]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - fs/btrfs/delayed-ref.h
Merge tag 'kvm-x86-misc-6.7' of https://github.com/kvm-x86/linux into HEAD
[thirdparty/kernel/stable.git] / fs / btrfs / delayed-ref.h
CommitLineData
9888c340 1/* SPDX-License-Identifier: GPL-2.0 */
56bec294
CM
2/*
3 * Copyright (C) 2008 Oracle. All rights reserved.
56bec294 4 */
9888c340
DS
5
6#ifndef BTRFS_DELAYED_REF_H
7#define BTRFS_DELAYED_REF_H
56bec294 8
6df8cdf5
ER
9#include <linux/refcount.h>
10
44a075bd 11/* these are the possible values of struct btrfs_delayed_ref_node->action */
56bec294
CM
12#define BTRFS_ADD_DELAYED_REF 1 /* add one backref to the tree */
13#define BTRFS_DROP_DELAYED_REF 2 /* delete one backref from the tree */
14#define BTRFS_ADD_DELAYED_EXTENT 3 /* record a full extent allocation */
1a81af4d 15#define BTRFS_UPDATE_DELAYED_HEAD 4 /* not changing ref count on head ref */
56bec294
CM
16
17struct btrfs_delayed_ref_node {
0e0adbcf 18 struct rb_node ref_node;
1d57ee94
WX
19 /*
20 * If action is BTRFS_ADD_DELAYED_REF, also link this node to
21 * ref_head->ref_add_list, then we do not need to iterate the
22 * whole ref_head->ref_list to find BTRFS_ADD_DELAYED_REF nodes.
23 */
24 struct list_head add_list;
c6fc2454 25
56bec294
CM
26 /* the starting bytenr of the extent */
27 u64 bytenr;
28
56bec294
CM
29 /* the size of the extent */
30 u64 num_bytes;
31
00f04b88
AJ
32 /* seq number to keep track of insertion order */
33 u64 seq;
34
56bec294 35 /* ref count on this data structure */
6df8cdf5 36 refcount_t refs;
56bec294
CM
37
38 /*
39 * how many refs is this entry adding or deleting. For
40 * head refs, this may be a negative number because it is keeping
41 * track of the total mods done to the reference count.
42 * For individual refs, this will always be a positive number
43 *
44 * It may be more than one, since it is possible for a single
45 * parent to have more than one ref on an extent
46 */
47 int ref_mod;
48
5d4f98a2
YZ
49 unsigned int action:8;
50 unsigned int type:8;
56bec294
CM
51};
52
5d4f98a2
YZ
53struct btrfs_delayed_extent_op {
54 struct btrfs_disk_key key;
35b3ad50
DS
55 u8 level;
56 bool update_key;
57 bool update_flags;
5d4f98a2 58 u64 flags_to_set;
5d4f98a2
YZ
59};
60
56bec294
CM
61/*
62 * the head refs are used to hold a lock on a given extent, which allows us
63 * to make sure that only one process is running the delayed refs
64 * at a time for a single extent. They also store the sum of all the
65 * reference count modifications we've queued up.
66 */
67struct btrfs_delayed_ref_head {
d278850e
JB
68 u64 bytenr;
69 u64 num_bytes;
315dd5cc
FM
70 /*
71 * For insertion into struct btrfs_delayed_ref_root::href_root.
72 * Keep it in the same cache line as 'bytenr' for more efficient
73 * searches in the rbtree.
74 */
75 struct rb_node href_node;
56bec294
CM
76 /*
77 * the mutex is held while running the refs, and it is also
78 * held when checking the sum of reference modifications.
79 */
80 struct mutex mutex;
81
315dd5cc
FM
82 refcount_t refs;
83
84 /* Protects 'ref_tree' and 'ref_add_list'. */
d7df2c79 85 spinlock_t lock;
e3d03965 86 struct rb_root_cached ref_tree;
1d57ee94
WX
87 /* accumulate add BTRFS_ADD_DELAYED_REF nodes to this ref_add_list. */
88 struct list_head ref_add_list;
c3e69d58 89
5d4f98a2 90 struct btrfs_delayed_extent_op *extent_op;
1262133b
JB
91
92 /*
93 * This is used to track the final ref_mod from all the refs associated
94 * with this head ref, this is not adjusted as delayed refs are run,
95 * this is meant to track if we need to do the csum accounting or not.
96 */
97 int total_ref_mod;
98
d278850e
JB
99 /*
100 * This is the current outstanding mod references for this bytenr. This
101 * is used with lookup_extent_info to get an accurate reference count
102 * for a bytenr, so it is adjusted as delayed refs are run so that any
103 * on disk reference count + ref_mod is accurate.
104 */
105 int ref_mod;
106
56bec294
CM
107 /*
108 * when a new extent is allocated, it is just reserved in memory
109 * The actual extent isn't inserted into the extent allocation tree
110 * until the delayed ref is processed. must_insert_reserved is
111 * used to flag a delayed ref so the accounting can be updated
112 * when a full insert is done.
113 *
114 * It is possible the extent will be freed before it is ever
115 * inserted into the extent allocation tree. In this case
116 * we need to update the in ram accounting to properly reflect
117 * the free has happened.
118 */
61c681fe
FM
119 bool must_insert_reserved;
120 bool is_data;
121 bool is_system;
122 bool processing;
56bec294
CM
123};
124
5d4f98a2 125struct btrfs_delayed_tree_ref {
56bec294 126 struct btrfs_delayed_ref_node node;
eebe063b
AJ
127 u64 root;
128 u64 parent;
5d4f98a2
YZ
129 int level;
130};
56bec294 131
5d4f98a2
YZ
132struct btrfs_delayed_data_ref {
133 struct btrfs_delayed_ref_node node;
eebe063b
AJ
134 u64 root;
135 u64 parent;
5d4f98a2
YZ
136 u64 objectid;
137 u64 offset;
56bec294
CM
138};
139
e19eb11f
JB
140enum btrfs_delayed_ref_flags {
141 /* Indicate that we are flushing delayed refs for the commit */
142 BTRFS_DELAYED_REFS_FLUSHING,
143};
144
56bec294 145struct btrfs_delayed_ref_root {
c46effa6 146 /* head ref rbtree */
5c9d028b 147 struct rb_root_cached href_root;
c46effa6 148
3368d001
QW
149 /* dirty extent records */
150 struct rb_root dirty_extent_root;
151
56bec294
CM
152 /* this spin lock protects the rbtree and the entries inside */
153 spinlock_t lock;
154
155 /* how many delayed ref updates we've queued, used by the
156 * throttling code
157 */
d7df2c79 158 atomic_t num_entries;
56bec294 159
c3e69d58
CM
160 /* total number of head nodes in tree */
161 unsigned long num_heads;
162
163 /* total number of head nodes ready for processing */
164 unsigned long num_heads_ready;
165
1262133b
JB
166 u64 pending_csums;
167
e19eb11f 168 unsigned long flags;
c3e69d58
CM
169
170 u64 run_delayed_start;
9086db86
QW
171
172 /*
173 * To make qgroup to skip given root.
01327610 174 * This is for snapshot, as btrfs_qgroup_inherit() will manually
9086db86
QW
175 * modify counters for snapshot and its source, so we should skip
176 * the snapshot in new_root/old_roots or it will get calculated twice
177 */
178 u64 qgroup_to_skip;
56bec294
CM
179};
180
b28b1f0c
QW
181enum btrfs_ref_type {
182 BTRFS_REF_NOT_SET,
183 BTRFS_REF_DATA,
184 BTRFS_REF_METADATA,
185 BTRFS_REF_LAST,
186};
187
188struct btrfs_data_ref {
189 /* For EXTENT_DATA_REF */
190
113479d5
NB
191 /* Original root this data extent belongs to */
192 u64 owning_root;
b28b1f0c
QW
193
194 /* Inode which refers to this data extent */
195 u64 ino;
196
197 /*
198 * file_offset - extent_offset
199 *
200 * file_offset is the key.offset of the EXTENT_DATA key.
201 * extent_offset is btrfs_file_extent_offset() of the EXTENT_DATA data.
202 */
203 u64 offset;
204};
205
206struct btrfs_tree_ref {
207 /*
208 * Level of this tree block
209 *
210 * Shared for skinny (TREE_BLOCK_REF) and normal tree ref.
211 */
212 int level;
213
214 /*
113479d5 215 * Root which owns this tree block.
b28b1f0c
QW
216 *
217 * For TREE_BLOCK_REF (skinny metadata, either inline or keyed)
218 */
113479d5 219 u64 owning_root;
b28b1f0c
QW
220
221 /* For non-skinny metadata, no special member needed */
222};
223
224struct btrfs_ref {
225 enum btrfs_ref_type type;
226 int action;
227
228 /*
229 * Whether this extent should go through qgroup record.
230 *
231 * Normally false, but for certain cases like delayed subtree scan,
232 * setting this flag can hugely reduce qgroup overhead.
233 */
234 bool skip_qgroup;
235
eed2037f
NB
236#ifdef CONFIG_BTRFS_FS_REF_VERIFY
237 /* Through which root is this modification. */
b28b1f0c 238 u64 real_root;
eed2037f 239#endif
b28b1f0c
QW
240 u64 bytenr;
241 u64 len;
242
243 /* Bytenr of the parent tree block */
244 u64 parent;
245 union {
246 struct btrfs_data_ref data_ref;
247 struct btrfs_tree_ref tree_ref;
248 };
249};
250
78a6184a
MX
251extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
252extern struct kmem_cache *btrfs_delayed_tree_ref_cachep;
253extern struct kmem_cache *btrfs_delayed_data_ref_cachep;
254extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
255
f5c29bd9 256int __init btrfs_delayed_ref_init(void);
e67c718b 257void __cold btrfs_delayed_ref_exit(void);
78a6184a 258
0e55a545
FM
259static inline u64 btrfs_calc_delayed_ref_bytes(const struct btrfs_fs_info *fs_info,
260 int num_delayed_refs)
261{
262 u64 num_bytes;
263
264 num_bytes = btrfs_calc_insert_metadata_size(fs_info, num_delayed_refs);
265
266 /*
267 * We have to check the mount option here because we could be enabling
268 * the free space tree for the first time and don't have the compat_ro
269 * option set yet.
270 *
271 * We need extra reservations if we have the free space tree because
272 * we'll have to modify that tree as well.
273 */
274 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
275 num_bytes *= 2;
276
277 return num_bytes;
278}
279
b28b1f0c
QW
280static inline void btrfs_init_generic_ref(struct btrfs_ref *generic_ref,
281 int action, u64 bytenr, u64 len, u64 parent)
282{
283 generic_ref->action = action;
284 generic_ref->bytenr = bytenr;
285 generic_ref->len = len;
286 generic_ref->parent = parent;
287}
288
289static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
f42c5da6 290 int level, u64 root, u64 mod_root, bool skip_qgroup)
b28b1f0c 291{
eed2037f 292#ifdef CONFIG_BTRFS_FS_REF_VERIFY
b28b1f0c 293 /* If @real_root not set, use @root as fallback */
eed2037f
NB
294 generic_ref->real_root = mod_root ?: root;
295#endif
b28b1f0c 296 generic_ref->tree_ref.level = level;
113479d5 297 generic_ref->tree_ref.owning_root = root;
b28b1f0c 298 generic_ref->type = BTRFS_REF_METADATA;
681145d4
NB
299 if (skip_qgroup || !(is_fstree(root) &&
300 (!mod_root || is_fstree(mod_root))))
301 generic_ref->skip_qgroup = true;
302 else
303 generic_ref->skip_qgroup = false;
304
b28b1f0c
QW
305}
306
307static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
f42c5da6
NB
308 u64 ref_root, u64 ino, u64 offset, u64 mod_root,
309 bool skip_qgroup)
b28b1f0c 310{
eed2037f 311#ifdef CONFIG_BTRFS_FS_REF_VERIFY
b28b1f0c 312 /* If @real_root not set, use @root as fallback */
eed2037f
NB
313 generic_ref->real_root = mod_root ?: ref_root;
314#endif
113479d5 315 generic_ref->data_ref.owning_root = ref_root;
b28b1f0c
QW
316 generic_ref->data_ref.ino = ino;
317 generic_ref->data_ref.offset = offset;
318 generic_ref->type = BTRFS_REF_DATA;
681145d4
NB
319 if (skip_qgroup || !(is_fstree(ref_root) &&
320 (!mod_root || is_fstree(mod_root))))
321 generic_ref->skip_qgroup = true;
322 else
323 generic_ref->skip_qgroup = false;
b28b1f0c
QW
324}
325
78a6184a
MX
326static inline struct btrfs_delayed_extent_op *
327btrfs_alloc_delayed_extent_op(void)
328{
329 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
330}
331
332static inline void
333btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
334{
335 if (op)
336 kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
337}
338
56bec294
CM
339static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
340{
6df8cdf5
ER
341 WARN_ON(refcount_read(&ref->refs) == 0);
342 if (refcount_dec_and_test(&ref->refs)) {
4d34ad34 343 WARN_ON(!RB_EMPTY_NODE(&ref->ref_node));
78a6184a
MX
344 switch (ref->type) {
345 case BTRFS_TREE_BLOCK_REF_KEY:
346 case BTRFS_SHARED_BLOCK_REF_KEY:
347 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
348 break;
349 case BTRFS_EXTENT_DATA_REF_KEY:
350 case BTRFS_SHARED_DATA_REF_KEY:
351 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
352 break;
78a6184a
MX
353 default:
354 BUG();
355 }
56bec294
CM
356 }
357}
358
2187374f
JB
359static inline u64 btrfs_ref_head_to_space_flags(
360 struct btrfs_delayed_ref_head *head_ref)
361{
362 if (head_ref->is_data)
363 return BTRFS_BLOCK_GROUP_DATA;
364 else if (head_ref->is_system)
365 return BTRFS_BLOCK_GROUP_SYSTEM;
366 return BTRFS_BLOCK_GROUP_METADATA;
367}
368
d278850e
JB
369static inline void btrfs_put_delayed_ref_head(struct btrfs_delayed_ref_head *head)
370{
371 if (refcount_dec_and_test(&head->refs))
372 kmem_cache_free(btrfs_delayed_ref_head_cachep, head);
373}
374
44e1c47d 375int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
ed4f255b 376 struct btrfs_ref *generic_ref,
2187374f 377 struct btrfs_delayed_extent_op *extent_op);
88a979c6 378int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
76675593 379 struct btrfs_ref *generic_ref,
2187374f 380 u64 reserved);
c6e340bc 381int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
5d4f98a2
YZ
382 u64 bytenr, u64 num_bytes,
383 struct btrfs_delayed_extent_op *extent_op);
0c555c97 384void btrfs_merge_delayed_refs(struct btrfs_fs_info *fs_info,
ae1e206b
JB
385 struct btrfs_delayed_ref_root *delayed_refs,
386 struct btrfs_delayed_ref_head *head);
56bec294 387
1887be66 388struct btrfs_delayed_ref_head *
f72ad18e
LB
389btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
390 u64 bytenr);
9e920a6f 391int btrfs_delayed_ref_lock(struct btrfs_delayed_ref_root *delayed_refs,
c3e69d58 392 struct btrfs_delayed_ref_head *head);
093486c4
MX
393static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
394{
395 mutex_unlock(&head->mutex);
396}
d7baffda
JB
397void btrfs_delete_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
398 struct btrfs_delayed_ref_head *head);
d7df2c79 399
5637c74b
LF
400struct btrfs_delayed_ref_head *btrfs_select_ref_head(
401 struct btrfs_delayed_ref_root *delayed_refs);
00f04b88 402
41d0bd3b 403int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq);
00f04b88 404
6ef03deb
JB
405void btrfs_delayed_refs_rsv_release(struct btrfs_fs_info *fs_info, int nr);
406void btrfs_update_delayed_refs_rsv(struct btrfs_trans_handle *trans);
407int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
408 enum btrfs_reserve_flush_enum flush);
409void btrfs_migrate_to_delayed_refs_rsv(struct btrfs_fs_info *fs_info,
6ef03deb 410 u64 num_bytes);
6ef03deb
JB
411bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info);
412
56bec294
CM
413/*
414 * helper functions to cast a node into its container
415 */
5d4f98a2
YZ
416static inline struct btrfs_delayed_tree_ref *
417btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node)
56bec294 418{
5d4f98a2
YZ
419 return container_of(node, struct btrfs_delayed_tree_ref, node);
420}
56bec294 421
5d4f98a2
YZ
422static inline struct btrfs_delayed_data_ref *
423btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node)
424{
5d4f98a2 425 return container_of(node, struct btrfs_delayed_data_ref, node);
56bec294 426}
9888c340 427
56bec294 428#endif