| 1 | /* |
| 2 | * Copyright (C) 2007 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 | |
| 19 | #ifndef __BTRFS_I__ |
| 20 | #define __BTRFS_I__ |
| 21 | |
| 22 | #include <linux/hash.h> |
| 23 | #include "extent_map.h" |
| 24 | #include "extent_io.h" |
| 25 | #include "ordered-data.h" |
| 26 | #include "delayed-inode.h" |
| 27 | |
| 28 | /* |
| 29 | * ordered_data_close is set by truncate when a file that used |
| 30 | * to have good data has been truncated to zero. When it is set |
| 31 | * the btrfs file release call will add this inode to the |
| 32 | * ordered operations list so that we make sure to flush out any |
| 33 | * new data the application may have written before commit. |
| 34 | */ |
| 35 | #define BTRFS_INODE_ORDERED_DATA_CLOSE 0 |
| 36 | #define BTRFS_INODE_ORPHAN_META_RESERVED 1 |
| 37 | #define BTRFS_INODE_DUMMY 2 |
| 38 | #define BTRFS_INODE_IN_DEFRAG 3 |
| 39 | #define BTRFS_INODE_DELALLOC_META_RESERVED 4 |
| 40 | #define BTRFS_INODE_HAS_ORPHAN_ITEM 5 |
| 41 | #define BTRFS_INODE_HAS_ASYNC_EXTENT 6 |
| 42 | #define BTRFS_INODE_NEEDS_FULL_SYNC 7 |
| 43 | #define BTRFS_INODE_COPY_EVERYTHING 8 |
| 44 | #define BTRFS_INODE_IN_DELALLOC_LIST 9 |
| 45 | #define BTRFS_INODE_READDIO_NEED_LOCK 10 |
| 46 | #define BTRFS_INODE_HAS_PROPS 11 |
| 47 | /* |
| 48 | * The following 3 bits are meant only for the btree inode. |
| 49 | * When any of them is set, it means an error happened while writing an |
| 50 | * extent buffer belonging to: |
| 51 | * 1) a non-log btree |
| 52 | * 2) a log btree and first log sub-transaction |
| 53 | * 3) a log btree and second log sub-transaction |
| 54 | */ |
| 55 | #define BTRFS_INODE_BTREE_ERR 12 |
| 56 | #define BTRFS_INODE_BTREE_LOG1_ERR 13 |
| 57 | #define BTRFS_INODE_BTREE_LOG2_ERR 14 |
| 58 | |
| 59 | /* in memory btrfs inode */ |
| 60 | struct btrfs_inode { |
| 61 | /* which subvolume this inode belongs to */ |
| 62 | struct btrfs_root *root; |
| 63 | |
| 64 | /* key used to find this inode on disk. This is used by the code |
| 65 | * to read in roots of subvolumes |
| 66 | */ |
| 67 | struct btrfs_key location; |
| 68 | |
| 69 | /* |
| 70 | * Lock for counters and all fields used to determine if the inode is in |
| 71 | * the log or not (last_trans, last_sub_trans, last_log_commit, |
| 72 | * logged_trans). |
| 73 | */ |
| 74 | spinlock_t lock; |
| 75 | |
| 76 | /* the extent_tree has caches of all the extent mappings to disk */ |
| 77 | struct extent_map_tree extent_tree; |
| 78 | |
| 79 | /* the io_tree does range state (DIRTY, LOCKED etc) */ |
| 80 | struct extent_io_tree io_tree; |
| 81 | |
| 82 | /* special utility tree used to record which mirrors have already been |
| 83 | * tried when checksums fail for a given block |
| 84 | */ |
| 85 | struct extent_io_tree io_failure_tree; |
| 86 | |
| 87 | /* held while logging the inode in tree-log.c */ |
| 88 | struct mutex log_mutex; |
| 89 | |
| 90 | /* held while doing delalloc reservations */ |
| 91 | struct mutex delalloc_mutex; |
| 92 | |
| 93 | /* used to order data wrt metadata */ |
| 94 | struct btrfs_ordered_inode_tree ordered_tree; |
| 95 | |
| 96 | /* list of all the delalloc inodes in the FS. There are times we need |
| 97 | * to write all the delalloc pages to disk, and this list is used |
| 98 | * to walk them all. |
| 99 | */ |
| 100 | struct list_head delalloc_inodes; |
| 101 | |
| 102 | /* node for the red-black tree that links inodes in subvolume root */ |
| 103 | struct rb_node rb_node; |
| 104 | |
| 105 | unsigned long runtime_flags; |
| 106 | |
| 107 | /* Keep track of who's O_SYNC/fsyncing currently */ |
| 108 | atomic_t sync_writers; |
| 109 | |
| 110 | /* full 64 bit generation number, struct vfs_inode doesn't have a big |
| 111 | * enough field for this. |
| 112 | */ |
| 113 | u64 generation; |
| 114 | |
| 115 | /* |
| 116 | * transid of the trans_handle that last modified this inode |
| 117 | */ |
| 118 | u64 last_trans; |
| 119 | |
| 120 | /* |
| 121 | * transid that last logged this inode |
| 122 | */ |
| 123 | u64 logged_trans; |
| 124 | |
| 125 | /* |
| 126 | * log transid when this inode was last modified |
| 127 | */ |
| 128 | int last_sub_trans; |
| 129 | |
| 130 | /* a local copy of root's last_log_commit */ |
| 131 | int last_log_commit; |
| 132 | |
| 133 | /* total number of bytes pending delalloc, used by stat to calc the |
| 134 | * real block usage of the file |
| 135 | */ |
| 136 | u64 delalloc_bytes; |
| 137 | |
| 138 | /* |
| 139 | * total number of bytes pending defrag, used by stat to check whether |
| 140 | * it needs COW. |
| 141 | */ |
| 142 | u64 defrag_bytes; |
| 143 | |
| 144 | /* |
| 145 | * the size of the file stored in the metadata on disk. data=ordered |
| 146 | * means the in-memory i_size might be larger than the size on disk |
| 147 | * because not all the blocks are written yet. |
| 148 | */ |
| 149 | u64 disk_i_size; |
| 150 | |
| 151 | /* |
| 152 | * if this is a directory then index_cnt is the counter for the index |
| 153 | * number for new files that are created |
| 154 | */ |
| 155 | u64 index_cnt; |
| 156 | |
| 157 | /* Cache the directory index number to speed the dir/file remove */ |
| 158 | u64 dir_index; |
| 159 | |
| 160 | /* the fsync log has some corner cases that mean we have to check |
| 161 | * directories to see if any unlinks have been done before |
| 162 | * the directory was logged. See tree-log.c for all the |
| 163 | * details |
| 164 | */ |
| 165 | u64 last_unlink_trans; |
| 166 | |
| 167 | /* |
| 168 | * Number of bytes outstanding that are going to need csums. This is |
| 169 | * used in ENOSPC accounting. |
| 170 | */ |
| 171 | u64 csum_bytes; |
| 172 | |
| 173 | /* flags field from the on disk inode */ |
| 174 | u32 flags; |
| 175 | |
| 176 | /* |
| 177 | * Counters to keep track of the number of extent item's we may use due |
| 178 | * to delalloc and such. outstanding_extents is the number of extent |
| 179 | * items we think we'll end up using, and reserved_extents is the number |
| 180 | * of extent items we've reserved metadata for. |
| 181 | */ |
| 182 | unsigned outstanding_extents; |
| 183 | unsigned reserved_extents; |
| 184 | |
| 185 | /* |
| 186 | * always compress this one file |
| 187 | */ |
| 188 | unsigned force_compress; |
| 189 | |
| 190 | struct btrfs_delayed_node *delayed_node; |
| 191 | |
| 192 | /* File creation time. */ |
| 193 | struct timespec i_otime; |
| 194 | |
| 195 | struct inode vfs_inode; |
| 196 | }; |
| 197 | |
| 198 | extern unsigned char btrfs_filetype_table[]; |
| 199 | |
| 200 | static inline struct btrfs_inode *BTRFS_I(struct inode *inode) |
| 201 | { |
| 202 | return container_of(inode, struct btrfs_inode, vfs_inode); |
| 203 | } |
| 204 | |
| 205 | static inline unsigned long btrfs_inode_hash(u64 objectid, |
| 206 | const struct btrfs_root *root) |
| 207 | { |
| 208 | u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME); |
| 209 | |
| 210 | #if BITS_PER_LONG == 32 |
| 211 | h = (h >> 32) ^ (h & 0xffffffff); |
| 212 | #endif |
| 213 | |
| 214 | return (unsigned long)h; |
| 215 | } |
| 216 | |
| 217 | static inline void btrfs_insert_inode_hash(struct inode *inode) |
| 218 | { |
| 219 | unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root); |
| 220 | |
| 221 | __insert_inode_hash(inode, h); |
| 222 | } |
| 223 | |
| 224 | static inline u64 btrfs_ino(struct inode *inode) |
| 225 | { |
| 226 | u64 ino = BTRFS_I(inode)->location.objectid; |
| 227 | |
| 228 | /* |
| 229 | * !ino: btree_inode |
| 230 | * type == BTRFS_ROOT_ITEM_KEY: subvol dir |
| 231 | */ |
| 232 | if (!ino || BTRFS_I(inode)->location.type == BTRFS_ROOT_ITEM_KEY) |
| 233 | ino = inode->i_ino; |
| 234 | return ino; |
| 235 | } |
| 236 | |
| 237 | static inline void btrfs_i_size_write(struct inode *inode, u64 size) |
| 238 | { |
| 239 | i_size_write(inode, size); |
| 240 | BTRFS_I(inode)->disk_i_size = size; |
| 241 | } |
| 242 | |
| 243 | static inline bool btrfs_is_free_space_inode(struct inode *inode) |
| 244 | { |
| 245 | struct btrfs_root *root = BTRFS_I(inode)->root; |
| 246 | |
| 247 | if (root == root->fs_info->tree_root && |
| 248 | btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) |
| 249 | return true; |
| 250 | if (BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID) |
| 251 | return true; |
| 252 | return false; |
| 253 | } |
| 254 | |
| 255 | static inline int btrfs_inode_in_log(struct inode *inode, u64 generation) |
| 256 | { |
| 257 | int ret = 0; |
| 258 | |
| 259 | spin_lock(&BTRFS_I(inode)->lock); |
| 260 | if (BTRFS_I(inode)->logged_trans == generation && |
| 261 | BTRFS_I(inode)->last_sub_trans <= |
| 262 | BTRFS_I(inode)->last_log_commit && |
| 263 | BTRFS_I(inode)->last_sub_trans <= |
| 264 | BTRFS_I(inode)->root->last_log_commit) { |
| 265 | /* |
| 266 | * After a ranged fsync we might have left some extent maps |
| 267 | * (that fall outside the fsync's range). So return false |
| 268 | * here if the list isn't empty, to make sure btrfs_log_inode() |
| 269 | * will be called and process those extent maps. |
| 270 | */ |
| 271 | smp_mb(); |
| 272 | if (list_empty(&BTRFS_I(inode)->extent_tree.modified_extents)) |
| 273 | ret = 1; |
| 274 | } |
| 275 | spin_unlock(&BTRFS_I(inode)->lock); |
| 276 | return ret; |
| 277 | } |
| 278 | |
| 279 | #define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1 |
| 280 | |
| 281 | struct btrfs_dio_private { |
| 282 | struct inode *inode; |
| 283 | unsigned long flags; |
| 284 | u64 logical_offset; |
| 285 | u64 disk_bytenr; |
| 286 | u64 bytes; |
| 287 | void *private; |
| 288 | |
| 289 | /* number of bios pending for this dio */ |
| 290 | atomic_t pending_bios; |
| 291 | |
| 292 | /* IO errors */ |
| 293 | int errors; |
| 294 | |
| 295 | /* orig_bio is our btrfs_io_bio */ |
| 296 | struct bio *orig_bio; |
| 297 | |
| 298 | /* dio_bio came from fs/direct-io.c */ |
| 299 | struct bio *dio_bio; |
| 300 | |
| 301 | /* |
| 302 | * The original bio may be splited to several sub-bios, this is |
| 303 | * done during endio of sub-bios |
| 304 | */ |
| 305 | int (*subio_endio)(struct inode *, struct btrfs_io_bio *, int); |
| 306 | }; |
| 307 | |
| 308 | /* |
| 309 | * Disable DIO read nolock optimization, so new dio readers will be forced |
| 310 | * to grab i_mutex. It is used to avoid the endless truncate due to |
| 311 | * nonlocked dio read. |
| 312 | */ |
| 313 | static inline void btrfs_inode_block_unlocked_dio(struct inode *inode) |
| 314 | { |
| 315 | set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags); |
| 316 | smp_mb(); |
| 317 | } |
| 318 | |
| 319 | static inline void btrfs_inode_resume_unlocked_dio(struct inode *inode) |
| 320 | { |
| 321 | smp_mb__before_atomic(); |
| 322 | clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, |
| 323 | &BTRFS_I(inode)->runtime_flags); |
| 324 | } |
| 325 | |
| 326 | bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end); |
| 327 | |
| 328 | #endif |