1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2011 STRATO. All rights reserved.
6 #ifndef BTRFS_BACKREF_H
7 #define BTRFS_BACKREF_H
9 #include <linux/btrfs.h>
13 #include "extent_io.h"
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
23 #define BTRFS_ITERATE_EXTENT_INODES_STOP 5
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
31 typedef int (iterate_extent_inodes_t)(u64 inum, u64 offset, u64 num_bytes,
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.
39 struct btrfs_backref_walk_ctx {
41 * The address of the extent for which we are doing backref walking.
42 * Can be either a data extent or a metadata extent.
44 * Must always be set by the top level caller.
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.
54 * Must always be set by the top level caller.
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.
62 bool ignore_extent_item_pos;
63 /* A valid transaction handle or NULL. */
64 struct btrfs_trans_handle *trans;
66 * The file system's info object, can not be NULL.
68 * Must always be set by the top level caller.
70 struct btrfs_fs_info *fs_info;
72 * Time sequence acquired from btrfs_get_tree_mod_seq(), in case the
73 * caller joined the tree mod log to get a consistent view of b+trees
74 * while we do backref walking, or BTRFS_SEQ_LAST.
75 * When using BTRFS_SEQ_LAST, delayed refs are not checked and it uses
76 * commit roots when searching b+trees - this is a special case for
77 * qgroups used during a transaction commit.
81 * Used to collect the bytenr of metadata extents that point to the
86 * List used to collect the IDs of the roots from which the target
87 * extent is accessible. Can be NULL in case the caller does not care
88 * about collecting root IDs.
92 * Used by iterate_extent_inodes() and the main backref walk code
93 * (find_parent_nodes()). Lookup and store functions for an optional
94 * cache which maps the logical address (bytenr) of leaves to an array
97 bool (*cache_lookup)(u64 leaf_bytenr, void *user_ctx,
98 const u64 **root_ids_ret, int *root_count_ret);
99 void (*cache_store)(u64 leaf_bytenr, const struct ulist *root_ids,
102 * If this is not NULL, then the backref walking code will call this
103 * for each indirect data extent reference as soon as it finds one,
104 * before collecting all the remaining backrefs and before resolving
105 * indirect backrefs. This allows for the caller to terminate backref
106 * walking as soon as it finds one backref that matches some specific
107 * criteria. The @cache_lookup and @cache_store callbacks should not
108 * be NULL in order to use this callback.
110 iterate_extent_inodes_t *indirect_ref_iterator;
112 * If this is not NULL, then the backref walking code will call this for
113 * each extent item it's meant to process before it actually starts
114 * processing it. If this returns anything other than 0, then it stops
115 * the backref walking code immediately.
117 int (*check_extent_item)(u64 bytenr, const struct btrfs_extent_item *ei,
118 const struct extent_buffer *leaf, void *user_ctx);
120 * If this is not NULL, then the backref walking code will call this for
121 * each extent data ref it finds (BTRFS_EXTENT_DATA_REF_KEY keys) before
122 * processing that data ref. If this callback return false, then it will
123 * ignore this data ref and it will never resolve the indirect data ref,
124 * saving time searching for leaves in a fs tree with file extent items
125 * matching the data ref.
127 bool (*skip_data_ref)(u64 root, u64 ino, u64 offset, void *user_ctx);
128 /* Context object to pass to the callbacks defined above. */
132 struct inode_fs_paths {
133 struct btrfs_path *btrfs_path;
134 struct btrfs_root *fs_root;
135 struct btrfs_data_container *fspath;
138 struct btrfs_backref_shared_cache_entry {
144 #define BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE 8
146 struct btrfs_backref_share_check_ctx {
147 /* Ulists used during backref walking. */
150 * The current leaf the caller of btrfs_is_data_extent_shared() is at.
151 * Typically the caller (at the moment only fiemap) tries to determine
152 * the sharedness of data extents point by file extent items from entire
155 u64 curr_leaf_bytenr;
157 * The previous leaf the caller was at in the previous call to
158 * btrfs_is_data_extent_shared(). This may be the same as the current
159 * leaf. On the first call it must be 0.
161 u64 prev_leaf_bytenr;
163 * A path from a root to a leaf that has a file extent item pointing to
164 * a given data extent should never exceed the maximum b+tree height.
166 struct btrfs_backref_shared_cache_entry path_cache_entries[BTRFS_MAX_LEVEL];
169 * Cache the sharedness result for the last few extents we have found,
170 * but only for extents for which we have multiple file extent items
171 * that point to them.
172 * It's very common to have several file extent items that point to the
173 * same extent (bytenr) but with different offsets and lengths. This
174 * typically happens for COW writes, partial writes into prealloc
175 * extents, NOCOW writes after snapshoting a root, hole punching or
176 * reflinking within the same file (less common perhaps).
177 * So keep a small cache with the lookup results for the extent pointed
178 * by the last few file extent items. This cache is checked, with a
179 * linear scan, whenever btrfs_is_data_extent_shared() is called, so
180 * it must be small so that it does not negatively affect performance in
181 * case we don't have multiple file extent items that point to the same
187 } prev_extents_cache[BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE];
189 * The slot in the prev_extents_cache array that will be used for
190 * storing the sharedness result of a new data extent.
192 int prev_extents_cache_slot;
195 struct btrfs_backref_share_check_ctx *btrfs_alloc_backref_share_check_ctx(void);
196 void btrfs_free_backref_share_ctx(struct btrfs_backref_share_check_ctx *ctx);
198 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
199 struct btrfs_path *path, struct btrfs_key *found_key,
202 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
203 struct btrfs_key *key, struct btrfs_extent_item *ei,
204 u32 item_size, u64 *out_root, u8 *out_level);
206 int iterate_extent_inodes(struct btrfs_backref_walk_ctx *ctx,
207 bool search_commit_root,
208 iterate_extent_inodes_t *iterate, void *user_ctx);
210 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
211 struct btrfs_path *path, void *ctx,
214 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath);
216 int btrfs_find_all_leafs(struct btrfs_backref_walk_ctx *ctx);
217 int btrfs_find_all_roots(struct btrfs_backref_walk_ctx *ctx,
218 bool skip_commit_root_sem);
219 char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
220 u32 name_len, unsigned long name_off,
221 struct extent_buffer *eb_in, u64 parent,
222 char *dest, u32 size);
224 struct btrfs_data_container *init_data_container(u32 total_bytes);
225 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
226 struct btrfs_path *path);
227 void free_ipath(struct inode_fs_paths *ipath);
229 int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
230 u64 start_off, struct btrfs_path *path,
231 struct btrfs_inode_extref **ret_extref,
233 int btrfs_is_data_extent_shared(struct btrfs_inode *inode, u64 bytenr,
235 struct btrfs_backref_share_check_ctx *ctx);
237 int __init btrfs_prelim_ref_init(void);
238 void __cold btrfs_prelim_ref_exit(void);
241 struct rb_node rbnode;
243 struct btrfs_key key_for_search;
246 struct extent_inode_elem *inode_list;
248 u64 wanted_disk_byte;
252 * Iterate backrefs of one extent.
254 * Now it only supports iteration of tree block in commit root.
256 struct btrfs_backref_iter {
258 struct btrfs_path *path;
259 struct btrfs_fs_info *fs_info;
260 struct btrfs_key cur_key;
266 struct btrfs_backref_iter *btrfs_backref_iter_alloc(struct btrfs_fs_info *fs_info);
268 static inline void btrfs_backref_iter_free(struct btrfs_backref_iter *iter)
272 btrfs_free_path(iter->path);
276 static inline struct extent_buffer *btrfs_backref_get_eb(
277 struct btrfs_backref_iter *iter)
281 return iter->path->nodes[0];
285 * For metadata with EXTENT_ITEM key (non-skinny) case, the first inline data
286 * is btrfs_tree_block_info, without a btrfs_extent_inline_ref header.
288 * This helper determines if that's the case.
290 static inline bool btrfs_backref_has_tree_block_info(
291 struct btrfs_backref_iter *iter)
293 if (iter->cur_key.type == BTRFS_EXTENT_ITEM_KEY &&
294 iter->cur_ptr - iter->item_ptr == sizeof(struct btrfs_extent_item))
299 int btrfs_backref_iter_start(struct btrfs_backref_iter *iter, u64 bytenr);
301 int btrfs_backref_iter_next(struct btrfs_backref_iter *iter);
303 static inline bool btrfs_backref_iter_is_inline_ref(
304 struct btrfs_backref_iter *iter)
306 if (iter->cur_key.type == BTRFS_EXTENT_ITEM_KEY ||
307 iter->cur_key.type == BTRFS_METADATA_ITEM_KEY)
312 static inline void btrfs_backref_iter_release(struct btrfs_backref_iter *iter)
318 btrfs_release_path(iter->path);
319 memset(&iter->cur_key, 0, sizeof(iter->cur_key));
323 * Backref cache related structures
325 * The whole objective of backref_cache is to build a bi-directional map
326 * of tree blocks (represented by backref_node) and all their parents.
330 * Represent a tree block in the backref cache
332 struct btrfs_backref_node {
334 struct rb_node rb_node;
336 }; /* Use rb_simple_node for search/insert */
339 /* Objectid of tree block owner, can be not uptodate */
341 /* Link to pending, changed or detached list */
342 struct list_head list;
344 /* List of upper level edges, which link this node to its parents */
345 struct list_head upper;
346 /* List of lower level edges, which link this node to its children */
347 struct list_head lower;
349 /* NULL if this node is not tree root */
350 struct btrfs_root *root;
351 /* Extent buffer got by COWing the block */
352 struct extent_buffer *eb;
353 /* Level of the tree block */
354 unsigned int level:8;
355 /* Is the block in a non-shareable tree */
356 unsigned int cowonly:1;
357 /* 1 if no child node is in the cache */
358 unsigned int lowest:1;
359 /* Is the extent buffer locked */
360 unsigned int locked:1;
361 /* Has the block been processed */
362 unsigned int processed:1;
363 /* Have backrefs of this block been checked */
364 unsigned int checked:1;
366 * 1 if corresponding block has been COWed but some upper level block
367 * pointers may not point to the new location
369 unsigned int pending:1;
370 /* 1 if the backref node isn't connected to any other backref node */
371 unsigned int detached:1;
374 * For generic purpose backref cache, where we only care if it's a reloc
375 * root, doesn't care the source subvolid.
377 unsigned int is_reloc_root:1;
384 * Represent an edge connecting upper and lower backref nodes.
386 struct btrfs_backref_edge {
388 * list[LOWER] is linked to btrfs_backref_node::upper of lower level
389 * node, and list[UPPER] is linked to btrfs_backref_node::lower of
392 * Also, build_backref_tree() uses list[UPPER] for pending edges, before
393 * linking list[UPPER] to its upper level nodes.
395 struct list_head list[2];
397 /* Two related nodes */
398 struct btrfs_backref_node *node[2];
401 struct btrfs_backref_cache {
402 /* Red black tree of all backref nodes in the cache */
403 struct rb_root rb_root;
404 /* For passing backref nodes to btrfs_reloc_cow_block */
405 struct btrfs_backref_node *path[BTRFS_MAX_LEVEL];
407 * List of blocks that have been COWed but some block pointers in upper
408 * level blocks may not reflect the new location
410 struct list_head pending[BTRFS_MAX_LEVEL];
411 /* List of backref nodes with no child node */
412 struct list_head leaves;
413 /* List of blocks that have been COWed in current transaction */
414 struct list_head changed;
415 /* List of detached backref node. */
416 struct list_head detached;
423 /* List of unchecked backref edges during backref cache build */
424 struct list_head pending_edge;
426 /* List of useless backref nodes during backref cache build */
427 struct list_head useless_node;
429 struct btrfs_fs_info *fs_info;
432 * Whether this cache is for relocation
434 * Reloction backref cache require more info for reloc root compared
435 * to generic backref cache.
437 unsigned int is_reloc;
440 void btrfs_backref_init_cache(struct btrfs_fs_info *fs_info,
441 struct btrfs_backref_cache *cache, int is_reloc);
442 struct btrfs_backref_node *btrfs_backref_alloc_node(
443 struct btrfs_backref_cache *cache, u64 bytenr, int level);
444 struct btrfs_backref_edge *btrfs_backref_alloc_edge(
445 struct btrfs_backref_cache *cache);
447 #define LINK_LOWER (1 << 0)
448 #define LINK_UPPER (1 << 1)
449 static inline void btrfs_backref_link_edge(struct btrfs_backref_edge *edge,
450 struct btrfs_backref_node *lower,
451 struct btrfs_backref_node *upper,
454 ASSERT(upper && lower && upper->level == lower->level + 1);
455 edge->node[LOWER] = lower;
456 edge->node[UPPER] = upper;
457 if (link_which & LINK_LOWER)
458 list_add_tail(&edge->list[LOWER], &lower->upper);
459 if (link_which & LINK_UPPER)
460 list_add_tail(&edge->list[UPPER], &upper->lower);
463 static inline void btrfs_backref_free_node(struct btrfs_backref_cache *cache,
464 struct btrfs_backref_node *node)
467 ASSERT(list_empty(&node->list));
468 ASSERT(list_empty(&node->lower));
469 ASSERT(node->eb == NULL);
471 btrfs_put_root(node->root);
476 static inline void btrfs_backref_free_edge(struct btrfs_backref_cache *cache,
477 struct btrfs_backref_edge *edge)
485 static inline void btrfs_backref_unlock_node_buffer(
486 struct btrfs_backref_node *node)
489 btrfs_tree_unlock(node->eb);
494 static inline void btrfs_backref_drop_node_buffer(
495 struct btrfs_backref_node *node)
498 btrfs_backref_unlock_node_buffer(node);
499 free_extent_buffer(node->eb);
505 * Drop the backref node from cache without cleaning up its children
508 * This can only be called on node without parent edges.
509 * The children edges are still kept as is.
511 static inline void btrfs_backref_drop_node(struct btrfs_backref_cache *tree,
512 struct btrfs_backref_node *node)
514 ASSERT(list_empty(&node->upper));
516 btrfs_backref_drop_node_buffer(node);
517 list_del_init(&node->list);
518 list_del_init(&node->lower);
519 if (!RB_EMPTY_NODE(&node->rb_node))
520 rb_erase(&node->rb_node, &tree->rb_root);
521 btrfs_backref_free_node(tree, node);
524 void btrfs_backref_cleanup_node(struct btrfs_backref_cache *cache,
525 struct btrfs_backref_node *node);
527 void btrfs_backref_release_cache(struct btrfs_backref_cache *cache);
529 static inline void btrfs_backref_panic(struct btrfs_fs_info *fs_info,
530 u64 bytenr, int errno)
532 btrfs_panic(fs_info, errno,
533 "Inconsistency in backref cache found at offset %llu",
537 int btrfs_backref_add_tree_node(struct btrfs_backref_cache *cache,
538 struct btrfs_path *path,
539 struct btrfs_backref_iter *iter,
540 struct btrfs_key *node_key,
541 struct btrfs_backref_node *cur);
543 int btrfs_backref_finish_upper_links(struct btrfs_backref_cache *cache,
544 struct btrfs_backref_node *start);
546 void btrfs_backref_error_cleanup(struct btrfs_backref_cache *cache,
547 struct btrfs_backref_node *node);