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 = 0,
  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 };
  33
  34 /* in memory btrfs inode */
  35 struct btrfs_inode {
  36         /* which subvolume this inode belongs to */
  37         struct btrfs_root *root;
  38
  39         /* key used to find this inode on disk.  This is used by the code
  40          * to read in roots of subvolumes
  41          */
  42         struct btrfs_key location;
  43
  44         /*
  45          * Lock for counters and all fields used to determine if the inode is in
  46          * the log or not (last_trans, last_sub_trans, last_log_commit,
  47          * logged_trans).
  48          */
  49         spinlock_t lock;
  50
  51         /* the extent_tree has caches of all the extent mappings to disk */
  52         struct extent_map_tree extent_tree;
  53
  54         /* the io_tree does range state (DIRTY, LOCKED etc) */
  55         struct extent_io_tree io_tree;
  56
  57         /* special utility tree used to record which mirrors have already been
  58          * tried when checksums fail for a given block
  59          */
  60         struct extent_io_tree io_failure_tree;
  61
  62         /* held while logging the inode in tree-log.c */
  63         struct mutex log_mutex;
  64
  65         /* held while doing delalloc reservations */
  66         struct mutex delalloc_mutex;
  67
  68         /* used to order data wrt metadata */
  69         struct btrfs_ordered_inode_tree ordered_tree;
  70
  71         /* list of all the delalloc inodes in the FS.  There are times we need
  72          * to write all the delalloc pages to disk, and this list is used
  73          * to walk them all.
  74          */
  75         struct list_head delalloc_inodes;
  76
  77         /* node for the red-black tree that links inodes in subvolume root */
  78         struct rb_node rb_node;
  79
  80         unsigned long runtime_flags;
  81
  82         /* Keep track of who's O_SYNC/fsyncing currently */
  83         atomic_t sync_writers;
  84
  85         /* full 64 bit generation number, struct vfs_inode doesn't have a big
  86          * enough field for this.
  87          */
  88         u64 generation;
  89
  90         /*
  91          * transid of the trans_handle that last modified this inode
  92          */
  93         u64 last_trans;
  94
  95         /*
  96          * transid that last logged this inode
  97          */
  98         u64 logged_trans;
  99
 100         /*
 101          * log transid when this inode was last modified
 102          */
 103         int last_sub_trans;
 104
 105         /* a local copy of root's last_log_commit */
 106         int last_log_commit;
 107
 108         /* total number of bytes pending delalloc, used by stat to calc the
 109          * real block usage of the file
 110          */
 111         u64 delalloc_bytes;
 112
 113         /*
 114          * Total number of bytes pending delalloc that fall within a file
 115          * range that is either a hole or beyond EOF (and no prealloc extent
 116          * exists in the range). This is always <= delalloc_bytes.
 117          */
 118         u64 new_delalloc_bytes;
 119
 120         /*
 121          * total number of bytes pending defrag, used by stat to check whether
 122          * it needs COW.
 123          */
 124         u64 defrag_bytes;
 125
 126         /*
 127          * the size of the file stored in the metadata on disk.  data=ordered
 128          * means the in-memory i_size might be larger than the size on disk
 129          * because not all the blocks are written yet.
 130          */
 131         u64 disk_i_size;
 132
 133         /*
 134          * if this is a directory then index_cnt is the counter for the index
 135          * number for new files that are created
 136          */
 137         u64 index_cnt;
 138
 139         /* Cache the directory index number to speed the dir/file remove */
 140         u64 dir_index;
 141
 142         /* the fsync log has some corner cases that mean we have to check
 143          * directories to see if any unlinks have been done before
 144          * the directory was logged.  See tree-log.c for all the
 145          * details
 146          */
 147         u64 last_unlink_trans;
 148
 149         /*
 150          * Number of bytes outstanding that are going to need csums.  This is
 151          * used in ENOSPC accounting.
 152          */
 153         u64 csum_bytes;
 154
 155         /* flags field from the on disk inode */
 156         u32 flags;
 157
 158         /*
 159          * Counters to keep track of the number of extent item's we may use due
 160          * to delalloc and such.  outstanding_extents is the number of extent
 161          * items we think we'll end up using, and reserved_extents is the number
 162          * of extent items we've reserved metadata for.
 163          */
 164         unsigned outstanding_extents;
 165
 166         struct btrfs_block_rsv block_rsv;
 167
 168         /*
 169          * Cached values of inode properties
 170          */
 171         unsigned prop_compress;         /* per-file compression algorithm */
 172         /*
 173          * Force compression on the file using the defrag ioctl, could be
 174          * different from prop_compress and takes precedence if set
 175          */
 176         unsigned defrag_compress;
 177
 178         struct btrfs_delayed_node *delayed_node;
 179
 180         /* File creation time. */
 181         struct timespec64 i_otime;
 182
 183         /* Hook into fs_info->delayed_iputs */
 184         struct list_head delayed_iput;
 185
 186         /*
 187          * To avoid races between lockless (i_mutex not held) direct IO writes
 188          * and concurrent fsync requests. Direct IO writes must acquire read
 189          * access on this semaphore for creating an extent map and its
 190          * corresponding ordered extent. The fast fsync path must acquire write
 191          * access on this semaphore before it collects ordered extents and
 192          * extent maps.
 193          */
 194         struct rw_semaphore dio_sem;
 195
 196         struct inode vfs_inode;
 197 };
 198
 199 extern unsigned char btrfs_filetype_table[];
 200
 201 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
 202 {
 203         return container_of(inode, struct btrfs_inode, vfs_inode);
 204 }
 205
 206 static inline unsigned long btrfs_inode_hash(u64 objectid,
 207                                              const struct btrfs_root *root)
 208 {
 209         u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
 210
 211 #if BITS_PER_LONG == 32
 212         h = (h >> 32) ^ (h & 0xffffffff);
 213 #endif
 214
 215         return (unsigned long)h;
 216 }
 217
 218 static inline void btrfs_insert_inode_hash(struct inode *inode)
 219 {
 220         unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
 221
 222         __insert_inode_hash(inode, h);
 223 }
 224
 225 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
 226 {
 227         u64 ino = inode->location.objectid;
 228
 229         /*
 230          * !ino: btree_inode
 231          * type == BTRFS_ROOT_ITEM_KEY: subvol dir
 232          */
 233         if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
 234                 ino = inode->vfs_inode.i_ino;
 235         return ino;
 236 }
 237
 238 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
 239 {
 240         i_size_write(&inode->vfs_inode, size);
 241         inode->disk_i_size = size;
 242 }
 243
 244 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
 245 {
 246         struct btrfs_root *root = inode->root;
 247
 248         if (root == root->fs_info->tree_root &&
 249             btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
 250                 return true;
 251         if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
 252                 return true;
 253         return false;
 254 }
 255
 256 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
 257                                                  int mod)
 258 {
 259         lockdep_assert_held(&inode->lock);
 260         inode->outstanding_extents += mod;
 261         if (btrfs_is_free_space_inode(inode))
 262                 return;
 263         trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
 264                                                   mod);
 265 }
 266
 267 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
 268 {
 269         int ret = 0;
 270
 271         spin_lock(&inode->lock);
 272         if (inode->logged_trans == generation &&
 273             inode->last_sub_trans <= inode->last_log_commit &&
 274             inode->last_sub_trans <= inode->root->last_log_commit) {
 275                 /*
 276                  * After a ranged fsync we might have left some extent maps
 277                  * (that fall outside the fsync's range). So return false
 278                  * here if the list isn't empty, to make sure btrfs_log_inode()
 279                  * will be called and process those extent maps.
 280                  */
 281                 smp_mb();
 282                 if (list_empty(&inode->extent_tree.modified_extents))
 283                         ret = 1;
 284         }
 285         spin_unlock(&inode->lock);
 286         return ret;
 287 }
 288
 289 #define BTRFS_DIO_ORIG_BIO_SUBMITTED    0x1
 290
 291 struct btrfs_dio_private {
 292         struct inode *inode;
 293         unsigned long flags;
 294         u64 logical_offset;
 295         u64 disk_bytenr;
 296         u64 bytes;
 297         void *private;
 298
 299         /* number of bios pending for this dio */
 300         atomic_t pending_bios;
 301
 302         /* IO errors */
 303         int errors;
 304
 305         /* orig_bio is our btrfs_io_bio */
 306         struct bio *orig_bio;
 307
 308         /* dio_bio came from fs/direct-io.c */
 309         struct bio *dio_bio;
 310
 311         /*
 312          * The original bio may be split to several sub-bios, this is
 313          * done during endio of sub-bios
 314          */
 315         blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
 316                         blk_status_t);
 317 };
 318
 319 /*
 320  * Disable DIO read nolock optimization, so new dio readers will be forced
 321  * to grab i_mutex. It is used to avoid the endless truncate due to
 322  * nonlocked dio read.
 323  */
 324 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
 325 {
 326         set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
 327         smp_mb();
 328 }
 329
 330 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
 331 {
 332         smp_mb__before_atomic();
 333         clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
 334 }
 335
 336 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
 337                 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
 338 {
 339         struct btrfs_root *root = inode->root;
 340
 341         /* Output minus objectid, which is more meaningful */
 342         if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID)
 343                 btrfs_warn_rl(root->fs_info,
 344         "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
 345                         root->root_key.objectid, btrfs_ino(inode),
 346                         logical_start, csum, csum_expected, mirror_num);
 347         else
 348                 btrfs_warn_rl(root->fs_info,
 349         "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
 350                         root->root_key.objectid, btrfs_ino(inode),
 351                         logical_start, csum, csum_expected, mirror_num);
 352 }
 353
 354 #endif