libxfs/xfs_inode_buf.c

   1 /*
   2  * Copyright (c) 2000-2006 Silicon Graphics, Inc.
   3  * All Rights Reserved.
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License as
   7  * published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it would be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License
  15  * along with this program; if not, write the Free Software Foundation,
  16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  17  */
  18
  19 #include <xfs.h>
  20
  21 /*
  22  * Check that none of the inode's in the buffer have a next
  23  * unlinked field of 0.
  24  */
  25 #if defined(DEBUG)
  26 void
  27 xfs_inobp_check(
  28         xfs_mount_t     *mp,
  29         xfs_buf_t       *bp)
  30 {
  31         int             i;
  32         int             j;
  33         xfs_dinode_t    *dip;
  34
  35         j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
  36
  37         for (i = 0; i < j; i++) {
  38                 dip = (xfs_dinode_t *)xfs_buf_offset(bp,
  39                                         i * mp->m_sb.sb_inodesize);
  40                 if (!dip->di_next_unlinked)  {
  41                         xfs_alert(mp,
  42         "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
  43                                 i, (long long)bp->b_bn);
  44                 }
  45         }
  46 }
  47 #endif
  48
  49 /*
  50  * If we are doing readahead on an inode buffer, we might be in log recovery
  51  * reading an inode allocation buffer that hasn't yet been replayed, and hence
  52  * has not had the inode cores stamped into it. Hence for readahead, the buffer
  53  * may be potentially invalid.
  54  *
  55  * If the readahead buffer is invalid, we don't want to mark it with an error,
  56  * but we do want to clear the DONE status of the buffer so that a followup read
  57  * will re-read it from disk. This will ensure that we don't get an unnecessary
  58  * warnings during log recovery and we don't get unnecessary panics on debug
  59  * kernels.
  60  */
  61 static void
  62 xfs_inode_buf_verify(
  63         struct xfs_buf  *bp,
  64         bool            readahead)
  65 {
  66         struct xfs_mount *mp = bp->b_target->bt_mount;
  67         int             i;
  68         int             ni;
  69
  70         /*
  71          * Validate the magic number and version of every inode in the buffer
  72          */
  73         ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
  74         for (i = 0; i < ni; i++) {
  75                 int             di_ok;
  76                 xfs_dinode_t    *dip;
  77
  78                 dip = (struct xfs_dinode *)xfs_buf_offset(bp,
  79                                         (i << mp->m_sb.sb_inodelog));
  80                 di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
  81                             XFS_DINODE_GOOD_VERSION(dip->di_version);
  82                 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
  83                                                 XFS_ERRTAG_ITOBP_INOTOBP,
  84                                                 XFS_RANDOM_ITOBP_INOTOBP))) {
  85                         if (readahead) {
  86                                 bp->b_flags &= ~XBF_DONE;
  87                                 return;
  88                         }
  89
  90                         xfs_buf_ioerror(bp, EFSCORRUPTED);
  91                         xfs_verifier_error(bp);
  92 #ifdef DEBUG
  93                         xfs_alert(mp,
  94                                 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
  95                                 (unsigned long long)bp->b_bn, i,
  96                                 be16_to_cpu(dip->di_magic));
  97 #endif
  98                 }
  99         }
 100         xfs_inobp_check(mp, bp);
 101 }
 102
 103
 104 static void
 105 xfs_inode_buf_read_verify(
 106         struct xfs_buf  *bp)
 107 {
 108         xfs_inode_buf_verify(bp, false);
 109 }
 110
 111 static void
 112 xfs_inode_buf_readahead_verify(
 113         struct xfs_buf  *bp)
 114 {
 115         xfs_inode_buf_verify(bp, true);
 116 }
 117
 118 static void
 119 xfs_inode_buf_write_verify(
 120         struct xfs_buf  *bp)
 121 {
 122         xfs_inode_buf_verify(bp, false);
 123 }
 124
 125 const struct xfs_buf_ops xfs_inode_buf_ops = {
 126         .verify_read = xfs_inode_buf_read_verify,
 127         .verify_write = xfs_inode_buf_write_verify,
 128 };
 129
 130 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
 131         .verify_read = xfs_inode_buf_readahead_verify,
 132         .verify_write = xfs_inode_buf_write_verify,
 133 };
 134
 135
 136 /*
 137  * This routine is called to map an inode to the buffer containing the on-disk
 138  * version of the inode.  It returns a pointer to the buffer containing the
 139  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
 140  * pointer to the on-disk inode within that buffer.
 141  *
 142  * If a non-zero error is returned, then the contents of bpp and dipp are
 143  * undefined.
 144  */
 145 int
 146 xfs_imap_to_bp(
 147         struct xfs_mount        *mp,
 148         struct xfs_trans        *tp,
 149         struct xfs_imap         *imap,
 150         struct xfs_dinode       **dipp,
 151         struct xfs_buf          **bpp,
 152         uint                    buf_flags,
 153         uint                    iget_flags)
 154 {
 155         struct xfs_buf          *bp;
 156         int                     error;
 157
 158         buf_flags |= XBF_UNMAPPED;
 159         error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
 160                                    (int)imap->im_len, buf_flags, &bp,
 161                                    &xfs_inode_buf_ops);
 162         if (error) {
 163                 if (error == EAGAIN) {
 164                         ASSERT(buf_flags & XBF_TRYLOCK);
 165                         return error;
 166                 }
 167
 168                 if (error == EFSCORRUPTED &&
 169                     (iget_flags & XFS_IGET_UNTRUSTED))
 170                         return XFS_ERROR(EINVAL);
 171
 172                 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
 173                         __func__, error);
 174                 return error;
 175         }
 176
 177         *bpp = bp;
 178         *dipp = (struct xfs_dinode *)xfs_buf_offset(bp, imap->im_boffset);
 179         return 0;
 180 }
 181
 182 void
 183 xfs_dinode_from_disk(
 184         xfs_icdinode_t          *to,
 185         xfs_dinode_t            *from)
 186 {
 187         to->di_magic = be16_to_cpu(from->di_magic);
 188         to->di_mode = be16_to_cpu(from->di_mode);
 189         to->di_version = from ->di_version;
 190         to->di_format = from->di_format;
 191         to->di_onlink = be16_to_cpu(from->di_onlink);
 192         to->di_uid = be32_to_cpu(from->di_uid);
 193         to->di_gid = be32_to_cpu(from->di_gid);
 194         to->di_nlink = be32_to_cpu(from->di_nlink);
 195         to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
 196         to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
 197         memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
 198         to->di_flushiter = be16_to_cpu(from->di_flushiter);
 199         to->di_atime.t_sec = be32_to_cpu(from->di_atime.t_sec);
 200         to->di_atime.t_nsec = be32_to_cpu(from->di_atime.t_nsec);
 201         to->di_mtime.t_sec = be32_to_cpu(from->di_mtime.t_sec);
 202         to->di_mtime.t_nsec = be32_to_cpu(from->di_mtime.t_nsec);
 203         to->di_ctime.t_sec = be32_to_cpu(from->di_ctime.t_sec);
 204         to->di_ctime.t_nsec = be32_to_cpu(from->di_ctime.t_nsec);
 205         to->di_size = be64_to_cpu(from->di_size);
 206         to->di_nblocks = be64_to_cpu(from->di_nblocks);
 207         to->di_extsize = be32_to_cpu(from->di_extsize);
 208         to->di_nextents = be32_to_cpu(from->di_nextents);
 209         to->di_anextents = be16_to_cpu(from->di_anextents);
 210         to->di_forkoff = from->di_forkoff;
 211         to->di_aformat  = from->di_aformat;
 212         to->di_dmevmask = be32_to_cpu(from->di_dmevmask);
 213         to->di_dmstate  = be16_to_cpu(from->di_dmstate);
 214         to->di_flags    = be16_to_cpu(from->di_flags);
 215         to->di_gen      = be32_to_cpu(from->di_gen);
 216
 217         if (to->di_version == 3) {
 218                 to->di_changecount = be64_to_cpu(from->di_changecount);
 219                 to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
 220                 to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
 221                 to->di_flags2 = be64_to_cpu(from->di_flags2);
 222                 to->di_ino = be64_to_cpu(from->di_ino);
 223                 to->di_lsn = be64_to_cpu(from->di_lsn);
 224                 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 225                 uuid_copy(&to->di_uuid, &from->di_uuid);
 226         }
 227 }
 228
 229 void
 230 xfs_dinode_to_disk(
 231         xfs_dinode_t            *to,
 232         xfs_icdinode_t          *from)
 233 {
 234         to->di_magic = cpu_to_be16(from->di_magic);
 235         to->di_mode = cpu_to_be16(from->di_mode);
 236         to->di_version = from ->di_version;
 237         to->di_format = from->di_format;
 238         to->di_onlink = cpu_to_be16(from->di_onlink);
 239         to->di_uid = cpu_to_be32(from->di_uid);
 240         to->di_gid = cpu_to_be32(from->di_gid);
 241         to->di_nlink = cpu_to_be32(from->di_nlink);
 242         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 243         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 244         memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
 245         to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
 246         to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
 247         to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
 248         to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
 249         to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
 250         to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
 251         to->di_size = cpu_to_be64(from->di_size);
 252         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 253         to->di_extsize = cpu_to_be32(from->di_extsize);
 254         to->di_nextents = cpu_to_be32(from->di_nextents);
 255         to->di_anextents = cpu_to_be16(from->di_anextents);
 256         to->di_forkoff = from->di_forkoff;
 257         to->di_aformat = from->di_aformat;
 258         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 259         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 260         to->di_flags = cpu_to_be16(from->di_flags);
 261         to->di_gen = cpu_to_be32(from->di_gen);
 262
 263         if (from->di_version == 3) {
 264                 to->di_changecount = cpu_to_be64(from->di_changecount);
 265                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 266                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 267                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 268                 to->di_ino = cpu_to_be64(from->di_ino);
 269                 to->di_lsn = cpu_to_be64(from->di_lsn);
 270                 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 271                 uuid_copy(&to->di_uuid, &from->di_uuid);
 272                 to->di_flushiter = 0;
 273         } else {
 274                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 275         }
 276 }
 277
 278 bool
 279 xfs_dinode_verify(
 280         struct xfs_mount        *mp,
 281         xfs_ino_t               ino,
 282         struct xfs_dinode       *dip)
 283 {
 284         if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
 285                 return false;
 286
 287         /* only version 3 or greater inodes are extensively verified here */
 288         if (dip->di_version < 3)
 289                 return true;
 290
 291         if (!xfs_sb_version_hascrc(&mp->m_sb))
 292                 return false;
 293         if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
 294                               XFS_DINODE_CRC_OFF))
 295                 return false;
 296         if (be64_to_cpu(dip->di_ino) != ino)
 297                 return false;
 298         if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_uuid))
 299                 return false;
 300         return true;
 301 }
 302
 303 void
 304 xfs_dinode_calc_crc(
 305         struct xfs_mount        *mp,
 306         struct xfs_dinode       *dip)
 307 {
 308         __uint32_t              crc;
 309
 310         if (dip->di_version < 3)
 311                 return;
 312
 313         ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
 314         crc = xfs_start_cksum((char *)dip, mp->m_sb.sb_inodesize,
 315                               XFS_DINODE_CRC_OFF);
 316         dip->di_crc = xfs_end_cksum(crc);
 317 }
 318
 319 /*
 320  * Read the disk inode attributes into the in-core inode structure.
 321  */
 322 int
 323 xfs_iread(
 324         xfs_mount_t     *mp,
 325         xfs_trans_t     *tp,
 326         xfs_inode_t     *ip,
 327         uint            iget_flags)
 328 {
 329         xfs_buf_t       *bp;
 330         xfs_dinode_t    *dip;
 331         int             error;
 332
 333         /*
 334          * Fill in the location information in the in-core inode.
 335          */
 336         error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
 337         if (error)
 338                 return error;
 339
 340         /*
 341          * Get pointers to the on-disk inode and the buffer containing it.
 342          */
 343         error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
 344         if (error)
 345                 return error;
 346
 347         /* even unallocated inodes are verified */
 348         if (!xfs_dinode_verify(mp, ip->i_ino, dip)) {
 349                 xfs_alert(mp, "%s: validation failed for inode %lld failed",
 350                                 __func__, ip->i_ino);
 351
 352                 XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, dip);
 353                 error = XFS_ERROR(EFSCORRUPTED);
 354                 goto out_brelse;
 355         }
 356
 357         /*
 358          * If the on-disk inode is already linked to a directory
 359          * entry, copy all of the inode into the in-core inode.
 360          * xfs_iformat_fork() handles copying in the inode format
 361          * specific information.
 362          * Otherwise, just get the truly permanent information.
 363          */
 364         if (dip->di_mode) {
 365                 xfs_dinode_from_disk(&ip->i_d, dip);
 366                 error = xfs_iformat_fork(ip, dip);
 367                 if (error)  {
 368 #ifdef DEBUG
 369                         xfs_alert(mp, "%s: xfs_iformat() returned error %d",
 370                                 __func__, error);
 371 #endif /* DEBUG */
 372                         goto out_brelse;
 373                 }
 374         } else {
 375                 /*
 376                  * Partial initialisation of the in-core inode. Just the bits
 377                  * that xfs_ialloc won't overwrite or relies on being correct.
 378                  */
 379                 ip->i_d.di_magic = be16_to_cpu(dip->di_magic);
 380                 ip->i_d.di_version = dip->di_version;
 381                 ip->i_d.di_gen = be32_to_cpu(dip->di_gen);
 382                 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
 383
 384                 if (dip->di_version == 3) {
 385                         ip->i_d.di_ino = be64_to_cpu(dip->di_ino);
 386                         uuid_copy(&ip->i_d.di_uuid, &dip->di_uuid);
 387                 }
 388
 389                 /*
 390                  * Make sure to pull in the mode here as well in
 391                  * case the inode is released without being used.
 392                  * This ensures that xfs_inactive() will see that
 393                  * the inode is already free and not try to mess
 394                  * with the uninitialized part of it.
 395                  */
 396                 ip->i_d.di_mode = 0;
 397         }
 398
 399         /*
 400          * The inode format changed when we moved the link count and
 401          * made it 32 bits long.  If this is an old format inode,
 402          * convert it in memory to look like a new one.  If it gets
 403          * flushed to disk we will convert back before flushing or
 404          * logging it.  We zero out the new projid field and the old link
 405          * count field.  We'll handle clearing the pad field (the remains
 406          * of the old uuid field) when we actually convert the inode to
 407          * the new format. We don't change the version number so that we
 408          * can distinguish this from a real new format inode.
 409          */
 410         if (ip->i_d.di_version == 1) {
 411                 ip->i_d.di_nlink = ip->i_d.di_onlink;
 412                 ip->i_d.di_onlink = 0;
 413                 xfs_set_projid(&ip->i_d, 0);
 414         }
 415
 416         ip->i_delayed_blks = 0;
 417
 418         /*
 419          * Mark the buffer containing the inode as something to keep
 420          * around for a while.  This helps to keep recently accessed
 421          * meta-data in-core longer.
 422          */
 423         xfs_buf_set_ref(bp, XFS_INO_REF);
 424
 425         /*
 426          * Use xfs_trans_brelse() to release the buffer containing the on-disk
 427          * inode, because it was acquired with xfs_trans_read_buf() in
 428          * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
 429          * brelse().  If we're within a transaction, then xfs_trans_brelse()
 430          * will only release the buffer if it is not dirty within the
 431          * transaction.  It will be OK to release the buffer in this case,
 432          * because inodes on disk are never destroyed and we will be locking the
 433          * new in-core inode before putting it in the cache where other
 434          * processes can find it.  Thus we don't have to worry about the inode
 435          * being changed just because we released the buffer.
 436          */
 437  out_brelse:
 438         xfs_trans_brelse(tp, bp);
 439         return error;
 440 }