039  *
1040  * returns 0 on success, <0 on error
1041  */
1042 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
1043 				struct btrfs_fs_info *fs_info, u64 bytenr,
1044 				u64 time_seq, struct ulist **leafs,
1045 				const u64 *extent_item_pos)
1046 {
1047 	struct ulist *tmp;
1048 	int ret;
1049 
1050 	tmp = ulist_alloc(GFP_NOFS);
1051 	if (!tmp)
1052 		return -ENOMEM;
1053 	*leafs = ulist_alloc(GFP_NOFS);
1054 	if (!*leafs) {
1055 		ulist_free(tmp);
1056 		return -ENOMEM;
1057 	}
1058 
1059 	ret = find_parent_nodes(trans, fs_info, bytenr,
1060 				time_seq, *leafs, tmp, extent_item_pos);
1061 	ulist_free(tmp);
1062 
1063 	if (ret < 0 && ret != -ENOENT) {
1064 		free_leaf_list(*leafs);
1065 		return ret;
1066 	}
1067 
1068 	return 0;
1069 }
1070 
1071 /*
1072  * walk all backrefs for a given extent to find all roots that reference this
1073  * extent. Walking a backref means finding all extents that reference this
1074  * extent and in turn walk the backrefs of those, too. Naturally this is a
1075  * recursive process, but here it is implemented in an iterative fashion: We
1076  * find all referencing extents for the extent in question and put them on a
1077  * list. In turn, we find all referencing extents for those, further appending
1078  * to the list. The way we iterate the list allows adding more elements after
1079  * the current while iterating. The process stops when we reach the end of the
1080  * list. Found roots are added to the roots list.
1081  *
1082  * returns 0 on success, < 0 on error.
1083  */
1084 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1085 				struct btrfs_fs_info *fs_info, u64 bytenr,
1086 				u64 time_seq, struct ulist **roots)
1087 {
1088 	struct ulist *tmp;
1089 	struct ulist_node *node = NULL;
1090 	struct ulist_iterator uiter;
1091 	int ret;
1092 
1093 	tmp = ulist_alloc(GFP_NOFS);
1094 	if (!tmp)
1095 		return -ENOMEM;
1096 	*roots = ulist_alloc(GFP_NOFS);
1097 	if (!*roots) {
1098 		ulist_free(tmp);
1099 		return -ENOMEM;
1100 	}
1101 
1102 	ULIST_ITER_INIT(&uiter);
1103 	while (1) {
1104 		ret = find_parent_nodes(trans, fs_info, bytenr,
1105 					time_seq, tmp, *roots, NULL);
1106 		if (ret < 0 && ret != -ENOENT) {
1107 			ulist_free(tmp);
1108 			ulist_free(*roots);
1109 			return ret;
1110 		}
1111 		node = ulist_next(tmp, &uiter);
1112 		if (!node)
1113 			break;
1114 		bytenr = node->val;
1115 	}
1116 
1117 	ulist_free(tmp);
1118 	return 0;
1119 }
1120 
1121 /*
1122  * this makes the path point to (inum INODE_ITEM ioff)
1123  */
1124 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1125 			struct btrfs_path *path)
1126 {
1127 	struct btrfs_key key;
1128 	return btrfs_find_item(fs_root, path, inum, ioff,
1129 			BTRFS_INODE_ITEM_KEY, &key);
1130 }
1131 
1132 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1133 				struct btrfs_path *path,
1134 				struct btrfs_key *found_key)
1135 {
1136 	return btrfs_find_item(fs_root, path, inum, ioff,
1137 			BTRFS_INODE_REF_KEY, found_key);
1138 }
1139 
1140 int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
1141 			  u64 start_off, struct btrfs_path *path,
1142 			  struct btrfs_inode_extref **ret_extref,
1143 			  u64 *found_off)
1144 {
1145 	int ret, slot;
1146 	struct btrfs_key key;
1147 	struct btrfs_key found_key;
1148 	struct btrfs_inode_extref *extref;
1149 	struct extent_buffer *leaf;
1150 	unsigned long ptr;
1151 
1152 	key.objectid = inode_objectid;
1153 	btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY);
1154 	key.offset = start_off;
1155 
1156 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1157 	if (ret < 0)
1158 		return ret;
1159 
1160 	while (1) {
1161 		leaf = path->nodes[0];
1162 		slot = path->slots[0];
1163 		if (slot >= btrfs_header_nritems(leaf)) {
1164 			/*
1165 			 * If the item at offset is not found,
1166 			 * btrfs_search_slot will point us to the slot
1167 			 * where it should be inserted. In our case
1168 			 * that will be the slot directly before the
1169 			 * next INODE_REF_KEY_V2 item. In the case
1170 			 * that we're pointing to the last slot in a
1171 			 * leaf, we must move one leaf over.
1172 			 */
1173 			ret = btrfs_next_leaf(root, path);
1174 			if (ret) {
1175 				if (ret >= 1)
1176 					ret = -ENOENT;
1177 				break;
1178 			}
1179 			continue;
1180 		}
1181 
1182 		btrfs_item_key_to_cpu(leaf, &found_key, slot);
1183 
1184 		/*
1185 		 * Check that we're still looking at an extended ref key for
1186 		 * this particular objectid. If we have different
1187 		 * objectid or type then there are no more to be found
1188 		 * in the tree and we can exit.
1189 		 */
1190 		ret = -ENOENT;
1191 		if (found_key.objectid != inode_objectid)
1192 			break;
1193 		if (btrfs_key_type(&found_key) != BTRFS_INODE_EXTREF_KEY)
1194 			break;
1195 
1196 		ret = 0;
1197 		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1198 		extref = (struct btrfs_inode_extref *)ptr;
1199 		*ret_extref = extref;
1200 		if (found_off)
1201 			*found_off = found_key.offset;
1202 		break;
1203 	}
1204 
1205 	return ret;
1206 }
1207 
1208 /*
1209  * this iterates to turn a name (from iref/extref) into a full filesystem path.
1210  * Elements of the path are separated by '/' and the path is guaranteed to be
1211  * 0-terminated. the path is only given within the current file system.
1212  * Therefore, it never starts with a '/'. the caller is responsible to provide
1213  * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1214  * the start point of the resulting string is returned. this pointer is within
1215  * dest, normally.
1216  * in case the path buffer would overflow, the pointer is decremented further
1217  * as if output was written to the buffer, though no more output is actually
1218  * generated. that way, the caller can determine how much space would be
1219  * required for the path to fit into the buffer. in that case, the returned
1220  * value will be smaller than dest. callers must check this!
1221  */
1222 char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1223 			u32 name_len, unsigned long name_off,
1224 			struct extent_buffer *eb_in, u64 parent,
1225 			char *dest, u32 size)
1226 {
1227 	int slot;
1228 	u64 next_inum;
1229 	int ret;
1230 	s64 bytes_left = ((s64)size) - 1;
1231 	struct extent_buffer *eb = eb_in;
1232 	struct btrfs_key found_key;
1233 	int leave_spinning = path->leave_spinning;
1234 	struct btrfs_inode_ref *iref;
1235 
1236 	if (bytes_left >= 0)
1237 		dest[bytes_left] = '\0';
1238 
1239 	path->leave_spinning = 1;
1240 	while (1) {
1241 		bytes_left -= name_len;
1242 		if (bytes_left >= 0)
1243 			read_extent_buffer(eb, dest + bytes_left,
1244 					   name_off, name_len);
1245 		if (eb != eb_in) {
1246 			btrfs_tree_read_unlock_blocking(eb);
1247 			free_extent_buffer(eb);
1248 		}
1249 		ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
1250 		if (ret > 0)
1251 			ret = -ENOENT;
1252 		if (ret)
1253 			break;
1254 
1255 		next_inum = found_key.offset;
1256 
1257 		/* regular exit ahead */
1258 		if (parent == next_inum)
1259 			break;
1260 
1261 		slot = path->slots[0];
1262 		eb = path->nodes[0];
1263 		/* make sure we can use eb after releasing the path */
1264 		if (eb != eb_in) {
1265 			atomic_inc(&eb->refs);
1266 			btrfs_tree_read_lock(eb);
1267 			btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1268 		}
1269 		btrfs_release_path(path);
1270 		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1271 
1272 		name_len = btrfs_inode_ref_name_len(eb, iref);
1273 		name_off = (unsigned long)(iref + 1);
1274 
1275 		parent = next_inum;
1276 		--bytes_left;
1277 		if (bytes_left >= 0)
1278 			dest[bytes_left] = '/';
1279 	}
1280 
1281 	btrfs_release_path(path);
1282 	path->leave_spinning = leave_spinning;
1283 
1284 	if (ret)
1285 		return ERR_PTR(ret);
1286 
1287 	return dest + bytes_left;
1288 }
1289 
1290 /*
1291  * this makes the path point to (logical EXTENT_ITEM *)
1292  * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1293  * tree blocks and <0 on error.
1294  */
1295 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1296 			struct btrfs_path *path, struct btrfs_key *found_key,
1297 			u64 *flags_ret)
1298 {
1299 	int ret;
1300 	u64 flags;
1301 	u64 size = 0;
1302 	u32 item_size;
1303 	struct extent_buffer *eb;
1304 	struct btrfs_extent_item *ei;
1305 	struct btrfs_key key;
1306 
1307 	if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1308 		key.type = BTRFS_METADATA_ITEM_KEY;
1309 	else
1310 		key.type = BTRFS_EXTENT_ITEM_KEY;
1311 	key.objectid = logical;
1312 	key.offset = (u64)-1;
1313 
1314 	ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1315 	if (ret < 0)
1316 		return ret;
1317 
1318 	while (1) {
1319 		u32 nritems;
1320 		if (path->slots[0] == 0) {
1321 			btrfs_set_path_blocking(path);
1322 			ret = btrfs_prev_leaf(fs_info->extent_root, path);
1323 			if (ret != 0) {
1324 				if (ret > 0) {
1325 					pr_debug("logical %llu is not within "
1326 						 "any extent\n", logical);
1327 					ret = -ENOENT;
1328 				}
1329 				return ret;
1330 			}
1331 		} else {
1332 			path->slots[0]--;
1333 		}
1334 		nritems = btrfs_header_nritems(path->nodes[0]);
1335 		if (nritems == 0) {
1336 			pr_debug("logical %llu is not within any extent\n",
1337 				 logical);
1338 			return -ENOENT;
1339 		}
1340 		if (path->slots[0] == nritems)
1341 			path->slots[0]--;
1342 
1343 		btrfs_item_key_to_cpu(path->nodes[0], found_key,
1344 				      path->slots[0]);
1345 		if (found_key->type == BTRFS_EXTENT_ITEM_KEY ||
1346 		    found_key->type == BTRFS_METADATA_ITEM_KEY)
1347 			break;
1348 	}
1349 
1350 	if (found_key->type == BTRFS_METADATA_ITEM_KEY)
1351 		size = fs_info->extent_root->leafsize;
1352 	else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
1353 		size = found_key->offset;
1354 
1355 	if (found_key->objectid > logical ||
1356 	    found_key->objectid + size <= logical) {
1357 		pr_debug("logical %llu is not within any extent\n", logical);
1358 		return -ENOENT;
1359 	}
1360 
1361 	eb = path->nodes[0];
1362 	item_size = btrfs_item_size_nr(eb, path->slots[0]);
1363 	BUG_ON(item_size < sizeof(*ei));
1364 
1365 	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1366 	flags = btrfs_extent_flags(eb, ei);
1367 
1368 	pr_debug("logical %llu is at position %llu within the extent (%llu "
1369 		 "EXTENT_ITEM %llu) flags %#llx size %u\n",
1370 		 logical, logical - found_key->objectid, found_key->objectid,
1371 		 found_key->offset, flags, item_size);
1372 
1373 	WARN_ON(!flags_ret);
1374 	if (flags_ret) {
1375 		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1376 			*flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
1377 		else if (flags & BTRFS_EXTENT_FLAG_DATA)
1378 			*flags_ret = BTRFS_EXTENT_FLAG_DATA;
1379 		else
1380 			BUG_ON(1);
1381 		return 0;
1382 	}
1383 
1384 	return -EIO;
1385 }
1386 
1387 /*
1388  * helper function to iterate extent inline refs. ptr must point to a 0 value
1389  * for the first call and may be modified. it is used to track state.
1390  * if more refs exist, 0 is returned and the next call to
1391  * __get_extent_inline_ref must pass the modified ptr parameter to get the
1392  * next ref. after the last ref was processed, 1 is returned.
1393  * returns <0 on error
1394  */
1395 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1396 				struct btrfs_extent_item *ei, u32 item_size,
1397 				struct btrfs_extent_inline_ref **out_eiref,
1398 				int *out_type)
1399 {
1400 	unsigned long end;
1401 	u64 flags;
1402 	struct btrfs_tree_block_info *info;
1403 
1404 	if (!*ptr) {
1405 		/* first call */
1406 		flags = btrfs_extent_flags(eb, ei);
1407 		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1408 			info = (struct btrfs_tree_block_info *)(ei + 1);
1409 			*out_eiref =
1410 				(struct btrfs_extent_inline_ref *)(info + 1);
1411 		} else {
1412 			*out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1413 		}
1414 		*ptr = (unsigned long)*out_eiref;
1415 		if ((void *)*ptr >= (void *)ei + item_size)
1416 			return -ENOENT;
1417 	}
1418 
1419 	end = (unsigned long)ei + item_size;
1420 	*out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1421 	*out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1422 
1423 	*ptr += btrfs_extent_inline_ref_size(*out_type);
1424 	WARN_ON(*ptr > end);
1425 	if (*ptr == end)
1426 		return 1; /* last */
1427 
1428 	return 0;
1429 }
1430 
1431 /*
1432  * reads the tree block backref for an extent. tree level and root are returned
1433  * through out_level and out_root. ptr must point to a 0 value for the first
1434  * call and may be modified (see __get_extent_inline_ref comment).
1435  * returns 0 if data was provided, 1 if there was no more data to provide or
1436  * <0 on error.
1437  */
1438 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1439 				struct btrfs_extent_item *ei, u32 item_size,
1440 				u64 *out_root, u8 *out_level)
1441 {
1442 	int ret;
1443 	int type;
1444 	struct btrfs_tree_block_info *info;
1445 	struct btrfs_extent_inline_ref *eiref;
1446 
1447 	if (*ptr == (unsigned long)-1)
1448 		return 1;
1449 
1450 	while (1) {
1451 		ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1452 						&eiref, &type);
1453 		if (ret < 0)
1454 			return ret;
1455 
1456 		if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1457 		    type == BTRFS_SHARED_BLOCK_REF_KEY)
1458 			break;
1459 
1460 		if (ret == 1)
1461 			return 1;
1462 	}
1463 
1464 	/* we can treat both ref types equally here */
1465 	info = (struct btrfs_tree_block_info *)(ei + 1);
1466 	*out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1467 	*out_level = btrfs_tree_block_level(eb, info);
1468 
1469 	if (ret == 1)
1470 		*ptr = (unsigned long)-1;
1471 
1472 	return 0;
1473 }
1474 
1475 static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
1476 				u64 root, u64 extent_item_objectid,
1477 				iterate_extent_inodes_t *iterate, void *ctx)
1478 {
1479 	struct extent_inode_elem *eie;
1480 	int ret = 0;
1481 
1482 	for (eie = inode_list; eie; eie = eie->next) {
1483 		pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1484 			 "root %llu\n", extent_item_objectid,
1485 			 eie->inum, eie->offset, root);
1486 		ret = iterate(eie->inum, eie->offset, root, ctx);
1487 		if (ret) {
1488 			pr_debug("stopping iteration for %llu due to ret=%d\n",
1489 				 extent_item_objectid, ret);
1490 			break;
1491 		}
1492 	}
1493 
1494 	return ret;
1495 }
1496 
1497 /*
1498  * calls iterate() for every inode that references the extent identified by
1499  * the given parameters.
1500  * when the iterator function returns a non-zero value, iteration stops.
1501  */
1502 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1503 				u64 extent_item_objectid, u64 extent_item_pos,
1504 				int search_commit_root,
1505 				iterate_extent_inodes_t *iterate, void *ctx)
1506 {
1507 	int ret;
1508 	struct btrfs_trans_handle *trans = NULL;
1509 	struct ulist *refs = NULL;
1510 	struct ulist *roots = NULL;
1511 	struct ulist_node *ref_node = NULL;
1512 	struct ulist_node *root_node = NULL;
1513 	struct seq_list tree_mod_seq_elem = {};
1514 	struct ulist_iterator ref_uiter;
1515 	struct ulist_iterator root_uiter;
1516 
1517 	pr_debug("resolving all inodes for extent %llu\n",
1518 			extent_item_objectid);
1519 
1520 	if (!search_commit_root) {
1521 		trans = btrfs_join_transaction(fs_info->extent_root);
1522 		if (IS_ERR(trans))
1523 			return PTR_ERR(trans);
1524 		btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1525 	}
1526 
1527 	ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1528 				   tree_mod_seq_elem.seq, &refs,
1529 				   &extent_item_pos);
1530 	if (ret)
1531 		goto out;
1532 
1533 	ULIST_ITER_INIT(&ref_uiter);
1534 	while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1535 		ret = btrfs_find_all_roots(trans, fs_info, ref_node->val,
1536 					   tree_mod_seq_elem.seq, &roots);
1537 		if (ret)
1538 			break;
1539 		ULIST_ITER_INIT(&root_uiter);
1540 		while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1541 			pr_debug("root %llu references leaf %llu, data list "
1542 				 "%#llx\n", root_node->val, ref_node->val,
1543 				 ref_node->aux);
1544 			ret = iterate_leaf_refs((struct extent_inode_elem *)
1545 						(uintptr_t)ref_node->aux,
1546 						root_node->val,
1547 						extent_item_objectid,
1548 						iterate, ctx);
1549 		}
1550 		ulist_free(roots);
1551 	}
1552 
1553 	free_leaf_list(refs);
1554 out:
1555 	if (!search_commit_root) {
1556 		btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1557 		btrfs_end_transaction(trans, fs_info->extent_root);
1558 	}
1559 
1560 	return ret;
1561 }
1562 
1563 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1564 				struct btrfs_path *path,
1565 				iterate_extent_inodes_t *iterate, void *ctx)
1566 {
1567 	int ret;
1568 	u64 extent_item_pos;
1569 	u64 flags = 0;
1570 	struct btrfs_key found_key;
1571 	int search_commit_root = path->search_commit_root;
1572 
1573 	ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
1574 	btrfs_release_path(path);
1575 	if (ret < 0)
1576 		return ret;
1577 	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1578 		return -EINVAL;
1579 
1580 	extent_item_pos = logical - found_key.objectid;
1581 	ret = iterate_extent_inodes(fs_info, found_key.objectid,
1582 					extent_item_pos, search_commit_root,
1583 					iterate, ctx);
1584 
1585 	return ret;
1586 }
1587 
1588 typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
1589 			      struct extent_buffer *eb, void *ctx);
1590 
1591 static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
1592 			      struct btrfs_path *path,
1593 			      iterate_irefs_t *iterate, void *ctx)
1594 {
1595 	int ret = 0;
1596 	int slot;
1597 	u32 cur;
1598 	u32 len;
1599 	u32 name_len;
1600 	u64 parent = 0;
1601 	int found = 0;
1602 	struct extent_buffer *eb;
1603 	struct btrfs_item *item;
1604 	struct btrfs_inode_ref *iref;
1605 	struct btrfs_key found_key;
1606 
1607 	while (!ret) {
1608 		ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1609 				     &found_key);
1610 		if (ret < 0)
1611 			break;
1612 		if (ret) {
1613 			ret = found ? 0 : -ENOENT;
1614 			break;
1615 		}
1616 		++found;
1617 
1618 		parent = found_key.offset;
1619 		slot = path->slots[0];
1620 		eb = btrfs_clone_extent_buffer(path->nodes[0]);
1621 		if (!eb) {
1622 			ret = -ENOMEM;
1623 			break;
1624 		}
1625 		extent_buffer_get(eb);
1626 		btrfs_tree_read_lock(eb);
1627 		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1628 		btrfs_release_path(path);
1629 
1630 		item = btrfs_item_nr(slot);
1631 		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1632 
1633 		for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1634 			name_len = btrfs_inode_ref_name_len(eb, iref);
1635 			/* path must be released before calling iterate()! */
1636 			pr_debug("following ref at offset %u for inode %llu in "
1637 				 "tree %llu\n", cur, found_key.objectid,
1638 				 fs_root->objectid);
1639 			ret = iterate(parent, name_len,
1640 				      (unsigned long)(iref + 1), eb, ctx);
1641 			if (ret)
1642 				break;
1643 			len = sizeof(*iref) + name_len;
1644 			iref = (struct btrfs_inode_ref *)((char *)iref + len);
1645 		}
1646 		btrfs_tree_read_unlock_blocking(eb);
1647 		free_extent_buffer(eb);
1648 	}
1649 
1650 	btrfs_release_path(path);
1651 
1652 	return ret;
1653 }
1654 
1655 static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
1656 				 struct btrfs_path *path,
1657 				 iterate_irefs_t *iterate, void *ctx)
1658 {
1659 	int ret;
1660 	int slot;
1661 	u64 offset = 0;
1662 	u64 parent;
1663 	int found = 0;
1664 	struct extent_buffer *eb;
1665 	struct btrfs_inode_extref *extref;
1666 	struct extent_buffer *leaf;
1667 	u32 item_size;
1668 	u32 cur_offset;
1669 	unsigned long ptr;
1670 
1671 	while (1) {
1672 		ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
1673 					    &offset);
1674 		if (ret < 0)
1675 			break;
1676 		if (ret) {
1677 			ret = found ? 0 : -ENOENT;
1678 			break;
1679 		}
1680 		++found;
1681 
1682 		slot = path->slots[0];
1683 		eb = btrfs_clone_extent_buffer(path->nodes[0]);
1684 		if (!eb) {
1685 			ret = -ENOMEM;
1686 			break;
1687 		}
1688 		extent_buffer_get(eb);
1689 
1690 		btrfs_tree_read_lock(eb);
1691 		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1692 		btrfs_release_path(path);
1693 
1694 		leaf = path->nodes[0];
1695 		item_size = btrfs_item_size_nr(leaf, slot);
1696 		ptr = btrfs_item_ptr_offset(leaf, slot);
1697 		cur_offset = 0;
1698 
1699 		while (cur_offset < item_size) {
1700 			u32 name_len;
1701 
1702 			extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
1703 			parent = btrfs_inode_extref_parent(eb, extref);
1704 			name_len = btrfs_inode_extref_name_len(eb, extref);
1705 			ret = iterate(parent, name_len,
1706 				      (unsigned long)&extref->name, eb, ctx);
1707 			if (ret)
1708 				break;
1709 
1710 			cur_offset += btrfs_inode_extref_name_len(leaf, extref);
1711 			cur_offset += sizeof(*extref);
1712 		}
1713 		btrfs_tree_read_unlock_blocking(eb);
1714 		free_extent_buffer(eb);
1715 
1716 		offset++;
1717 	}
1718 
1719 	btrfs_release_path(path);
1720 
1721 	return ret;
1722 }
1723 
1724 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1725 			 struct btrfs_path *path, iterate_irefs_t *iterate,
1726 			 void *ctx)
1727 {
1728 	int ret;
1729 	int found_refs = 0;
1730 
1731 	ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
1732 	if (!ret)
1733 		++found_refs;
1734 	else if (ret != -ENOENT)
1735 		return ret;
1736 
1737 	ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
1738 	if (ret == -ENOENT && found_refs)
1739 		return 0;
1740 
1741 	return ret;
1742 }
1743 
1744 /*
1745  * returns 0 if the path could be dumped (probably truncated)
1746  * returns <0 in case of an error
1747  */
1748 static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
1749 			 struct extent_buffer *eb, void *ctx)
1750 {
1751 	struct inode_fs_paths *ipath = ctx;
1752 	char *fspath;
1753 	char *fspath_min;
1754 	int i = ipath->fspath->elem_cnt;
1755 	const int s_ptr = sizeof(char *);
1756 	u32 bytes_left;
1757 
1758 	bytes_left = ipath->fspath->bytes_left > s_ptr ?
1759 					ipath->fspath->bytes_left - s_ptr : 0;
1760 
1761 	fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1762 	fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
1763 				   name_off, eb, inum, fspath_min, bytes_left);
1764 	if (IS_ERR(fspath))
1765 		return PTR_ERR(fspath);
1766 
1767 	if (fspath > fspath_min) {
1768 		ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1769 		++ipath->fspath->elem_cnt;
1770 		ipath->fspath->bytes_left = fspath - fspath_min;
1771 	} else {
1772 		++ipath->fspath->elem_missed;
1773 		ipath->fspath->bytes_missing += fspath_min - fspath;
1774 		ipath->fspath->bytes_left = 0;
1775 	}
1776 
1777 	return 0;
1778 }
1779 
1780 /*
1781  * this dumps all file system paths to the inode into the ipath struct, provided
1782  * is has been created large enough. each path is zero-terminated and accessed
1783  * from ipath->fspath->val[i].
1784  * when it returns, there are ipath->fspath->elem_cnt number of paths available
1785  * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1786  * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1787  * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1788  * have been needed to return all paths.
1789  */
1790 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1791 {
1792 	return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1793 			     inode_to_path, ipath);
1794 }
1795 
1796 struct btrfs_data_container *init_data_container(u32 total_bytes)
1797 {
1798 	struct btrfs_data_container *data;
1799 	size_t alloc_bytes;
1800 
1801 	alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1802 	data = vmalloc(alloc_bytes);
1803 	if (!data)
1804 		return ERR_PTR(-ENOMEM);
1805 
1806 	if (total_bytes >= sizeof(*data)) {
1807 		data->bytes_left = total_bytes - sizeof(*data);
1808 		data->bytes_missing = 0;
1809 	} else {
1810 		data->bytes_missing = sizeof(*data) - total_bytes;
1811 		data->bytes_left = 0;
1812 	}
1813 
1814 	data->elem_cnt = 0;
1815 	data->elem_missed = 0;
1816 
1817 	return data;
1818 }
1819 
1820 /*
1821  * allocates space to return multiple file system paths for an inode.
1822  * total_bytes to allocate are passed, note that space usable for actual path
1823  * information will be total_bytes - sizeof(struct inode_fs_paths).
1824  * the returned pointer must be freed with free_ipath() in the end.
1825  */
1826 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1827 					struct btrfs_path *path)
1828 {
1829 	struct inode_fs_paths *ifp;
1830 	struct btrfs_data_container *fspath;
1831 
1832 	fspath = init_data_container(total_bytes);
1833 	if (IS_ERR(fspath))
1834 		return (void *)fspath;
1835 
1836 	ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1837 	if (!ifp) {
1838 		kfree(fspath);
1839 		return ERR_PTR(-ENOMEM);
1840 	}
1841 
1842 	ifp->btrfs_path = path;
1843 	ifp->fspath = fspath;
1844 	ifp->fs_root = fs_root;
1845 
1846 	return ifp;
1847 }
1848 
1849 void free_ipath(struct inode_fs_paths *ipath)
1850 {
1851 	if (!ipath)
1852 		return;
1853 	vfree(ipath->fspath);
1854 	kfree(ipath);
1855 }
1856