backref.c 59.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2011 STRATO.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

19
#include <linux/mm.h>
20
#include <linux/rbtree.h>
21
#include <trace/events/btrfs.h>
22 23 24
#include "ctree.h"
#include "disk-io.h"
#include "backref.h"
25 26 27
#include "ulist.h"
#include "transaction.h"
#include "delayed-ref.h"
28
#include "locking.h"
29

30 31 32
/* Just an arbitrary number so we can be sure this happened */
#define BACKREF_FOUND_SHARED 6

33 34 35 36 37 38
struct extent_inode_elem {
	u64 inum;
	u64 offset;
	struct extent_inode_elem *next;
};

39 40 41 42
static int check_extent_in_eb(const struct btrfs_key *key,
			      const struct extent_buffer *eb,
			      const struct btrfs_file_extent_item *fi,
			      u64 extent_item_pos,
43 44
			      struct extent_inode_elem **eie,
			      bool ignore_offset)
45
{
46
	u64 offset = 0;
47 48
	struct extent_inode_elem *e;

49 50
	if (!ignore_offset &&
	    !btrfs_file_extent_compression(eb, fi) &&
51 52 53 54
	    !btrfs_file_extent_encryption(eb, fi) &&
	    !btrfs_file_extent_other_encoding(eb, fi)) {
		u64 data_offset;
		u64 data_len;
55

56 57 58 59 60 61 62 63
		data_offset = btrfs_file_extent_offset(eb, fi);
		data_len = btrfs_file_extent_num_bytes(eb, fi);

		if (extent_item_pos < data_offset ||
		    extent_item_pos >= data_offset + data_len)
			return 1;
		offset = extent_item_pos - data_offset;
	}
64 65 66 67 68 69 70

	e = kmalloc(sizeof(*e), GFP_NOFS);
	if (!e)
		return -ENOMEM;

	e->next = *eie;
	e->inum = key->objectid;
71
	e->offset = key->offset + offset;
72 73 74 75 76
	*eie = e;

	return 0;
}

77 78 79 80 81 82 83 84 85 86
static void free_inode_elem_list(struct extent_inode_elem *eie)
{
	struct extent_inode_elem *eie_next;

	for (; eie; eie = eie_next) {
		eie_next = eie->next;
		kfree(eie);
	}
}

87 88
static int find_extent_in_eb(const struct extent_buffer *eb,
			     u64 wanted_disk_byte, u64 extent_item_pos,
89 90
			     struct extent_inode_elem **eie,
			     bool ignore_offset)
91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
{
	u64 disk_byte;
	struct btrfs_key key;
	struct btrfs_file_extent_item *fi;
	int slot;
	int nritems;
	int extent_type;
	int ret;

	/*
	 * from the shared data ref, we only have the leaf but we need
	 * the key. thus, we must look into all items and see that we
	 * find one (some) with a reference to our extent item.
	 */
	nritems = btrfs_header_nritems(eb);
	for (slot = 0; slot < nritems; ++slot) {
		btrfs_item_key_to_cpu(eb, &key, slot);
		if (key.type != BTRFS_EXTENT_DATA_KEY)
			continue;
		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
		extent_type = btrfs_file_extent_type(eb, fi);
		if (extent_type == BTRFS_FILE_EXTENT_INLINE)
			continue;
		/* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
		if (disk_byte != wanted_disk_byte)
			continue;

119
		ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie, ignore_offset);
120 121 122 123 124 125 126
		if (ret < 0)
			return ret;
	}

	return 0;
}

127 128
struct preftree {
	struct rb_root root;
129
	unsigned int count;
130 131
};

132
#define PREFTREE_INIT	{ .root = RB_ROOT, .count = 0 }
133 134 135 136 137 138 139

struct preftrees {
	struct preftree direct;    /* BTRFS_SHARED_[DATA|BLOCK]_REF_KEY */
	struct preftree indirect;  /* BTRFS_[TREE_BLOCK|EXTENT_DATA]_REF_KEY */
	struct preftree indirect_missing_keys;
};

140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158
/*
 * Checks for a shared extent during backref search.
 *
 * The share_count tracks prelim_refs (direct and indirect) having a
 * ref->count >0:
 *  - incremented when a ref->count transitions to >0
 *  - decremented when a ref->count transitions to <1
 */
struct share_check {
	u64 root_objectid;
	u64 inum;
	int share_count;
};

static inline int extent_is_shared(struct share_check *sc)
{
	return (sc && sc->share_count > 1) ? BACKREF_FOUND_SHARED : 0;
}

159 160 161 162 163
static struct kmem_cache *btrfs_prelim_ref_cache;

int __init btrfs_prelim_ref_init(void)
{
	btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
164
					sizeof(struct prelim_ref),
165
					0,
166
					SLAB_MEM_SPREAD,
167 168 169 170 171 172 173 174
					NULL);
	if (!btrfs_prelim_ref_cache)
		return -ENOMEM;
	return 0;
}

void btrfs_prelim_ref_exit(void)
{
175
	kmem_cache_destroy(btrfs_prelim_ref_cache);
176 177
}

178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218
static void free_pref(struct prelim_ref *ref)
{
	kmem_cache_free(btrfs_prelim_ref_cache, ref);
}

/*
 * Return 0 when both refs are for the same block (and can be merged).
 * A -1 return indicates ref1 is a 'lower' block than ref2, while 1
 * indicates a 'higher' block.
 */
static int prelim_ref_compare(struct prelim_ref *ref1,
			      struct prelim_ref *ref2)
{
	if (ref1->level < ref2->level)
		return -1;
	if (ref1->level > ref2->level)
		return 1;
	if (ref1->root_id < ref2->root_id)
		return -1;
	if (ref1->root_id > ref2->root_id)
		return 1;
	if (ref1->key_for_search.type < ref2->key_for_search.type)
		return -1;
	if (ref1->key_for_search.type > ref2->key_for_search.type)
		return 1;
	if (ref1->key_for_search.objectid < ref2->key_for_search.objectid)
		return -1;
	if (ref1->key_for_search.objectid > ref2->key_for_search.objectid)
		return 1;
	if (ref1->key_for_search.offset < ref2->key_for_search.offset)
		return -1;
	if (ref1->key_for_search.offset > ref2->key_for_search.offset)
		return 1;
	if (ref1->parent < ref2->parent)
		return -1;
	if (ref1->parent > ref2->parent)
		return 1;

	return 0;
}

219 220
static void update_share_count(struct share_check *sc, int oldcount,
			       int newcount)
221 222 223 224 225 226 227 228 229 230
{
	if ((!sc) || (oldcount == 0 && newcount < 1))
		return;

	if (oldcount > 0 && newcount < 1)
		sc->share_count--;
	else if (oldcount < 1 && newcount > 0)
		sc->share_count++;
}

231 232 233
/*
 * Add @newref to the @root rbtree, merging identical refs.
 *
234
 * Callers should assume that newref has been freed after calling.
235
 */
236 237
static void prelim_ref_insert(const struct btrfs_fs_info *fs_info,
			      struct preftree *preftree,
238 239
			      struct prelim_ref *newref,
			      struct share_check *sc)
240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268
{
	struct rb_root *root;
	struct rb_node **p;
	struct rb_node *parent = NULL;
	struct prelim_ref *ref;
	int result;

	root = &preftree->root;
	p = &root->rb_node;

	while (*p) {
		parent = *p;
		ref = rb_entry(parent, struct prelim_ref, rbnode);
		result = prelim_ref_compare(ref, newref);
		if (result < 0) {
			p = &(*p)->rb_left;
		} else if (result > 0) {
			p = &(*p)->rb_right;
		} else {
			/* Identical refs, merge them and free @newref */
			struct extent_inode_elem *eie = ref->inode_list;

			while (eie && eie->next)
				eie = eie->next;

			if (!eie)
				ref->inode_list = newref->inode_list;
			else
				eie->next = newref->inode_list;
269 270
			trace_btrfs_prelim_ref_merge(fs_info, ref, newref,
						     preftree->count);
271 272 273 274 275 276 277
			/*
			 * A delayed ref can have newref->count < 0.
			 * The ref->count is updated to follow any
			 * BTRFS_[ADD|DROP]_DELAYED_REF actions.
			 */
			update_share_count(sc, ref->count,
					   ref->count + newref->count);
278 279 280 281 282 283
			ref->count += newref->count;
			free_pref(newref);
			return;
		}
	}

284
	update_share_count(sc, 0, newref->count);
285
	preftree->count++;
286
	trace_btrfs_prelim_ref_insert(fs_info, newref, NULL, preftree->count);
287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
	rb_link_node(&newref->rbnode, parent, p);
	rb_insert_color(&newref->rbnode, root);
}

/*
 * Release the entire tree.  We don't care about internal consistency so
 * just free everything and then reset the tree root.
 */
static void prelim_release(struct preftree *preftree)
{
	struct prelim_ref *ref, *next_ref;

	rbtree_postorder_for_each_entry_safe(ref, next_ref, &preftree->root,
					     rbnode)
		free_pref(ref);

	preftree->root = RB_ROOT;
304
	preftree->count = 0;
305 306
}

307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
/*
 * the rules for all callers of this function are:
 * - obtaining the parent is the goal
 * - if you add a key, you must know that it is a correct key
 * - if you cannot add the parent or a correct key, then we will look into the
 *   block later to set a correct key
 *
 * delayed refs
 * ============
 *        backref type | shared | indirect | shared | indirect
 * information         |   tree |     tree |   data |     data
 * --------------------+--------+----------+--------+----------
 *      parent logical |    y   |     -    |    -   |     -
 *      key to resolve |    -   |     y    |    y   |     y
 *  tree block logical |    -   |     -    |    -   |     -
 *  root for resolving |    y   |     y    |    y   |     y
 *
 * - column 1:       we've the parent -> done
 * - column 2, 3, 4: we use the key to find the parent
 *
 * on disk refs (inline or keyed)
 * ==============================
 *        backref type | shared | indirect | shared | indirect
 * information         |   tree |     tree |   data |     data
 * --------------------+--------+----------+--------+----------
 *      parent logical |    y   |     -    |    y   |     -
 *      key to resolve |    -   |     -    |    -   |     y
 *  tree block logical |    y   |     y    |    y   |     y
 *  root for resolving |    -   |     y    |    y   |     y
 *
 * - column 1, 3: we've the parent -> done
 * - column 2:    we take the first key from the block to find the parent
339
 *                (see add_missing_keys)
340 341 342 343 344
 * - column 4:    we use the key to find the parent
 *
 * additional information that's available but not required to find the parent
 * block might help in merging entries to gain some speed.
 */
345 346
static int add_prelim_ref(const struct btrfs_fs_info *fs_info,
			  struct preftree *preftree, u64 root_id,
347
			  const struct btrfs_key *key, int level, u64 parent,
348 349
			  u64 wanted_disk_byte, int count,
			  struct share_check *sc, gfp_t gfp_mask)
350
{
351
	struct prelim_ref *ref;
352

353 354 355
	if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
		return 0;

356
	ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
357 358 359 360
	if (!ref)
		return -ENOMEM;

	ref->root_id = root_id;
361
	if (key) {
362
		ref->key_for_search = *key;
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
		/*
		 * We can often find data backrefs with an offset that is too
		 * large (>= LLONG_MAX, maximum allowed file offset) due to
		 * underflows when subtracting a file's offset with the data
		 * offset of its corresponding extent data item. This can
		 * happen for example in the clone ioctl.
		 * So if we detect such case we set the search key's offset to
		 * zero to make sure we will find the matching file extent item
		 * at add_all_parents(), otherwise we will miss it because the
		 * offset taken form the backref is much larger then the offset
		 * of the file extent item. This can make us scan a very large
		 * number of file extent items, but at least it will not make
		 * us miss any.
		 * This is an ugly workaround for a behaviour that should have
		 * never existed, but it does and a fix for the clone ioctl
		 * would touch a lot of places, cause backwards incompatibility
		 * and would not fix the problem for extents cloned with older
		 * kernels.
		 */
		if (ref->key_for_search.type == BTRFS_EXTENT_DATA_KEY &&
		    ref->key_for_search.offset >= LLONG_MAX)
			ref->key_for_search.offset = 0;
	} else {
386
		memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
387
	}
388

389
	ref->inode_list = NULL;
390 391 392 393
	ref->level = level;
	ref->count = count;
	ref->parent = parent;
	ref->wanted_disk_byte = wanted_disk_byte;
394 395
	prelim_ref_insert(fs_info, preftree, ref, sc);
	return extent_is_shared(sc);
396 397
}

398
/* direct refs use root == 0, key == NULL */
399 400
static int add_direct_ref(const struct btrfs_fs_info *fs_info,
			  struct preftrees *preftrees, int level, u64 parent,
401 402
			  u64 wanted_disk_byte, int count,
			  struct share_check *sc, gfp_t gfp_mask)
403
{
404
	return add_prelim_ref(fs_info, &preftrees->direct, 0, NULL, level,
405
			      parent, wanted_disk_byte, count, sc, gfp_mask);
406 407 408
}

/* indirect refs use parent == 0 */
409 410
static int add_indirect_ref(const struct btrfs_fs_info *fs_info,
			    struct preftrees *preftrees, u64 root_id,
411
			    const struct btrfs_key *key, int level,
412 413
			    u64 wanted_disk_byte, int count,
			    struct share_check *sc, gfp_t gfp_mask)
414 415 416 417 418
{
	struct preftree *tree = &preftrees->indirect;

	if (!key)
		tree = &preftrees->indirect_missing_keys;
419
	return add_prelim_ref(fs_info, tree, root_id, key, level, 0,
420
			      wanted_disk_byte, count, sc, gfp_mask);
421 422
}

423
static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
424
			   struct ulist *parents, struct prelim_ref *ref,
425
			   int level, u64 time_seq, const u64 *extent_item_pos,
426
			   u64 total_refs, bool ignore_offset)
427
{
428 429 430 431
	int ret = 0;
	int slot;
	struct extent_buffer *eb;
	struct btrfs_key key;
432
	struct btrfs_key *key_for_search = &ref->key_for_search;
433
	struct btrfs_file_extent_item *fi;
434
	struct extent_inode_elem *eie = NULL, *old = NULL;
435
	u64 disk_byte;
436 437
	u64 wanted_disk_byte = ref->wanted_disk_byte;
	u64 count = 0;
438

439 440 441
	if (level != 0) {
		eb = path->nodes[level];
		ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
442 443
		if (ret < 0)
			return ret;
444
		return 0;
445
	}
446 447

	/*
448 449 450
	 * We normally enter this function with the path already pointing to
	 * the first item to check. But sometimes, we may enter it with
	 * slot==nritems. In that case, go to the next leaf before we continue.
451
	 */
452
	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
453
		if (time_seq == SEQ_LAST)
454 455 456 457
			ret = btrfs_next_leaf(root, path);
		else
			ret = btrfs_next_old_leaf(root, path, time_seq);
	}
458

459
	while (!ret && count < total_refs) {
460
		eb = path->nodes[0];
461 462 463 464 465 466 467 468 469 470 471 472 473
		slot = path->slots[0];

		btrfs_item_key_to_cpu(eb, &key, slot);

		if (key.objectid != key_for_search->objectid ||
		    key.type != BTRFS_EXTENT_DATA_KEY)
			break;

		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);

		if (disk_byte == wanted_disk_byte) {
			eie = NULL;
474
			old = NULL;
475
			count++;
476 477 478
			if (extent_item_pos) {
				ret = check_extent_in_eb(&key, eb, fi,
						*extent_item_pos,
479
						&eie, ignore_offset);
480 481 482
				if (ret < 0)
					break;
			}
483 484
			if (ret > 0)
				goto next;
485 486
			ret = ulist_add_merge_ptr(parents, eb->start,
						  eie, (void **)&old, GFP_NOFS);
487 488 489 490 491 492
			if (ret < 0)
				break;
			if (!ret && extent_item_pos) {
				while (old->next)
					old = old->next;
				old->next = eie;
493
			}
494
			eie = NULL;
495
		}
496
next:
497
		if (time_seq == SEQ_LAST)
498 499 500
			ret = btrfs_next_item(root, path);
		else
			ret = btrfs_next_old_item(root, path, time_seq);
501 502
	}

503 504
	if (ret > 0)
		ret = 0;
505 506
	else if (ret < 0)
		free_inode_elem_list(eie);
507
	return ret;
508 509 510 511 512 513
}

/*
 * resolve an indirect backref in the form (root_id, key, level)
 * to a logical address
 */
514 515 516
static int resolve_indirect_ref(struct btrfs_fs_info *fs_info,
				struct btrfs_path *path, u64 time_seq,
				struct prelim_ref *ref, struct ulist *parents,
517 518
				const u64 *extent_item_pos, u64 total_refs,
				bool ignore_offset)
519 520 521 522 523 524 525
{
	struct btrfs_root *root;
	struct btrfs_key root_key;
	struct extent_buffer *eb;
	int ret = 0;
	int root_level;
	int level = ref->level;
526
	int index;
527 528 529 530

	root_key.objectid = ref->root_id;
	root_key.type = BTRFS_ROOT_ITEM_KEY;
	root_key.offset = (u64)-1;
531 532 533

	index = srcu_read_lock(&fs_info->subvol_srcu);

534
	root = btrfs_get_fs_root(fs_info, &root_key, false);
535
	if (IS_ERR(root)) {
536
		srcu_read_unlock(&fs_info->subvol_srcu, index);
537 538 539 540
		ret = PTR_ERR(root);
		goto out;
	}

541
	if (btrfs_is_testing(fs_info)) {
J
Josef Bacik 已提交
542 543 544 545 546
		srcu_read_unlock(&fs_info->subvol_srcu, index);
		ret = -ENOENT;
		goto out;
	}

547 548
	if (path->search_commit_root)
		root_level = btrfs_header_level(root->commit_root);
549
	else if (time_seq == SEQ_LAST)
550
		root_level = btrfs_header_level(root->node);
551 552
	else
		root_level = btrfs_old_root_level(root, time_seq);
553

554 555
	if (root_level + 1 == level) {
		srcu_read_unlock(&fs_info->subvol_srcu, index);
556
		goto out;
557
	}
558 559

	path->lowest_level = level;
560
	if (time_seq == SEQ_LAST)
561 562 563 564 565
		ret = btrfs_search_slot(NULL, root, &ref->key_for_search, path,
					0, 0);
	else
		ret = btrfs_search_old_slot(root, &ref->key_for_search, path,
					    time_seq);
566 567 568 569

	/* root node has been locked, we can release @subvol_srcu safely here */
	srcu_read_unlock(&fs_info->subvol_srcu, index);

570 571
	btrfs_debug(fs_info,
		"search slot in root %llu (level %d, ref count %d) returned %d for key (%llu %u %llu)",
572 573 574
		 ref->root_id, level, ref->count, ret,
		 ref->key_for_search.objectid, ref->key_for_search.type,
		 ref->key_for_search.offset);
575 576 577 578
	if (ret < 0)
		goto out;

	eb = path->nodes[level];
579
	while (!eb) {
580
		if (WARN_ON(!level)) {
581 582 583 584 585
			ret = 1;
			goto out;
		}
		level--;
		eb = path->nodes[level];
586 587
	}

588
	ret = add_all_parents(root, path, parents, ref, level, time_seq,
589
			      extent_item_pos, total_refs, ignore_offset);
590
out:
591 592
	path->lowest_level = 0;
	btrfs_release_path(path);
593 594 595
	return ret;
}

596 597 598 599 600 601 602 603
static struct extent_inode_elem *
unode_aux_to_inode_list(struct ulist_node *node)
{
	if (!node)
		return NULL;
	return (struct extent_inode_elem *)(uintptr_t)node->aux;
}

604
/*
605 606 607 608 609 610 611 612 613 614 615 616 617 618
 * We maintain three seperate rbtrees: one for direct refs, one for
 * indirect refs which have a key, and one for indirect refs which do not
 * have a key. Each tree does merge on insertion.
 *
 * Once all of the references are located, we iterate over the tree of
 * indirect refs with missing keys. An appropriate key is located and
 * the ref is moved onto the tree for indirect refs. After all missing
 * keys are thus located, we iterate over the indirect ref tree, resolve
 * each reference, and then insert the resolved reference onto the
 * direct tree (merging there too).
 *
 * New backrefs (i.e., for parent nodes) are added to the appropriate
 * rbtree as they are encountered. The new backrefs are subsequently
 * resolved as above.
619
 */
620 621
static int resolve_indirect_refs(struct btrfs_fs_info *fs_info,
				 struct btrfs_path *path, u64 time_seq,
622
				 struct preftrees *preftrees,
623
				 const u64 *extent_item_pos, u64 total_refs,
624
				 struct share_check *sc, bool ignore_offset)
625 626 627 628 629
{
	int err;
	int ret = 0;
	struct ulist *parents;
	struct ulist_node *node;
J
Jan Schmidt 已提交
630
	struct ulist_iterator uiter;
631
	struct rb_node *rnode;
632 633 634 635 636 637

	parents = ulist_alloc(GFP_NOFS);
	if (!parents)
		return -ENOMEM;

	/*
638 639 640 641
	 * We could trade memory usage for performance here by iterating
	 * the tree, allocating new refs for each insertion, and then
	 * freeing the entire indirect tree when we're done.  In some test
	 * cases, the tree can grow quite large (~200k objects).
642
	 */
643 644 645 646 647 648 649 650 651 652 653
	while ((rnode = rb_first(&preftrees->indirect.root))) {
		struct prelim_ref *ref;

		ref = rb_entry(rnode, struct prelim_ref, rbnode);
		if (WARN(ref->parent,
			 "BUG: direct ref found in indirect tree")) {
			ret = -EINVAL;
			goto out;
		}

		rb_erase(&ref->rbnode, &preftrees->indirect.root);
654
		preftrees->indirect.count--;
655 656 657

		if (ref->count == 0) {
			free_pref(ref);
658
			continue;
659 660
		}

661 662
		if (sc && sc->root_objectid &&
		    ref->root_id != sc->root_objectid) {
663
			free_pref(ref);
664 665 666
			ret = BACKREF_FOUND_SHARED;
			goto out;
		}
667 668
		err = resolve_indirect_ref(fs_info, path, time_seq, ref,
					   parents, extent_item_pos,
669
					   total_refs, ignore_offset);
670 671 672 673 674
		/*
		 * we can only tolerate ENOENT,otherwise,we should catch error
		 * and return directly.
		 */
		if (err == -ENOENT) {
675 676
			prelim_ref_insert(fs_info, &preftrees->direct, ref,
					  NULL);
677
			continue;
678
		} else if (err) {
679
			free_pref(ref);
680 681 682
			ret = err;
			goto out;
		}
683 684

		/* we put the first parent into the ref at hand */
J
Jan Schmidt 已提交
685 686
		ULIST_ITER_INIT(&uiter);
		node = ulist_next(parents, &uiter);
687
		ref->parent = node ? node->val : 0;
688
		ref->inode_list = unode_aux_to_inode_list(node);
689

690
		/* Add a prelim_ref(s) for any other parent(s). */
J
Jan Schmidt 已提交
691
		while ((node = ulist_next(parents, &uiter))) {
692 693
			struct prelim_ref *new_ref;

694 695
			new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
						   GFP_NOFS);
696
			if (!new_ref) {
697
				free_pref(ref);
698
				ret = -ENOMEM;
699
				goto out;
700 701 702
			}
			memcpy(new_ref, ref, sizeof(*ref));
			new_ref->parent = node->val;
703
			new_ref->inode_list = unode_aux_to_inode_list(node);
704 705
			prelim_ref_insert(fs_info, &preftrees->direct,
					  new_ref, NULL);
706
		}
707

708 709 710 711 712
		/*
		 * Now it's a direct ref, put it in the the direct tree. We must
		 * do this last because the ref could be merged/freed here.
		 */
		prelim_ref_insert(fs_info, &preftrees->direct, ref, NULL);
713

714
		ulist_reinit(parents);
715
		cond_resched();
716
	}
717
out:
718 719 720 721
	ulist_free(parents);
	return ret;
}

722 723 724
/*
 * read tree blocks and add keys where required.
 */
725
static int add_missing_keys(struct btrfs_fs_info *fs_info,
726
			    struct preftrees *preftrees)
727
{
728
	struct prelim_ref *ref;
729
	struct extent_buffer *eb;
730 731
	struct preftree *tree = &preftrees->indirect_missing_keys;
	struct rb_node *node;
732

733 734 735 736 737 738
	while ((node = rb_first(&tree->root))) {
		ref = rb_entry(node, struct prelim_ref, rbnode);
		rb_erase(node, &tree->root);

		BUG_ON(ref->parent);	/* should not be a direct ref */
		BUG_ON(ref->key_for_search.type);
739
		BUG_ON(!ref->wanted_disk_byte);
740

741
		eb = read_tree_block(fs_info, ref->wanted_disk_byte, 0);
742
		if (IS_ERR(eb)) {
743
			free_pref(ref);
744 745
			return PTR_ERR(eb);
		} else if (!extent_buffer_uptodate(eb)) {
746
			free_pref(ref);
747 748 749
			free_extent_buffer(eb);
			return -EIO;
		}
750 751 752 753 754 755 756
		btrfs_tree_read_lock(eb);
		if (btrfs_header_level(eb) == 0)
			btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
		else
			btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
		btrfs_tree_read_unlock(eb);
		free_extent_buffer(eb);
757
		prelim_ref_insert(fs_info, &preftrees->indirect, ref, NULL);
758
		cond_resched();
759 760 761 762
	}
	return 0;
}

763 764 765 766
/*
 * add all currently queued delayed refs from this head whose seq nr is
 * smaller or equal that seq to the list
 */
767 768
static int add_delayed_refs(const struct btrfs_fs_info *fs_info,
			    struct btrfs_delayed_ref_head *head, u64 seq,
769
			    struct preftrees *preftrees, u64 *total_refs,
770
			    struct share_check *sc)
771
{
772
	struct btrfs_delayed_ref_node *node;
773
	struct btrfs_delayed_extent_op *extent_op = head->extent_op;
774
	struct btrfs_key key;
775 776
	struct btrfs_key tmp_op_key;
	struct btrfs_key *op_key = NULL;
777
	struct rb_node *n;
778
	int count;
779
	int ret = 0;
780

781 782 783 784
	if (extent_op && extent_op->update_key) {
		btrfs_disk_key_to_cpu(&tmp_op_key, &extent_op->key);
		op_key = &tmp_op_key;
	}
785

786
	spin_lock(&head->lock);
787 788 789
	for (n = rb_first(&head->ref_tree); n; n = rb_next(n)) {
		node = rb_entry(n, struct btrfs_delayed_ref_node,
				ref_node);
790 791 792 793 794 795 796 797 798
		if (node->seq > seq)
			continue;

		switch (node->action) {
		case BTRFS_ADD_DELAYED_EXTENT:
		case BTRFS_UPDATE_DELAYED_HEAD:
			WARN_ON(1);
			continue;
		case BTRFS_ADD_DELAYED_REF:
799
			count = node->ref_mod;
800 801
			break;
		case BTRFS_DROP_DELAYED_REF:
802
			count = node->ref_mod * -1;
803 804 805 806
			break;
		default:
			BUG_ON(1);
		}
807
		*total_refs += count;
808 809
		switch (node->type) {
		case BTRFS_TREE_BLOCK_REF_KEY: {
810
			/* NORMAL INDIRECT METADATA backref */
811 812 813
			struct btrfs_delayed_tree_ref *ref;

			ref = btrfs_delayed_node_to_tree_ref(node);
814 815
			ret = add_indirect_ref(fs_info, preftrees, ref->root,
					       &tmp_op_key, ref->level + 1,
816 817
					       node->bytenr, count, sc,
					       GFP_ATOMIC);
818 819 820
			break;
		}
		case BTRFS_SHARED_BLOCK_REF_KEY: {
821
			/* SHARED DIRECT METADATA backref */
822 823 824
			struct btrfs_delayed_tree_ref *ref;

			ref = btrfs_delayed_node_to_tree_ref(node);
825

826 827
			ret = add_direct_ref(fs_info, preftrees, ref->level + 1,
					     ref->parent, node->bytenr, count,
828
					     sc, GFP_ATOMIC);
829 830 831
			break;
		}
		case BTRFS_EXTENT_DATA_REF_KEY: {
832
			/* NORMAL INDIRECT DATA backref */
833 834 835 836 837 838
			struct btrfs_delayed_data_ref *ref;
			ref = btrfs_delayed_node_to_data_ref(node);

			key.objectid = ref->objectid;
			key.type = BTRFS_EXTENT_DATA_KEY;
			key.offset = ref->offset;
839 840 841 842 843

			/*
			 * Found a inum that doesn't match our known inum, we
			 * know it's shared.
			 */
844
			if (sc && sc->inum && ref->objectid != sc->inum) {
845
				ret = BACKREF_FOUND_SHARED;
846
				goto out;
847 848
			}

849
			ret = add_indirect_ref(fs_info, preftrees, ref->root,
850 851
					       &key, 0, node->bytenr, count, sc,
					       GFP_ATOMIC);
852 853 854
			break;
		}
		case BTRFS_SHARED_DATA_REF_KEY: {
855
			/* SHARED DIRECT FULL backref */
856 857 858
			struct btrfs_delayed_data_ref *ref;

			ref = btrfs_delayed_node_to_data_ref(node);
859

860 861 862
			ret = add_direct_ref(fs_info, preftrees, 0, ref->parent,
					     node->bytenr, count, sc,
					     GFP_ATOMIC);
863 864 865 866 867
			break;
		}
		default:
			WARN_ON(1);
		}
868 869 870 871 872
		/*
		 * We must ignore BACKREF_FOUND_SHARED until all delayed
		 * refs have been checked.
		 */
		if (ret && (ret != BACKREF_FOUND_SHARED))
873
			break;
874
	}
875 876 877
	if (!ret)
		ret = extent_is_shared(sc);
out:
878 879
	spin_unlock(&head->lock);
	return ret;
880 881 882 883
}

/*
 * add all inline backrefs for bytenr to the list
884 885
 *
 * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
886
 */
887 888
static int add_inline_refs(const struct btrfs_fs_info *fs_info,
			   struct btrfs_path *path, u64 bytenr,
889
			   int *info_level, struct preftrees *preftrees,
890
			   u64 *total_refs, struct share_check *sc)
891
{
892
	int ret = 0;
893 894 895
	int slot;
	struct extent_buffer *leaf;
	struct btrfs_key key;
896
	struct btrfs_key found_key;
897 898 899 900 901 902 903 904 905 906
	unsigned long ptr;
	unsigned long end;
	struct btrfs_extent_item *ei;
	u64 flags;
	u64 item_size;

	/*
	 * enumerate all inline refs
	 */
	leaf = path->nodes[0];
907
	slot = path->slots[0];
908 909 910 911 912 913

	item_size = btrfs_item_size_nr(leaf, slot);
	BUG_ON(item_size < sizeof(*ei));

	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
	flags = btrfs_extent_flags(leaf, ei);
914
	*total_refs += btrfs_extent_refs(leaf, ei);
915
	btrfs_item_key_to_cpu(leaf, &found_key, slot);
916 917 918 919

	ptr = (unsigned long)(ei + 1);
	end = (unsigned long)ei + item_size;

920 921
	if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
	    flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
922 923 924 925 926 927
		struct btrfs_tree_block_info *info;

		info = (struct btrfs_tree_block_info *)ptr;
		*info_level = btrfs_tree_block_level(leaf, info);
		ptr += sizeof(struct btrfs_tree_block_info);
		BUG_ON(ptr > end);
928 929
	} else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
		*info_level = found_key.offset;
930 931 932 933 934 935 936 937 938 939
	} else {
		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
	}

	while (ptr < end) {
		struct btrfs_extent_inline_ref *iref;
		u64 offset;
		int type;

		iref = (struct btrfs_extent_inline_ref *)ptr;
940 941 942 943 944
		type = btrfs_get_extent_inline_ref_type(leaf, iref,
							BTRFS_REF_TYPE_ANY);
		if (type == BTRFS_REF_TYPE_INVALID)
			return -EINVAL;

945 946 947 948
		offset = btrfs_extent_inline_ref_offset(leaf, iref);

		switch (type) {
		case BTRFS_SHARED_BLOCK_REF_KEY:
949 950
			ret = add_direct_ref(fs_info, preftrees,
					     *info_level + 1, offset,
951
					     bytenr, 1, NULL, GFP_NOFS);
952 953 954 955 956 957 958
			break;
		case BTRFS_SHARED_DATA_REF_KEY: {
			struct btrfs_shared_data_ref *sdref;
			int count;

			sdref = (struct btrfs_shared_data_ref *)(iref + 1);
			count = btrfs_shared_data_ref_count(leaf, sdref);
959

960
			ret = add_direct_ref(fs_info, preftrees, 0, offset,
961
					     bytenr, count, sc, GFP_NOFS);
962 963 964
			break;
		}
		case BTRFS_TREE_BLOCK_REF_KEY:
965 966
			ret = add_indirect_ref(fs_info, preftrees, offset,
					       NULL, *info_level + 1,
967
					       bytenr, 1, NULL, GFP_NOFS);
968 969 970 971 972 973 974 975 976 977 978 979
			break;
		case BTRFS_EXTENT_DATA_REF_KEY: {
			struct btrfs_extent_data_ref *dref;
			int count;
			u64 root;

			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
			count = btrfs_extent_data_ref_count(leaf, dref);
			key.objectid = btrfs_extent_data_ref_objectid(leaf,
								      dref);
			key.type = BTRFS_EXTENT_DATA_KEY;
			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
980

981
			if (sc && sc->inum && key.objectid != sc->inum) {
982 983 984 985
				ret = BACKREF_FOUND_SHARED;
				break;
			}

986
			root = btrfs_extent_data_ref_root(leaf, dref);
987

988 989
			ret = add_indirect_ref(fs_info, preftrees, root,
					       &key, 0, bytenr, count,
990
					       sc, GFP_NOFS);
991 992 993 994 995
			break;
		}
		default:
			WARN_ON(1);
		}
996 997
		if (ret)
			return ret;
998 999 1000 1001 1002 1003 1004 1005
		ptr += btrfs_extent_inline_ref_size(type);
	}

	return 0;
}

/*
 * add all non-inline backrefs for bytenr to the list
1006 1007
 *
 * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
1008
 */
1009 1010
static int add_keyed_refs(struct btrfs_fs_info *fs_info,
			  struct btrfs_path *path, u64 bytenr,
1011
			  int info_level, struct preftrees *preftrees,
1012
			  struct share_check *sc)
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
{
	struct btrfs_root *extent_root = fs_info->extent_root;
	int ret;
	int slot;
	struct extent_buffer *leaf;
	struct btrfs_key key;

	while (1) {
		ret = btrfs_next_item(extent_root, path);
		if (ret < 0)
			break;
		if (ret) {
			ret = 0;
			break;
		}

		slot = path->slots[0];
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, slot);

		if (key.objectid != bytenr)
			break;
		if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
			continue;
		if (key.type > BTRFS_SHARED_DATA_REF_KEY)
			break;

		switch (key.type) {
		case BTRFS_SHARED_BLOCK_REF_KEY:
1042
			/* SHARED DIRECT METADATA backref */
1043 1044
			ret = add_direct_ref(fs_info, preftrees,
					     info_level + 1, key.offset,
1045
					     bytenr, 1, NULL, GFP_NOFS);
1046 1047
			break;
		case BTRFS_SHARED_DATA_REF_KEY: {
1048
			/* SHARED DIRECT FULL backref */
1049 1050 1051 1052 1053 1054
			struct btrfs_shared_data_ref *sdref;
			int count;

			sdref = btrfs_item_ptr(leaf, slot,
					      struct btrfs_shared_data_ref);
			count = btrfs_shared_data_ref_count(leaf, sdref);
1055 1056
			ret = add_direct_ref(fs_info, preftrees, 0,
					     key.offset, bytenr, count,
1057
					     sc, GFP_NOFS);
1058 1059 1060
			break;
		}
		case BTRFS_TREE_BLOCK_REF_KEY:
1061
			/* NORMAL INDIRECT METADATA backref */
1062 1063
			ret = add_indirect_ref(fs_info, preftrees, key.offset,
					       NULL, info_level + 1, bytenr,
1064
					       1, NULL, GFP_NOFS);
1065 1066
			break;
		case BTRFS_EXTENT_DATA_REF_KEY: {
1067
			/* NORMAL INDIRECT DATA backref */
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
			struct btrfs_extent_data_ref *dref;
			int count;
			u64 root;

			dref = btrfs_item_ptr(leaf, slot,
					      struct btrfs_extent_data_ref);
			count = btrfs_extent_data_ref_count(leaf, dref);
			key.objectid = btrfs_extent_data_ref_objectid(leaf,
								      dref);
			key.type = BTRFS_EXTENT_DATA_KEY;
			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
1079

1080
			if (sc && sc->inum && key.objectid != sc->inum) {
1081 1082 1083 1084
				ret = BACKREF_FOUND_SHARED;
				break;
			}

1085
			root = btrfs_extent_data_ref_root(leaf, dref);
1086 1087
			ret = add_indirect_ref(fs_info, preftrees, root,
					       &key, 0, bytenr, count,
1088
					       sc, GFP_NOFS);
1089 1090 1091 1092 1093
			break;
		}
		default:
			WARN_ON(1);
		}
1094 1095 1096
		if (ret)
			return ret;

1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
	}

	return ret;
}

/*
 * this adds all existing backrefs (inline backrefs, backrefs and delayed
 * refs) for the given bytenr to the refs list, merges duplicates and resolves
 * indirect refs to their parent bytenr.
 * When roots are found, they're added to the roots list
 *
1108
 * If time_seq is set to SEQ_LAST, it will not search delayed_refs, and behave
1109 1110 1111 1112
 * much like trans == NULL case, the difference only lies in it will not
 * commit root.
 * The special case is for qgroup to search roots in commit_transaction().
 *
1113 1114 1115 1116 1117
 * @sc - if !NULL, then immediately return BACKREF_FOUND_SHARED when a
 * shared extent is detected.
 *
 * Otherwise this returns 0 for success and <0 for an error.
 *
1118 1119 1120 1121
 * If ignore_offset is set to false, only extent refs whose offsets match
 * extent_item_pos are returned.  If true, every extent ref is returned
 * and extent_item_pos is ignored.
 *
1122 1123 1124 1125
 * FIXME some caching might speed things up
 */
static int find_parent_nodes(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info, u64 bytenr,
1126
			     u64 time_seq, struct ulist *refs,
1127
			     struct ulist *roots, const u64 *extent_item_pos,
1128
			     struct share_check *sc, bool ignore_offset)
1129 1130 1131 1132
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct btrfs_delayed_ref_root *delayed_refs = NULL;
1133
	struct btrfs_delayed_ref_head *head;
1134 1135
	int info_level = 0;
	int ret;
1136
	struct prelim_ref *ref;
1137
	struct rb_node *node;
1138
	struct extent_inode_elem *eie = NULL;
1139
	/* total of both direct AND indirect refs! */
1140
	u64 total_refs = 0;
1141 1142 1143 1144 1145
	struct preftrees preftrees = {
		.direct = PREFTREE_INIT,
		.indirect = PREFTREE_INIT,
		.indirect_missing_keys = PREFTREE_INIT
	};
1146 1147 1148

	key.objectid = bytenr;
	key.offset = (u64)-1;
1149 1150 1151 1152
	if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
		key.type = BTRFS_METADATA_ITEM_KEY;
	else
		key.type = BTRFS_EXTENT_ITEM_KEY;
1153 1154 1155 1156

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
1157
	if (!trans) {
1158
		path->search_commit_root = 1;
1159 1160
		path->skip_locking = 1;
	}
1161

1162
	if (time_seq == SEQ_LAST)
1163 1164
		path->skip_locking = 1;

1165 1166 1167 1168 1169 1170
	/*
	 * grab both a lock on the path and a lock on the delayed ref head.
	 * We need both to get a consistent picture of how the refs look
	 * at a specified point in time
	 */
again:
1171 1172
	head = NULL;

1173 1174 1175 1176 1177
	ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
	if (ret < 0)
		goto out;
	BUG_ON(ret == 0);

1178
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1179
	if (trans && likely(trans->type != __TRANS_DUMMY) &&
1180
	    time_seq != SEQ_LAST) {
1181
#else
1182
	if (trans && time_seq != SEQ_LAST) {
1183
#endif
1184 1185 1186 1187 1188 1189
		/*
		 * look if there are updates for this ref queued and lock the
		 * head
		 */
		delayed_refs = &trans->transaction->delayed_refs;
		spin_lock(&delayed_refs->lock);
1190
		head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
1191 1192
		if (head) {
			if (!mutex_trylock(&head->mutex)) {
1193
				refcount_inc(&head->refs);
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
				spin_unlock(&delayed_refs->lock);

				btrfs_release_path(path);

				/*
				 * Mutex was contended, block until it's
				 * released and try again
				 */
				mutex_lock(&head->mutex);
				mutex_unlock(&head->mutex);
1204
				btrfs_put_delayed_ref_head(head);
1205 1206
				goto again;
			}
1207
			spin_unlock(&delayed_refs->lock);
1208
			ret = add_delayed_refs(fs_info, head, time_seq,
1209
					       &preftrees, &total_refs, sc);
1210
			mutex_unlock(&head->mutex);
1211
			if (ret)
1212
				goto out;
1213 1214
		} else {
			spin_unlock(&delayed_refs->lock);
1215
		}
1216 1217 1218 1219 1220 1221
	}

	if (path->slots[0]) {
		struct extent_buffer *leaf;
		int slot;

1222
		path->slots[0]--;
1223
		leaf = path->nodes[0];
1224
		slot = path->slots[0];
1225 1226
		btrfs_item_key_to_cpu(leaf, &key, slot);
		if (key.objectid == bytenr &&
1227 1228
		    (key.type == BTRFS_EXTENT_ITEM_KEY ||
		     key.type == BTRFS_METADATA_ITEM_KEY)) {
1229 1230
			ret = add_inline_refs(fs_info, path, bytenr,
					      &info_level, &preftrees,
1231
					      &total_refs, sc);
1232 1233
			if (ret)
				goto out;
1234
			ret = add_keyed_refs(fs_info, path, bytenr, info_level,
1235
					     &preftrees, sc);
1236 1237 1238 1239 1240
			if (ret)
				goto out;
		}
	}

1241
	btrfs_release_path(path);
1242

1243
	ret = add_missing_keys(fs_info, &preftrees);
1244 1245 1246
	if (ret)
		goto out;

1247
	WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect_missing_keys.root));
1248

1249
	ret = resolve_indirect_refs(fs_info, path, time_seq, &preftrees,
1250
				    extent_item_pos, total_refs, sc, ignore_offset);
1251 1252 1253
	if (ret)
		goto out;

1254
	WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect.root));
1255

1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
	/*
	 * This walks the tree of merged and resolved refs. Tree blocks are
	 * read in as needed. Unique entries are added to the ulist, and
	 * the list of found roots is updated.
	 *
	 * We release the entire tree in one go before returning.
	 */
	node = rb_first(&preftrees.direct.root);
	while (node) {
		ref = rb_entry(node, struct prelim_ref, rbnode);
		node = rb_next(&ref->rbnode);
1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
		/*
		 * ref->count < 0 can happen here if there are delayed
		 * refs with a node->action of BTRFS_DROP_DELAYED_REF.
		 * prelim_ref_insert() relies on this when merging
		 * identical refs to keep the overall count correct.
		 * prelim_ref_insert() will merge only those refs
		 * which compare identically.  Any refs having
		 * e.g. different offsets would not be merged,
		 * and would retain their original ref->count < 0.
		 */
1277
		if (roots && ref->count && ref->root_id && ref->parent == 0) {
1278 1279
			if (sc && sc->root_objectid &&
			    ref->root_id != sc->root_objectid) {
1280 1281 1282 1283
				ret = BACKREF_FOUND_SHARED;
				goto out;
			}

1284 1285
			/* no parent == root of tree */
			ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
1286 1287
			if (ret < 0)
				goto out;
1288 1289
		}
		if (ref->count && ref->parent) {
1290 1291
			if (extent_item_pos && !ref->inode_list &&
			    ref->level == 0) {
1292
				struct extent_buffer *eb;
1293

1294
				eb = read_tree_block(fs_info, ref->parent, 0);
1295 1296 1297 1298
				if (IS_ERR(eb)) {
					ret = PTR_ERR(eb);
					goto out;
				} else if (!extent_buffer_uptodate(eb)) {
1299
					free_extent_buffer(eb);
1300 1301
					ret = -EIO;
					goto out;
1302
				}
1303 1304
				btrfs_tree_read_lock(eb);
				btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1305
				ret = find_extent_in_eb(eb, bytenr,
1306
							*extent_item_pos, &eie, ignore_offset);
1307
				btrfs_tree_read_unlock_blocking(eb);
1308
				free_extent_buffer(eb);
1309 1310 1311
				if (ret < 0)
					goto out;
				ref->inode_list = eie;
1312
			}
1313 1314 1315
			ret = ulist_add_merge_ptr(refs, ref->parent,
						  ref->inode_list,
						  (void **)&eie, GFP_NOFS);
1316 1317
			if (ret < 0)
				goto out;
1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
			if (!ret && extent_item_pos) {
				/*
				 * we've recorded that parent, so we must extend
				 * its inode list here
				 */
				BUG_ON(!eie);
				while (eie->next)
					eie = eie->next;
				eie->next = ref->inode_list;
			}
1328
			eie = NULL;
1329
		}
1330
		cond_resched();
1331 1332 1333 1334
	}

out:
	btrfs_free_path(path);
1335 1336 1337 1338 1339

	prelim_release(&preftrees.direct);
	prelim_release(&preftrees.indirect);
	prelim_release(&preftrees.indirect_missing_keys);

1340 1341
	if (ret < 0)
		free_inode_elem_list(eie);
1342 1343 1344
	return ret;
}

1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
static void free_leaf_list(struct ulist *blocks)
{
	struct ulist_node *node = NULL;
	struct extent_inode_elem *eie;
	struct ulist_iterator uiter;

	ULIST_ITER_INIT(&uiter);
	while ((node = ulist_next(blocks, &uiter))) {
		if (!node->aux)
			continue;
1355
		eie = unode_aux_to_inode_list(node);
1356
		free_inode_elem_list(eie);
1357 1358 1359 1360 1361 1362
		node->aux = 0;
	}

	ulist_free(blocks);
}

1363 1364 1365 1366 1367 1368 1369 1370 1371 1372
/*
 * Finds all leafs with a reference to the specified combination of bytenr and
 * offset. key_list_head will point to a list of corresponding keys (caller must
 * free each list element). The leafs will be stored in the leafs ulist, which
 * must be freed with ulist_free.
 *
 * returns 0 on success, <0 on error
 */
static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
				struct btrfs_fs_info *fs_info, u64 bytenr,
1373
				u64 time_seq, struct ulist **leafs,
1374
				const u64 *extent_item_pos, bool ignore_offset)
1375 1376 1377 1378
{
	int ret;

	*leafs = ulist_alloc(GFP_NOFS);
1379
	if (!*leafs)
1380 1381
		return -ENOMEM;

1382
	ret = find_parent_nodes(trans, fs_info, bytenr, time_seq,
1383
				*leafs, NULL, extent_item_pos, NULL, ignore_offset);
1384
	if (ret < 0 && ret != -ENOENT) {
1385
		free_leaf_list(*leafs);
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404
		return ret;
	}

	return 0;
}

/*
 * walk all backrefs for a given extent to find all roots that reference this
 * extent. Walking a backref means finding all extents that reference this
 * extent and in turn walk the backrefs of those, too. Naturally this is a
 * recursive process, but here it is implemented in an iterative fashion: We
 * find all referencing extents for the extent in question and put them on a
 * list. In turn, we find all referencing extents for those, further appending
 * to the list. The way we iterate the list allows adding more elements after
 * the current while iterating. The process stops when we reach the end of the
 * list. Found roots are added to the roots list.
 *
 * returns 0 on success, < 0 on error.
 */
1405 1406
static int btrfs_find_all_roots_safe(struct btrfs_trans_handle *trans,
				     struct btrfs_fs_info *fs_info, u64 bytenr,
1407 1408
				     u64 time_seq, struct ulist **roots,
				     bool ignore_offset)
1409 1410 1411
{
	struct ulist *tmp;
	struct ulist_node *node = NULL;
J
Jan Schmidt 已提交
1412
	struct ulist_iterator uiter;
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423
	int ret;

	tmp = ulist_alloc(GFP_NOFS);
	if (!tmp)
		return -ENOMEM;
	*roots = ulist_alloc(GFP_NOFS);
	if (!*roots) {
		ulist_free(tmp);
		return -ENOMEM;
	}

J
Jan Schmidt 已提交
1424
	ULIST_ITER_INIT(&uiter);
1425
	while (1) {
1426
		ret = find_parent_nodes(trans, fs_info, bytenr, time_seq,
1427
					tmp, *roots, NULL, NULL, ignore_offset);
1428 1429 1430 1431 1432
		if (ret < 0 && ret != -ENOENT) {
			ulist_free(tmp);
			ulist_free(*roots);
			return ret;
		}
J
Jan Schmidt 已提交
1433
		node = ulist_next(tmp, &uiter);
1434 1435 1436
		if (!node)
			break;
		bytenr = node->val;
1437
		cond_resched();
1438 1439 1440 1441 1442 1443
	}

	ulist_free(tmp);
	return 0;
}

1444 1445
int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
			 struct btrfs_fs_info *fs_info, u64 bytenr,
1446 1447
			 u64 time_seq, struct ulist **roots,
			 bool ignore_offset)
1448 1449 1450 1451 1452
{
	int ret;

	if (!trans)
		down_read(&fs_info->commit_root_sem);
1453
	ret = btrfs_find_all_roots_safe(trans, fs_info, bytenr,
1454
					time_seq, roots, ignore_offset);
1455 1456 1457 1458 1459
	if (!trans)
		up_read(&fs_info->commit_root_sem);
	return ret;
}

1460 1461 1462 1463 1464 1465 1466 1467 1468
/**
 * btrfs_check_shared - tell us whether an extent is shared
 *
 * btrfs_check_shared uses the backref walking code but will short
 * circuit as soon as it finds a root or inode that doesn't match the
 * one passed in. This provides a significant performance benefit for
 * callers (such as fiemap) which want to know whether the extent is
 * shared but do not need a ref count.
 *
1469 1470 1471
 * This attempts to allocate a transaction in order to account for
 * delayed refs, but continues on even when the alloc fails.
 *
1472 1473
 * Return: 0 if extent is not shared, 1 if it is shared, < 0 on error.
 */
1474
int btrfs_check_shared(struct btrfs_root *root, u64 inum, u64 bytenr)
1475
{
1476 1477
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_trans_handle *trans;
1478 1479 1480 1481
	struct ulist *tmp = NULL;
	struct ulist *roots = NULL;
	struct ulist_iterator uiter;
	struct ulist_node *node;
1482
	struct seq_list elem = SEQ_LIST_INIT(elem);
1483
	int ret = 0;
1484 1485 1486 1487 1488
	struct share_check shared = {
		.root_objectid = root->objectid,
		.inum = inum,
		.share_count = 0,
	};
1489 1490 1491 1492 1493 1494 1495 1496 1497

	tmp = ulist_alloc(GFP_NOFS);
	roots = ulist_alloc(GFP_NOFS);
	if (!tmp || !roots) {
		ulist_free(tmp);
		ulist_free(roots);
		return -ENOMEM;
	}

1498 1499 1500
	trans = btrfs_join_transaction(root);
	if (IS_ERR(trans)) {
		trans = NULL;
1501
		down_read(&fs_info->commit_root_sem);
1502 1503 1504 1505
	} else {
		btrfs_get_tree_mod_seq(fs_info, &elem);
	}

1506 1507 1508
	ULIST_ITER_INIT(&uiter);
	while (1) {
		ret = find_parent_nodes(trans, fs_info, bytenr, elem.seq, tmp,
1509
					roots, NULL, &shared, false);
1510
		if (ret == BACKREF_FOUND_SHARED) {
1511
			/* this is the only condition under which we return 1 */
1512 1513 1514 1515 1516
			ret = 1;
			break;
		}
		if (ret < 0 && ret != -ENOENT)
			break;
1517
		ret = 0;
1518 1519 1520 1521 1522 1523
		node = ulist_next(tmp, &uiter);
		if (!node)
			break;
		bytenr = node->val;
		cond_resched();
	}
1524 1525

	if (trans) {
1526
		btrfs_put_tree_mod_seq(fs_info, &elem);
1527 1528
		btrfs_end_transaction(trans);
	} else {
1529
		up_read(&fs_info->commit_root_sem);
1530
	}
1531 1532 1533 1534 1535
	ulist_free(tmp);
	ulist_free(roots);
	return ret;
}

M
Mark Fasheh 已提交
1536 1537 1538 1539 1540 1541 1542 1543 1544
int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
			  u64 start_off, struct btrfs_path *path,
			  struct btrfs_inode_extref **ret_extref,
			  u64 *found_off)
{
	int ret, slot;
	struct btrfs_key key;
	struct btrfs_key found_key;
	struct btrfs_inode_extref *extref;
1545
	const struct extent_buffer *leaf;
M
Mark Fasheh 已提交
1546 1547 1548
	unsigned long ptr;

	key.objectid = inode_objectid;
1549
	key.type = BTRFS_INODE_EXTREF_KEY;
M
Mark Fasheh 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
	key.offset = start_off;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		return ret;

	while (1) {
		leaf = path->nodes[0];
		slot = path->slots[0];
		if (slot >= btrfs_header_nritems(leaf)) {
			/*
			 * If the item at offset is not found,
			 * btrfs_search_slot will point us to the slot
			 * where it should be inserted. In our case
			 * that will be the slot directly before the
			 * next INODE_REF_KEY_V2 item. In the case
			 * that we're pointing to the last slot in a
			 * leaf, we must move one leaf over.
			 */
			ret = btrfs_next_leaf(root, path);
			if (ret) {
				if (ret >= 1)
					ret = -ENOENT;
				break;
			}
			continue;
		}

		btrfs_item_key_to_cpu(leaf, &found_key, slot);

		/*
		 * Check that we're still looking at an extended ref key for
		 * this particular objectid. If we have different
		 * objectid or type then there are no more to be found
		 * in the tree and we can exit.
		 */
		ret = -ENOENT;
		if (found_key.objectid != inode_objectid)
			break;
1589
		if (found_key.type != BTRFS_INODE_EXTREF_KEY)
M
Mark Fasheh 已提交
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603
			break;

		ret = 0;
		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
		extref = (struct btrfs_inode_extref *)ptr;
		*ret_extref = extref;
		if (found_off)
			*found_off = found_key.offset;
		break;
	}

	return ret;
}

1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
/*
 * this iterates to turn a name (from iref/extref) into a full filesystem path.
 * Elements of the path are separated by '/' and the path is guaranteed to be
 * 0-terminated. the path is only given within the current file system.
 * Therefore, it never starts with a '/'. the caller is responsible to provide
 * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
 * the start point of the resulting string is returned. this pointer is within
 * dest, normally.
 * in case the path buffer would overflow, the pointer is decremented further
 * as if output was written to the buffer, though no more output is actually
 * generated. that way, the caller can determine how much space would be
 * required for the path to fit into the buffer. in that case, the returned
 * value will be smaller than dest. callers must check this!
 */
1618 1619 1620 1621
char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
			u32 name_len, unsigned long name_off,
			struct extent_buffer *eb_in, u64 parent,
			char *dest, u32 size)
1622 1623 1624 1625
{
	int slot;
	u64 next_inum;
	int ret;
1626
	s64 bytes_left = ((s64)size) - 1;
1627 1628
	struct extent_buffer *eb = eb_in;
	struct btrfs_key found_key;
1629
	int leave_spinning = path->leave_spinning;
M
Mark Fasheh 已提交
1630
	struct btrfs_inode_ref *iref;
1631 1632 1633 1634

	if (bytes_left >= 0)
		dest[bytes_left] = '\0';

1635
	path->leave_spinning = 1;
1636
	while (1) {
M
Mark Fasheh 已提交
1637
		bytes_left -= name_len;
1638 1639
		if (bytes_left >= 0)
			read_extent_buffer(eb, dest + bytes_left,
M
Mark Fasheh 已提交
1640
					   name_off, name_len);
1641
		if (eb != eb_in) {
1642 1643
			if (!path->skip_locking)
				btrfs_tree_read_unlock_blocking(eb);
1644
			free_extent_buffer(eb);
1645
		}
1646 1647
		ret = btrfs_find_item(fs_root, path, parent, 0,
				BTRFS_INODE_REF_KEY, &found_key);
1648 1649
		if (ret > 0)
			ret = -ENOENT;
1650 1651
		if (ret)
			break;
M
Mark Fasheh 已提交
1652

1653 1654 1655 1656 1657 1658 1659 1660 1661
		next_inum = found_key.offset;

		/* regular exit ahead */
		if (parent == next_inum)
			break;

		slot = path->slots[0];
		eb = path->nodes[0];
		/* make sure we can use eb after releasing the path */
1662
		if (eb != eb_in) {
1663 1664 1665 1666
			if (!path->skip_locking)
				btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
			path->nodes[0] = NULL;
			path->locks[0] = 0;
1667
		}
1668 1669
		btrfs_release_path(path);
		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
M
Mark Fasheh 已提交
1670 1671 1672 1673

		name_len = btrfs_inode_ref_name_len(eb, iref);
		name_off = (unsigned long)(iref + 1);

1674 1675 1676 1677 1678 1679 1680
		parent = next_inum;
		--bytes_left;
		if (bytes_left >= 0)
			dest[bytes_left] = '/';
	}

	btrfs_release_path(path);
1681
	path->leave_spinning = leave_spinning;
1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694

	if (ret)
		return ERR_PTR(ret);

	return dest + bytes_left;
}

/*
 * this makes the path point to (logical EXTENT_ITEM *)
 * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
 * tree blocks and <0 on error.
 */
int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1695 1696
			struct btrfs_path *path, struct btrfs_key *found_key,
			u64 *flags_ret)
1697 1698 1699
{
	int ret;
	u64 flags;
1700
	u64 size = 0;
1701
	u32 item_size;
1702
	const struct extent_buffer *eb;
1703 1704 1705
	struct btrfs_extent_item *ei;
	struct btrfs_key key;

1706 1707 1708 1709
	if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
		key.type = BTRFS_METADATA_ITEM_KEY;
	else
		key.type = BTRFS_EXTENT_ITEM_KEY;
1710 1711 1712 1713 1714 1715 1716
	key.objectid = logical;
	key.offset = (u64)-1;

	ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
	if (ret < 0)
		return ret;

1717 1718 1719 1720 1721
	ret = btrfs_previous_extent_item(fs_info->extent_root, path, 0);
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		return ret;
1722
	}
1723
	btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
1724
	if (found_key->type == BTRFS_METADATA_ITEM_KEY)
1725
		size = fs_info->nodesize;
1726 1727 1728
	else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
		size = found_key->offset;

1729
	if (found_key->objectid > logical ||
1730
	    found_key->objectid + size <= logical) {
1731 1732
		btrfs_debug(fs_info,
			"logical %llu is not within any extent", logical);
1733
		return -ENOENT;
J
Jan Schmidt 已提交
1734
	}
1735 1736 1737 1738 1739 1740 1741 1742

	eb = path->nodes[0];
	item_size = btrfs_item_size_nr(eb, path->slots[0]);
	BUG_ON(item_size < sizeof(*ei));

	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
	flags = btrfs_extent_flags(eb, ei);

1743 1744
	btrfs_debug(fs_info,
		"logical %llu is at position %llu within the extent (%llu EXTENT_ITEM %llu) flags %#llx size %u",
1745 1746
		 logical, logical - found_key->objectid, found_key->objectid,
		 found_key->offset, flags, item_size);
1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757

	WARN_ON(!flags_ret);
	if (flags_ret) {
		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
			*flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
		else if (flags & BTRFS_EXTENT_FLAG_DATA)
			*flags_ret = BTRFS_EXTENT_FLAG_DATA;
		else
			BUG_ON(1);
		return 0;
	}
1758 1759 1760 1761 1762 1763 1764 1765

	return -EIO;
}

/*
 * helper function to iterate extent inline refs. ptr must point to a 0 value
 * for the first call and may be modified. it is used to track state.
 * if more refs exist, 0 is returned and the next call to
1766
 * get_extent_inline_ref must pass the modified ptr parameter to get the
1767 1768 1769
 * next ref. after the last ref was processed, 1 is returned.
 * returns <0 on error
 */
1770 1771 1772 1773 1774 1775 1776
static int get_extent_inline_ref(unsigned long *ptr,
				 const struct extent_buffer *eb,
				 const struct btrfs_key *key,
				 const struct btrfs_extent_item *ei,
				 u32 item_size,
				 struct btrfs_extent_inline_ref **out_eiref,
				 int *out_type)
1777 1778 1779 1780 1781 1782 1783 1784 1785
{
	unsigned long end;
	u64 flags;
	struct btrfs_tree_block_info *info;

	if (!*ptr) {
		/* first call */
		flags = btrfs_extent_flags(eb, ei);
		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
			if (key->type == BTRFS_METADATA_ITEM_KEY) {
				/* a skinny metadata extent */
				*out_eiref =
				     (struct btrfs_extent_inline_ref *)(ei + 1);
			} else {
				WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
				info = (struct btrfs_tree_block_info *)(ei + 1);
				*out_eiref =
				   (struct btrfs_extent_inline_ref *)(info + 1);
			}
1796 1797 1798 1799
		} else {
			*out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
		}
		*ptr = (unsigned long)*out_eiref;
1800
		if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
1801 1802 1803 1804
			return -ENOENT;
	}

	end = (unsigned long)ei + item_size;
1805
	*out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
1806 1807 1808 1809
	*out_type = btrfs_get_extent_inline_ref_type(eb, *out_eiref,
						     BTRFS_REF_TYPE_ANY);
	if (*out_type == BTRFS_REF_TYPE_INVALID)
		return -EINVAL;
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821

	*ptr += btrfs_extent_inline_ref_size(*out_type);
	WARN_ON(*ptr > end);
	if (*ptr == end)
		return 1; /* last */

	return 0;
}

/*
 * reads the tree block backref for an extent. tree level and root are returned
 * through out_level and out_root. ptr must point to a 0 value for the first
1822
 * call and may be modified (see get_extent_inline_ref comment).
1823 1824 1825 1826
 * returns 0 if data was provided, 1 if there was no more data to provide or
 * <0 on error.
 */
int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1827 1828
			    struct btrfs_key *key, struct btrfs_extent_item *ei,
			    u32 item_size, u64 *out_root, u8 *out_level)
1829 1830 1831 1832 1833 1834 1835 1836 1837
{
	int ret;
	int type;
	struct btrfs_extent_inline_ref *eiref;

	if (*ptr == (unsigned long)-1)
		return 1;

	while (1) {
1838
		ret = get_extent_inline_ref(ptr, eb, key, ei, item_size,
1839
					      &eiref, &type);
1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
		if (ret < 0)
			return ret;

		if (type == BTRFS_TREE_BLOCK_REF_KEY ||
		    type == BTRFS_SHARED_BLOCK_REF_KEY)
			break;

		if (ret == 1)
			return 1;
	}

	/* we can treat both ref types equally here */
	*out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862

	if (key->type == BTRFS_EXTENT_ITEM_KEY) {
		struct btrfs_tree_block_info *info;

		info = (struct btrfs_tree_block_info *)(ei + 1);
		*out_level = btrfs_tree_block_level(eb, info);
	} else {
		ASSERT(key->type == BTRFS_METADATA_ITEM_KEY);
		*out_level = (u8)key->offset;
	}
1863 1864 1865 1866 1867 1868 1869

	if (ret == 1)
		*ptr = (unsigned long)-1;

	return 0;
}

1870 1871 1872 1873
static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
			     struct extent_inode_elem *inode_list,
			     u64 root, u64 extent_item_objectid,
			     iterate_extent_inodes_t *iterate, void *ctx)
1874
{
1875
	struct extent_inode_elem *eie;
J
Jan Schmidt 已提交
1876 1877
	int ret = 0;

1878
	for (eie = inode_list; eie; eie = eie->next) {
1879 1880 1881 1882
		btrfs_debug(fs_info,
			    "ref for %llu resolved, key (%llu EXTEND_DATA %llu), root %llu",
			    extent_item_objectid, eie->inum,
			    eie->offset, root);
1883
		ret = iterate(eie->inum, eie->offset, root, ctx);
J
Jan Schmidt 已提交
1884
		if (ret) {
1885 1886 1887
			btrfs_debug(fs_info,
				    "stopping iteration for %llu due to ret=%d",
				    extent_item_objectid, ret);
J
Jan Schmidt 已提交
1888 1889
			break;
		}
1890 1891 1892 1893 1894 1895 1896
	}

	return ret;
}

/*
 * calls iterate() for every inode that references the extent identified by
J
Jan Schmidt 已提交
1897
 * the given parameters.
1898 1899 1900
 * when the iterator function returns a non-zero value, iteration stops.
 */
int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
J
Jan Schmidt 已提交
1901
				u64 extent_item_objectid, u64 extent_item_pos,
1902
				int search_commit_root,
1903 1904
				iterate_extent_inodes_t *iterate, void *ctx,
				bool ignore_offset)
1905 1906
{
	int ret;
1907
	struct btrfs_trans_handle *trans = NULL;
1908 1909
	struct ulist *refs = NULL;
	struct ulist *roots = NULL;
J
Jan Schmidt 已提交
1910 1911
	struct ulist_node *ref_node = NULL;
	struct ulist_node *root_node = NULL;
1912
	struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
J
Jan Schmidt 已提交
1913 1914
	struct ulist_iterator ref_uiter;
	struct ulist_iterator root_uiter;
1915

1916
	btrfs_debug(fs_info, "resolving all inodes for extent %llu",
J
Jan Schmidt 已提交
1917
			extent_item_objectid);
1918

1919
	if (!search_commit_root) {
1920 1921 1922
		trans = btrfs_join_transaction(fs_info->extent_root);
		if (IS_ERR(trans))
			return PTR_ERR(trans);
1923
		btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1924 1925
	} else {
		down_read(&fs_info->commit_root_sem);
1926
	}
1927

J
Jan Schmidt 已提交
1928
	ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1929
				   tree_mod_seq_elem.seq, &refs,
1930
				   &extent_item_pos, ignore_offset);
J
Jan Schmidt 已提交
1931 1932
	if (ret)
		goto out;
1933

J
Jan Schmidt 已提交
1934 1935
	ULIST_ITER_INIT(&ref_uiter);
	while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1936
		ret = btrfs_find_all_roots_safe(trans, fs_info, ref_node->val,
1937 1938
						tree_mod_seq_elem.seq, &roots,
						ignore_offset);
J
Jan Schmidt 已提交
1939 1940
		if (ret)
			break;
J
Jan Schmidt 已提交
1941 1942
		ULIST_ITER_INIT(&root_uiter);
		while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1943 1944 1945 1946 1947 1948
			btrfs_debug(fs_info,
				    "root %llu references leaf %llu, data list %#llx",
				    root_node->val, ref_node->val,
				    ref_node->aux);
			ret = iterate_leaf_refs(fs_info,
						(struct extent_inode_elem *)
1949 1950 1951 1952
						(uintptr_t)ref_node->aux,
						root_node->val,
						extent_item_objectid,
						iterate, ctx);
J
Jan Schmidt 已提交
1953
		}
1954
		ulist_free(roots);
1955 1956
	}

1957
	free_leaf_list(refs);
J
Jan Schmidt 已提交
1958
out:
1959
	if (!search_commit_root) {
1960
		btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1961
		btrfs_end_transaction(trans);
1962 1963
	} else {
		up_read(&fs_info->commit_root_sem);
1964 1965
	}

1966 1967 1968 1969 1970
	return ret;
}

int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
				struct btrfs_path *path,
1971 1972
				iterate_extent_inodes_t *iterate, void *ctx,
				bool ignore_offset)
1973 1974
{
	int ret;
J
Jan Schmidt 已提交
1975
	u64 extent_item_pos;
1976
	u64 flags = 0;
1977
	struct btrfs_key found_key;
1978
	int search_commit_root = path->search_commit_root;
1979

1980
	ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
J
Jan Schmidt 已提交
1981
	btrfs_release_path(path);
1982 1983
	if (ret < 0)
		return ret;
1984
	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1985
		return -EINVAL;
1986

J
Jan Schmidt 已提交
1987
	extent_item_pos = logical - found_key.objectid;
1988 1989
	ret = iterate_extent_inodes(fs_info, found_key.objectid,
					extent_item_pos, search_commit_root,
1990
					iterate, ctx, ignore_offset);
1991 1992 1993 1994

	return ret;
}

M
Mark Fasheh 已提交
1995 1996 1997 1998 1999 2000
typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
			      struct extent_buffer *eb, void *ctx);

static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
			      struct btrfs_path *path,
			      iterate_irefs_t *iterate, void *ctx)
2001
{
2002
	int ret = 0;
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
	int slot;
	u32 cur;
	u32 len;
	u32 name_len;
	u64 parent = 0;
	int found = 0;
	struct extent_buffer *eb;
	struct btrfs_item *item;
	struct btrfs_inode_ref *iref;
	struct btrfs_key found_key;

2014
	while (!ret) {
2015 2016 2017 2018
		ret = btrfs_find_item(fs_root, path, inum,
				parent ? parent + 1 : 0, BTRFS_INODE_REF_KEY,
				&found_key);

2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
		if (ret < 0)
			break;
		if (ret) {
			ret = found ? 0 : -ENOENT;
			break;
		}
		++found;

		parent = found_key.offset;
		slot = path->slots[0];
2029 2030 2031 2032 2033 2034
		eb = btrfs_clone_extent_buffer(path->nodes[0]);
		if (!eb) {
			ret = -ENOMEM;
			break;
		}
		extent_buffer_get(eb);
2035 2036
		btrfs_tree_read_lock(eb);
		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
2037 2038
		btrfs_release_path(path);

2039
		item = btrfs_item_nr(slot);
2040 2041 2042 2043 2044
		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);

		for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
			name_len = btrfs_inode_ref_name_len(eb, iref);
			/* path must be released before calling iterate()! */
2045 2046 2047
			btrfs_debug(fs_root->fs_info,
				"following ref at offset %u for inode %llu in tree %llu",
				cur, found_key.objectid, fs_root->objectid);
M
Mark Fasheh 已提交
2048 2049
			ret = iterate(parent, name_len,
				      (unsigned long)(iref + 1), eb, ctx);
2050
			if (ret)
2051 2052 2053 2054
				break;
			len = sizeof(*iref) + name_len;
			iref = (struct btrfs_inode_ref *)((char *)iref + len);
		}
2055
		btrfs_tree_read_unlock_blocking(eb);
2056 2057 2058 2059 2060 2061 2062 2063
		free_extent_buffer(eb);
	}

	btrfs_release_path(path);

	return ret;
}

M
Mark Fasheh 已提交
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
				 struct btrfs_path *path,
				 iterate_irefs_t *iterate, void *ctx)
{
	int ret;
	int slot;
	u64 offset = 0;
	u64 parent;
	int found = 0;
	struct extent_buffer *eb;
	struct btrfs_inode_extref *extref;
	u32 item_size;
	u32 cur_offset;
	unsigned long ptr;

	while (1) {
		ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
					    &offset);
		if (ret < 0)
			break;
		if (ret) {
			ret = found ? 0 : -ENOENT;
			break;
		}
		++found;

		slot = path->slots[0];
2091 2092 2093 2094 2095 2096
		eb = btrfs_clone_extent_buffer(path->nodes[0]);
		if (!eb) {
			ret = -ENOMEM;
			break;
		}
		extent_buffer_get(eb);
M
Mark Fasheh 已提交
2097 2098 2099 2100 2101

		btrfs_tree_read_lock(eb);
		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
		btrfs_release_path(path);

2102 2103
		item_size = btrfs_item_size_nr(eb, slot);
		ptr = btrfs_item_ptr_offset(eb, slot);
M
Mark Fasheh 已提交
2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
		cur_offset = 0;

		while (cur_offset < item_size) {
			u32 name_len;

			extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
			parent = btrfs_inode_extref_parent(eb, extref);
			name_len = btrfs_inode_extref_name_len(eb, extref);
			ret = iterate(parent, name_len,
				      (unsigned long)&extref->name, eb, ctx);
			if (ret)
				break;

2117
			cur_offset += btrfs_inode_extref_name_len(eb, extref);
M
Mark Fasheh 已提交
2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150
			cur_offset += sizeof(*extref);
		}
		btrfs_tree_read_unlock_blocking(eb);
		free_extent_buffer(eb);

		offset++;
	}

	btrfs_release_path(path);

	return ret;
}

static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
			 struct btrfs_path *path, iterate_irefs_t *iterate,
			 void *ctx)
{
	int ret;
	int found_refs = 0;

	ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
	if (!ret)
		++found_refs;
	else if (ret != -ENOENT)
		return ret;

	ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
	if (ret == -ENOENT && found_refs)
		return 0;

	return ret;
}

2151 2152 2153 2154
/*
 * returns 0 if the path could be dumped (probably truncated)
 * returns <0 in case of an error
 */
M
Mark Fasheh 已提交
2155 2156
static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
			 struct extent_buffer *eb, void *ctx)
2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167
{
	struct inode_fs_paths *ipath = ctx;
	char *fspath;
	char *fspath_min;
	int i = ipath->fspath->elem_cnt;
	const int s_ptr = sizeof(char *);
	u32 bytes_left;

	bytes_left = ipath->fspath->bytes_left > s_ptr ?
					ipath->fspath->bytes_left - s_ptr : 0;

2168
	fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
2169 2170
	fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
				   name_off, eb, inum, fspath_min, bytes_left);
2171 2172 2173 2174
	if (IS_ERR(fspath))
		return PTR_ERR(fspath);

	if (fspath > fspath_min) {
2175
		ipath->fspath->val[i] = (u64)(unsigned long)fspath;
2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
		++ipath->fspath->elem_cnt;
		ipath->fspath->bytes_left = fspath - fspath_min;
	} else {
		++ipath->fspath->elem_missed;
		ipath->fspath->bytes_missing += fspath_min - fspath;
		ipath->fspath->bytes_left = 0;
	}

	return 0;
}

/*
 * this dumps all file system paths to the inode into the ipath struct, provided
 * is has been created large enough. each path is zero-terminated and accessed
2190
 * from ipath->fspath->val[i].
2191
 * when it returns, there are ipath->fspath->elem_cnt number of paths available
2192
 * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
2193
 * number of missed paths is recorded in ipath->fspath->elem_missed, otherwise,
2194 2195 2196 2197 2198 2199
 * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
 * have been needed to return all paths.
 */
int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
{
	return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
M
Mark Fasheh 已提交
2200
			     inode_to_path, ipath);
2201 2202 2203 2204 2205 2206 2207 2208
}

struct btrfs_data_container *init_data_container(u32 total_bytes)
{
	struct btrfs_data_container *data;
	size_t alloc_bytes;

	alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
2209
	data = kvmalloc(alloc_bytes, GFP_KERNEL);
2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242
	if (!data)
		return ERR_PTR(-ENOMEM);

	if (total_bytes >= sizeof(*data)) {
		data->bytes_left = total_bytes - sizeof(*data);
		data->bytes_missing = 0;
	} else {
		data->bytes_missing = sizeof(*data) - total_bytes;
		data->bytes_left = 0;
	}

	data->elem_cnt = 0;
	data->elem_missed = 0;

	return data;
}

/*
 * allocates space to return multiple file system paths for an inode.
 * total_bytes to allocate are passed, note that space usable for actual path
 * information will be total_bytes - sizeof(struct inode_fs_paths).
 * the returned pointer must be freed with free_ipath() in the end.
 */
struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
					struct btrfs_path *path)
{
	struct inode_fs_paths *ifp;
	struct btrfs_data_container *fspath;

	fspath = init_data_container(total_bytes);
	if (IS_ERR(fspath))
		return (void *)fspath;

2243
	ifp = kmalloc(sizeof(*ifp), GFP_KERNEL);
2244
	if (!ifp) {
2245
		kvfree(fspath);
2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257
		return ERR_PTR(-ENOMEM);
	}

	ifp->btrfs_path = path;
	ifp->fspath = fspath;
	ifp->fs_root = fs_root;

	return ifp;
}

void free_ipath(struct inode_fs_paths *ipath)
{
2258 2259
	if (!ipath)
		return;
2260
	kvfree(ipath->fspath);
2261 2262
	kfree(ipath);
}