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