2 * Copyright (C) 2008 Oracle. All rights reserved.
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.
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.
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.
18 #ifndef __DELAYED_REF__
19 #define __DELAYED_REF__
21 /* these are the possible values of struct btrfs_delayed_ref_node->action */
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 */
25 #define BTRFS_UPDATE_DELAYED_HEAD 4 /* not changing ref count on head ref */
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.
36 struct btrfs_delayed_ref_node
{
38 * ref_head use rb tree, stored in ref_root->href.
41 struct rb_node rb_node
;
43 /*data/tree ref use list, stored in ref_head->ref_list. */
44 struct list_head list
;
46 /* the starting bytenr of the extent */
49 /* the size of the extent */
52 /* seq number to keep track of insertion order */
55 /* ref count on this data structure */
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
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
69 unsigned int action
:8;
71 unsigned int no_quota
:1;
72 /* is this node still in the rbtree? */
73 unsigned int is_head
:1;
74 unsigned int in_tree
:1;
77 struct btrfs_delayed_extent_op
{
78 struct btrfs_disk_key key
;
81 unsigned int update_key
:1;
82 unsigned int update_flags
:1;
83 unsigned int is_data
:1;
87 * the head refs are used to hold a lock on a given extent, which allows us
88 * to make sure that only one process is running the delayed refs
89 * at a time for a single extent. They also store the sum of all the
90 * reference count modifications we've queued up.
92 struct btrfs_delayed_ref_head
{
93 struct btrfs_delayed_ref_node node
;
96 * the mutex is held while running the refs, and it is also
97 * held when checking the sum of reference modifications.
102 struct list_head ref_list
;
104 struct rb_node href_node
;
106 struct btrfs_delayed_extent_op
*extent_op
;
109 * This is used to track the final ref_mod from all the refs associated
110 * with this head ref, this is not adjusted as delayed refs are run,
111 * this is meant to track if we need to do the csum accounting or not.
116 * For qgroup reserved space freeing.
118 * ref_root and reserved will be recorded after
119 * BTRFS_ADD_DELAYED_EXTENT is called.
120 * And will be used to free reserved qgroup space at
121 * run_delayed_refs() time.
127 * when a new extent is allocated, it is just reserved in memory
128 * The actual extent isn't inserted into the extent allocation tree
129 * until the delayed ref is processed. must_insert_reserved is
130 * used to flag a delayed ref so the accounting can be updated
131 * when a full insert is done.
133 * It is possible the extent will be freed before it is ever
134 * inserted into the extent allocation tree. In this case
135 * we need to update the in ram accounting to properly reflect
136 * the free has happened.
138 unsigned int must_insert_reserved
:1;
139 unsigned int is_data
:1;
140 unsigned int processing
:1;
143 struct btrfs_delayed_tree_ref
{
144 struct btrfs_delayed_ref_node node
;
150 struct btrfs_delayed_data_ref
{
151 struct btrfs_delayed_ref_node node
;
158 struct btrfs_delayed_ref_root
{
159 /* head ref rbtree */
160 struct rb_root href_root
;
162 /* dirty extent records */
163 struct rb_root dirty_extent_root
;
165 /* this spin lock protects the rbtree and the entries inside */
168 /* how many delayed ref updates we've queued, used by the
171 atomic_t num_entries
;
173 /* total number of head nodes in tree */
174 unsigned long num_heads
;
176 /* total number of head nodes ready for processing */
177 unsigned long num_heads_ready
;
182 * set when the tree is flushing before a transaction commit,
183 * used by the throttling code to decide if new updates need
184 * to be run right away
188 u64 run_delayed_start
;
191 * To make qgroup to skip given root.
192 * This is for snapshot, as btrfs_qgroup_inherit() will manully
193 * modify counters for snapshot and its source, so we should skip
194 * the snapshot in new_root/old_roots or it will get calculated twice
199 extern struct kmem_cache
*btrfs_delayed_ref_head_cachep
;
200 extern struct kmem_cache
*btrfs_delayed_tree_ref_cachep
;
201 extern struct kmem_cache
*btrfs_delayed_data_ref_cachep
;
202 extern struct kmem_cache
*btrfs_delayed_extent_op_cachep
;
204 int btrfs_delayed_ref_init(void);
205 void btrfs_delayed_ref_exit(void);
207 static inline struct btrfs_delayed_extent_op
*
208 btrfs_alloc_delayed_extent_op(void)
210 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep
, GFP_NOFS
);
214 btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op
*op
)
217 kmem_cache_free(btrfs_delayed_extent_op_cachep
, op
);
220 static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node
*ref
)
222 WARN_ON(atomic_read(&ref
->refs
) == 0);
223 if (atomic_dec_and_test(&ref
->refs
)) {
224 WARN_ON(ref
->in_tree
);
226 case BTRFS_TREE_BLOCK_REF_KEY
:
227 case BTRFS_SHARED_BLOCK_REF_KEY
:
228 kmem_cache_free(btrfs_delayed_tree_ref_cachep
, ref
);
230 case BTRFS_EXTENT_DATA_REF_KEY
:
231 case BTRFS_SHARED_DATA_REF_KEY
:
232 kmem_cache_free(btrfs_delayed_data_ref_cachep
, ref
);
235 kmem_cache_free(btrfs_delayed_ref_head_cachep
, ref
);
243 int btrfs_add_delayed_tree_ref(struct btrfs_fs_info
*fs_info
,
244 struct btrfs_trans_handle
*trans
,
245 u64 bytenr
, u64 num_bytes
, u64 parent
,
246 u64 ref_root
, int level
, int action
,
247 struct btrfs_delayed_extent_op
*extent_op
,
249 int btrfs_add_delayed_data_ref(struct btrfs_fs_info
*fs_info
,
250 struct btrfs_trans_handle
*trans
,
251 u64 bytenr
, u64 num_bytes
,
252 u64 parent
, u64 ref_root
,
253 u64 owner
, u64 offset
, int action
,
254 struct btrfs_delayed_extent_op
*extent_op
,
256 int btrfs_add_delayed_qgroup_reserve(struct btrfs_fs_info
*fs_info
,
257 struct btrfs_trans_handle
*trans
,
258 u64 ref_root
, u64 bytenr
, u64 num_bytes
);
259 int btrfs_add_delayed_extent_op(struct btrfs_fs_info
*fs_info
,
260 struct btrfs_trans_handle
*trans
,
261 u64 bytenr
, u64 num_bytes
,
262 struct btrfs_delayed_extent_op
*extent_op
);
263 void btrfs_merge_delayed_refs(struct btrfs_trans_handle
*trans
,
264 struct btrfs_fs_info
*fs_info
,
265 struct btrfs_delayed_ref_root
*delayed_refs
,
266 struct btrfs_delayed_ref_head
*head
);
268 struct btrfs_delayed_ref_head
*
269 btrfs_find_delayed_ref_head(struct btrfs_trans_handle
*trans
, u64 bytenr
);
270 int btrfs_delayed_ref_lock(struct btrfs_trans_handle
*trans
,
271 struct btrfs_delayed_ref_head
*head
);
272 static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head
*head
)
274 mutex_unlock(&head
->mutex
);
278 struct btrfs_delayed_ref_head
*
279 btrfs_select_ref_head(struct btrfs_trans_handle
*trans
);
281 int btrfs_check_delayed_seq(struct btrfs_fs_info
*fs_info
,
282 struct btrfs_delayed_ref_root
*delayed_refs
,
286 * a node might live in a head or a regular ref, this lets you
287 * test for the proper type to use.
289 static int btrfs_delayed_ref_is_head(struct btrfs_delayed_ref_node
*node
)
291 return node
->is_head
;
295 * helper functions to cast a node into its container
297 static inline struct btrfs_delayed_tree_ref
*
298 btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node
*node
)
300 WARN_ON(btrfs_delayed_ref_is_head(node
));
301 return container_of(node
, struct btrfs_delayed_tree_ref
, node
);
304 static inline struct btrfs_delayed_data_ref
*
305 btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node
*node
)
307 WARN_ON(btrfs_delayed_ref_is_head(node
));
308 return container_of(node
, struct btrfs_delayed_data_ref
, node
);
311 static inline struct btrfs_delayed_ref_head
*
312 btrfs_delayed_node_to_head(struct btrfs_delayed_ref_node
*node
)
314 WARN_ON(!btrfs_delayed_ref_is_head(node
));
315 return container_of(node
, struct btrfs_delayed_ref_head
, node
);