aboutsummaryrefslogtreecommitdiffstats
path: root/fs/btrfs/backref.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs/backref.c')
-rw-r--r--fs/btrfs/backref.c1131
1 files changed, 878 insertions, 253 deletions
diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 22c64fff1bd5..b9a843226de8 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -19,18 +19,789 @@
19#include "ctree.h" 19#include "ctree.h"
20#include "disk-io.h" 20#include "disk-io.h"
21#include "backref.h" 21#include "backref.h"
22#include "ulist.h"
23#include "transaction.h"
24#include "delayed-ref.h"
22 25
23struct __data_ref { 26/*
27 * this structure records all encountered refs on the way up to the root
28 */
29struct __prelim_ref {
24 struct list_head list; 30 struct list_head list;
25 u64 inum; 31 u64 root_id;
26 u64 root; 32 struct btrfs_key key;
27 u64 extent_data_item_offset; 33 int level;
34 int count;
35 u64 parent;
36 u64 wanted_disk_byte;
28}; 37};
29 38
30struct __shared_ref { 39static int __add_prelim_ref(struct list_head *head, u64 root_id,
31 struct list_head list; 40 struct btrfs_key *key, int level, u64 parent,
41 u64 wanted_disk_byte, int count)
42{
43 struct __prelim_ref *ref;
44
45 /* in case we're adding delayed refs, we're holding the refs spinlock */
46 ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
47 if (!ref)
48 return -ENOMEM;
49
50 ref->root_id = root_id;
51 if (key)
52 ref->key = *key;
53 else
54 memset(&ref->key, 0, sizeof(ref->key));
55
56 ref->level = level;
57 ref->count = count;
58 ref->parent = parent;
59 ref->wanted_disk_byte = wanted_disk_byte;
60 list_add_tail(&ref->list, head);
61
62 return 0;
63}
64
65static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
66 struct ulist *parents,
67 struct extent_buffer *eb, int level,
68 u64 wanted_objectid, u64 wanted_disk_byte)
69{
70 int ret;
71 int slot;
72 struct btrfs_file_extent_item *fi;
73 struct btrfs_key key;
32 u64 disk_byte; 74 u64 disk_byte;
33}; 75
76add_parent:
77 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
78 if (ret < 0)
79 return ret;
80
81 if (level != 0)
82 return 0;
83
84 /*
85 * if the current leaf is full with EXTENT_DATA items, we must
86 * check the next one if that holds a reference as well.
87 * ref->count cannot be used to skip this check.
88 * repeat this until we don't find any additional EXTENT_DATA items.
89 */
90 while (1) {
91 ret = btrfs_next_leaf(root, path);
92 if (ret < 0)
93 return ret;
94 if (ret)
95 return 0;
96
97 eb = path->nodes[0];
98 for (slot = 0; slot < btrfs_header_nritems(eb); ++slot) {
99 btrfs_item_key_to_cpu(eb, &key, slot);
100 if (key.objectid != wanted_objectid ||
101 key.type != BTRFS_EXTENT_DATA_KEY)
102 return 0;
103 fi = btrfs_item_ptr(eb, slot,
104 struct btrfs_file_extent_item);
105 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
106 if (disk_byte == wanted_disk_byte)
107 goto add_parent;
108 }
109 }
110
111 return 0;
112}
113
114/*
115 * resolve an indirect backref in the form (root_id, key, level)
116 * to a logical address
117 */
118static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
119 struct __prelim_ref *ref,
120 struct ulist *parents)
121{
122 struct btrfs_path *path;
123 struct btrfs_root *root;
124 struct btrfs_key root_key;
125 struct btrfs_key key = {0};
126 struct extent_buffer *eb;
127 int ret = 0;
128 int root_level;
129 int level = ref->level;
130
131 path = btrfs_alloc_path();
132 if (!path)
133 return -ENOMEM;
134
135 root_key.objectid = ref->root_id;
136 root_key.type = BTRFS_ROOT_ITEM_KEY;
137 root_key.offset = (u64)-1;
138 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
139 if (IS_ERR(root)) {
140 ret = PTR_ERR(root);
141 goto out;
142 }
143
144 rcu_read_lock();
145 root_level = btrfs_header_level(root->node);
146 rcu_read_unlock();
147
148 if (root_level + 1 == level)
149 goto out;
150
151 path->lowest_level = level;
152 ret = btrfs_search_slot(NULL, root, &ref->key, path, 0, 0);
153 pr_debug("search slot in root %llu (level %d, ref count %d) returned "
154 "%d for key (%llu %u %llu)\n",
155 (unsigned long long)ref->root_id, level, ref->count, ret,
156 (unsigned long long)ref->key.objectid, ref->key.type,
157 (unsigned long long)ref->key.offset);
158 if (ret < 0)
159 goto out;
160
161 eb = path->nodes[level];
162 if (!eb) {
163 WARN_ON(1);
164 ret = 1;
165 goto out;
166 }
167
168 if (level == 0) {
169 if (ret == 1 && path->slots[0] >= btrfs_header_nritems(eb)) {
170 ret = btrfs_next_leaf(root, path);
171 if (ret)
172 goto out;
173 eb = path->nodes[0];
174 }
175
176 btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
177 }
178
179 /* the last two parameters will only be used for level == 0 */
180 ret = add_all_parents(root, path, parents, eb, level, key.objectid,
181 ref->wanted_disk_byte);
182out:
183 btrfs_free_path(path);
184 return ret;
185}
186
187/*
188 * resolve all indirect backrefs from the list
189 */
190static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
191 struct list_head *head)
192{
193 int err;
194 int ret = 0;
195 struct __prelim_ref *ref;
196 struct __prelim_ref *ref_safe;
197 struct __prelim_ref *new_ref;
198 struct ulist *parents;
199 struct ulist_node *node;
200
201 parents = ulist_alloc(GFP_NOFS);
202 if (!parents)
203 return -ENOMEM;
204
205 /*
206 * _safe allows us to insert directly after the current item without
207 * iterating over the newly inserted items.
208 * we're also allowed to re-assign ref during iteration.
209 */
210 list_for_each_entry_safe(ref, ref_safe, head, list) {
211 if (ref->parent) /* already direct */
212 continue;
213 if (ref->count == 0)
214 continue;
215 err = __resolve_indirect_ref(fs_info, ref, parents);
216 if (err) {
217 if (ret == 0)
218 ret = err;
219 continue;
220 }
221
222 /* we put the first parent into the ref at hand */
223 node = ulist_next(parents, NULL);
224 ref->parent = node ? node->val : 0;
225
226 /* additional parents require new refs being added here */
227 while ((node = ulist_next(parents, node))) {
228 new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
229 if (!new_ref) {
230 ret = -ENOMEM;
231 break;
232 }
233 memcpy(new_ref, ref, sizeof(*ref));
234 new_ref->parent = node->val;
235 list_add(&new_ref->list, &ref->list);
236 }
237 ulist_reinit(parents);
238 }
239
240 ulist_free(parents);
241 return ret;
242}
243
244/*
245 * merge two lists of backrefs and adjust counts accordingly
246 *
247 * mode = 1: merge identical keys, if key is set
248 * mode = 2: merge identical parents
249 */
250static int __merge_refs(struct list_head *head, int mode)
251{
252 struct list_head *pos1;
253
254 list_for_each(pos1, head) {
255 struct list_head *n2;
256 struct list_head *pos2;
257 struct __prelim_ref *ref1;
258
259 ref1 = list_entry(pos1, struct __prelim_ref, list);
260
261 if (mode == 1 && ref1->key.type == 0)
262 continue;
263 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
264 pos2 = n2, n2 = pos2->next) {
265 struct __prelim_ref *ref2;
266
267 ref2 = list_entry(pos2, struct __prelim_ref, list);
268
269 if (mode == 1) {
270 if (memcmp(&ref1->key, &ref2->key,
271 sizeof(ref1->key)) ||
272 ref1->level != ref2->level ||
273 ref1->root_id != ref2->root_id)
274 continue;
275 ref1->count += ref2->count;
276 } else {
277 if (ref1->parent != ref2->parent)
278 continue;
279 ref1->count += ref2->count;
280 }
281 list_del(&ref2->list);
282 kfree(ref2);
283 }
284
285 }
286 return 0;
287}
288
289/*
290 * add all currently queued delayed refs from this head whose seq nr is
291 * smaller or equal that seq to the list
292 */
293static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
294 struct btrfs_key *info_key,
295 struct list_head *prefs)
296{
297 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
298 struct rb_node *n = &head->node.rb_node;
299 int sgn;
300 int ret;
301
302 if (extent_op && extent_op->update_key)
303 btrfs_disk_key_to_cpu(info_key, &extent_op->key);
304
305 while ((n = rb_prev(n))) {
306 struct btrfs_delayed_ref_node *node;
307 node = rb_entry(n, struct btrfs_delayed_ref_node,
308 rb_node);
309 if (node->bytenr != head->node.bytenr)
310 break;
311 WARN_ON(node->is_head);
312
313 if (node->seq > seq)
314 continue;
315
316 switch (node->action) {
317 case BTRFS_ADD_DELAYED_EXTENT:
318 case BTRFS_UPDATE_DELAYED_HEAD:
319 WARN_ON(1);
320 continue;
321 case BTRFS_ADD_DELAYED_REF:
322 sgn = 1;
323 break;
324 case BTRFS_DROP_DELAYED_REF:
325 sgn = -1;
326 break;
327 default:
328 BUG_ON(1);
329 }
330 switch (node->type) {
331 case BTRFS_TREE_BLOCK_REF_KEY: {
332 struct btrfs_delayed_tree_ref *ref;
333
334 ref = btrfs_delayed_node_to_tree_ref(node);
335 ret = __add_prelim_ref(prefs, ref->root, info_key,
336 ref->level + 1, 0, node->bytenr,
337 node->ref_mod * sgn);
338 break;
339 }
340 case BTRFS_SHARED_BLOCK_REF_KEY: {
341 struct btrfs_delayed_tree_ref *ref;
342
343 ref = btrfs_delayed_node_to_tree_ref(node);
344 ret = __add_prelim_ref(prefs, ref->root, info_key,
345 ref->level + 1, ref->parent,
346 node->bytenr,
347 node->ref_mod * sgn);
348 break;
349 }
350 case BTRFS_EXTENT_DATA_REF_KEY: {
351 struct btrfs_delayed_data_ref *ref;
352 struct btrfs_key key;
353
354 ref = btrfs_delayed_node_to_data_ref(node);
355
356 key.objectid = ref->objectid;
357 key.type = BTRFS_EXTENT_DATA_KEY;
358 key.offset = ref->offset;
359 ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
360 node->bytenr,
361 node->ref_mod * sgn);
362 break;
363 }
364 case BTRFS_SHARED_DATA_REF_KEY: {
365 struct btrfs_delayed_data_ref *ref;
366 struct btrfs_key key;
367
368 ref = btrfs_delayed_node_to_data_ref(node);
369
370 key.objectid = ref->objectid;
371 key.type = BTRFS_EXTENT_DATA_KEY;
372 key.offset = ref->offset;
373 ret = __add_prelim_ref(prefs, ref->root, &key, 0,
374 ref->parent, node->bytenr,
375 node->ref_mod * sgn);
376 break;
377 }
378 default:
379 WARN_ON(1);
380 }
381 BUG_ON(ret);
382 }
383
384 return 0;
385}
386
387/*
388 * add all inline backrefs for bytenr to the list
389 */
390static int __add_inline_refs(struct btrfs_fs_info *fs_info,
391 struct btrfs_path *path, u64 bytenr,
392 struct btrfs_key *info_key, int *info_level,
393 struct list_head *prefs)
394{
395 int ret;
396 int slot;
397 struct extent_buffer *leaf;
398 struct btrfs_key key;
399 unsigned long ptr;
400 unsigned long end;
401 struct btrfs_extent_item *ei;
402 u64 flags;
403 u64 item_size;
404
405 /*
406 * enumerate all inline refs
407 */
408 leaf = path->nodes[0];
409 slot = path->slots[0] - 1;
410
411 item_size = btrfs_item_size_nr(leaf, slot);
412 BUG_ON(item_size < sizeof(*ei));
413
414 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
415 flags = btrfs_extent_flags(leaf, ei);
416
417 ptr = (unsigned long)(ei + 1);
418 end = (unsigned long)ei + item_size;
419
420 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
421 struct btrfs_tree_block_info *info;
422 struct btrfs_disk_key disk_key;
423
424 info = (struct btrfs_tree_block_info *)ptr;
425 *info_level = btrfs_tree_block_level(leaf, info);
426 btrfs_tree_block_key(leaf, info, &disk_key);
427 btrfs_disk_key_to_cpu(info_key, &disk_key);
428 ptr += sizeof(struct btrfs_tree_block_info);
429 BUG_ON(ptr > end);
430 } else {
431 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
432 }
433
434 while (ptr < end) {
435 struct btrfs_extent_inline_ref *iref;
436 u64 offset;
437 int type;
438
439 iref = (struct btrfs_extent_inline_ref *)ptr;
440 type = btrfs_extent_inline_ref_type(leaf, iref);
441 offset = btrfs_extent_inline_ref_offset(leaf, iref);
442
443 switch (type) {
444 case BTRFS_SHARED_BLOCK_REF_KEY:
445 ret = __add_prelim_ref(prefs, 0, info_key,
446 *info_level + 1, offset,
447 bytenr, 1);
448 break;
449 case BTRFS_SHARED_DATA_REF_KEY: {
450 struct btrfs_shared_data_ref *sdref;
451 int count;
452
453 sdref = (struct btrfs_shared_data_ref *)(iref + 1);
454 count = btrfs_shared_data_ref_count(leaf, sdref);
455 ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
456 bytenr, count);
457 break;
458 }
459 case BTRFS_TREE_BLOCK_REF_KEY:
460 ret = __add_prelim_ref(prefs, offset, info_key,
461 *info_level + 1, 0, bytenr, 1);
462 break;
463 case BTRFS_EXTENT_DATA_REF_KEY: {
464 struct btrfs_extent_data_ref *dref;
465 int count;
466 u64 root;
467
468 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
469 count = btrfs_extent_data_ref_count(leaf, dref);
470 key.objectid = btrfs_extent_data_ref_objectid(leaf,
471 dref);
472 key.type = BTRFS_EXTENT_DATA_KEY;
473 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
474 root = btrfs_extent_data_ref_root(leaf, dref);
475 ret = __add_prelim_ref(prefs, root, &key, 0, 0, bytenr,
476 count);
477 break;
478 }
479 default:
480 WARN_ON(1);
481 }
482 BUG_ON(ret);
483 ptr += btrfs_extent_inline_ref_size(type);
484 }
485
486 return 0;
487}
488
489/*
490 * add all non-inline backrefs for bytenr to the list
491 */
492static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
493 struct btrfs_path *path, u64 bytenr,
494 struct btrfs_key *info_key, int info_level,
495 struct list_head *prefs)
496{
497 struct btrfs_root *extent_root = fs_info->extent_root;
498 int ret;
499 int slot;
500 struct extent_buffer *leaf;
501 struct btrfs_key key;
502
503 while (1) {
504 ret = btrfs_next_item(extent_root, path);
505 if (ret < 0)
506 break;
507 if (ret) {
508 ret = 0;
509 break;
510 }
511
512 slot = path->slots[0];
513 leaf = path->nodes[0];
514 btrfs_item_key_to_cpu(leaf, &key, slot);
515
516 if (key.objectid != bytenr)
517 break;
518 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
519 continue;
520 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
521 break;
522
523 switch (key.type) {
524 case BTRFS_SHARED_BLOCK_REF_KEY:
525 ret = __add_prelim_ref(prefs, 0, info_key,
526 info_level + 1, key.offset,
527 bytenr, 1);
528 break;
529 case BTRFS_SHARED_DATA_REF_KEY: {
530 struct btrfs_shared_data_ref *sdref;
531 int count;
532
533 sdref = btrfs_item_ptr(leaf, slot,
534 struct btrfs_shared_data_ref);
535 count = btrfs_shared_data_ref_count(leaf, sdref);
536 ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
537 bytenr, count);
538 break;
539 }
540 case BTRFS_TREE_BLOCK_REF_KEY:
541 ret = __add_prelim_ref(prefs, key.offset, info_key,
542 info_level + 1, 0, bytenr, 1);
543 break;
544 case BTRFS_EXTENT_DATA_REF_KEY: {
545 struct btrfs_extent_data_ref *dref;
546 int count;
547 u64 root;
548
549 dref = btrfs_item_ptr(leaf, slot,
550 struct btrfs_extent_data_ref);
551 count = btrfs_extent_data_ref_count(leaf, dref);
552 key.objectid = btrfs_extent_data_ref_objectid(leaf,
553 dref);
554 key.type = BTRFS_EXTENT_DATA_KEY;
555 key.offset = btrfs_extent_data_ref_offset(leaf, dref);
556 root = btrfs_extent_data_ref_root(leaf, dref);
557 ret = __add_prelim_ref(prefs, root, &key, 0, 0,
558 bytenr, count);
559 break;
560 }
561 default:
562 WARN_ON(1);
563 }
564 BUG_ON(ret);
565 }
566
567 return ret;
568}
569
570/*
571 * this adds all existing backrefs (inline backrefs, backrefs and delayed
572 * refs) for the given bytenr to the refs list, merges duplicates and resolves
573 * indirect refs to their parent bytenr.
574 * When roots are found, they're added to the roots list
575 *
576 * FIXME some caching might speed things up
577 */
578static int find_parent_nodes(struct btrfs_trans_handle *trans,
579 struct btrfs_fs_info *fs_info, u64 bytenr,
580 u64 seq, struct ulist *refs, struct ulist *roots)
581{
582 struct btrfs_key key;
583 struct btrfs_path *path;
584 struct btrfs_key info_key = { 0 };
585 struct btrfs_delayed_ref_root *delayed_refs = NULL;
586 struct btrfs_delayed_ref_head *head = NULL;
587 int info_level = 0;
588 int ret;
589 struct list_head prefs_delayed;
590 struct list_head prefs;
591 struct __prelim_ref *ref;
592
593 INIT_LIST_HEAD(&prefs);
594 INIT_LIST_HEAD(&prefs_delayed);
595
596 key.objectid = bytenr;
597 key.type = BTRFS_EXTENT_ITEM_KEY;
598 key.offset = (u64)-1;
599
600 path = btrfs_alloc_path();
601 if (!path)
602 return -ENOMEM;
603
604 /*
605 * grab both a lock on the path and a lock on the delayed ref head.
606 * We need both to get a consistent picture of how the refs look
607 * at a specified point in time
608 */
609again:
610 ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
611 if (ret < 0)
612 goto out;
613 BUG_ON(ret == 0);
614
615 /*
616 * look if there are updates for this ref queued and lock the head
617 */
618 delayed_refs = &trans->transaction->delayed_refs;
619 spin_lock(&delayed_refs->lock);
620 head = btrfs_find_delayed_ref_head(trans, bytenr);
621 if (head) {
622 if (!mutex_trylock(&head->mutex)) {
623 atomic_inc(&head->node.refs);
624 spin_unlock(&delayed_refs->lock);
625
626 btrfs_release_path(path);
627
628 /*
629 * Mutex was contended, block until it's
630 * released and try again
631 */
632 mutex_lock(&head->mutex);
633 mutex_unlock(&head->mutex);
634 btrfs_put_delayed_ref(&head->node);
635 goto again;
636 }
637 ret = __add_delayed_refs(head, seq, &info_key, &prefs_delayed);
638 if (ret)
639 goto out;
640 }
641 spin_unlock(&delayed_refs->lock);
642
643 if (path->slots[0]) {
644 struct extent_buffer *leaf;
645 int slot;
646
647 leaf = path->nodes[0];
648 slot = path->slots[0] - 1;
649 btrfs_item_key_to_cpu(leaf, &key, slot);
650 if (key.objectid == bytenr &&
651 key.type == BTRFS_EXTENT_ITEM_KEY) {
652 ret = __add_inline_refs(fs_info, path, bytenr,
653 &info_key, &info_level, &prefs);
654 if (ret)
655 goto out;
656 ret = __add_keyed_refs(fs_info, path, bytenr, &info_key,
657 info_level, &prefs);
658 if (ret)
659 goto out;
660 }
661 }
662 btrfs_release_path(path);
663
664 /*
665 * when adding the delayed refs above, the info_key might not have
666 * been known yet. Go over the list and replace the missing keys
667 */
668 list_for_each_entry(ref, &prefs_delayed, list) {
669 if ((ref->key.offset | ref->key.type | ref->key.objectid) == 0)
670 memcpy(&ref->key, &info_key, sizeof(ref->key));
671 }
672 list_splice_init(&prefs_delayed, &prefs);
673
674 ret = __merge_refs(&prefs, 1);
675 if (ret)
676 goto out;
677
678 ret = __resolve_indirect_refs(fs_info, &prefs);
679 if (ret)
680 goto out;
681
682 ret = __merge_refs(&prefs, 2);
683 if (ret)
684 goto out;
685
686 while (!list_empty(&prefs)) {
687 ref = list_first_entry(&prefs, struct __prelim_ref, list);
688 list_del(&ref->list);
689 if (ref->count < 0)
690 WARN_ON(1);
691 if (ref->count && ref->root_id && ref->parent == 0) {
692 /* no parent == root of tree */
693 ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
694 BUG_ON(ret < 0);
695 }
696 if (ref->count && ref->parent) {
697 ret = ulist_add(refs, ref->parent, 0, GFP_NOFS);
698 BUG_ON(ret < 0);
699 }
700 kfree(ref);
701 }
702
703out:
704 if (head)
705 mutex_unlock(&head->mutex);
706 btrfs_free_path(path);
707 while (!list_empty(&prefs)) {
708 ref = list_first_entry(&prefs, struct __prelim_ref, list);
709 list_del(&ref->list);
710 kfree(ref);
711 }
712 while (!list_empty(&prefs_delayed)) {
713 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
714 list);
715 list_del(&ref->list);
716 kfree(ref);
717 }
718
719 return ret;
720}
721
722/*
723 * Finds all leafs with a reference to the specified combination of bytenr and
724 * offset. key_list_head will point to a list of corresponding keys (caller must
725 * free each list element). The leafs will be stored in the leafs ulist, which
726 * must be freed with ulist_free.
727 *
728 * returns 0 on success, <0 on error
729 */
730static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
731 struct btrfs_fs_info *fs_info, u64 bytenr,
732 u64 num_bytes, u64 seq, struct ulist **leafs)
733{
734 struct ulist *tmp;
735 int ret;
736
737 tmp = ulist_alloc(GFP_NOFS);
738 if (!tmp)
739 return -ENOMEM;
740 *leafs = ulist_alloc(GFP_NOFS);
741 if (!*leafs) {
742 ulist_free(tmp);
743 return -ENOMEM;
744 }
745
746 ret = find_parent_nodes(trans, fs_info, bytenr, seq, *leafs, tmp);
747 ulist_free(tmp);
748
749 if (ret < 0 && ret != -ENOENT) {
750 ulist_free(*leafs);
751 return ret;
752 }
753
754 return 0;
755}
756
757/*
758 * walk all backrefs for a given extent to find all roots that reference this
759 * extent. Walking a backref means finding all extents that reference this
760 * extent and in turn walk the backrefs of those, too. Naturally this is a
761 * recursive process, but here it is implemented in an iterative fashion: We
762 * find all referencing extents for the extent in question and put them on a
763 * list. In turn, we find all referencing extents for those, further appending
764 * to the list. The way we iterate the list allows adding more elements after
765 * the current while iterating. The process stops when we reach the end of the
766 * list. Found roots are added to the roots list.
767 *
768 * returns 0 on success, < 0 on error.
769 */
770int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
771 struct btrfs_fs_info *fs_info, u64 bytenr,
772 u64 num_bytes, u64 seq, struct ulist **roots)
773{
774 struct ulist *tmp;
775 struct ulist_node *node = NULL;
776 int ret;
777
778 tmp = ulist_alloc(GFP_NOFS);
779 if (!tmp)
780 return -ENOMEM;
781 *roots = ulist_alloc(GFP_NOFS);
782 if (!*roots) {
783 ulist_free(tmp);
784 return -ENOMEM;
785 }
786
787 while (1) {
788 ret = find_parent_nodes(trans, fs_info, bytenr, seq,
789 tmp, *roots);
790 if (ret < 0 && ret != -ENOENT) {
791 ulist_free(tmp);
792 ulist_free(*roots);
793 return ret;
794 }
795 node = ulist_next(tmp, node);
796 if (!node)
797 break;
798 bytenr = node->val;
799 }
800
801 ulist_free(tmp);
802 return 0;
803}
804
34 805
35static int __inode_info(u64 inum, u64 ioff, u8 key_type, 806static int __inode_info(u64 inum, u64 ioff, u8 key_type,
36 struct btrfs_root *fs_root, struct btrfs_path *path, 807 struct btrfs_root *fs_root, struct btrfs_path *path,
@@ -181,8 +952,11 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
181 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]); 952 btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
182 if (found_key->type != BTRFS_EXTENT_ITEM_KEY || 953 if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
183 found_key->objectid > logical || 954 found_key->objectid > logical ||
184 found_key->objectid + found_key->offset <= logical) 955 found_key->objectid + found_key->offset <= logical) {
956 pr_debug("logical %llu is not within any extent\n",
957 (unsigned long long)logical);
185 return -ENOENT; 958 return -ENOENT;
959 }
186 960
187 eb = path->nodes[0]; 961 eb = path->nodes[0];
188 item_size = btrfs_item_size_nr(eb, path->slots[0]); 962 item_size = btrfs_item_size_nr(eb, path->slots[0]);
@@ -191,6 +965,13 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
191 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item); 965 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
192 flags = btrfs_extent_flags(eb, ei); 966 flags = btrfs_extent_flags(eb, ei);
193 967
968 pr_debug("logical %llu is at position %llu within the extent (%llu "
969 "EXTENT_ITEM %llu) flags %#llx size %u\n",
970 (unsigned long long)logical,
971 (unsigned long long)(logical - found_key->objectid),
972 (unsigned long long)found_key->objectid,
973 (unsigned long long)found_key->offset,
974 (unsigned long long)flags, item_size);
194 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) 975 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
195 return BTRFS_EXTENT_FLAG_TREE_BLOCK; 976 return BTRFS_EXTENT_FLAG_TREE_BLOCK;
196 if (flags & BTRFS_EXTENT_FLAG_DATA) 977 if (flags & BTRFS_EXTENT_FLAG_DATA)
@@ -287,128 +1068,11 @@ int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
287 return 0; 1068 return 0;
288} 1069}
289 1070
290static int __data_list_add(struct list_head *head, u64 inum, 1071static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
291 u64 extent_data_item_offset, u64 root) 1072 struct btrfs_path *path, u64 logical,
292{ 1073 u64 orig_extent_item_objectid,
293 struct __data_ref *ref; 1074 u64 extent_item_pos, u64 root,
294 1075 iterate_extent_inodes_t *iterate, void *ctx)
295 ref = kmalloc(sizeof(*ref), GFP_NOFS);
296 if (!ref)
297 return -ENOMEM;
298
299 ref->inum = inum;
300 ref->extent_data_item_offset = extent_data_item_offset;
301 ref->root = root;
302 list_add_tail(&ref->list, head);
303
304 return 0;
305}
306
307static int __data_list_add_eb(struct list_head *head, struct extent_buffer *eb,
308 struct btrfs_extent_data_ref *dref)
309{
310 return __data_list_add(head, btrfs_extent_data_ref_objectid(eb, dref),
311 btrfs_extent_data_ref_offset(eb, dref),
312 btrfs_extent_data_ref_root(eb, dref));
313}
314
315static int __shared_list_add(struct list_head *head, u64 disk_byte)
316{
317 struct __shared_ref *ref;
318
319 ref = kmalloc(sizeof(*ref), GFP_NOFS);
320 if (!ref)
321 return -ENOMEM;
322
323 ref->disk_byte = disk_byte;
324 list_add_tail(&ref->list, head);
325
326 return 0;
327}
328
329static int __iter_shared_inline_ref_inodes(struct btrfs_fs_info *fs_info,
330 u64 logical, u64 inum,
331 u64 extent_data_item_offset,
332 u64 extent_offset,
333 struct btrfs_path *path,
334 struct list_head *data_refs,
335 iterate_extent_inodes_t *iterate,
336 void *ctx)
337{
338 u64 ref_root;
339 u32 item_size;
340 struct btrfs_key key;
341 struct extent_buffer *eb;
342 struct btrfs_extent_item *ei;
343 struct btrfs_extent_inline_ref *eiref;
344 struct __data_ref *ref;
345 int ret;
346 int type;
347 int last;
348 unsigned long ptr = 0;
349
350 WARN_ON(!list_empty(data_refs));
351 ret = extent_from_logical(fs_info, logical, path, &key);
352 if (ret & BTRFS_EXTENT_FLAG_DATA)
353 ret = -EIO;
354 if (ret < 0)
355 goto out;
356
357 eb = path->nodes[0];
358 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
359 item_size = btrfs_item_size_nr(eb, path->slots[0]);
360
361 ret = 0;
362 ref_root = 0;
363 /*
364 * as done in iterate_extent_inodes, we first build a list of refs to
365 * iterate, then free the path and then iterate them to avoid deadlocks.
366 */
367 do {
368 last = __get_extent_inline_ref(&ptr, eb, ei, item_size,
369 &eiref, &type);
370 if (last < 0) {
371 ret = last;
372 goto out;
373 }
374 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
375 type == BTRFS_SHARED_BLOCK_REF_KEY) {
376 ref_root = btrfs_extent_inline_ref_offset(eb, eiref);
377 ret = __data_list_add(data_refs, inum,
378 extent_data_item_offset,
379 ref_root);
380 }
381 } while (!ret && !last);
382
383 btrfs_release_path(path);
384
385 if (ref_root == 0) {
386 printk(KERN_ERR "btrfs: failed to find tree block ref "
387 "for shared data backref %llu\n", logical);
388 WARN_ON(1);
389 ret = -EIO;
390 }
391
392out:
393 while (!list_empty(data_refs)) {
394 ref = list_first_entry(data_refs, struct __data_ref, list);
395 list_del(&ref->list);
396 if (!ret)
397 ret = iterate(ref->inum, extent_offset +
398 ref->extent_data_item_offset,
399 ref->root, ctx);
400 kfree(ref);
401 }
402
403 return ret;
404}
405
406static int __iter_shared_inline_ref(struct btrfs_fs_info *fs_info,
407 u64 logical, u64 orig_extent_item_objectid,
408 u64 extent_offset, struct btrfs_path *path,
409 struct list_head *data_refs,
410 iterate_extent_inodes_t *iterate,
411 void *ctx)
412{ 1076{
413 u64 disk_byte; 1077 u64 disk_byte;
414 struct btrfs_key key; 1078 struct btrfs_key key;
@@ -416,8 +1080,10 @@ static int __iter_shared_inline_ref(struct btrfs_fs_info *fs_info,
416 struct extent_buffer *eb; 1080 struct extent_buffer *eb;
417 int slot; 1081 int slot;
418 int nritems; 1082 int nritems;
419 int ret; 1083 int ret = 0;
420 int found = 0; 1084 int extent_type;
1085 u64 data_offset;
1086 u64 data_len;
421 1087
422 eb = read_tree_block(fs_info->tree_root, logical, 1088 eb = read_tree_block(fs_info->tree_root, logical,
423 fs_info->tree_root->leafsize, 0); 1089 fs_info->tree_root->leafsize, 0);
@@ -435,149 +1101,99 @@ static int __iter_shared_inline_ref(struct btrfs_fs_info *fs_info,
435 if (key.type != BTRFS_EXTENT_DATA_KEY) 1101 if (key.type != BTRFS_EXTENT_DATA_KEY)
436 continue; 1102 continue;
437 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item); 1103 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
438 if (!fi) { 1104 extent_type = btrfs_file_extent_type(eb, fi);
439 free_extent_buffer(eb); 1105 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
440 return -EIO; 1106 continue;
441 } 1107 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
442 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi); 1108 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
443 if (disk_byte != orig_extent_item_objectid) { 1109 if (disk_byte != orig_extent_item_objectid)
444 if (found) 1110 continue;
445 break;
446 else
447 continue;
448 }
449 ++found;
450 ret = __iter_shared_inline_ref_inodes(fs_info, logical,
451 key.objectid,
452 key.offset,
453 extent_offset, path,
454 data_refs,
455 iterate, ctx);
456 if (ret)
457 break;
458 }
459 1111
460 if (!found) { 1112 data_offset = btrfs_file_extent_offset(eb, fi);
461 printk(KERN_ERR "btrfs: failed to follow shared data backref " 1113 data_len = btrfs_file_extent_num_bytes(eb, fi);
462 "to parent %llu\n", logical); 1114
463 WARN_ON(1); 1115 if (extent_item_pos < data_offset ||
464 ret = -EIO; 1116 extent_item_pos >= data_offset + data_len)
1117 continue;
1118
1119 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1120 "root %llu\n", orig_extent_item_objectid,
1121 key.objectid, key.offset, root);
1122 ret = iterate(key.objectid,
1123 key.offset + (extent_item_pos - data_offset),
1124 root, ctx);
1125 if (ret) {
1126 pr_debug("stopping iteration because ret=%d\n", ret);
1127 break;
1128 }
465 } 1129 }
466 1130
467 free_extent_buffer(eb); 1131 free_extent_buffer(eb);
1132
468 return ret; 1133 return ret;
469} 1134}
470 1135
471/* 1136/*
472 * calls iterate() for every inode that references the extent identified by 1137 * calls iterate() for every inode that references the extent identified by
473 * the given parameters. will use the path given as a parameter and return it 1138 * the given parameters.
474 * released.
475 * when the iterator function returns a non-zero value, iteration stops. 1139 * when the iterator function returns a non-zero value, iteration stops.
1140 * path is guaranteed to be in released state when iterate() is called.
476 */ 1141 */
477int iterate_extent_inodes(struct btrfs_fs_info *fs_info, 1142int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
478 struct btrfs_path *path, 1143 struct btrfs_path *path,
479 u64 extent_item_objectid, 1144 u64 extent_item_objectid, u64 extent_item_pos,
480 u64 extent_offset,
481 iterate_extent_inodes_t *iterate, void *ctx) 1145 iterate_extent_inodes_t *iterate, void *ctx)
482{ 1146{
483 unsigned long ptr = 0;
484 int last;
485 int ret; 1147 int ret;
486 int type;
487 u64 logical;
488 u32 item_size;
489 struct btrfs_extent_inline_ref *eiref;
490 struct btrfs_extent_data_ref *dref;
491 struct extent_buffer *eb;
492 struct btrfs_extent_item *ei;
493 struct btrfs_key key;
494 struct list_head data_refs = LIST_HEAD_INIT(data_refs); 1148 struct list_head data_refs = LIST_HEAD_INIT(data_refs);
495 struct list_head shared_refs = LIST_HEAD_INIT(shared_refs); 1149 struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
496 struct __data_ref *ref_d; 1150 struct btrfs_trans_handle *trans;
497 struct __shared_ref *ref_s; 1151 struct ulist *refs;
498 1152 struct ulist *roots;
499 eb = path->nodes[0]; 1153 struct ulist_node *ref_node = NULL;
500 ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item); 1154 struct ulist_node *root_node = NULL;
501 item_size = btrfs_item_size_nr(eb, path->slots[0]); 1155 struct seq_list seq_elem;
502 1156 struct btrfs_delayed_ref_root *delayed_refs;
503 /* first we iterate the inline refs, ... */ 1157
504 do { 1158 trans = btrfs_join_transaction(fs_info->extent_root);
505 last = __get_extent_inline_ref(&ptr, eb, ei, item_size, 1159 if (IS_ERR(trans))
506 &eiref, &type); 1160 return PTR_ERR(trans);
507 if (last == -ENOENT) { 1161
508 ret = 0; 1162 pr_debug("resolving all inodes for extent %llu\n",
509 break; 1163 extent_item_objectid);
510 } 1164
511 if (last < 0) { 1165 delayed_refs = &trans->transaction->delayed_refs;
512 ret = last; 1166 spin_lock(&delayed_refs->lock);
513 break; 1167 btrfs_get_delayed_seq(delayed_refs, &seq_elem);
514 } 1168 spin_unlock(&delayed_refs->lock);
1169
1170 ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1171 extent_item_pos, seq_elem.seq,
1172 &refs);
515 1173
516 if (type == BTRFS_EXTENT_DATA_REF_KEY) { 1174 if (ret)
517 dref = (struct btrfs_extent_data_ref *)(&eiref->offset); 1175 goto out;
518 ret = __data_list_add_eb(&data_refs, eb, dref);
519 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
520 logical = btrfs_extent_inline_ref_offset(eb, eiref);
521 ret = __shared_list_add(&shared_refs, logical);
522 }
523 } while (!ret && !last);
524 1176
525 /* ... then we proceed to in-tree references and ... */ 1177 while (!ret && (ref_node = ulist_next(refs, ref_node))) {
526 while (!ret) { 1178 ret = btrfs_find_all_roots(trans, fs_info, ref_node->val, -1,
527 ++path->slots[0]; 1179 seq_elem.seq, &roots);
528 if (path->slots[0] > btrfs_header_nritems(eb)) { 1180 if (ret)
529 ret = btrfs_next_leaf(fs_info->extent_root, path);
530 if (ret) {
531 if (ret == 1)
532 ret = 0; /* we're done */
533 break;
534 }
535 eb = path->nodes[0];
536 }
537 btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
538 if (key.objectid != extent_item_objectid)
539 break; 1181 break;
540 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) { 1182 while (!ret && (root_node = ulist_next(roots, root_node))) {
541 dref = btrfs_item_ptr(eb, path->slots[0], 1183 pr_debug("root %llu references leaf %llu\n",
542 struct btrfs_extent_data_ref); 1184 root_node->val, ref_node->val);
543 ret = __data_list_add_eb(&data_refs, eb, dref); 1185 ret = iterate_leaf_refs(fs_info, path, ref_node->val,
544 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) { 1186 extent_item_objectid,
545 ret = __shared_list_add(&shared_refs, key.offset); 1187 extent_item_pos, root_node->val,
1188 iterate, ctx);
546 } 1189 }
547 } 1190 }
548 1191
549 btrfs_release_path(path); 1192 ulist_free(refs);
550 1193 ulist_free(roots);
551 /* 1194out:
552 * ... only at the very end we can process the refs we found. this is 1195 btrfs_put_delayed_seq(delayed_refs, &seq_elem);
553 * because the iterator function we call is allowed to make tree lookups 1196 btrfs_end_transaction(trans, fs_info->extent_root);
554 * and we have to avoid deadlocks. additionally, we need more tree
555 * lookups ourselves for shared data refs.
556 */
557 while (!list_empty(&data_refs)) {
558 ref_d = list_first_entry(&data_refs, struct __data_ref, list);
559 list_del(&ref_d->list);
560 if (!ret)
561 ret = iterate(ref_d->inum, extent_offset +
562 ref_d->extent_data_item_offset,
563 ref_d->root, ctx);
564 kfree(ref_d);
565 }
566
567 while (!list_empty(&shared_refs)) {
568 ref_s = list_first_entry(&shared_refs, struct __shared_ref,
569 list);
570 list_del(&ref_s->list);
571 if (!ret)
572 ret = __iter_shared_inline_ref(fs_info,
573 ref_s->disk_byte,
574 extent_item_objectid,
575 extent_offset, path,
576 &data_refs,
577 iterate, ctx);
578 kfree(ref_s);
579 }
580
581 return ret; 1197 return ret;
582} 1198}
583 1199
@@ -586,19 +1202,20 @@ int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
586 iterate_extent_inodes_t *iterate, void *ctx) 1202 iterate_extent_inodes_t *iterate, void *ctx)
587{ 1203{
588 int ret; 1204 int ret;
589 u64 offset; 1205 u64 extent_item_pos;
590 struct btrfs_key found_key; 1206 struct btrfs_key found_key;
591 1207
592 ret = extent_from_logical(fs_info, logical, path, 1208 ret = extent_from_logical(fs_info, logical, path,
593 &found_key); 1209 &found_key);
1210 btrfs_release_path(path);
594 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK) 1211 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
595 ret = -EINVAL; 1212 ret = -EINVAL;
596 if (ret < 0) 1213 if (ret < 0)
597 return ret; 1214 return ret;
598 1215
599 offset = logical - found_key.objectid; 1216 extent_item_pos = logical - found_key.objectid;
600 ret = iterate_extent_inodes(fs_info, path, found_key.objectid, 1217 ret = iterate_extent_inodes(fs_info, path, found_key.objectid,
601 offset, iterate, ctx); 1218 extent_item_pos, iterate, ctx);
602 1219
603 return ret; 1220 return ret;
604} 1221}
@@ -643,6 +1260,10 @@ static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
643 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) { 1260 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
644 name_len = btrfs_inode_ref_name_len(eb, iref); 1261 name_len = btrfs_inode_ref_name_len(eb, iref);
645 /* path must be released before calling iterate()! */ 1262 /* path must be released before calling iterate()! */
1263 pr_debug("following ref at offset %u for inode %llu in "
1264 "tree %llu\n", cur,
1265 (unsigned long long)found_key.objectid,
1266 (unsigned long long)fs_root->objectid);
646 ret = iterate(parent, iref, eb, ctx); 1267 ret = iterate(parent, iref, eb, ctx);
647 if (ret) { 1268 if (ret) {
648 free_extent_buffer(eb); 1269 free_extent_buffer(eb);
@@ -683,10 +1304,14 @@ static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
683 return PTR_ERR(fspath); 1304 return PTR_ERR(fspath);
684 1305
685 if (fspath > fspath_min) { 1306 if (fspath > fspath_min) {
1307 pr_debug("path resolved: %s\n", fspath);
686 ipath->fspath->val[i] = (u64)(unsigned long)fspath; 1308 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
687 ++ipath->fspath->elem_cnt; 1309 ++ipath->fspath->elem_cnt;
688 ipath->fspath->bytes_left = fspath - fspath_min; 1310 ipath->fspath->bytes_left = fspath - fspath_min;
689 } else { 1311 } else {
1312 pr_debug("missed path, not enough space. missing bytes: %lu, "
1313 "constructed so far: %s\n",
1314 (unsigned long)(fspath_min - fspath), fspath_min);
690 ++ipath->fspath->elem_missed; 1315 ++ipath->fspath->elem_missed;
691 ipath->fspath->bytes_missing += fspath_min - fspath; 1316 ipath->fspath->bytes_missing += fspath_min - fspath;
692 ipath->fspath->bytes_left = 0; 1317 ipath->fspath->bytes_left = 0;
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-04 14:17:33 -0500