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