blob: c24b653c7343018a89931ef8344e0f71c2c1bf97 [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
Qu Wenruoc6fc2452015-03-30 17:03:00 +080027/*
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 */
Chris Mason56bec292009-03-13 10:10:06 -040036struct btrfs_delayed_ref_node {
Qu Wenruoc6fc2452015-03-30 17:03:00 +080037 /*
38 * ref_head use rb tree, stored in ref_root->href.
39 * indexed by bytenr
40 */
Chris Mason56bec292009-03-13 10:10:06 -040041 struct rb_node rb_node;
42
Qu Wenruoc6fc2452015-03-30 17:03:00 +080043 /*data/tree ref use list, stored in ref_head->ref_list. */
44 struct list_head list;
45
Chris Mason56bec292009-03-13 10:10:06 -040046 /* the starting bytenr of the extent */
47 u64 bytenr;
48
Chris Mason56bec292009-03-13 10:10:06 -040049 /* the size of the extent */
50 u64 num_bytes;
51
Arne Jansen00f04b82011-09-14 12:37:00 +020052 /* seq number to keep track of insertion order */
53 u64 seq;
54
Chris Mason56bec292009-03-13 10:10:06 -040055 /* ref count on this data structure */
56 atomic_t refs;
57
58 /*
59 * how many refs is this entry adding or deleting. For
60 * head refs, this may be a negative number because it is keeping
61 * track of the total mods done to the reference count.
62 * For individual refs, this will always be a positive number
63 *
64 * It may be more than one, since it is possible for a single
65 * parent to have more than one ref on an extent
66 */
67 int ref_mod;
68
Yan Zheng5d4f98a2009-06-10 10:45:14 -040069 unsigned int action:8;
70 unsigned int type:8;
Chris Mason56bec292009-03-13 10:10:06 -040071 /* is this node still in the rbtree? */
Yan Zheng5d4f98a2009-06-10 10:45:14 -040072 unsigned int is_head:1;
Chris Mason56bec292009-03-13 10:10:06 -040073 unsigned int in_tree:1;
74};
75
Yan Zheng5d4f98a2009-06-10 10:45:14 -040076struct btrfs_delayed_extent_op {
77 struct btrfs_disk_key key;
David Sterba35b3ad52015-11-30 16:51:29 +010078 u8 level;
79 bool update_key;
80 bool update_flags;
81 bool is_data;
Yan Zheng5d4f98a2009-06-10 10:45:14 -040082 u64 flags_to_set;
Yan Zheng5d4f98a2009-06-10 10:45:14 -040083};
84
Chris Mason56bec292009-03-13 10:10:06 -040085/*
86 * the head refs are used to hold a lock on a given extent, which allows us
87 * to make sure that only one process is running the delayed refs
88 * at a time for a single extent. They also store the sum of all the
89 * reference count modifications we've queued up.
90 */
91struct btrfs_delayed_ref_head {
92 struct btrfs_delayed_ref_node node;
93
94 /*
95 * the mutex is held while running the refs, and it is also
96 * held when checking the sum of reference modifications.
97 */
98 struct mutex mutex;
99
Josef Bacikd7df2c72014-01-23 09:21:38 -0500100 spinlock_t lock;
Qu Wenruoc6fc2452015-03-30 17:03:00 +0800101 struct list_head ref_list;
Chris Masonc3e69d52009-03-13 10:17:05 -0400102
Liu Boc46effa2013-10-14 12:59:45 +0800103 struct rb_node href_node;
104
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400105 struct btrfs_delayed_extent_op *extent_op;
Josef Bacik12621332015-02-03 07:50:16 -0800106
107 /*
108 * This is used to track the final ref_mod from all the refs associated
109 * with this head ref, this is not adjusted as delayed refs are run,
110 * this is meant to track if we need to do the csum accounting or not.
111 */
112 int total_ref_mod;
113
Chris Mason56bec292009-03-13 10:10:06 -0400114 /*
Qu Wenruof64d5ca2015-09-08 17:08:36 +0800115 * For qgroup reserved space freeing.
116 *
117 * ref_root and reserved will be recorded after
118 * BTRFS_ADD_DELAYED_EXTENT is called.
119 * And will be used to free reserved qgroup space at
120 * run_delayed_refs() time.
121 */
122 u64 qgroup_ref_root;
123 u64 qgroup_reserved;
124
125 /*
Chris Mason56bec292009-03-13 10:10:06 -0400126 * when a new extent is allocated, it is just reserved in memory
127 * The actual extent isn't inserted into the extent allocation tree
128 * until the delayed ref is processed. must_insert_reserved is
129 * used to flag a delayed ref so the accounting can be updated
130 * when a full insert is done.
131 *
132 * It is possible the extent will be freed before it is ever
133 * inserted into the extent allocation tree. In this case
134 * we need to update the in ram accounting to properly reflect
135 * the free has happened.
136 */
137 unsigned int must_insert_reserved:1;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400138 unsigned int is_data:1;
Josef Bacikd7df2c72014-01-23 09:21:38 -0500139 unsigned int processing:1;
Chris Mason56bec292009-03-13 10:10:06 -0400140};
141
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400142struct btrfs_delayed_tree_ref {
Chris Mason56bec292009-03-13 10:10:06 -0400143 struct btrfs_delayed_ref_node node;
Arne Janseneebe0632011-09-14 14:01:24 +0200144 u64 root;
145 u64 parent;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400146 int level;
147};
Chris Mason56bec292009-03-13 10:10:06 -0400148
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400149struct btrfs_delayed_data_ref {
150 struct btrfs_delayed_ref_node node;
Arne Janseneebe0632011-09-14 14:01:24 +0200151 u64 root;
152 u64 parent;
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400153 u64 objectid;
154 u64 offset;
Chris Mason56bec292009-03-13 10:10:06 -0400155};
156
157struct btrfs_delayed_ref_root {
Liu Boc46effa2013-10-14 12:59:45 +0800158 /* head ref rbtree */
159 struct rb_root href_root;
160
Qu Wenruo3368d002015-04-16 14:34:17 +0800161 /* dirty extent records */
162 struct rb_root dirty_extent_root;
163
Chris Mason56bec292009-03-13 10:10:06 -0400164 /* this spin lock protects the rbtree and the entries inside */
165 spinlock_t lock;
166
167 /* how many delayed ref updates we've queued, used by the
168 * throttling code
169 */
Josef Bacikd7df2c72014-01-23 09:21:38 -0500170 atomic_t num_entries;
Chris Mason56bec292009-03-13 10:10:06 -0400171
Chris Masonc3e69d52009-03-13 10:17:05 -0400172 /* total number of head nodes in tree */
173 unsigned long num_heads;
174
175 /* total number of head nodes ready for processing */
176 unsigned long num_heads_ready;
177
Josef Bacik12621332015-02-03 07:50:16 -0800178 u64 pending_csums;
179
Chris Mason56bec292009-03-13 10:10:06 -0400180 /*
181 * set when the tree is flushing before a transaction commit,
182 * used by the throttling code to decide if new updates need
183 * to be run right away
184 */
185 int flushing;
Chris Masonc3e69d52009-03-13 10:17:05 -0400186
187 u64 run_delayed_start;
Qu Wenruo9086db82015-04-20 09:53:50 +0800188
189 /*
190 * To make qgroup to skip given root.
191 * This is for snapshot, as btrfs_qgroup_inherit() will manully
192 * modify counters for snapshot and its source, so we should skip
193 * the snapshot in new_root/old_roots or it will get calculated twice
194 */
195 u64 qgroup_to_skip;
Chris Mason56bec292009-03-13 10:10:06 -0400196};
197
Miao Xie78a61842012-11-21 02:21:28 +0000198extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
199extern struct kmem_cache *btrfs_delayed_tree_ref_cachep;
200extern struct kmem_cache *btrfs_delayed_data_ref_cachep;
201extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
202
203int btrfs_delayed_ref_init(void);
204void btrfs_delayed_ref_exit(void);
205
206static inline struct btrfs_delayed_extent_op *
207btrfs_alloc_delayed_extent_op(void)
208{
209 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
210}
211
212static inline void
213btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
214{
215 if (op)
216 kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
217}
218
Chris Mason56bec292009-03-13 10:10:06 -0400219static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
220{
221 WARN_ON(atomic_read(&ref->refs) == 0);
222 if (atomic_dec_and_test(&ref->refs)) {
223 WARN_ON(ref->in_tree);
Miao Xie78a61842012-11-21 02:21:28 +0000224 switch (ref->type) {
225 case BTRFS_TREE_BLOCK_REF_KEY:
226 case BTRFS_SHARED_BLOCK_REF_KEY:
227 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
228 break;
229 case BTRFS_EXTENT_DATA_REF_KEY:
230 case BTRFS_SHARED_DATA_REF_KEY:
231 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
232 break;
233 case 0:
234 kmem_cache_free(btrfs_delayed_ref_head_cachep, ref);
235 break;
236 default:
237 BUG();
238 }
Chris Mason56bec292009-03-13 10:10:06 -0400239 }
240}
241
Arne Jansen66d7e7f2011-09-12 15:26:38 +0200242int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
243 struct btrfs_trans_handle *trans,
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400244 u64 bytenr, u64 num_bytes, u64 parent,
245 u64 ref_root, int level, int action,
Filipe Mananab06c4bf2015-10-23 07:52:54 +0100246 struct btrfs_delayed_extent_op *extent_op);
Arne Jansen66d7e7f2011-09-12 15:26:38 +0200247int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
248 struct btrfs_trans_handle *trans,
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400249 u64 bytenr, u64 num_bytes,
250 u64 parent, u64 ref_root,
Qu Wenruo5846a3c2015-10-26 14:11:18 +0800251 u64 owner, u64 offset, u64 reserved, int action,
Filipe Mananab06c4bf2015-10-23 07:52:54 +0100252 struct btrfs_delayed_extent_op *extent_op);
Qu Wenruof64d5ca2015-09-08 17:08:36 +0800253int btrfs_add_delayed_qgroup_reserve(struct btrfs_fs_info *fs_info,
254 struct btrfs_trans_handle *trans,
255 u64 ref_root, u64 bytenr, u64 num_bytes);
Arne Jansen66d7e7f2011-09-12 15:26:38 +0200256int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
257 struct btrfs_trans_handle *trans,
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400258 u64 bytenr, u64 num_bytes,
259 struct btrfs_delayed_extent_op *extent_op);
Josef Bacikae1e2062012-08-07 16:00:32 -0400260void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
261 struct btrfs_fs_info *fs_info,
262 struct btrfs_delayed_ref_root *delayed_refs,
263 struct btrfs_delayed_ref_head *head);
Chris Mason56bec292009-03-13 10:10:06 -0400264
Chris Mason1887be62009-03-13 10:11:24 -0400265struct btrfs_delayed_ref_head *
266btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr);
Chris Masonc3e69d52009-03-13 10:17:05 -0400267int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
268 struct btrfs_delayed_ref_head *head);
Miao Xie093486c2012-12-19 08:10:10 +0000269static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
270{
271 mutex_unlock(&head->mutex);
272}
273
Josef Bacikd7df2c72014-01-23 09:21:38 -0500274
275struct btrfs_delayed_ref_head *
276btrfs_select_ref_head(struct btrfs_trans_handle *trans);
Arne Jansen00f04b82011-09-14 12:37:00 +0200277
Jan Schmidt097b8a72012-06-21 11:08:04 +0200278int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
279 struct btrfs_delayed_ref_root *delayed_refs,
Arne Jansen00f04b82011-09-14 12:37:00 +0200280 u64 seq);
281
282/*
Chris Mason56bec292009-03-13 10:10:06 -0400283 * a node might live in a head or a regular ref, this lets you
284 * test for the proper type to use.
285 */
286static int btrfs_delayed_ref_is_head(struct btrfs_delayed_ref_node *node)
287{
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400288 return node->is_head;
Chris Mason56bec292009-03-13 10:10:06 -0400289}
290
291/*
292 * helper functions to cast a node into its container
293 */
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400294static inline struct btrfs_delayed_tree_ref *
295btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node)
Chris Mason56bec292009-03-13 10:10:06 -0400296{
297 WARN_ON(btrfs_delayed_ref_is_head(node));
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400298 return container_of(node, struct btrfs_delayed_tree_ref, node);
299}
Chris Mason56bec292009-03-13 10:10:06 -0400300
Yan Zheng5d4f98a2009-06-10 10:45:14 -0400301static inline struct btrfs_delayed_data_ref *
302btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node)
303{
304 WARN_ON(btrfs_delayed_ref_is_head(node));
305 return container_of(node, struct btrfs_delayed_data_ref, node);
Chris Mason56bec292009-03-13 10:10:06 -0400306}
307
308static inline struct btrfs_delayed_ref_head *
309btrfs_delayed_node_to_head(struct btrfs_delayed_ref_node *node)
310{
311 WARN_ON(!btrfs_delayed_ref_is_head(node));
312 return container_of(node, struct btrfs_delayed_ref_head, node);
Chris Mason56bec292009-03-13 10:10:06 -0400313}
314#endif