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 defrag, used by stat to check whether
 129          * it needs COW.
 130          */
 131         u64 defrag_bytes;
 132
 133         /*
 134          * the size of the file stored in the metadata on disk.  data=ordered
 135          * means the in-memory i_size might be larger than the size on disk
 136          * because not all the blocks are written yet.
 137          */
 138         u64 disk_i_size;
 139
 140         /*
 141          * if this is a directory then index_cnt is the counter for the index
 142          * number for new files that are created
 143          */
 144         u64 index_cnt;
 145
 146         /* Cache the directory index number to speed the dir/file remove */
 147         u64 dir_index;
 148
 149         /* the fsync log has some corner cases that mean we have to check
 150          * directories to see if any unlinks have been done before
 151          * the directory was logged.  See tree-log.c for all the
 152          * details
 153          */
 154         u64 last_unlink_trans;
 155
 156         /*
 157          * Number of bytes outstanding that are going to need csums.  This is
 158          * used in ENOSPC accounting.
 159          */
 160         u64 csum_bytes;
 161
 162         /* flags field from the on disk inode */
 163         u32 flags;
 164
 165         /*
 166          * Counters to keep track of the number of extent item's we may use due
 167          * to delalloc and such.  outstanding_extents is the number of extent
 168          * items we think we'll end up using, and reserved_extents is the number
 169          * of extent items we've reserved metadata for.
 170          */
 171         unsigned outstanding_extents;
 172         unsigned reserved_extents;
 173
 174         /*
 175          * always compress this one file
 176          */
 177         unsigned force_compress;
 178
 179         struct btrfs_delayed_node *delayed_node;
 180
 181         /* File creation time. */
 182         struct timespec i_otime;
 183
 184         /* Hook into fs_info->delayed_iputs */
 185         struct list_head delayed_iput;
 186         long delayed_iput_count;
 187
 188         /*
 189          * To avoid races between lockless (i_mutex not held) direct IO writes
 190          * and concurrent fsync requests. Direct IO writes must acquire read
 191          * access on this semaphore for creating an extent map and its
 192          * corresponding ordered extent. The fast fsync path must acquire write
 193          * access on this semaphore before it collects ordered extents and
 194          * extent maps.
 195          */
 196         struct rw_semaphore dio_sem;
 197
 198         struct inode vfs_inode;
 199 };
 200
 201 extern unsigned char btrfs_filetype_table[];
 202
 203 static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
 204 {
 205         return container_of(inode, struct btrfs_inode, vfs_inode);
 206 }
 207
 208 static inline unsigned long btrfs_inode_hash(u64 objectid,
 209                                              const struct btrfs_root *root)
 210 {
 211         u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
 212
 213 #if BITS_PER_LONG == 32
 214         h = (h >> 32) ^ (h & 0xffffffff);
 215 #endif
 216
 217         return (unsigned long)h;
 218 }
 219
 220 static inline void btrfs_insert_inode_hash(struct inode *inode)
 221 {
 222         unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
 223
 224         __insert_inode_hash(inode, h);
 225 }
 226
 227 static inline u64 btrfs_ino(struct btrfs_inode *inode)
 228 {
 229         u64 ino = inode->location.objectid;
 230
 231         /*
 232          * !ino: btree_inode
 233          * type == BTRFS_ROOT_ITEM_KEY: subvol dir
 234          */
 235         if (!ino || inode->location.type == BTRFS_ROOT_ITEM_KEY)
 236                 ino = inode->vfs_inode.i_ino;
 237         return ino;
 238 }
 239
 240 static inline void btrfs_i_size_write(struct inode *inode, u64 size)
 241 {
 242         i_size_write(inode, size);
 243         BTRFS_I(inode)->disk_i_size = size;
 244 }
 245
 246 static inline bool btrfs_is_free_space_inode(struct inode *inode)
 247 {
 248         struct btrfs_root *root = BTRFS_I(inode)->root;
 249
 250         if (root == root->fs_info->tree_root &&
 251             btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID)
 252                 return true;
 253         if (BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
 254                 return true;
 255         return false;
 256 }
 257
 258 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
 259 {
 260         int ret = 0;
 261
 262         spin_lock(&inode->lock);
 263         if (inode->logged_trans == generation &&
 264             inode->last_sub_trans <= inode->last_log_commit &&
 265             inode->last_sub_trans <= inode->root->last_log_commit) {
 266                 /*
 267                  * After a ranged fsync we might have left some extent maps
 268                  * (that fall outside the fsync's range). So return false
 269                  * here if the list isn't empty, to make sure btrfs_log_inode()
 270                  * will be called and process those extent maps.
 271                  */
 272                 smp_mb();
 273                 if (list_empty(&inode->extent_tree.modified_extents))
 274                         ret = 1;
 275         }
 276         spin_unlock(&inode->lock);
 277         return ret;
 278 }
 279
 280 #define BTRFS_DIO_ORIG_BIO_SUBMITTED    0x1
 281
 282 struct btrfs_dio_private {
 283         struct inode *inode;
 284         unsigned long flags;
 285         u64 logical_offset;
 286         u64 disk_bytenr;
 287         u64 bytes;
 288         void *private;
 289
 290         /* number of bios pending for this dio */
 291         atomic_t pending_bios;
 292
 293         /* IO errors */
 294         int errors;
 295
 296         /* orig_bio is our btrfs_io_bio */
 297         struct bio *orig_bio;
 298
 299         /* dio_bio came from fs/direct-io.c */
 300         struct bio *dio_bio;
 301
 302         /*
 303          * The original bio may be split to several sub-bios, this is
 304          * done during endio of sub-bios
 305          */
 306         int (*subio_endio)(struct inode *, struct btrfs_io_bio *, int);
 307 };
 308
 309 /*
 310  * Disable DIO read nolock optimization, so new dio readers will be forced
 311  * to grab i_mutex. It is used to avoid the endless truncate due to
 312  * nonlocked dio read.
 313  */
 314 static inline void btrfs_inode_block_unlocked_dio(struct inode *inode)
 315 {
 316         set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags);
 317         smp_mb();
 318 }
 319
 320 static inline void btrfs_inode_resume_unlocked_dio(struct inode *inode)
 321 {
 322         smp_mb__before_atomic();
 323         clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
 324                   &BTRFS_I(inode)->runtime_flags);
 325 }
 326
 327 static inline void btrfs_print_data_csum_error(struct inode *inode,
 328                 u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
 329 {
 330         struct btrfs_root *root = BTRFS_I(inode)->root;
 331
 332         /* Output minus objectid, which is more meaningful */
 333         if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
 334                 btrfs_warn_rl(root->fs_info,
 335         "csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
 336                         root->objectid, btrfs_ino(BTRFS_I(inode)),
 337                         logical_start, csum, csum_expected, mirror_num);
 338         else
 339                 btrfs_warn_rl(root->fs_info,
 340         "csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
 341                         root->objectid, btrfs_ino(BTRFS_I(inode)),
 342                         logical_start, csum, csum_expected, mirror_num);
 343 }
 344
 345 bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
 346
 347 #endif