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 <linux/refcount.h> 11 #include "extent_map.h" 12 #include "extent_io.h" 13 #include "ordered-data.h" 14 #include "delayed-inode.h" 15 16 /* 17 * ordered_data_close is set by truncate when a file that used 18 * to have good data has been truncated to zero. When it is set 19 * the btrfs file release call will add this inode to the 20 * ordered operations list so that we make sure to flush out any 21 * new data the application may have written before commit. 22 */ 23 enum { 24 BTRFS_INODE_FLUSH_ON_CLOSE, 25 BTRFS_INODE_DUMMY, 26 BTRFS_INODE_IN_DEFRAG, 27 BTRFS_INODE_HAS_ASYNC_EXTENT, 28 /* 29 * Always set under the VFS' inode lock, otherwise it can cause races 30 * during fsync (we start as a fast fsync and then end up in a full 31 * fsync racing with ordered extent completion). 32 */ 33 BTRFS_INODE_NEEDS_FULL_SYNC, 34 BTRFS_INODE_COPY_EVERYTHING, 35 BTRFS_INODE_IN_DELALLOC_LIST, 36 BTRFS_INODE_HAS_PROPS, 37 BTRFS_INODE_SNAPSHOT_FLUSH, 38 }; 39 40 /* in memory btrfs inode */ 41 struct btrfs_inode { 42 /* which subvolume this inode belongs to */ 43 struct btrfs_root *root; 44 45 /* key used to find this inode on disk. This is used by the code 46 * to read in roots of subvolumes 47 */ 48 struct btrfs_key location; 49 50 /* 51 * Lock for counters and all fields used to determine if the inode is in 52 * the log or not (last_trans, last_sub_trans, last_log_commit, 53 * logged_trans), to access/update new_delalloc_bytes and to update the 54 * VFS' inode number of bytes used. 55 */ 56 spinlock_t lock; 57 58 /* the extent_tree has caches of all the extent mappings to disk */ 59 struct extent_map_tree extent_tree; 60 61 /* the io_tree does range state (DIRTY, LOCKED etc) */ 62 struct extent_io_tree io_tree; 63 64 /* special utility tree used to record which mirrors have already been 65 * tried when checksums fail for a given block 66 */ 67 struct extent_io_tree io_failure_tree; 68 69 /* 70 * Keep track of where the inode has extent items mapped in order to 71 * make sure the i_size adjustments are accurate 72 */ 73 struct extent_io_tree file_extent_tree; 74 75 /* held while logging the inode in tree-log.c */ 76 struct mutex log_mutex; 77 78 /* used to order data wrt metadata */ 79 struct btrfs_ordered_inode_tree ordered_tree; 80 81 /* list of all the delalloc inodes in the FS. There are times we need 82 * to write all the delalloc pages to disk, and this list is used 83 * to walk them all. 84 */ 85 struct list_head delalloc_inodes; 86 87 /* node for the red-black tree that links inodes in subvolume root */ 88 struct rb_node rb_node; 89 90 unsigned long runtime_flags; 91 92 /* Keep track of who's O_SYNC/fsyncing currently */ 93 atomic_t sync_writers; 94 95 /* full 64 bit generation number, struct vfs_inode doesn't have a big 96 * enough field for this. 97 */ 98 u64 generation; 99 100 /* 101 * transid of the trans_handle that last modified this inode 102 */ 103 u64 last_trans; 104 105 /* 106 * transid that last logged this inode 107 */ 108 u64 logged_trans; 109 110 /* 111 * log transid when this inode was last modified 112 */ 113 int last_sub_trans; 114 115 /* a local copy of root's last_log_commit */ 116 int last_log_commit; 117 118 /* total number of bytes pending delalloc, used by stat to calc the 119 * real block usage of the file 120 */ 121 u64 delalloc_bytes; 122 123 /* 124 * Total number of bytes pending delalloc that fall within a file 125 * range that is either a hole or beyond EOF (and no prealloc extent 126 * exists in the range). This is always <= delalloc_bytes. 127 */ 128 u64 new_delalloc_bytes; 129 130 /* 131 * total number of bytes pending defrag, used by stat to check whether 132 * it needs COW. 133 */ 134 u64 defrag_bytes; 135 136 /* 137 * the size of the file stored in the metadata on disk. data=ordered 138 * means the in-memory i_size might be larger than the size on disk 139 * because not all the blocks are written yet. 140 */ 141 u64 disk_i_size; 142 143 /* 144 * if this is a directory then index_cnt is the counter for the index 145 * number for new files that are created 146 */ 147 u64 index_cnt; 148 149 /* Cache the directory index number to speed the dir/file remove */ 150 u64 dir_index; 151 152 /* the fsync log has some corner cases that mean we have to check 153 * directories to see if any unlinks have been done before 154 * the directory was logged. See tree-log.c for all the 155 * details 156 */ 157 u64 last_unlink_trans; 158 159 /* 160 * The id/generation of the last transaction where this inode was 161 * either the source or the destination of a clone/dedupe operation. 162 * Used when logging an inode to know if there are shared extents that 163 * need special care when logging checksum items, to avoid duplicate 164 * checksum items in a log (which can lead to a corruption where we end 165 * up with missing checksum ranges after log replay). 166 * Protected by the vfs inode lock. 167 */ 168 u64 last_reflink_trans; 169 170 /* 171 * Number of bytes outstanding that are going to need csums. This is 172 * used in ENOSPC accounting. 173 */ 174 u64 csum_bytes; 175 176 /* flags field from the on disk inode */ 177 u32 flags; 178 179 /* 180 * Counters to keep track of the number of extent item's we may use due 181 * to delalloc and such. outstanding_extents is the number of extent 182 * items we think we'll end up using, and reserved_extents is the number 183 * of extent items we've reserved metadata for. 184 */ 185 unsigned outstanding_extents; 186 187 struct btrfs_block_rsv block_rsv; 188 189 /* 190 * Cached values of inode properties 191 */ 192 unsigned prop_compress; /* per-file compression algorithm */ 193 /* 194 * Force compression on the file using the defrag ioctl, could be 195 * different from prop_compress and takes precedence if set 196 */ 197 unsigned defrag_compress; 198 199 struct btrfs_delayed_node *delayed_node; 200 201 /* File creation time. */ 202 struct timespec64 i_otime; 203 204 /* Hook into fs_info->delayed_iputs */ 205 struct list_head delayed_iput; 206 207 struct inode vfs_inode; 208 }; 209 210 static inline u32 btrfs_inode_sectorsize(const struct btrfs_inode *inode) 211 { 212 return inode->root->fs_info->sectorsize; 213 } 214 215 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode) 216 { 217 return container_of(inode, struct btrfs_inode, vfs_inode); 218 } 219 220 static inline unsigned long btrfs_inode_hash(u64 objectid, 221 const struct btrfs_root *root) 222 { 223 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME); 224 225 #if BITS_PER_LONG == 32 226 h = (h >> 32) ^ (h & 0xffffffff); 227 #endif 228 229 return (unsigned long)h; 230 } 231 232 static inline void btrfs_insert_inode_hash(struct inode *inode) 233 { 234 unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root); 235 236 __insert_inode_hash(inode, h); 237 } 238 239 static inline u64 btrfs_ino(const struct btrfs_inode *inode) 240 { 241 u64 ino = inode->location.objectid; 242 243 /* 244 * !ino: btree_inode 245 * type == BTRFS_ROOT_ITEM_KEY: subvol dir 246 */ 247 if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY) 248 ino = inode->vfs_inode.i_ino; 249 return ino; 250 } 251 252 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size) 253 { 254 i_size_write(&inode->vfs_inode, size); 255 inode->disk_i_size = size; 256 } 257 258 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode) 259 { 260 struct btrfs_root *root = inode->root; 261 262 if (root == root->fs_info->tree_root && 263 btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID) 264 return true; 265 if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID) 266 return true; 267 return false; 268 } 269 270 static inline bool is_data_inode(struct inode *inode) 271 { 272 return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID; 273 } 274 275 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode, 276 int mod) 277 { 278 lockdep_assert_held(&inode->lock); 279 inode->outstanding_extents += mod; 280 if (btrfs_is_free_space_inode(inode)) 281 return; 282 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode), 283 mod); 284 } 285 286 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation) 287 { 288 int ret = 0; 289 290 spin_lock(&inode->lock); 291 if (inode->logged_trans == generation && 292 inode->last_sub_trans <= inode->last_log_commit && 293 inode->last_sub_trans <= inode->root->last_log_commit) { 294 /* 295 * After a ranged fsync we might have left some extent maps 296 * (that fall outside the fsync's range). So return false 297 * here if the list isn't empty, to make sure btrfs_log_inode() 298 * will be called and process those extent maps. 299 */ 300 smp_mb(); 301 if (list_empty(&inode->extent_tree.modified_extents)) 302 ret = 1; 303 } 304 spin_unlock(&inode->lock); 305 return ret; 306 } 307 308 struct btrfs_dio_private { 309 struct inode *inode; 310 u64 logical_offset; 311 u64 disk_bytenr; 312 u64 bytes; 313 314 /* 315 * References to this structure. There is one reference per in-flight 316 * bio plus one while we're still setting up. 317 */ 318 refcount_t refs; 319 320 /* dio_bio came from fs/direct-io.c */ 321 struct bio *dio_bio; 322 323 /* Array of checksums */ 324 u8 csums[]; 325 }; 326 327 /* Array of bytes with variable length, hexadecimal format 0x1234 */ 328 #define CSUM_FMT "0x%*phN" 329 #define CSUM_FMT_VALUE(size, bytes) size, bytes 330 331 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode, 332 u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num) 333 { 334 struct btrfs_root *root = inode->root; 335 const u32 csum_size = root->fs_info->csum_size; 336 337 /* Output minus objectid, which is more meaningful */ 338 if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID) 339 btrfs_warn_rl(root->fs_info, 340 "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", 341 root->root_key.objectid, btrfs_ino(inode), 342 logical_start, 343 CSUM_FMT_VALUE(csum_size, csum), 344 CSUM_FMT_VALUE(csum_size, csum_expected), 345 mirror_num); 346 else 347 btrfs_warn_rl(root->fs_info, 348 "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d", 349 root->root_key.objectid, btrfs_ino(inode), 350 logical_start, 351 CSUM_FMT_VALUE(csum_size, csum), 352 CSUM_FMT_VALUE(csum_size, csum_expected), 353 mirror_num); 354 } 355 356 #endif 357