1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2008 Oracle. All rights reserved.
6 #ifndef BTRFS_DELAYED_REF_H
7 #define BTRFS_DELAYED_REF_H
9 #include <linux/types.h>
10 #include <linux/refcount.h>
11 #include <linux/list.h>
12 #include <linux/rbtree.h>
13 #include <linux/mutex.h>
14 #include <linux/spinlock.h>
15 #include <linux/slab.h>
16 #include <uapi/linux/btrfs_tree.h>
18 struct btrfs_trans_handle;
21 /* these are the possible values of struct btrfs_delayed_ref_node->action */
22 enum btrfs_delayed_ref_action {
23 /* Add one backref to the tree */
24 BTRFS_ADD_DELAYED_REF = 1,
25 /* Delete one backref from the tree */
26 BTRFS_DROP_DELAYED_REF,
27 /* Record a full extent allocation */
28 BTRFS_ADD_DELAYED_EXTENT,
29 /* Not changing ref count on head ref */
30 BTRFS_UPDATE_DELAYED_HEAD,
33 struct btrfs_data_ref {
34 /* For EXTENT_DATA_REF */
36 /* Inode which refers to this data extent */
40 * file_offset - extent_offset
42 * file_offset is the key.offset of the EXTENT_DATA key.
43 * extent_offset is btrfs_file_extent_offset() of the EXTENT_DATA data.
48 struct btrfs_tree_ref {
50 * Level of this tree block.
52 * Shared for skinny (TREE_BLOCK_REF) and normal tree ref.
56 /* For non-skinny metadata, no special member needed */
59 struct btrfs_delayed_ref_node {
60 struct rb_node ref_node;
62 * If action is BTRFS_ADD_DELAYED_REF, also link this node to
63 * ref_head->ref_add_list, then we do not need to iterate the
64 * whole ref_head->ref_list to find BTRFS_ADD_DELAYED_REF nodes.
66 struct list_head add_list;
68 /* the starting bytenr of the extent */
71 /* the size of the extent */
74 /* seq number to keep track of insertion order */
77 /* The ref_root for this ref */
81 * The parent for this ref, if this isn't set the ref_root is the
86 /* ref count on this data structure */
90 * how many refs is this entry adding or deleting. For
91 * head refs, this may be a negative number because it is keeping
92 * track of the total mods done to the reference count.
93 * For individual refs, this will always be a positive number
95 * It may be more than one, since it is possible for a single
96 * parent to have more than one ref on an extent
100 unsigned int action:8;
104 struct btrfs_tree_ref tree_ref;
105 struct btrfs_data_ref data_ref;
109 struct btrfs_delayed_extent_op {
110 struct btrfs_disk_key key;
118 * the head refs are used to hold a lock on a given extent, which allows us
119 * to make sure that only one process is running the delayed refs
120 * at a time for a single extent. They also store the sum of all the
121 * reference count modifications we've queued up.
123 struct btrfs_delayed_ref_head {
127 * For insertion into struct btrfs_delayed_ref_root::href_root.
128 * Keep it in the same cache line as 'bytenr' for more efficient
129 * searches in the rbtree.
131 struct rb_node href_node;
133 * the mutex is held while running the refs, and it is also
134 * held when checking the sum of reference modifications.
140 /* Protects 'ref_tree' and 'ref_add_list'. */
142 struct rb_root_cached ref_tree;
143 /* accumulate add BTRFS_ADD_DELAYED_REF nodes to this ref_add_list. */
144 struct list_head ref_add_list;
146 struct btrfs_delayed_extent_op *extent_op;
149 * This is used to track the final ref_mod from all the refs associated
150 * with this head ref, this is not adjusted as delayed refs are run,
151 * this is meant to track if we need to do the csum accounting or not.
156 * This is the current outstanding mod references for this bytenr. This
157 * is used with lookup_extent_info to get an accurate reference count
158 * for a bytenr, so it is adjusted as delayed refs are run so that any
159 * on disk reference count + ref_mod is accurate.
164 * The root that triggered the allocation when must_insert_reserved is
170 * Track reserved bytes when setting must_insert_reserved. On success
171 * or cleanup, we will need to free the reservation.
176 * when a new extent is allocated, it is just reserved in memory
177 * The actual extent isn't inserted into the extent allocation tree
178 * until the delayed ref is processed. must_insert_reserved is
179 * used to flag a delayed ref so the accounting can be updated
180 * when a full insert is done.
182 * It is possible the extent will be freed before it is ever
183 * inserted into the extent allocation tree. In this case
184 * we need to update the in ram accounting to properly reflect
185 * the free has happened.
187 bool must_insert_reserved;
194 enum btrfs_delayed_ref_flags {
195 /* Indicate that we are flushing delayed refs for the commit */
196 BTRFS_DELAYED_REFS_FLUSHING,
199 struct btrfs_delayed_ref_root {
200 /* head ref rbtree */
201 struct rb_root_cached href_root;
203 /* dirty extent records */
204 struct rb_root dirty_extent_root;
206 /* this spin lock protects the rbtree and the entries inside */
209 /* how many delayed ref updates we've queued, used by the
212 atomic_t num_entries;
214 /* total number of head nodes in tree */
215 unsigned long num_heads;
217 /* total number of head nodes ready for processing */
218 unsigned long num_heads_ready;
224 u64 run_delayed_start;
227 * To make qgroup to skip given root.
228 * This is for snapshot, as btrfs_qgroup_inherit() will manually
229 * modify counters for snapshot and its source, so we should skip
230 * the snapshot in new_root/old_roots or it will get calculated twice
235 enum btrfs_ref_type {
243 enum btrfs_ref_type type;
244 enum btrfs_delayed_ref_action action;
247 * Whether this extent should go through qgroup record.
249 * Normally false, but for certain cases like delayed subtree scan,
250 * setting this flag can hugely reduce qgroup overhead.
254 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
255 /* Through which root is this modification. */
263 * The root that owns the reference for this reference, this will be set
264 * or ->parent will be set, depending on what type of reference this is.
268 /* Bytenr of the parent tree block */
271 struct btrfs_data_ref data_ref;
272 struct btrfs_tree_ref tree_ref;
276 extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
277 extern struct kmem_cache *btrfs_delayed_ref_node_cachep;
278 extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
280 int __init btrfs_delayed_ref_init(void);
281 void __cold btrfs_delayed_ref_exit(void);
283 static inline u64 btrfs_calc_delayed_ref_bytes(const struct btrfs_fs_info *fs_info,
284 int num_delayed_refs)
288 num_bytes = btrfs_calc_insert_metadata_size(fs_info, num_delayed_refs);
291 * We have to check the mount option here because we could be enabling
292 * the free space tree for the first time and don't have the compat_ro
295 * We need extra reservations if we have the free space tree because
296 * we'll have to modify that tree as well.
298 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
304 static inline u64 btrfs_calc_delayed_ref_csum_bytes(const struct btrfs_fs_info *fs_info,
308 * Deleting csum items does not result in new nodes/leaves and does not
309 * require changing the free space tree, only the csum tree, so this is
312 return btrfs_calc_metadata_size(fs_info, num_csum_items);
315 void btrfs_init_tree_ref(struct btrfs_ref *generic_ref, int level, u64 mod_root,
317 void btrfs_init_data_ref(struct btrfs_ref *generic_ref, u64 ino, u64 offset,
318 u64 mod_root, bool skip_qgroup);
320 static inline struct btrfs_delayed_extent_op *
321 btrfs_alloc_delayed_extent_op(void)
323 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
327 btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
330 kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
333 void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref);
335 static inline u64 btrfs_ref_head_to_space_flags(
336 struct btrfs_delayed_ref_head *head_ref)
338 if (head_ref->is_data)
339 return BTRFS_BLOCK_GROUP_DATA;
340 else if (head_ref->is_system)
341 return BTRFS_BLOCK_GROUP_SYSTEM;
342 return BTRFS_BLOCK_GROUP_METADATA;
345 static inline void btrfs_put_delayed_ref_head(struct btrfs_delayed_ref_head *head)
347 if (refcount_dec_and_test(&head->refs))
348 kmem_cache_free(btrfs_delayed_ref_head_cachep, head);
351 int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
352 struct btrfs_ref *generic_ref,
353 struct btrfs_delayed_extent_op *extent_op);
354 int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
355 struct btrfs_ref *generic_ref,
357 int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
358 u64 bytenr, u64 num_bytes,
359 struct btrfs_delayed_extent_op *extent_op);
360 void btrfs_merge_delayed_refs(struct btrfs_fs_info *fs_info,
361 struct btrfs_delayed_ref_root *delayed_refs,
362 struct btrfs_delayed_ref_head *head);
364 struct btrfs_delayed_ref_head *
365 btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
367 int btrfs_delayed_ref_lock(struct btrfs_delayed_ref_root *delayed_refs,
368 struct btrfs_delayed_ref_head *head);
369 static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
371 mutex_unlock(&head->mutex);
373 void btrfs_delete_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
374 struct btrfs_delayed_ref_head *head);
376 struct btrfs_delayed_ref_head *btrfs_select_ref_head(
377 struct btrfs_delayed_ref_root *delayed_refs);
379 int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq);
381 void btrfs_delayed_refs_rsv_release(struct btrfs_fs_info *fs_info, int nr_refs, int nr_csums);
382 void btrfs_update_delayed_refs_rsv(struct btrfs_trans_handle *trans);
383 void btrfs_inc_delayed_refs_rsv_bg_inserts(struct btrfs_fs_info *fs_info);
384 void btrfs_dec_delayed_refs_rsv_bg_inserts(struct btrfs_fs_info *fs_info);
385 void btrfs_inc_delayed_refs_rsv_bg_updates(struct btrfs_fs_info *fs_info);
386 void btrfs_dec_delayed_refs_rsv_bg_updates(struct btrfs_fs_info *fs_info);
387 int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
388 enum btrfs_reserve_flush_enum flush);
389 void btrfs_migrate_to_delayed_refs_rsv(struct btrfs_fs_info *fs_info,
391 bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info);
393 static inline u64 btrfs_delayed_ref_owner(struct btrfs_delayed_ref_node *node)
395 if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
396 node->type == BTRFS_SHARED_DATA_REF_KEY)
397 return node->data_ref.objectid;
398 return node->tree_ref.level;
401 static inline u64 btrfs_delayed_ref_offset(struct btrfs_delayed_ref_node *node)
403 if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
404 node->type == BTRFS_SHARED_DATA_REF_KEY)
405 return node->data_ref.offset;
409 static inline u8 btrfs_ref_type(struct btrfs_ref *ref)
411 ASSERT(ref->type == BTRFS_REF_DATA || ref->type == BTRFS_REF_METADATA);
413 if (ref->type == BTRFS_REF_DATA) {
415 return BTRFS_SHARED_DATA_REF_KEY;
417 return BTRFS_EXTENT_DATA_REF_KEY;
420 return BTRFS_SHARED_BLOCK_REF_KEY;
422 return BTRFS_TREE_BLOCK_REF_KEY;