blob: 70b962cc177d973688b0d2deb56965bffbc423ae [file] [log] [blame]
Chris Mason56bec292009-03-13 10:10:06 -04001/*
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
Wang Sheng-Hui44a075b2012-09-21 08:21:01 +080021/* these are the possible values of struct btrfs_delayed_ref_node->action */
Chris Mason56bec292009-03-13 10:10:06 -040022#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 */
Chris Mason1a81af42009-03-25 09:55:11 -040025#define BTRFS_UPDATE_DELAYED_HEAD 4 /* not changing ref count on head ref */
Chris Mason56bec292009-03-13 10:10:06 -040026
27struct btrfs_delayed_ref_node {
28 struct rb_node rb_node;
29
30 /* the starting bytenr of the extent */
31 u64 bytenr;
32
Chris Mason56bec292009-03-13 10:10:06 -040033 /* the size of the extent */
34 u64 num_bytes;
35
Arne Jansen00f04b82011-09-14 12:37:00 +020036 /* seq number to keep track of insertion order */
37 u64 seq;
38
Chris Mason56bec292009-03-13 10:10:06 -040039 /* ref count on this data structure */
40 atomic_t refs;
41
42 /*
43 * how many refs is this entry adding or deleting. For
44 * head refs, this may be a negative number because it is keeping
45 * track of the total mods done to the reference count.
46 * For individual refs, this will always be a positive number
47 *
48 * It may be more than one, since it is possible for a single
49 * parent to have more than one ref on an extent
50 */
51 int ref_mod;
52
Yan Zheng5d4f98a2009-06-10 10:45:14 -040053 unsigned int action:8;
54 unsigned int type:8;
Chris Mason56bec292009-03-13 10:10:06 -040055 /* is this node still in the rbtree? */
Yan Zheng5d4f98a2009-06-10 10:45:14 -040056 unsigned int is_head:1;
Chris Mason56bec292009-03-13 10:10:06 -040057 unsigned int in_tree:1;
58};
59
Yan Zheng5d4f98a2009-06-10 10:45:14 -040060struct btrfs_delayed_extent_op {
61 struct btrfs_disk_key key;
62 u64 flags_to_set;
Josef Bacikb1c79e02013-05-09 13:49:30 -040063 int level;
Yan Zheng5d4f98a2009-06-10 10:45:14 -040064 unsigned int update_key:1;
65 unsigned int update_flags:1;
66 unsigned int is_data:1;
67};
68
Chris Mason56bec292009-03-13 10:10:06 -040069/*
70 * the head refs are used to hold a lock on a given extent, which allows us
71 * to make sure that only one process is running the delayed refs
72 * at a time for a single extent. They also store the sum of all the
73 * reference count modifications we've queued up.
74 */
75struct btrfs_delayed_ref_head {
76 struct btrfs_delayed_ref_node node;
77
78 /*
79 * the mutex is held while running the refs, and it is also
80 * held when checking the sum of reference modifications.
81 */
82 struct mutex mutex;
83
Chris Masonc3e69d52009-03-13 10:17:05 -040084 struct list_head cluster;
85
Yan Zheng5d4f98a2009-06-10 10:45:14 -040086 struct btrfs_delayed_extent_op *extent_op;
Chris Mason56bec292009-03-13 10:10:06 -040087 /*
88 * when a new extent is allocated, it is just reserved in memory
89 * The actual extent isn't inserted into the extent allocation tree
90 * until the delayed ref is processed. must_insert_reserved is
91 * used to flag a delayed ref so the accounting can be updated
92 * when a full insert is done.
93 *
94 * It is possible the extent will be freed before it is ever
95 * inserted into the extent allocation tree. In this case
96 * we need to update the in ram accounting to properly reflect
97 * the free has happened.
98 */
99 unsigned int must_insert_reserved:1;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400100 unsigned int is_data:1;
Chris Mason56bec292009-03-13 10:10:06 -0400101};
102
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400103struct btrfs_delayed_tree_ref {
Chris Mason56bec292009-03-13 10:10:06 -0400104 struct btrfs_delayed_ref_node node;
Arne Janseneebe0632011-09-14 14:01:24 +0200105 u64 root;
106 u64 parent;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400107 int level;
108};
Chris Mason56bec292009-03-13 10:10:06 -0400109
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400110struct btrfs_delayed_data_ref {
111 struct btrfs_delayed_ref_node node;
Arne Janseneebe0632011-09-14 14:01:24 +0200112 u64 root;
113 u64 parent;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400114 u64 objectid;
115 u64 offset;
Chris Mason56bec292009-03-13 10:10:06 -0400116};
117
118struct btrfs_delayed_ref_root {
119 struct rb_root root;
120
121 /* this spin lock protects the rbtree and the entries inside */
122 spinlock_t lock;
123
124 /* how many delayed ref updates we've queued, used by the
125 * throttling code
126 */
127 unsigned long num_entries;
128
Chris Masonc3e69d52009-03-13 10:17:05 -0400129 /* total number of head nodes in tree */
130 unsigned long num_heads;
131
132 /* total number of head nodes ready for processing */
133 unsigned long num_heads_ready;
134
Chris Mason56bec292009-03-13 10:10:06 -0400135 /*
Chris Masonbb721702013-01-29 18:44:12 -0500136 * bumped when someone is making progress on the delayed
137 * refs, so that other procs know they are just adding to
138 * contention intead of helping
139 */
140 atomic_t procs_running_refs;
141 atomic_t ref_seq;
142 wait_queue_head_t wait;
143
144 /*
Chris Mason56bec292009-03-13 10:10:06 -0400145 * set when the tree is flushing before a transaction commit,
146 * used by the throttling code to decide if new updates need
147 * to be run right away
148 */
149 int flushing;
Chris Masonc3e69d52009-03-13 10:17:05 -0400150
151 u64 run_delayed_start;
Chris Mason56bec292009-03-13 10:10:06 -0400152};
153
Miao Xie78a61842012-11-21 02:21:28 +0000154extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
155extern struct kmem_cache *btrfs_delayed_tree_ref_cachep;
156extern struct kmem_cache *btrfs_delayed_data_ref_cachep;
157extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
158
159int btrfs_delayed_ref_init(void);
160void btrfs_delayed_ref_exit(void);
161
162static inline struct btrfs_delayed_extent_op *
163btrfs_alloc_delayed_extent_op(void)
164{
165 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
166}
167
168static inline void
169btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
170{
171 if (op)
172 kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
173}
174
Chris Mason56bec292009-03-13 10:10:06 -0400175static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
176{
177 WARN_ON(atomic_read(&ref->refs) == 0);
178 if (atomic_dec_and_test(&ref->refs)) {
179 WARN_ON(ref->in_tree);
Miao Xie78a61842012-11-21 02:21:28 +0000180 switch (ref->type) {
181 case BTRFS_TREE_BLOCK_REF_KEY:
182 case BTRFS_SHARED_BLOCK_REF_KEY:
183 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
184 break;
185 case BTRFS_EXTENT_DATA_REF_KEY:
186 case BTRFS_SHARED_DATA_REF_KEY:
187 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
188 break;
189 case 0:
190 kmem_cache_free(btrfs_delayed_ref_head_cachep, ref);
191 break;
192 default:
193 BUG();
194 }
Chris Mason56bec292009-03-13 10:10:06 -0400195 }
196}
197
Arne Jansen66d7e7f2011-09-12 15:26:38 +0200198int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
199 struct btrfs_trans_handle *trans,
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400200 u64 bytenr, u64 num_bytes, u64 parent,
201 u64 ref_root, int level, int action,
Arne Jansen66d7e7f2011-09-12 15:26:38 +0200202 struct btrfs_delayed_extent_op *extent_op,
203 int for_cow);
204int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
205 struct btrfs_trans_handle *trans,
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400206 u64 bytenr, u64 num_bytes,
207 u64 parent, u64 ref_root,
208 u64 owner, u64 offset, int action,
Arne Jansen66d7e7f2011-09-12 15:26:38 +0200209 struct btrfs_delayed_extent_op *extent_op,
210 int for_cow);
211int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
212 struct btrfs_trans_handle *trans,
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400213 u64 bytenr, u64 num_bytes,
214 struct btrfs_delayed_extent_op *extent_op);
Josef Bacikae1e2062012-08-07 16:00:32 -0400215void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
216 struct btrfs_fs_info *fs_info,
217 struct btrfs_delayed_ref_root *delayed_refs,
218 struct btrfs_delayed_ref_head *head);
Chris Mason56bec292009-03-13 10:10:06 -0400219
Chris Mason1887be62009-03-13 10:11:24 -0400220struct btrfs_delayed_ref_head *
221btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr);
Chris Masonc3e69d52009-03-13 10:17:05 -0400222int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
223 struct btrfs_delayed_ref_head *head);
Miao Xie093486c2012-12-19 08:10:10 +0000224static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
225{
226 mutex_unlock(&head->mutex);
227}
228
Chris Masonc3e69d52009-03-13 10:17:05 -0400229int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
230 struct list_head *cluster, u64 search_start);
Miao Xie093486c2012-12-19 08:10:10 +0000231void btrfs_release_ref_cluster(struct list_head *cluster);
Arne Jansen00f04b82011-09-14 12:37:00 +0200232
Jan Schmidt097b8a72012-06-21 11:08:04 +0200233int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
234 struct btrfs_delayed_ref_root *delayed_refs,
Arne Jansen00f04b82011-09-14 12:37:00 +0200235 u64 seq);
236
237/*
Jan Schmidt546adb02012-06-14 16:37:44 +0200238 * delayed refs with a ref_seq > 0 must be held back during backref walking.
239 * this only applies to items in one of the fs-trees. for_cow items never need
240 * to be held back, so they won't get a ref_seq number.
241 */
242static inline int need_ref_seq(int for_cow, u64 rootid)
243{
244 if (for_cow)
245 return 0;
246
247 if (rootid == BTRFS_FS_TREE_OBJECTID)
248 return 1;
249
250 if ((s64)rootid >= (s64)BTRFS_FIRST_FREE_OBJECTID)
251 return 1;
252
253 return 0;
254}
255
256/*
Chris Mason56bec292009-03-13 10:10:06 -0400257 * a node might live in a head or a regular ref, this lets you
258 * test for the proper type to use.
259 */
260static int btrfs_delayed_ref_is_head(struct btrfs_delayed_ref_node *node)
261{
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400262 return node->is_head;
Chris Mason56bec292009-03-13 10:10:06 -0400263}
264
265/*
266 * helper functions to cast a node into its container
267 */
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400268static inline struct btrfs_delayed_tree_ref *
269btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node)
Chris Mason56bec292009-03-13 10:10:06 -0400270{
271 WARN_ON(btrfs_delayed_ref_is_head(node));
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400272 return container_of(node, struct btrfs_delayed_tree_ref, node);
273}
Chris Mason56bec292009-03-13 10:10:06 -0400274
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400275static inline struct btrfs_delayed_data_ref *
276btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node)
277{
278 WARN_ON(btrfs_delayed_ref_is_head(node));
279 return container_of(node, struct btrfs_delayed_data_ref, node);
Chris Mason56bec292009-03-13 10:10:06 -0400280}
281
282static inline struct btrfs_delayed_ref_head *
283btrfs_delayed_node_to_head(struct btrfs_delayed_ref_node *node)
284{
285 WARN_ON(!btrfs_delayed_ref_is_head(node));
286 return container_of(node, struct btrfs_delayed_ref_head, node);
Chris Mason56bec292009-03-13 10:10:06 -0400287}
288#endif