]>
Commit | Line | Data |
---|---|---|
a542ad1b JS |
1 | /* |
2 | * Copyright (C) 2011 STRATO. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
425d17a2 | 19 | #include <linux/vmalloc.h> |
a542ad1b JS |
20 | #include "ctree.h" |
21 | #include "disk-io.h" | |
22 | #include "backref.h" | |
8da6d581 JS |
23 | #include "ulist.h" |
24 | #include "transaction.h" | |
25 | #include "delayed-ref.h" | |
b916a59a | 26 | #include "locking.h" |
a542ad1b | 27 | |
976b1908 JS |
28 | struct extent_inode_elem { |
29 | u64 inum; | |
30 | u64 offset; | |
31 | struct extent_inode_elem *next; | |
32 | }; | |
33 | ||
34 | static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb, | |
35 | struct btrfs_file_extent_item *fi, | |
36 | u64 extent_item_pos, | |
37 | struct extent_inode_elem **eie) | |
38 | { | |
8ca15e05 | 39 | u64 offset = 0; |
976b1908 JS |
40 | struct extent_inode_elem *e; |
41 | ||
8ca15e05 JB |
42 | if (!btrfs_file_extent_compression(eb, fi) && |
43 | !btrfs_file_extent_encryption(eb, fi) && | |
44 | !btrfs_file_extent_other_encoding(eb, fi)) { | |
45 | u64 data_offset; | |
46 | u64 data_len; | |
976b1908 | 47 | |
8ca15e05 JB |
48 | data_offset = btrfs_file_extent_offset(eb, fi); |
49 | data_len = btrfs_file_extent_num_bytes(eb, fi); | |
50 | ||
51 | if (extent_item_pos < data_offset || | |
52 | extent_item_pos >= data_offset + data_len) | |
53 | return 1; | |
54 | offset = extent_item_pos - data_offset; | |
55 | } | |
976b1908 JS |
56 | |
57 | e = kmalloc(sizeof(*e), GFP_NOFS); | |
58 | if (!e) | |
59 | return -ENOMEM; | |
60 | ||
61 | e->next = *eie; | |
62 | e->inum = key->objectid; | |
8ca15e05 | 63 | e->offset = key->offset + offset; |
976b1908 JS |
64 | *eie = e; |
65 | ||
66 | return 0; | |
67 | } | |
68 | ||
69 | static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte, | |
70 | u64 extent_item_pos, | |
71 | struct extent_inode_elem **eie) | |
72 | { | |
73 | u64 disk_byte; | |
74 | struct btrfs_key key; | |
75 | struct btrfs_file_extent_item *fi; | |
76 | int slot; | |
77 | int nritems; | |
78 | int extent_type; | |
79 | int ret; | |
80 | ||
81 | /* | |
82 | * from the shared data ref, we only have the leaf but we need | |
83 | * the key. thus, we must look into all items and see that we | |
84 | * find one (some) with a reference to our extent item. | |
85 | */ | |
86 | nritems = btrfs_header_nritems(eb); | |
87 | for (slot = 0; slot < nritems; ++slot) { | |
88 | btrfs_item_key_to_cpu(eb, &key, slot); | |
89 | if (key.type != BTRFS_EXTENT_DATA_KEY) | |
90 | continue; | |
91 | fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item); | |
92 | extent_type = btrfs_file_extent_type(eb, fi); | |
93 | if (extent_type == BTRFS_FILE_EXTENT_INLINE) | |
94 | continue; | |
95 | /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */ | |
96 | disk_byte = btrfs_file_extent_disk_bytenr(eb, fi); | |
97 | if (disk_byte != wanted_disk_byte) | |
98 | continue; | |
99 | ||
100 | ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie); | |
101 | if (ret < 0) | |
102 | return ret; | |
103 | } | |
104 | ||
105 | return 0; | |
106 | } | |
107 | ||
8da6d581 JS |
108 | /* |
109 | * this structure records all encountered refs on the way up to the root | |
110 | */ | |
111 | struct __prelim_ref { | |
112 | struct list_head list; | |
113 | u64 root_id; | |
d5c88b73 | 114 | struct btrfs_key key_for_search; |
8da6d581 JS |
115 | int level; |
116 | int count; | |
3301958b | 117 | struct extent_inode_elem *inode_list; |
8da6d581 JS |
118 | u64 parent; |
119 | u64 wanted_disk_byte; | |
120 | }; | |
121 | ||
d5c88b73 JS |
122 | /* |
123 | * the rules for all callers of this function are: | |
124 | * - obtaining the parent is the goal | |
125 | * - if you add a key, you must know that it is a correct key | |
126 | * - if you cannot add the parent or a correct key, then we will look into the | |
127 | * block later to set a correct key | |
128 | * | |
129 | * delayed refs | |
130 | * ============ | |
131 | * backref type | shared | indirect | shared | indirect | |
132 | * information | tree | tree | data | data | |
133 | * --------------------+--------+----------+--------+---------- | |
134 | * parent logical | y | - | - | - | |
135 | * key to resolve | - | y | y | y | |
136 | * tree block logical | - | - | - | - | |
137 | * root for resolving | y | y | y | y | |
138 | * | |
139 | * - column 1: we've the parent -> done | |
140 | * - column 2, 3, 4: we use the key to find the parent | |
141 | * | |
142 | * on disk refs (inline or keyed) | |
143 | * ============================== | |
144 | * backref type | shared | indirect | shared | indirect | |
145 | * information | tree | tree | data | data | |
146 | * --------------------+--------+----------+--------+---------- | |
147 | * parent logical | y | - | y | - | |
148 | * key to resolve | - | - | - | y | |
149 | * tree block logical | y | y | y | y | |
150 | * root for resolving | - | y | y | y | |
151 | * | |
152 | * - column 1, 3: we've the parent -> done | |
153 | * - column 2: we take the first key from the block to find the parent | |
154 | * (see __add_missing_keys) | |
155 | * - column 4: we use the key to find the parent | |
156 | * | |
157 | * additional information that's available but not required to find the parent | |
158 | * block might help in merging entries to gain some speed. | |
159 | */ | |
160 | ||
8da6d581 | 161 | static int __add_prelim_ref(struct list_head *head, u64 root_id, |
d5c88b73 JS |
162 | struct btrfs_key *key, int level, |
163 | u64 parent, u64 wanted_disk_byte, int count) | |
8da6d581 JS |
164 | { |
165 | struct __prelim_ref *ref; | |
166 | ||
167 | /* in case we're adding delayed refs, we're holding the refs spinlock */ | |
168 | ref = kmalloc(sizeof(*ref), GFP_ATOMIC); | |
169 | if (!ref) | |
170 | return -ENOMEM; | |
171 | ||
172 | ref->root_id = root_id; | |
173 | if (key) | |
d5c88b73 | 174 | ref->key_for_search = *key; |
8da6d581 | 175 | else |
d5c88b73 | 176 | memset(&ref->key_for_search, 0, sizeof(ref->key_for_search)); |
8da6d581 | 177 | |
3301958b | 178 | ref->inode_list = NULL; |
8da6d581 JS |
179 | ref->level = level; |
180 | ref->count = count; | |
181 | ref->parent = parent; | |
182 | ref->wanted_disk_byte = wanted_disk_byte; | |
183 | list_add_tail(&ref->list, head); | |
184 | ||
185 | return 0; | |
186 | } | |
187 | ||
188 | static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path, | |
976b1908 | 189 | struct ulist *parents, int level, |
69bca40d | 190 | struct btrfs_key *key_for_search, u64 time_seq, |
3d7806ec | 191 | u64 wanted_disk_byte, |
976b1908 | 192 | const u64 *extent_item_pos) |
8da6d581 | 193 | { |
69bca40d AB |
194 | int ret = 0; |
195 | int slot; | |
196 | struct extent_buffer *eb; | |
197 | struct btrfs_key key; | |
8da6d581 | 198 | struct btrfs_file_extent_item *fi; |
ed8c4913 | 199 | struct extent_inode_elem *eie = NULL, *old = NULL; |
8da6d581 JS |
200 | u64 disk_byte; |
201 | ||
69bca40d AB |
202 | if (level != 0) { |
203 | eb = path->nodes[level]; | |
204 | ret = ulist_add(parents, eb->start, 0, GFP_NOFS); | |
3301958b JS |
205 | if (ret < 0) |
206 | return ret; | |
8da6d581 | 207 | return 0; |
69bca40d | 208 | } |
8da6d581 JS |
209 | |
210 | /* | |
69bca40d AB |
211 | * We normally enter this function with the path already pointing to |
212 | * the first item to check. But sometimes, we may enter it with | |
213 | * slot==nritems. In that case, go to the next leaf before we continue. | |
8da6d581 | 214 | */ |
69bca40d | 215 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) |
3d7806ec | 216 | ret = btrfs_next_old_leaf(root, path, time_seq); |
8da6d581 | 217 | |
69bca40d | 218 | while (!ret) { |
8da6d581 | 219 | eb = path->nodes[0]; |
69bca40d AB |
220 | slot = path->slots[0]; |
221 | ||
222 | btrfs_item_key_to_cpu(eb, &key, slot); | |
223 | ||
224 | if (key.objectid != key_for_search->objectid || | |
225 | key.type != BTRFS_EXTENT_DATA_KEY) | |
226 | break; | |
227 | ||
228 | fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item); | |
229 | disk_byte = btrfs_file_extent_disk_bytenr(eb, fi); | |
230 | ||
231 | if (disk_byte == wanted_disk_byte) { | |
232 | eie = NULL; | |
ed8c4913 | 233 | old = NULL; |
69bca40d AB |
234 | if (extent_item_pos) { |
235 | ret = check_extent_in_eb(&key, eb, fi, | |
236 | *extent_item_pos, | |
237 | &eie); | |
238 | if (ret < 0) | |
239 | break; | |
240 | } | |
ed8c4913 JB |
241 | if (ret > 0) |
242 | goto next; | |
243 | ret = ulist_add_merge(parents, eb->start, | |
244 | (uintptr_t)eie, | |
245 | (u64 *)&old, GFP_NOFS); | |
246 | if (ret < 0) | |
247 | break; | |
248 | if (!ret && extent_item_pos) { | |
249 | while (old->next) | |
250 | old = old->next; | |
251 | old->next = eie; | |
69bca40d | 252 | } |
8da6d581 | 253 | } |
ed8c4913 | 254 | next: |
69bca40d | 255 | ret = btrfs_next_old_item(root, path, time_seq); |
8da6d581 JS |
256 | } |
257 | ||
69bca40d AB |
258 | if (ret > 0) |
259 | ret = 0; | |
260 | return ret; | |
8da6d581 JS |
261 | } |
262 | ||
263 | /* | |
264 | * resolve an indirect backref in the form (root_id, key, level) | |
265 | * to a logical address | |
266 | */ | |
267 | static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info, | |
da61d31a JB |
268 | struct btrfs_path *path, u64 time_seq, |
269 | struct __prelim_ref *ref, | |
270 | struct ulist *parents, | |
271 | const u64 *extent_item_pos) | |
8da6d581 | 272 | { |
8da6d581 JS |
273 | struct btrfs_root *root; |
274 | struct btrfs_key root_key; | |
8da6d581 JS |
275 | struct extent_buffer *eb; |
276 | int ret = 0; | |
277 | int root_level; | |
278 | int level = ref->level; | |
279 | ||
8da6d581 JS |
280 | root_key.objectid = ref->root_id; |
281 | root_key.type = BTRFS_ROOT_ITEM_KEY; | |
282 | root_key.offset = (u64)-1; | |
283 | root = btrfs_read_fs_root_no_name(fs_info, &root_key); | |
284 | if (IS_ERR(root)) { | |
285 | ret = PTR_ERR(root); | |
286 | goto out; | |
287 | } | |
288 | ||
5b6602e7 | 289 | root_level = btrfs_old_root_level(root, time_seq); |
8da6d581 JS |
290 | |
291 | if (root_level + 1 == level) | |
292 | goto out; | |
293 | ||
294 | path->lowest_level = level; | |
8445f61c | 295 | ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq); |
8da6d581 JS |
296 | pr_debug("search slot in root %llu (level %d, ref count %d) returned " |
297 | "%d for key (%llu %u %llu)\n", | |
298 | (unsigned long long)ref->root_id, level, ref->count, ret, | |
d5c88b73 JS |
299 | (unsigned long long)ref->key_for_search.objectid, |
300 | ref->key_for_search.type, | |
301 | (unsigned long long)ref->key_for_search.offset); | |
8da6d581 JS |
302 | if (ret < 0) |
303 | goto out; | |
304 | ||
305 | eb = path->nodes[level]; | |
9345457f JS |
306 | while (!eb) { |
307 | if (!level) { | |
308 | WARN_ON(1); | |
309 | ret = 1; | |
310 | goto out; | |
311 | } | |
312 | level--; | |
313 | eb = path->nodes[level]; | |
8da6d581 JS |
314 | } |
315 | ||
69bca40d AB |
316 | ret = add_all_parents(root, path, parents, level, &ref->key_for_search, |
317 | time_seq, ref->wanted_disk_byte, | |
318 | extent_item_pos); | |
8da6d581 | 319 | out: |
da61d31a JB |
320 | path->lowest_level = 0; |
321 | btrfs_release_path(path); | |
8da6d581 JS |
322 | return ret; |
323 | } | |
324 | ||
325 | /* | |
326 | * resolve all indirect backrefs from the list | |
327 | */ | |
328 | static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info, | |
da61d31a | 329 | struct btrfs_path *path, u64 time_seq, |
976b1908 JS |
330 | struct list_head *head, |
331 | const u64 *extent_item_pos) | |
8da6d581 JS |
332 | { |
333 | int err; | |
334 | int ret = 0; | |
335 | struct __prelim_ref *ref; | |
336 | struct __prelim_ref *ref_safe; | |
337 | struct __prelim_ref *new_ref; | |
338 | struct ulist *parents; | |
339 | struct ulist_node *node; | |
cd1b413c | 340 | struct ulist_iterator uiter; |
8da6d581 JS |
341 | |
342 | parents = ulist_alloc(GFP_NOFS); | |
343 | if (!parents) | |
344 | return -ENOMEM; | |
345 | ||
346 | /* | |
347 | * _safe allows us to insert directly after the current item without | |
348 | * iterating over the newly inserted items. | |
349 | * we're also allowed to re-assign ref during iteration. | |
350 | */ | |
351 | list_for_each_entry_safe(ref, ref_safe, head, list) { | |
352 | if (ref->parent) /* already direct */ | |
353 | continue; | |
354 | if (ref->count == 0) | |
355 | continue; | |
da61d31a JB |
356 | err = __resolve_indirect_ref(fs_info, path, time_seq, ref, |
357 | parents, extent_item_pos); | |
e36902d4 WS |
358 | if (err == -ENOMEM) |
359 | goto out; | |
ca60ebfa | 360 | if (err) |
8da6d581 | 361 | continue; |
8da6d581 JS |
362 | |
363 | /* we put the first parent into the ref at hand */ | |
cd1b413c JS |
364 | ULIST_ITER_INIT(&uiter); |
365 | node = ulist_next(parents, &uiter); | |
8da6d581 | 366 | ref->parent = node ? node->val : 0; |
995e01b7 JS |
367 | ref->inode_list = node ? |
368 | (struct extent_inode_elem *)(uintptr_t)node->aux : 0; | |
8da6d581 JS |
369 | |
370 | /* additional parents require new refs being added here */ | |
cd1b413c | 371 | while ((node = ulist_next(parents, &uiter))) { |
8da6d581 JS |
372 | new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS); |
373 | if (!new_ref) { | |
374 | ret = -ENOMEM; | |
e36902d4 | 375 | goto out; |
8da6d581 JS |
376 | } |
377 | memcpy(new_ref, ref, sizeof(*ref)); | |
378 | new_ref->parent = node->val; | |
995e01b7 JS |
379 | new_ref->inode_list = (struct extent_inode_elem *) |
380 | (uintptr_t)node->aux; | |
8da6d581 JS |
381 | list_add(&new_ref->list, &ref->list); |
382 | } | |
383 | ulist_reinit(parents); | |
384 | } | |
e36902d4 | 385 | out: |
8da6d581 JS |
386 | ulist_free(parents); |
387 | return ret; | |
388 | } | |
389 | ||
d5c88b73 JS |
390 | static inline int ref_for_same_block(struct __prelim_ref *ref1, |
391 | struct __prelim_ref *ref2) | |
392 | { | |
393 | if (ref1->level != ref2->level) | |
394 | return 0; | |
395 | if (ref1->root_id != ref2->root_id) | |
396 | return 0; | |
397 | if (ref1->key_for_search.type != ref2->key_for_search.type) | |
398 | return 0; | |
399 | if (ref1->key_for_search.objectid != ref2->key_for_search.objectid) | |
400 | return 0; | |
401 | if (ref1->key_for_search.offset != ref2->key_for_search.offset) | |
402 | return 0; | |
403 | if (ref1->parent != ref2->parent) | |
404 | return 0; | |
405 | ||
406 | return 1; | |
407 | } | |
408 | ||
409 | /* | |
410 | * read tree blocks and add keys where required. | |
411 | */ | |
412 | static int __add_missing_keys(struct btrfs_fs_info *fs_info, | |
413 | struct list_head *head) | |
414 | { | |
415 | struct list_head *pos; | |
416 | struct extent_buffer *eb; | |
417 | ||
418 | list_for_each(pos, head) { | |
419 | struct __prelim_ref *ref; | |
420 | ref = list_entry(pos, struct __prelim_ref, list); | |
421 | ||
422 | if (ref->parent) | |
423 | continue; | |
424 | if (ref->key_for_search.type) | |
425 | continue; | |
426 | BUG_ON(!ref->wanted_disk_byte); | |
427 | eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte, | |
428 | fs_info->tree_root->leafsize, 0); | |
416bc658 JB |
429 | if (!eb || !extent_buffer_uptodate(eb)) { |
430 | free_extent_buffer(eb); | |
431 | return -EIO; | |
432 | } | |
d5c88b73 JS |
433 | btrfs_tree_read_lock(eb); |
434 | if (btrfs_header_level(eb) == 0) | |
435 | btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0); | |
436 | else | |
437 | btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0); | |
438 | btrfs_tree_read_unlock(eb); | |
439 | free_extent_buffer(eb); | |
440 | } | |
441 | return 0; | |
442 | } | |
443 | ||
8da6d581 JS |
444 | /* |
445 | * merge two lists of backrefs and adjust counts accordingly | |
446 | * | |
447 | * mode = 1: merge identical keys, if key is set | |
d5c88b73 JS |
448 | * FIXME: if we add more keys in __add_prelim_ref, we can merge more here. |
449 | * additionally, we could even add a key range for the blocks we | |
450 | * looked into to merge even more (-> replace unresolved refs by those | |
451 | * having a parent). | |
8da6d581 JS |
452 | * mode = 2: merge identical parents |
453 | */ | |
692206b1 | 454 | static void __merge_refs(struct list_head *head, int mode) |
8da6d581 JS |
455 | { |
456 | struct list_head *pos1; | |
457 | ||
458 | list_for_each(pos1, head) { | |
459 | struct list_head *n2; | |
460 | struct list_head *pos2; | |
461 | struct __prelim_ref *ref1; | |
462 | ||
463 | ref1 = list_entry(pos1, struct __prelim_ref, list); | |
464 | ||
8da6d581 JS |
465 | for (pos2 = pos1->next, n2 = pos2->next; pos2 != head; |
466 | pos2 = n2, n2 = pos2->next) { | |
467 | struct __prelim_ref *ref2; | |
d5c88b73 | 468 | struct __prelim_ref *xchg; |
3ef5969c | 469 | struct extent_inode_elem *eie; |
8da6d581 JS |
470 | |
471 | ref2 = list_entry(pos2, struct __prelim_ref, list); | |
472 | ||
473 | if (mode == 1) { | |
d5c88b73 | 474 | if (!ref_for_same_block(ref1, ref2)) |
8da6d581 | 475 | continue; |
d5c88b73 JS |
476 | if (!ref1->parent && ref2->parent) { |
477 | xchg = ref1; | |
478 | ref1 = ref2; | |
479 | ref2 = xchg; | |
480 | } | |
8da6d581 JS |
481 | } else { |
482 | if (ref1->parent != ref2->parent) | |
483 | continue; | |
8da6d581 | 484 | } |
3ef5969c AB |
485 | |
486 | eie = ref1->inode_list; | |
487 | while (eie && eie->next) | |
488 | eie = eie->next; | |
489 | if (eie) | |
490 | eie->next = ref2->inode_list; | |
491 | else | |
492 | ref1->inode_list = ref2->inode_list; | |
493 | ref1->count += ref2->count; | |
494 | ||
8da6d581 JS |
495 | list_del(&ref2->list); |
496 | kfree(ref2); | |
497 | } | |
498 | ||
499 | } | |
8da6d581 JS |
500 | } |
501 | ||
502 | /* | |
503 | * add all currently queued delayed refs from this head whose seq nr is | |
504 | * smaller or equal that seq to the list | |
505 | */ | |
506 | static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq, | |
8da6d581 JS |
507 | struct list_head *prefs) |
508 | { | |
509 | struct btrfs_delayed_extent_op *extent_op = head->extent_op; | |
510 | struct rb_node *n = &head->node.rb_node; | |
d5c88b73 JS |
511 | struct btrfs_key key; |
512 | struct btrfs_key op_key = {0}; | |
8da6d581 | 513 | int sgn; |
b1375d64 | 514 | int ret = 0; |
8da6d581 JS |
515 | |
516 | if (extent_op && extent_op->update_key) | |
d5c88b73 | 517 | btrfs_disk_key_to_cpu(&op_key, &extent_op->key); |
8da6d581 JS |
518 | |
519 | while ((n = rb_prev(n))) { | |
520 | struct btrfs_delayed_ref_node *node; | |
521 | node = rb_entry(n, struct btrfs_delayed_ref_node, | |
522 | rb_node); | |
523 | if (node->bytenr != head->node.bytenr) | |
524 | break; | |
525 | WARN_ON(node->is_head); | |
526 | ||
527 | if (node->seq > seq) | |
528 | continue; | |
529 | ||
530 | switch (node->action) { | |
531 | case BTRFS_ADD_DELAYED_EXTENT: | |
532 | case BTRFS_UPDATE_DELAYED_HEAD: | |
533 | WARN_ON(1); | |
534 | continue; | |
535 | case BTRFS_ADD_DELAYED_REF: | |
536 | sgn = 1; | |
537 | break; | |
538 | case BTRFS_DROP_DELAYED_REF: | |
539 | sgn = -1; | |
540 | break; | |
541 | default: | |
542 | BUG_ON(1); | |
543 | } | |
544 | switch (node->type) { | |
545 | case BTRFS_TREE_BLOCK_REF_KEY: { | |
546 | struct btrfs_delayed_tree_ref *ref; | |
547 | ||
548 | ref = btrfs_delayed_node_to_tree_ref(node); | |
d5c88b73 | 549 | ret = __add_prelim_ref(prefs, ref->root, &op_key, |
8da6d581 JS |
550 | ref->level + 1, 0, node->bytenr, |
551 | node->ref_mod * sgn); | |
552 | break; | |
553 | } | |
554 | case BTRFS_SHARED_BLOCK_REF_KEY: { | |
555 | struct btrfs_delayed_tree_ref *ref; | |
556 | ||
557 | ref = btrfs_delayed_node_to_tree_ref(node); | |
d5c88b73 | 558 | ret = __add_prelim_ref(prefs, ref->root, NULL, |
8da6d581 JS |
559 | ref->level + 1, ref->parent, |
560 | node->bytenr, | |
561 | node->ref_mod * sgn); | |
562 | break; | |
563 | } | |
564 | case BTRFS_EXTENT_DATA_REF_KEY: { | |
565 | struct btrfs_delayed_data_ref *ref; | |
8da6d581 JS |
566 | ref = btrfs_delayed_node_to_data_ref(node); |
567 | ||
568 | key.objectid = ref->objectid; | |
569 | key.type = BTRFS_EXTENT_DATA_KEY; | |
570 | key.offset = ref->offset; | |
571 | ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0, | |
572 | node->bytenr, | |
573 | node->ref_mod * sgn); | |
574 | break; | |
575 | } | |
576 | case BTRFS_SHARED_DATA_REF_KEY: { | |
577 | struct btrfs_delayed_data_ref *ref; | |
8da6d581 JS |
578 | |
579 | ref = btrfs_delayed_node_to_data_ref(node); | |
580 | ||
581 | key.objectid = ref->objectid; | |
582 | key.type = BTRFS_EXTENT_DATA_KEY; | |
583 | key.offset = ref->offset; | |
584 | ret = __add_prelim_ref(prefs, ref->root, &key, 0, | |
585 | ref->parent, node->bytenr, | |
586 | node->ref_mod * sgn); | |
587 | break; | |
588 | } | |
589 | default: | |
590 | WARN_ON(1); | |
591 | } | |
1149ab6b WS |
592 | if (ret) |
593 | return ret; | |
8da6d581 JS |
594 | } |
595 | ||
596 | return 0; | |
597 | } | |
598 | ||
599 | /* | |
600 | * add all inline backrefs for bytenr to the list | |
601 | */ | |
602 | static int __add_inline_refs(struct btrfs_fs_info *fs_info, | |
603 | struct btrfs_path *path, u64 bytenr, | |
d5c88b73 | 604 | int *info_level, struct list_head *prefs) |
8da6d581 | 605 | { |
b1375d64 | 606 | int ret = 0; |
8da6d581 JS |
607 | int slot; |
608 | struct extent_buffer *leaf; | |
609 | struct btrfs_key key; | |
261c84b6 | 610 | struct btrfs_key found_key; |
8da6d581 JS |
611 | unsigned long ptr; |
612 | unsigned long end; | |
613 | struct btrfs_extent_item *ei; | |
614 | u64 flags; | |
615 | u64 item_size; | |
616 | ||
617 | /* | |
618 | * enumerate all inline refs | |
619 | */ | |
620 | leaf = path->nodes[0]; | |
dadcaf78 | 621 | slot = path->slots[0]; |
8da6d581 JS |
622 | |
623 | item_size = btrfs_item_size_nr(leaf, slot); | |
624 | BUG_ON(item_size < sizeof(*ei)); | |
625 | ||
626 | ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item); | |
627 | flags = btrfs_extent_flags(leaf, ei); | |
261c84b6 | 628 | btrfs_item_key_to_cpu(leaf, &found_key, slot); |
8da6d581 JS |
629 | |
630 | ptr = (unsigned long)(ei + 1); | |
631 | end = (unsigned long)ei + item_size; | |
632 | ||
261c84b6 JB |
633 | if (found_key.type == BTRFS_EXTENT_ITEM_KEY && |
634 | flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) { | |
8da6d581 | 635 | struct btrfs_tree_block_info *info; |
8da6d581 JS |
636 | |
637 | info = (struct btrfs_tree_block_info *)ptr; | |
638 | *info_level = btrfs_tree_block_level(leaf, info); | |
8da6d581 JS |
639 | ptr += sizeof(struct btrfs_tree_block_info); |
640 | BUG_ON(ptr > end); | |
261c84b6 JB |
641 | } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) { |
642 | *info_level = found_key.offset; | |
8da6d581 JS |
643 | } else { |
644 | BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA)); | |
645 | } | |
646 | ||
647 | while (ptr < end) { | |
648 | struct btrfs_extent_inline_ref *iref; | |
649 | u64 offset; | |
650 | int type; | |
651 | ||
652 | iref = (struct btrfs_extent_inline_ref *)ptr; | |
653 | type = btrfs_extent_inline_ref_type(leaf, iref); | |
654 | offset = btrfs_extent_inline_ref_offset(leaf, iref); | |
655 | ||
656 | switch (type) { | |
657 | case BTRFS_SHARED_BLOCK_REF_KEY: | |
d5c88b73 | 658 | ret = __add_prelim_ref(prefs, 0, NULL, |
8da6d581 JS |
659 | *info_level + 1, offset, |
660 | bytenr, 1); | |
661 | break; | |
662 | case BTRFS_SHARED_DATA_REF_KEY: { | |
663 | struct btrfs_shared_data_ref *sdref; | |
664 | int count; | |
665 | ||
666 | sdref = (struct btrfs_shared_data_ref *)(iref + 1); | |
667 | count = btrfs_shared_data_ref_count(leaf, sdref); | |
668 | ret = __add_prelim_ref(prefs, 0, NULL, 0, offset, | |
669 | bytenr, count); | |
670 | break; | |
671 | } | |
672 | case BTRFS_TREE_BLOCK_REF_KEY: | |
d5c88b73 JS |
673 | ret = __add_prelim_ref(prefs, offset, NULL, |
674 | *info_level + 1, 0, | |
675 | bytenr, 1); | |
8da6d581 JS |
676 | break; |
677 | case BTRFS_EXTENT_DATA_REF_KEY: { | |
678 | struct btrfs_extent_data_ref *dref; | |
679 | int count; | |
680 | u64 root; | |
681 | ||
682 | dref = (struct btrfs_extent_data_ref *)(&iref->offset); | |
683 | count = btrfs_extent_data_ref_count(leaf, dref); | |
684 | key.objectid = btrfs_extent_data_ref_objectid(leaf, | |
685 | dref); | |
686 | key.type = BTRFS_EXTENT_DATA_KEY; | |
687 | key.offset = btrfs_extent_data_ref_offset(leaf, dref); | |
688 | root = btrfs_extent_data_ref_root(leaf, dref); | |
d5c88b73 JS |
689 | ret = __add_prelim_ref(prefs, root, &key, 0, 0, |
690 | bytenr, count); | |
8da6d581 JS |
691 | break; |
692 | } | |
693 | default: | |
694 | WARN_ON(1); | |
695 | } | |
1149ab6b WS |
696 | if (ret) |
697 | return ret; | |
8da6d581 JS |
698 | ptr += btrfs_extent_inline_ref_size(type); |
699 | } | |
700 | ||
701 | return 0; | |
702 | } | |
703 | ||
704 | /* | |
705 | * add all non-inline backrefs for bytenr to the list | |
706 | */ | |
707 | static int __add_keyed_refs(struct btrfs_fs_info *fs_info, | |
708 | struct btrfs_path *path, u64 bytenr, | |
d5c88b73 | 709 | int info_level, struct list_head *prefs) |
8da6d581 JS |
710 | { |
711 | struct btrfs_root *extent_root = fs_info->extent_root; | |
712 | int ret; | |
713 | int slot; | |
714 | struct extent_buffer *leaf; | |
715 | struct btrfs_key key; | |
716 | ||
717 | while (1) { | |
718 | ret = btrfs_next_item(extent_root, path); | |
719 | if (ret < 0) | |
720 | break; | |
721 | if (ret) { | |
722 | ret = 0; | |
723 | break; | |
724 | } | |
725 | ||
726 | slot = path->slots[0]; | |
727 | leaf = path->nodes[0]; | |
728 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
729 | ||
730 | if (key.objectid != bytenr) | |
731 | break; | |
732 | if (key.type < BTRFS_TREE_BLOCK_REF_KEY) | |
733 | continue; | |
734 | if (key.type > BTRFS_SHARED_DATA_REF_KEY) | |
735 | break; | |
736 | ||
737 | switch (key.type) { | |
738 | case BTRFS_SHARED_BLOCK_REF_KEY: | |
d5c88b73 | 739 | ret = __add_prelim_ref(prefs, 0, NULL, |
8da6d581 JS |
740 | info_level + 1, key.offset, |
741 | bytenr, 1); | |
742 | break; | |
743 | case BTRFS_SHARED_DATA_REF_KEY: { | |
744 | struct btrfs_shared_data_ref *sdref; | |
745 | int count; | |
746 | ||
747 | sdref = btrfs_item_ptr(leaf, slot, | |
748 | struct btrfs_shared_data_ref); | |
749 | count = btrfs_shared_data_ref_count(leaf, sdref); | |
750 | ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset, | |
751 | bytenr, count); | |
752 | break; | |
753 | } | |
754 | case BTRFS_TREE_BLOCK_REF_KEY: | |
d5c88b73 JS |
755 | ret = __add_prelim_ref(prefs, key.offset, NULL, |
756 | info_level + 1, 0, | |
757 | bytenr, 1); | |
8da6d581 JS |
758 | break; |
759 | case BTRFS_EXTENT_DATA_REF_KEY: { | |
760 | struct btrfs_extent_data_ref *dref; | |
761 | int count; | |
762 | u64 root; | |
763 | ||
764 | dref = btrfs_item_ptr(leaf, slot, | |
765 | struct btrfs_extent_data_ref); | |
766 | count = btrfs_extent_data_ref_count(leaf, dref); | |
767 | key.objectid = btrfs_extent_data_ref_objectid(leaf, | |
768 | dref); | |
769 | key.type = BTRFS_EXTENT_DATA_KEY; | |
770 | key.offset = btrfs_extent_data_ref_offset(leaf, dref); | |
771 | root = btrfs_extent_data_ref_root(leaf, dref); | |
772 | ret = __add_prelim_ref(prefs, root, &key, 0, 0, | |
d5c88b73 | 773 | bytenr, count); |
8da6d581 JS |
774 | break; |
775 | } | |
776 | default: | |
777 | WARN_ON(1); | |
778 | } | |
1149ab6b WS |
779 | if (ret) |
780 | return ret; | |
781 | ||
8da6d581 JS |
782 | } |
783 | ||
784 | return ret; | |
785 | } | |
786 | ||
787 | /* | |
788 | * this adds all existing backrefs (inline backrefs, backrefs and delayed | |
789 | * refs) for the given bytenr to the refs list, merges duplicates and resolves | |
790 | * indirect refs to their parent bytenr. | |
791 | * When roots are found, they're added to the roots list | |
792 | * | |
793 | * FIXME some caching might speed things up | |
794 | */ | |
795 | static int find_parent_nodes(struct btrfs_trans_handle *trans, | |
796 | struct btrfs_fs_info *fs_info, u64 bytenr, | |
097b8a7c JS |
797 | u64 time_seq, struct ulist *refs, |
798 | struct ulist *roots, const u64 *extent_item_pos) | |
8da6d581 JS |
799 | { |
800 | struct btrfs_key key; | |
801 | struct btrfs_path *path; | |
8da6d581 | 802 | struct btrfs_delayed_ref_root *delayed_refs = NULL; |
d3b01064 | 803 | struct btrfs_delayed_ref_head *head; |
8da6d581 JS |
804 | int info_level = 0; |
805 | int ret; | |
806 | struct list_head prefs_delayed; | |
807 | struct list_head prefs; | |
808 | struct __prelim_ref *ref; | |
809 | ||
810 | INIT_LIST_HEAD(&prefs); | |
811 | INIT_LIST_HEAD(&prefs_delayed); | |
812 | ||
813 | key.objectid = bytenr; | |
8da6d581 | 814 | key.offset = (u64)-1; |
261c84b6 JB |
815 | if (btrfs_fs_incompat(fs_info, SKINNY_METADATA)) |
816 | key.type = BTRFS_METADATA_ITEM_KEY; | |
817 | else | |
818 | key.type = BTRFS_EXTENT_ITEM_KEY; | |
8da6d581 JS |
819 | |
820 | path = btrfs_alloc_path(); | |
821 | if (!path) | |
822 | return -ENOMEM; | |
da61d31a JB |
823 | if (!trans) |
824 | path->search_commit_root = 1; | |
8da6d581 JS |
825 | |
826 | /* | |
827 | * grab both a lock on the path and a lock on the delayed ref head. | |
828 | * We need both to get a consistent picture of how the refs look | |
829 | * at a specified point in time | |
830 | */ | |
831 | again: | |
d3b01064 LZ |
832 | head = NULL; |
833 | ||
8da6d581 JS |
834 | ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0); |
835 | if (ret < 0) | |
836 | goto out; | |
837 | BUG_ON(ret == 0); | |
838 | ||
da61d31a | 839 | if (trans) { |
7a3ae2f8 JS |
840 | /* |
841 | * look if there are updates for this ref queued and lock the | |
842 | * head | |
843 | */ | |
844 | delayed_refs = &trans->transaction->delayed_refs; | |
845 | spin_lock(&delayed_refs->lock); | |
846 | head = btrfs_find_delayed_ref_head(trans, bytenr); | |
847 | if (head) { | |
848 | if (!mutex_trylock(&head->mutex)) { | |
849 | atomic_inc(&head->node.refs); | |
850 | spin_unlock(&delayed_refs->lock); | |
851 | ||
852 | btrfs_release_path(path); | |
853 | ||
854 | /* | |
855 | * Mutex was contended, block until it's | |
856 | * released and try again | |
857 | */ | |
858 | mutex_lock(&head->mutex); | |
859 | mutex_unlock(&head->mutex); | |
860 | btrfs_put_delayed_ref(&head->node); | |
861 | goto again; | |
862 | } | |
097b8a7c | 863 | ret = __add_delayed_refs(head, time_seq, |
8445f61c | 864 | &prefs_delayed); |
155725c9 | 865 | mutex_unlock(&head->mutex); |
7a3ae2f8 JS |
866 | if (ret) { |
867 | spin_unlock(&delayed_refs->lock); | |
868 | goto out; | |
869 | } | |
d3b01064 | 870 | } |
7a3ae2f8 | 871 | spin_unlock(&delayed_refs->lock); |
8da6d581 | 872 | } |
8da6d581 JS |
873 | |
874 | if (path->slots[0]) { | |
875 | struct extent_buffer *leaf; | |
876 | int slot; | |
877 | ||
dadcaf78 | 878 | path->slots[0]--; |
8da6d581 | 879 | leaf = path->nodes[0]; |
dadcaf78 | 880 | slot = path->slots[0]; |
8da6d581 JS |
881 | btrfs_item_key_to_cpu(leaf, &key, slot); |
882 | if (key.objectid == bytenr && | |
261c84b6 JB |
883 | (key.type == BTRFS_EXTENT_ITEM_KEY || |
884 | key.type == BTRFS_METADATA_ITEM_KEY)) { | |
8da6d581 | 885 | ret = __add_inline_refs(fs_info, path, bytenr, |
d5c88b73 | 886 | &info_level, &prefs); |
8da6d581 JS |
887 | if (ret) |
888 | goto out; | |
d5c88b73 | 889 | ret = __add_keyed_refs(fs_info, path, bytenr, |
8da6d581 JS |
890 | info_level, &prefs); |
891 | if (ret) | |
892 | goto out; | |
893 | } | |
894 | } | |
895 | btrfs_release_path(path); | |
896 | ||
8da6d581 JS |
897 | list_splice_init(&prefs_delayed, &prefs); |
898 | ||
d5c88b73 JS |
899 | ret = __add_missing_keys(fs_info, &prefs); |
900 | if (ret) | |
901 | goto out; | |
902 | ||
692206b1 | 903 | __merge_refs(&prefs, 1); |
8da6d581 | 904 | |
da61d31a JB |
905 | ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs, |
906 | extent_item_pos); | |
8da6d581 JS |
907 | if (ret) |
908 | goto out; | |
909 | ||
692206b1 | 910 | __merge_refs(&prefs, 2); |
8da6d581 JS |
911 | |
912 | while (!list_empty(&prefs)) { | |
913 | ref = list_first_entry(&prefs, struct __prelim_ref, list); | |
6c1500f2 | 914 | WARN_ON(ref->count < 0); |
8da6d581 JS |
915 | if (ref->count && ref->root_id && ref->parent == 0) { |
916 | /* no parent == root of tree */ | |
917 | ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS); | |
f1723939 WS |
918 | if (ret < 0) |
919 | goto out; | |
8da6d581 JS |
920 | } |
921 | if (ref->count && ref->parent) { | |
976b1908 | 922 | struct extent_inode_elem *eie = NULL; |
3301958b | 923 | if (extent_item_pos && !ref->inode_list) { |
976b1908 JS |
924 | u32 bsz; |
925 | struct extent_buffer *eb; | |
926 | bsz = btrfs_level_size(fs_info->extent_root, | |
927 | info_level); | |
928 | eb = read_tree_block(fs_info->extent_root, | |
929 | ref->parent, bsz, 0); | |
416bc658 JB |
930 | if (!eb || !extent_buffer_uptodate(eb)) { |
931 | free_extent_buffer(eb); | |
c16c2e2e WS |
932 | ret = -EIO; |
933 | goto out; | |
416bc658 | 934 | } |
976b1908 JS |
935 | ret = find_extent_in_eb(eb, bytenr, |
936 | *extent_item_pos, &eie); | |
937 | free_extent_buffer(eb); | |
f5929cd8 FDBM |
938 | if (ret < 0) |
939 | goto out; | |
940 | ref->inode_list = eie; | |
976b1908 | 941 | } |
3301958b | 942 | ret = ulist_add_merge(refs, ref->parent, |
995e01b7 | 943 | (uintptr_t)ref->inode_list, |
34d73f54 | 944 | (u64 *)&eie, GFP_NOFS); |
f1723939 WS |
945 | if (ret < 0) |
946 | goto out; | |
3301958b JS |
947 | if (!ret && extent_item_pos) { |
948 | /* | |
949 | * we've recorded that parent, so we must extend | |
950 | * its inode list here | |
951 | */ | |
952 | BUG_ON(!eie); | |
953 | while (eie->next) | |
954 | eie = eie->next; | |
955 | eie->next = ref->inode_list; | |
956 | } | |
8da6d581 | 957 | } |
a4fdb61e | 958 | list_del(&ref->list); |
8da6d581 JS |
959 | kfree(ref); |
960 | } | |
961 | ||
962 | out: | |
8da6d581 JS |
963 | btrfs_free_path(path); |
964 | while (!list_empty(&prefs)) { | |
965 | ref = list_first_entry(&prefs, struct __prelim_ref, list); | |
966 | list_del(&ref->list); | |
967 | kfree(ref); | |
968 | } | |
969 | while (!list_empty(&prefs_delayed)) { | |
970 | ref = list_first_entry(&prefs_delayed, struct __prelim_ref, | |
971 | list); | |
972 | list_del(&ref->list); | |
973 | kfree(ref); | |
974 | } | |
975 | ||
976 | return ret; | |
977 | } | |
978 | ||
976b1908 JS |
979 | static void free_leaf_list(struct ulist *blocks) |
980 | { | |
981 | struct ulist_node *node = NULL; | |
982 | struct extent_inode_elem *eie; | |
983 | struct extent_inode_elem *eie_next; | |
984 | struct ulist_iterator uiter; | |
985 | ||
986 | ULIST_ITER_INIT(&uiter); | |
987 | while ((node = ulist_next(blocks, &uiter))) { | |
988 | if (!node->aux) | |
989 | continue; | |
995e01b7 | 990 | eie = (struct extent_inode_elem *)(uintptr_t)node->aux; |
976b1908 JS |
991 | for (; eie; eie = eie_next) { |
992 | eie_next = eie->next; | |
993 | kfree(eie); | |
994 | } | |
995 | node->aux = 0; | |
996 | } | |
997 | ||
998 | ulist_free(blocks); | |
999 | } | |
1000 | ||
8da6d581 JS |
1001 | /* |
1002 | * Finds all leafs with a reference to the specified combination of bytenr and | |
1003 | * offset. key_list_head will point to a list of corresponding keys (caller must | |
1004 | * free each list element). The leafs will be stored in the leafs ulist, which | |
1005 | * must be freed with ulist_free. | |
1006 | * | |
1007 | * returns 0 on success, <0 on error | |
1008 | */ | |
1009 | static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans, | |
1010 | struct btrfs_fs_info *fs_info, u64 bytenr, | |
097b8a7c | 1011 | u64 time_seq, struct ulist **leafs, |
976b1908 | 1012 | const u64 *extent_item_pos) |
8da6d581 JS |
1013 | { |
1014 | struct ulist *tmp; | |
1015 | int ret; | |
1016 | ||
1017 | tmp = ulist_alloc(GFP_NOFS); | |
1018 | if (!tmp) | |
1019 | return -ENOMEM; | |
1020 | *leafs = ulist_alloc(GFP_NOFS); | |
1021 | if (!*leafs) { | |
1022 | ulist_free(tmp); | |
1023 | return -ENOMEM; | |
1024 | } | |
1025 | ||
097b8a7c | 1026 | ret = find_parent_nodes(trans, fs_info, bytenr, |
8445f61c | 1027 | time_seq, *leafs, tmp, extent_item_pos); |
8da6d581 JS |
1028 | ulist_free(tmp); |
1029 | ||
1030 | if (ret < 0 && ret != -ENOENT) { | |
976b1908 | 1031 | free_leaf_list(*leafs); |
8da6d581 JS |
1032 | return ret; |
1033 | } | |
1034 | ||
1035 | return 0; | |
1036 | } | |
1037 | ||
1038 | /* | |
1039 | * walk all backrefs for a given extent to find all roots that reference this | |
1040 | * extent. Walking a backref means finding all extents that reference this | |
1041 | * extent and in turn walk the backrefs of those, too. Naturally this is a | |
1042 | * recursive process, but here it is implemented in an iterative fashion: We | |
1043 | * find all referencing extents for the extent in question and put them on a | |
1044 | * list. In turn, we find all referencing extents for those, further appending | |
1045 | * to the list. The way we iterate the list allows adding more elements after | |
1046 | * the current while iterating. The process stops when we reach the end of the | |
1047 | * list. Found roots are added to the roots list. | |
1048 | * | |
1049 | * returns 0 on success, < 0 on error. | |
1050 | */ | |
1051 | int btrfs_find_all_roots(struct btrfs_trans_handle *trans, | |
1052 | struct btrfs_fs_info *fs_info, u64 bytenr, | |
097b8a7c | 1053 | u64 time_seq, struct ulist **roots) |
8da6d581 JS |
1054 | { |
1055 | struct ulist *tmp; | |
1056 | struct ulist_node *node = NULL; | |
cd1b413c | 1057 | struct ulist_iterator uiter; |
8da6d581 JS |
1058 | int ret; |
1059 | ||
1060 | tmp = ulist_alloc(GFP_NOFS); | |
1061 | if (!tmp) | |
1062 | return -ENOMEM; | |
1063 | *roots = ulist_alloc(GFP_NOFS); | |
1064 | if (!*roots) { | |
1065 | ulist_free(tmp); | |
1066 | return -ENOMEM; | |
1067 | } | |
1068 | ||
cd1b413c | 1069 | ULIST_ITER_INIT(&uiter); |
8da6d581 | 1070 | while (1) { |
097b8a7c | 1071 | ret = find_parent_nodes(trans, fs_info, bytenr, |
8445f61c | 1072 | time_seq, tmp, *roots, NULL); |
8da6d581 JS |
1073 | if (ret < 0 && ret != -ENOENT) { |
1074 | ulist_free(tmp); | |
1075 | ulist_free(*roots); | |
1076 | return ret; | |
1077 | } | |
cd1b413c | 1078 | node = ulist_next(tmp, &uiter); |
8da6d581 JS |
1079 | if (!node) |
1080 | break; | |
1081 | bytenr = node->val; | |
1082 | } | |
1083 | ||
1084 | ulist_free(tmp); | |
1085 | return 0; | |
1086 | } | |
1087 | ||
1088 | ||
a542ad1b JS |
1089 | static int __inode_info(u64 inum, u64 ioff, u8 key_type, |
1090 | struct btrfs_root *fs_root, struct btrfs_path *path, | |
1091 | struct btrfs_key *found_key) | |
1092 | { | |
1093 | int ret; | |
1094 | struct btrfs_key key; | |
1095 | struct extent_buffer *eb; | |
1096 | ||
1097 | key.type = key_type; | |
1098 | key.objectid = inum; | |
1099 | key.offset = ioff; | |
1100 | ||
1101 | ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0); | |
1102 | if (ret < 0) | |
1103 | return ret; | |
1104 | ||
1105 | eb = path->nodes[0]; | |
1106 | if (ret && path->slots[0] >= btrfs_header_nritems(eb)) { | |
1107 | ret = btrfs_next_leaf(fs_root, path); | |
1108 | if (ret) | |
1109 | return ret; | |
1110 | eb = path->nodes[0]; | |
1111 | } | |
1112 | ||
1113 | btrfs_item_key_to_cpu(eb, found_key, path->slots[0]); | |
1114 | if (found_key->type != key.type || found_key->objectid != key.objectid) | |
1115 | return 1; | |
1116 | ||
1117 | return 0; | |
1118 | } | |
1119 | ||
1120 | /* | |
1121 | * this makes the path point to (inum INODE_ITEM ioff) | |
1122 | */ | |
1123 | int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root, | |
1124 | struct btrfs_path *path) | |
1125 | { | |
1126 | struct btrfs_key key; | |
1127 | return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path, | |
1128 | &key); | |
1129 | } | |
1130 | ||
1131 | static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root, | |
1132 | struct btrfs_path *path, | |
1133 | struct btrfs_key *found_key) | |
1134 | { | |
1135 | return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path, | |
1136 | found_key); | |
1137 | } | |
1138 | ||
f186373f MF |
1139 | int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid, |
1140 | u64 start_off, struct btrfs_path *path, | |
1141 | struct btrfs_inode_extref **ret_extref, | |
1142 | u64 *found_off) | |
1143 | { | |
1144 | int ret, slot; | |
1145 | struct btrfs_key key; | |
1146 | struct btrfs_key found_key; | |
1147 | struct btrfs_inode_extref *extref; | |
1148 | struct extent_buffer *leaf; | |
1149 | unsigned long ptr; | |
1150 | ||
1151 | key.objectid = inode_objectid; | |
1152 | btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY); | |
1153 | key.offset = start_off; | |
1154 | ||
1155 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1156 | if (ret < 0) | |
1157 | return ret; | |
1158 | ||
1159 | while (1) { | |
1160 | leaf = path->nodes[0]; | |
1161 | slot = path->slots[0]; | |
1162 | if (slot >= btrfs_header_nritems(leaf)) { | |
1163 | /* | |
1164 | * If the item at offset is not found, | |
1165 | * btrfs_search_slot will point us to the slot | |
1166 | * where it should be inserted. In our case | |
1167 | * that will be the slot directly before the | |
1168 | * next INODE_REF_KEY_V2 item. In the case | |
1169 | * that we're pointing to the last slot in a | |
1170 | * leaf, we must move one leaf over. | |
1171 | */ | |
1172 | ret = btrfs_next_leaf(root, path); | |
1173 | if (ret) { | |
1174 | if (ret >= 1) | |
1175 | ret = -ENOENT; | |
1176 | break; | |
1177 | } | |
1178 | continue; | |
1179 | } | |
1180 | ||
1181 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1182 | ||
1183 | /* | |
1184 | * Check that we're still looking at an extended ref key for | |
1185 | * this particular objectid. If we have different | |
1186 | * objectid or type then there are no more to be found | |
1187 | * in the tree and we can exit. | |
1188 | */ | |
1189 | ret = -ENOENT; | |
1190 | if (found_key.objectid != inode_objectid) | |
1191 | break; | |
1192 | if (btrfs_key_type(&found_key) != BTRFS_INODE_EXTREF_KEY) | |
1193 | break; | |
1194 | ||
1195 | ret = 0; | |
1196 | ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); | |
1197 | extref = (struct btrfs_inode_extref *)ptr; | |
1198 | *ret_extref = extref; | |
1199 | if (found_off) | |
1200 | *found_off = found_key.offset; | |
1201 | break; | |
1202 | } | |
1203 | ||
1204 | return ret; | |
1205 | } | |
1206 | ||
48a3b636 ES |
1207 | /* |
1208 | * this iterates to turn a name (from iref/extref) into a full filesystem path. | |
1209 | * Elements of the path are separated by '/' and the path is guaranteed to be | |
1210 | * 0-terminated. the path is only given within the current file system. | |
1211 | * Therefore, it never starts with a '/'. the caller is responsible to provide | |
1212 | * "size" bytes in "dest". the dest buffer will be filled backwards. finally, | |
1213 | * the start point of the resulting string is returned. this pointer is within | |
1214 | * dest, normally. | |
1215 | * in case the path buffer would overflow, the pointer is decremented further | |
1216 | * as if output was written to the buffer, though no more output is actually | |
1217 | * generated. that way, the caller can determine how much space would be | |
1218 | * required for the path to fit into the buffer. in that case, the returned | |
1219 | * value will be smaller than dest. callers must check this! | |
1220 | */ | |
96b5bd77 JS |
1221 | char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path, |
1222 | u32 name_len, unsigned long name_off, | |
1223 | struct extent_buffer *eb_in, u64 parent, | |
1224 | char *dest, u32 size) | |
a542ad1b | 1225 | { |
a542ad1b JS |
1226 | int slot; |
1227 | u64 next_inum; | |
1228 | int ret; | |
661bec6b | 1229 | s64 bytes_left = ((s64)size) - 1; |
a542ad1b JS |
1230 | struct extent_buffer *eb = eb_in; |
1231 | struct btrfs_key found_key; | |
b916a59a | 1232 | int leave_spinning = path->leave_spinning; |
d24bec3a | 1233 | struct btrfs_inode_ref *iref; |
a542ad1b JS |
1234 | |
1235 | if (bytes_left >= 0) | |
1236 | dest[bytes_left] = '\0'; | |
1237 | ||
b916a59a | 1238 | path->leave_spinning = 1; |
a542ad1b | 1239 | while (1) { |
d24bec3a | 1240 | bytes_left -= name_len; |
a542ad1b JS |
1241 | if (bytes_left >= 0) |
1242 | read_extent_buffer(eb, dest + bytes_left, | |
d24bec3a | 1243 | name_off, name_len); |
b916a59a JS |
1244 | if (eb != eb_in) { |
1245 | btrfs_tree_read_unlock_blocking(eb); | |
a542ad1b | 1246 | free_extent_buffer(eb); |
b916a59a | 1247 | } |
a542ad1b | 1248 | ret = inode_ref_info(parent, 0, fs_root, path, &found_key); |
8f24b496 JS |
1249 | if (ret > 0) |
1250 | ret = -ENOENT; | |
a542ad1b JS |
1251 | if (ret) |
1252 | break; | |
d24bec3a | 1253 | |
a542ad1b JS |
1254 | next_inum = found_key.offset; |
1255 | ||
1256 | /* regular exit ahead */ | |
1257 | if (parent == next_inum) | |
1258 | break; | |
1259 | ||
1260 | slot = path->slots[0]; | |
1261 | eb = path->nodes[0]; | |
1262 | /* make sure we can use eb after releasing the path */ | |
b916a59a | 1263 | if (eb != eb_in) { |
a542ad1b | 1264 | atomic_inc(&eb->refs); |
b916a59a JS |
1265 | btrfs_tree_read_lock(eb); |
1266 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
1267 | } | |
a542ad1b | 1268 | btrfs_release_path(path); |
a542ad1b | 1269 | iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref); |
d24bec3a MF |
1270 | |
1271 | name_len = btrfs_inode_ref_name_len(eb, iref); | |
1272 | name_off = (unsigned long)(iref + 1); | |
1273 | ||
a542ad1b JS |
1274 | parent = next_inum; |
1275 | --bytes_left; | |
1276 | if (bytes_left >= 0) | |
1277 | dest[bytes_left] = '/'; | |
1278 | } | |
1279 | ||
1280 | btrfs_release_path(path); | |
b916a59a | 1281 | path->leave_spinning = leave_spinning; |
a542ad1b JS |
1282 | |
1283 | if (ret) | |
1284 | return ERR_PTR(ret); | |
1285 | ||
1286 | return dest + bytes_left; | |
1287 | } | |
1288 | ||
1289 | /* | |
1290 | * this makes the path point to (logical EXTENT_ITEM *) | |
1291 | * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for | |
1292 | * tree blocks and <0 on error. | |
1293 | */ | |
1294 | int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical, | |
69917e43 LB |
1295 | struct btrfs_path *path, struct btrfs_key *found_key, |
1296 | u64 *flags_ret) | |
a542ad1b JS |
1297 | { |
1298 | int ret; | |
1299 | u64 flags; | |
261c84b6 | 1300 | u64 size = 0; |
a542ad1b JS |
1301 | u32 item_size; |
1302 | struct extent_buffer *eb; | |
1303 | struct btrfs_extent_item *ei; | |
1304 | struct btrfs_key key; | |
1305 | ||
261c84b6 JB |
1306 | if (btrfs_fs_incompat(fs_info, SKINNY_METADATA)) |
1307 | key.type = BTRFS_METADATA_ITEM_KEY; | |
1308 | else | |
1309 | key.type = BTRFS_EXTENT_ITEM_KEY; | |
a542ad1b JS |
1310 | key.objectid = logical; |
1311 | key.offset = (u64)-1; | |
1312 | ||
1313 | ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0); | |
1314 | if (ret < 0) | |
1315 | return ret; | |
1316 | ret = btrfs_previous_item(fs_info->extent_root, path, | |
1317 | 0, BTRFS_EXTENT_ITEM_KEY); | |
1318 | if (ret < 0) | |
1319 | return ret; | |
1320 | ||
1321 | btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]); | |
261c84b6 JB |
1322 | if (found_key->type == BTRFS_METADATA_ITEM_KEY) |
1323 | size = fs_info->extent_root->leafsize; | |
1324 | else if (found_key->type == BTRFS_EXTENT_ITEM_KEY) | |
1325 | size = found_key->offset; | |
1326 | ||
1327 | if ((found_key->type != BTRFS_EXTENT_ITEM_KEY && | |
1328 | found_key->type != BTRFS_METADATA_ITEM_KEY) || | |
a542ad1b | 1329 | found_key->objectid > logical || |
261c84b6 | 1330 | found_key->objectid + size <= logical) { |
4692cf58 JS |
1331 | pr_debug("logical %llu is not within any extent\n", |
1332 | (unsigned long long)logical); | |
a542ad1b | 1333 | return -ENOENT; |
4692cf58 | 1334 | } |
a542ad1b JS |
1335 | |
1336 | eb = path->nodes[0]; | |
1337 | item_size = btrfs_item_size_nr(eb, path->slots[0]); | |
1338 | BUG_ON(item_size < sizeof(*ei)); | |
1339 | ||
1340 | ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item); | |
1341 | flags = btrfs_extent_flags(eb, ei); | |
1342 | ||
4692cf58 JS |
1343 | pr_debug("logical %llu is at position %llu within the extent (%llu " |
1344 | "EXTENT_ITEM %llu) flags %#llx size %u\n", | |
1345 | (unsigned long long)logical, | |
1346 | (unsigned long long)(logical - found_key->objectid), | |
1347 | (unsigned long long)found_key->objectid, | |
1348 | (unsigned long long)found_key->offset, | |
1349 | (unsigned long long)flags, item_size); | |
69917e43 LB |
1350 | |
1351 | WARN_ON(!flags_ret); | |
1352 | if (flags_ret) { | |
1353 | if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) | |
1354 | *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK; | |
1355 | else if (flags & BTRFS_EXTENT_FLAG_DATA) | |
1356 | *flags_ret = BTRFS_EXTENT_FLAG_DATA; | |
1357 | else | |
1358 | BUG_ON(1); | |
1359 | return 0; | |
1360 | } | |
a542ad1b JS |
1361 | |
1362 | return -EIO; | |
1363 | } | |
1364 | ||
1365 | /* | |
1366 | * helper function to iterate extent inline refs. ptr must point to a 0 value | |
1367 | * for the first call and may be modified. it is used to track state. | |
1368 | * if more refs exist, 0 is returned and the next call to | |
1369 | * __get_extent_inline_ref must pass the modified ptr parameter to get the | |
1370 | * next ref. after the last ref was processed, 1 is returned. | |
1371 | * returns <0 on error | |
1372 | */ | |
1373 | static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb, | |
1374 | struct btrfs_extent_item *ei, u32 item_size, | |
1375 | struct btrfs_extent_inline_ref **out_eiref, | |
1376 | int *out_type) | |
1377 | { | |
1378 | unsigned long end; | |
1379 | u64 flags; | |
1380 | struct btrfs_tree_block_info *info; | |
1381 | ||
1382 | if (!*ptr) { | |
1383 | /* first call */ | |
1384 | flags = btrfs_extent_flags(eb, ei); | |
1385 | if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) { | |
1386 | info = (struct btrfs_tree_block_info *)(ei + 1); | |
1387 | *out_eiref = | |
1388 | (struct btrfs_extent_inline_ref *)(info + 1); | |
1389 | } else { | |
1390 | *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1); | |
1391 | } | |
1392 | *ptr = (unsigned long)*out_eiref; | |
1393 | if ((void *)*ptr >= (void *)ei + item_size) | |
1394 | return -ENOENT; | |
1395 | } | |
1396 | ||
1397 | end = (unsigned long)ei + item_size; | |
1398 | *out_eiref = (struct btrfs_extent_inline_ref *)*ptr; | |
1399 | *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref); | |
1400 | ||
1401 | *ptr += btrfs_extent_inline_ref_size(*out_type); | |
1402 | WARN_ON(*ptr > end); | |
1403 | if (*ptr == end) | |
1404 | return 1; /* last */ | |
1405 | ||
1406 | return 0; | |
1407 | } | |
1408 | ||
1409 | /* | |
1410 | * reads the tree block backref for an extent. tree level and root are returned | |
1411 | * through out_level and out_root. ptr must point to a 0 value for the first | |
1412 | * call and may be modified (see __get_extent_inline_ref comment). | |
1413 | * returns 0 if data was provided, 1 if there was no more data to provide or | |
1414 | * <0 on error. | |
1415 | */ | |
1416 | int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb, | |
1417 | struct btrfs_extent_item *ei, u32 item_size, | |
1418 | u64 *out_root, u8 *out_level) | |
1419 | { | |
1420 | int ret; | |
1421 | int type; | |
1422 | struct btrfs_tree_block_info *info; | |
1423 | struct btrfs_extent_inline_ref *eiref; | |
1424 | ||
1425 | if (*ptr == (unsigned long)-1) | |
1426 | return 1; | |
1427 | ||
1428 | while (1) { | |
1429 | ret = __get_extent_inline_ref(ptr, eb, ei, item_size, | |
1430 | &eiref, &type); | |
1431 | if (ret < 0) | |
1432 | return ret; | |
1433 | ||
1434 | if (type == BTRFS_TREE_BLOCK_REF_KEY || | |
1435 | type == BTRFS_SHARED_BLOCK_REF_KEY) | |
1436 | break; | |
1437 | ||
1438 | if (ret == 1) | |
1439 | return 1; | |
1440 | } | |
1441 | ||
1442 | /* we can treat both ref types equally here */ | |
1443 | info = (struct btrfs_tree_block_info *)(ei + 1); | |
1444 | *out_root = btrfs_extent_inline_ref_offset(eb, eiref); | |
1445 | *out_level = btrfs_tree_block_level(eb, info); | |
1446 | ||
1447 | if (ret == 1) | |
1448 | *ptr = (unsigned long)-1; | |
1449 | ||
1450 | return 0; | |
1451 | } | |
1452 | ||
976b1908 JS |
1453 | static int iterate_leaf_refs(struct extent_inode_elem *inode_list, |
1454 | u64 root, u64 extent_item_objectid, | |
4692cf58 | 1455 | iterate_extent_inodes_t *iterate, void *ctx) |
a542ad1b | 1456 | { |
976b1908 | 1457 | struct extent_inode_elem *eie; |
4692cf58 | 1458 | int ret = 0; |
4692cf58 | 1459 | |
976b1908 | 1460 | for (eie = inode_list; eie; eie = eie->next) { |
4692cf58 | 1461 | pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), " |
976b1908 JS |
1462 | "root %llu\n", extent_item_objectid, |
1463 | eie->inum, eie->offset, root); | |
1464 | ret = iterate(eie->inum, eie->offset, root, ctx); | |
4692cf58 | 1465 | if (ret) { |
976b1908 JS |
1466 | pr_debug("stopping iteration for %llu due to ret=%d\n", |
1467 | extent_item_objectid, ret); | |
4692cf58 JS |
1468 | break; |
1469 | } | |
a542ad1b JS |
1470 | } |
1471 | ||
a542ad1b JS |
1472 | return ret; |
1473 | } | |
1474 | ||
1475 | /* | |
1476 | * calls iterate() for every inode that references the extent identified by | |
4692cf58 | 1477 | * the given parameters. |
a542ad1b JS |
1478 | * when the iterator function returns a non-zero value, iteration stops. |
1479 | */ | |
1480 | int iterate_extent_inodes(struct btrfs_fs_info *fs_info, | |
4692cf58 | 1481 | u64 extent_item_objectid, u64 extent_item_pos, |
7a3ae2f8 | 1482 | int search_commit_root, |
a542ad1b JS |
1483 | iterate_extent_inodes_t *iterate, void *ctx) |
1484 | { | |
a542ad1b | 1485 | int ret; |
da61d31a | 1486 | struct btrfs_trans_handle *trans = NULL; |
7a3ae2f8 JS |
1487 | struct ulist *refs = NULL; |
1488 | struct ulist *roots = NULL; | |
4692cf58 JS |
1489 | struct ulist_node *ref_node = NULL; |
1490 | struct ulist_node *root_node = NULL; | |
8445f61c | 1491 | struct seq_list tree_mod_seq_elem = {}; |
cd1b413c JS |
1492 | struct ulist_iterator ref_uiter; |
1493 | struct ulist_iterator root_uiter; | |
a542ad1b | 1494 | |
4692cf58 JS |
1495 | pr_debug("resolving all inodes for extent %llu\n", |
1496 | extent_item_objectid); | |
a542ad1b | 1497 | |
da61d31a | 1498 | if (!search_commit_root) { |
7a3ae2f8 JS |
1499 | trans = btrfs_join_transaction(fs_info->extent_root); |
1500 | if (IS_ERR(trans)) | |
1501 | return PTR_ERR(trans); | |
8445f61c | 1502 | btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem); |
7a3ae2f8 | 1503 | } |
a542ad1b | 1504 | |
4692cf58 | 1505 | ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid, |
097b8a7c | 1506 | tree_mod_seq_elem.seq, &refs, |
8445f61c | 1507 | &extent_item_pos); |
4692cf58 JS |
1508 | if (ret) |
1509 | goto out; | |
a542ad1b | 1510 | |
cd1b413c JS |
1511 | ULIST_ITER_INIT(&ref_uiter); |
1512 | while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) { | |
976b1908 | 1513 | ret = btrfs_find_all_roots(trans, fs_info, ref_node->val, |
097b8a7c | 1514 | tree_mod_seq_elem.seq, &roots); |
4692cf58 JS |
1515 | if (ret) |
1516 | break; | |
cd1b413c JS |
1517 | ULIST_ITER_INIT(&root_uiter); |
1518 | while (!ret && (root_node = ulist_next(roots, &root_uiter))) { | |
976b1908 | 1519 | pr_debug("root %llu references leaf %llu, data list " |
34d73f54 | 1520 | "%#llx\n", root_node->val, ref_node->val, |
995e01b7 JS |
1521 | (long long)ref_node->aux); |
1522 | ret = iterate_leaf_refs((struct extent_inode_elem *) | |
1523 | (uintptr_t)ref_node->aux, | |
1524 | root_node->val, | |
1525 | extent_item_objectid, | |
1526 | iterate, ctx); | |
4692cf58 | 1527 | } |
976b1908 | 1528 | ulist_free(roots); |
a542ad1b JS |
1529 | } |
1530 | ||
976b1908 | 1531 | free_leaf_list(refs); |
4692cf58 | 1532 | out: |
7a3ae2f8 | 1533 | if (!search_commit_root) { |
8445f61c | 1534 | btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem); |
7a3ae2f8 JS |
1535 | btrfs_end_transaction(trans, fs_info->extent_root); |
1536 | } | |
1537 | ||
a542ad1b JS |
1538 | return ret; |
1539 | } | |
1540 | ||
1541 | int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info, | |
1542 | struct btrfs_path *path, | |
1543 | iterate_extent_inodes_t *iterate, void *ctx) | |
1544 | { | |
1545 | int ret; | |
4692cf58 | 1546 | u64 extent_item_pos; |
69917e43 | 1547 | u64 flags = 0; |
a542ad1b | 1548 | struct btrfs_key found_key; |
7a3ae2f8 | 1549 | int search_commit_root = path->search_commit_root; |
a542ad1b | 1550 | |
69917e43 | 1551 | ret = extent_from_logical(fs_info, logical, path, &found_key, &flags); |
4692cf58 | 1552 | btrfs_release_path(path); |
a542ad1b JS |
1553 | if (ret < 0) |
1554 | return ret; | |
69917e43 | 1555 | if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) |
3627bf45 | 1556 | return -EINVAL; |
a542ad1b | 1557 | |
4692cf58 | 1558 | extent_item_pos = logical - found_key.objectid; |
7a3ae2f8 JS |
1559 | ret = iterate_extent_inodes(fs_info, found_key.objectid, |
1560 | extent_item_pos, search_commit_root, | |
1561 | iterate, ctx); | |
a542ad1b JS |
1562 | |
1563 | return ret; | |
1564 | } | |
1565 | ||
d24bec3a MF |
1566 | typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off, |
1567 | struct extent_buffer *eb, void *ctx); | |
1568 | ||
1569 | static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root, | |
1570 | struct btrfs_path *path, | |
1571 | iterate_irefs_t *iterate, void *ctx) | |
a542ad1b | 1572 | { |
aefc1eb1 | 1573 | int ret = 0; |
a542ad1b JS |
1574 | int slot; |
1575 | u32 cur; | |
1576 | u32 len; | |
1577 | u32 name_len; | |
1578 | u64 parent = 0; | |
1579 | int found = 0; | |
1580 | struct extent_buffer *eb; | |
1581 | struct btrfs_item *item; | |
1582 | struct btrfs_inode_ref *iref; | |
1583 | struct btrfs_key found_key; | |
1584 | ||
aefc1eb1 | 1585 | while (!ret) { |
b916a59a | 1586 | path->leave_spinning = 1; |
a542ad1b | 1587 | ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path, |
d24bec3a | 1588 | &found_key); |
a542ad1b JS |
1589 | if (ret < 0) |
1590 | break; | |
1591 | if (ret) { | |
1592 | ret = found ? 0 : -ENOENT; | |
1593 | break; | |
1594 | } | |
1595 | ++found; | |
1596 | ||
1597 | parent = found_key.offset; | |
1598 | slot = path->slots[0]; | |
1599 | eb = path->nodes[0]; | |
1600 | /* make sure we can use eb after releasing the path */ | |
1601 | atomic_inc(&eb->refs); | |
b916a59a JS |
1602 | btrfs_tree_read_lock(eb); |
1603 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
a542ad1b JS |
1604 | btrfs_release_path(path); |
1605 | ||
1606 | item = btrfs_item_nr(eb, slot); | |
1607 | iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref); | |
1608 | ||
1609 | for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) { | |
1610 | name_len = btrfs_inode_ref_name_len(eb, iref); | |
1611 | /* path must be released before calling iterate()! */ | |
4692cf58 JS |
1612 | pr_debug("following ref at offset %u for inode %llu in " |
1613 | "tree %llu\n", cur, | |
1614 | (unsigned long long)found_key.objectid, | |
1615 | (unsigned long long)fs_root->objectid); | |
d24bec3a MF |
1616 | ret = iterate(parent, name_len, |
1617 | (unsigned long)(iref + 1), eb, ctx); | |
aefc1eb1 | 1618 | if (ret) |
a542ad1b | 1619 | break; |
a542ad1b JS |
1620 | len = sizeof(*iref) + name_len; |
1621 | iref = (struct btrfs_inode_ref *)((char *)iref + len); | |
1622 | } | |
b916a59a | 1623 | btrfs_tree_read_unlock_blocking(eb); |
a542ad1b JS |
1624 | free_extent_buffer(eb); |
1625 | } | |
1626 | ||
1627 | btrfs_release_path(path); | |
1628 | ||
1629 | return ret; | |
1630 | } | |
1631 | ||
d24bec3a MF |
1632 | static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root, |
1633 | struct btrfs_path *path, | |
1634 | iterate_irefs_t *iterate, void *ctx) | |
1635 | { | |
1636 | int ret; | |
1637 | int slot; | |
1638 | u64 offset = 0; | |
1639 | u64 parent; | |
1640 | int found = 0; | |
1641 | struct extent_buffer *eb; | |
1642 | struct btrfs_inode_extref *extref; | |
1643 | struct extent_buffer *leaf; | |
1644 | u32 item_size; | |
1645 | u32 cur_offset; | |
1646 | unsigned long ptr; | |
1647 | ||
1648 | while (1) { | |
1649 | ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref, | |
1650 | &offset); | |
1651 | if (ret < 0) | |
1652 | break; | |
1653 | if (ret) { | |
1654 | ret = found ? 0 : -ENOENT; | |
1655 | break; | |
1656 | } | |
1657 | ++found; | |
1658 | ||
1659 | slot = path->slots[0]; | |
1660 | eb = path->nodes[0]; | |
1661 | /* make sure we can use eb after releasing the path */ | |
1662 | atomic_inc(&eb->refs); | |
1663 | ||
1664 | btrfs_tree_read_lock(eb); | |
1665 | btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); | |
1666 | btrfs_release_path(path); | |
1667 | ||
1668 | leaf = path->nodes[0]; | |
1669 | item_size = btrfs_item_size_nr(leaf, path->slots[0]); | |
1670 | ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); | |
1671 | cur_offset = 0; | |
1672 | ||
1673 | while (cur_offset < item_size) { | |
1674 | u32 name_len; | |
1675 | ||
1676 | extref = (struct btrfs_inode_extref *)(ptr + cur_offset); | |
1677 | parent = btrfs_inode_extref_parent(eb, extref); | |
1678 | name_len = btrfs_inode_extref_name_len(eb, extref); | |
1679 | ret = iterate(parent, name_len, | |
1680 | (unsigned long)&extref->name, eb, ctx); | |
1681 | if (ret) | |
1682 | break; | |
1683 | ||
1684 | cur_offset += btrfs_inode_extref_name_len(leaf, extref); | |
1685 | cur_offset += sizeof(*extref); | |
1686 | } | |
1687 | btrfs_tree_read_unlock_blocking(eb); | |
1688 | free_extent_buffer(eb); | |
1689 | ||
1690 | offset++; | |
1691 | } | |
1692 | ||
1693 | btrfs_release_path(path); | |
1694 | ||
1695 | return ret; | |
1696 | } | |
1697 | ||
1698 | static int iterate_irefs(u64 inum, struct btrfs_root *fs_root, | |
1699 | struct btrfs_path *path, iterate_irefs_t *iterate, | |
1700 | void *ctx) | |
1701 | { | |
1702 | int ret; | |
1703 | int found_refs = 0; | |
1704 | ||
1705 | ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx); | |
1706 | if (!ret) | |
1707 | ++found_refs; | |
1708 | else if (ret != -ENOENT) | |
1709 | return ret; | |
1710 | ||
1711 | ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx); | |
1712 | if (ret == -ENOENT && found_refs) | |
1713 | return 0; | |
1714 | ||
1715 | return ret; | |
1716 | } | |
1717 | ||
a542ad1b JS |
1718 | /* |
1719 | * returns 0 if the path could be dumped (probably truncated) | |
1720 | * returns <0 in case of an error | |
1721 | */ | |
d24bec3a MF |
1722 | static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off, |
1723 | struct extent_buffer *eb, void *ctx) | |
a542ad1b JS |
1724 | { |
1725 | struct inode_fs_paths *ipath = ctx; | |
1726 | char *fspath; | |
1727 | char *fspath_min; | |
1728 | int i = ipath->fspath->elem_cnt; | |
1729 | const int s_ptr = sizeof(char *); | |
1730 | u32 bytes_left; | |
1731 | ||
1732 | bytes_left = ipath->fspath->bytes_left > s_ptr ? | |
1733 | ipath->fspath->bytes_left - s_ptr : 0; | |
1734 | ||
740c3d22 | 1735 | fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr; |
96b5bd77 JS |
1736 | fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len, |
1737 | name_off, eb, inum, fspath_min, bytes_left); | |
a542ad1b JS |
1738 | if (IS_ERR(fspath)) |
1739 | return PTR_ERR(fspath); | |
1740 | ||
1741 | if (fspath > fspath_min) { | |
745c4d8e | 1742 | ipath->fspath->val[i] = (u64)(unsigned long)fspath; |
a542ad1b JS |
1743 | ++ipath->fspath->elem_cnt; |
1744 | ipath->fspath->bytes_left = fspath - fspath_min; | |
1745 | } else { | |
1746 | ++ipath->fspath->elem_missed; | |
1747 | ipath->fspath->bytes_missing += fspath_min - fspath; | |
1748 | ipath->fspath->bytes_left = 0; | |
1749 | } | |
1750 | ||
1751 | return 0; | |
1752 | } | |
1753 | ||
1754 | /* | |
1755 | * this dumps all file system paths to the inode into the ipath struct, provided | |
1756 | * is has been created large enough. each path is zero-terminated and accessed | |
740c3d22 | 1757 | * from ipath->fspath->val[i]. |
a542ad1b | 1758 | * when it returns, there are ipath->fspath->elem_cnt number of paths available |
740c3d22 | 1759 | * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the |
a542ad1b JS |
1760 | * number of missed paths in recored in ipath->fspath->elem_missed, otherwise, |
1761 | * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would | |
1762 | * have been needed to return all paths. | |
1763 | */ | |
1764 | int paths_from_inode(u64 inum, struct inode_fs_paths *ipath) | |
1765 | { | |
1766 | return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path, | |
d24bec3a | 1767 | inode_to_path, ipath); |
a542ad1b JS |
1768 | } |
1769 | ||
a542ad1b JS |
1770 | struct btrfs_data_container *init_data_container(u32 total_bytes) |
1771 | { | |
1772 | struct btrfs_data_container *data; | |
1773 | size_t alloc_bytes; | |
1774 | ||
1775 | alloc_bytes = max_t(size_t, total_bytes, sizeof(*data)); | |
425d17a2 | 1776 | data = vmalloc(alloc_bytes); |
a542ad1b JS |
1777 | if (!data) |
1778 | return ERR_PTR(-ENOMEM); | |
1779 | ||
1780 | if (total_bytes >= sizeof(*data)) { | |
1781 | data->bytes_left = total_bytes - sizeof(*data); | |
1782 | data->bytes_missing = 0; | |
1783 | } else { | |
1784 | data->bytes_missing = sizeof(*data) - total_bytes; | |
1785 | data->bytes_left = 0; | |
1786 | } | |
1787 | ||
1788 | data->elem_cnt = 0; | |
1789 | data->elem_missed = 0; | |
1790 | ||
1791 | return data; | |
1792 | } | |
1793 | ||
1794 | /* | |
1795 | * allocates space to return multiple file system paths for an inode. | |
1796 | * total_bytes to allocate are passed, note that space usable for actual path | |
1797 | * information will be total_bytes - sizeof(struct inode_fs_paths). | |
1798 | * the returned pointer must be freed with free_ipath() in the end. | |
1799 | */ | |
1800 | struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root, | |
1801 | struct btrfs_path *path) | |
1802 | { | |
1803 | struct inode_fs_paths *ifp; | |
1804 | struct btrfs_data_container *fspath; | |
1805 | ||
1806 | fspath = init_data_container(total_bytes); | |
1807 | if (IS_ERR(fspath)) | |
1808 | return (void *)fspath; | |
1809 | ||
1810 | ifp = kmalloc(sizeof(*ifp), GFP_NOFS); | |
1811 | if (!ifp) { | |
1812 | kfree(fspath); | |
1813 | return ERR_PTR(-ENOMEM); | |
1814 | } | |
1815 | ||
1816 | ifp->btrfs_path = path; | |
1817 | ifp->fspath = fspath; | |
1818 | ifp->fs_root = fs_root; | |
1819 | ||
1820 | return ifp; | |
1821 | } | |
1822 | ||
1823 | void free_ipath(struct inode_fs_paths *ipath) | |
1824 | { | |
4735fb28 JJ |
1825 | if (!ipath) |
1826 | return; | |
425d17a2 | 1827 | vfree(ipath->fspath); |
a542ad1b JS |
1828 | kfree(ipath); |
1829 | } |