overlayfs: Implement splice-read
[linux-block.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 51 /* is this node still in the rbtree? */
5d4f98a2 52 unsigned int is_head:1;
56bec294
CM
53 unsigned int in_tree:1;
54};
55
5d4f98a2
YZ
56struct btrfs_delayed_extent_op {
57 struct btrfs_disk_key key;
35b3ad50
DS
58 u8 level;
59 bool update_key;
60 bool update_flags;
5d4f98a2 61 u64 flags_to_set;
5d4f98a2
YZ
62};
63
56bec294
CM
64/*
65 * the head refs are used to hold a lock on a given extent, which allows us
66 * to make sure that only one process is running the delayed refs
67 * at a time for a single extent. They also store the sum of all the
68 * reference count modifications we've queued up.
69 */
70struct btrfs_delayed_ref_head {
d278850e
JB
71 u64 bytenr;
72 u64 num_bytes;
73 refcount_t refs;
56bec294
CM
74 /*
75 * the mutex is held while running the refs, and it is also
76 * held when checking the sum of reference modifications.
77 */
78 struct mutex mutex;
79
d7df2c79 80 spinlock_t lock;
e3d03965 81 struct rb_root_cached ref_tree;
1d57ee94
WX
82 /* accumulate add BTRFS_ADD_DELAYED_REF nodes to this ref_add_list. */
83 struct list_head ref_add_list;
c3e69d58 84
c46effa6
LB
85 struct rb_node href_node;
86
5d4f98a2 87 struct btrfs_delayed_extent_op *extent_op;
1262133b
JB
88
89 /*
90 * This is used to track the final ref_mod from all the refs associated
91 * with this head ref, this is not adjusted as delayed refs are run,
92 * this is meant to track if we need to do the csum accounting or not.
93 */
94 int total_ref_mod;
95
d278850e
JB
96 /*
97 * This is the current outstanding mod references for this bytenr. This
98 * is used with lookup_extent_info to get an accurate reference count
99 * for a bytenr, so it is adjusted as delayed refs are run so that any
100 * on disk reference count + ref_mod is accurate.
101 */
102 int ref_mod;
103
56bec294
CM
104 /*
105 * when a new extent is allocated, it is just reserved in memory
106 * The actual extent isn't inserted into the extent allocation tree
107 * until the delayed ref is processed. must_insert_reserved is
108 * used to flag a delayed ref so the accounting can be updated
109 * when a full insert is done.
110 *
111 * It is possible the extent will be freed before it is ever
112 * inserted into the extent allocation tree. In this case
113 * we need to update the in ram accounting to properly reflect
114 * the free has happened.
115 */
116 unsigned int must_insert_reserved:1;
5d4f98a2 117 unsigned int is_data:1;
5e388e95 118 unsigned int is_system:1;
d7df2c79 119 unsigned int processing:1;
56bec294
CM
120};
121
5d4f98a2 122struct btrfs_delayed_tree_ref {
56bec294 123 struct btrfs_delayed_ref_node node;
eebe063b
AJ
124 u64 root;
125 u64 parent;
5d4f98a2
YZ
126 int level;
127};
56bec294 128
5d4f98a2
YZ
129struct btrfs_delayed_data_ref {
130 struct btrfs_delayed_ref_node node;
eebe063b
AJ
131 u64 root;
132 u64 parent;
5d4f98a2
YZ
133 u64 objectid;
134 u64 offset;
56bec294
CM
135};
136
e19eb11f
JB
137enum btrfs_delayed_ref_flags {
138 /* Indicate that we are flushing delayed refs for the commit */
139 BTRFS_DELAYED_REFS_FLUSHING,
140};
141
56bec294 142struct btrfs_delayed_ref_root {
c46effa6 143 /* head ref rbtree */
5c9d028b 144 struct rb_root_cached href_root;
c46effa6 145
3368d001
QW
146 /* dirty extent records */
147 struct rb_root dirty_extent_root;
148
56bec294
CM
149 /* this spin lock protects the rbtree and the entries inside */
150 spinlock_t lock;
151
152 /* how many delayed ref updates we've queued, used by the
153 * throttling code
154 */
d7df2c79 155 atomic_t num_entries;
56bec294 156
c3e69d58
CM
157 /* total number of head nodes in tree */
158 unsigned long num_heads;
159
160 /* total number of head nodes ready for processing */
161 unsigned long num_heads_ready;
162
1262133b
JB
163 u64 pending_csums;
164
e19eb11f 165 unsigned long flags;
c3e69d58
CM
166
167 u64 run_delayed_start;
9086db86
QW
168
169 /*
170 * To make qgroup to skip given root.
01327610 171 * This is for snapshot, as btrfs_qgroup_inherit() will manually
9086db86
QW
172 * modify counters for snapshot and its source, so we should skip
173 * the snapshot in new_root/old_roots or it will get calculated twice
174 */
175 u64 qgroup_to_skip;
56bec294
CM
176};
177
b28b1f0c
QW
178enum btrfs_ref_type {
179 BTRFS_REF_NOT_SET,
180 BTRFS_REF_DATA,
181 BTRFS_REF_METADATA,
182 BTRFS_REF_LAST,
183};
184
185struct btrfs_data_ref {
186 /* For EXTENT_DATA_REF */
187
113479d5
NB
188 /* Original root this data extent belongs to */
189 u64 owning_root;
b28b1f0c
QW
190
191 /* Inode which refers to this data extent */
192 u64 ino;
193
194 /*
195 * file_offset - extent_offset
196 *
197 * file_offset is the key.offset of the EXTENT_DATA key.
198 * extent_offset is btrfs_file_extent_offset() of the EXTENT_DATA data.
199 */
200 u64 offset;
201};
202
203struct btrfs_tree_ref {
204 /*
205 * Level of this tree block
206 *
207 * Shared for skinny (TREE_BLOCK_REF) and normal tree ref.
208 */
209 int level;
210
211 /*
113479d5 212 * Root which owns this tree block.
b28b1f0c
QW
213 *
214 * For TREE_BLOCK_REF (skinny metadata, either inline or keyed)
215 */
113479d5 216 u64 owning_root;
b28b1f0c
QW
217
218 /* For non-skinny metadata, no special member needed */
219};
220
221struct btrfs_ref {
222 enum btrfs_ref_type type;
223 int action;
224
225 /*
226 * Whether this extent should go through qgroup record.
227 *
228 * Normally false, but for certain cases like delayed subtree scan,
229 * setting this flag can hugely reduce qgroup overhead.
230 */
231 bool skip_qgroup;
232
eed2037f
NB
233#ifdef CONFIG_BTRFS_FS_REF_VERIFY
234 /* Through which root is this modification. */
b28b1f0c 235 u64 real_root;
eed2037f 236#endif
b28b1f0c
QW
237 u64 bytenr;
238 u64 len;
239
240 /* Bytenr of the parent tree block */
241 u64 parent;
242 union {
243 struct btrfs_data_ref data_ref;
244 struct btrfs_tree_ref tree_ref;
245 };
246};
247
78a6184a
MX
248extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
249extern struct kmem_cache *btrfs_delayed_tree_ref_cachep;
250extern struct kmem_cache *btrfs_delayed_data_ref_cachep;
251extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
252
f5c29bd9 253int __init btrfs_delayed_ref_init(void);
e67c718b 254void __cold btrfs_delayed_ref_exit(void);
78a6184a 255
0e55a545
FM
256static inline u64 btrfs_calc_delayed_ref_bytes(const struct btrfs_fs_info *fs_info,
257 int num_delayed_refs)
258{
259 u64 num_bytes;
260
261 num_bytes = btrfs_calc_insert_metadata_size(fs_info, num_delayed_refs);
262
263 /*
264 * We have to check the mount option here because we could be enabling
265 * the free space tree for the first time and don't have the compat_ro
266 * option set yet.
267 *
268 * We need extra reservations if we have the free space tree because
269 * we'll have to modify that tree as well.
270 */
271 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
272 num_bytes *= 2;
273
274 return num_bytes;
275}
276
b28b1f0c
QW
277static inline void btrfs_init_generic_ref(struct btrfs_ref *generic_ref,
278 int action, u64 bytenr, u64 len, u64 parent)
279{
280 generic_ref->action = action;
281 generic_ref->bytenr = bytenr;
282 generic_ref->len = len;
283 generic_ref->parent = parent;
284}
285
286static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
f42c5da6 287 int level, u64 root, u64 mod_root, bool skip_qgroup)
b28b1f0c 288{
eed2037f 289#ifdef CONFIG_BTRFS_FS_REF_VERIFY
b28b1f0c 290 /* If @real_root not set, use @root as fallback */
eed2037f
NB
291 generic_ref->real_root = mod_root ?: root;
292#endif
b28b1f0c 293 generic_ref->tree_ref.level = level;
113479d5 294 generic_ref->tree_ref.owning_root = root;
b28b1f0c 295 generic_ref->type = BTRFS_REF_METADATA;
681145d4
NB
296 if (skip_qgroup || !(is_fstree(root) &&
297 (!mod_root || is_fstree(mod_root))))
298 generic_ref->skip_qgroup = true;
299 else
300 generic_ref->skip_qgroup = false;
301
b28b1f0c
QW
302}
303
304static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
f42c5da6
NB
305 u64 ref_root, u64 ino, u64 offset, u64 mod_root,
306 bool skip_qgroup)
b28b1f0c 307{
eed2037f 308#ifdef CONFIG_BTRFS_FS_REF_VERIFY
b28b1f0c 309 /* If @real_root not set, use @root as fallback */
eed2037f
NB
310 generic_ref->real_root = mod_root ?: ref_root;
311#endif
113479d5 312 generic_ref->data_ref.owning_root = ref_root;
b28b1f0c
QW
313 generic_ref->data_ref.ino = ino;
314 generic_ref->data_ref.offset = offset;
315 generic_ref->type = BTRFS_REF_DATA;
681145d4
NB
316 if (skip_qgroup || !(is_fstree(ref_root) &&
317 (!mod_root || is_fstree(mod_root))))
318 generic_ref->skip_qgroup = true;
319 else
320 generic_ref->skip_qgroup = false;
b28b1f0c
QW
321}
322
78a6184a
MX
323static inline struct btrfs_delayed_extent_op *
324btrfs_alloc_delayed_extent_op(void)
325{
326 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
327}
328
329static inline void
330btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
331{
332 if (op)
333 kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
334}
335
56bec294
CM
336static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
337{
6df8cdf5
ER
338 WARN_ON(refcount_read(&ref->refs) == 0);
339 if (refcount_dec_and_test(&ref->refs)) {
56bec294 340 WARN_ON(ref->in_tree);
78a6184a
MX
341 switch (ref->type) {
342 case BTRFS_TREE_BLOCK_REF_KEY:
343 case BTRFS_SHARED_BLOCK_REF_KEY:
344 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
345 break;
346 case BTRFS_EXTENT_DATA_REF_KEY:
347 case BTRFS_SHARED_DATA_REF_KEY:
348 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
349 break;
78a6184a
MX
350 default:
351 BUG();
352 }
56bec294
CM
353 }
354}
355
2187374f
JB
356static inline u64 btrfs_ref_head_to_space_flags(
357 struct btrfs_delayed_ref_head *head_ref)
358{
359 if (head_ref->is_data)
360 return BTRFS_BLOCK_GROUP_DATA;
361 else if (head_ref->is_system)
362 return BTRFS_BLOCK_GROUP_SYSTEM;
363 return BTRFS_BLOCK_GROUP_METADATA;
364}
365
d278850e
JB
366static inline void btrfs_put_delayed_ref_head(struct btrfs_delayed_ref_head *head)
367{
368 if (refcount_dec_and_test(&head->refs))
369 kmem_cache_free(btrfs_delayed_ref_head_cachep, head);
370}
371
44e1c47d 372int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
ed4f255b 373 struct btrfs_ref *generic_ref,
2187374f 374 struct btrfs_delayed_extent_op *extent_op);
88a979c6 375int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
76675593 376 struct btrfs_ref *generic_ref,
2187374f 377 u64 reserved);
c6e340bc 378int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
5d4f98a2
YZ
379 u64 bytenr, u64 num_bytes,
380 struct btrfs_delayed_extent_op *extent_op);
0c555c97 381void btrfs_merge_delayed_refs(struct btrfs_fs_info *fs_info,
ae1e206b
JB
382 struct btrfs_delayed_ref_root *delayed_refs,
383 struct btrfs_delayed_ref_head *head);
56bec294 384
1887be66 385struct btrfs_delayed_ref_head *
f72ad18e
LB
386btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
387 u64 bytenr);
9e920a6f 388int btrfs_delayed_ref_lock(struct btrfs_delayed_ref_root *delayed_refs,
c3e69d58 389 struct btrfs_delayed_ref_head *head);
093486c4
MX
390static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
391{
392 mutex_unlock(&head->mutex);
393}
d7baffda
JB
394void btrfs_delete_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
395 struct btrfs_delayed_ref_head *head);
d7df2c79 396
5637c74b
LF
397struct btrfs_delayed_ref_head *btrfs_select_ref_head(
398 struct btrfs_delayed_ref_root *delayed_refs);
00f04b88 399
41d0bd3b 400int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq);
00f04b88 401
6ef03deb
JB
402void btrfs_delayed_refs_rsv_release(struct btrfs_fs_info *fs_info, int nr);
403void btrfs_update_delayed_refs_rsv(struct btrfs_trans_handle *trans);
404int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
405 enum btrfs_reserve_flush_enum flush);
406void btrfs_migrate_to_delayed_refs_rsv(struct btrfs_fs_info *fs_info,
407 struct btrfs_block_rsv *src,
408 u64 num_bytes);
6ef03deb
JB
409bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info);
410
56bec294
CM
411/*
412 * helper functions to cast a node into its container
413 */
5d4f98a2
YZ
414static inline struct btrfs_delayed_tree_ref *
415btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node)
56bec294 416{
5d4f98a2
YZ
417 return container_of(node, struct btrfs_delayed_tree_ref, node);
418}
56bec294 419
5d4f98a2
YZ
420static inline struct btrfs_delayed_data_ref *
421btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node)
422{
5d4f98a2 423 return container_of(node, struct btrfs_delayed_data_ref, node);
56bec294 424}
9888c340 425
56bec294 426#endif