fs/btrfs/btrfs_inode.h

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #ifndef __BTRFS_I__
  20 #define __BTRFS_I__
  21
  22 #include <linux/hash.h>
  23 #include "extent_map.h"
  24 #include "extent_io.h"
  25 #include "ordered-data.h"
  26 #include "delayed-inode.h"
  27
  28 /*
  29  * ordered_data_close is set by truncate when a file that used
  30  * to have good data has been truncated to zero.  When it is set
  31  * the btrfs file release call will add this inode to the
  32  * ordered operations list so that we make sure to flush out any
  33  * new data the application may have written before commit.
  34  */
  35 #define BTRFS_INODE_ORDERED_DATA_CLOSE          0
  36 #define BTRFS_INODE_ORPHAN_META_RESERVED        1
  37 #define BTRFS_INODE_DUMMY                       2
  38 #define BTRFS_INODE_IN_DEFRAG                   3
  39 #define BTRFS_INODE_HAS_ORPHAN_ITEM             4
  40 #define BTRFS_INODE_HAS_ASYNC_EXTENT            5
  41 #define BTRFS_INODE_NEEDS_FULL_SYNC             6
  42 #define BTRFS_INODE_COPY_EVERYTHING             7
  43 #define BTRFS_INODE_IN_DELALLOC_LIST            8
  44 #define BTRFS_INODE_READDIO_NEED_LOCK           9
  45 #define BTRFS_INODE_HAS_PROPS                   10
  46
  47 /* in memory btrfs inode */
  48 struct btrfs_inode {
  49         /* which subvolume this inode belongs to */
  50         struct btrfs_root *root;
  51
  52         /* key used to find this inode on disk.  This is used by the code
  53          * to read in roots of subvolumes
  54          */
  55         struct btrfs_key location;
  56
  57         /*
  58          * Lock for counters and all fields used to determine if the inode is in
  59          * the log or not (last_trans, last_sub_trans, last_log_commit,
  60          * logged_trans).
  61          */
  62         spinlock_t lock;
  63
  64         /* the extent_tree has caches of all the extent mappings to disk */
  65         struct extent_map_tree extent_tree;
  66
  67         /* the io_tree does range state (DIRTY, LOCKED etc) */
  68         struct extent_io_tree io_tree;
  69
  70         /* special utility tree used to record which mirrors have already been
  71          * tried when checksums fail for a given block
  72          */
  73         struct extent_io_tree io_failure_tree;
  74
  75         /* held while logging the inode in tree-log.c */
  76         struct mutex log_mutex;
  77
  78         /* held while doing delalloc reservations */
  79         struct mutex delalloc_mutex;
  80
  81         /* used to order data wrt metadata */
  82         struct btrfs_ordered_inode_tree ordered_tree;
  83
  84         /* list of all the delalloc inodes in the FS.  There are times we need
  85          * to write all the delalloc pages to disk, and this list is used
  86          * to walk them all.
  87          */
  88         struct list_head delalloc_inodes;
  89
  90         /* node for the red-black tree that links inodes in subvolume root */
  91         struct rb_node rb_node;
  92
  93         unsigned long runtime_flags;
  94
  95         /* Keep track of who's O_SYNC/fsyncing currently */
  96         atomic_t sync_writers;
  97
  98         /* full 64 bit generation number, struct vfs_inode doesn't have a big
  99          * enough field for this.
 100          */
 101         u64 generation;
 102
 103         /*
 104          * transid of the trans_handle that last modified this inode
 105          */
 106         u64 last_trans;
 107
 108         /*
 109          * transid that last logged this inode
 110          */
 111         u64 logged_trans;
 112
 113         /*
 114          * log transid when this inode was last modified
 115          */
 116         int last_sub_trans;
 117
 118         /* a local copy of root's last_log_commit */
 119         int last_log_commit;
 120
 121         /* total number of bytes pending delalloc, used by stat to calc the
 122          * real block usage of the file
 123          */
 124         u64 delalloc_bytes;
 125
 126         /*
 127          * Total number of bytes pending delalloc that fall within a file
 128          * range that is either a hole or beyond EOF (and no prealloc extent
 129          * exists in the range). This is always <= delalloc_bytes.
 130          */
 131         u64 new_delalloc_bytes;
 132
 133         /*
 134          * total number of bytes pending defrag, used by stat to check whether
 135          * it needs COW.
 136          */
 137         u64 defrag_bytes;
 138
 139         /*
 140          * the size of the file stored in the metadata on disk.  data=ordered
 141          * means the in-memory i_size might be larger than the size on disk
 142          * because not all the blocks are written yet.
 143          */
 144         u64 disk_i_size;
 145
 146         /*
 147          * if this is a directory then index_cnt is the counter for the index
 148          * number for new files that are created
 149          */
 150         u64 index_cnt;
 151
 152         /* Cache the directory index number to speed the dir/file remove */
 153         u64 dir_index;
 154
 155         /* the fsync log has some corner cases that mean we have to check
 156          * directories to see if any unlinks have been done before
 157          * the directory was logged.  See tree-log.c for all the
 158          * details
 159          */
 160         u64 last_unlink_trans;
 161
 162         /*
 163          * Number of bytes outstanding that are going to need csums.  This is
 164          * used in ENOSPC accounting.
 165          */
 166         u64 csum_bytes;
 167
 168         /* flags field from the on disk inode */
 169         u32 flags;
 170
 171         /*
 172          * Counters to keep track of the number of extent item's we may use due
 173          * to delalloc and such.  outstanding_extents is the number of extent
 174          * items we think we'll end up using, and reserved_extents is the number
 175          * of extent items we've reserved metadata for.
 176          */
 177         unsigned outstanding_extents;
 178
 179         struct btrfs_block_rsv block_rsv;
 180
 181         /*
 182          * Cached values of inode properties
 183          */
 184         unsigned prop_compress;         /* per-file compression algorithm */
 185         /*
 186          * Force compression on the file using the defrag ioctl, could be
 187          * different from prop_compress and takes precedence if set
 188          */
 189         unsigned defrag_compress;
 190
 191         struct btrfs_delayed_node *delayed_node;
 192
 193         /* File creation time. */
 194         struct timespec i_otime;
 195
 196         /* Hook into fs_info->delayed_iputs */
 197         struct list_head delayed_iput;
 198         long delayed_iput_count;
 199
 200         /*
 201          * To avoid races between lockless (i_mutex not held) direct IO writes
 202          * and concurrent fsync requests. Direct IO writes must acquire read
 203          * access on this semaphore for creating an extent map and its
 204          * corresponding ordered extent. The fast fsync path must acquire write
 205          * access on this semaphore before it collects ordered extents and
 206          * extent maps.
 207          */
 208         struct rw_semaphore dio_sem;
 209
 210         struct inode vfs_inode;
 211 };
 212
 213 extern unsigned char btrfs_filetype_table[];
 214
 215 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
 216 {
 217         return container_of(inode, struct btrfs_inode, vfs_inode);
 218 }
 219
 220 static inline unsigned long btrfs_inode_hash(u64 objectid,
 221                                              const struct btrfs_root *root)
 222 {
 223         u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
 224
 225 #if BITS_PER_LONG == 32
 226         h = (h >> 32) ^ (h & 0xffffffff);
 227 #endif
 228
 229         return (unsigned long)h;
 230 }
 231
 232 static inline void btrfs_insert_inode_hash(struct inode *inode)
 233 {
 234         unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
 235
 236         __insert_inode_hash(inode, h);
 237 }
 238
 239 static inline u64 btrfs_ino(const struct btrfs_inode *inode)
 240 {
 241         u64 ino = inode->location.objectid;
 242
 243         /*
 244          * !ino: btree_inode
 245          * type == BTRFS_ROOT_ITEM_KEY: subvol dir
 246          */
 247         if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
 248                 ino = inode->vfs_inode.i_ino;
 249         return ino;
 250 }
 251
 252 static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)
 253 {
 254         i_size_write(&inode->vfs_inode, size);
 255         inode->disk_i_size = size;
 256 }
 257
 258 static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
 259 {
 260         struct btrfs_root *root = inode->root;
 261
 262         if (root == root->fs_info->tree_root &&
 263             btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
 264                 return true;
 265         if (inode->location.objectid == BTRFS_FREE_INO_OBJECTID)
 266                 return true;
 267         return false;
 268 }
 269
 270 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
 271                                                  int mod)
 272 {
 273         lockdep_assert_held(&inode->lock);
 274         inode->outstanding_extents += mod;
 275         if (btrfs_is_free_space_inode(inode))
 276                 return;
 277         trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
 278                                                   mod);
 279 }
 280
 281 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
 282 {
 283         int ret = 0;
 284
 285         spin_lock(&inode->lock);
 286         if (inode->logged_trans == generation &&
 287             inode->last_sub_trans <= inode->last_log_commit &&
 288             inode->last_sub_trans <= inode->root->last_log_commit) {
 289                 /*
 290                  * After a ranged fsync we might have left some extent maps
 291                  * (that fall outside the fsync's range). So return false
 292                  * here if the list isn't empty, to make sure btrfs_log_inode()
 293                  * will be called and process those extent maps.
 294                  */
 295                 smp_mb();
 296                 if (list_empty(&inode->extent_tree.modified_extents))
 297                         ret = 1;
 298         }
 299         spin_unlock(&inode->lock);
 300         return ret;
 301 }
 302
 303 #define BTRFS_DIO_ORIG_BIO_SUBMITTED    0x1
 304
 305 struct btrfs_dio_private {
 306         struct inode *inode;
 307         unsigned long flags;
 308         u64 logical_offset;
 309         u64 disk_bytenr;
 310         u64 bytes;
 311         void *private;
 312
 313         /* number of bios pending for this dio */
 314         atomic_t pending_bios;
 315
 316         /* IO errors */
 317         int errors;
 318
 319         /* orig_bio is our btrfs_io_bio */
 320         struct bio *orig_bio;
 321
 322         /* dio_bio came from fs/direct-io.c */
 323         struct bio *dio_bio;
 324
 325         /*
 326          * The original bio may be split to several sub-bios, this is
 327          * done during endio of sub-bios
 328          */
 329         blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
 330                         blk_status_t);
 331 };
 332
 333 /*
 334  * Disable DIO read nolock optimization, so new dio readers will be forced
 335  * to grab i_mutex. It is used to avoid the endless truncate due to
 336  * nonlocked dio read.
 337  */
 338 static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
 339 {
 340         set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
 341         smp_mb();
 342 }
 343
 344 static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
 345 {
 346         smp_mb__before_atomic();
 347         clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
 348 }
 349
 350 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
 351                 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
 352 {
 353         struct btrfs_root *root = inode->root;
 354
 355         /* Output minus objectid, which is more meaningful */
 356         if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
 357                 btrfs_warn_rl(root->fs_info,
 358         "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
 359                         root->objectid, btrfs_ino(inode),
 360                         logical_start, csum, csum_expected, mirror_num);
 361         else
 362                 btrfs_warn_rl(root->fs_info,
 363         "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
 364                         root->objectid, btrfs_ino(inode),
 365                         logical_start, csum, csum_expected, mirror_num);
 366 }
 367
 368 bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
 369
 370 #endif