fs/btrfs/btrfs_inode.h

   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 enum {
  23         BTRFS_INODE_ORDERED_DATA_CLOSE,
  24         BTRFS_INODE_DUMMY,
  25         BTRFS_INODE_IN_DEFRAG,
  26         BTRFS_INODE_HAS_ASYNC_EXTENT,
  27         BTRFS_INODE_NEEDS_FULL_SYNC,
  28         BTRFS_INODE_COPY_EVERYTHING,
  29         BTRFS_INODE_IN_DELALLOC_LIST,
  30         BTRFS_INODE_READDIO_NEED_LOCK,
  31         BTRFS_INODE_HAS_PROPS,
  32         BTRFS_INODE_SNAPSHOT_FLUSH,
  33 };
  34
  35 /* in memory btrfs inode */
  36 struct btrfs_inode {
  37         /* which subvolume this inode belongs to */
  38         struct btrfs_root *root;
  39
  40         /* key used to find this inode on disk.  This is used by the code
  41          * to read in roots of subvolumes
  42          */
  43         struct btrfs_key location;
  44
  45         /*
  46          * Lock for counters and all fields used to determine if the inode is in
  47          * the log or not (last_trans, last_sub_trans, last_log_commit,
  48          * logged_trans).
  49          */
  50         spinlock_t lock;
  51
  52         /* the extent_tree has caches of all the extent mappings to disk */
  53         struct extent_map_tree extent_tree;
  54
  55         /* the io_tree does range state (DIRTY, LOCKED etc) */
  56         struct extent_io_tree io_tree;
  57
  58         /* special utility tree used to record which mirrors have already been
  59          * tried when checksums fail for a given block
  60          */
  61         struct extent_io_tree io_failure_tree;
  62
  63         /*
  64          * Keep track of where the inode has extent items mapped in order to
  65          * make sure the i_size adjustments are accurate
  66          */
  67         struct extent_io_tree file_extent_tree;
  68
  69         /* held while logging the inode in tree-log.c */
  70         struct mutex log_mutex;
  71
  72         /* used to order data wrt metadata */
  73         struct btrfs_ordered_inode_tree ordered_tree;
  74
  75         /* list of all the delalloc inodes in the FS.  There are times we need
  76          * to write all the delalloc pages to disk, and this list is used
  77          * to walk them all.
  78          */
  79         struct list_head delalloc_inodes;
  80
  81         /* node for the red-black tree that links inodes in subvolume root */
  82         struct rb_node rb_node;
  83
  84         unsigned long runtime_flags;
  85
  86         /* Keep track of who's O_SYNC/fsyncing currently */
  87         atomic_t sync_writers;
  88
  89         /* full 64 bit generation number, struct vfs_inode doesn't have a big
  90          * enough field for this.
  91          */
  92         u64 generation;
  93
  94         /*
  95          * transid of the trans_handle that last modified this inode
  96          */
  97         u64 last_trans;
  98
  99         /*
 100          * transid that last logged this inode
 101          */
 102         u64 logged_trans;
 103
 104         /*
 105          * log transid when this inode was last modified
 106          */
 107         int last_sub_trans;
 108
 109         /* a local copy of root's last_log_commit */
 110         int last_log_commit;
 111
 112         /* total number of bytes pending delalloc, used by stat to calc the
 113          * real block usage of the file
 114          */
 115         u64 delalloc_bytes;
 116
 117         /*
 118          * Total number of bytes pending delalloc that fall within a file
 119          * range that is either a hole or beyond EOF (and no prealloc extent
 120          * exists in the range). This is always <= delalloc_bytes.
 121          */
 122         u64 new_delalloc_bytes;
 123
 124         /*
 125          * total number of bytes pending defrag, used by stat to check whether
 126          * it needs COW.
 127          */
 128         u64 defrag_bytes;
 129
 130         /*
 131          * the size of the file stored in the metadata on disk.  data=ordered
 132          * means the in-memory i_size might be larger than the size on disk
 133          * because not all the blocks are written yet.
 134          */
 135         u64 disk_i_size;
 136
 137         /*
 138          * if this is a directory then index_cnt is the counter for the index
 139          * number for new files that are created
 140          */
 141         u64 index_cnt;
 142
 143         /* Cache the directory index number to speed the dir/file remove */
 144         u64 dir_index;
 145
 146         /* the fsync log has some corner cases that mean we have to check
 147          * directories to see if any unlinks have been done before
 148          * the directory was logged.  See tree-log.c for all the
 149          * details
 150          */
 151         u64 last_unlink_trans;
 152
 153         /*
 154          * Number of bytes outstanding that are going to need csums.  This is
 155          * used in ENOSPC accounting.
 156          */
 157         u64 csum_bytes;
 158
 159         /* flags field from the on disk inode */
 160         u32 flags;
 161
 162         /*
 163          * Counters to keep track of the number of extent item's we may use due
 164          * to delalloc and such.  outstanding_extents is the number of extent
 165          * items we think we'll end up using, and reserved_extents is the number
 166          * of extent items we've reserved metadata for.
 167          */
 168         unsigned outstanding_extents;
 169
 170         struct btrfs_block_rsv block_rsv;
 171
 172         /*
 173          * Cached values of inode properties
 174          */
 175         unsigned prop_compress;         /* per-file compression algorithm */
 176         /*
 177          * Force compression on the file using the defrag ioctl, could be
 178          * different from prop_compress and takes precedence if set
 179          */
 180         unsigned defrag_compress;
 181
 182         struct btrfs_delayed_node *delayed_node;
 183
 184         /* File creation time. */
 185         struct timespec64 i_otime;
 186
 187         /* Hook into fs_info->delayed_iputs */
 188         struct list_head delayed_iput;
 189
 190         /*
 191          * To avoid races between lockless (i_mutex not held) direct IO writes
 192          * and concurrent fsync requests. Direct IO writes must acquire read
 193          * access on this semaphore for creating an extent map and its
 194          * corresponding ordered extent. The fast fsync path must acquire write
 195          * access on this semaphore before it collects ordered extents and
 196          * extent maps.
 197          */
 198         struct rw_semaphore dio_sem;
 199
 200         struct inode vfs_inode;
 201 };
 202
 203 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
 204 {
 205         return container_of(inode, struct btrfs_inode, vfs_inode);
 206 }
 207
 208 static inline unsigned long btrfs_inode_hash(u64 objectid,
 209                                              const struct btrfs_root *root)
 210 {
 211         u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
 212
 213 #if BITS_PER_LONG == 32
 214         h = (h >> 32) ^ (h & 0xffffffff);
 215 #endif
 216
 217         return (unsigned long)h;
 218 }
 219
 220 static inline void btrfs_insert_inode_hash(struct inode *inode)
 221 {
 222         unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
 223
 224         __insert_inode_hash(inode, h);
 225 }
 226
 227 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
 228 {
 229         u64 ino = inode->location.objectid;
 230
 231         /*
 232          * !ino: btree_inode
 233          * type == BTRFS_ROOT_ITEM_KEY: subvol dir
 234          */
 235         if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
 236                 ino = inode->vfs_inode.i_ino;
 237         return ino;
 238 }
 239
 240 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
 241 {
 242         i_size_write(&inode->vfs_inode, size);
 243         inode->disk_i_size = size;
 244 }
 245
 246 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
 247 {
 248         struct btrfs_root *root = inode->root;
 249
 250         if (root == root->fs_info->tree_root &&
 251             btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
 252                 return true;
 253         if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
 254                 return true;
 255         return false;
 256 }
 257
 258 static inline bool is_data_inode(struct inode *inode)
 259 {
 260         return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
 261 }
 262
 263 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
 264                                                  int mod)
 265 {
 266         lockdep_assert_held(&inode->lock);
 267         inode->outstanding_extents += mod;
 268         if (btrfs_is_free_space_inode(inode))
 269                 return;
 270         trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
 271                                                   mod);
 272 }
 273
 274 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
 275 {
 276         int ret = 0;
 277
 278         spin_lock(&inode->lock);
 279         if (inode->logged_trans == generation &&
 280             inode->last_sub_trans <= inode->last_log_commit &&
 281             inode->last_sub_trans <= inode->root->last_log_commit) {
 282                 /*
 283                  * After a ranged fsync we might have left some extent maps
 284                  * (that fall outside the fsync's range). So return false
 285                  * here if the list isn't empty, to make sure btrfs_log_inode()
 286                  * will be called and process those extent maps.
 287                  */
 288                 smp_mb();
 289                 if (list_empty(&inode->extent_tree.modified_extents))
 290                         ret = 1;
 291         }
 292         spin_unlock(&inode->lock);
 293         return ret;
 294 }
 295
 296 #define BTRFS_DIO_ORIG_BIO_SUBMITTED    0x1
 297
 298 struct btrfs_dio_private {
 299         struct inode *inode;
 300         unsigned long flags;
 301         u64 logical_offset;
 302         u64 disk_bytenr;
 303         u64 bytes;
 304         void *private;
 305
 306         /* number of bios pending for this dio */
 307         atomic_t pending_bios;
 308
 309         /* IO errors */
 310         int errors;
 311
 312         /* orig_bio is our btrfs_io_bio */
 313         struct bio *orig_bio;
 314
 315         /* dio_bio came from fs/direct-io.c */
 316         struct bio *dio_bio;
 317
 318         /*
 319          * The original bio may be split to several sub-bios, this is
 320          * done during endio of sub-bios
 321          */
 322         blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
 323                         blk_status_t);
 324 };
 325
 326 /*
 327  * Disable DIO read nolock optimization, so new dio readers will be forced
 328  * to grab i_mutex. It is used to avoid the endless truncate due to
 329  * nonlocked dio read.
 330  */
 331 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
 332 {
 333         set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
 334         smp_mb();
 335 }
 336
 337 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
 338 {
 339         smp_mb__before_atomic();
 340         clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
 341 }
 342
 343 /* Array of bytes with variable length, hexadecimal format 0x1234 */
 344 #define CSUM_FMT                                "0x%*phN"
 345 #define CSUM_FMT_VALUE(size, bytes)             size, bytes
 346
 347 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
 348                 u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num)
 349 {
 350         struct btrfs_root *root = inode->root;
 351         struct btrfs_super_block *sb = root->fs_info->super_copy;
 352         const u16 csum_size = btrfs_super_csum_size(sb);
 353
 354         /* Output minus objectid, which is more meaningful */
 355         if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID)
 356                 btrfs_warn_rl(root->fs_info,
 357 "csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
 358                         root->root_key.objectid, btrfs_ino(inode),
 359                         logical_start,
 360                         CSUM_FMT_VALUE(csum_size, csum),
 361                         CSUM_FMT_VALUE(csum_size, csum_expected),
 362                         mirror_num);
 363         else
 364                 btrfs_warn_rl(root->fs_info,
 365 "csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
 366                         root->root_key.objectid, btrfs_ino(inode),
 367                         logical_start,
 368                         CSUM_FMT_VALUE(csum_size, csum),
 369                         CSUM_FMT_VALUE(csum_size, csum_expected),
 370                         mirror_num);
 371 }
 372
 373 #endif