1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2 | /* |
3 | * Copyright (C) 2011 STRATO. All rights reserved. |
4 | */ |
5 | |
6 | #ifndef BTRFS_BACKREF_H |
7 | #define BTRFS_BACKREF_H |
8 | |
9 | #include <linux/btrfs.h> |
10 | #include "messages.h" |
11 | #include "ulist.h" |
12 | #include "disk-io.h" |
13 | #include "extent_io.h" |
14 | |
15 | /* |
16 | * Used by implementations of iterate_extent_inodes_t (see definition below) to |
17 | * signal that backref iteration can stop immediately and no error happened. |
18 | * The value must be non-negative and must not be 0, 1 (which is a common return |
19 | * value from things like btrfs_search_slot() and used internally in the backref |
20 | * walking code) and different from BACKREF_FOUND_SHARED and |
21 | * BACKREF_FOUND_NOT_SHARED |
22 | */ |
23 | #define BTRFS_ITERATE_EXTENT_INODES_STOP 5 |
24 | |
25 | /* |
26 | * Should return 0 if no errors happened and iteration of backrefs should |
27 | * continue. Can return BTRFS_ITERATE_EXTENT_INODES_STOP or any other non-zero |
28 | * value to immediately stop iteration and possibly signal an error back to |
29 | * the caller. |
30 | */ |
31 | typedef int (iterate_extent_inodes_t)(u64 inum, u64 offset, u64 num_bytes, |
32 | u64 root, void *ctx); |
33 | |
34 | /* |
35 | * Context and arguments for backref walking functions. Some of the fields are |
36 | * to be filled by the caller of such functions while other are filled by the |
37 | * functions themselves, as described below. |
38 | */ |
39 | struct btrfs_backref_walk_ctx { |
40 | /* |
41 | * The address of the extent for which we are doing backref walking. |
42 | * Can be either a data extent or a metadata extent. |
43 | * |
44 | * Must always be set by the top level caller. |
45 | */ |
46 | u64 bytenr; |
47 | /* |
48 | * Offset relative to the target extent. This is only used for data |
49 | * extents, and it's meaningful because we can have file extent items |
50 | * that point only to a section of a data extent ("bookend" extents), |
51 | * and we want to filter out any that don't point to a section of the |
52 | * data extent containing the given offset. |
53 | * |
54 | * Must always be set by the top level caller. |
55 | */ |
56 | u64 extent_item_pos; |
57 | /* |
58 | * If true and bytenr corresponds to a data extent, then references from |
59 | * all file extent items that point to the data extent are considered, |
60 | * @extent_item_pos is ignored. |
61 | */ |
62 | bool ignore_extent_item_pos; |
63 | /* |
64 | * If true and bytenr corresponds to a data extent, then the inode list |
65 | * (each member describing inode number, file offset and root) is not |
66 | * added to each reference added to the @refs ulist. |
67 | */ |
68 | bool skip_inode_ref_list; |
69 | /* A valid transaction handle or NULL. */ |
70 | struct btrfs_trans_handle *trans; |
71 | /* |
72 | * The file system's info object, can not be NULL. |
73 | * |
74 | * Must always be set by the top level caller. |
75 | */ |
76 | struct btrfs_fs_info *fs_info; |
77 | /* |
78 | * Time sequence acquired from btrfs_get_tree_mod_seq(), in case the |
79 | * caller joined the tree mod log to get a consistent view of b+trees |
80 | * while we do backref walking, or BTRFS_SEQ_LAST. |
81 | * When using BTRFS_SEQ_LAST, delayed refs are not checked and it uses |
82 | * commit roots when searching b+trees - this is a special case for |
83 | * qgroups used during a transaction commit. |
84 | */ |
85 | u64 time_seq; |
86 | /* |
87 | * Used to collect the bytenr of metadata extents that point to the |
88 | * target extent. |
89 | */ |
90 | struct ulist *refs; |
91 | /* |
92 | * List used to collect the IDs of the roots from which the target |
93 | * extent is accessible. Can be NULL in case the caller does not care |
94 | * about collecting root IDs. |
95 | */ |
96 | struct ulist *roots; |
97 | /* |
98 | * Used by iterate_extent_inodes() and the main backref walk code |
99 | * (find_parent_nodes()). Lookup and store functions for an optional |
100 | * cache which maps the logical address (bytenr) of leaves to an array |
101 | * of root IDs. |
102 | */ |
103 | bool (*cache_lookup)(u64 leaf_bytenr, void *user_ctx, |
104 | const u64 **root_ids_ret, int *root_count_ret); |
105 | void (*cache_store)(u64 leaf_bytenr, const struct ulist *root_ids, |
106 | void *user_ctx); |
107 | /* |
108 | * If this is not NULL, then the backref walking code will call this |
109 | * for each indirect data extent reference as soon as it finds one, |
110 | * before collecting all the remaining backrefs and before resolving |
111 | * indirect backrefs. This allows for the caller to terminate backref |
112 | * walking as soon as it finds one backref that matches some specific |
113 | * criteria. The @cache_lookup and @cache_store callbacks should not |
114 | * be NULL in order to use this callback. |
115 | */ |
116 | iterate_extent_inodes_t *indirect_ref_iterator; |
117 | /* |
118 | * If this is not NULL, then the backref walking code will call this for |
119 | * each extent item it's meant to process before it actually starts |
120 | * processing it. If this returns anything other than 0, then it stops |
121 | * the backref walking code immediately. |
122 | */ |
123 | int (*check_extent_item)(u64 bytenr, const struct btrfs_extent_item *ei, |
124 | const struct extent_buffer *leaf, void *user_ctx); |
125 | /* |
126 | * If this is not NULL, then the backref walking code will call this for |
127 | * each extent data ref it finds (BTRFS_EXTENT_DATA_REF_KEY keys) before |
128 | * processing that data ref. If this callback return false, then it will |
129 | * ignore this data ref and it will never resolve the indirect data ref, |
130 | * saving time searching for leaves in a fs tree with file extent items |
131 | * matching the data ref. |
132 | */ |
133 | bool (*skip_data_ref)(u64 root, u64 ino, u64 offset, void *user_ctx); |
134 | /* Context object to pass to the callbacks defined above. */ |
135 | void *user_ctx; |
136 | }; |
137 | |
138 | struct inode_fs_paths { |
139 | struct btrfs_path *btrfs_path; |
140 | struct btrfs_root *fs_root; |
141 | struct btrfs_data_container *fspath; |
142 | }; |
143 | |
144 | struct btrfs_backref_shared_cache_entry { |
145 | u64 bytenr; |
146 | u64 gen; |
147 | bool is_shared; |
148 | }; |
149 | |
150 | #define BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE 8 |
151 | |
152 | struct btrfs_backref_share_check_ctx { |
153 | /* Ulists used during backref walking. */ |
154 | struct ulist refs; |
155 | /* |
156 | * The current leaf the caller of btrfs_is_data_extent_shared() is at. |
157 | * Typically the caller (at the moment only fiemap) tries to determine |
158 | * the sharedness of data extents point by file extent items from entire |
159 | * leaves. |
160 | */ |
161 | u64 curr_leaf_bytenr; |
162 | /* |
163 | * The previous leaf the caller was at in the previous call to |
164 | * btrfs_is_data_extent_shared(). This may be the same as the current |
165 | * leaf. On the first call it must be 0. |
166 | */ |
167 | u64 prev_leaf_bytenr; |
168 | /* |
169 | * A path from a root to a leaf that has a file extent item pointing to |
170 | * a given data extent should never exceed the maximum b+tree height. |
171 | */ |
172 | struct btrfs_backref_shared_cache_entry path_cache_entries[BTRFS_MAX_LEVEL]; |
173 | bool use_path_cache; |
174 | /* |
175 | * Cache the sharedness result for the last few extents we have found, |
176 | * but only for extents for which we have multiple file extent items |
177 | * that point to them. |
178 | * It's very common to have several file extent items that point to the |
179 | * same extent (bytenr) but with different offsets and lengths. This |
180 | * typically happens for COW writes, partial writes into prealloc |
181 | * extents, NOCOW writes after snapshoting a root, hole punching or |
182 | * reflinking within the same file (less common perhaps). |
183 | * So keep a small cache with the lookup results for the extent pointed |
184 | * by the last few file extent items. This cache is checked, with a |
185 | * linear scan, whenever btrfs_is_data_extent_shared() is called, so |
186 | * it must be small so that it does not negatively affect performance in |
187 | * case we don't have multiple file extent items that point to the same |
188 | * data extent. |
189 | */ |
190 | struct { |
191 | u64 bytenr; |
192 | bool is_shared; |
193 | } prev_extents_cache[BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE]; |
194 | /* |
195 | * The slot in the prev_extents_cache array that will be used for |
196 | * storing the sharedness result of a new data extent. |
197 | */ |
198 | int prev_extents_cache_slot; |
199 | }; |
200 | |
201 | struct btrfs_backref_share_check_ctx *btrfs_alloc_backref_share_check_ctx(void); |
202 | void btrfs_free_backref_share_ctx(struct btrfs_backref_share_check_ctx *ctx); |
203 | |
204 | int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical, |
205 | struct btrfs_path *path, struct btrfs_key *found_key, |
206 | u64 *flags); |
207 | |
208 | int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb, |
209 | struct btrfs_key *key, struct btrfs_extent_item *ei, |
210 | u32 item_size, u64 *out_root, u8 *out_level); |
211 | |
212 | int iterate_extent_inodes(struct btrfs_backref_walk_ctx *ctx, |
213 | bool search_commit_root, |
214 | iterate_extent_inodes_t *iterate, void *user_ctx); |
215 | |
216 | int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info, |
217 | struct btrfs_path *path, void *ctx, |
218 | bool ignore_offset); |
219 | |
220 | int paths_from_inode(u64 inum, struct inode_fs_paths *ipath); |
221 | |
222 | int btrfs_find_all_leafs(struct btrfs_backref_walk_ctx *ctx); |
223 | int btrfs_find_all_roots(struct btrfs_backref_walk_ctx *ctx, |
224 | bool skip_commit_root_sem); |
225 | char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path, |
226 | u32 name_len, unsigned long name_off, |
227 | struct extent_buffer *eb_in, u64 parent, |
228 | char *dest, u32 size); |
229 | |
230 | struct btrfs_data_container *init_data_container(u32 total_bytes); |
231 | struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root, |
232 | struct btrfs_path *path); |
233 | void free_ipath(struct inode_fs_paths *ipath); |
234 | |
235 | int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid, |
236 | u64 start_off, struct btrfs_path *path, |
237 | struct btrfs_inode_extref **ret_extref, |
238 | u64 *found_off); |
239 | int btrfs_is_data_extent_shared(struct btrfs_inode *inode, u64 bytenr, |
240 | u64 extent_gen, |
241 | struct btrfs_backref_share_check_ctx *ctx); |
242 | |
243 | int __init btrfs_prelim_ref_init(void); |
244 | void __cold btrfs_prelim_ref_exit(void); |
245 | |
246 | struct prelim_ref { |
247 | struct rb_node rbnode; |
248 | u64 root_id; |
249 | struct btrfs_key key_for_search; |
250 | u8 level; |
251 | int count; |
252 | struct extent_inode_elem *inode_list; |
253 | u64 parent; |
254 | u64 wanted_disk_byte; |
255 | }; |
256 | |
257 | /* |
258 | * Iterate backrefs of one extent. |
259 | * |
260 | * Now it only supports iteration of tree block in commit root. |
261 | */ |
262 | struct btrfs_backref_iter { |
263 | u64 bytenr; |
264 | struct btrfs_path *path; |
265 | struct btrfs_fs_info *fs_info; |
266 | struct btrfs_key cur_key; |
267 | u32 item_ptr; |
268 | u32 cur_ptr; |
269 | u32 end_ptr; |
270 | }; |
271 | |
272 | struct btrfs_backref_iter *btrfs_backref_iter_alloc(struct btrfs_fs_info *fs_info); |
273 | |
274 | static inline void btrfs_backref_iter_free(struct btrfs_backref_iter *iter) |
275 | { |
276 | if (!iter) |
277 | return; |
278 | btrfs_free_path(p: iter->path); |
279 | kfree(objp: iter); |
280 | } |
281 | |
282 | static inline struct extent_buffer *btrfs_backref_get_eb( |
283 | struct btrfs_backref_iter *iter) |
284 | { |
285 | if (!iter) |
286 | return NULL; |
287 | return iter->path->nodes[0]; |
288 | } |
289 | |
290 | /* |
291 | * For metadata with EXTENT_ITEM key (non-skinny) case, the first inline data |
292 | * is btrfs_tree_block_info, without a btrfs_extent_inline_ref header. |
293 | * |
294 | * This helper determines if that's the case. |
295 | */ |
296 | static inline bool btrfs_backref_has_tree_block_info( |
297 | struct btrfs_backref_iter *iter) |
298 | { |
299 | if (iter->cur_key.type == BTRFS_EXTENT_ITEM_KEY && |
300 | iter->cur_ptr - iter->item_ptr == sizeof(struct btrfs_extent_item)) |
301 | return true; |
302 | return false; |
303 | } |
304 | |
305 | int btrfs_backref_iter_start(struct btrfs_backref_iter *iter, u64 bytenr); |
306 | |
307 | int btrfs_backref_iter_next(struct btrfs_backref_iter *iter); |
308 | |
309 | static inline bool btrfs_backref_iter_is_inline_ref( |
310 | struct btrfs_backref_iter *iter) |
311 | { |
312 | if (iter->cur_key.type == BTRFS_EXTENT_ITEM_KEY || |
313 | iter->cur_key.type == BTRFS_METADATA_ITEM_KEY) |
314 | return true; |
315 | return false; |
316 | } |
317 | |
318 | static inline void btrfs_backref_iter_release(struct btrfs_backref_iter *iter) |
319 | { |
320 | iter->bytenr = 0; |
321 | iter->item_ptr = 0; |
322 | iter->cur_ptr = 0; |
323 | iter->end_ptr = 0; |
324 | btrfs_release_path(p: iter->path); |
325 | memset(&iter->cur_key, 0, sizeof(iter->cur_key)); |
326 | } |
327 | |
328 | /* |
329 | * Backref cache related structures |
330 | * |
331 | * The whole objective of backref_cache is to build a bi-directional map |
332 | * of tree blocks (represented by backref_node) and all their parents. |
333 | */ |
334 | |
335 | /* |
336 | * Represent a tree block in the backref cache |
337 | */ |
338 | struct btrfs_backref_node { |
339 | struct { |
340 | struct rb_node rb_node; |
341 | u64 bytenr; |
342 | }; /* Use rb_simple_node for search/insert */ |
343 | |
344 | u64 new_bytenr; |
345 | /* Objectid of tree block owner, can be not uptodate */ |
346 | u64 owner; |
347 | /* Link to pending, changed or detached list */ |
348 | struct list_head list; |
349 | |
350 | /* List of upper level edges, which link this node to its parents */ |
351 | struct list_head upper; |
352 | /* List of lower level edges, which link this node to its children */ |
353 | struct list_head lower; |
354 | |
355 | /* NULL if this node is not tree root */ |
356 | struct btrfs_root *root; |
357 | /* Extent buffer got by COWing the block */ |
358 | struct extent_buffer *eb; |
359 | /* Level of the tree block */ |
360 | unsigned int level:8; |
361 | /* Is the block in a non-shareable tree */ |
362 | unsigned int cowonly:1; |
363 | /* 1 if no child node is in the cache */ |
364 | unsigned int lowest:1; |
365 | /* Is the extent buffer locked */ |
366 | unsigned int locked:1; |
367 | /* Has the block been processed */ |
368 | unsigned int processed:1; |
369 | /* Have backrefs of this block been checked */ |
370 | unsigned int checked:1; |
371 | /* |
372 | * 1 if corresponding block has been COWed but some upper level block |
373 | * pointers may not point to the new location |
374 | */ |
375 | unsigned int pending:1; |
376 | /* 1 if the backref node isn't connected to any other backref node */ |
377 | unsigned int detached:1; |
378 | |
379 | /* |
380 | * For generic purpose backref cache, where we only care if it's a reloc |
381 | * root, doesn't care the source subvolid. |
382 | */ |
383 | unsigned int is_reloc_root:1; |
384 | }; |
385 | |
386 | #define LOWER 0 |
387 | #define UPPER 1 |
388 | |
389 | /* |
390 | * Represent an edge connecting upper and lower backref nodes. |
391 | */ |
392 | struct btrfs_backref_edge { |
393 | /* |
394 | * list[LOWER] is linked to btrfs_backref_node::upper of lower level |
395 | * node, and list[UPPER] is linked to btrfs_backref_node::lower of |
396 | * upper level node. |
397 | * |
398 | * Also, build_backref_tree() uses list[UPPER] for pending edges, before |
399 | * linking list[UPPER] to its upper level nodes. |
400 | */ |
401 | struct list_head list[2]; |
402 | |
403 | /* Two related nodes */ |
404 | struct btrfs_backref_node *node[2]; |
405 | }; |
406 | |
407 | struct btrfs_backref_cache { |
408 | /* Red black tree of all backref nodes in the cache */ |
409 | struct rb_root rb_root; |
410 | /* For passing backref nodes to btrfs_reloc_cow_block */ |
411 | struct btrfs_backref_node *path[BTRFS_MAX_LEVEL]; |
412 | /* |
413 | * List of blocks that have been COWed but some block pointers in upper |
414 | * level blocks may not reflect the new location |
415 | */ |
416 | struct list_head pending[BTRFS_MAX_LEVEL]; |
417 | /* List of backref nodes with no child node */ |
418 | struct list_head leaves; |
419 | /* List of blocks that have been COWed in current transaction */ |
420 | struct list_head changed; |
421 | /* List of detached backref node. */ |
422 | struct list_head detached; |
423 | |
424 | u64 last_trans; |
425 | |
426 | int nr_nodes; |
427 | int nr_edges; |
428 | |
429 | /* List of unchecked backref edges during backref cache build */ |
430 | struct list_head pending_edge; |
431 | |
432 | /* List of useless backref nodes during backref cache build */ |
433 | struct list_head useless_node; |
434 | |
435 | struct btrfs_fs_info *fs_info; |
436 | |
437 | /* |
438 | * Whether this cache is for relocation |
439 | * |
440 | * Reloction backref cache require more info for reloc root compared |
441 | * to generic backref cache. |
442 | */ |
443 | bool is_reloc; |
444 | }; |
445 | |
446 | void btrfs_backref_init_cache(struct btrfs_fs_info *fs_info, |
447 | struct btrfs_backref_cache *cache, bool is_reloc); |
448 | struct btrfs_backref_node *btrfs_backref_alloc_node( |
449 | struct btrfs_backref_cache *cache, u64 bytenr, int level); |
450 | struct btrfs_backref_edge *btrfs_backref_alloc_edge( |
451 | struct btrfs_backref_cache *cache); |
452 | |
453 | #define LINK_LOWER (1 << 0) |
454 | #define LINK_UPPER (1 << 1) |
455 | static inline void btrfs_backref_link_edge(struct btrfs_backref_edge *edge, |
456 | struct btrfs_backref_node *lower, |
457 | struct btrfs_backref_node *upper, |
458 | int link_which) |
459 | { |
460 | ASSERT(upper && lower && upper->level == lower->level + 1); |
461 | edge->node[LOWER] = lower; |
462 | edge->node[UPPER] = upper; |
463 | if (link_which & LINK_LOWER) |
464 | list_add_tail(new: &edge->list[LOWER], head: &lower->upper); |
465 | if (link_which & LINK_UPPER) |
466 | list_add_tail(new: &edge->list[UPPER], head: &upper->lower); |
467 | } |
468 | |
469 | static inline void btrfs_backref_free_node(struct btrfs_backref_cache *cache, |
470 | struct btrfs_backref_node *node) |
471 | { |
472 | if (node) { |
473 | ASSERT(list_empty(&node->list)); |
474 | ASSERT(list_empty(&node->lower)); |
475 | ASSERT(node->eb == NULL); |
476 | cache->nr_nodes--; |
477 | btrfs_put_root(root: node->root); |
478 | kfree(objp: node); |
479 | } |
480 | } |
481 | |
482 | static inline void btrfs_backref_free_edge(struct btrfs_backref_cache *cache, |
483 | struct btrfs_backref_edge *edge) |
484 | { |
485 | if (edge) { |
486 | cache->nr_edges--; |
487 | kfree(objp: edge); |
488 | } |
489 | } |
490 | |
491 | static inline void btrfs_backref_unlock_node_buffer( |
492 | struct btrfs_backref_node *node) |
493 | { |
494 | if (node->locked) { |
495 | btrfs_tree_unlock(eb: node->eb); |
496 | node->locked = 0; |
497 | } |
498 | } |
499 | |
500 | static inline void btrfs_backref_drop_node_buffer( |
501 | struct btrfs_backref_node *node) |
502 | { |
503 | if (node->eb) { |
504 | btrfs_backref_unlock_node_buffer(node); |
505 | free_extent_buffer(eb: node->eb); |
506 | node->eb = NULL; |
507 | } |
508 | } |
509 | |
510 | /* |
511 | * Drop the backref node from cache without cleaning up its children |
512 | * edges. |
513 | * |
514 | * This can only be called on node without parent edges. |
515 | * The children edges are still kept as is. |
516 | */ |
517 | static inline void btrfs_backref_drop_node(struct btrfs_backref_cache *tree, |
518 | struct btrfs_backref_node *node) |
519 | { |
520 | ASSERT(list_empty(&node->upper)); |
521 | |
522 | btrfs_backref_drop_node_buffer(node); |
523 | list_del_init(entry: &node->list); |
524 | list_del_init(entry: &node->lower); |
525 | if (!RB_EMPTY_NODE(&node->rb_node)) |
526 | rb_erase(&node->rb_node, &tree->rb_root); |
527 | btrfs_backref_free_node(cache: tree, node); |
528 | } |
529 | |
530 | void btrfs_backref_cleanup_node(struct btrfs_backref_cache *cache, |
531 | struct btrfs_backref_node *node); |
532 | |
533 | void btrfs_backref_release_cache(struct btrfs_backref_cache *cache); |
534 | |
535 | static inline void btrfs_backref_panic(struct btrfs_fs_info *fs_info, |
536 | u64 bytenr, int error) |
537 | { |
538 | btrfs_panic(fs_info, error, |
539 | "Inconsistency in backref cache found at offset %llu" , |
540 | bytenr); |
541 | } |
542 | |
543 | int btrfs_backref_add_tree_node(struct btrfs_trans_handle *trans, |
544 | struct btrfs_backref_cache *cache, |
545 | struct btrfs_path *path, |
546 | struct btrfs_backref_iter *iter, |
547 | struct btrfs_key *node_key, |
548 | struct btrfs_backref_node *cur); |
549 | |
550 | int btrfs_backref_finish_upper_links(struct btrfs_backref_cache *cache, |
551 | struct btrfs_backref_node *start); |
552 | |
553 | void btrfs_backref_error_cleanup(struct btrfs_backref_cache *cache, |
554 | struct btrfs_backref_node *node); |
555 | |
556 | #endif |
557 | |