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