backref.h

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