1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Copyright (C) 2007 Oracle. All rights reserved. 4 */ 5 6 #ifndef BTRFS_INODE_H 7 #define BTRFS_INODE_H 8 9 #include <linux/hash.h> 10 #include "extent_map.h" 11 #include "extent_io.h" 12 #include "ordered-data.h" 13 #include "delayed-inode.h" 14 15 /* 16 * ordered_data_close is set by truncate when a file that used 17 * to have good data has been truncated to zero. When it is set 18 * the btrfs file release call will add this inode to the 19 * ordered operations list so that we make sure to flush out any 20 * new data the application may have written before commit. 21 */ 22 #define BTRFS_INODE_ORDERED_DATA_CLOSE 0 23 #define BTRFS_INODE_ORPHAN_META_RESERVED 1 24 #define BTRFS_INODE_DUMMY 2 25 #define BTRFS_INODE_IN_DEFRAG 3 26 #define BTRFS_INODE_HAS_ASYNC_EXTENT 5 27 #define BTRFS_INODE_NEEDS_FULL_SYNC 6 28 #define BTRFS_INODE_COPY_EVERYTHING 7 29 #define BTRFS_INODE_IN_DELALLOC_LIST 8 30 #define BTRFS_INODE_READDIO_NEED_LOCK 9 31 #define BTRFS_INODE_HAS_PROPS 10 32 33 /* in memory btrfs inode */ 34 struct btrfs_inode { 35 /* which subvolume this inode belongs to */ 36 struct btrfs_root *root; 37 38 /* key used to find this inode on disk. This is used by the code 39 * to read in roots of subvolumes 40 */ 41 struct btrfs_key location; 42 43 /* 44 * Lock for counters and all fields used to determine if the inode is in 45 * the log or not (last_trans, last_sub_trans, last_log_commit, 46 * logged_trans). 47 */ 48 spinlock_t lock; 49 50 /* the extent_tree has caches of all the extent mappings to disk */ 51 struct extent_map_tree extent_tree; 52 53 /* the io_tree does range state (DIRTY, LOCKED etc) */ 54 struct extent_io_tree io_tree; 55 56 /* special utility tree used to record which mirrors have already been 57 * tried when checksums fail for a given block 58 */ 59 struct extent_io_tree io_failure_tree; 60 61 /* held while logging the inode in tree-log.c */ 62 struct mutex log_mutex; 63 64 /* held while doing delalloc reservations */ 65 struct mutex delalloc_mutex; 66 67 /* used to order data wrt metadata */ 68 struct btrfs_ordered_inode_tree ordered_tree; 69 70 /* list of all the delalloc inodes in the FS. There are times we need 71 * to write all the delalloc pages to disk, and this list is used 72 * to walk them all. 73 */ 74 struct list_head delalloc_inodes; 75 76 /* node for the red-black tree that links inodes in subvolume root */ 77 struct rb_node rb_node; 78 79 unsigned long runtime_flags; 80 81 /* Keep track of who's O_SYNC/fsyncing currently */ 82 atomic_t sync_writers; 83 84 /* full 64 bit generation number, struct vfs_inode doesn't have a big 85 * enough field for this. 86 */ 87 u64 generation; 88 89 /* 90 * transid of the trans_handle that last modified this inode 91 */ 92 u64 last_trans; 93 94 /* 95 * transid that last logged this inode 96 */ 97 u64 logged_trans; 98 99 /* 100 * log transid when this inode was last modified 101 */ 102 int last_sub_trans; 103 104 /* a local copy of root's last_log_commit */ 105 int last_log_commit; 106 107 /* total number of bytes pending delalloc, used by stat to calc the 108 * real block usage of the file 109 */ 110 u64 delalloc_bytes; 111 112 /* 113 * Total number of bytes pending delalloc that fall within a file 114 * range that is either a hole or beyond EOF (and no prealloc extent 115 * exists in the range). This is always <= delalloc_bytes. 116 */ 117 u64 new_delalloc_bytes; 118 119 /* 120 * total number of bytes pending defrag, used by stat to check whether 121 * it needs COW. 122 */ 123 u64 defrag_bytes; 124 125 /* 126 * the size of the file stored in the metadata on disk. data=ordered 127 * means the in-memory i_size might be larger than the size on disk 128 * because not all the blocks are written yet. 129 */ 130 u64 disk_i_size; 131 132 /* 133 * if this is a directory then index_cnt is the counter for the index 134 * number for new files that are created 135 */ 136 u64 index_cnt; 137 138 /* Cache the directory index number to speed the dir/file remove */ 139 u64 dir_index; 140 141 /* the fsync log has some corner cases that mean we have to check 142 * directories to see if any unlinks have been done before 143 * the directory was logged. See tree-log.c for all the 144 * details 145 */ 146 u64 last_unlink_trans; 147 148 /* 149 * Number of bytes outstanding that are going to need csums. This is 150 * used in ENOSPC accounting. 151 */ 152 u64 csum_bytes; 153 154 /* flags field from the on disk inode */ 155 u32 flags; 156 157 /* 158 * Counters to keep track of the number of extent item's we may use due 159 * to delalloc and such. outstanding_extents is the number of extent 160 * items we think we'll end up using, and reserved_extents is the number 161 * of extent items we've reserved metadata for. 162 */ 163 unsigned outstanding_extents; 164 165 struct btrfs_block_rsv block_rsv; 166 167 /* 168 * Cached values of inode properties 169 */ 170 unsigned prop_compress; /* per-file compression algorithm */ 171 /* 172 * Force compression on the file using the defrag ioctl, could be 173 * different from prop_compress and takes precedence if set 174 */ 175 unsigned defrag_compress; 176 177 struct btrfs_delayed_node *delayed_node; 178 179 /* File creation time. */ 180 struct timespec i_otime; 181 182 /* Hook into fs_info->delayed_iputs */ 183 struct list_head delayed_iput; 184 185 /* 186 * To avoid races between lockless (i_mutex not held) direct IO writes 187 * and concurrent fsync requests. Direct IO writes must acquire read 188 * access on this semaphore for creating an extent map and its 189 * corresponding ordered extent. The fast fsync path must acquire write 190 * access on this semaphore before it collects ordered extents and 191 * extent maps. 192 */ 193 struct rw_semaphore dio_sem; 194 195 struct inode vfs_inode; 196 }; 197 198 extern unsigned char btrfs_filetype_table[]; 199 200 static inline struct btrfs_inode *BTRFS_I(const 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(const struct btrfs_inode *inode) 225 { 226 u64 ino = inode->location.objectid; 227 228 /* 229 * !ino: btree_inode 230 * type == BTRFS_ROOT_ITEM_KEY: subvol dir 231 */ 232 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY) 233 ino = inode->vfs_inode.i_ino; 234 return ino; 235 } 236 237 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size) 238 { 239 i_size_write(&inode->vfs_inode, size); 240 inode->disk_i_size = size; 241 } 242 243 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode) 244 { 245 struct btrfs_root *root = inode->root; 246 247 if (root == root->fs_info->tree_root && 248 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) 249 return true; 250 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID) 251 return true; 252 return false; 253 } 254 255 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode, 256 int mod) 257 { 258 lockdep_assert_held(&inode->lock); 259 inode->outstanding_extents += mod; 260 if (btrfs_is_free_space_inode(inode)) 261 return; 262 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode), 263 mod); 264 } 265 266 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation) 267 { 268 int ret = 0; 269 270 spin_lock(&inode->lock); 271 if (inode->logged_trans == generation && 272 inode->last_sub_trans <= inode->last_log_commit && 273 inode->last_sub_trans <= inode->root->last_log_commit) { 274 /* 275 * After a ranged fsync we might have left some extent maps 276 * (that fall outside the fsync's range). So return false 277 * here if the list isn't empty, to make sure btrfs_log_inode() 278 * will be called and process those extent maps. 279 */ 280 smp_mb(); 281 if (list_empty(&inode->extent_tree.modified_extents)) 282 ret = 1; 283 } 284 spin_unlock(&inode->lock); 285 return ret; 286 } 287 288 #define BTRFS_DIO_ORIG_BIO_SUBMITTED 0x1 289 290 struct btrfs_dio_private { 291 struct inode *inode; 292 unsigned long flags; 293 u64 logical_offset; 294 u64 disk_bytenr; 295 u64 bytes; 296 void *private; 297 298 /* number of bios pending for this dio */ 299 atomic_t pending_bios; 300 301 /* IO errors */ 302 int errors; 303 304 /* orig_bio is our btrfs_io_bio */ 305 struct bio *orig_bio; 306 307 /* dio_bio came from fs/direct-io.c */ 308 struct bio *dio_bio; 309 310 /* 311 * The original bio may be split to several sub-bios, this is 312 * done during endio of sub-bios 313 */ 314 blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *, 315 blk_status_t); 316 }; 317 318 /* 319 * Disable DIO read nolock optimization, so new dio readers will be forced 320 * to grab i_mutex. It is used to avoid the endless truncate due to 321 * nonlocked dio read. 322 */ 323 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode) 324 { 325 set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags); 326 smp_mb(); 327 } 328 329 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode) 330 { 331 smp_mb__before_atomic(); 332 clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags); 333 } 334 335 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode, 336 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num) 337 { 338 struct btrfs_root *root = inode->root; 339 340 /* Output minus objectid, which is more meaningful */ 341 if (root->objectid >= BTRFS_LAST_FREE_OBJECTID) 342 btrfs_warn_rl(root->fs_info, 343 "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d", 344 root->objectid, btrfs_ino(inode), 345 logical_start, csum, csum_expected, mirror_num); 346 else 347 btrfs_warn_rl(root->fs_info, 348 "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d", 349 root->objectid, btrfs_ino(inode), 350 logical_start, csum, csum_expected, mirror_num); 351 } 352 353 #endif 354