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 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 */ 26 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 */ 36 struct btrfs_delayed_ref_node { 37 /* 38 * ref_head use rb tree, stored in ref_root->href. 39 * indexed by bytenr 40 */ 41 struct rb_node rb_node; 42 43 /*data/tree ref use list, stored in ref_head->ref_list. */ 44 struct list_head list; 45 46 /* the starting bytenr of the extent */ 47 u64 bytenr; 48 49 /* the size of the extent */ 50 u64 num_bytes; 51 52 /* seq number to keep track of insertion order */ 53 u64 seq; 54 55 /* 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 69 unsigned int action:8; 70 unsigned int type:8; 71 /* is this node still in the rbtree? */ 72 unsigned int is_head:1; 73 unsigned int in_tree:1; 74 }; 75 76 struct btrfs_delayed_extent_op { 77 struct btrfs_disk_key key; 78 u8 level; 79 bool update_key; 80 bool update_flags; 81 bool is_data; 82 u64 flags_to_set; 83 }; 84 85 /* 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 */ 91 struct 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; 99 100 spinlock_t lock; 101 struct list_head ref_list; 102 103 struct rb_node href_node; 104 105 struct btrfs_delayed_extent_op *extent_op; 106 107 /* 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; 113 114 /* 115 * 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 /* 126 * 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; 138 unsigned int is_data:1; 139 unsigned int processing:1; 140 }; 141 142 struct btrfs_delayed_tree_ref { 143 struct btrfs_delayed_ref_node node; 144 u64 root; 145 u64 parent; 146 int level; 147 }; 148 149 struct btrfs_delayed_data_ref { 150 struct btrfs_delayed_ref_node node; 151 u64 root; 152 u64 parent; 153 u64 objectid; 154 u64 offset; 155 }; 156 157 struct btrfs_delayed_ref_root { 158 /* head ref rbtree */ 159 struct rb_root href_root; 160 161 /* dirty extent records */ 162 struct rb_root dirty_extent_root; 163 164 /* 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 */ 170 atomic_t num_entries; 171 172 /* 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; 177 178 u64 pending_csums; 179 180 /* 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; 186 187 u64 run_delayed_start; 188 189 /* 190 * To make qgroup to skip given root. 191 * This is for snapshot, as btrfs_qgroup_inherit() will manually 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; 196 }; 197 198 extern struct kmem_cache *btrfs_delayed_ref_head_cachep; 199 extern struct kmem_cache *btrfs_delayed_tree_ref_cachep; 200 extern struct kmem_cache *btrfs_delayed_data_ref_cachep; 201 extern struct kmem_cache *btrfs_delayed_extent_op_cachep; 202 203 int btrfs_delayed_ref_init(void); 204 void btrfs_delayed_ref_exit(void); 205 206 static inline struct btrfs_delayed_extent_op * 207 btrfs_alloc_delayed_extent_op(void) 208 { 209 return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS); 210 } 211 212 static inline void 213 btrfs_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 } 218 219 static 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); 224 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 } 239 } 240 } 241 242 int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info, 243 struct btrfs_trans_handle *trans, 244 u64 bytenr, u64 num_bytes, u64 parent, 245 u64 ref_root, int level, int action, 246 struct btrfs_delayed_extent_op *extent_op); 247 int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info, 248 struct btrfs_trans_handle *trans, 249 u64 bytenr, u64 num_bytes, 250 u64 parent, u64 ref_root, 251 u64 owner, u64 offset, u64 reserved, int action, 252 struct btrfs_delayed_extent_op *extent_op); 253 int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info, 254 struct btrfs_trans_handle *trans, 255 u64 bytenr, u64 num_bytes, 256 struct btrfs_delayed_extent_op *extent_op); 257 void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, 258 struct btrfs_fs_info *fs_info, 259 struct btrfs_delayed_ref_root *delayed_refs, 260 struct btrfs_delayed_ref_head *head); 261 262 struct btrfs_delayed_ref_head * 263 btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr); 264 int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans, 265 struct btrfs_delayed_ref_head *head); 266 static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head) 267 { 268 mutex_unlock(&head->mutex); 269 } 270 271 272 struct btrfs_delayed_ref_head * 273 btrfs_select_ref_head(struct btrfs_trans_handle *trans); 274 275 int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, 276 struct btrfs_delayed_ref_root *delayed_refs, 277 u64 seq); 278 279 /* 280 * a node might live in a head or a regular ref, this lets you 281 * test for the proper type to use. 282 */ 283 static int btrfs_delayed_ref_is_head(struct btrfs_delayed_ref_node *node) 284 { 285 return node->is_head; 286 } 287 288 /* 289 * helper functions to cast a node into its container 290 */ 291 static inline struct btrfs_delayed_tree_ref * 292 btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node) 293 { 294 WARN_ON(btrfs_delayed_ref_is_head(node)); 295 return container_of(node, struct btrfs_delayed_tree_ref, node); 296 } 297 298 static inline struct btrfs_delayed_data_ref * 299 btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node) 300 { 301 WARN_ON(btrfs_delayed_ref_is_head(node)); 302 return container_of(node, struct btrfs_delayed_data_ref, node); 303 } 304 305 static inline struct btrfs_delayed_ref_head * 306 btrfs_delayed_node_to_head(struct btrfs_delayed_ref_node *node) 307 { 308 WARN_ON(!btrfs_delayed_ref_is_head(node)); 309 return container_of(node, struct btrfs_delayed_ref_head, node); 310 } 311 #endif 312