1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2007 Oracle. All rights reserved.
9 #include <linux/pagemap.h>
13 struct btrfs_trans_handle;
14 struct btrfs_transaction;
15 struct btrfs_pending_snapshot;
16 struct btrfs_delayed_ref_root;
17 struct btrfs_space_info;
18 struct btrfs_block_group;
19 struct btrfs_ordered_sum;
22 struct btrfs_ioctl_encoded_io_args;
24 struct btrfs_fs_devices;
25 struct btrfs_balance_control;
26 struct btrfs_delayed_root;
29 /* Read ahead values for struct btrfs_path.reada */
35 * Similar to READA_FORWARD but unlike it:
37 * 1) It will trigger readahead even for leaves that are not close to
39 * 2) It also triggers readahead for nodes;
40 * 3) During a search, even when a node or leaf is already in memory, it
41 * will still trigger readahead for other nodes and leaves that follow
44 * This is meant to be used only when we know we are iterating over the
45 * entire tree or a very large part of it.
51 * btrfs_paths remember the path taken from the root down to the leaf.
52 * level 0 is always the leaf, and nodes[1...BTRFS_MAX_LEVEL] will point
53 * to any other levels that are present.
55 * The slots array records the index of the item or block pointer
56 * used while walking the tree.
59 struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
60 int slots[BTRFS_MAX_LEVEL];
61 /* if there is real range locking, this locks field will change */
62 u8 locks[BTRFS_MAX_LEVEL];
64 /* keep some upper locks as we walk down */
68 * set by btrfs_split_item, tells search_slot to keep all locks
69 * and to force calls to keep space in the nodes
71 unsigned int search_for_split:1;
72 unsigned int keep_locks:1;
73 unsigned int skip_locking:1;
74 unsigned int search_commit_root:1;
75 unsigned int need_commit_sem:1;
76 unsigned int skip_release_on_error:1;
78 * Indicate that new item (btrfs_search_slot) is extending already
79 * existing item and ins_len contains only the data size and not item
80 * header (ie. sizeof(struct btrfs_item) is not included).
82 unsigned int search_for_extension:1;
83 /* Stop search if any locks need to be taken (for read) */
84 unsigned int nowait:1;
88 * The state of btrfs root
92 * btrfs_record_root_in_trans is a multi-step process, and it can race
93 * with the balancing code. But the race is very small, and only the
94 * first time the root is added to each transaction. So IN_TRANS_SETUP
95 * is used to tell us when more checks are required
97 BTRFS_ROOT_IN_TRANS_SETUP,
100 * Set if tree blocks of this root can be shared by other roots.
101 * Only subvolume trees and their reloc trees have this bit set.
102 * Conflicts with TRACK_DIRTY bit.
104 * This affects two things:
106 * - How balance works
107 * For shareable roots, we need to use reloc tree and do path
108 * replacement for balance, and need various pre/post hooks for
109 * snapshot creation to handle them.
111 * While for non-shareable trees, we just simply do a tree search
114 * - How dirty roots are tracked
115 * For shareable roots, btrfs_record_root_in_trans() is needed to
116 * track them, while non-subvolume roots have TRACK_DIRTY bit, they
117 * don't need to set this manually.
119 BTRFS_ROOT_SHAREABLE,
120 BTRFS_ROOT_TRACK_DIRTY,
122 BTRFS_ROOT_ORPHAN_ITEM_INSERTED,
123 BTRFS_ROOT_DEFRAG_RUNNING,
124 BTRFS_ROOT_FORCE_COW,
125 BTRFS_ROOT_MULTI_LOG_TASKS,
130 * Reloc tree is orphan, only kept here for qgroup delayed subtree scan
132 * Set for the subvolume tree owning the reloc tree.
134 BTRFS_ROOT_DEAD_RELOC_TREE,
135 /* Mark dead root stored on device whose cleanup needs to be resumed */
136 BTRFS_ROOT_DEAD_TREE,
137 /* The root has a log tree. Used for subvolume roots and the tree root. */
138 BTRFS_ROOT_HAS_LOG_TREE,
139 /* Qgroup flushing is in progress */
140 BTRFS_ROOT_QGROUP_FLUSHING,
141 /* We started the orphan cleanup for this root. */
142 BTRFS_ROOT_ORPHAN_CLEANUP,
143 /* This root has a drop operation that was started previously. */
144 BTRFS_ROOT_UNFINISHED_DROP,
145 /* This reloc root needs to have its buffers lockdep class reset. */
146 BTRFS_ROOT_RESET_LOCKDEP_CLASS,
150 * Record swapped tree blocks of a subvolume tree for delayed subtree trace
151 * code. For detail check comment in fs/btrfs/qgroup.c.
153 struct btrfs_qgroup_swapped_blocks {
155 /* RM_EMPTY_ROOT() of above blocks[] */
157 struct rb_root blocks[BTRFS_MAX_LEVEL];
161 * in ram representation of the tree. extent_root is used for all allocations
162 * and for the extent tree extent_root root.
165 struct rb_node rb_node;
167 struct extent_buffer *node;
169 struct extent_buffer *commit_root;
170 struct btrfs_root *log_root;
171 struct btrfs_root *reloc_root;
174 struct btrfs_root_item root_item;
175 struct btrfs_key root_key;
176 struct btrfs_fs_info *fs_info;
177 struct extent_io_tree dirty_log_pages;
179 struct mutex objectid_mutex;
181 spinlock_t accounting_lock;
182 struct btrfs_block_rsv *block_rsv;
184 struct mutex log_mutex;
185 wait_queue_head_t log_writer_wait;
186 wait_queue_head_t log_commit_wait[2];
187 struct list_head log_ctxs[2];
188 /* Used only for log trees of subvolumes, not for the log root tree */
189 atomic_t log_writers;
190 atomic_t log_commit[2];
191 /* Used only for log trees of subvolumes, not for the log root tree */
194 /* No matter the commit succeeds or not*/
195 int log_transid_committed;
196 /* Just be updated when the commit succeeds. */
206 struct btrfs_key defrag_progress;
207 struct btrfs_key defrag_max;
209 /* The dirty list is only used by non-shareable roots */
210 struct list_head dirty_list;
212 struct list_head root_list;
214 spinlock_t log_extents_lock[2];
215 struct list_head logged_list[2];
217 spinlock_t inode_lock;
218 /* red-black tree that keeps track of in-memory inodes */
219 struct rb_root inode_tree;
222 * radix tree that keeps track of delayed nodes of every inode,
223 * protected by inode_lock
225 struct radix_tree_root delayed_nodes_tree;
227 * right now this just gets used so that a root has its own devid
228 * for stat. It may be used for more later
232 spinlock_t root_item_lock;
235 struct mutex delalloc_mutex;
236 spinlock_t delalloc_lock;
238 * all of the inodes that have delalloc bytes. It is possible for
239 * this list to be empty even when there is still dirty data=ordered
240 * extents waiting to finish IO.
242 struct list_head delalloc_inodes;
243 struct list_head delalloc_root;
244 u64 nr_delalloc_inodes;
246 struct mutex ordered_extent_mutex;
248 * this is used by the balancing code to wait for all the pending
251 spinlock_t ordered_extent_lock;
254 * all of the data=ordered extents pending writeback
255 * these can span multiple transactions and basically include
256 * every dirty data page that isn't from nodatacow
258 struct list_head ordered_extents;
259 struct list_head ordered_root;
260 u64 nr_ordered_extents;
263 * Not empty if this subvolume root has gone through tree block swap
266 * Will be used by reloc_control::dirty_subvol_roots.
268 struct list_head reloc_dirty_list;
271 * Number of currently running SEND ioctls to prevent
272 * manipulation with the read-only status via SUBVOL_SETFLAGS
274 int send_in_progress;
276 * Number of currently running deduplication operations that have a
277 * destination inode belonging to this root. Protected by the lock
280 int dedupe_in_progress;
281 /* For exclusion of snapshot creation and nocow writes */
282 struct btrfs_drew_lock snapshot_lock;
284 atomic_t snapshot_force_cow;
286 /* For qgroup metadata reserved space */
287 spinlock_t qgroup_meta_rsv_lock;
288 u64 qgroup_meta_rsv_pertrans;
289 u64 qgroup_meta_rsv_prealloc;
290 wait_queue_head_t qgroup_flush_wait;
292 /* Number of active swapfiles */
293 atomic_t nr_swapfiles;
295 /* Record pairs of swapped blocks for qgroup */
296 struct btrfs_qgroup_swapped_blocks swapped_blocks;
298 /* Used only by log trees, when logging csum items */
299 struct extent_io_tree log_csum_range;
301 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
305 #ifdef CONFIG_BTRFS_DEBUG
306 struct list_head leak_list;
310 static inline bool btrfs_root_readonly(const struct btrfs_root *root)
312 /* Byte-swap the constant at compile time, root_item::flags is LE */
313 return (root->root_item.flags & cpu_to_le64(BTRFS_ROOT_SUBVOL_RDONLY)) != 0;
316 static inline bool btrfs_root_dead(const struct btrfs_root *root)
318 /* Byte-swap the constant at compile time, root_item::flags is LE */
319 return (root->root_item.flags & cpu_to_le64(BTRFS_ROOT_SUBVOL_DEAD)) != 0;
322 static inline u64 btrfs_root_id(const struct btrfs_root *root)
324 return root->root_key.objectid;
328 * Structure that conveys information about an extent that is going to replace
329 * all the extents in a file range.
331 struct btrfs_replace_extent_info {
337 /* Pointer to a file extent item of type regular or prealloc. */
340 * Set to true when attempting to replace a file range with a new extent
341 * described by this structure, set to false when attempting to clone an
342 * existing extent into a file range.
345 /* Indicate if we should update the inode's mtime and ctime. */
347 /* Meaningful only if is_new_extent is true. */
350 * Meaningful only if is_new_extent is true.
351 * Used to track how many extent items we have already inserted in a
352 * subvolume tree that refer to the extent described by this structure,
353 * so that we know when to create a new delayed ref or update an existing
359 /* Arguments for btrfs_drop_extents() */
360 struct btrfs_drop_extents_args {
361 /* Input parameters */
364 * If NULL, btrfs_drop_extents() will allocate and free its own path.
365 * If 'replace_extent' is true, this must not be NULL. Also the path
366 * is always released except if 'replace_extent' is true and
367 * btrfs_drop_extents() sets 'extent_inserted' to true, in which case
368 * the path is kept locked.
370 struct btrfs_path *path;
371 /* Start offset of the range to drop extents from */
373 /* End (exclusive, last byte + 1) of the range to drop extents from */
375 /* If true drop all the extent maps in the range */
378 * If true it means we want to insert a new extent after dropping all
379 * the extents in the range. If this is true, the 'extent_item_size'
380 * parameter must be set as well and the 'extent_inserted' field will
381 * be set to true by btrfs_drop_extents() if it could insert the new
383 * Note: when this is set to true the path must not be NULL.
387 * Used if 'replace_extent' is true. Size of the file extent item to
388 * insert after dropping all existing extents in the range
390 u32 extent_item_size;
392 /* Output parameters */
395 * Set to the minimum between the input parameter 'end' and the end
396 * (exclusive, last byte + 1) of the last dropped extent. This is always
397 * set even if btrfs_drop_extents() returns an error.
401 * The number of allocated bytes found in the range. This can be smaller
402 * than the range's length when there are holes in the range.
406 * Only set if 'replace_extent' is true. Set to true if we were able
407 * to insert a replacement extent after dropping all extents in the
408 * range, otherwise set to false by btrfs_drop_extents().
409 * Also, if btrfs_drop_extents() has set this to true it means it
410 * returned with the path locked, otherwise if it has set this to
411 * false it has returned with the path released.
413 bool extent_inserted;
416 struct btrfs_file_private {
419 struct extent_state *llseek_cached_state;
422 static inline u32 BTRFS_LEAF_DATA_SIZE(const struct btrfs_fs_info *info)
424 return info->nodesize - sizeof(struct btrfs_header);
427 static inline u32 BTRFS_MAX_ITEM_SIZE(const struct btrfs_fs_info *info)
429 return BTRFS_LEAF_DATA_SIZE(info) - sizeof(struct btrfs_item);
432 static inline u32 BTRFS_NODEPTRS_PER_BLOCK(const struct btrfs_fs_info *info)
434 return BTRFS_LEAF_DATA_SIZE(info) / sizeof(struct btrfs_key_ptr);
437 static inline u32 BTRFS_MAX_XATTR_SIZE(const struct btrfs_fs_info *info)
439 return BTRFS_MAX_ITEM_SIZE(info) - sizeof(struct btrfs_dir_item);
442 #define BTRFS_BYTES_TO_BLKS(fs_info, bytes) \
443 ((bytes) >> (fs_info)->sectorsize_bits)
445 static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping)
447 return mapping_gfp_constraint(mapping, ~__GFP_FS);
450 int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
452 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
453 u64 num_bytes, u64 *actual_bytes);
454 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range);
457 int __init btrfs_ctree_init(void);
458 void __cold btrfs_ctree_exit(void);
460 int btrfs_bin_search(struct extent_buffer *eb, int first_slot,
461 const struct btrfs_key *key, int *slot);
463 int __pure btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2);
464 int btrfs_previous_item(struct btrfs_root *root,
465 struct btrfs_path *path, u64 min_objectid,
467 int btrfs_previous_extent_item(struct btrfs_root *root,
468 struct btrfs_path *path, u64 min_objectid);
469 void btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
470 struct btrfs_path *path,
471 const struct btrfs_key *new_key);
472 struct extent_buffer *btrfs_root_node(struct btrfs_root *root);
473 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
474 struct btrfs_key *key, int lowest_level,
476 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
477 struct btrfs_path *path,
479 struct extent_buffer *btrfs_read_node_slot(struct extent_buffer *parent,
482 int btrfs_cow_block(struct btrfs_trans_handle *trans,
483 struct btrfs_root *root, struct extent_buffer *buf,
484 struct extent_buffer *parent, int parent_slot,
485 struct extent_buffer **cow_ret,
486 enum btrfs_lock_nesting nest);
487 int btrfs_copy_root(struct btrfs_trans_handle *trans,
488 struct btrfs_root *root,
489 struct extent_buffer *buf,
490 struct extent_buffer **cow_ret, u64 new_root_objectid);
491 int btrfs_block_can_be_shared(struct btrfs_root *root,
492 struct extent_buffer *buf);
493 int btrfs_del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
494 struct btrfs_path *path, int level, int slot);
495 void btrfs_extend_item(struct btrfs_trans_handle *trans,
496 struct btrfs_path *path, u32 data_size);
497 void btrfs_truncate_item(struct btrfs_trans_handle *trans,
498 struct btrfs_path *path, u32 new_size, int from_end);
499 int btrfs_split_item(struct btrfs_trans_handle *trans,
500 struct btrfs_root *root,
501 struct btrfs_path *path,
502 const struct btrfs_key *new_key,
503 unsigned long split_offset);
504 int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
505 struct btrfs_root *root,
506 struct btrfs_path *path,
507 const struct btrfs_key *new_key);
508 int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path,
509 u64 inum, u64 ioff, u8 key_type, struct btrfs_key *found_key);
510 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
511 const struct btrfs_key *key, struct btrfs_path *p,
512 int ins_len, int cow);
513 int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key,
514 struct btrfs_path *p, u64 time_seq);
515 int btrfs_search_slot_for_read(struct btrfs_root *root,
516 const struct btrfs_key *key,
517 struct btrfs_path *p, int find_higher,
519 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
520 struct btrfs_root *root, struct extent_buffer *parent,
521 int start_slot, u64 *last_ret,
522 struct btrfs_key *progress);
523 void btrfs_release_path(struct btrfs_path *p);
524 struct btrfs_path *btrfs_alloc_path(void);
525 void btrfs_free_path(struct btrfs_path *p);
527 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
528 struct btrfs_path *path, int slot, int nr);
529 static inline int btrfs_del_item(struct btrfs_trans_handle *trans,
530 struct btrfs_root *root,
531 struct btrfs_path *path)
533 return btrfs_del_items(trans, root, path, path->slots[0], 1);
537 * Describes a batch of items to insert in a btree. This is used by
538 * btrfs_insert_empty_items().
540 struct btrfs_item_batch {
542 * Pointer to an array containing the keys of the items to insert (in
545 const struct btrfs_key *keys;
546 /* Pointer to an array containing the data size for each item to insert. */
547 const u32 *data_sizes;
549 * The sum of data sizes for all items. The caller can compute this while
550 * setting up the data_sizes array, so it ends up being more efficient
551 * than having btrfs_insert_empty_items() or setup_item_for_insert()
552 * doing it, as it would avoid an extra loop over a potentially large
553 * array, and in the case of setup_item_for_insert(), we would be doing
554 * it while holding a write lock on a leaf and often on upper level nodes
555 * too, unnecessarily increasing the size of a critical section.
558 /* Size of the keys and data_sizes arrays (number of items in the batch). */
562 void btrfs_setup_item_for_insert(struct btrfs_trans_handle *trans,
563 struct btrfs_root *root,
564 struct btrfs_path *path,
565 const struct btrfs_key *key,
567 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
568 const struct btrfs_key *key, void *data, u32 data_size);
569 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
570 struct btrfs_root *root,
571 struct btrfs_path *path,
572 const struct btrfs_item_batch *batch);
574 static inline int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
575 struct btrfs_root *root,
576 struct btrfs_path *path,
577 const struct btrfs_key *key,
580 struct btrfs_item_batch batch;
583 batch.data_sizes = &data_size;
584 batch.total_data_size = data_size;
587 return btrfs_insert_empty_items(trans, root, path, &batch);
590 int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
593 int btrfs_search_backwards(struct btrfs_root *root, struct btrfs_key *key,
594 struct btrfs_path *path);
596 int btrfs_get_next_valid_item(struct btrfs_root *root, struct btrfs_key *key,
597 struct btrfs_path *path);
600 * Search in @root for a given @key, and store the slot found in @found_key.
602 * @root: The root node of the tree.
603 * @key: The key we are looking for.
604 * @found_key: Will hold the found item.
605 * @path: Holds the current slot/leaf.
606 * @iter_ret: Contains the value returned from btrfs_search_slot or
607 * btrfs_get_next_valid_item, whichever was executed last.
609 * The @iter_ret is an output variable that will contain the return value of
610 * btrfs_search_slot, if it encountered an error, or the value returned from
611 * btrfs_get_next_valid_item otherwise. That return value can be 0, if a valid
612 * slot was found, 1 if there were no more leaves, and <0 if there was an error.
614 * It's recommended to use a separate variable for iter_ret and then use it to
615 * set the function return value so there's no confusion of the 0/1/errno
616 * values stemming from btrfs_search_slot.
618 #define btrfs_for_each_slot(root, key, found_key, path, iter_ret) \
619 for (iter_ret = btrfs_search_slot(NULL, (root), (key), (path), 0, 0); \
621 (iter_ret = btrfs_get_next_valid_item((root), (found_key), (path))) == 0; \
625 int btrfs_next_old_item(struct btrfs_root *root, struct btrfs_path *path, u64 time_seq);
628 * Search the tree again to find a leaf with greater keys.
630 * Returns 0 if it found something or 1 if there are no greater leaves.
631 * Returns < 0 on error.
633 static inline int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
635 return btrfs_next_old_leaf(root, path, 0);
638 static inline int btrfs_next_item(struct btrfs_root *root, struct btrfs_path *p)
640 return btrfs_next_old_item(root, p, 0);
642 int btrfs_leaf_free_space(const struct extent_buffer *leaf);
644 static inline int is_fstree(u64 rootid)
646 if (rootid == BTRFS_FS_TREE_OBJECTID ||
647 ((s64)rootid >= (s64)BTRFS_FIRST_FREE_OBJECTID &&
648 !btrfs_qgroup_level(rootid)))
653 static inline bool btrfs_is_data_reloc_root(const struct btrfs_root *root)
655 return root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID;
658 u16 btrfs_csum_type_size(u16 type);
659 int btrfs_super_csum_size(const struct btrfs_super_block *s);
660 const char *btrfs_super_csum_name(u16 csum_type);
661 const char *btrfs_super_csum_driver(u16 csum_type);
662 size_t __attribute_const__ btrfs_get_num_csums(void);
665 * We use page status Private2 to indicate there is an ordered extent with
668 * Rename the Private2 accessors to Ordered, to improve readability.
670 #define PageOrdered(page) PagePrivate2(page)
671 #define SetPageOrdered(page) SetPagePrivate2(page)
672 #define ClearPageOrdered(page) ClearPagePrivate2(page)
673 #define folio_test_ordered(folio) folio_test_private_2(folio)
674 #define folio_set_ordered(folio) folio_set_private_2(folio)
675 #define folio_clear_ordered(folio) folio_clear_private_2(folio)