xref: /openbmc/linux/fs/btrfs/ctree.h (revision d5e09e38)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #ifndef BTRFS_CTREE_H
7 #define BTRFS_CTREE_H
8 
9 #include <linux/mm.h>
10 #include <linux/sched/signal.h>
11 #include <linux/highmem.h>
12 #include <linux/fs.h>
13 #include <linux/rwsem.h>
14 #include <linux/semaphore.h>
15 #include <linux/completion.h>
16 #include <linux/backing-dev.h>
17 #include <linux/wait.h>
18 #include <linux/slab.h>
19 #include <trace/events/btrfs.h>
20 #include <asm/unaligned.h>
21 #include <linux/pagemap.h>
22 #include <linux/btrfs.h>
23 #include <linux/btrfs_tree.h>
24 #include <linux/workqueue.h>
25 #include <linux/security.h>
26 #include <linux/sizes.h>
27 #include <linux/dynamic_debug.h>
28 #include <linux/refcount.h>
29 #include <linux/crc32c.h>
30 #include <linux/iomap.h>
31 #include <linux/fscrypt.h>
32 #include "extent-io-tree.h"
33 #include "extent_io.h"
34 #include "extent_map.h"
35 #include "async-thread.h"
36 #include "block-rsv.h"
37 #include "locking.h"
38 #include "misc.h"
39 #include "fs.h"
40 
41 struct btrfs_trans_handle;
42 struct btrfs_transaction;
43 struct btrfs_pending_snapshot;
44 struct btrfs_delayed_ref_root;
45 struct btrfs_space_info;
46 struct btrfs_block_group;
47 struct btrfs_ordered_sum;
48 struct btrfs_ref;
49 struct btrfs_bio;
50 struct btrfs_ioctl_encoded_io_args;
51 struct btrfs_device;
52 struct btrfs_fs_devices;
53 struct btrfs_balance_control;
54 struct btrfs_delayed_root;
55 struct reloc_control;
56 
57 /* Read ahead values for struct btrfs_path.reada */
58 enum {
59 	READA_NONE,
60 	READA_BACK,
61 	READA_FORWARD,
62 	/*
63 	 * Similar to READA_FORWARD but unlike it:
64 	 *
65 	 * 1) It will trigger readahead even for leaves that are not close to
66 	 *    each other on disk;
67 	 * 2) It also triggers readahead for nodes;
68 	 * 3) During a search, even when a node or leaf is already in memory, it
69 	 *    will still trigger readahead for other nodes and leaves that follow
70 	 *    it.
71 	 *
72 	 * This is meant to be used only when we know we are iterating over the
73 	 * entire tree or a very large part of it.
74 	 */
75 	READA_FORWARD_ALWAYS,
76 };
77 
78 /*
79  * btrfs_paths remember the path taken from the root down to the leaf.
80  * level 0 is always the leaf, and nodes[1...BTRFS_MAX_LEVEL] will point
81  * to any other levels that are present.
82  *
83  * The slots array records the index of the item or block pointer
84  * used while walking the tree.
85  */
86 struct btrfs_path {
87 	struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
88 	int slots[BTRFS_MAX_LEVEL];
89 	/* if there is real range locking, this locks field will change */
90 	u8 locks[BTRFS_MAX_LEVEL];
91 	u8 reada;
92 	/* keep some upper locks as we walk down */
93 	u8 lowest_level;
94 
95 	/*
96 	 * set by btrfs_split_item, tells search_slot to keep all locks
97 	 * and to force calls to keep space in the nodes
98 	 */
99 	unsigned int search_for_split:1;
100 	unsigned int keep_locks:1;
101 	unsigned int skip_locking:1;
102 	unsigned int search_commit_root:1;
103 	unsigned int need_commit_sem:1;
104 	unsigned int skip_release_on_error:1;
105 	/*
106 	 * Indicate that new item (btrfs_search_slot) is extending already
107 	 * existing item and ins_len contains only the data size and not item
108 	 * header (ie. sizeof(struct btrfs_item) is not included).
109 	 */
110 	unsigned int search_for_extension:1;
111 	/* Stop search if any locks need to be taken (for read) */
112 	unsigned int nowait:1;
113 };
114 
115 /*
116  * The state of btrfs root
117  */
118 enum {
119 	/*
120 	 * btrfs_record_root_in_trans is a multi-step process, and it can race
121 	 * with the balancing code.   But the race is very small, and only the
122 	 * first time the root is added to each transaction.  So IN_TRANS_SETUP
123 	 * is used to tell us when more checks are required
124 	 */
125 	BTRFS_ROOT_IN_TRANS_SETUP,
126 
127 	/*
128 	 * Set if tree blocks of this root can be shared by other roots.
129 	 * Only subvolume trees and their reloc trees have this bit set.
130 	 * Conflicts with TRACK_DIRTY bit.
131 	 *
132 	 * This affects two things:
133 	 *
134 	 * - How balance works
135 	 *   For shareable roots, we need to use reloc tree and do path
136 	 *   replacement for balance, and need various pre/post hooks for
137 	 *   snapshot creation to handle them.
138 	 *
139 	 *   While for non-shareable trees, we just simply do a tree search
140 	 *   with COW.
141 	 *
142 	 * - How dirty roots are tracked
143 	 *   For shareable roots, btrfs_record_root_in_trans() is needed to
144 	 *   track them, while non-subvolume roots have TRACK_DIRTY bit, they
145 	 *   don't need to set this manually.
146 	 */
147 	BTRFS_ROOT_SHAREABLE,
148 	BTRFS_ROOT_TRACK_DIRTY,
149 	BTRFS_ROOT_IN_RADIX,
150 	BTRFS_ROOT_ORPHAN_ITEM_INSERTED,
151 	BTRFS_ROOT_DEFRAG_RUNNING,
152 	BTRFS_ROOT_FORCE_COW,
153 	BTRFS_ROOT_MULTI_LOG_TASKS,
154 	BTRFS_ROOT_DIRTY,
155 	BTRFS_ROOT_DELETING,
156 
157 	/*
158 	 * Reloc tree is orphan, only kept here for qgroup delayed subtree scan
159 	 *
160 	 * Set for the subvolume tree owning the reloc tree.
161 	 */
162 	BTRFS_ROOT_DEAD_RELOC_TREE,
163 	/* Mark dead root stored on device whose cleanup needs to be resumed */
164 	BTRFS_ROOT_DEAD_TREE,
165 	/* The root has a log tree. Used for subvolume roots and the tree root. */
166 	BTRFS_ROOT_HAS_LOG_TREE,
167 	/* Qgroup flushing is in progress */
168 	BTRFS_ROOT_QGROUP_FLUSHING,
169 	/* We started the orphan cleanup for this root. */
170 	BTRFS_ROOT_ORPHAN_CLEANUP,
171 	/* This root has a drop operation that was started previously. */
172 	BTRFS_ROOT_UNFINISHED_DROP,
173 	/* This reloc root needs to have its buffers lockdep class reset. */
174 	BTRFS_ROOT_RESET_LOCKDEP_CLASS,
175 };
176 
177 /*
178  * Record swapped tree blocks of a subvolume tree for delayed subtree trace
179  * code. For detail check comment in fs/btrfs/qgroup.c.
180  */
181 struct btrfs_qgroup_swapped_blocks {
182 	spinlock_t lock;
183 	/* RM_EMPTY_ROOT() of above blocks[] */
184 	bool swapped;
185 	struct rb_root blocks[BTRFS_MAX_LEVEL];
186 };
187 
188 /*
189  * in ram representation of the tree.  extent_root is used for all allocations
190  * and for the extent tree extent_root root.
191  */
192 struct btrfs_root {
193 	struct rb_node rb_node;
194 
195 	struct extent_buffer *node;
196 
197 	struct extent_buffer *commit_root;
198 	struct btrfs_root *log_root;
199 	struct btrfs_root *reloc_root;
200 
201 	unsigned long state;
202 	struct btrfs_root_item root_item;
203 	struct btrfs_key root_key;
204 	struct btrfs_fs_info *fs_info;
205 	struct extent_io_tree dirty_log_pages;
206 
207 	struct mutex objectid_mutex;
208 
209 	spinlock_t accounting_lock;
210 	struct btrfs_block_rsv *block_rsv;
211 
212 	struct mutex log_mutex;
213 	wait_queue_head_t log_writer_wait;
214 	wait_queue_head_t log_commit_wait[2];
215 	struct list_head log_ctxs[2];
216 	/* Used only for log trees of subvolumes, not for the log root tree */
217 	atomic_t log_writers;
218 	atomic_t log_commit[2];
219 	/* Used only for log trees of subvolumes, not for the log root tree */
220 	atomic_t log_batch;
221 	int log_transid;
222 	/* No matter the commit succeeds or not*/
223 	int log_transid_committed;
224 	/* Just be updated when the commit succeeds. */
225 	int last_log_commit;
226 	pid_t log_start_pid;
227 
228 	u64 last_trans;
229 
230 	u32 type;
231 
232 	u64 free_objectid;
233 
234 	struct btrfs_key defrag_progress;
235 	struct btrfs_key defrag_max;
236 
237 	/* The dirty list is only used by non-shareable roots */
238 	struct list_head dirty_list;
239 
240 	struct list_head root_list;
241 
242 	spinlock_t log_extents_lock[2];
243 	struct list_head logged_list[2];
244 
245 	spinlock_t inode_lock;
246 	/* red-black tree that keeps track of in-memory inodes */
247 	struct rb_root inode_tree;
248 
249 	/*
250 	 * radix tree that keeps track of delayed nodes of every inode,
251 	 * protected by inode_lock
252 	 */
253 	struct radix_tree_root delayed_nodes_tree;
254 	/*
255 	 * right now this just gets used so that a root has its own devid
256 	 * for stat.  It may be used for more later
257 	 */
258 	dev_t anon_dev;
259 
260 	spinlock_t root_item_lock;
261 	refcount_t refs;
262 
263 	struct mutex delalloc_mutex;
264 	spinlock_t delalloc_lock;
265 	/*
266 	 * all of the inodes that have delalloc bytes.  It is possible for
267 	 * this list to be empty even when there is still dirty data=ordered
268 	 * extents waiting to finish IO.
269 	 */
270 	struct list_head delalloc_inodes;
271 	struct list_head delalloc_root;
272 	u64 nr_delalloc_inodes;
273 
274 	struct mutex ordered_extent_mutex;
275 	/*
276 	 * this is used by the balancing code to wait for all the pending
277 	 * ordered extents
278 	 */
279 	spinlock_t ordered_extent_lock;
280 
281 	/*
282 	 * all of the data=ordered extents pending writeback
283 	 * these can span multiple transactions and basically include
284 	 * every dirty data page that isn't from nodatacow
285 	 */
286 	struct list_head ordered_extents;
287 	struct list_head ordered_root;
288 	u64 nr_ordered_extents;
289 
290 	/*
291 	 * Not empty if this subvolume root has gone through tree block swap
292 	 * (relocation)
293 	 *
294 	 * Will be used by reloc_control::dirty_subvol_roots.
295 	 */
296 	struct list_head reloc_dirty_list;
297 
298 	/*
299 	 * Number of currently running SEND ioctls to prevent
300 	 * manipulation with the read-only status via SUBVOL_SETFLAGS
301 	 */
302 	int send_in_progress;
303 	/*
304 	 * Number of currently running deduplication operations that have a
305 	 * destination inode belonging to this root. Protected by the lock
306 	 * root_item_lock.
307 	 */
308 	int dedupe_in_progress;
309 	/* For exclusion of snapshot creation and nocow writes */
310 	struct btrfs_drew_lock snapshot_lock;
311 
312 	atomic_t snapshot_force_cow;
313 
314 	/* For qgroup metadata reserved space */
315 	spinlock_t qgroup_meta_rsv_lock;
316 	u64 qgroup_meta_rsv_pertrans;
317 	u64 qgroup_meta_rsv_prealloc;
318 	wait_queue_head_t qgroup_flush_wait;
319 
320 	/* Number of active swapfiles */
321 	atomic_t nr_swapfiles;
322 
323 	/* Record pairs of swapped blocks for qgroup */
324 	struct btrfs_qgroup_swapped_blocks swapped_blocks;
325 
326 	/* Used only by log trees, when logging csum items */
327 	struct extent_io_tree log_csum_range;
328 
329 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
330 	u64 alloc_bytenr;
331 #endif
332 
333 #ifdef CONFIG_BTRFS_DEBUG
334 	struct list_head leak_list;
335 #endif
336 };
337 
btrfs_root_readonly(const struct btrfs_root * root)338 static inline bool btrfs_root_readonly(const struct btrfs_root *root)
339 {
340 	/* Byte-swap the constant at compile time, root_item::flags is LE */
341 	return (root->root_item.flags & cpu_to_le64(BTRFS_ROOT_SUBVOL_RDONLY)) != 0;
342 }
343 
btrfs_root_dead(const struct btrfs_root * root)344 static inline bool btrfs_root_dead(const struct btrfs_root *root)
345 {
346 	/* Byte-swap the constant at compile time, root_item::flags is LE */
347 	return (root->root_item.flags & cpu_to_le64(BTRFS_ROOT_SUBVOL_DEAD)) != 0;
348 }
349 
btrfs_root_id(const struct btrfs_root * root)350 static inline u64 btrfs_root_id(const struct btrfs_root *root)
351 {
352 	return root->root_key.objectid;
353 }
354 
355 /*
356  * Structure that conveys information about an extent that is going to replace
357  * all the extents in a file range.
358  */
359 struct btrfs_replace_extent_info {
360 	u64 disk_offset;
361 	u64 disk_len;
362 	u64 data_offset;
363 	u64 data_len;
364 	u64 file_offset;
365 	/* Pointer to a file extent item of type regular or prealloc. */
366 	char *extent_buf;
367 	/*
368 	 * Set to true when attempting to replace a file range with a new extent
369 	 * described by this structure, set to false when attempting to clone an
370 	 * existing extent into a file range.
371 	 */
372 	bool is_new_extent;
373 	/* Indicate if we should update the inode's mtime and ctime. */
374 	bool update_times;
375 	/* Meaningful only if is_new_extent is true. */
376 	int qgroup_reserved;
377 	/*
378 	 * Meaningful only if is_new_extent is true.
379 	 * Used to track how many extent items we have already inserted in a
380 	 * subvolume tree that refer to the extent described by this structure,
381 	 * so that we know when to create a new delayed ref or update an existing
382 	 * one.
383 	 */
384 	int insertions;
385 };
386 
387 /* Arguments for btrfs_drop_extents() */
388 struct btrfs_drop_extents_args {
389 	/* Input parameters */
390 
391 	/*
392 	 * If NULL, btrfs_drop_extents() will allocate and free its own path.
393 	 * If 'replace_extent' is true, this must not be NULL. Also the path
394 	 * is always released except if 'replace_extent' is true and
395 	 * btrfs_drop_extents() sets 'extent_inserted' to true, in which case
396 	 * the path is kept locked.
397 	 */
398 	struct btrfs_path *path;
399 	/* Start offset of the range to drop extents from */
400 	u64 start;
401 	/* End (exclusive, last byte + 1) of the range to drop extents from */
402 	u64 end;
403 	/* If true drop all the extent maps in the range */
404 	bool drop_cache;
405 	/*
406 	 * If true it means we want to insert a new extent after dropping all
407 	 * the extents in the range. If this is true, the 'extent_item_size'
408 	 * parameter must be set as well and the 'extent_inserted' field will
409 	 * be set to true by btrfs_drop_extents() if it could insert the new
410 	 * extent.
411 	 * Note: when this is set to true the path must not be NULL.
412 	 */
413 	bool replace_extent;
414 	/*
415 	 * Used if 'replace_extent' is true. Size of the file extent item to
416 	 * insert after dropping all existing extents in the range
417 	 */
418 	u32 extent_item_size;
419 
420 	/* Output parameters */
421 
422 	/*
423 	 * Set to the minimum between the input parameter 'end' and the end
424 	 * (exclusive, last byte + 1) of the last dropped extent. This is always
425 	 * set even if btrfs_drop_extents() returns an error.
426 	 */
427 	u64 drop_end;
428 	/*
429 	 * The number of allocated bytes found in the range. This can be smaller
430 	 * than the range's length when there are holes in the range.
431 	 */
432 	u64 bytes_found;
433 	/*
434 	 * Only set if 'replace_extent' is true. Set to true if we were able
435 	 * to insert a replacement extent after dropping all extents in the
436 	 * range, otherwise set to false by btrfs_drop_extents().
437 	 * Also, if btrfs_drop_extents() has set this to true it means it
438 	 * returned with the path locked, otherwise if it has set this to
439 	 * false it has returned with the path released.
440 	 */
441 	bool extent_inserted;
442 };
443 
444 struct btrfs_file_private {
445 	void *filldir_buf;
446 	u64 last_index;
447 	struct extent_state *llseek_cached_state;
448 };
449 
BTRFS_LEAF_DATA_SIZE(const struct btrfs_fs_info * info)450 static inline u32 BTRFS_LEAF_DATA_SIZE(const struct btrfs_fs_info *info)
451 {
452 	return info->nodesize - sizeof(struct btrfs_header);
453 }
454 
BTRFS_MAX_ITEM_SIZE(const struct btrfs_fs_info * info)455 static inline u32 BTRFS_MAX_ITEM_SIZE(const struct btrfs_fs_info *info)
456 {
457 	return BTRFS_LEAF_DATA_SIZE(info) - sizeof(struct btrfs_item);
458 }
459 
BTRFS_NODEPTRS_PER_BLOCK(const struct btrfs_fs_info * info)460 static inline u32 BTRFS_NODEPTRS_PER_BLOCK(const struct btrfs_fs_info *info)
461 {
462 	return BTRFS_LEAF_DATA_SIZE(info) / sizeof(struct btrfs_key_ptr);
463 }
464 
BTRFS_MAX_XATTR_SIZE(const struct btrfs_fs_info * info)465 static inline u32 BTRFS_MAX_XATTR_SIZE(const struct btrfs_fs_info *info)
466 {
467 	return BTRFS_MAX_ITEM_SIZE(info) - sizeof(struct btrfs_dir_item);
468 }
469 
470 #define BTRFS_BYTES_TO_BLKS(fs_info, bytes) \
471 				((bytes) >> (fs_info)->sectorsize_bits)
472 
btrfs_crc32c(u32 crc,const void * address,unsigned length)473 static inline u32 btrfs_crc32c(u32 crc, const void *address, unsigned length)
474 {
475 	return crc32c(crc, address, length);
476 }
477 
btrfs_crc32c_final(u32 crc,u8 * result)478 static inline void btrfs_crc32c_final(u32 crc, u8 *result)
479 {
480 	put_unaligned_le32(~crc, result);
481 }
482 
btrfs_name_hash(const char * name,int len)483 static inline u64 btrfs_name_hash(const char *name, int len)
484 {
485        return crc32c((u32)~1, name, len);
486 }
487 
488 /*
489  * Figure the key offset of an extended inode ref
490  */
btrfs_extref_hash(u64 parent_objectid,const char * name,int len)491 static inline u64 btrfs_extref_hash(u64 parent_objectid, const char *name,
492                                    int len)
493 {
494        return (u64) crc32c(parent_objectid, name, len);
495 }
496 
btrfs_alloc_write_mask(struct address_space * mapping)497 static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping)
498 {
499 	return mapping_gfp_constraint(mapping, ~__GFP_FS);
500 }
501 
502 int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
503 				   u64 start, u64 end);
504 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
505 			 u64 num_bytes, u64 *actual_bytes);
506 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range);
507 
508 /* ctree.c */
509 int __init btrfs_ctree_init(void);
510 void __cold btrfs_ctree_exit(void);
511 
512 int btrfs_bin_search(struct extent_buffer *eb, int first_slot,
513 		     const struct btrfs_key *key, int *slot);
514 
515 int __pure btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2);
516 int btrfs_previous_item(struct btrfs_root *root,
517 			struct btrfs_path *path, u64 min_objectid,
518 			int type);
519 int btrfs_previous_extent_item(struct btrfs_root *root,
520 			struct btrfs_path *path, u64 min_objectid);
521 void btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
522 			     struct btrfs_path *path,
523 			     const struct btrfs_key *new_key);
524 struct extent_buffer *btrfs_root_node(struct btrfs_root *root);
525 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
526 			struct btrfs_key *key, int lowest_level,
527 			u64 min_trans);
528 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
529 			 struct btrfs_path *path,
530 			 u64 min_trans);
531 struct extent_buffer *btrfs_read_node_slot(struct extent_buffer *parent,
532 					   int slot);
533 
534 int btrfs_cow_block(struct btrfs_trans_handle *trans,
535 		    struct btrfs_root *root, struct extent_buffer *buf,
536 		    struct extent_buffer *parent, int parent_slot,
537 		    struct extent_buffer **cow_ret,
538 		    enum btrfs_lock_nesting nest);
539 int btrfs_copy_root(struct btrfs_trans_handle *trans,
540 		      struct btrfs_root *root,
541 		      struct extent_buffer *buf,
542 		      struct extent_buffer **cow_ret, u64 new_root_objectid);
543 int btrfs_block_can_be_shared(struct btrfs_trans_handle *trans,
544 			      struct btrfs_root *root,
545 			      struct extent_buffer *buf);
546 int btrfs_del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
547 		  struct btrfs_path *path, int level, int slot);
548 void btrfs_extend_item(struct btrfs_trans_handle *trans,
549 		       struct btrfs_path *path, u32 data_size);
550 void btrfs_truncate_item(struct btrfs_trans_handle *trans,
551 			 struct btrfs_path *path, u32 new_size, int from_end);
552 int btrfs_split_item(struct btrfs_trans_handle *trans,
553 		     struct btrfs_root *root,
554 		     struct btrfs_path *path,
555 		     const struct btrfs_key *new_key,
556 		     unsigned long split_offset);
557 int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
558 			 struct btrfs_root *root,
559 			 struct btrfs_path *path,
560 			 const struct btrfs_key *new_key);
561 int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path,
562 		u64 inum, u64 ioff, u8 key_type, struct btrfs_key *found_key);
563 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
564 		      const struct btrfs_key *key, struct btrfs_path *p,
565 		      int ins_len, int cow);
566 int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key,
567 			  struct btrfs_path *p, u64 time_seq);
568 int btrfs_search_slot_for_read(struct btrfs_root *root,
569 			       const struct btrfs_key *key,
570 			       struct btrfs_path *p, int find_higher,
571 			       int return_any);
572 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
573 		       struct btrfs_root *root, struct extent_buffer *parent,
574 		       int start_slot, u64 *last_ret,
575 		       struct btrfs_key *progress);
576 void btrfs_release_path(struct btrfs_path *p);
577 struct btrfs_path *btrfs_alloc_path(void);
578 void btrfs_free_path(struct btrfs_path *p);
579 
580 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
581 		   struct btrfs_path *path, int slot, int nr);
btrfs_del_item(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_path * path)582 static inline int btrfs_del_item(struct btrfs_trans_handle *trans,
583 				 struct btrfs_root *root,
584 				 struct btrfs_path *path)
585 {
586 	return btrfs_del_items(trans, root, path, path->slots[0], 1);
587 }
588 
589 /*
590  * Describes a batch of items to insert in a btree. This is used by
591  * btrfs_insert_empty_items().
592  */
593 struct btrfs_item_batch {
594 	/*
595 	 * Pointer to an array containing the keys of the items to insert (in
596 	 * sorted order).
597 	 */
598 	const struct btrfs_key *keys;
599 	/* Pointer to an array containing the data size for each item to insert. */
600 	const u32 *data_sizes;
601 	/*
602 	 * The sum of data sizes for all items. The caller can compute this while
603 	 * setting up the data_sizes array, so it ends up being more efficient
604 	 * than having btrfs_insert_empty_items() or setup_item_for_insert()
605 	 * doing it, as it would avoid an extra loop over a potentially large
606 	 * array, and in the case of setup_item_for_insert(), we would be doing
607 	 * it while holding a write lock on a leaf and often on upper level nodes
608 	 * too, unnecessarily increasing the size of a critical section.
609 	 */
610 	u32 total_data_size;
611 	/* Size of the keys and data_sizes arrays (number of items in the batch). */
612 	int nr;
613 };
614 
615 void btrfs_setup_item_for_insert(struct btrfs_trans_handle *trans,
616 				 struct btrfs_root *root,
617 				 struct btrfs_path *path,
618 				 const struct btrfs_key *key,
619 				 u32 data_size);
620 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
621 		      const struct btrfs_key *key, void *data, u32 data_size);
622 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
623 			     struct btrfs_root *root,
624 			     struct btrfs_path *path,
625 			     const struct btrfs_item_batch *batch);
626 
btrfs_insert_empty_item(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_path * path,const struct btrfs_key * key,u32 data_size)627 static inline int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
628 					  struct btrfs_root *root,
629 					  struct btrfs_path *path,
630 					  const struct btrfs_key *key,
631 					  u32 data_size)
632 {
633 	struct btrfs_item_batch batch;
634 
635 	batch.keys = key;
636 	batch.data_sizes = &data_size;
637 	batch.total_data_size = data_size;
638 	batch.nr = 1;
639 
640 	return btrfs_insert_empty_items(trans, root, path, &batch);
641 }
642 
643 int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
644 			u64 time_seq);
645 
646 int btrfs_search_backwards(struct btrfs_root *root, struct btrfs_key *key,
647 			   struct btrfs_path *path);
648 
649 int btrfs_get_next_valid_item(struct btrfs_root *root, struct btrfs_key *key,
650 			      struct btrfs_path *path);
651 
652 /*
653  * Search in @root for a given @key, and store the slot found in @found_key.
654  *
655  * @root:	The root node of the tree.
656  * @key:	The key we are looking for.
657  * @found_key:	Will hold the found item.
658  * @path:	Holds the current slot/leaf.
659  * @iter_ret:	Contains the value returned from btrfs_search_slot or
660  * 		btrfs_get_next_valid_item, whichever was executed last.
661  *
662  * The @iter_ret is an output variable that will contain the return value of
663  * btrfs_search_slot, if it encountered an error, or the value returned from
664  * btrfs_get_next_valid_item otherwise. That return value can be 0, if a valid
665  * slot was found, 1 if there were no more leaves, and <0 if there was an error.
666  *
667  * It's recommended to use a separate variable for iter_ret and then use it to
668  * set the function return value so there's no confusion of the 0/1/errno
669  * values stemming from btrfs_search_slot.
670  */
671 #define btrfs_for_each_slot(root, key, found_key, path, iter_ret)		\
672 	for (iter_ret = btrfs_search_slot(NULL, (root), (key), (path), 0, 0);	\
673 		(iter_ret) >= 0 &&						\
674 		(iter_ret = btrfs_get_next_valid_item((root), (found_key), (path))) == 0; \
675 		(path)->slots[0]++						\
676 	)
677 
678 int btrfs_next_old_item(struct btrfs_root *root, struct btrfs_path *path, u64 time_seq);
679 
680 /*
681  * Search the tree again to find a leaf with greater keys.
682  *
683  * Returns 0 if it found something or 1 if there are no greater leaves.
684  * Returns < 0 on error.
685  */
btrfs_next_leaf(struct btrfs_root * root,struct btrfs_path * path)686 static inline int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
687 {
688 	return btrfs_next_old_leaf(root, path, 0);
689 }
690 
btrfs_next_item(struct btrfs_root * root,struct btrfs_path * p)691 static inline int btrfs_next_item(struct btrfs_root *root, struct btrfs_path *p)
692 {
693 	return btrfs_next_old_item(root, p, 0);
694 }
695 int btrfs_leaf_free_space(const struct extent_buffer *leaf);
696 
is_fstree(u64 rootid)697 static inline int is_fstree(u64 rootid)
698 {
699 	if (rootid == BTRFS_FS_TREE_OBJECTID ||
700 	    ((s64)rootid >= (s64)BTRFS_FIRST_FREE_OBJECTID &&
701 	      !btrfs_qgroup_level(rootid)))
702 		return 1;
703 	return 0;
704 }
705 
btrfs_is_data_reloc_root(const struct btrfs_root * root)706 static inline bool btrfs_is_data_reloc_root(const struct btrfs_root *root)
707 {
708 	return root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID;
709 }
710 
711 u16 btrfs_csum_type_size(u16 type);
712 int btrfs_super_csum_size(const struct btrfs_super_block *s);
713 const char *btrfs_super_csum_name(u16 csum_type);
714 const char *btrfs_super_csum_driver(u16 csum_type);
715 size_t __attribute_const__ btrfs_get_num_csums(void);
716 
717 /*
718  * We use page status Private2 to indicate there is an ordered extent with
719  * unfinished IO.
720  *
721  * Rename the Private2 accessors to Ordered, to improve readability.
722  */
723 #define PageOrdered(page)		PagePrivate2(page)
724 #define SetPageOrdered(page)		SetPagePrivate2(page)
725 #define ClearPageOrdered(page)		ClearPagePrivate2(page)
726 #define folio_test_ordered(folio)	folio_test_private_2(folio)
727 #define folio_set_ordered(folio)	folio_set_private_2(folio)
728 #define folio_clear_ordered(folio)	folio_clear_private_2(folio)
729 
730 #endif
731