2 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
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.
19 #define _XOPEN_SOURCE 500
25 #include <sys/types.h>
29 #include <uuid/uuid.h>
34 #include "print-tree.h"
35 #include "transaction.h"
39 #include "free-space-cache.h"
42 static u64 bytes_used = 0;
43 static u64 total_csum_bytes = 0;
44 static u64 total_btree_bytes = 0;
45 static u64 total_fs_tree_bytes = 0;
46 static u64 total_extent_tree_bytes = 0;
47 static u64 btree_space_waste = 0;
48 static u64 data_bytes_allocated = 0;
49 static u64 data_bytes_referenced = 0;
50 static int found_old_backref = 0;
51 static LIST_HEAD(duplicate_extents);
52 static LIST_HEAD(delete_items);
53 static int repair = 0;
54 static int no_holes = 0;
56 struct extent_backref {
57 struct list_head list;
58 unsigned int is_data:1;
59 unsigned int found_extent_tree:1;
60 unsigned int full_backref:1;
61 unsigned int found_ref:1;
62 unsigned int broken:1;
66 struct extent_backref node;
81 struct extent_backref node;
88 struct extent_record {
89 struct list_head backrefs;
90 struct list_head dups;
91 struct list_head list;
92 struct cache_extent cache;
93 struct btrfs_disk_key parent_key;
94 unsigned int found_rec;
101 u64 parent_generation;
105 unsigned int content_checked:1;
106 unsigned int owner_ref_checked:1;
107 unsigned int is_root:1;
108 unsigned int metadata:1;
111 struct inode_backref {
112 struct list_head list;
113 unsigned int found_dir_item:1;
114 unsigned int found_dir_index:1;
115 unsigned int found_inode_ref:1;
116 unsigned int filetype:8;
118 unsigned int ref_type;
125 struct dropping_root_item_record {
126 struct list_head list;
127 struct btrfs_root_item ri;
128 struct btrfs_key found_key;
131 #define REF_ERR_NO_DIR_ITEM (1 << 0)
132 #define REF_ERR_NO_DIR_INDEX (1 << 1)
133 #define REF_ERR_NO_INODE_REF (1 << 2)
134 #define REF_ERR_DUP_DIR_ITEM (1 << 3)
135 #define REF_ERR_DUP_DIR_INDEX (1 << 4)
136 #define REF_ERR_DUP_INODE_REF (1 << 5)
137 #define REF_ERR_INDEX_UNMATCH (1 << 6)
138 #define REF_ERR_FILETYPE_UNMATCH (1 << 7)
139 #define REF_ERR_NAME_TOO_LONG (1 << 8) // 100
140 #define REF_ERR_NO_ROOT_REF (1 << 9)
141 #define REF_ERR_NO_ROOT_BACKREF (1 << 10)
142 #define REF_ERR_DUP_ROOT_REF (1 << 11)
143 #define REF_ERR_DUP_ROOT_BACKREF (1 << 12)
145 struct inode_record {
146 struct list_head backrefs;
147 unsigned int checked:1;
148 unsigned int merging:1;
149 unsigned int found_inode_item:1;
150 unsigned int found_dir_item:1;
151 unsigned int found_file_extent:1;
152 unsigned int found_csum_item:1;
153 unsigned int some_csum_missing:1;
154 unsigned int nodatasum:1;
167 u64 first_extent_gap;
172 #define I_ERR_NO_INODE_ITEM (1 << 0)
173 #define I_ERR_NO_ORPHAN_ITEM (1 << 1)
174 #define I_ERR_DUP_INODE_ITEM (1 << 2)
175 #define I_ERR_DUP_DIR_INDEX (1 << 3)
176 #define I_ERR_ODD_DIR_ITEM (1 << 4)
177 #define I_ERR_ODD_FILE_EXTENT (1 << 5)
178 #define I_ERR_BAD_FILE_EXTENT (1 << 6)
179 #define I_ERR_FILE_EXTENT_OVERLAP (1 << 7)
180 #define I_ERR_FILE_EXTENT_DISCOUNT (1 << 8) // 100
181 #define I_ERR_DIR_ISIZE_WRONG (1 << 9)
182 #define I_ERR_FILE_NBYTES_WRONG (1 << 10) // 400
183 #define I_ERR_ODD_CSUM_ITEM (1 << 11)
184 #define I_ERR_SOME_CSUM_MISSING (1 << 12)
185 #define I_ERR_LINK_COUNT_WRONG (1 << 13)
187 struct root_backref {
188 struct list_head list;
189 unsigned int found_dir_item:1;
190 unsigned int found_dir_index:1;
191 unsigned int found_back_ref:1;
192 unsigned int found_forward_ref:1;
193 unsigned int reachable:1;
203 struct list_head backrefs;
204 struct cache_extent cache;
205 unsigned int found_root_item:1;
211 struct cache_extent cache;
216 struct cache_extent cache;
217 struct cache_tree root_cache;
218 struct cache_tree inode_cache;
219 struct inode_record *current;
228 struct walk_control {
229 struct cache_tree shared;
230 struct shared_node *nodes[BTRFS_MAX_LEVEL];
236 struct btrfs_key key;
238 struct list_head list;
241 static void reset_cached_block_groups(struct btrfs_fs_info *fs_info);
243 static u8 imode_to_type(u32 imode)
246 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
247 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
248 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
249 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
250 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
251 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
252 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
253 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
256 return btrfs_type_by_mode[(imode & S_IFMT) >> S_SHIFT];
260 static int device_record_compare(struct rb_node *node1, struct rb_node *node2)
262 struct device_record *rec1;
263 struct device_record *rec2;
265 rec1 = rb_entry(node1, struct device_record, node);
266 rec2 = rb_entry(node2, struct device_record, node);
267 if (rec1->devid > rec2->devid)
269 else if (rec1->devid < rec2->devid)
275 static struct inode_record *clone_inode_rec(struct inode_record *orig_rec)
277 struct inode_record *rec;
278 struct inode_backref *backref;
279 struct inode_backref *orig;
282 rec = malloc(sizeof(*rec));
283 memcpy(rec, orig_rec, sizeof(*rec));
285 INIT_LIST_HEAD(&rec->backrefs);
287 list_for_each_entry(orig, &orig_rec->backrefs, list) {
288 size = sizeof(*orig) + orig->namelen + 1;
289 backref = malloc(size);
290 memcpy(backref, orig, size);
291 list_add_tail(&backref->list, &rec->backrefs);
296 static void print_inode_error(int errors)
298 if (errors & I_ERR_NO_INODE_ITEM)
299 fprintf(stderr, ", no inode item");
300 if (errors & I_ERR_NO_ORPHAN_ITEM)
301 fprintf(stderr, ", no orphan item");
302 if (errors & I_ERR_DUP_INODE_ITEM)
303 fprintf(stderr, ", dup inode item");
304 if (errors & I_ERR_DUP_DIR_INDEX)
305 fprintf(stderr, ", dup dir index");
306 if (errors & I_ERR_ODD_DIR_ITEM)
307 fprintf(stderr, ", odd dir item");
308 if (errors & I_ERR_ODD_FILE_EXTENT)
309 fprintf(stderr, ", odd file extent");
310 if (errors & I_ERR_BAD_FILE_EXTENT)
311 fprintf(stderr, ", bad file extent");
312 if (errors & I_ERR_FILE_EXTENT_OVERLAP)
313 fprintf(stderr, ", file extent overlap");
314 if (errors & I_ERR_FILE_EXTENT_DISCOUNT)
315 fprintf(stderr, ", file extent discount");
316 if (errors & I_ERR_DIR_ISIZE_WRONG)
317 fprintf(stderr, ", dir isize wrong");
318 if (errors & I_ERR_FILE_NBYTES_WRONG)
319 fprintf(stderr, ", nbytes wrong");
320 if (errors & I_ERR_ODD_CSUM_ITEM)
321 fprintf(stderr, ", odd csum item");
322 if (errors & I_ERR_SOME_CSUM_MISSING)
323 fprintf(stderr, ", some csum missing");
324 if (errors & I_ERR_LINK_COUNT_WRONG)
325 fprintf(stderr, ", link count wrong");
326 fprintf(stderr, "\n");
329 static void print_ref_error(int errors)
331 if (errors & REF_ERR_NO_DIR_ITEM)
332 fprintf(stderr, ", no dir item");
333 if (errors & REF_ERR_NO_DIR_INDEX)
334 fprintf(stderr, ", no dir index");
335 if (errors & REF_ERR_NO_INODE_REF)
336 fprintf(stderr, ", no inode ref");
337 if (errors & REF_ERR_DUP_DIR_ITEM)
338 fprintf(stderr, ", dup dir item");
339 if (errors & REF_ERR_DUP_DIR_INDEX)
340 fprintf(stderr, ", dup dir index");
341 if (errors & REF_ERR_DUP_INODE_REF)
342 fprintf(stderr, ", dup inode ref");
343 if (errors & REF_ERR_INDEX_UNMATCH)
344 fprintf(stderr, ", index unmatch");
345 if (errors & REF_ERR_FILETYPE_UNMATCH)
346 fprintf(stderr, ", filetype unmatch");
347 if (errors & REF_ERR_NAME_TOO_LONG)
348 fprintf(stderr, ", name too long");
349 if (errors & REF_ERR_NO_ROOT_REF)
350 fprintf(stderr, ", no root ref");
351 if (errors & REF_ERR_NO_ROOT_BACKREF)
352 fprintf(stderr, ", no root backref");
353 if (errors & REF_ERR_DUP_ROOT_REF)
354 fprintf(stderr, ", dup root ref");
355 if (errors & REF_ERR_DUP_ROOT_BACKREF)
356 fprintf(stderr, ", dup root backref");
357 fprintf(stderr, "\n");
360 static struct inode_record *get_inode_rec(struct cache_tree *inode_cache,
363 struct ptr_node *node;
364 struct cache_extent *cache;
365 struct inode_record *rec = NULL;
368 cache = lookup_cache_extent(inode_cache, ino, 1);
370 node = container_of(cache, struct ptr_node, cache);
372 if (mod && rec->refs > 1) {
373 node->data = clone_inode_rec(rec);
378 rec = calloc(1, sizeof(*rec));
380 rec->extent_start = (u64)-1;
381 rec->first_extent_gap = (u64)-1;
383 INIT_LIST_HEAD(&rec->backrefs);
385 node = malloc(sizeof(*node));
386 node->cache.start = ino;
387 node->cache.size = 1;
390 if (ino == BTRFS_FREE_INO_OBJECTID)
393 ret = insert_cache_extent(inode_cache, &node->cache);
399 static void free_inode_rec(struct inode_record *rec)
401 struct inode_backref *backref;
406 while (!list_empty(&rec->backrefs)) {
407 backref = list_entry(rec->backrefs.next,
408 struct inode_backref, list);
409 list_del(&backref->list);
415 static int can_free_inode_rec(struct inode_record *rec)
417 if (!rec->errors && rec->checked && rec->found_inode_item &&
418 rec->nlink == rec->found_link && list_empty(&rec->backrefs))
423 static void maybe_free_inode_rec(struct cache_tree *inode_cache,
424 struct inode_record *rec)
426 struct cache_extent *cache;
427 struct inode_backref *tmp, *backref;
428 struct ptr_node *node;
429 unsigned char filetype;
431 if (!rec->found_inode_item)
434 filetype = imode_to_type(rec->imode);
435 list_for_each_entry_safe(backref, tmp, &rec->backrefs, list) {
436 if (backref->found_dir_item && backref->found_dir_index) {
437 if (backref->filetype != filetype)
438 backref->errors |= REF_ERR_FILETYPE_UNMATCH;
439 if (!backref->errors && backref->found_inode_ref) {
440 list_del(&backref->list);
446 if (!rec->checked || rec->merging)
449 if (S_ISDIR(rec->imode)) {
450 if (rec->found_size != rec->isize)
451 rec->errors |= I_ERR_DIR_ISIZE_WRONG;
452 if (rec->found_file_extent)
453 rec->errors |= I_ERR_ODD_FILE_EXTENT;
454 } else if (S_ISREG(rec->imode) || S_ISLNK(rec->imode)) {
455 if (rec->found_dir_item)
456 rec->errors |= I_ERR_ODD_DIR_ITEM;
457 if (rec->found_size != rec->nbytes)
458 rec->errors |= I_ERR_FILE_NBYTES_WRONG;
459 if (rec->extent_start == (u64)-1 || rec->extent_start > 0)
460 rec->first_extent_gap = 0;
461 if (rec->nlink > 0 && !no_holes &&
462 (rec->extent_end < rec->isize ||
463 rec->first_extent_gap < rec->isize))
464 rec->errors |= I_ERR_FILE_EXTENT_DISCOUNT;
467 if (S_ISREG(rec->imode) || S_ISLNK(rec->imode)) {
468 if (rec->found_csum_item && rec->nodatasum)
469 rec->errors |= I_ERR_ODD_CSUM_ITEM;
470 if (rec->some_csum_missing && !rec->nodatasum)
471 rec->errors |= I_ERR_SOME_CSUM_MISSING;
474 BUG_ON(rec->refs != 1);
475 if (can_free_inode_rec(rec)) {
476 cache = lookup_cache_extent(inode_cache, rec->ino, 1);
477 node = container_of(cache, struct ptr_node, cache);
478 BUG_ON(node->data != rec);
479 remove_cache_extent(inode_cache, &node->cache);
485 static int check_orphan_item(struct btrfs_root *root, u64 ino)
487 struct btrfs_path path;
488 struct btrfs_key key;
491 key.objectid = BTRFS_ORPHAN_OBJECTID;
492 key.type = BTRFS_ORPHAN_ITEM_KEY;
495 btrfs_init_path(&path);
496 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
497 btrfs_release_path(&path);
503 static int process_inode_item(struct extent_buffer *eb,
504 int slot, struct btrfs_key *key,
505 struct shared_node *active_node)
507 struct inode_record *rec;
508 struct btrfs_inode_item *item;
510 rec = active_node->current;
511 BUG_ON(rec->ino != key->objectid || rec->refs > 1);
512 if (rec->found_inode_item) {
513 rec->errors |= I_ERR_DUP_INODE_ITEM;
516 item = btrfs_item_ptr(eb, slot, struct btrfs_inode_item);
517 rec->nlink = btrfs_inode_nlink(eb, item);
518 rec->isize = btrfs_inode_size(eb, item);
519 rec->nbytes = btrfs_inode_nbytes(eb, item);
520 rec->imode = btrfs_inode_mode(eb, item);
521 if (btrfs_inode_flags(eb, item) & BTRFS_INODE_NODATASUM)
523 rec->found_inode_item = 1;
525 rec->errors |= I_ERR_NO_ORPHAN_ITEM;
526 maybe_free_inode_rec(&active_node->inode_cache, rec);
530 static struct inode_backref *get_inode_backref(struct inode_record *rec,
532 int namelen, u64 dir)
534 struct inode_backref *backref;
536 list_for_each_entry(backref, &rec->backrefs, list) {
537 if (backref->dir != dir || backref->namelen != namelen)
539 if (memcmp(name, backref->name, namelen))
544 backref = malloc(sizeof(*backref) + namelen + 1);
545 memset(backref, 0, sizeof(*backref));
547 backref->namelen = namelen;
548 memcpy(backref->name, name, namelen);
549 backref->name[namelen] = '\0';
550 list_add_tail(&backref->list, &rec->backrefs);
554 static int add_inode_backref(struct cache_tree *inode_cache,
555 u64 ino, u64 dir, u64 index,
556 const char *name, int namelen,
557 int filetype, int itemtype, int errors)
559 struct inode_record *rec;
560 struct inode_backref *backref;
562 rec = get_inode_rec(inode_cache, ino, 1);
563 backref = get_inode_backref(rec, name, namelen, dir);
565 backref->errors |= errors;
566 if (itemtype == BTRFS_DIR_INDEX_KEY) {
567 if (backref->found_dir_index)
568 backref->errors |= REF_ERR_DUP_DIR_INDEX;
569 if (backref->found_inode_ref && backref->index != index)
570 backref->errors |= REF_ERR_INDEX_UNMATCH;
571 if (backref->found_dir_item && backref->filetype != filetype)
572 backref->errors |= REF_ERR_FILETYPE_UNMATCH;
574 backref->index = index;
575 backref->filetype = filetype;
576 backref->found_dir_index = 1;
577 } else if (itemtype == BTRFS_DIR_ITEM_KEY) {
579 if (backref->found_dir_item)
580 backref->errors |= REF_ERR_DUP_DIR_ITEM;
581 if (backref->found_dir_index && backref->filetype != filetype)
582 backref->errors |= REF_ERR_FILETYPE_UNMATCH;
584 backref->filetype = filetype;
585 backref->found_dir_item = 1;
586 } else if ((itemtype == BTRFS_INODE_REF_KEY) ||
587 (itemtype == BTRFS_INODE_EXTREF_KEY)) {
588 if (backref->found_inode_ref)
589 backref->errors |= REF_ERR_DUP_INODE_REF;
590 if (backref->found_dir_index && backref->index != index)
591 backref->errors |= REF_ERR_INDEX_UNMATCH;
593 backref->ref_type = itemtype;
594 backref->index = index;
595 backref->found_inode_ref = 1;
600 maybe_free_inode_rec(inode_cache, rec);
604 static int merge_inode_recs(struct inode_record *src, struct inode_record *dst,
605 struct cache_tree *dst_cache)
607 struct inode_backref *backref;
611 list_for_each_entry(backref, &src->backrefs, list) {
612 if (backref->found_dir_index) {
613 add_inode_backref(dst_cache, dst->ino, backref->dir,
614 backref->index, backref->name,
615 backref->namelen, backref->filetype,
616 BTRFS_DIR_INDEX_KEY, backref->errors);
618 if (backref->found_dir_item) {
620 add_inode_backref(dst_cache, dst->ino,
621 backref->dir, 0, backref->name,
622 backref->namelen, backref->filetype,
623 BTRFS_DIR_ITEM_KEY, backref->errors);
625 if (backref->found_inode_ref) {
626 add_inode_backref(dst_cache, dst->ino,
627 backref->dir, backref->index,
628 backref->name, backref->namelen, 0,
629 backref->ref_type, backref->errors);
633 if (src->found_dir_item)
634 dst->found_dir_item = 1;
635 if (src->found_file_extent)
636 dst->found_file_extent = 1;
637 if (src->found_csum_item)
638 dst->found_csum_item = 1;
639 if (src->some_csum_missing)
640 dst->some_csum_missing = 1;
641 if (dst->first_extent_gap > src->first_extent_gap)
642 dst->first_extent_gap = src->first_extent_gap;
644 BUG_ON(src->found_link < dir_count);
645 dst->found_link += src->found_link - dir_count;
646 dst->found_size += src->found_size;
647 if (src->extent_start != (u64)-1) {
648 if (dst->extent_start == (u64)-1) {
649 dst->extent_start = src->extent_start;
650 dst->extent_end = src->extent_end;
652 if (dst->extent_end > src->extent_start)
653 dst->errors |= I_ERR_FILE_EXTENT_OVERLAP;
654 else if (dst->extent_end < src->extent_start &&
655 dst->extent_end < dst->first_extent_gap)
656 dst->first_extent_gap = dst->extent_end;
657 if (dst->extent_end < src->extent_end)
658 dst->extent_end = src->extent_end;
662 dst->errors |= src->errors;
663 if (src->found_inode_item) {
664 if (!dst->found_inode_item) {
665 dst->nlink = src->nlink;
666 dst->isize = src->isize;
667 dst->nbytes = src->nbytes;
668 dst->imode = src->imode;
669 dst->nodatasum = src->nodatasum;
670 dst->found_inode_item = 1;
672 dst->errors |= I_ERR_DUP_INODE_ITEM;
680 static int splice_shared_node(struct shared_node *src_node,
681 struct shared_node *dst_node)
683 struct cache_extent *cache;
684 struct ptr_node *node, *ins;
685 struct cache_tree *src, *dst;
686 struct inode_record *rec, *conflict;
691 if (--src_node->refs == 0)
693 if (src_node->current)
694 current_ino = src_node->current->ino;
696 src = &src_node->root_cache;
697 dst = &dst_node->root_cache;
699 cache = search_cache_extent(src, 0);
701 node = container_of(cache, struct ptr_node, cache);
703 cache = next_cache_extent(cache);
706 remove_cache_extent(src, &node->cache);
709 ins = malloc(sizeof(*ins));
710 ins->cache.start = node->cache.start;
711 ins->cache.size = node->cache.size;
715 ret = insert_cache_extent(dst, &ins->cache);
716 if (ret == -EEXIST) {
717 conflict = get_inode_rec(dst, rec->ino, 1);
718 merge_inode_recs(rec, conflict, dst);
720 conflict->checked = 1;
721 if (dst_node->current == conflict)
722 dst_node->current = NULL;
724 maybe_free_inode_rec(dst, conflict);
732 if (src == &src_node->root_cache) {
733 src = &src_node->inode_cache;
734 dst = &dst_node->inode_cache;
738 if (current_ino > 0 && (!dst_node->current ||
739 current_ino > dst_node->current->ino)) {
740 if (dst_node->current) {
741 dst_node->current->checked = 1;
742 maybe_free_inode_rec(dst, dst_node->current);
744 dst_node->current = get_inode_rec(dst, current_ino, 1);
749 static void free_inode_ptr(struct cache_extent *cache)
751 struct ptr_node *node;
752 struct inode_record *rec;
754 node = container_of(cache, struct ptr_node, cache);
760 FREE_EXTENT_CACHE_BASED_TREE(inode_recs, free_inode_ptr);
762 static struct shared_node *find_shared_node(struct cache_tree *shared,
765 struct cache_extent *cache;
766 struct shared_node *node;
768 cache = lookup_cache_extent(shared, bytenr, 1);
770 node = container_of(cache, struct shared_node, cache);
776 static int add_shared_node(struct cache_tree *shared, u64 bytenr, u32 refs)
779 struct shared_node *node;
781 node = calloc(1, sizeof(*node));
782 node->cache.start = bytenr;
783 node->cache.size = 1;
784 cache_tree_init(&node->root_cache);
785 cache_tree_init(&node->inode_cache);
788 ret = insert_cache_extent(shared, &node->cache);
793 static int enter_shared_node(struct btrfs_root *root, u64 bytenr, u32 refs,
794 struct walk_control *wc, int level)
796 struct shared_node *node;
797 struct shared_node *dest;
799 if (level == wc->active_node)
802 BUG_ON(wc->active_node <= level);
803 node = find_shared_node(&wc->shared, bytenr);
805 add_shared_node(&wc->shared, bytenr, refs);
806 node = find_shared_node(&wc->shared, bytenr);
807 wc->nodes[level] = node;
808 wc->active_node = level;
812 if (wc->root_level == wc->active_node &&
813 btrfs_root_refs(&root->root_item) == 0) {
814 if (--node->refs == 0) {
815 free_inode_recs_tree(&node->root_cache);
816 free_inode_recs_tree(&node->inode_cache);
817 remove_cache_extent(&wc->shared, &node->cache);
823 dest = wc->nodes[wc->active_node];
824 splice_shared_node(node, dest);
825 if (node->refs == 0) {
826 remove_cache_extent(&wc->shared, &node->cache);
832 static int leave_shared_node(struct btrfs_root *root,
833 struct walk_control *wc, int level)
835 struct shared_node *node;
836 struct shared_node *dest;
839 if (level == wc->root_level)
842 for (i = level + 1; i < BTRFS_MAX_LEVEL; i++) {
846 BUG_ON(i >= BTRFS_MAX_LEVEL);
848 node = wc->nodes[wc->active_node];
849 wc->nodes[wc->active_node] = NULL;
852 dest = wc->nodes[wc->active_node];
853 if (wc->active_node < wc->root_level ||
854 btrfs_root_refs(&root->root_item) > 0) {
855 BUG_ON(node->refs <= 1);
856 splice_shared_node(node, dest);
858 BUG_ON(node->refs < 2);
864 static int is_child_root(struct btrfs_root *root, u64 parent_root_id,
867 struct btrfs_path path;
868 struct btrfs_key key;
869 struct extent_buffer *leaf;
873 btrfs_init_path(&path);
875 key.objectid = parent_root_id;
876 key.type = BTRFS_ROOT_REF_KEY;
877 key.offset = child_root_id;
878 ret = btrfs_search_slot(NULL, root->fs_info->tree_root, &key, &path,
881 btrfs_release_path(&path);
885 key.objectid = child_root_id;
886 key.type = BTRFS_ROOT_BACKREF_KEY;
888 ret = btrfs_search_slot(NULL, root->fs_info->tree_root, &key, &path,
893 leaf = path.nodes[0];
894 if (path.slots[0] >= btrfs_header_nritems(leaf)) {
895 ret = btrfs_next_leaf(root->fs_info->tree_root, &path);
900 leaf = path.nodes[0];
903 btrfs_item_key_to_cpu(leaf, &key, path.slots[0]);
904 if (key.objectid != child_root_id ||
905 key.type != BTRFS_ROOT_BACKREF_KEY)
910 if (key.offset == parent_root_id) {
911 btrfs_release_path(&path);
918 btrfs_release_path(&path);
919 return has_parent? 0 : -1;
922 static int process_dir_item(struct btrfs_root *root,
923 struct extent_buffer *eb,
924 int slot, struct btrfs_key *key,
925 struct shared_node *active_node)
935 struct btrfs_dir_item *di;
936 struct inode_record *rec;
937 struct cache_tree *root_cache;
938 struct cache_tree *inode_cache;
939 struct btrfs_key location;
940 char namebuf[BTRFS_NAME_LEN];
942 root_cache = &active_node->root_cache;
943 inode_cache = &active_node->inode_cache;
944 rec = active_node->current;
945 rec->found_dir_item = 1;
947 di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
948 total = btrfs_item_size_nr(eb, slot);
949 while (cur < total) {
951 btrfs_dir_item_key_to_cpu(eb, di, &location);
952 name_len = btrfs_dir_name_len(eb, di);
953 data_len = btrfs_dir_data_len(eb, di);
954 filetype = btrfs_dir_type(eb, di);
956 rec->found_size += name_len;
957 if (name_len <= BTRFS_NAME_LEN) {
961 len = BTRFS_NAME_LEN;
962 error = REF_ERR_NAME_TOO_LONG;
964 read_extent_buffer(eb, namebuf, (unsigned long)(di + 1), len);
966 if (location.type == BTRFS_INODE_ITEM_KEY) {
967 add_inode_backref(inode_cache, location.objectid,
968 key->objectid, key->offset, namebuf,
969 len, filetype, key->type, error);
970 } else if (location.type == BTRFS_ROOT_ITEM_KEY) {
971 add_inode_backref(root_cache, location.objectid,
972 key->objectid, key->offset,
973 namebuf, len, filetype,
976 fprintf(stderr, "warning line %d\n", __LINE__);
979 len = sizeof(*di) + name_len + data_len;
980 di = (struct btrfs_dir_item *)((char *)di + len);
983 if (key->type == BTRFS_DIR_INDEX_KEY && nritems > 1)
984 rec->errors |= I_ERR_DUP_DIR_INDEX;
989 static int process_inode_ref(struct extent_buffer *eb,
990 int slot, struct btrfs_key *key,
991 struct shared_node *active_node)
999 struct cache_tree *inode_cache;
1000 struct btrfs_inode_ref *ref;
1001 char namebuf[BTRFS_NAME_LEN];
1003 inode_cache = &active_node->inode_cache;
1005 ref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1006 total = btrfs_item_size_nr(eb, slot);
1007 while (cur < total) {
1008 name_len = btrfs_inode_ref_name_len(eb, ref);
1009 index = btrfs_inode_ref_index(eb, ref);
1010 if (name_len <= BTRFS_NAME_LEN) {
1014 len = BTRFS_NAME_LEN;
1015 error = REF_ERR_NAME_TOO_LONG;
1017 read_extent_buffer(eb, namebuf, (unsigned long)(ref + 1), len);
1018 add_inode_backref(inode_cache, key->objectid, key->offset,
1019 index, namebuf, len, 0, key->type, error);
1021 len = sizeof(*ref) + name_len;
1022 ref = (struct btrfs_inode_ref *)((char *)ref + len);
1028 static int process_inode_extref(struct extent_buffer *eb,
1029 int slot, struct btrfs_key *key,
1030 struct shared_node *active_node)
1039 struct cache_tree *inode_cache;
1040 struct btrfs_inode_extref *extref;
1041 char namebuf[BTRFS_NAME_LEN];
1043 inode_cache = &active_node->inode_cache;
1045 extref = btrfs_item_ptr(eb, slot, struct btrfs_inode_extref);
1046 total = btrfs_item_size_nr(eb, slot);
1047 while (cur < total) {
1048 name_len = btrfs_inode_extref_name_len(eb, extref);
1049 index = btrfs_inode_extref_index(eb, extref);
1050 parent = btrfs_inode_extref_parent(eb, extref);
1051 if (name_len <= BTRFS_NAME_LEN) {
1055 len = BTRFS_NAME_LEN;
1056 error = REF_ERR_NAME_TOO_LONG;
1058 read_extent_buffer(eb, namebuf,
1059 (unsigned long)(extref + 1), len);
1060 add_inode_backref(inode_cache, key->objectid, parent,
1061 index, namebuf, len, 0, key->type, error);
1063 len = sizeof(*extref) + name_len;
1064 extref = (struct btrfs_inode_extref *)((char *)extref + len);
1071 static u64 count_csum_range(struct btrfs_root *root, u64 start, u64 len)
1073 struct btrfs_key key;
1074 struct btrfs_path path;
1075 struct extent_buffer *leaf;
1080 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
1082 btrfs_init_path(&path);
1084 key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
1086 key.type = BTRFS_EXTENT_CSUM_KEY;
1088 ret = btrfs_search_slot(NULL, root->fs_info->csum_root,
1091 if (ret > 0 && path.slots[0] > 0) {
1092 leaf = path.nodes[0];
1093 btrfs_item_key_to_cpu(leaf, &key, path.slots[0] - 1);
1094 if (key.objectid == BTRFS_EXTENT_CSUM_OBJECTID &&
1095 key.type == BTRFS_EXTENT_CSUM_KEY)
1100 leaf = path.nodes[0];
1101 if (path.slots[0] >= btrfs_header_nritems(leaf)) {
1102 ret = btrfs_next_leaf(root->fs_info->csum_root, &path);
1106 leaf = path.nodes[0];
1109 btrfs_item_key_to_cpu(leaf, &key, path.slots[0]);
1110 if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
1111 key.type != BTRFS_EXTENT_CSUM_KEY)
1114 btrfs_item_key_to_cpu(leaf, &key, path.slots[0]);
1115 if (key.offset >= start + len)
1118 if (key.offset > start)
1121 size = btrfs_item_size_nr(leaf, path.slots[0]);
1122 csum_end = key.offset + (size / csum_size) * root->sectorsize;
1123 if (csum_end > start) {
1124 size = min(csum_end - start, len);
1132 btrfs_release_path(&path);
1136 static int process_file_extent(struct btrfs_root *root,
1137 struct extent_buffer *eb,
1138 int slot, struct btrfs_key *key,
1139 struct shared_node *active_node)
1141 struct inode_record *rec;
1142 struct btrfs_file_extent_item *fi;
1144 u64 disk_bytenr = 0;
1145 u64 extent_offset = 0;
1146 u64 mask = root->sectorsize - 1;
1149 rec = active_node->current;
1150 BUG_ON(rec->ino != key->objectid || rec->refs > 1);
1151 rec->found_file_extent = 1;
1153 if (rec->extent_start == (u64)-1) {
1154 rec->extent_start = key->offset;
1155 rec->extent_end = key->offset;
1158 if (rec->extent_end > key->offset)
1159 rec->errors |= I_ERR_FILE_EXTENT_OVERLAP;
1160 else if (rec->extent_end < key->offset &&
1161 rec->extent_end < rec->first_extent_gap)
1162 rec->first_extent_gap = rec->extent_end;
1164 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
1165 extent_type = btrfs_file_extent_type(eb, fi);
1167 if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1168 num_bytes = btrfs_file_extent_inline_len(eb, fi);
1170 rec->errors |= I_ERR_BAD_FILE_EXTENT;
1171 rec->found_size += num_bytes;
1172 num_bytes = (num_bytes + mask) & ~mask;
1173 } else if (extent_type == BTRFS_FILE_EXTENT_REG ||
1174 extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
1175 num_bytes = btrfs_file_extent_num_bytes(eb, fi);
1176 disk_bytenr = btrfs_file_extent_disk_bytenr(eb, fi);
1177 extent_offset = btrfs_file_extent_offset(eb, fi);
1178 if (num_bytes == 0 || (num_bytes & mask))
1179 rec->errors |= I_ERR_BAD_FILE_EXTENT;
1180 if (num_bytes + extent_offset >
1181 btrfs_file_extent_ram_bytes(eb, fi))
1182 rec->errors |= I_ERR_BAD_FILE_EXTENT;
1183 if (extent_type == BTRFS_FILE_EXTENT_PREALLOC &&
1184 (btrfs_file_extent_compression(eb, fi) ||
1185 btrfs_file_extent_encryption(eb, fi) ||
1186 btrfs_file_extent_other_encoding(eb, fi)))
1187 rec->errors |= I_ERR_BAD_FILE_EXTENT;
1188 if (disk_bytenr > 0)
1189 rec->found_size += num_bytes;
1191 rec->errors |= I_ERR_BAD_FILE_EXTENT;
1193 rec->extent_end = key->offset + num_bytes;
1195 if (disk_bytenr > 0) {
1197 if (btrfs_file_extent_compression(eb, fi))
1198 num_bytes = btrfs_file_extent_disk_num_bytes(eb, fi);
1200 disk_bytenr += extent_offset;
1202 found = count_csum_range(root, disk_bytenr, num_bytes);
1203 if (extent_type == BTRFS_FILE_EXTENT_REG) {
1205 rec->found_csum_item = 1;
1206 if (found < num_bytes)
1207 rec->some_csum_missing = 1;
1208 } else if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
1210 rec->errors |= I_ERR_ODD_CSUM_ITEM;
1216 static int process_one_leaf(struct btrfs_root *root, struct extent_buffer *eb,
1217 struct walk_control *wc)
1219 struct btrfs_key key;
1224 struct cache_tree *inode_cache;
1225 struct shared_node *active_node;
1227 if (wc->root_level == wc->active_node &&
1228 btrfs_root_refs(&root->root_item) == 0)
1231 active_node = wc->nodes[wc->active_node];
1232 inode_cache = &active_node->inode_cache;
1233 nritems = btrfs_header_nritems(eb);
1234 for (i = 0; i < nritems; i++) {
1235 btrfs_item_key_to_cpu(eb, &key, i);
1237 if (key.objectid == BTRFS_FREE_SPACE_OBJECTID)
1239 if (key.type == BTRFS_ORPHAN_ITEM_KEY)
1242 if (active_node->current == NULL ||
1243 active_node->current->ino < key.objectid) {
1244 if (active_node->current) {
1245 active_node->current->checked = 1;
1246 maybe_free_inode_rec(inode_cache,
1247 active_node->current);
1249 active_node->current = get_inode_rec(inode_cache,
1253 case BTRFS_DIR_ITEM_KEY:
1254 case BTRFS_DIR_INDEX_KEY:
1255 ret = process_dir_item(root, eb, i, &key, active_node);
1257 case BTRFS_INODE_REF_KEY:
1258 ret = process_inode_ref(eb, i, &key, active_node);
1260 case BTRFS_INODE_EXTREF_KEY:
1261 ret = process_inode_extref(eb, i, &key, active_node);
1263 case BTRFS_INODE_ITEM_KEY:
1264 ret = process_inode_item(eb, i, &key, active_node);
1266 case BTRFS_EXTENT_DATA_KEY:
1267 ret = process_file_extent(root, eb, i, &key,
1279 static void reada_walk_down(struct btrfs_root *root,
1280 struct extent_buffer *node, int slot)
1290 level = btrfs_header_level(node);
1294 nritems = btrfs_header_nritems(node);
1295 blocksize = btrfs_level_size(root, level - 1);
1296 for (i = slot; i < nritems; i++) {
1297 bytenr = btrfs_node_blockptr(node, i);
1298 ptr_gen = btrfs_node_ptr_generation(node, i);
1299 ret = readahead_tree_block(root, bytenr, blocksize, ptr_gen);
1305 static int walk_down_tree(struct btrfs_root *root, struct btrfs_path *path,
1306 struct walk_control *wc, int *level)
1310 struct extent_buffer *next;
1311 struct extent_buffer *cur;
1316 WARN_ON(*level < 0);
1317 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1318 ret = btrfs_lookup_extent_info(NULL, root,
1319 path->nodes[*level]->start,
1320 *level, 1, &refs, NULL);
1327 ret = enter_shared_node(root, path->nodes[*level]->start,
1335 while (*level >= 0) {
1336 WARN_ON(*level < 0);
1337 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1338 cur = path->nodes[*level];
1340 if (btrfs_header_level(cur) != *level)
1343 if (path->slots[*level] >= btrfs_header_nritems(cur))
1346 ret = process_one_leaf(root, cur, wc);
1349 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
1350 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
1351 blocksize = btrfs_level_size(root, *level - 1);
1352 ret = btrfs_lookup_extent_info(NULL, root, bytenr, *level - 1,
1358 ret = enter_shared_node(root, bytenr, refs,
1361 path->slots[*level]++;
1366 next = btrfs_find_tree_block(root, bytenr, blocksize);
1367 if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
1368 free_extent_buffer(next);
1369 reada_walk_down(root, cur, path->slots[*level]);
1370 next = read_tree_block(root, bytenr, blocksize,
1378 *level = *level - 1;
1379 free_extent_buffer(path->nodes[*level]);
1380 path->nodes[*level] = next;
1381 path->slots[*level] = 0;
1384 path->slots[*level] = btrfs_header_nritems(path->nodes[*level]);
1388 static int walk_up_tree(struct btrfs_root *root, struct btrfs_path *path,
1389 struct walk_control *wc, int *level)
1392 struct extent_buffer *leaf;
1394 for (i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
1395 leaf = path->nodes[i];
1396 if (path->slots[i] + 1 < btrfs_header_nritems(leaf)) {
1401 free_extent_buffer(path->nodes[*level]);
1402 path->nodes[*level] = NULL;
1403 BUG_ON(*level > wc->active_node);
1404 if (*level == wc->active_node)
1405 leave_shared_node(root, wc, *level);
1412 static int check_root_dir(struct inode_record *rec)
1414 struct inode_backref *backref;
1417 if (!rec->found_inode_item || rec->errors)
1419 if (rec->nlink != 1 || rec->found_link != 0)
1421 if (list_empty(&rec->backrefs))
1423 backref = list_entry(rec->backrefs.next, struct inode_backref, list);
1424 if (!backref->found_inode_ref)
1426 if (backref->index != 0 || backref->namelen != 2 ||
1427 memcmp(backref->name, "..", 2))
1429 if (backref->found_dir_index || backref->found_dir_item)
1436 static int repair_inode_isize(struct btrfs_trans_handle *trans,
1437 struct btrfs_root *root, struct btrfs_path *path,
1438 struct inode_record *rec)
1440 struct btrfs_inode_item *ei;
1441 struct btrfs_key key;
1444 key.objectid = rec->ino;
1445 key.type = BTRFS_INODE_ITEM_KEY;
1446 key.offset = (u64)-1;
1448 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1452 if (!path->slots[0]) {
1459 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1460 if (key.objectid != rec->ino) {
1465 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
1466 struct btrfs_inode_item);
1467 btrfs_set_inode_size(path->nodes[0], ei, rec->found_size);
1468 btrfs_mark_buffer_dirty(path->nodes[0]);
1469 rec->errors &= ~I_ERR_DIR_ISIZE_WRONG;
1470 printf("reset isize for dir %Lu root %Lu\n", rec->ino,
1471 root->root_key.objectid);
1473 btrfs_release_path(path);
1477 static int repair_inode_orphan_item(struct btrfs_trans_handle *trans,
1478 struct btrfs_root *root,
1479 struct btrfs_path *path,
1480 struct inode_record *rec)
1482 struct btrfs_key key;
1485 key.objectid = BTRFS_ORPHAN_OBJECTID;
1486 key.type = BTRFS_ORPHAN_ITEM_KEY;
1487 key.offset = rec->ino;
1489 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1490 btrfs_release_path(path);
1492 rec->errors &= ~I_ERR_NO_ORPHAN_ITEM;
1496 static int try_repair_inode(struct btrfs_root *root, struct inode_record *rec)
1498 struct btrfs_trans_handle *trans;
1499 struct btrfs_path *path;
1502 /* So far we just fix dir isize wrong */
1503 if (!(rec->errors & (I_ERR_DIR_ISIZE_WRONG | I_ERR_NO_ORPHAN_ITEM)))
1506 path = btrfs_alloc_path();
1510 trans = btrfs_start_transaction(root, 1);
1511 if (IS_ERR(trans)) {
1512 btrfs_free_path(path);
1513 return PTR_ERR(trans);
1516 if (rec->errors & I_ERR_DIR_ISIZE_WRONG)
1517 ret = repair_inode_isize(trans, root, path, rec);
1518 if (!ret && rec->errors & I_ERR_NO_ORPHAN_ITEM)
1519 ret = repair_inode_orphan_item(trans, root, path, rec);
1520 btrfs_commit_transaction(trans, root);
1521 btrfs_free_path(path);
1525 static int check_inode_recs(struct btrfs_root *root,
1526 struct cache_tree *inode_cache)
1528 struct cache_extent *cache;
1529 struct ptr_node *node;
1530 struct inode_record *rec;
1531 struct inode_backref *backref;
1534 u64 root_dirid = btrfs_root_dirid(&root->root_item);
1536 if (btrfs_root_refs(&root->root_item) == 0) {
1537 if (!cache_tree_empty(inode_cache))
1538 fprintf(stderr, "warning line %d\n", __LINE__);
1542 rec = get_inode_rec(inode_cache, root_dirid, 0);
1544 ret = check_root_dir(rec);
1546 fprintf(stderr, "root %llu root dir %llu error\n",
1547 (unsigned long long)root->root_key.objectid,
1548 (unsigned long long)root_dirid);
1552 fprintf(stderr, "root %llu root dir %llu not found\n",
1553 (unsigned long long)root->root_key.objectid,
1554 (unsigned long long)root_dirid);
1558 cache = search_cache_extent(inode_cache, 0);
1561 node = container_of(cache, struct ptr_node, cache);
1563 remove_cache_extent(inode_cache, &node->cache);
1565 if (rec->ino == root_dirid ||
1566 rec->ino == BTRFS_ORPHAN_OBJECTID) {
1567 free_inode_rec(rec);
1571 if (rec->errors & I_ERR_NO_ORPHAN_ITEM) {
1572 ret = check_orphan_item(root, rec->ino);
1574 rec->errors &= ~I_ERR_NO_ORPHAN_ITEM;
1575 if (can_free_inode_rec(rec)) {
1576 free_inode_rec(rec);
1582 ret = try_repair_inode(root, rec);
1583 if (ret == 0 && can_free_inode_rec(rec)) {
1584 free_inode_rec(rec);
1591 if (!rec->found_inode_item)
1592 rec->errors |= I_ERR_NO_INODE_ITEM;
1593 if (rec->found_link != rec->nlink)
1594 rec->errors |= I_ERR_LINK_COUNT_WRONG;
1595 fprintf(stderr, "root %llu inode %llu errors %x",
1596 (unsigned long long) root->root_key.objectid,
1597 (unsigned long long) rec->ino, rec->errors);
1598 print_inode_error(rec->errors);
1599 list_for_each_entry(backref, &rec->backrefs, list) {
1600 if (!backref->found_dir_item)
1601 backref->errors |= REF_ERR_NO_DIR_ITEM;
1602 if (!backref->found_dir_index)
1603 backref->errors |= REF_ERR_NO_DIR_INDEX;
1604 if (!backref->found_inode_ref)
1605 backref->errors |= REF_ERR_NO_INODE_REF;
1606 fprintf(stderr, "\tunresolved ref dir %llu index %llu"
1607 " namelen %u name %s filetype %d error %x",
1608 (unsigned long long)backref->dir,
1609 (unsigned long long)backref->index,
1610 backref->namelen, backref->name,
1611 backref->filetype, backref->errors);
1612 print_ref_error(backref->errors);
1614 free_inode_rec(rec);
1616 return (error > 0) ? -1 : 0;
1619 static struct root_record *get_root_rec(struct cache_tree *root_cache,
1622 struct cache_extent *cache;
1623 struct root_record *rec = NULL;
1626 cache = lookup_cache_extent(root_cache, objectid, 1);
1628 rec = container_of(cache, struct root_record, cache);
1630 rec = calloc(1, sizeof(*rec));
1631 rec->objectid = objectid;
1632 INIT_LIST_HEAD(&rec->backrefs);
1633 rec->cache.start = objectid;
1634 rec->cache.size = 1;
1636 ret = insert_cache_extent(root_cache, &rec->cache);
1642 static struct root_backref *get_root_backref(struct root_record *rec,
1643 u64 ref_root, u64 dir, u64 index,
1644 const char *name, int namelen)
1646 struct root_backref *backref;
1648 list_for_each_entry(backref, &rec->backrefs, list) {
1649 if (backref->ref_root != ref_root || backref->dir != dir ||
1650 backref->namelen != namelen)
1652 if (memcmp(name, backref->name, namelen))
1657 backref = malloc(sizeof(*backref) + namelen + 1);
1658 memset(backref, 0, sizeof(*backref));
1659 backref->ref_root = ref_root;
1661 backref->index = index;
1662 backref->namelen = namelen;
1663 memcpy(backref->name, name, namelen);
1664 backref->name[namelen] = '\0';
1665 list_add_tail(&backref->list, &rec->backrefs);
1669 static void free_root_record(struct cache_extent *cache)
1671 struct root_record *rec;
1672 struct root_backref *backref;
1674 rec = container_of(cache, struct root_record, cache);
1675 while (!list_empty(&rec->backrefs)) {
1676 backref = list_entry(rec->backrefs.next,
1677 struct root_backref, list);
1678 list_del(&backref->list);
1685 FREE_EXTENT_CACHE_BASED_TREE(root_recs, free_root_record);
1687 static int add_root_backref(struct cache_tree *root_cache,
1688 u64 root_id, u64 ref_root, u64 dir, u64 index,
1689 const char *name, int namelen,
1690 int item_type, int errors)
1692 struct root_record *rec;
1693 struct root_backref *backref;
1695 rec = get_root_rec(root_cache, root_id);
1696 backref = get_root_backref(rec, ref_root, dir, index, name, namelen);
1698 backref->errors |= errors;
1700 if (item_type != BTRFS_DIR_ITEM_KEY) {
1701 if (backref->found_dir_index || backref->found_back_ref ||
1702 backref->found_forward_ref) {
1703 if (backref->index != index)
1704 backref->errors |= REF_ERR_INDEX_UNMATCH;
1706 backref->index = index;
1710 if (item_type == BTRFS_DIR_ITEM_KEY) {
1711 if (backref->found_forward_ref)
1713 backref->found_dir_item = 1;
1714 } else if (item_type == BTRFS_DIR_INDEX_KEY) {
1715 backref->found_dir_index = 1;
1716 } else if (item_type == BTRFS_ROOT_REF_KEY) {
1717 if (backref->found_forward_ref)
1718 backref->errors |= REF_ERR_DUP_ROOT_REF;
1719 else if (backref->found_dir_item)
1721 backref->found_forward_ref = 1;
1722 } else if (item_type == BTRFS_ROOT_BACKREF_KEY) {
1723 if (backref->found_back_ref)
1724 backref->errors |= REF_ERR_DUP_ROOT_BACKREF;
1725 backref->found_back_ref = 1;
1730 if (backref->found_forward_ref && backref->found_dir_item)
1731 backref->reachable = 1;
1735 static int merge_root_recs(struct btrfs_root *root,
1736 struct cache_tree *src_cache,
1737 struct cache_tree *dst_cache)
1739 struct cache_extent *cache;
1740 struct ptr_node *node;
1741 struct inode_record *rec;
1742 struct inode_backref *backref;
1744 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
1745 free_inode_recs_tree(src_cache);
1750 cache = search_cache_extent(src_cache, 0);
1753 node = container_of(cache, struct ptr_node, cache);
1755 remove_cache_extent(src_cache, &node->cache);
1758 if (!is_child_root(root, root->objectid, rec->ino))
1761 list_for_each_entry(backref, &rec->backrefs, list) {
1762 BUG_ON(backref->found_inode_ref);
1763 if (backref->found_dir_item)
1764 add_root_backref(dst_cache, rec->ino,
1765 root->root_key.objectid, backref->dir,
1766 backref->index, backref->name,
1767 backref->namelen, BTRFS_DIR_ITEM_KEY,
1769 if (backref->found_dir_index)
1770 add_root_backref(dst_cache, rec->ino,
1771 root->root_key.objectid, backref->dir,
1772 backref->index, backref->name,
1773 backref->namelen, BTRFS_DIR_INDEX_KEY,
1777 free_inode_rec(rec);
1782 static int check_root_refs(struct btrfs_root *root,
1783 struct cache_tree *root_cache)
1785 struct root_record *rec;
1786 struct root_record *ref_root;
1787 struct root_backref *backref;
1788 struct cache_extent *cache;
1794 rec = get_root_rec(root_cache, BTRFS_FS_TREE_OBJECTID);
1797 /* fixme: this can not detect circular references */
1800 cache = search_cache_extent(root_cache, 0);
1804 rec = container_of(cache, struct root_record, cache);
1805 cache = next_cache_extent(cache);
1807 if (rec->found_ref == 0)
1810 list_for_each_entry(backref, &rec->backrefs, list) {
1811 if (!backref->reachable)
1814 ref_root = get_root_rec(root_cache,
1816 if (ref_root->found_ref > 0)
1819 backref->reachable = 0;
1821 if (rec->found_ref == 0)
1827 cache = search_cache_extent(root_cache, 0);
1831 rec = container_of(cache, struct root_record, cache);
1832 cache = next_cache_extent(cache);
1834 if (rec->found_ref == 0 &&
1835 rec->objectid >= BTRFS_FIRST_FREE_OBJECTID &&
1836 rec->objectid <= BTRFS_LAST_FREE_OBJECTID) {
1837 ret = check_orphan_item(root->fs_info->tree_root,
1843 * If we don't have a root item then we likely just have
1844 * a dir item in a snapshot for this root but no actual
1845 * ref key or anything so it's meaningless.
1847 if (!rec->found_root_item)
1850 fprintf(stderr, "fs tree %llu not referenced\n",
1851 (unsigned long long)rec->objectid);
1855 if (rec->found_ref > 0 && !rec->found_root_item)
1857 list_for_each_entry(backref, &rec->backrefs, list) {
1858 if (!backref->found_dir_item)
1859 backref->errors |= REF_ERR_NO_DIR_ITEM;
1860 if (!backref->found_dir_index)
1861 backref->errors |= REF_ERR_NO_DIR_INDEX;
1862 if (!backref->found_back_ref)
1863 backref->errors |= REF_ERR_NO_ROOT_BACKREF;
1864 if (!backref->found_forward_ref)
1865 backref->errors |= REF_ERR_NO_ROOT_REF;
1866 if (backref->reachable && backref->errors)
1873 fprintf(stderr, "fs tree %llu refs %u %s\n",
1874 (unsigned long long)rec->objectid, rec->found_ref,
1875 rec->found_root_item ? "" : "not found");
1877 list_for_each_entry(backref, &rec->backrefs, list) {
1878 if (!backref->reachable)
1880 if (!backref->errors && rec->found_root_item)
1882 fprintf(stderr, "\tunresolved ref root %llu dir %llu"
1883 " index %llu namelen %u name %s error %x\n",
1884 (unsigned long long)backref->ref_root,
1885 (unsigned long long)backref->dir,
1886 (unsigned long long)backref->index,
1887 backref->namelen, backref->name,
1891 return errors > 0 ? 1 : 0;
1894 static int process_root_ref(struct extent_buffer *eb, int slot,
1895 struct btrfs_key *key,
1896 struct cache_tree *root_cache)
1902 struct btrfs_root_ref *ref;
1903 char namebuf[BTRFS_NAME_LEN];
1906 ref = btrfs_item_ptr(eb, slot, struct btrfs_root_ref);
1908 dirid = btrfs_root_ref_dirid(eb, ref);
1909 index = btrfs_root_ref_sequence(eb, ref);
1910 name_len = btrfs_root_ref_name_len(eb, ref);
1912 if (name_len <= BTRFS_NAME_LEN) {
1916 len = BTRFS_NAME_LEN;
1917 error = REF_ERR_NAME_TOO_LONG;
1919 read_extent_buffer(eb, namebuf, (unsigned long)(ref + 1), len);
1921 if (key->type == BTRFS_ROOT_REF_KEY) {
1922 add_root_backref(root_cache, key->offset, key->objectid, dirid,
1923 index, namebuf, len, key->type, error);
1925 add_root_backref(root_cache, key->objectid, key->offset, dirid,
1926 index, namebuf, len, key->type, error);
1931 static int check_fs_root(struct btrfs_root *root,
1932 struct cache_tree *root_cache,
1933 struct walk_control *wc)
1938 struct btrfs_path path;
1939 struct shared_node root_node;
1940 struct root_record *rec;
1941 struct btrfs_root_item *root_item = &root->root_item;
1943 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
1944 rec = get_root_rec(root_cache, root->root_key.objectid);
1945 if (btrfs_root_refs(root_item) > 0)
1946 rec->found_root_item = 1;
1949 btrfs_init_path(&path);
1950 memset(&root_node, 0, sizeof(root_node));
1951 cache_tree_init(&root_node.root_cache);
1952 cache_tree_init(&root_node.inode_cache);
1954 level = btrfs_header_level(root->node);
1955 memset(wc->nodes, 0, sizeof(wc->nodes));
1956 wc->nodes[level] = &root_node;
1957 wc->active_node = level;
1958 wc->root_level = level;
1960 if (btrfs_root_refs(root_item) > 0 ||
1961 btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
1962 path.nodes[level] = root->node;
1963 extent_buffer_get(root->node);
1964 path.slots[level] = 0;
1966 struct btrfs_key key;
1967 struct btrfs_disk_key found_key;
1969 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
1970 level = root_item->drop_level;
1971 path.lowest_level = level;
1972 wret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
1974 btrfs_node_key(path.nodes[level], &found_key,
1976 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
1977 sizeof(found_key)));
1981 wret = walk_down_tree(root, &path, wc, &level);
1987 wret = walk_up_tree(root, &path, wc, &level);
1993 btrfs_release_path(&path);
1995 merge_root_recs(root, &root_node.root_cache, root_cache);
1997 if (root_node.current) {
1998 root_node.current->checked = 1;
1999 maybe_free_inode_rec(&root_node.inode_cache,
2003 ret = check_inode_recs(root, &root_node.inode_cache);
2007 static int fs_root_objectid(u64 objectid)
2009 if (objectid == BTRFS_FS_TREE_OBJECTID ||
2010 objectid == BTRFS_TREE_RELOC_OBJECTID ||
2011 objectid == BTRFS_DATA_RELOC_TREE_OBJECTID ||
2012 (objectid >= BTRFS_FIRST_FREE_OBJECTID &&
2013 objectid <= BTRFS_LAST_FREE_OBJECTID))
2018 static int check_fs_roots(struct btrfs_root *root,
2019 struct cache_tree *root_cache)
2021 struct btrfs_path path;
2022 struct btrfs_key key;
2023 struct walk_control wc;
2024 struct extent_buffer *leaf;
2025 struct btrfs_root *tmp_root;
2026 struct btrfs_root *tree_root = root->fs_info->tree_root;
2031 * Just in case we made any changes to the extent tree that weren't
2032 * reflected into the free space cache yet.
2035 reset_cached_block_groups(root->fs_info);
2036 memset(&wc, 0, sizeof(wc));
2037 cache_tree_init(&wc.shared);
2038 btrfs_init_path(&path);
2042 key.type = BTRFS_ROOT_ITEM_KEY;
2043 ret = btrfs_search_slot(NULL, tree_root, &key, &path, 0, 0);
2046 leaf = path.nodes[0];
2047 if (path.slots[0] >= btrfs_header_nritems(leaf)) {
2048 ret = btrfs_next_leaf(tree_root, &path);
2051 leaf = path.nodes[0];
2053 btrfs_item_key_to_cpu(leaf, &key, path.slots[0]);
2054 if (key.type == BTRFS_ROOT_ITEM_KEY &&
2055 fs_root_objectid(key.objectid)) {
2056 key.offset = (u64)-1;
2057 tmp_root = btrfs_read_fs_root(root->fs_info, &key);
2058 if (IS_ERR(tmp_root)) {
2062 ret = check_fs_root(tmp_root, root_cache, &wc);
2065 } else if (key.type == BTRFS_ROOT_REF_KEY ||
2066 key.type == BTRFS_ROOT_BACKREF_KEY) {
2067 process_root_ref(leaf, path.slots[0], &key,
2073 btrfs_release_path(&path);
2075 if (!cache_tree_empty(&wc.shared))
2076 fprintf(stderr, "warning line %d\n", __LINE__);
2081 static int all_backpointers_checked(struct extent_record *rec, int print_errs)
2083 struct list_head *cur = rec->backrefs.next;
2084 struct extent_backref *back;
2085 struct tree_backref *tback;
2086 struct data_backref *dback;
2090 while(cur != &rec->backrefs) {
2091 back = list_entry(cur, struct extent_backref, list);
2093 if (!back->found_extent_tree) {
2097 if (back->is_data) {
2098 dback = (struct data_backref *)back;
2099 fprintf(stderr, "Backref %llu %s %llu"
2100 " owner %llu offset %llu num_refs %lu"
2101 " not found in extent tree\n",
2102 (unsigned long long)rec->start,
2103 back->full_backref ?
2105 back->full_backref ?
2106 (unsigned long long)dback->parent:
2107 (unsigned long long)dback->root,
2108 (unsigned long long)dback->owner,
2109 (unsigned long long)dback->offset,
2110 (unsigned long)dback->num_refs);
2112 tback = (struct tree_backref *)back;
2113 fprintf(stderr, "Backref %llu parent %llu"
2114 " root %llu not found in extent tree\n",
2115 (unsigned long long)rec->start,
2116 (unsigned long long)tback->parent,
2117 (unsigned long long)tback->root);
2120 if (!back->is_data && !back->found_ref) {
2124 tback = (struct tree_backref *)back;
2125 fprintf(stderr, "Backref %llu %s %llu not referenced back %p\n",
2126 (unsigned long long)rec->start,
2127 back->full_backref ? "parent" : "root",
2128 back->full_backref ?
2129 (unsigned long long)tback->parent :
2130 (unsigned long long)tback->root, back);
2132 if (back->is_data) {
2133 dback = (struct data_backref *)back;
2134 if (dback->found_ref != dback->num_refs) {
2138 fprintf(stderr, "Incorrect local backref count"
2139 " on %llu %s %llu owner %llu"
2140 " offset %llu found %u wanted %u back %p\n",
2141 (unsigned long long)rec->start,
2142 back->full_backref ?
2144 back->full_backref ?
2145 (unsigned long long)dback->parent:
2146 (unsigned long long)dback->root,
2147 (unsigned long long)dback->owner,
2148 (unsigned long long)dback->offset,
2149 dback->found_ref, dback->num_refs, back);
2151 if (dback->disk_bytenr != rec->start) {
2155 fprintf(stderr, "Backref disk bytenr does not"
2156 " match extent record, bytenr=%llu, "
2157 "ref bytenr=%llu\n",
2158 (unsigned long long)rec->start,
2159 (unsigned long long)dback->disk_bytenr);
2162 if (dback->bytes != rec->nr) {
2166 fprintf(stderr, "Backref bytes do not match "
2167 "extent backref, bytenr=%llu, ref "
2168 "bytes=%llu, backref bytes=%llu\n",
2169 (unsigned long long)rec->start,
2170 (unsigned long long)rec->nr,
2171 (unsigned long long)dback->bytes);
2174 if (!back->is_data) {
2177 dback = (struct data_backref *)back;
2178 found += dback->found_ref;
2181 if (found != rec->refs) {
2185 fprintf(stderr, "Incorrect global backref count "
2186 "on %llu found %llu wanted %llu\n",
2187 (unsigned long long)rec->start,
2188 (unsigned long long)found,
2189 (unsigned long long)rec->refs);
2195 static int free_all_extent_backrefs(struct extent_record *rec)
2197 struct extent_backref *back;
2198 struct list_head *cur;
2199 while (!list_empty(&rec->backrefs)) {
2200 cur = rec->backrefs.next;
2201 back = list_entry(cur, struct extent_backref, list);
2208 static void free_extent_record_cache(struct btrfs_fs_info *fs_info,
2209 struct cache_tree *extent_cache)
2211 struct cache_extent *cache;
2212 struct extent_record *rec;
2215 cache = first_cache_extent(extent_cache);
2218 rec = container_of(cache, struct extent_record, cache);
2219 btrfs_unpin_extent(fs_info, rec->start, rec->max_size);
2220 remove_cache_extent(extent_cache, cache);
2221 free_all_extent_backrefs(rec);
2226 static int maybe_free_extent_rec(struct cache_tree *extent_cache,
2227 struct extent_record *rec)
2229 if (rec->content_checked && rec->owner_ref_checked &&
2230 rec->extent_item_refs == rec->refs && rec->refs > 0 &&
2231 rec->num_duplicates == 0 && !all_backpointers_checked(rec, 0)) {
2232 remove_cache_extent(extent_cache, &rec->cache);
2233 free_all_extent_backrefs(rec);
2234 list_del_init(&rec->list);
2240 static int check_owner_ref(struct btrfs_root *root,
2241 struct extent_record *rec,
2242 struct extent_buffer *buf)
2244 struct extent_backref *node;
2245 struct tree_backref *back;
2246 struct btrfs_root *ref_root;
2247 struct btrfs_key key;
2248 struct btrfs_path path;
2249 struct extent_buffer *parent;
2254 list_for_each_entry(node, &rec->backrefs, list) {
2257 if (!node->found_ref)
2259 if (node->full_backref)
2261 back = (struct tree_backref *)node;
2262 if (btrfs_header_owner(buf) == back->root)
2265 BUG_ON(rec->is_root);
2267 /* try to find the block by search corresponding fs tree */
2268 key.objectid = btrfs_header_owner(buf);
2269 key.type = BTRFS_ROOT_ITEM_KEY;
2270 key.offset = (u64)-1;
2272 ref_root = btrfs_read_fs_root(root->fs_info, &key);
2273 if (IS_ERR(ref_root))
2276 level = btrfs_header_level(buf);
2278 btrfs_item_key_to_cpu(buf, &key, 0);
2280 btrfs_node_key_to_cpu(buf, &key, 0);
2282 btrfs_init_path(&path);
2283 path.lowest_level = level + 1;
2284 ret = btrfs_search_slot(NULL, ref_root, &key, &path, 0, 0);
2288 parent = path.nodes[level + 1];
2289 if (parent && buf->start == btrfs_node_blockptr(parent,
2290 path.slots[level + 1]))
2293 btrfs_release_path(&path);
2294 return found ? 0 : 1;
2297 static int is_extent_tree_record(struct extent_record *rec)
2299 struct list_head *cur = rec->backrefs.next;
2300 struct extent_backref *node;
2301 struct tree_backref *back;
2304 while(cur != &rec->backrefs) {
2305 node = list_entry(cur, struct extent_backref, list);
2309 back = (struct tree_backref *)node;
2310 if (node->full_backref)
2312 if (back->root == BTRFS_EXTENT_TREE_OBJECTID)
2319 static int record_bad_block_io(struct btrfs_fs_info *info,
2320 struct cache_tree *extent_cache,
2323 struct extent_record *rec;
2324 struct cache_extent *cache;
2325 struct btrfs_key key;
2327 cache = lookup_cache_extent(extent_cache, start, len);
2331 rec = container_of(cache, struct extent_record, cache);
2332 if (!is_extent_tree_record(rec))
2335 btrfs_disk_key_to_cpu(&key, &rec->parent_key);
2336 return btrfs_add_corrupt_extent_record(info, &key, start, len, 0);
2339 static int swap_values(struct btrfs_root *root, struct btrfs_path *path,
2340 struct extent_buffer *buf, int slot)
2342 if (btrfs_header_level(buf)) {
2343 struct btrfs_key_ptr ptr1, ptr2;
2345 read_extent_buffer(buf, &ptr1, btrfs_node_key_ptr_offset(slot),
2346 sizeof(struct btrfs_key_ptr));
2347 read_extent_buffer(buf, &ptr2,
2348 btrfs_node_key_ptr_offset(slot + 1),
2349 sizeof(struct btrfs_key_ptr));
2350 write_extent_buffer(buf, &ptr1,
2351 btrfs_node_key_ptr_offset(slot + 1),
2352 sizeof(struct btrfs_key_ptr));
2353 write_extent_buffer(buf, &ptr2,
2354 btrfs_node_key_ptr_offset(slot),
2355 sizeof(struct btrfs_key_ptr));
2357 struct btrfs_disk_key key;
2358 btrfs_node_key(buf, &key, 0);
2359 btrfs_fixup_low_keys(root, path, &key,
2360 btrfs_header_level(buf) + 1);
2363 struct btrfs_item *item1, *item2;
2364 struct btrfs_key k1, k2;
2365 char *item1_data, *item2_data;
2366 u32 item1_offset, item2_offset, item1_size, item2_size;
2368 item1 = btrfs_item_nr(slot);
2369 item2 = btrfs_item_nr(slot + 1);
2370 btrfs_item_key_to_cpu(buf, &k1, slot);
2371 btrfs_item_key_to_cpu(buf, &k2, slot + 1);
2372 item1_offset = btrfs_item_offset(buf, item1);
2373 item2_offset = btrfs_item_offset(buf, item2);
2374 item1_size = btrfs_item_size(buf, item1);
2375 item2_size = btrfs_item_size(buf, item2);
2377 item1_data = malloc(item1_size);
2380 item2_data = malloc(item2_size);
2386 read_extent_buffer(buf, item1_data, item1_offset, item1_size);
2387 read_extent_buffer(buf, item2_data, item2_offset, item2_size);
2389 write_extent_buffer(buf, item1_data, item2_offset, item2_size);
2390 write_extent_buffer(buf, item2_data, item1_offset, item1_size);
2394 btrfs_set_item_offset(buf, item1, item2_offset);
2395 btrfs_set_item_offset(buf, item2, item1_offset);
2396 btrfs_set_item_size(buf, item1, item2_size);
2397 btrfs_set_item_size(buf, item2, item1_size);
2399 path->slots[0] = slot;
2400 btrfs_set_item_key_unsafe(root, path, &k2);
2401 path->slots[0] = slot + 1;
2402 btrfs_set_item_key_unsafe(root, path, &k1);
2408 * Attempt to fix basic block failures. Currently we only handle bad key
2409 * orders, we will cycle through the keys and swap them if necessary.
2411 static int try_to_fix_bad_block(struct btrfs_trans_handle *trans,
2412 struct btrfs_root *root,
2413 struct extent_buffer *buf,
2414 struct btrfs_disk_key *parent_key,
2415 enum btrfs_tree_block_status status)
2417 struct btrfs_path *path;
2418 struct btrfs_key k1, k2;
2422 if (status != BTRFS_TREE_BLOCK_BAD_KEY_ORDER)
2425 k1.objectid = btrfs_header_owner(buf);
2426 k1.type = BTRFS_ROOT_ITEM_KEY;
2427 k1.offset = (u64)-1;
2429 root = btrfs_read_fs_root(root->fs_info, &k1);
2433 path = btrfs_alloc_path();
2437 path->lowest_level = btrfs_header_level(buf);
2438 path->skip_check_block = 1;
2439 if (btrfs_header_level(buf))
2440 btrfs_node_key_to_cpu(buf, &k1, 0);
2442 btrfs_item_key_to_cpu(buf, &k1, 0);
2444 ret = btrfs_search_slot(trans, root, &k1, path, 0, 1);
2446 btrfs_free_path(path);
2450 buf = path->nodes[0];
2451 for (i = 0; i < btrfs_header_nritems(buf) - 1; i++) {
2452 if (btrfs_header_level(buf)) {
2453 btrfs_node_key_to_cpu(buf, &k1, i);
2454 btrfs_node_key_to_cpu(buf, &k2, i + 1);
2456 btrfs_item_key_to_cpu(buf, &k1, i);
2457 btrfs_item_key_to_cpu(buf, &k2, i + 1);
2459 if (btrfs_comp_cpu_keys(&k1, &k2) < 0)
2461 ret = swap_values(root, path, buf, i);
2464 btrfs_mark_buffer_dirty(buf);
2468 btrfs_free_path(path);
2472 static int check_block(struct btrfs_trans_handle *trans,
2473 struct btrfs_root *root,
2474 struct cache_tree *extent_cache,
2475 struct extent_buffer *buf, u64 flags)
2477 struct extent_record *rec;
2478 struct cache_extent *cache;
2479 struct btrfs_key key;
2480 enum btrfs_tree_block_status status;
2484 cache = lookup_cache_extent(extent_cache, buf->start, buf->len);
2487 rec = container_of(cache, struct extent_record, cache);
2488 rec->generation = btrfs_header_generation(buf);
2490 level = btrfs_header_level(buf);
2491 if (btrfs_header_nritems(buf) > 0) {
2494 btrfs_item_key_to_cpu(buf, &key, 0);
2496 btrfs_node_key_to_cpu(buf, &key, 0);
2498 rec->info_objectid = key.objectid;
2500 rec->info_level = level;
2502 if (btrfs_is_leaf(buf))
2503 status = btrfs_check_leaf(root, &rec->parent_key, buf);
2505 status = btrfs_check_node(root, &rec->parent_key, buf);
2507 if (status != BTRFS_TREE_BLOCK_CLEAN) {
2509 status = try_to_fix_bad_block(trans, root, buf,
2512 if (status != BTRFS_TREE_BLOCK_CLEAN) {
2514 fprintf(stderr, "bad block %llu\n",
2515 (unsigned long long)buf->start);
2518 * Signal to callers we need to start the scan over
2519 * again since we'll have cow'ed blocks.
2524 rec->content_checked = 1;
2525 if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)
2526 rec->owner_ref_checked = 1;
2528 ret = check_owner_ref(root, rec, buf);
2530 rec->owner_ref_checked = 1;
2534 maybe_free_extent_rec(extent_cache, rec);
2538 static struct tree_backref *find_tree_backref(struct extent_record *rec,
2539 u64 parent, u64 root)
2541 struct list_head *cur = rec->backrefs.next;
2542 struct extent_backref *node;
2543 struct tree_backref *back;
2545 while(cur != &rec->backrefs) {
2546 node = list_entry(cur, struct extent_backref, list);
2550 back = (struct tree_backref *)node;
2552 if (!node->full_backref)
2554 if (parent == back->parent)
2557 if (node->full_backref)
2559 if (back->root == root)
2566 static struct tree_backref *alloc_tree_backref(struct extent_record *rec,
2567 u64 parent, u64 root)
2569 struct tree_backref *ref = malloc(sizeof(*ref));
2570 memset(&ref->node, 0, sizeof(ref->node));
2572 ref->parent = parent;
2573 ref->node.full_backref = 1;
2576 ref->node.full_backref = 0;
2578 list_add_tail(&ref->node.list, &rec->backrefs);
2583 static struct data_backref *find_data_backref(struct extent_record *rec,
2584 u64 parent, u64 root,
2585 u64 owner, u64 offset,
2587 u64 disk_bytenr, u64 bytes)
2589 struct list_head *cur = rec->backrefs.next;
2590 struct extent_backref *node;
2591 struct data_backref *back;
2593 while(cur != &rec->backrefs) {
2594 node = list_entry(cur, struct extent_backref, list);
2598 back = (struct data_backref *)node;
2600 if (!node->full_backref)
2602 if (parent == back->parent)
2605 if (node->full_backref)
2607 if (back->root == root && back->owner == owner &&
2608 back->offset == offset) {
2609 if (found_ref && node->found_ref &&
2610 (back->bytes != bytes ||
2611 back->disk_bytenr != disk_bytenr))
2620 static struct data_backref *alloc_data_backref(struct extent_record *rec,
2621 u64 parent, u64 root,
2622 u64 owner, u64 offset,
2625 struct data_backref *ref = malloc(sizeof(*ref));
2626 memset(&ref->node, 0, sizeof(ref->node));
2627 ref->node.is_data = 1;
2630 ref->parent = parent;
2633 ref->node.full_backref = 1;
2637 ref->offset = offset;
2638 ref->node.full_backref = 0;
2640 ref->bytes = max_size;
2643 list_add_tail(&ref->node.list, &rec->backrefs);
2644 if (max_size > rec->max_size)
2645 rec->max_size = max_size;
2649 static int add_extent_rec(struct cache_tree *extent_cache,
2650 struct btrfs_key *parent_key, u64 parent_gen,
2651 u64 start, u64 nr, u64 extent_item_refs,
2652 int is_root, int inc_ref, int set_checked,
2653 int metadata, int extent_rec, u64 max_size)
2655 struct extent_record *rec;
2656 struct cache_extent *cache;
2660 cache = lookup_cache_extent(extent_cache, start, nr);
2662 rec = container_of(cache, struct extent_record, cache);
2666 rec->nr = max(nr, max_size);
2669 * We need to make sure to reset nr to whatever the extent
2670 * record says was the real size, this way we can compare it to
2674 if (start != rec->start || rec->found_rec) {
2675 struct extent_record *tmp;
2678 if (list_empty(&rec->list))
2679 list_add_tail(&rec->list,
2680 &duplicate_extents);
2683 * We have to do this song and dance in case we
2684 * find an extent record that falls inside of
2685 * our current extent record but does not have
2686 * the same objectid.
2688 tmp = malloc(sizeof(*tmp));
2692 tmp->max_size = max_size;
2695 tmp->metadata = metadata;
2696 tmp->extent_item_refs = extent_item_refs;
2697 INIT_LIST_HEAD(&tmp->list);
2698 list_add_tail(&tmp->list, &rec->dups);
2699 rec->num_duplicates++;
2706 if (extent_item_refs && !dup) {
2707 if (rec->extent_item_refs) {
2708 fprintf(stderr, "block %llu rec "
2709 "extent_item_refs %llu, passed %llu\n",
2710 (unsigned long long)start,
2711 (unsigned long long)
2712 rec->extent_item_refs,
2713 (unsigned long long)extent_item_refs);
2715 rec->extent_item_refs = extent_item_refs;
2720 rec->content_checked = 1;
2721 rec->owner_ref_checked = 1;
2725 btrfs_cpu_key_to_disk(&rec->parent_key, parent_key);
2727 rec->parent_generation = parent_gen;
2729 if (rec->max_size < max_size)
2730 rec->max_size = max_size;
2732 maybe_free_extent_rec(extent_cache, rec);
2735 rec = malloc(sizeof(*rec));
2737 rec->max_size = max_size;
2738 rec->nr = max(nr, max_size);
2739 rec->found_rec = extent_rec;
2740 rec->content_checked = 0;
2741 rec->owner_ref_checked = 0;
2742 rec->num_duplicates = 0;
2743 rec->metadata = metadata;
2744 INIT_LIST_HEAD(&rec->backrefs);
2745 INIT_LIST_HEAD(&rec->dups);
2746 INIT_LIST_HEAD(&rec->list);
2758 if (extent_item_refs)
2759 rec->extent_item_refs = extent_item_refs;
2761 rec->extent_item_refs = 0;
2764 btrfs_cpu_key_to_disk(&rec->parent_key, parent_key);
2766 memset(&rec->parent_key, 0, sizeof(*parent_key));
2769 rec->parent_generation = parent_gen;
2771 rec->parent_generation = 0;
2773 rec->cache.start = start;
2774 rec->cache.size = nr;
2775 ret = insert_cache_extent(extent_cache, &rec->cache);
2779 rec->content_checked = 1;
2780 rec->owner_ref_checked = 1;
2785 static int add_tree_backref(struct cache_tree *extent_cache, u64 bytenr,
2786 u64 parent, u64 root, int found_ref)
2788 struct extent_record *rec;
2789 struct tree_backref *back;
2790 struct cache_extent *cache;
2792 cache = lookup_cache_extent(extent_cache, bytenr, 1);
2794 add_extent_rec(extent_cache, NULL, 0, bytenr,
2795 1, 0, 0, 0, 0, 1, 0, 0);
2796 cache = lookup_cache_extent(extent_cache, bytenr, 1);
2801 rec = container_of(cache, struct extent_record, cache);
2802 if (rec->start != bytenr) {
2806 back = find_tree_backref(rec, parent, root);
2808 back = alloc_tree_backref(rec, parent, root);
2811 if (back->node.found_ref) {
2812 fprintf(stderr, "Extent back ref already exists "
2813 "for %llu parent %llu root %llu \n",
2814 (unsigned long long)bytenr,
2815 (unsigned long long)parent,
2816 (unsigned long long)root);
2818 back->node.found_ref = 1;
2820 if (back->node.found_extent_tree) {
2821 fprintf(stderr, "Extent back ref already exists "
2822 "for %llu parent %llu root %llu \n",
2823 (unsigned long long)bytenr,
2824 (unsigned long long)parent,
2825 (unsigned long long)root);
2827 back->node.found_extent_tree = 1;
2832 static int add_data_backref(struct cache_tree *extent_cache, u64 bytenr,
2833 u64 parent, u64 root, u64 owner, u64 offset,
2834 u32 num_refs, int found_ref, u64 max_size)
2836 struct extent_record *rec;
2837 struct data_backref *back;
2838 struct cache_extent *cache;
2840 cache = lookup_cache_extent(extent_cache, bytenr, 1);
2842 add_extent_rec(extent_cache, NULL, 0, bytenr, 1, 0, 0, 0, 0,
2844 cache = lookup_cache_extent(extent_cache, bytenr, 1);
2849 rec = container_of(cache, struct extent_record, cache);
2850 if (rec->max_size < max_size)
2851 rec->max_size = max_size;
2854 * If found_ref is set then max_size is the real size and must match the
2855 * existing refs. So if we have already found a ref then we need to
2856 * make sure that this ref matches the existing one, otherwise we need
2857 * to add a new backref so we can notice that the backrefs don't match
2858 * and we need to figure out who is telling the truth. This is to
2859 * account for that awful fsync bug I introduced where we'd end up with
2860 * a btrfs_file_extent_item that would have its length include multiple
2861 * prealloc extents or point inside of a prealloc extent.
2863 back = find_data_backref(rec, parent, root, owner, offset, found_ref,
2866 back = alloc_data_backref(rec, parent, root, owner, offset,
2870 BUG_ON(num_refs != 1);
2871 if (back->node.found_ref)
2872 BUG_ON(back->bytes != max_size);
2873 back->node.found_ref = 1;
2874 back->found_ref += 1;
2875 back->bytes = max_size;
2876 back->disk_bytenr = bytenr;
2878 rec->content_checked = 1;
2879 rec->owner_ref_checked = 1;
2881 if (back->node.found_extent_tree) {
2882 fprintf(stderr, "Extent back ref already exists "
2883 "for %llu parent %llu root %llu"
2884 "owner %llu offset %llu num_refs %lu\n",
2885 (unsigned long long)bytenr,
2886 (unsigned long long)parent,
2887 (unsigned long long)root,
2888 (unsigned long long)owner,
2889 (unsigned long long)offset,
2890 (unsigned long)num_refs);
2892 back->num_refs = num_refs;
2893 back->node.found_extent_tree = 1;
2898 static int add_pending(struct cache_tree *pending,
2899 struct cache_tree *seen, u64 bytenr, u32 size)
2902 ret = add_cache_extent(seen, bytenr, size);
2905 add_cache_extent(pending, bytenr, size);
2909 static int pick_next_pending(struct cache_tree *pending,
2910 struct cache_tree *reada,
2911 struct cache_tree *nodes,
2912 u64 last, struct block_info *bits, int bits_nr,
2915 unsigned long node_start = last;
2916 struct cache_extent *cache;
2919 cache = search_cache_extent(reada, 0);
2921 bits[0].start = cache->start;
2922 bits[1].size = cache->size;
2927 if (node_start > 32768)
2928 node_start -= 32768;
2930 cache = search_cache_extent(nodes, node_start);
2932 cache = search_cache_extent(nodes, 0);
2935 cache = search_cache_extent(pending, 0);
2940 bits[ret].start = cache->start;
2941 bits[ret].size = cache->size;
2942 cache = next_cache_extent(cache);
2944 } while (cache && ret < bits_nr);
2950 bits[ret].start = cache->start;
2951 bits[ret].size = cache->size;
2952 cache = next_cache_extent(cache);
2954 } while (cache && ret < bits_nr);
2956 if (bits_nr - ret > 8) {
2957 u64 lookup = bits[0].start + bits[0].size;
2958 struct cache_extent *next;
2959 next = search_cache_extent(pending, lookup);
2961 if (next->start - lookup > 32768)
2963 bits[ret].start = next->start;
2964 bits[ret].size = next->size;
2965 lookup = next->start + next->size;
2969 next = next_cache_extent(next);
2977 static void free_chunk_record(struct cache_extent *cache)
2979 struct chunk_record *rec;
2981 rec = container_of(cache, struct chunk_record, cache);
2985 void free_chunk_cache_tree(struct cache_tree *chunk_cache)
2987 cache_tree_free_extents(chunk_cache, free_chunk_record);
2990 static void free_device_record(struct rb_node *node)
2992 struct device_record *rec;
2994 rec = container_of(node, struct device_record, node);
2998 FREE_RB_BASED_TREE(device_cache, free_device_record);
3000 int insert_block_group_record(struct block_group_tree *tree,
3001 struct block_group_record *bg_rec)
3005 ret = insert_cache_extent(&tree->tree, &bg_rec->cache);
3009 list_add_tail(&bg_rec->list, &tree->block_groups);
3013 static void free_block_group_record(struct cache_extent *cache)
3015 struct block_group_record *rec;
3017 rec = container_of(cache, struct block_group_record, cache);
3021 void free_block_group_tree(struct block_group_tree *tree)
3023 cache_tree_free_extents(&tree->tree, free_block_group_record);
3026 int insert_device_extent_record(struct device_extent_tree *tree,
3027 struct device_extent_record *de_rec)
3032 * Device extent is a bit different from the other extents, because
3033 * the extents which belong to the different devices may have the
3034 * same start and size, so we need use the special extent cache
3035 * search/insert functions.
3037 ret = insert_cache_extent2(&tree->tree, &de_rec->cache);
3041 list_add_tail(&de_rec->chunk_list, &tree->no_chunk_orphans);
3042 list_add_tail(&de_rec->device_list, &tree->no_device_orphans);
3046 static void free_device_extent_record(struct cache_extent *cache)
3048 struct device_extent_record *rec;
3050 rec = container_of(cache, struct device_extent_record, cache);
3054 void free_device_extent_tree(struct device_extent_tree *tree)
3056 cache_tree_free_extents(&tree->tree, free_device_extent_record);
3059 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3060 static int process_extent_ref_v0(struct cache_tree *extent_cache,
3061 struct extent_buffer *leaf, int slot)
3063 struct btrfs_extent_ref_v0 *ref0;
3064 struct btrfs_key key;
3066 btrfs_item_key_to_cpu(leaf, &key, slot);
3067 ref0 = btrfs_item_ptr(leaf, slot, struct btrfs_extent_ref_v0);
3068 if (btrfs_ref_objectid_v0(leaf, ref0) < BTRFS_FIRST_FREE_OBJECTID) {
3069 add_tree_backref(extent_cache, key.objectid, key.offset, 0, 0);
3071 add_data_backref(extent_cache, key.objectid, key.offset, 0,
3072 0, 0, btrfs_ref_count_v0(leaf, ref0), 0, 0);
3078 struct chunk_record *btrfs_new_chunk_record(struct extent_buffer *leaf,
3079 struct btrfs_key *key,
3082 struct btrfs_chunk *ptr;
3083 struct chunk_record *rec;
3086 ptr = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
3087 num_stripes = btrfs_chunk_num_stripes(leaf, ptr);
3089 rec = malloc(btrfs_chunk_record_size(num_stripes));
3091 fprintf(stderr, "memory allocation failed\n");
3095 memset(rec, 0, btrfs_chunk_record_size(num_stripes));
3097 INIT_LIST_HEAD(&rec->list);
3098 INIT_LIST_HEAD(&rec->dextents);
3101 rec->cache.start = key->offset;
3102 rec->cache.size = btrfs_chunk_length(leaf, ptr);
3104 rec->generation = btrfs_header_generation(leaf);
3106 rec->objectid = key->objectid;
3107 rec->type = key->type;
3108 rec->offset = key->offset;
3110 rec->length = rec->cache.size;
3111 rec->owner = btrfs_chunk_owner(leaf, ptr);
3112 rec->stripe_len = btrfs_chunk_stripe_len(leaf, ptr);
3113 rec->type_flags = btrfs_chunk_type(leaf, ptr);
3114 rec->io_width = btrfs_chunk_io_width(leaf, ptr);
3115 rec->io_align = btrfs_chunk_io_align(leaf, ptr);
3116 rec->sector_size = btrfs_chunk_sector_size(leaf, ptr);
3117 rec->num_stripes = num_stripes;
3118 rec->sub_stripes = btrfs_chunk_sub_stripes(leaf, ptr);
3120 for (i = 0; i < rec->num_stripes; ++i) {
3121 rec->stripes[i].devid =
3122 btrfs_stripe_devid_nr(leaf, ptr, i);
3123 rec->stripes[i].offset =
3124 btrfs_stripe_offset_nr(leaf, ptr, i);
3125 read_extent_buffer(leaf, rec->stripes[i].dev_uuid,
3126 (unsigned long)btrfs_stripe_dev_uuid_nr(ptr, i),
3133 static int process_chunk_item(struct cache_tree *chunk_cache,
3134 struct btrfs_key *key, struct extent_buffer *eb,
3137 struct chunk_record *rec;
3140 rec = btrfs_new_chunk_record(eb, key, slot);
3141 ret = insert_cache_extent(chunk_cache, &rec->cache);
3143 fprintf(stderr, "Chunk[%llu, %llu] existed.\n",
3144 rec->offset, rec->length);
3151 static int process_device_item(struct rb_root *dev_cache,
3152 struct btrfs_key *key, struct extent_buffer *eb, int slot)
3154 struct btrfs_dev_item *ptr;
3155 struct device_record *rec;
3158 ptr = btrfs_item_ptr(eb,
3159 slot, struct btrfs_dev_item);
3161 rec = malloc(sizeof(*rec));
3163 fprintf(stderr, "memory allocation failed\n");
3167 rec->devid = key->offset;
3168 rec->generation = btrfs_header_generation(eb);
3170 rec->objectid = key->objectid;
3171 rec->type = key->type;
3172 rec->offset = key->offset;
3174 rec->devid = btrfs_device_id(eb, ptr);
3175 rec->total_byte = btrfs_device_total_bytes(eb, ptr);
3176 rec->byte_used = btrfs_device_bytes_used(eb, ptr);
3178 ret = rb_insert(dev_cache, &rec->node, device_record_compare);
3180 fprintf(stderr, "Device[%llu] existed.\n", rec->devid);
3187 struct block_group_record *
3188 btrfs_new_block_group_record(struct extent_buffer *leaf, struct btrfs_key *key,
3191 struct btrfs_block_group_item *ptr;
3192 struct block_group_record *rec;
3194 rec = malloc(sizeof(*rec));
3196 fprintf(stderr, "memory allocation failed\n");
3199 memset(rec, 0, sizeof(*rec));
3201 rec->cache.start = key->objectid;
3202 rec->cache.size = key->offset;
3204 rec->generation = btrfs_header_generation(leaf);
3206 rec->objectid = key->objectid;
3207 rec->type = key->type;
3208 rec->offset = key->offset;
3210 ptr = btrfs_item_ptr(leaf, slot, struct btrfs_block_group_item);
3211 rec->flags = btrfs_disk_block_group_flags(leaf, ptr);
3213 INIT_LIST_HEAD(&rec->list);
3218 static int process_block_group_item(struct block_group_tree *block_group_cache,
3219 struct btrfs_key *key,
3220 struct extent_buffer *eb, int slot)
3222 struct block_group_record *rec;
3225 rec = btrfs_new_block_group_record(eb, key, slot);
3226 ret = insert_block_group_record(block_group_cache, rec);
3228 fprintf(stderr, "Block Group[%llu, %llu] existed.\n",
3229 rec->objectid, rec->offset);
3236 struct device_extent_record *
3237 btrfs_new_device_extent_record(struct extent_buffer *leaf,
3238 struct btrfs_key *key, int slot)
3240 struct device_extent_record *rec;
3241 struct btrfs_dev_extent *ptr;
3243 rec = malloc(sizeof(*rec));
3245 fprintf(stderr, "memory allocation failed\n");
3248 memset(rec, 0, sizeof(*rec));
3250 rec->cache.objectid = key->objectid;
3251 rec->cache.start = key->offset;
3253 rec->generation = btrfs_header_generation(leaf);
3255 rec->objectid = key->objectid;
3256 rec->type = key->type;
3257 rec->offset = key->offset;
3259 ptr = btrfs_item_ptr(leaf, slot, struct btrfs_dev_extent);
3260 rec->chunk_objecteid =
3261 btrfs_dev_extent_chunk_objectid(leaf, ptr);
3263 btrfs_dev_extent_chunk_offset(leaf, ptr);
3264 rec->length = btrfs_dev_extent_length(leaf, ptr);
3265 rec->cache.size = rec->length;
3267 INIT_LIST_HEAD(&rec->chunk_list);
3268 INIT_LIST_HEAD(&rec->device_list);
3274 process_device_extent_item(struct device_extent_tree *dev_extent_cache,
3275 struct btrfs_key *key, struct extent_buffer *eb,
3278 struct device_extent_record *rec;
3281 rec = btrfs_new_device_extent_record(eb, key, slot);
3282 ret = insert_device_extent_record(dev_extent_cache, rec);
3285 "Device extent[%llu, %llu, %llu] existed.\n",
3286 rec->objectid, rec->offset, rec->length);
3293 static int process_extent_item(struct btrfs_root *root,
3294 struct cache_tree *extent_cache,
3295 struct extent_buffer *eb, int slot)
3297 struct btrfs_extent_item *ei;
3298 struct btrfs_extent_inline_ref *iref;
3299 struct btrfs_extent_data_ref *dref;
3300 struct btrfs_shared_data_ref *sref;
3301 struct btrfs_key key;
3305 u32 item_size = btrfs_item_size_nr(eb, slot);
3311 btrfs_item_key_to_cpu(eb, &key, slot);
3313 if (key.type == BTRFS_METADATA_ITEM_KEY) {
3315 num_bytes = root->leafsize;
3317 num_bytes = key.offset;
3320 if (item_size < sizeof(*ei)) {
3321 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3322 struct btrfs_extent_item_v0 *ei0;
3323 BUG_ON(item_size != sizeof(*ei0));
3324 ei0 = btrfs_item_ptr(eb, slot, struct btrfs_extent_item_v0);
3325 refs = btrfs_extent_refs_v0(eb, ei0);
3329 return add_extent_rec(extent_cache, NULL, 0, key.objectid,
3330 num_bytes, refs, 0, 0, 0, metadata, 1,
3334 ei = btrfs_item_ptr(eb, slot, struct btrfs_extent_item);
3335 refs = btrfs_extent_refs(eb, ei);
3337 add_extent_rec(extent_cache, NULL, 0, key.objectid, num_bytes,
3338 refs, 0, 0, 0, metadata, 1, num_bytes);
3340 ptr = (unsigned long)(ei + 1);
3341 if (btrfs_extent_flags(eb, ei) & BTRFS_EXTENT_FLAG_TREE_BLOCK &&
3342 key.type == BTRFS_EXTENT_ITEM_KEY)
3343 ptr += sizeof(struct btrfs_tree_block_info);
3345 end = (unsigned long)ei + item_size;
3347 iref = (struct btrfs_extent_inline_ref *)ptr;
3348 type = btrfs_extent_inline_ref_type(eb, iref);
3349 offset = btrfs_extent_inline_ref_offset(eb, iref);
3351 case BTRFS_TREE_BLOCK_REF_KEY:
3352 add_tree_backref(extent_cache, key.objectid,
3355 case BTRFS_SHARED_BLOCK_REF_KEY:
3356 add_tree_backref(extent_cache, key.objectid,
3359 case BTRFS_EXTENT_DATA_REF_KEY:
3360 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
3361 add_data_backref(extent_cache, key.objectid, 0,
3362 btrfs_extent_data_ref_root(eb, dref),
3363 btrfs_extent_data_ref_objectid(eb,
3365 btrfs_extent_data_ref_offset(eb, dref),
3366 btrfs_extent_data_ref_count(eb, dref),
3369 case BTRFS_SHARED_DATA_REF_KEY:
3370 sref = (struct btrfs_shared_data_ref *)(iref + 1);
3371 add_data_backref(extent_cache, key.objectid, offset,
3373 btrfs_shared_data_ref_count(eb, sref),
3377 fprintf(stderr, "corrupt extent record: key %Lu %u %Lu\n",
3378 key.objectid, key.type, num_bytes);
3381 ptr += btrfs_extent_inline_ref_size(type);
3388 static int check_cache_range(struct btrfs_root *root,
3389 struct btrfs_block_group_cache *cache,
3390 u64 offset, u64 bytes)
3392 struct btrfs_free_space *entry;
3398 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
3399 bytenr = btrfs_sb_offset(i);
3400 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
3401 cache->key.objectid, bytenr, 0,
3402 &logical, &nr, &stripe_len);
3407 if (logical[nr] + stripe_len <= offset)
3409 if (offset + bytes <= logical[nr])
3411 if (logical[nr] == offset) {
3412 if (stripe_len >= bytes) {
3416 bytes -= stripe_len;
3417 offset += stripe_len;
3418 } else if (logical[nr] < offset) {
3419 if (logical[nr] + stripe_len >=
3424 bytes = (offset + bytes) -
3425 (logical[nr] + stripe_len);
3426 offset = logical[nr] + stripe_len;
3429 * Could be tricky, the super may land in the
3430 * middle of the area we're checking. First
3431 * check the easiest case, it's at the end.
3433 if (logical[nr] + stripe_len >=
3435 bytes = logical[nr] - offset;
3439 /* Check the left side */
3440 ret = check_cache_range(root, cache,
3442 logical[nr] - offset);
3448 /* Now we continue with the right side */
3449 bytes = (offset + bytes) -
3450 (logical[nr] + stripe_len);
3451 offset = logical[nr] + stripe_len;
3458 entry = btrfs_find_free_space(cache->free_space_ctl, offset, bytes);
3460 fprintf(stderr, "There is no free space entry for %Lu-%Lu\n",
3461 offset, offset+bytes);
3465 if (entry->offset != offset) {
3466 fprintf(stderr, "Wanted offset %Lu, found %Lu\n", offset,
3471 if (entry->bytes != bytes) {
3472 fprintf(stderr, "Wanted bytes %Lu, found %Lu for off %Lu\n",
3473 bytes, entry->bytes, offset);
3477 unlink_free_space(cache->free_space_ctl, entry);
3482 static int verify_space_cache(struct btrfs_root *root,
3483 struct btrfs_block_group_cache *cache)
3485 struct btrfs_path *path;
3486 struct extent_buffer *leaf;
3487 struct btrfs_key key;
3491 path = btrfs_alloc_path();
3495 root = root->fs_info->extent_root;
3497 last = max_t(u64, cache->key.objectid, BTRFS_SUPER_INFO_OFFSET);
3499 key.objectid = last;
3501 key.type = BTRFS_EXTENT_ITEM_KEY;
3503 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3508 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3509 ret = btrfs_next_leaf(root, path);
3517 leaf = path->nodes[0];
3518 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3519 if (key.objectid >= cache->key.offset + cache->key.objectid)
3521 if (key.type != BTRFS_EXTENT_ITEM_KEY &&
3522 key.type != BTRFS_METADATA_ITEM_KEY) {
3527 if (last == key.objectid) {
3528 if (key.type == BTRFS_EXTENT_ITEM_KEY)
3529 last = key.objectid + key.offset;
3531 last = key.objectid + root->leafsize;
3536 ret = check_cache_range(root, cache, last,
3537 key.objectid - last);
3540 if (key.type == BTRFS_EXTENT_ITEM_KEY)
3541 last = key.objectid + key.offset;
3543 last = key.objectid + root->leafsize;
3547 if (last < cache->key.objectid + cache->key.offset)
3548 ret = check_cache_range(root, cache, last,
3549 cache->key.objectid +
3550 cache->key.offset - last);
3553 btrfs_free_path(path);
3556 !RB_EMPTY_ROOT(&cache->free_space_ctl->free_space_offset)) {
3557 fprintf(stderr, "There are still entries left in the space "
3565 static int check_space_cache(struct btrfs_root *root)
3567 struct btrfs_block_group_cache *cache;
3568 u64 start = BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE;
3572 if (btrfs_super_cache_generation(root->fs_info->super_copy) != -1ULL &&
3573 btrfs_super_generation(root->fs_info->super_copy) !=
3574 btrfs_super_cache_generation(root->fs_info->super_copy)) {
3575 printf("cache and super generation don't match, space cache "
3576 "will be invalidated\n");
3581 cache = btrfs_lookup_first_block_group(root->fs_info, start);
3585 start = cache->key.objectid + cache->key.offset;
3586 if (!cache->free_space_ctl) {
3587 if (btrfs_init_free_space_ctl(cache,
3588 root->sectorsize)) {
3593 btrfs_remove_free_space_cache(cache);
3596 ret = load_free_space_cache(root->fs_info, cache);
3600 ret = verify_space_cache(root, cache);
3602 fprintf(stderr, "cache appears valid but isnt %Lu\n",
3603 cache->key.objectid);
3608 return error ? -EINVAL : 0;
3611 static int check_extent_exists(struct btrfs_root *root, u64 bytenr,
3614 struct btrfs_path *path;
3615 struct extent_buffer *leaf;
3616 struct btrfs_key key;
3619 path = btrfs_alloc_path();
3621 fprintf(stderr, "Error allocing path\n");
3625 key.objectid = bytenr;
3626 key.type = BTRFS_EXTENT_ITEM_KEY;
3631 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
3634 fprintf(stderr, "Error looking up extent record %d\n", ret);
3635 btrfs_free_path(path);
3641 btrfs_prev_leaf(root, path);
3644 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3647 * Block group items come before extent items if they have the same
3648 * bytenr, so walk back one more just in case. Dear future traveler,
3649 * first congrats on mastering time travel. Now if it's not too much
3650 * trouble could you go back to 2006 and tell Chris to make the
3651 * BLOCK_GROUP_ITEM_KEY lower than the EXTENT_ITEM_KEY please?
3653 if (key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
3657 btrfs_prev_leaf(root, path);
3661 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3662 ret = btrfs_next_leaf(root, path);
3664 fprintf(stderr, "Error going to next leaf "
3666 btrfs_free_path(path);
3672 leaf = path->nodes[0];
3673 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3674 if (key.type != BTRFS_EXTENT_ITEM_KEY) {
3678 if (key.objectid + key.offset < bytenr) {
3682 if (key.objectid > bytenr + num_bytes)
3685 if (key.objectid == bytenr) {
3686 if (key.offset >= num_bytes) {
3690 num_bytes -= key.offset;
3691 bytenr += key.offset;
3692 } else if (key.objectid < bytenr) {
3693 if (key.objectid + key.offset >= bytenr + num_bytes) {
3697 num_bytes = (bytenr + num_bytes) -
3698 (key.objectid + key.offset);
3699 bytenr = key.objectid + key.offset;
3701 if (key.objectid + key.offset < bytenr + num_bytes) {
3702 u64 new_start = key.objectid + key.offset;
3703 u64 new_bytes = bytenr + num_bytes - new_start;
3706 * Weird case, the extent is in the middle of
3707 * our range, we'll have to search one side
3708 * and then the other. Not sure if this happens
3709 * in real life, but no harm in coding it up
3710 * anyway just in case.
3712 btrfs_release_path(path);
3713 ret = check_extent_exists(root, new_start,
3716 fprintf(stderr, "Right section didn't "
3720 num_bytes = key.objectid - bytenr;
3723 num_bytes = key.objectid - bytenr;
3730 fprintf(stderr, "There are no extents for csum range "
3731 "%Lu-%Lu\n", bytenr, bytenr+num_bytes);
3735 btrfs_free_path(path);
3739 static int check_csums(struct btrfs_root *root)
3741 struct btrfs_path *path;
3742 struct extent_buffer *leaf;
3743 struct btrfs_key key;
3744 u64 offset = 0, num_bytes = 0;
3745 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
3749 root = root->fs_info->csum_root;
3751 key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
3752 key.type = BTRFS_EXTENT_CSUM_KEY;
3755 path = btrfs_alloc_path();
3759 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3761 fprintf(stderr, "Error searching csum tree %d\n", ret);
3762 btrfs_free_path(path);
3766 if (ret > 0 && path->slots[0])
3771 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3772 ret = btrfs_next_leaf(root, path);
3774 fprintf(stderr, "Error going to next leaf "
3781 leaf = path->nodes[0];
3783 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3784 if (key.type != BTRFS_EXTENT_CSUM_KEY) {
3790 offset = key.offset;
3791 } else if (key.offset != offset + num_bytes) {
3792 ret = check_extent_exists(root, offset, num_bytes);
3794 fprintf(stderr, "Csum exists for %Lu-%Lu but "
3795 "there is no extent record\n",
3796 offset, offset+num_bytes);
3799 offset = key.offset;
3803 num_bytes += (btrfs_item_size_nr(leaf, path->slots[0]) /
3804 csum_size) * root->sectorsize;
3808 btrfs_free_path(path);
3812 static int is_dropped_key(struct btrfs_key *key,
3813 struct btrfs_key *drop_key) {
3814 if (key->objectid < drop_key->objectid)
3816 else if (key->objectid == drop_key->objectid) {
3817 if (key->type < drop_key->type)
3819 else if (key->type == drop_key->type) {
3820 if (key->offset < drop_key->offset)
3827 static int run_next_block(struct btrfs_trans_handle *trans,
3828 struct btrfs_root *root,
3829 struct block_info *bits,
3832 struct cache_tree *pending,
3833 struct cache_tree *seen,
3834 struct cache_tree *reada,
3835 struct cache_tree *nodes,
3836 struct cache_tree *extent_cache,
3837 struct cache_tree *chunk_cache,
3838 struct rb_root *dev_cache,
3839 struct block_group_tree *block_group_cache,
3840 struct device_extent_tree *dev_extent_cache,
3841 struct btrfs_root_item *ri)
3843 struct extent_buffer *buf;
3854 struct btrfs_key key;
3855 struct cache_extent *cache;
3858 nritems = pick_next_pending(pending, reada, nodes, *last, bits,
3859 bits_nr, &reada_bits);
3864 for(i = 0; i < nritems; i++) {
3865 ret = add_cache_extent(reada, bits[i].start,
3870 /* fixme, get the parent transid */
3871 readahead_tree_block(root, bits[i].start,
3875 *last = bits[0].start;
3876 bytenr = bits[0].start;
3877 size = bits[0].size;
3879 cache = lookup_cache_extent(pending, bytenr, size);
3881 remove_cache_extent(pending, cache);
3884 cache = lookup_cache_extent(reada, bytenr, size);
3886 remove_cache_extent(reada, cache);
3889 cache = lookup_cache_extent(nodes, bytenr, size);
3891 remove_cache_extent(nodes, cache);
3894 cache = lookup_cache_extent(seen, bytenr, size);
3896 remove_cache_extent(seen, cache);
3900 cache = lookup_cache_extent(extent_cache, bytenr, size);
3902 struct extent_record *rec;
3904 rec = container_of(cache, struct extent_record, cache);
3905 gen = rec->parent_generation;
3908 /* fixme, get the real parent transid */
3909 buf = read_tree_block(root, bytenr, size, gen);
3910 if (!extent_buffer_uptodate(buf)) {
3911 record_bad_block_io(root->fs_info,
3912 extent_cache, bytenr, size);
3916 nritems = btrfs_header_nritems(buf);
3918 ret = btrfs_lookup_extent_info(NULL, root, bytenr,
3919 btrfs_header_level(buf), 1, NULL,
3922 flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
3924 if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
3929 owner = btrfs_header_owner(buf);
3932 ret = check_block(trans, root, extent_cache, buf, flags);
3936 if (btrfs_is_leaf(buf)) {
3937 btree_space_waste += btrfs_leaf_free_space(root, buf);
3938 for (i = 0; i < nritems; i++) {
3939 struct btrfs_file_extent_item *fi;
3940 btrfs_item_key_to_cpu(buf, &key, i);
3941 if (key.type == BTRFS_EXTENT_ITEM_KEY) {
3942 process_extent_item(root, extent_cache, buf,
3946 if (key.type == BTRFS_METADATA_ITEM_KEY) {
3947 process_extent_item(root, extent_cache, buf,
3951 if (key.type == BTRFS_EXTENT_CSUM_KEY) {
3953 btrfs_item_size_nr(buf, i);
3956 if (key.type == BTRFS_CHUNK_ITEM_KEY) {
3957 process_chunk_item(chunk_cache, &key, buf, i);
3960 if (key.type == BTRFS_DEV_ITEM_KEY) {
3961 process_device_item(dev_cache, &key, buf, i);
3964 if (key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
3965 process_block_group_item(block_group_cache,
3969 if (key.type == BTRFS_DEV_EXTENT_KEY) {
3970 process_device_extent_item(dev_extent_cache,
3975 if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
3976 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3977 process_extent_ref_v0(extent_cache, buf, i);
3984 if (key.type == BTRFS_TREE_BLOCK_REF_KEY) {
3985 add_tree_backref(extent_cache, key.objectid, 0,
3989 if (key.type == BTRFS_SHARED_BLOCK_REF_KEY) {
3990 add_tree_backref(extent_cache, key.objectid,
3994 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
3995 struct btrfs_extent_data_ref *ref;
3996 ref = btrfs_item_ptr(buf, i,
3997 struct btrfs_extent_data_ref);
3998 add_data_backref(extent_cache,
4000 btrfs_extent_data_ref_root(buf, ref),
4001 btrfs_extent_data_ref_objectid(buf,
4003 btrfs_extent_data_ref_offset(buf, ref),
4004 btrfs_extent_data_ref_count(buf, ref),
4005 0, root->sectorsize);
4008 if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
4009 struct btrfs_shared_data_ref *ref;
4010 ref = btrfs_item_ptr(buf, i,
4011 struct btrfs_shared_data_ref);
4012 add_data_backref(extent_cache,
4013 key.objectid, key.offset, 0, 0, 0,
4014 btrfs_shared_data_ref_count(buf, ref),
4015 0, root->sectorsize);
4018 if (key.type == BTRFS_ORPHAN_ITEM_KEY) {
4019 struct bad_item *bad;
4021 if (key.objectid == BTRFS_ORPHAN_OBJECTID)
4025 bad = malloc(sizeof(struct bad_item));
4028 INIT_LIST_HEAD(&bad->list);
4029 memcpy(&bad->key, &key,
4030 sizeof(struct btrfs_key));
4031 bad->root_id = owner;
4032 list_add_tail(&bad->list, &delete_items);
4035 if (key.type != BTRFS_EXTENT_DATA_KEY)
4037 fi = btrfs_item_ptr(buf, i,
4038 struct btrfs_file_extent_item);
4039 if (btrfs_file_extent_type(buf, fi) ==
4040 BTRFS_FILE_EXTENT_INLINE)
4042 if (btrfs_file_extent_disk_bytenr(buf, fi) == 0)
4045 data_bytes_allocated +=
4046 btrfs_file_extent_disk_num_bytes(buf, fi);
4047 if (data_bytes_allocated < root->sectorsize) {
4050 data_bytes_referenced +=
4051 btrfs_file_extent_num_bytes(buf, fi);
4052 add_data_backref(extent_cache,
4053 btrfs_file_extent_disk_bytenr(buf, fi),
4054 parent, owner, key.objectid, key.offset -
4055 btrfs_file_extent_offset(buf, fi), 1, 1,
4056 btrfs_file_extent_disk_num_bytes(buf, fi));
4060 struct btrfs_key first_key;
4062 first_key.objectid = 0;
4065 btrfs_item_key_to_cpu(buf, &first_key, 0);
4066 level = btrfs_header_level(buf);
4067 for (i = 0; i < nritems; i++) {
4068 ptr = btrfs_node_blockptr(buf, i);
4069 size = btrfs_level_size(root, level - 1);
4070 btrfs_node_key_to_cpu(buf, &key, i);
4072 struct btrfs_key drop_key;
4073 btrfs_disk_key_to_cpu(&drop_key,
4074 &ri->drop_progress);
4075 if ((level == ri->drop_level)
4076 && is_dropped_key(&key, &drop_key)) {
4080 ret = add_extent_rec(extent_cache, &key,
4081 btrfs_node_ptr_generation(buf, i),
4082 ptr, size, 0, 0, 1, 0, 1, 0,
4086 add_tree_backref(extent_cache, ptr, parent, owner, 1);
4089 add_pending(nodes, seen, ptr, size);
4091 add_pending(pending, seen, ptr, size);
4094 btree_space_waste += (BTRFS_NODEPTRS_PER_BLOCK(root) -
4095 nritems) * sizeof(struct btrfs_key_ptr);
4097 total_btree_bytes += buf->len;
4098 if (fs_root_objectid(btrfs_header_owner(buf)))
4099 total_fs_tree_bytes += buf->len;
4100 if (btrfs_header_owner(buf) == BTRFS_EXTENT_TREE_OBJECTID)
4101 total_extent_tree_bytes += buf->len;
4102 if (!found_old_backref &&
4103 btrfs_header_owner(buf) == BTRFS_TREE_RELOC_OBJECTID &&
4104 btrfs_header_backref_rev(buf) == BTRFS_MIXED_BACKREF_REV &&
4105 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC))
4106 found_old_backref = 1;
4108 free_extent_buffer(buf);
4112 static int add_root_to_pending(struct extent_buffer *buf,
4113 struct cache_tree *extent_cache,
4114 struct cache_tree *pending,
4115 struct cache_tree *seen,
4116 struct cache_tree *nodes,
4117 struct btrfs_key *root_key)
4119 if (btrfs_header_level(buf) > 0)
4120 add_pending(nodes, seen, buf->start, buf->len);
4122 add_pending(pending, seen, buf->start, buf->len);
4123 add_extent_rec(extent_cache, NULL, 0, buf->start, buf->len,
4124 0, 1, 1, 0, 1, 0, buf->len);
4126 if (root_key->objectid == BTRFS_TREE_RELOC_OBJECTID ||
4127 btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
4128 add_tree_backref(extent_cache, buf->start, buf->start,
4131 add_tree_backref(extent_cache, buf->start, 0,
4132 root_key->objectid, 1);
4136 /* as we fix the tree, we might be deleting blocks that
4137 * we're tracking for repair. This hook makes sure we
4138 * remove any backrefs for blocks as we are fixing them.
4140 static int free_extent_hook(struct btrfs_trans_handle *trans,
4141 struct btrfs_root *root,
4142 u64 bytenr, u64 num_bytes, u64 parent,
4143 u64 root_objectid, u64 owner, u64 offset,
4146 struct extent_record *rec;
4147 struct cache_extent *cache;
4149 struct cache_tree *extent_cache = root->fs_info->fsck_extent_cache;
4151 is_data = owner >= BTRFS_FIRST_FREE_OBJECTID;
4152 cache = lookup_cache_extent(extent_cache, bytenr, num_bytes);
4156 rec = container_of(cache, struct extent_record, cache);
4158 struct data_backref *back;
4159 back = find_data_backref(rec, parent, root_objectid, owner,
4160 offset, 1, bytenr, num_bytes);
4163 if (back->node.found_ref) {
4164 back->found_ref -= refs_to_drop;
4166 rec->refs -= refs_to_drop;
4168 if (back->node.found_extent_tree) {
4169 back->num_refs -= refs_to_drop;
4170 if (rec->extent_item_refs)
4171 rec->extent_item_refs -= refs_to_drop;
4173 if (back->found_ref == 0)
4174 back->node.found_ref = 0;
4175 if (back->num_refs == 0)
4176 back->node.found_extent_tree = 0;
4178 if (!back->node.found_extent_tree && back->node.found_ref) {
4179 list_del(&back->node.list);
4183 struct tree_backref *back;
4184 back = find_tree_backref(rec, parent, root_objectid);
4187 if (back->node.found_ref) {
4190 back->node.found_ref = 0;
4192 if (back->node.found_extent_tree) {
4193 if (rec->extent_item_refs)
4194 rec->extent_item_refs--;
4195 back->node.found_extent_tree = 0;
4197 if (!back->node.found_extent_tree && back->node.found_ref) {
4198 list_del(&back->node.list);
4202 maybe_free_extent_rec(extent_cache, rec);
4207 static int delete_extent_records(struct btrfs_trans_handle *trans,
4208 struct btrfs_root *root,
4209 struct btrfs_path *path,
4210 u64 bytenr, u64 new_len)
4212 struct btrfs_key key;
4213 struct btrfs_key found_key;
4214 struct extent_buffer *leaf;
4219 key.objectid = bytenr;
4221 key.offset = (u64)-1;
4224 ret = btrfs_search_slot(trans, root->fs_info->extent_root,
4231 if (path->slots[0] == 0)
4237 leaf = path->nodes[0];
4238 slot = path->slots[0];
4240 btrfs_item_key_to_cpu(leaf, &found_key, slot);
4241 if (found_key.objectid != bytenr)
4244 if (found_key.type != BTRFS_EXTENT_ITEM_KEY &&
4245 found_key.type != BTRFS_METADATA_ITEM_KEY &&
4246 found_key.type != BTRFS_TREE_BLOCK_REF_KEY &&
4247 found_key.type != BTRFS_EXTENT_DATA_REF_KEY &&
4248 found_key.type != BTRFS_EXTENT_REF_V0_KEY &&
4249 found_key.type != BTRFS_SHARED_BLOCK_REF_KEY &&
4250 found_key.type != BTRFS_SHARED_DATA_REF_KEY) {
4251 btrfs_release_path(path);
4252 if (found_key.type == 0) {
4253 if (found_key.offset == 0)
4255 key.offset = found_key.offset - 1;
4256 key.type = found_key.type;
4258 key.type = found_key.type - 1;
4259 key.offset = (u64)-1;
4263 fprintf(stderr, "repair deleting extent record: key %Lu %u %Lu\n",
4264 found_key.objectid, found_key.type, found_key.offset);
4266 ret = btrfs_del_item(trans, root->fs_info->extent_root, path);
4269 btrfs_release_path(path);
4271 if (found_key.type == BTRFS_EXTENT_ITEM_KEY ||
4272 found_key.type == BTRFS_METADATA_ITEM_KEY) {
4273 u64 bytes = (found_key.type == BTRFS_EXTENT_ITEM_KEY) ?
4274 found_key.offset : root->leafsize;
4276 ret = btrfs_update_block_group(trans, root, bytenr,
4283 btrfs_release_path(path);
4288 * for a single backref, this will allocate a new extent
4289 * and add the backref to it.
4291 static int record_extent(struct btrfs_trans_handle *trans,
4292 struct btrfs_fs_info *info,
4293 struct btrfs_path *path,
4294 struct extent_record *rec,
4295 struct extent_backref *back,
4296 int allocated, u64 flags)
4299 struct btrfs_root *extent_root = info->extent_root;
4300 struct extent_buffer *leaf;
4301 struct btrfs_key ins_key;
4302 struct btrfs_extent_item *ei;
4303 struct tree_backref *tback;
4304 struct data_backref *dback;
4305 struct btrfs_tree_block_info *bi;
4308 rec->max_size = max_t(u64, rec->max_size,
4309 info->extent_root->leafsize);
4312 u32 item_size = sizeof(*ei);
4315 item_size += sizeof(*bi);
4317 ins_key.objectid = rec->start;
4318 ins_key.offset = rec->max_size;
4319 ins_key.type = BTRFS_EXTENT_ITEM_KEY;
4321 ret = btrfs_insert_empty_item(trans, extent_root, path,
4322 &ins_key, item_size);
4326 leaf = path->nodes[0];
4327 ei = btrfs_item_ptr(leaf, path->slots[0],
4328 struct btrfs_extent_item);
4330 btrfs_set_extent_refs(leaf, ei, 0);
4331 btrfs_set_extent_generation(leaf, ei, rec->generation);
4333 if (back->is_data) {
4334 btrfs_set_extent_flags(leaf, ei,
4335 BTRFS_EXTENT_FLAG_DATA);
4337 struct btrfs_disk_key copy_key;;
4339 tback = (struct tree_backref *)back;
4340 bi = (struct btrfs_tree_block_info *)(ei + 1);
4341 memset_extent_buffer(leaf, 0, (unsigned long)bi,
4344 btrfs_set_disk_key_objectid(©_key,
4345 rec->info_objectid);
4346 btrfs_set_disk_key_type(©_key, 0);
4347 btrfs_set_disk_key_offset(©_key, 0);
4349 btrfs_set_tree_block_level(leaf, bi, rec->info_level);
4350 btrfs_set_tree_block_key(leaf, bi, ©_key);
4352 btrfs_set_extent_flags(leaf, ei,
4353 BTRFS_EXTENT_FLAG_TREE_BLOCK | flags);
4356 btrfs_mark_buffer_dirty(leaf);
4357 ret = btrfs_update_block_group(trans, extent_root, rec->start,
4358 rec->max_size, 1, 0);
4361 btrfs_release_path(path);
4364 if (back->is_data) {
4368 dback = (struct data_backref *)back;
4369 if (back->full_backref)
4370 parent = dback->parent;
4374 for (i = 0; i < dback->found_ref; i++) {
4375 /* if parent != 0, we're doing a full backref
4376 * passing BTRFS_FIRST_FREE_OBJECTID as the owner
4377 * just makes the backref allocator create a data
4380 ret = btrfs_inc_extent_ref(trans, info->extent_root,
4381 rec->start, rec->max_size,
4385 BTRFS_FIRST_FREE_OBJECTID :
4391 fprintf(stderr, "adding new data backref"
4392 " on %llu %s %llu owner %llu"
4393 " offset %llu found %d\n",
4394 (unsigned long long)rec->start,
4395 back->full_backref ?
4397 back->full_backref ?
4398 (unsigned long long)parent :
4399 (unsigned long long)dback->root,
4400 (unsigned long long)dback->owner,
4401 (unsigned long long)dback->offset,
4406 tback = (struct tree_backref *)back;
4407 if (back->full_backref)
4408 parent = tback->parent;
4412 ret = btrfs_inc_extent_ref(trans, info->extent_root,
4413 rec->start, rec->max_size,
4414 parent, tback->root, 0, 0);
4415 fprintf(stderr, "adding new tree backref on "
4416 "start %llu len %llu parent %llu root %llu\n",
4417 rec->start, rec->max_size, tback->parent, tback->root);
4422 btrfs_release_path(path);
4426 struct extent_entry {
4431 struct list_head list;
4434 static struct extent_entry *find_entry(struct list_head *entries,
4435 u64 bytenr, u64 bytes)
4437 struct extent_entry *entry = NULL;
4439 list_for_each_entry(entry, entries, list) {
4440 if (entry->bytenr == bytenr && entry->bytes == bytes)
4447 static struct extent_entry *find_most_right_entry(struct list_head *entries)
4449 struct extent_entry *entry, *best = NULL, *prev = NULL;
4451 list_for_each_entry(entry, entries, list) {
4458 * If there are as many broken entries as entries then we know
4459 * not to trust this particular entry.
4461 if (entry->broken == entry->count)
4465 * If our current entry == best then we can't be sure our best
4466 * is really the best, so we need to keep searching.
4468 if (best && best->count == entry->count) {
4474 /* Prev == entry, not good enough, have to keep searching */
4475 if (!prev->broken && prev->count == entry->count)
4479 best = (prev->count > entry->count) ? prev : entry;
4480 else if (best->count < entry->count)
4488 static int repair_ref(struct btrfs_trans_handle *trans,
4489 struct btrfs_fs_info *info, struct btrfs_path *path,
4490 struct data_backref *dback, struct extent_entry *entry)
4492 struct btrfs_root *root;
4493 struct btrfs_file_extent_item *fi;
4494 struct extent_buffer *leaf;
4495 struct btrfs_key key;
4499 key.objectid = dback->root;
4500 key.type = BTRFS_ROOT_ITEM_KEY;
4501 key.offset = (u64)-1;
4502 root = btrfs_read_fs_root(info, &key);
4504 fprintf(stderr, "Couldn't find root for our ref\n");
4509 * The backref points to the original offset of the extent if it was
4510 * split, so we need to search down to the offset we have and then walk
4511 * forward until we find the backref we're looking for.
4513 key.objectid = dback->owner;
4514 key.type = BTRFS_EXTENT_DATA_KEY;
4515 key.offset = dback->offset;
4516 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
4518 fprintf(stderr, "Error looking up ref %d\n", ret);
4523 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
4524 ret = btrfs_next_leaf(root, path);
4526 fprintf(stderr, "Couldn't find our ref, next\n");
4530 leaf = path->nodes[0];
4531 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
4532 if (key.objectid != dback->owner ||
4533 key.type != BTRFS_EXTENT_DATA_KEY) {
4534 fprintf(stderr, "Couldn't find our ref, search\n");
4537 fi = btrfs_item_ptr(leaf, path->slots[0],
4538 struct btrfs_file_extent_item);
4539 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
4540 bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
4542 if (bytenr == dback->disk_bytenr && bytes == dback->bytes)
4547 btrfs_release_path(path);
4550 * Have to make sure that this root gets updated when we commit the
4553 root->track_dirty = 1;
4554 if (root->last_trans != trans->transid) {
4555 root->last_trans = trans->transid;
4556 root->commit_root = root->node;
4557 extent_buffer_get(root->node);
4561 * Ok we have the key of the file extent we want to fix, now we can cow
4562 * down to the thing and fix it.
4564 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
4566 fprintf(stderr, "Error cowing down to ref [%Lu, %u, %Lu]: %d\n",
4567 key.objectid, key.type, key.offset, ret);
4571 fprintf(stderr, "Well that's odd, we just found this key "
4572 "[%Lu, %u, %Lu]\n", key.objectid, key.type,
4576 leaf = path->nodes[0];
4577 fi = btrfs_item_ptr(leaf, path->slots[0],
4578 struct btrfs_file_extent_item);
4580 if (btrfs_file_extent_compression(leaf, fi) &&
4581 dback->disk_bytenr != entry->bytenr) {
4582 fprintf(stderr, "Ref doesn't match the record start and is "
4583 "compressed, please take a btrfs-image of this file "
4584 "system and send it to a btrfs developer so they can "
4585 "complete this functionality for bytenr %Lu\n",
4586 dback->disk_bytenr);
4590 if (dback->node.broken && dback->disk_bytenr != entry->bytenr) {
4591 btrfs_set_file_extent_disk_bytenr(leaf, fi, entry->bytenr);
4592 } else if (dback->disk_bytenr > entry->bytenr) {
4593 u64 off_diff, offset;
4595 off_diff = dback->disk_bytenr - entry->bytenr;
4596 offset = btrfs_file_extent_offset(leaf, fi);
4597 if (dback->disk_bytenr + offset +
4598 btrfs_file_extent_num_bytes(leaf, fi) >
4599 entry->bytenr + entry->bytes) {
4600 fprintf(stderr, "Ref is past the entry end, please "
4601 "take a btrfs-image of this file system and "
4602 "send it to a btrfs developer, ref %Lu\n",
4603 dback->disk_bytenr);
4607 btrfs_set_file_extent_disk_bytenr(leaf, fi, entry->bytenr);
4608 btrfs_set_file_extent_offset(leaf, fi, offset);
4609 } else if (dback->disk_bytenr < entry->bytenr) {
4612 offset = btrfs_file_extent_offset(leaf, fi);
4613 if (dback->disk_bytenr + offset < entry->bytenr) {
4614 fprintf(stderr, "Ref is before the entry start, please"
4615 " take a btrfs-image of this file system and "
4616 "send it to a btrfs developer, ref %Lu\n",
4617 dback->disk_bytenr);
4621 offset += dback->disk_bytenr;
4622 offset -= entry->bytenr;
4623 btrfs_set_file_extent_disk_bytenr(leaf, fi, entry->bytenr);
4624 btrfs_set_file_extent_offset(leaf, fi, offset);
4627 btrfs_set_file_extent_disk_num_bytes(leaf, fi, entry->bytes);
4630 * Chances are if disk_num_bytes were wrong then so is ram_bytes, but
4631 * only do this if we aren't using compression, otherwise it's a
4634 if (!btrfs_file_extent_compression(leaf, fi))
4635 btrfs_set_file_extent_ram_bytes(leaf, fi, entry->bytes);
4637 printf("ram bytes may be wrong?\n");
4638 btrfs_mark_buffer_dirty(leaf);
4639 btrfs_release_path(path);
4643 static int verify_backrefs(struct btrfs_trans_handle *trans,
4644 struct btrfs_fs_info *info, struct btrfs_path *path,
4645 struct extent_record *rec)
4647 struct extent_backref *back;
4648 struct data_backref *dback;
4649 struct extent_entry *entry, *best = NULL;
4652 int broken_entries = 0;
4657 * Metadata is easy and the backrefs should always agree on bytenr and
4658 * size, if not we've got bigger issues.
4663 list_for_each_entry(back, &rec->backrefs, list) {
4664 dback = (struct data_backref *)back;
4666 * We only pay attention to backrefs that we found a real
4669 if (dback->found_ref == 0)
4671 if (back->full_backref)
4675 * For now we only catch when the bytes don't match, not the
4676 * bytenr. We can easily do this at the same time, but I want
4677 * to have a fs image to test on before we just add repair
4678 * functionality willy-nilly so we know we won't screw up the
4682 entry = find_entry(&entries, dback->disk_bytenr,
4685 entry = malloc(sizeof(struct extent_entry));
4690 memset(entry, 0, sizeof(*entry));
4691 entry->bytenr = dback->disk_bytenr;
4692 entry->bytes = dback->bytes;
4693 list_add_tail(&entry->list, &entries);
4698 * If we only have on entry we may think the entries agree when
4699 * in reality they don't so we have to do some extra checking.
4701 if (dback->disk_bytenr != rec->start ||
4702 dback->bytes != rec->nr || back->broken)
4713 /* Yay all the backrefs agree, carry on good sir */
4714 if (nr_entries <= 1 && !mismatch)
4717 fprintf(stderr, "attempting to repair backref discrepency for bytenr "
4718 "%Lu\n", rec->start);
4721 * First we want to see if the backrefs can agree amongst themselves who
4722 * is right, so figure out which one of the entries has the highest
4725 best = find_most_right_entry(&entries);
4728 * Ok so we may have an even split between what the backrefs think, so
4729 * this is where we use the extent ref to see what it thinks.
4732 entry = find_entry(&entries, rec->start, rec->nr);
4733 if (!entry && (!broken_entries || !rec->found_rec)) {
4734 fprintf(stderr, "Backrefs don't agree with eachother "
4735 "and extent record doesn't agree with anybody,"
4736 " so we can't fix bytenr %Lu bytes %Lu\n",
4737 rec->start, rec->nr);
4740 } else if (!entry) {
4742 * Ok our backrefs were broken, we'll assume this is the
4743 * correct value and add an entry for this range.
4745 entry = malloc(sizeof(struct extent_entry));
4750 memset(entry, 0, sizeof(*entry));
4751 entry->bytenr = rec->start;
4752 entry->bytes = rec->nr;
4753 list_add_tail(&entry->list, &entries);
4757 best = find_most_right_entry(&entries);
4759 fprintf(stderr, "Backrefs and extent record evenly "
4760 "split on who is right, this is going to "
4761 "require user input to fix bytenr %Lu bytes "
4762 "%Lu\n", rec->start, rec->nr);
4769 * I don't think this can happen currently as we'll abort() if we catch
4770 * this case higher up, but in case somebody removes that we still can't
4771 * deal with it properly here yet, so just bail out of that's the case.
4773 if (best->bytenr != rec->start) {
4774 fprintf(stderr, "Extent start and backref starts don't match, "
4775 "please use btrfs-image on this file system and send "
4776 "it to a btrfs developer so they can make fsck fix "
4777 "this particular case. bytenr is %Lu, bytes is %Lu\n",
4778 rec->start, rec->nr);
4784 * Ok great we all agreed on an extent record, let's go find the real
4785 * references and fix up the ones that don't match.
4787 list_for_each_entry(back, &rec->backrefs, list) {
4788 dback = (struct data_backref *)back;
4791 * Still ignoring backrefs that don't have a real ref attached
4794 if (dback->found_ref == 0)
4796 if (back->full_backref)
4799 if (dback->bytes == best->bytes &&
4800 dback->disk_bytenr == best->bytenr)
4803 ret = repair_ref(trans, info, path, dback, best);
4809 * Ok we messed with the actual refs, which means we need to drop our
4810 * entire cache and go back and rescan. I know this is a huge pain and
4811 * adds a lot of extra work, but it's the only way to be safe. Once all
4812 * the backrefs agree we may not need to do anything to the extent
4817 while (!list_empty(&entries)) {
4818 entry = list_entry(entries.next, struct extent_entry, list);
4819 list_del_init(&entry->list);
4825 static int process_duplicates(struct btrfs_root *root,
4826 struct cache_tree *extent_cache,
4827 struct extent_record *rec)
4829 struct extent_record *good, *tmp;
4830 struct cache_extent *cache;
4834 * If we found a extent record for this extent then return, or if we
4835 * have more than one duplicate we are likely going to need to delete
4838 if (rec->found_rec || rec->num_duplicates > 1)
4841 /* Shouldn't happen but just in case */
4842 BUG_ON(!rec->num_duplicates);
4845 * So this happens if we end up with a backref that doesn't match the
4846 * actual extent entry. So either the backref is bad or the extent
4847 * entry is bad. Either way we want to have the extent_record actually
4848 * reflect what we found in the extent_tree, so we need to take the
4849 * duplicate out and use that as the extent_record since the only way we
4850 * get a duplicate is if we find a real life BTRFS_EXTENT_ITEM_KEY.
4852 remove_cache_extent(extent_cache, &rec->cache);
4854 good = list_entry(rec->dups.next, struct extent_record, list);
4855 list_del_init(&good->list);
4856 INIT_LIST_HEAD(&good->backrefs);
4857 INIT_LIST_HEAD(&good->dups);
4858 good->cache.start = good->start;
4859 good->cache.size = good->nr;
4860 good->content_checked = 0;
4861 good->owner_ref_checked = 0;
4862 good->num_duplicates = 0;
4863 good->refs = rec->refs;
4864 list_splice_init(&rec->backrefs, &good->backrefs);
4866 cache = lookup_cache_extent(extent_cache, good->start,
4870 tmp = container_of(cache, struct extent_record, cache);
4873 * If we find another overlapping extent and it's found_rec is
4874 * set then it's a duplicate and we need to try and delete
4877 if (tmp->found_rec || tmp->num_duplicates > 0) {
4878 if (list_empty(&good->list))
4879 list_add_tail(&good->list,
4880 &duplicate_extents);
4881 good->num_duplicates += tmp->num_duplicates + 1;
4882 list_splice_init(&tmp->dups, &good->dups);
4883 list_del_init(&tmp->list);
4884 list_add_tail(&tmp->list, &good->dups);
4885 remove_cache_extent(extent_cache, &tmp->cache);
4890 * Ok we have another non extent item backed extent rec, so lets
4891 * just add it to this extent and carry on like we did above.
4893 good->refs += tmp->refs;
4894 list_splice_init(&tmp->backrefs, &good->backrefs);
4895 remove_cache_extent(extent_cache, &tmp->cache);
4898 ret = insert_cache_extent(extent_cache, &good->cache);
4901 return good->num_duplicates ? 0 : 1;
4904 static int delete_duplicate_records(struct btrfs_trans_handle *trans,
4905 struct btrfs_root *root,
4906 struct extent_record *rec)
4908 LIST_HEAD(delete_list);
4909 struct btrfs_path *path;
4910 struct extent_record *tmp, *good, *n;
4913 struct btrfs_key key;
4915 path = btrfs_alloc_path();
4922 /* Find the record that covers all of the duplicates. */
4923 list_for_each_entry(tmp, &rec->dups, list) {
4924 if (good->start < tmp->start)
4926 if (good->nr > tmp->nr)
4929 if (tmp->start + tmp->nr < good->start + good->nr) {
4930 fprintf(stderr, "Ok we have overlapping extents that "
4931 "aren't completely covered by eachother, this "
4932 "is going to require more careful thought. "
4933 "The extents are [%Lu-%Lu] and [%Lu-%Lu]\n",
4934 tmp->start, tmp->nr, good->start, good->nr);
4941 list_add_tail(&rec->list, &delete_list);
4943 list_for_each_entry_safe(tmp, n, &rec->dups, list) {
4946 list_move_tail(&tmp->list, &delete_list);
4949 root = root->fs_info->extent_root;
4950 list_for_each_entry(tmp, &delete_list, list) {
4951 if (tmp->found_rec == 0)
4953 key.objectid = tmp->start;
4954 key.type = BTRFS_EXTENT_ITEM_KEY;
4955 key.offset = tmp->nr;
4957 /* Shouldn't happen but just in case */
4958 if (tmp->metadata) {
4959 fprintf(stderr, "Well this shouldn't happen, extent "
4960 "record overlaps but is metadata? "
4961 "[%Lu, %Lu]\n", tmp->start, tmp->nr);
4965 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
4971 ret = btrfs_del_item(trans, root, path);
4974 btrfs_release_path(path);
4979 while (!list_empty(&delete_list)) {
4980 tmp = list_entry(delete_list.next, struct extent_record, list);
4981 list_del_init(&tmp->list);
4987 while (!list_empty(&rec->dups)) {
4988 tmp = list_entry(rec->dups.next, struct extent_record, list);
4989 list_del_init(&tmp->list);
4993 btrfs_free_path(path);
4995 if (!ret && !nr_del)
4996 rec->num_duplicates = 0;
4998 return ret ? ret : nr_del;
5001 static int find_possible_backrefs(struct btrfs_trans_handle *trans,
5002 struct btrfs_fs_info *info,
5003 struct btrfs_path *path,
5004 struct cache_tree *extent_cache,
5005 struct extent_record *rec)
5007 struct btrfs_root *root;
5008 struct extent_backref *back;
5009 struct data_backref *dback;
5010 struct cache_extent *cache;
5011 struct btrfs_file_extent_item *fi;
5012 struct btrfs_key key;
5016 list_for_each_entry(back, &rec->backrefs, list) {
5017 dback = (struct data_backref *)back;
5019 /* We found this one, we don't need to do a lookup */
5020 if (dback->found_ref)
5022 /* Don't care about full backrefs (poor unloved backrefs) */
5023 if (back->full_backref)
5025 key.objectid = dback->root;
5026 key.type = BTRFS_ROOT_ITEM_KEY;
5027 key.offset = (u64)-1;
5029 root = btrfs_read_fs_root(info, &key);
5031 /* No root, definitely a bad ref, skip */
5032 if (IS_ERR(root) && PTR_ERR(root) == -ENOENT)
5034 /* Other err, exit */
5036 return PTR_ERR(root);
5038 key.objectid = dback->owner;
5039 key.type = BTRFS_EXTENT_DATA_KEY;
5040 key.offset = dback->offset;
5041 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5043 btrfs_release_path(path);
5046 /* Didn't find it, we can carry on */
5051 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
5052 struct btrfs_file_extent_item);
5053 bytenr = btrfs_file_extent_disk_bytenr(path->nodes[0], fi);
5054 bytes = btrfs_file_extent_disk_num_bytes(path->nodes[0], fi);
5055 btrfs_release_path(path);
5056 cache = lookup_cache_extent(extent_cache, bytenr, 1);
5058 struct extent_record *tmp;
5059 tmp = container_of(cache, struct extent_record, cache);
5062 * If we found an extent record for the bytenr for this
5063 * particular backref then we can't add it to our
5064 * current extent record. We only want to add backrefs
5065 * that don't have a corresponding extent item in the
5066 * extent tree since they likely belong to this record
5067 * and we need to fix it if it doesn't match bytenrs.
5073 dback->found_ref += 1;
5074 dback->disk_bytenr = bytenr;
5075 dback->bytes = bytes;
5078 * Set this so the verify backref code knows not to trust the
5079 * values in this backref.
5088 * when an incorrect extent item is found, this will delete
5089 * all of the existing entries for it and recreate them
5090 * based on what the tree scan found.
5092 static int fixup_extent_refs(struct btrfs_trans_handle *trans,
5093 struct btrfs_fs_info *info,
5094 struct cache_tree *extent_cache,
5095 struct extent_record *rec)
5098 struct btrfs_path *path;
5099 struct list_head *cur = rec->backrefs.next;
5100 struct cache_extent *cache;
5101 struct extent_backref *back;
5105 /* remember our flags for recreating the extent */
5106 ret = btrfs_lookup_extent_info(NULL, info->extent_root, rec->start,
5107 rec->max_size, rec->metadata, NULL,
5110 flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5112 path = btrfs_alloc_path();
5116 if (rec->refs != rec->extent_item_refs && !rec->metadata) {
5118 * Sometimes the backrefs themselves are so broken they don't
5119 * get attached to any meaningful rec, so first go back and
5120 * check any of our backrefs that we couldn't find and throw
5121 * them into the list if we find the backref so that
5122 * verify_backrefs can figure out what to do.
5124 ret = find_possible_backrefs(trans, info, path, extent_cache,
5130 /* step one, make sure all of the backrefs agree */
5131 ret = verify_backrefs(trans, info, path, rec);
5135 /* step two, delete all the existing records */
5136 ret = delete_extent_records(trans, info->extent_root, path,
5137 rec->start, rec->max_size);
5142 /* was this block corrupt? If so, don't add references to it */
5143 cache = lookup_cache_extent(info->corrupt_blocks,
5144 rec->start, rec->max_size);
5150 /* step three, recreate all the refs we did find */
5151 while(cur != &rec->backrefs) {
5152 back = list_entry(cur, struct extent_backref, list);
5156 * if we didn't find any references, don't create a
5159 if (!back->found_ref)
5162 ret = record_extent(trans, info, path, rec, back, allocated, flags);
5169 btrfs_free_path(path);
5173 /* right now we only prune from the extent allocation tree */
5174 static int prune_one_block(struct btrfs_trans_handle *trans,
5175 struct btrfs_fs_info *info,
5176 struct btrfs_corrupt_block *corrupt)
5179 struct btrfs_path path;
5180 struct extent_buffer *eb;
5184 int level = corrupt->level + 1;
5186 btrfs_init_path(&path);
5188 /* we want to stop at the parent to our busted block */
5189 path.lowest_level = level;
5191 ret = btrfs_search_slot(trans, info->extent_root,
5192 &corrupt->key, &path, -1, 1);
5197 eb = path.nodes[level];
5204 * hopefully the search gave us the block we want to prune,
5205 * lets try that first
5207 slot = path.slots[level];
5208 found = btrfs_node_blockptr(eb, slot);
5209 if (found == corrupt->cache.start)
5212 nritems = btrfs_header_nritems(eb);
5214 /* the search failed, lets scan this node and hope we find it */
5215 for (slot = 0; slot < nritems; slot++) {
5216 found = btrfs_node_blockptr(eb, slot);
5217 if (found == corrupt->cache.start)
5221 * we couldn't find the bad block. TODO, search all the nodes for pointers
5224 if (eb == info->extent_root->node) {
5229 btrfs_release_path(&path);
5234 printk("deleting pointer to block %Lu\n", corrupt->cache.start);
5235 ret = btrfs_del_ptr(trans, info->extent_root, &path, level, slot);
5238 btrfs_release_path(&path);
5242 static int prune_corrupt_blocks(struct btrfs_trans_handle *trans,
5243 struct btrfs_fs_info *info)
5245 struct cache_extent *cache;
5246 struct btrfs_corrupt_block *corrupt;
5248 cache = search_cache_extent(info->corrupt_blocks, 0);
5252 corrupt = container_of(cache, struct btrfs_corrupt_block, cache);
5253 prune_one_block(trans, info, corrupt);
5254 cache = next_cache_extent(cache);
5259 static void free_corrupt_block(struct cache_extent *cache)
5261 struct btrfs_corrupt_block *corrupt;
5263 corrupt = container_of(cache, struct btrfs_corrupt_block, cache);
5267 FREE_EXTENT_CACHE_BASED_TREE(corrupt_blocks, free_corrupt_block);
5269 static void reset_cached_block_groups(struct btrfs_fs_info *fs_info)
5271 struct btrfs_block_group_cache *cache;
5276 ret = find_first_extent_bit(&fs_info->free_space_cache, 0,
5277 &start, &end, EXTENT_DIRTY);
5280 clear_extent_dirty(&fs_info->free_space_cache, start, end,
5286 cache = btrfs_lookup_first_block_group(fs_info, start);
5291 start = cache->key.objectid + cache->key.offset;
5295 static int check_extent_refs(struct btrfs_trans_handle *trans,
5296 struct btrfs_root *root,
5297 struct cache_tree *extent_cache)
5299 struct extent_record *rec;
5300 struct cache_extent *cache;
5308 * if we're doing a repair, we have to make sure
5309 * we don't allocate from the problem extents.
5310 * In the worst case, this will be all the
5313 cache = search_cache_extent(extent_cache, 0);
5315 rec = container_of(cache, struct extent_record, cache);
5316 btrfs_pin_extent(root->fs_info,
5317 rec->start, rec->max_size);
5318 cache = next_cache_extent(cache);
5321 /* pin down all the corrupted blocks too */
5322 cache = search_cache_extent(root->fs_info->corrupt_blocks, 0);
5324 btrfs_pin_extent(root->fs_info,
5325 cache->start, cache->size);
5326 cache = next_cache_extent(cache);
5328 prune_corrupt_blocks(trans, root->fs_info);
5329 reset_cached_block_groups(root->fs_info);
5333 * We need to delete any duplicate entries we find first otherwise we
5334 * could mess up the extent tree when we have backrefs that actually
5335 * belong to a different extent item and not the weird duplicate one.
5337 while (repair && !list_empty(&duplicate_extents)) {
5338 rec = list_entry(duplicate_extents.next, struct extent_record,
5340 list_del_init(&rec->list);
5342 /* Sometimes we can find a backref before we find an actual
5343 * extent, so we need to process it a little bit to see if there
5344 * truly are multiple EXTENT_ITEM_KEY's for the same range, or
5345 * if this is a backref screwup. If we need to delete stuff
5346 * process_duplicates() will return 0, otherwise it will return
5349 if (process_duplicates(root, extent_cache, rec))
5351 ret = delete_duplicate_records(trans, root, rec);
5355 * delete_duplicate_records will return the number of entries
5356 * deleted, so if it's greater than 0 then we know we actually
5357 * did something and we need to remove.
5368 cache = search_cache_extent(extent_cache, 0);
5371 rec = container_of(cache, struct extent_record, cache);
5372 if (rec->num_duplicates) {
5373 fprintf(stderr, "extent item %llu has multiple extent "
5374 "items\n", (unsigned long long)rec->start);
5378 if (rec->refs != rec->extent_item_refs) {
5379 fprintf(stderr, "ref mismatch on [%llu %llu] ",
5380 (unsigned long long)rec->start,
5381 (unsigned long long)rec->nr);
5382 fprintf(stderr, "extent item %llu, found %llu\n",
5383 (unsigned long long)rec->extent_item_refs,
5384 (unsigned long long)rec->refs);
5385 if (!fixed && repair) {
5386 ret = fixup_extent_refs(trans, root->fs_info,
5395 if (all_backpointers_checked(rec, 1)) {
5396 fprintf(stderr, "backpointer mismatch on [%llu %llu]\n",
5397 (unsigned long long)rec->start,
5398 (unsigned long long)rec->nr);
5400 if (!fixed && repair) {
5401 ret = fixup_extent_refs(trans, root->fs_info,
5410 if (!rec->owner_ref_checked) {
5411 fprintf(stderr, "owner ref check failed [%llu %llu]\n",
5412 (unsigned long long)rec->start,
5413 (unsigned long long)rec->nr);
5414 if (!fixed && repair) {
5415 ret = fixup_extent_refs(trans, root->fs_info,
5424 remove_cache_extent(extent_cache, cache);
5425 free_all_extent_backrefs(rec);
5430 if (ret && ret != -EAGAIN) {
5431 fprintf(stderr, "failed to repair damaged filesystem, aborting\n");
5434 btrfs_fix_block_accounting(trans, root);
5437 fprintf(stderr, "repaired damaged extent references\n");
5443 u64 calc_stripe_length(u64 type, u64 length, int num_stripes)
5447 if (type & BTRFS_BLOCK_GROUP_RAID0) {
5448 stripe_size = length;
5449 stripe_size /= num_stripes;
5450 } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
5451 stripe_size = length * 2;
5452 stripe_size /= num_stripes;
5453 } else if (type & BTRFS_BLOCK_GROUP_RAID5) {
5454 stripe_size = length;
5455 stripe_size /= (num_stripes - 1);
5456 } else if (type & BTRFS_BLOCK_GROUP_RAID6) {
5457 stripe_size = length;
5458 stripe_size /= (num_stripes - 2);
5460 stripe_size = length;
5465 static int check_chunk_refs(struct chunk_record *chunk_rec,
5466 struct block_group_tree *block_group_cache,
5467 struct device_extent_tree *dev_extent_cache,
5470 struct cache_extent *block_group_item;
5471 struct block_group_record *block_group_rec;
5472 struct cache_extent *dev_extent_item;
5473 struct device_extent_record *dev_extent_rec;
5480 block_group_item = lookup_cache_extent(&block_group_cache->tree,
5483 if (block_group_item) {
5484 block_group_rec = container_of(block_group_item,
5485 struct block_group_record,
5487 if (chunk_rec->length != block_group_rec->offset ||
5488 chunk_rec->offset != block_group_rec->objectid ||
5489 chunk_rec->type_flags != block_group_rec->flags) {
5492 "Chunk[%llu, %u, %llu]: length(%llu), offset(%llu), type(%llu) mismatch with block group[%llu, %u, %llu]: offset(%llu), objectid(%llu), flags(%llu)\n",
5493 chunk_rec->objectid,
5498 chunk_rec->type_flags,
5499 block_group_rec->objectid,
5500 block_group_rec->type,
5501 block_group_rec->offset,
5502 block_group_rec->offset,
5503 block_group_rec->objectid,
5504 block_group_rec->flags);
5507 list_del_init(&block_group_rec->list);
5508 chunk_rec->bg_rec = block_group_rec;
5513 "Chunk[%llu, %u, %llu]: length(%llu), offset(%llu), type(%llu) is not found in block group\n",
5514 chunk_rec->objectid,
5519 chunk_rec->type_flags);
5523 length = calc_stripe_length(chunk_rec->type_flags, chunk_rec->length,
5524 chunk_rec->num_stripes);
5525 for (i = 0; i < chunk_rec->num_stripes; ++i) {
5526 devid = chunk_rec->stripes[i].devid;
5527 offset = chunk_rec->stripes[i].offset;
5528 dev_extent_item = lookup_cache_extent2(&dev_extent_cache->tree,
5529 devid, offset, length);
5530 if (dev_extent_item) {
5531 dev_extent_rec = container_of(dev_extent_item,
5532 struct device_extent_record,
5534 if (dev_extent_rec->objectid != devid ||
5535 dev_extent_rec->offset != offset ||
5536 dev_extent_rec->chunk_offset != chunk_rec->offset ||
5537 dev_extent_rec->length != length) {
5540 "Chunk[%llu, %u, %llu] stripe[%llu, %llu] dismatch dev extent[%llu, %llu, %llu]\n",
5541 chunk_rec->objectid,
5544 chunk_rec->stripes[i].devid,
5545 chunk_rec->stripes[i].offset,
5546 dev_extent_rec->objectid,
5547 dev_extent_rec->offset,
5548 dev_extent_rec->length);
5551 list_move(&dev_extent_rec->chunk_list,
5552 &chunk_rec->dextents);
5557 "Chunk[%llu, %u, %llu] stripe[%llu, %llu] is not found in dev extent\n",
5558 chunk_rec->objectid,
5561 chunk_rec->stripes[i].devid,
5562 chunk_rec->stripes[i].offset);
5569 /* check btrfs_chunk -> btrfs_dev_extent / btrfs_block_group_item */
5570 int check_chunks(struct cache_tree *chunk_cache,
5571 struct block_group_tree *block_group_cache,
5572 struct device_extent_tree *dev_extent_cache,
5573 struct list_head *good, struct list_head *bad, int silent)
5575 struct cache_extent *chunk_item;
5576 struct chunk_record *chunk_rec;
5577 struct block_group_record *bg_rec;
5578 struct device_extent_record *dext_rec;
5582 chunk_item = first_cache_extent(chunk_cache);
5583 while (chunk_item) {
5584 chunk_rec = container_of(chunk_item, struct chunk_record,
5586 err = check_chunk_refs(chunk_rec, block_group_cache,
5587 dev_extent_cache, silent);
5591 list_add_tail(&chunk_rec->list, bad);
5594 list_add_tail(&chunk_rec->list, good);
5597 chunk_item = next_cache_extent(chunk_item);
5600 list_for_each_entry(bg_rec, &block_group_cache->block_groups, list) {
5603 "Block group[%llu, %llu] (flags = %llu) didn't find the relative chunk.\n",
5611 list_for_each_entry(dext_rec, &dev_extent_cache->no_chunk_orphans,
5615 "Device extent[%llu, %llu, %llu] didn't find the relative chunk.\n",
5626 static int check_device_used(struct device_record *dev_rec,
5627 struct device_extent_tree *dext_cache)
5629 struct cache_extent *cache;
5630 struct device_extent_record *dev_extent_rec;
5633 cache = search_cache_extent2(&dext_cache->tree, dev_rec->devid, 0);
5635 dev_extent_rec = container_of(cache,
5636 struct device_extent_record,
5638 if (dev_extent_rec->objectid != dev_rec->devid)
5641 list_del(&dev_extent_rec->device_list);
5642 total_byte += dev_extent_rec->length;
5643 cache = next_cache_extent(cache);
5646 if (total_byte != dev_rec->byte_used) {
5648 "Dev extent's total-byte(%llu) is not equal to byte-used(%llu) in dev[%llu, %u, %llu]\n",
5649 total_byte, dev_rec->byte_used, dev_rec->objectid,
5650 dev_rec->type, dev_rec->offset);
5657 /* check btrfs_dev_item -> btrfs_dev_extent */
5658 static int check_devices(struct rb_root *dev_cache,
5659 struct device_extent_tree *dev_extent_cache)
5661 struct rb_node *dev_node;
5662 struct device_record *dev_rec;
5663 struct device_extent_record *dext_rec;
5667 dev_node = rb_first(dev_cache);
5669 dev_rec = container_of(dev_node, struct device_record, node);
5670 err = check_device_used(dev_rec, dev_extent_cache);
5674 dev_node = rb_next(dev_node);
5676 list_for_each_entry(dext_rec, &dev_extent_cache->no_device_orphans,
5679 "Device extent[%llu, %llu, %llu] didn't find its device.\n",
5680 dext_rec->objectid, dext_rec->offset, dext_rec->length);
5687 static int check_chunks_and_extents(struct btrfs_root *root)
5689 struct rb_root dev_cache;
5690 struct cache_tree chunk_cache;
5691 struct block_group_tree block_group_cache;
5692 struct device_extent_tree dev_extent_cache;
5693 struct cache_tree extent_cache;
5694 struct cache_tree seen;
5695 struct cache_tree pending;
5696 struct cache_tree reada;
5697 struct cache_tree nodes;
5698 struct cache_tree corrupt_blocks;
5699 struct btrfs_path path;
5700 struct btrfs_key key;
5701 struct btrfs_key found_key;
5704 struct block_info *bits;
5706 struct extent_buffer *leaf;
5707 struct btrfs_trans_handle *trans = NULL;
5709 struct btrfs_root_item ri;
5710 struct list_head dropping_trees;
5712 dev_cache = RB_ROOT;
5713 cache_tree_init(&chunk_cache);
5714 block_group_tree_init(&block_group_cache);
5715 device_extent_tree_init(&dev_extent_cache);
5717 cache_tree_init(&extent_cache);
5718 cache_tree_init(&seen);
5719 cache_tree_init(&pending);
5720 cache_tree_init(&nodes);
5721 cache_tree_init(&reada);
5722 cache_tree_init(&corrupt_blocks);
5723 INIT_LIST_HEAD(&dropping_trees);
5726 trans = btrfs_start_transaction(root, 1);
5727 if (IS_ERR(trans)) {
5728 fprintf(stderr, "Error starting transaction\n");
5729 return PTR_ERR(trans);
5731 root->fs_info->fsck_extent_cache = &extent_cache;
5732 root->fs_info->free_extent_hook = free_extent_hook;
5733 root->fs_info->corrupt_blocks = &corrupt_blocks;
5737 bits = malloc(bits_nr * sizeof(struct block_info));
5744 add_root_to_pending(root->fs_info->tree_root->node,
5745 &extent_cache, &pending, &seen, &nodes,
5746 &root->fs_info->tree_root->root_key);
5748 add_root_to_pending(root->fs_info->chunk_root->node,
5749 &extent_cache, &pending, &seen, &nodes,
5750 &root->fs_info->chunk_root->root_key);
5752 btrfs_init_path(&path);
5755 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
5756 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
5760 leaf = path.nodes[0];
5761 slot = path.slots[0];
5762 if (slot >= btrfs_header_nritems(path.nodes[0])) {
5763 ret = btrfs_next_leaf(root, &path);
5766 leaf = path.nodes[0];
5767 slot = path.slots[0];
5769 btrfs_item_key_to_cpu(leaf, &found_key, path.slots[0]);
5770 if (btrfs_key_type(&found_key) == BTRFS_ROOT_ITEM_KEY) {
5771 unsigned long offset;
5772 struct extent_buffer *buf;
5774 offset = btrfs_item_ptr_offset(leaf, path.slots[0]);
5775 read_extent_buffer(leaf, &ri, offset, sizeof(ri));
5776 if (btrfs_disk_key_objectid(&ri.drop_progress) == 0) {
5777 buf = read_tree_block(root->fs_info->tree_root,
5778 btrfs_root_bytenr(&ri),
5779 btrfs_level_size(root,
5780 btrfs_root_level(&ri)),
5786 add_root_to_pending(buf, &extent_cache,
5787 &pending, &seen, &nodes,
5789 free_extent_buffer(buf);
5791 struct dropping_root_item_record *dri_rec;
5792 dri_rec = malloc(sizeof(*dri_rec));
5797 memcpy(&dri_rec->ri, &ri, sizeof(ri));
5798 memcpy(&dri_rec->found_key, &found_key,
5800 list_add_tail(&dri_rec->list, &dropping_trees);
5805 btrfs_release_path(&path);
5807 ret = run_next_block(trans, root, bits, bits_nr, &last,
5808 &pending, &seen, &reada, &nodes,
5809 &extent_cache, &chunk_cache, &dev_cache,
5810 &block_group_cache, &dev_extent_cache,
5816 while (!list_empty(&dropping_trees)) {
5817 struct dropping_root_item_record *rec;
5818 struct extent_buffer *buf;
5819 rec = list_entry(dropping_trees.next,
5820 struct dropping_root_item_record, list);
5826 buf = read_tree_block(root->fs_info->tree_root,
5827 btrfs_root_bytenr(&rec->ri),
5828 btrfs_level_size(root,
5829 btrfs_root_level(&rec->ri)), 0);
5834 add_root_to_pending(buf, &extent_cache, &pending,
5835 &seen, &nodes, &rec->found_key);
5837 ret = run_next_block(trans, root, bits, bits_nr, &last,
5838 &pending, &seen, &reada,
5839 &nodes, &extent_cache,
5840 &chunk_cache, &dev_cache,
5847 free_extent_buffer(buf);
5848 list_del(&rec->list);
5853 ret = check_extent_refs(trans, root, &extent_cache);
5854 if (ret == -EAGAIN) {
5855 ret = btrfs_commit_transaction(trans, root);
5859 trans = btrfs_start_transaction(root, 1);
5860 if (IS_ERR(trans)) {
5861 ret = PTR_ERR(trans);
5865 free_corrupt_blocks_tree(root->fs_info->corrupt_blocks);
5866 free_extent_cache_tree(&seen);
5867 free_extent_cache_tree(&pending);
5868 free_extent_cache_tree(&reada);
5869 free_extent_cache_tree(&nodes);
5870 free_extent_record_cache(root->fs_info, &extent_cache);
5874 err = check_chunks(&chunk_cache, &block_group_cache,
5875 &dev_extent_cache, NULL, NULL, 0);
5879 err = check_devices(&dev_cache, &dev_extent_cache);
5884 err = btrfs_commit_transaction(trans, root);
5890 free_corrupt_blocks_tree(root->fs_info->corrupt_blocks);
5891 root->fs_info->fsck_extent_cache = NULL;
5892 root->fs_info->free_extent_hook = NULL;
5893 root->fs_info->corrupt_blocks = NULL;
5896 free_chunk_cache_tree(&chunk_cache);
5897 free_device_cache_tree(&dev_cache);
5898 free_block_group_tree(&block_group_cache);
5899 free_device_extent_tree(&dev_extent_cache);
5903 static int btrfs_fsck_reinit_root(struct btrfs_trans_handle *trans,
5904 struct btrfs_root *root, int overwrite)
5906 struct extent_buffer *c;
5907 struct extent_buffer *old = root->node;
5909 struct btrfs_disk_key disk_key = {0,0,0};
5915 extent_buffer_get(c);
5918 c = btrfs_alloc_free_block(trans, root,
5919 btrfs_level_size(root, 0),
5920 root->root_key.objectid,
5921 &disk_key, level, 0, 0);
5924 extent_buffer_get(c);
5927 memset_extent_buffer(c, 0, 0, sizeof(struct btrfs_header));
5928 btrfs_set_header_level(c, level);
5929 btrfs_set_header_bytenr(c, c->start);
5930 btrfs_set_header_generation(c, trans->transid);
5931 btrfs_set_header_backref_rev(c, BTRFS_MIXED_BACKREF_REV);
5932 btrfs_set_header_owner(c, root->root_key.objectid);
5934 write_extent_buffer(c, root->fs_info->fsid,
5935 btrfs_header_fsid(), BTRFS_FSID_SIZE);
5937 write_extent_buffer(c, root->fs_info->chunk_tree_uuid,
5938 btrfs_header_chunk_tree_uuid(c),
5941 btrfs_mark_buffer_dirty(c);
5943 free_extent_buffer(old);
5945 add_root_to_dirty_list(root);
5949 static int pin_down_tree_blocks(struct btrfs_fs_info *fs_info,
5950 struct extent_buffer *eb, int tree_root)
5952 struct extent_buffer *tmp;
5953 struct btrfs_root_item *ri;
5954 struct btrfs_key key;
5957 int level = btrfs_header_level(eb);
5962 btrfs_pin_extent(fs_info, eb->start, eb->len);
5964 leafsize = btrfs_super_leafsize(fs_info->super_copy);
5965 nritems = btrfs_header_nritems(eb);
5966 for (i = 0; i < nritems; i++) {
5968 btrfs_item_key_to_cpu(eb, &key, i);
5969 if (key.type != BTRFS_ROOT_ITEM_KEY)
5971 /* Skip the extent root and reloc roots */
5972 if (key.objectid == BTRFS_EXTENT_TREE_OBJECTID ||
5973 key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
5974 key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
5976 ri = btrfs_item_ptr(eb, i, struct btrfs_root_item);
5977 bytenr = btrfs_disk_root_bytenr(eb, ri);
5980 * If at any point we start needing the real root we
5981 * will have to build a stump root for the root we are
5982 * in, but for now this doesn't actually use the root so
5983 * just pass in extent_root.
5985 tmp = read_tree_block(fs_info->extent_root, bytenr,
5988 fprintf(stderr, "Error reading root block\n");
5991 ret = pin_down_tree_blocks(fs_info, tmp, 0);
5992 free_extent_buffer(tmp);
5996 bytenr = btrfs_node_blockptr(eb, i);
5998 /* If we aren't the tree root don't read the block */
5999 if (level == 1 && !tree_root) {
6000 btrfs_pin_extent(fs_info, bytenr, leafsize);
6004 tmp = read_tree_block(fs_info->extent_root, bytenr,
6007 fprintf(stderr, "Error reading tree block\n");
6010 ret = pin_down_tree_blocks(fs_info, tmp, tree_root);
6011 free_extent_buffer(tmp);
6020 static int pin_metadata_blocks(struct btrfs_fs_info *fs_info)
6024 ret = pin_down_tree_blocks(fs_info, fs_info->chunk_root->node, 0);
6028 return pin_down_tree_blocks(fs_info, fs_info->tree_root->node, 1);
6031 static int reset_block_groups(struct btrfs_fs_info *fs_info)
6033 struct btrfs_path *path;
6034 struct extent_buffer *leaf;
6035 struct btrfs_chunk *chunk;
6036 struct btrfs_key key;
6039 path = btrfs_alloc_path();
6044 key.type = BTRFS_CHUNK_ITEM_KEY;
6047 ret = btrfs_search_slot(NULL, fs_info->chunk_root, &key, path, 0, 0);
6049 btrfs_free_path(path);
6054 * We do this in case the block groups were screwed up and had alloc
6055 * bits that aren't actually set on the chunks. This happens with
6056 * restored images every time and could happen in real life I guess.
6058 fs_info->avail_data_alloc_bits = 0;
6059 fs_info->avail_metadata_alloc_bits = 0;
6060 fs_info->avail_system_alloc_bits = 0;
6062 /* First we need to create the in-memory block groups */
6064 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
6065 ret = btrfs_next_leaf(fs_info->chunk_root, path);
6067 btrfs_free_path(path);
6075 leaf = path->nodes[0];
6076 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
6077 if (key.type != BTRFS_CHUNK_ITEM_KEY) {
6082 chunk = btrfs_item_ptr(leaf, path->slots[0],
6083 struct btrfs_chunk);
6084 btrfs_add_block_group(fs_info, 0,
6085 btrfs_chunk_type(leaf, chunk),
6086 key.objectid, key.offset,
6087 btrfs_chunk_length(leaf, chunk));
6091 btrfs_free_path(path);
6095 static int reset_balance(struct btrfs_trans_handle *trans,
6096 struct btrfs_fs_info *fs_info)
6098 struct btrfs_root *root = fs_info->tree_root;
6099 struct btrfs_path *path;
6100 struct extent_buffer *leaf;
6101 struct btrfs_key key;
6102 int del_slot, del_nr = 0;
6106 path = btrfs_alloc_path();
6110 key.objectid = BTRFS_BALANCE_OBJECTID;
6111 key.type = BTRFS_BALANCE_ITEM_KEY;
6114 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
6121 ret = btrfs_del_item(trans, root, path);
6124 btrfs_release_path(path);
6126 key.objectid = BTRFS_TREE_RELOC_OBJECTID;
6127 key.type = BTRFS_ROOT_ITEM_KEY;
6130 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
6134 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
6139 ret = btrfs_del_items(trans, root, path,
6146 btrfs_release_path(path);
6149 ret = btrfs_search_slot(trans, root, &key, path,
6156 leaf = path->nodes[0];
6157 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
6158 if (key.objectid > BTRFS_TREE_RELOC_OBJECTID)
6160 if (key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
6165 del_slot = path->slots[0];
6174 ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
6178 btrfs_release_path(path);
6180 key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6181 key.type = BTRFS_ROOT_ITEM_KEY;
6182 key.offset = (u64)-1;
6183 root = btrfs_read_fs_root(fs_info, &key);
6185 fprintf(stderr, "Error reading data reloc tree\n");
6186 return PTR_ERR(root);
6188 root->track_dirty = 1;
6189 if (root->last_trans != trans->transid) {
6190 root->last_trans = trans->transid;
6191 root->commit_root = root->node;
6192 extent_buffer_get(root->node);
6194 ret = btrfs_fsck_reinit_root(trans, root, 0);
6196 btrfs_free_path(path);
6200 static int reinit_extent_tree(struct btrfs_trans_handle *trans,
6201 struct btrfs_fs_info *fs_info)
6207 * The only reason we don't do this is because right now we're just
6208 * walking the trees we find and pinning down their bytes, we don't look
6209 * at any of the leaves. In order to do mixed groups we'd have to check
6210 * the leaves of any fs roots and pin down the bytes for any file
6211 * extents we find. Not hard but why do it if we don't have to?
6213 if (btrfs_fs_incompat(fs_info, BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)) {
6214 fprintf(stderr, "We don't support re-initing the extent tree "
6215 "for mixed block groups yet, please notify a btrfs "
6216 "developer you want to do this so they can add this "
6217 "functionality.\n");
6222 * first we need to walk all of the trees except the extent tree and pin
6223 * down the bytes that are in use so we don't overwrite any existing
6226 ret = pin_metadata_blocks(fs_info);
6228 fprintf(stderr, "error pinning down used bytes\n");
6233 * Need to drop all the block groups since we're going to recreate all
6236 btrfs_free_block_groups(fs_info);
6237 ret = reset_block_groups(fs_info);
6239 fprintf(stderr, "error resetting the block groups\n");
6243 ret = reset_balance(trans, fs_info);
6245 fprintf(stderr, "error reseting the pending balance\n");
6249 /* Ok we can allocate now, reinit the extent root */
6250 ret = btrfs_fsck_reinit_root(trans, fs_info->extent_root, 0);
6252 fprintf(stderr, "extent root initialization failed\n");
6254 * When the transaction code is updated we should end the
6255 * transaction, but for now progs only knows about commit so
6256 * just return an error.
6262 * Now we have all the in-memory block groups setup so we can make
6263 * allocations properly, and the metadata we care about is safe since we
6264 * pinned all of it above.
6267 struct btrfs_block_group_cache *cache;
6269 cache = btrfs_lookup_first_block_group(fs_info, start);
6272 start = cache->key.objectid + cache->key.offset;
6273 ret = btrfs_insert_item(trans, fs_info->extent_root,
6274 &cache->key, &cache->item,
6275 sizeof(cache->item));
6277 fprintf(stderr, "Error adding block group\n");
6280 btrfs_extent_post_op(trans, fs_info->extent_root);
6286 static int recow_extent_buffer(struct btrfs_root *root, struct extent_buffer *eb)
6288 struct btrfs_path *path;
6289 struct btrfs_trans_handle *trans;
6290 struct btrfs_key key;
6293 printf("Recowing metadata block %llu\n", eb->start);
6294 key.objectid = btrfs_header_owner(eb);
6295 key.type = BTRFS_ROOT_ITEM_KEY;
6296 key.offset = (u64)-1;
6298 root = btrfs_read_fs_root(root->fs_info, &key);
6300 fprintf(stderr, "Couldn't find owner root %llu\n",
6302 return PTR_ERR(root);
6305 path = btrfs_alloc_path();
6309 trans = btrfs_start_transaction(root, 1);
6310 if (IS_ERR(trans)) {
6311 btrfs_free_path(path);
6312 return PTR_ERR(trans);
6315 path->lowest_level = btrfs_header_level(eb);
6316 if (path->lowest_level)
6317 btrfs_node_key_to_cpu(eb, &key, 0);
6319 btrfs_item_key_to_cpu(eb, &key, 0);
6321 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
6322 btrfs_commit_transaction(trans, root);
6323 btrfs_free_path(path);
6327 static int delete_bad_item(struct btrfs_root *root, struct bad_item *bad)
6329 struct btrfs_path *path;
6330 struct btrfs_trans_handle *trans;
6331 struct btrfs_key key;
6334 printf("Deleting bad item [%llu,%u,%llu]\n", bad->key.objectid,
6335 bad->key.type, bad->key.offset);
6336 key.objectid = bad->root_id;
6337 key.type = BTRFS_ROOT_ITEM_KEY;
6338 key.offset = (u64)-1;
6340 root = btrfs_read_fs_root(root->fs_info, &key);
6342 fprintf(stderr, "Couldn't find owner root %llu\n",
6344 return PTR_ERR(root);
6347 path = btrfs_alloc_path();
6351 trans = btrfs_start_transaction(root, 1);
6352 if (IS_ERR(trans)) {
6353 btrfs_free_path(path);
6354 return PTR_ERR(trans);
6357 ret = btrfs_search_slot(trans, root, &bad->key, path, -1, 1);
6363 ret = btrfs_del_item(trans, root, path);
6365 btrfs_commit_transaction(trans, root);
6366 btrfs_free_path(path);
6370 static struct option long_options[] = {
6371 { "super", 1, NULL, 's' },
6372 { "repair", 0, NULL, 0 },
6373 { "init-csum-tree", 0, NULL, 0 },
6374 { "init-extent-tree", 0, NULL, 0 },
6375 { "backup", 0, NULL, 0 },
6379 const char * const cmd_check_usage[] = {
6380 "btrfs check [options] <device>",
6381 "Check an unmounted btrfs filesystem.",
6383 "-s|--super <superblock> use this superblock copy",
6384 "-b|--backup use the backup root copy",
6385 "--repair try to repair the filesystem",
6386 "--init-csum-tree create a new CRC tree",
6387 "--init-extent-tree create a new extent tree",
6391 int cmd_check(int argc, char **argv)
6393 struct cache_tree root_cache;
6394 struct btrfs_root *root;
6395 struct btrfs_fs_info *info;
6397 char uuidbuf[BTRFS_UUID_UNPARSED_SIZE];
6400 int option_index = 0;
6401 int init_csum_tree = 0;
6402 int init_extent_tree = 0;
6403 enum btrfs_open_ctree_flags ctree_flags =
6404 OPEN_CTREE_PARTIAL | OPEN_CTREE_EXCLUSIVE;
6408 c = getopt_long(argc, argv, "as:b", long_options,
6413 case 'a': /* ignored */ break;
6415 ctree_flags |= OPEN_CTREE_BACKUP_ROOT;
6418 num = arg_strtou64(optarg);
6419 if (num >= BTRFS_SUPER_MIRROR_MAX) {
6421 "ERROR: super mirror should be less than: %d\n",
6422 BTRFS_SUPER_MIRROR_MAX);
6425 bytenr = btrfs_sb_offset(((int)num));
6426 printf("using SB copy %llu, bytenr %llu\n", num,
6427 (unsigned long long)bytenr);
6431 usage(cmd_check_usage);
6433 if (option_index == 1) {
6434 printf("enabling repair mode\n");
6436 ctree_flags |= OPEN_CTREE_WRITES;
6437 } else if (option_index == 2) {
6438 printf("Creating a new CRC tree\n");
6441 ctree_flags |= OPEN_CTREE_WRITES;
6442 } else if (option_index == 3) {
6443 init_extent_tree = 1;
6444 ctree_flags |= (OPEN_CTREE_WRITES |
6445 OPEN_CTREE_NO_BLOCK_GROUPS);
6450 argc = argc - optind;
6453 usage(cmd_check_usage);
6456 cache_tree_init(&root_cache);
6458 if((ret = check_mounted(argv[optind])) < 0) {
6459 fprintf(stderr, "Could not check mount status: %s\n", strerror(-ret));
6462 fprintf(stderr, "%s is currently mounted. Aborting.\n", argv[optind]);
6467 info = open_ctree_fs_info(argv[optind], bytenr, 0, ctree_flags);
6469 fprintf(stderr, "Couldn't open file system\n");
6474 root = info->fs_root;
6475 uuid_unparse(info->super_copy->fsid, uuidbuf);
6476 printf("Checking filesystem on %s\nUUID: %s\n", argv[optind], uuidbuf);
6478 if (!extent_buffer_uptodate(info->tree_root->node) ||
6479 !extent_buffer_uptodate(info->dev_root->node) ||
6480 !extent_buffer_uptodate(info->chunk_root->node)) {
6481 fprintf(stderr, "Critical roots corrupted, unable to fsck the FS\n");
6486 if (init_extent_tree || init_csum_tree) {
6487 struct btrfs_trans_handle *trans;
6489 trans = btrfs_start_transaction(info->extent_root, 0);
6490 if (IS_ERR(trans)) {
6491 fprintf(stderr, "Error starting transaction\n");
6492 ret = PTR_ERR(trans);
6496 if (init_extent_tree) {
6497 printf("Creating a new extent tree\n");
6498 ret = reinit_extent_tree(trans, info);
6503 if (init_csum_tree) {
6504 fprintf(stderr, "Reinit crc root\n");
6505 ret = btrfs_fsck_reinit_root(trans, info->csum_root, 0);
6507 fprintf(stderr, "crc root initialization failed\n");
6513 * Ok now we commit and run the normal fsck, which will add
6514 * extent entries for all of the items it finds.
6516 ret = btrfs_commit_transaction(trans, info->extent_root);
6520 if (!extent_buffer_uptodate(info->extent_root->node)) {
6521 fprintf(stderr, "Critical roots corrupted, unable to fsck the FS\n");
6526 fprintf(stderr, "checking extents\n");
6527 ret = check_chunks_and_extents(root);
6529 fprintf(stderr, "Errors found in extent allocation tree or chunk allocation\n");
6531 fprintf(stderr, "checking free space cache\n");
6532 ret = check_space_cache(root);
6537 * We used to have to have these hole extents in between our real
6538 * extents so if we don't have this flag set we need to make sure there
6539 * are no gaps in the file extents for inodes, otherwise we can just
6540 * ignore it when this happens.
6542 no_holes = btrfs_fs_incompat(root->fs_info,
6543 BTRFS_FEATURE_INCOMPAT_NO_HOLES);
6544 fprintf(stderr, "checking fs roots\n");
6545 ret = check_fs_roots(root, &root_cache);
6549 fprintf(stderr, "checking csums\n");
6550 ret = check_csums(root);
6554 fprintf(stderr, "checking root refs\n");
6555 ret = check_root_refs(root, &root_cache);
6559 while (repair && !list_empty(&root->fs_info->recow_ebs)) {
6560 struct extent_buffer *eb;
6562 eb = list_first_entry(&root->fs_info->recow_ebs,
6563 struct extent_buffer, recow);
6564 ret = recow_extent_buffer(root, eb);
6569 while (!list_empty(&delete_items)) {
6570 struct bad_item *bad;
6572 bad = list_first_entry(&delete_items, struct bad_item, list);
6573 list_del_init(&bad->list);
6575 ret = delete_bad_item(root, bad);
6579 if (!list_empty(&root->fs_info->recow_ebs)) {
6580 fprintf(stderr, "Transid errors in file system\n");
6584 if (found_old_backref) { /*
6585 * there was a disk format change when mixed
6586 * backref was in testing tree. The old format
6587 * existed about one week.
6589 printf("\n * Found old mixed backref format. "
6590 "The old format is not supported! *"
6591 "\n * Please mount the FS in readonly mode, "
6592 "backup data and re-format the FS. *\n\n");
6595 printf("found %llu bytes used err is %d\n",
6596 (unsigned long long)bytes_used, ret);
6597 printf("total csum bytes: %llu\n",(unsigned long long)total_csum_bytes);
6598 printf("total tree bytes: %llu\n",
6599 (unsigned long long)total_btree_bytes);
6600 printf("total fs tree bytes: %llu\n",
6601 (unsigned long long)total_fs_tree_bytes);
6602 printf("total extent tree bytes: %llu\n",
6603 (unsigned long long)total_extent_tree_bytes);
6604 printf("btree space waste bytes: %llu\n",
6605 (unsigned long long)btree_space_waste);
6606 printf("file data blocks allocated: %llu\n referenced %llu\n",
6607 (unsigned long long)data_bytes_allocated,
6608 (unsigned long long)data_bytes_referenced);
6609 printf("%s\n", BTRFS_BUILD_VERSION);
6611 free_root_recs_tree(&root_cache);