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