]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/btrfs/delayed-ref.h
btrfs: improve delayed refs iterations
[mirror_ubuntu-artful-kernel.git] / fs / btrfs / delayed-ref.h
CommitLineData
56bec294
CM
1/*
2 * Copyright (C) 2008 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18#ifndef __DELAYED_REF__
19#define __DELAYED_REF__
20
44a075bd 21/* these are the possible values of struct btrfs_delayed_ref_node->action */
56bec294
CM
22#define BTRFS_ADD_DELAYED_REF 1 /* add one backref to the tree */
23#define BTRFS_DROP_DELAYED_REF 2 /* delete one backref from the tree */
24#define BTRFS_ADD_DELAYED_EXTENT 3 /* record a full extent allocation */
1a81af4d 25#define BTRFS_UPDATE_DELAYED_HEAD 4 /* not changing ref count on head ref */
56bec294 26
c6fc2454
QW
27/*
28 * XXX: Qu: I really hate the design that ref_head and tree/data ref shares the
29 * same ref_node structure.
30 * Ref_head is in a higher logic level than tree/data ref, and duplicated
31 * bytenr/num_bytes in ref_node is really a waste or memory, they should be
32 * referred from ref_head.
33 * This gets more disgusting after we use list to store tree/data ref in
34 * ref_head. Must clean this mess up later.
35 */
56bec294 36struct btrfs_delayed_ref_node {
c6fc2454
QW
37 /*
38 * ref_head use rb tree, stored in ref_root->href.
39 * indexed by bytenr
40 */
56bec294
CM
41 struct rb_node rb_node;
42
c6fc2454
QW
43 /*data/tree ref use list, stored in ref_head->ref_list. */
44 struct list_head list;
1d57ee94
WX
45 /*
46 * If action is BTRFS_ADD_DELAYED_REF, also link this node to
47 * ref_head->ref_add_list, then we do not need to iterate the
48 * whole ref_head->ref_list to find BTRFS_ADD_DELAYED_REF nodes.
49 */
50 struct list_head add_list;
c6fc2454 51
56bec294
CM
52 /* the starting bytenr of the extent */
53 u64 bytenr;
54
56bec294
CM
55 /* the size of the extent */
56 u64 num_bytes;
57
00f04b88
AJ
58 /* seq number to keep track of insertion order */
59 u64 seq;
60
56bec294
CM
61 /* ref count on this data structure */
62 atomic_t refs;
63
64 /*
65 * how many refs is this entry adding or deleting. For
66 * head refs, this may be a negative number because it is keeping
67 * track of the total mods done to the reference count.
68 * For individual refs, this will always be a positive number
69 *
70 * It may be more than one, since it is possible for a single
71 * parent to have more than one ref on an extent
72 */
73 int ref_mod;
74
5d4f98a2
YZ
75 unsigned int action:8;
76 unsigned int type:8;
56bec294 77 /* is this node still in the rbtree? */
5d4f98a2 78 unsigned int is_head:1;
56bec294
CM
79 unsigned int in_tree:1;
80};
81
5d4f98a2
YZ
82struct btrfs_delayed_extent_op {
83 struct btrfs_disk_key key;
35b3ad50
DS
84 u8 level;
85 bool update_key;
86 bool update_flags;
87 bool is_data;
5d4f98a2 88 u64 flags_to_set;
5d4f98a2
YZ
89};
90
56bec294
CM
91/*
92 * the head refs are used to hold a lock on a given extent, which allows us
93 * to make sure that only one process is running the delayed refs
94 * at a time for a single extent. They also store the sum of all the
95 * reference count modifications we've queued up.
96 */
97struct btrfs_delayed_ref_head {
98 struct btrfs_delayed_ref_node node;
99
100 /*
101 * the mutex is held while running the refs, and it is also
102 * held when checking the sum of reference modifications.
103 */
104 struct mutex mutex;
105
d7df2c79 106 spinlock_t lock;
c6fc2454 107 struct list_head ref_list;
1d57ee94
WX
108 /* accumulate add BTRFS_ADD_DELAYED_REF nodes to this ref_add_list. */
109 struct list_head ref_add_list;
c3e69d58 110
c46effa6
LB
111 struct rb_node href_node;
112
5d4f98a2 113 struct btrfs_delayed_extent_op *extent_op;
1262133b
JB
114
115 /*
116 * This is used to track the final ref_mod from all the refs associated
117 * with this head ref, this is not adjusted as delayed refs are run,
118 * this is meant to track if we need to do the csum accounting or not.
119 */
120 int total_ref_mod;
121
f64d5ca8
QW
122 /*
123 * For qgroup reserved space freeing.
124 *
125 * ref_root and reserved will be recorded after
126 * BTRFS_ADD_DELAYED_EXTENT is called.
127 * And will be used to free reserved qgroup space at
128 * run_delayed_refs() time.
129 */
130 u64 qgroup_ref_root;
131 u64 qgroup_reserved;
132
56bec294
CM
133 /*
134 * when a new extent is allocated, it is just reserved in memory
135 * The actual extent isn't inserted into the extent allocation tree
136 * until the delayed ref is processed. must_insert_reserved is
137 * used to flag a delayed ref so the accounting can be updated
138 * when a full insert is done.
139 *
140 * It is possible the extent will be freed before it is ever
141 * inserted into the extent allocation tree. In this case
142 * we need to update the in ram accounting to properly reflect
143 * the free has happened.
144 */
145 unsigned int must_insert_reserved:1;
5d4f98a2 146 unsigned int is_data:1;
d7df2c79 147 unsigned int processing:1;
56bec294
CM
148};
149
5d4f98a2 150struct btrfs_delayed_tree_ref {
56bec294 151 struct btrfs_delayed_ref_node node;
eebe063b
AJ
152 u64 root;
153 u64 parent;
5d4f98a2
YZ
154 int level;
155};
56bec294 156
5d4f98a2
YZ
157struct btrfs_delayed_data_ref {
158 struct btrfs_delayed_ref_node node;
eebe063b
AJ
159 u64 root;
160 u64 parent;
5d4f98a2
YZ
161 u64 objectid;
162 u64 offset;
56bec294
CM
163};
164
165struct btrfs_delayed_ref_root {
c46effa6
LB
166 /* head ref rbtree */
167 struct rb_root href_root;
168
3368d001
QW
169 /* dirty extent records */
170 struct rb_root dirty_extent_root;
171
56bec294
CM
172 /* this spin lock protects the rbtree and the entries inside */
173 spinlock_t lock;
174
175 /* how many delayed ref updates we've queued, used by the
176 * throttling code
177 */
d7df2c79 178 atomic_t num_entries;
56bec294 179
c3e69d58
CM
180 /* total number of head nodes in tree */
181 unsigned long num_heads;
182
183 /* total number of head nodes ready for processing */
184 unsigned long num_heads_ready;
185
1262133b
JB
186 u64 pending_csums;
187
56bec294
CM
188 /*
189 * set when the tree is flushing before a transaction commit,
190 * used by the throttling code to decide if new updates need
191 * to be run right away
192 */
193 int flushing;
c3e69d58
CM
194
195 u64 run_delayed_start;
9086db86
QW
196
197 /*
198 * To make qgroup to skip given root.
01327610 199 * This is for snapshot, as btrfs_qgroup_inherit() will manually
9086db86
QW
200 * modify counters for snapshot and its source, so we should skip
201 * the snapshot in new_root/old_roots or it will get calculated twice
202 */
203 u64 qgroup_to_skip;
56bec294
CM
204};
205
78a6184a
MX
206extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
207extern struct kmem_cache *btrfs_delayed_tree_ref_cachep;
208extern struct kmem_cache *btrfs_delayed_data_ref_cachep;
209extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
210
211int btrfs_delayed_ref_init(void);
212void btrfs_delayed_ref_exit(void);
213
214static inline struct btrfs_delayed_extent_op *
215btrfs_alloc_delayed_extent_op(void)
216{
217 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
218}
219
220static inline void
221btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
222{
223 if (op)
224 kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
225}
226
56bec294
CM
227static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
228{
229 WARN_ON(atomic_read(&ref->refs) == 0);
230 if (atomic_dec_and_test(&ref->refs)) {
231 WARN_ON(ref->in_tree);
78a6184a
MX
232 switch (ref->type) {
233 case BTRFS_TREE_BLOCK_REF_KEY:
234 case BTRFS_SHARED_BLOCK_REF_KEY:
235 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
236 break;
237 case BTRFS_EXTENT_DATA_REF_KEY:
238 case BTRFS_SHARED_DATA_REF_KEY:
239 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
240 break;
241 case 0:
242 kmem_cache_free(btrfs_delayed_ref_head_cachep, ref);
243 break;
244 default:
245 BUG();
246 }
56bec294
CM
247 }
248}
249
66d7e7f0
AJ
250int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
251 struct btrfs_trans_handle *trans,
5d4f98a2
YZ
252 u64 bytenr, u64 num_bytes, u64 parent,
253 u64 ref_root, int level, int action,
b06c4bf5 254 struct btrfs_delayed_extent_op *extent_op);
66d7e7f0
AJ
255int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
256 struct btrfs_trans_handle *trans,
5d4f98a2
YZ
257 u64 bytenr, u64 num_bytes,
258 u64 parent, u64 ref_root,
5846a3c2 259 u64 owner, u64 offset, u64 reserved, int action,
b06c4bf5 260 struct btrfs_delayed_extent_op *extent_op);
66d7e7f0
AJ
261int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
262 struct btrfs_trans_handle *trans,
5d4f98a2
YZ
263 u64 bytenr, u64 num_bytes,
264 struct btrfs_delayed_extent_op *extent_op);
ae1e206b
JB
265void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
266 struct btrfs_fs_info *fs_info,
267 struct btrfs_delayed_ref_root *delayed_refs,
268 struct btrfs_delayed_ref_head *head);
56bec294 269
1887be66
CM
270struct btrfs_delayed_ref_head *
271btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr);
c3e69d58
CM
272int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
273 struct btrfs_delayed_ref_head *head);
093486c4
MX
274static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
275{
276 mutex_unlock(&head->mutex);
277}
278
d7df2c79
JB
279
280struct btrfs_delayed_ref_head *
281btrfs_select_ref_head(struct btrfs_trans_handle *trans);
00f04b88 282
097b8a7c
JS
283int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
284 struct btrfs_delayed_ref_root *delayed_refs,
00f04b88
AJ
285 u64 seq);
286
56bec294
CM
287/*
288 * a node might live in a head or a regular ref, this lets you
289 * test for the proper type to use.
290 */
291static int btrfs_delayed_ref_is_head(struct btrfs_delayed_ref_node *node)
292{
5d4f98a2 293 return node->is_head;
56bec294
CM
294}
295
296/*
297 * helper functions to cast a node into its container
298 */
5d4f98a2
YZ
299static inline struct btrfs_delayed_tree_ref *
300btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node)
56bec294
CM
301{
302 WARN_ON(btrfs_delayed_ref_is_head(node));
5d4f98a2
YZ
303 return container_of(node, struct btrfs_delayed_tree_ref, node);
304}
56bec294 305
5d4f98a2
YZ
306static inline struct btrfs_delayed_data_ref *
307btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node)
308{
309 WARN_ON(btrfs_delayed_ref_is_head(node));
310 return container_of(node, struct btrfs_delayed_data_ref, node);
56bec294
CM
311}
312
313static inline struct btrfs_delayed_ref_head *
314btrfs_delayed_node_to_head(struct btrfs_delayed_ref_node *node)
315{
316 WARN_ON(!btrfs_delayed_ref_is_head(node));
317 return container_of(node, struct btrfs_delayed_ref_head, node);
56bec294
CM
318}
319#endif