fs/xfs/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 #include "xfs.h"
  19 #include "xfs_fs.h"
  20 #include "xfs_shared.h"
  21 #include "xfs_format.h"
  22 #include "xfs_log_format.h"
  23 #include "xfs_trans_resv.h"
  24 #include "xfs_mount.h"
  25 #include "xfs_inode.h"
  26 #include "xfs_error.h"
  27 #include "xfs_cksum.h"
  28 #include "xfs_icache.h"
  29 #include "xfs_trans.h"
  30 #include "xfs_ialloc.h"
  31
  32 /*
  33  * Check that none of the inode's in the buffer have a next
  34  * unlinked field of 0.
  35  */
  36 #if defined(DEBUG)
  37 void
  38 xfs_inobp_check(
  39         xfs_mount_t     *mp,
  40         xfs_buf_t       *bp)
  41 {
  42         int             i;
  43         int             j;
  44         xfs_dinode_t    *dip;
  45
  46         j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
  47
  48         for (i = 0; i < j; i++) {
  49                 dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize);
  50                 if (!dip->di_next_unlinked)  {
  51                         xfs_alert(mp,
  52         "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
  53                                 i, (long long)bp->b_bn);
  54                 }
  55         }
  56 }
  57 #endif
  58
  59 /*
  60  * If we are doing readahead on an inode buffer, we might be in log recovery
  61  * reading an inode allocation buffer that hasn't yet been replayed, and hence
  62  * has not had the inode cores stamped into it. Hence for readahead, the buffer
  63  * may be potentially invalid.
  64  *
  65  * If the readahead buffer is invalid, we need to mark it with an error and
  66  * clear the DONE status of the buffer so that a followup read will re-read it
  67  * from disk. We don't report the error otherwise to avoid warnings during log
  68  * recovery and we don't get unnecssary panics on debug kernels. We use EIO here
  69  * because all we want to do is say readahead failed; there is no-one to report
  70  * the error to, so this will distinguish it from a non-ra verifier failure.
  71  * Changes to this readahead error behavour also need to be reflected in
  72  * xfs_dquot_buf_readahead_verify().
  73  */
  74 static void
  75 xfs_inode_buf_verify(
  76         struct xfs_buf  *bp,
  77         bool            readahead)
  78 {
  79         struct xfs_mount *mp = bp->b_target->bt_mount;
  80         int             i;
  81         int             ni;
  82
  83         /*
  84          * Validate the magic number and version of every inode in the buffer
  85          */
  86         ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
  87         for (i = 0; i < ni; i++) {
  88                 int             di_ok;
  89                 xfs_dinode_t    *dip;
  90
  91                 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
  92                 di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
  93                             XFS_DINODE_GOOD_VERSION(dip->di_version);
  94                 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
  95                                                 XFS_ERRTAG_ITOBP_INOTOBP,
  96                                                 XFS_RANDOM_ITOBP_INOTOBP))) {
  97                         if (readahead) {
  98                                 bp->b_flags &= ~XBF_DONE;
  99                                 xfs_buf_ioerror(bp, -EIO);
 100                                 return;
 101                         }
 102
 103                         xfs_buf_ioerror(bp, -EFSCORRUPTED);
 104                         xfs_verifier_error(bp);
 105 #ifdef DEBUG
 106                         xfs_alert(mp,
 107                                 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
 108                                 (unsigned long long)bp->b_bn, i,
 109                                 be16_to_cpu(dip->di_magic));
 110 #endif
 111                 }
 112         }
 113         xfs_inobp_check(mp, bp);
 114 }
 115
 116
 117 static void
 118 xfs_inode_buf_read_verify(
 119         struct xfs_buf  *bp)
 120 {
 121         xfs_inode_buf_verify(bp, false);
 122 }
 123
 124 static void
 125 xfs_inode_buf_readahead_verify(
 126         struct xfs_buf  *bp)
 127 {
 128         xfs_inode_buf_verify(bp, true);
 129 }
 130
 131 static void
 132 xfs_inode_buf_write_verify(
 133         struct xfs_buf  *bp)
 134 {
 135         xfs_inode_buf_verify(bp, false);
 136 }
 137
 138 const struct xfs_buf_ops xfs_inode_buf_ops = {
 139         .name = "xfs_inode",
 140         .verify_read = xfs_inode_buf_read_verify,
 141         .verify_write = xfs_inode_buf_write_verify,
 142 };
 143
 144 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
 145         .name = "xxfs_inode_ra",
 146         .verify_read = xfs_inode_buf_readahead_verify,
 147         .verify_write = xfs_inode_buf_write_verify,
 148 };
 149
 150
 151 /*
 152  * This routine is called to map an inode to the buffer containing the on-disk
 153  * version of the inode.  It returns a pointer to the buffer containing the
 154  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
 155  * pointer to the on-disk inode within that buffer.
 156  *
 157  * If a non-zero error is returned, then the contents of bpp and dipp are
 158  * undefined.
 159  */
 160 int
 161 xfs_imap_to_bp(
 162         struct xfs_mount        *mp,
 163         struct xfs_trans        *tp,
 164         struct xfs_imap         *imap,
 165         struct xfs_dinode       **dipp,
 166         struct xfs_buf          **bpp,
 167         uint                    buf_flags,
 168         uint                    iget_flags)
 169 {
 170         struct xfs_buf          *bp;
 171         int                     error;
 172
 173         buf_flags |= XBF_UNMAPPED;
 174         error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
 175                                    (int)imap->im_len, buf_flags, &bp,
 176                                    &xfs_inode_buf_ops);
 177         if (error) {
 178                 if (error == -EAGAIN) {
 179                         ASSERT(buf_flags & XBF_TRYLOCK);
 180                         return error;
 181                 }
 182
 183                 if (error == -EFSCORRUPTED &&
 184                     (iget_flags & XFS_IGET_UNTRUSTED))
 185                         return -EINVAL;
 186
 187                 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
 188                         __func__, error);
 189                 return error;
 190         }
 191
 192         *bpp = bp;
 193         *dipp = xfs_buf_offset(bp, imap->im_boffset);
 194         return 0;
 195 }
 196
 197 void
 198 xfs_inode_from_disk(
 199         struct xfs_inode        *ip,
 200         struct xfs_dinode       *from)
 201 {
 202         struct xfs_icdinode     *to = &ip->i_d;
 203         struct inode            *inode = VFS_I(ip);
 204
 205         to->di_mode = be16_to_cpu(from->di_mode);
 206         to->di_version = from ->di_version;
 207
 208         /*
 209          * Convert v1 inodes immediately to v2 inode format as this is the
 210          * minimum inode version format we support in the rest of the code.
 211          */
 212         if (to->di_version == 1) {
 213                 set_nlink(inode, be16_to_cpu(from->di_onlink));
 214                 to->di_projid_lo = 0;
 215                 to->di_projid_hi = 0;
 216                 to->di_version = 2;
 217         } else {
 218                 set_nlink(inode, be32_to_cpu(from->di_nlink));
 219                 to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
 220                 to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
 221         }
 222
 223         to->di_format = from->di_format;
 224         to->di_uid = be32_to_cpu(from->di_uid);
 225         to->di_gid = be32_to_cpu(from->di_gid);
 226         to->di_flushiter = be16_to_cpu(from->di_flushiter);
 227
 228         /*
 229          * Time is signed, so need to convert to signed 32 bit before
 230          * storing in inode timestamp which may be 64 bit. Otherwise
 231          * a time before epoch is converted to a time long after epoch
 232          * on 64 bit systems.
 233          */
 234         inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
 235         inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
 236         inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
 237         inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
 238         inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
 239         inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
 240
 241         to->di_size = be64_to_cpu(from->di_size);
 242         to->di_nblocks = be64_to_cpu(from->di_nblocks);
 243         to->di_extsize = be32_to_cpu(from->di_extsize);
 244         to->di_nextents = be32_to_cpu(from->di_nextents);
 245         to->di_anextents = be16_to_cpu(from->di_anextents);
 246         to->di_forkoff = from->di_forkoff;
 247         to->di_aformat  = from->di_aformat;
 248         to->di_dmevmask = be32_to_cpu(from->di_dmevmask);
 249         to->di_dmstate  = be16_to_cpu(from->di_dmstate);
 250         to->di_flags    = be16_to_cpu(from->di_flags);
 251         to->di_gen      = be32_to_cpu(from->di_gen);
 252
 253         if (to->di_version == 3) {
 254                 to->di_changecount = be64_to_cpu(from->di_changecount);
 255                 to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
 256                 to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
 257                 to->di_flags2 = be64_to_cpu(from->di_flags2);
 258         }
 259 }
 260
 261 void
 262 xfs_inode_to_disk(
 263         struct xfs_inode        *ip,
 264         struct xfs_dinode       *to,
 265         xfs_lsn_t               lsn)
 266 {
 267         struct xfs_icdinode     *from = &ip->i_d;
 268         struct inode            *inode = VFS_I(ip);
 269
 270         to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
 271         to->di_onlink = 0;
 272
 273         to->di_mode = cpu_to_be16(from->di_mode);
 274         to->di_version = from->di_version;
 275         to->di_format = from->di_format;
 276         to->di_uid = cpu_to_be32(from->di_uid);
 277         to->di_gid = cpu_to_be32(from->di_gid);
 278         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 279         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 280
 281         memset(to->di_pad, 0, sizeof(to->di_pad));
 282         to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
 283         to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
 284         to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
 285         to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
 286         to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
 287         to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
 288         to->di_nlink = cpu_to_be32(inode->i_nlink);
 289
 290         to->di_size = cpu_to_be64(from->di_size);
 291         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 292         to->di_extsize = cpu_to_be32(from->di_extsize);
 293         to->di_nextents = cpu_to_be32(from->di_nextents);
 294         to->di_anextents = cpu_to_be16(from->di_anextents);
 295         to->di_forkoff = from->di_forkoff;
 296         to->di_aformat = from->di_aformat;
 297         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 298         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 299         to->di_flags = cpu_to_be16(from->di_flags);
 300         to->di_gen = cpu_to_be32(from->di_gen);
 301
 302         if (from->di_version == 3) {
 303                 to->di_changecount = cpu_to_be64(from->di_changecount);
 304                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 305                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 306                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 307
 308                 to->di_ino = cpu_to_be64(ip->i_ino);
 309                 to->di_lsn = cpu_to_be64(lsn);
 310                 memset(to->di_pad2, 0, sizeof(to->di_pad2));
 311                 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
 312                 to->di_flushiter = 0;
 313         } else {
 314                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 315         }
 316 }
 317
 318 void
 319 xfs_log_dinode_to_disk(
 320         struct xfs_log_dinode   *from,
 321         struct xfs_dinode       *to)
 322 {
 323         to->di_magic = cpu_to_be16(from->di_magic);
 324         to->di_mode = cpu_to_be16(from->di_mode);
 325         to->di_version = from->di_version;
 326         to->di_format = from->di_format;
 327         to->di_onlink = 0;
 328         to->di_uid = cpu_to_be32(from->di_uid);
 329         to->di_gid = cpu_to_be32(from->di_gid);
 330         to->di_nlink = cpu_to_be32(from->di_nlink);
 331         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 332         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 333         memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
 334
 335         to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
 336         to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
 337         to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
 338         to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
 339         to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
 340         to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
 341
 342         to->di_size = cpu_to_be64(from->di_size);
 343         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 344         to->di_extsize = cpu_to_be32(from->di_extsize);
 345         to->di_nextents = cpu_to_be32(from->di_nextents);
 346         to->di_anextents = cpu_to_be16(from->di_anextents);
 347         to->di_forkoff = from->di_forkoff;
 348         to->di_aformat = from->di_aformat;
 349         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 350         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 351         to->di_flags = cpu_to_be16(from->di_flags);
 352         to->di_gen = cpu_to_be32(from->di_gen);
 353
 354         if (from->di_version == 3) {
 355                 to->di_changecount = cpu_to_be64(from->di_changecount);
 356                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 357                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 358                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 359                 to->di_ino = cpu_to_be64(from->di_ino);
 360                 to->di_lsn = cpu_to_be64(from->di_lsn);
 361                 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 362                 uuid_copy(&to->di_uuid, &from->di_uuid);
 363                 to->di_flushiter = 0;
 364         } else {
 365                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 366         }
 367 }
 368
 369 static bool
 370 xfs_dinode_verify(
 371         struct xfs_mount        *mp,
 372         struct xfs_inode        *ip,
 373         struct xfs_dinode       *dip)
 374 {
 375         if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
 376                 return false;
 377
 378         /* only version 3 or greater inodes are extensively verified here */
 379         if (dip->di_version < 3)
 380                 return true;
 381
 382         if (!xfs_sb_version_hascrc(&mp->m_sb))
 383                 return false;
 384         if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
 385                               XFS_DINODE_CRC_OFF))
 386                 return false;
 387         if (be64_to_cpu(dip->di_ino) != ip->i_ino)
 388                 return false;
 389         if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
 390                 return false;
 391         return true;
 392 }
 393
 394 void
 395 xfs_dinode_calc_crc(
 396         struct xfs_mount        *mp,
 397         struct xfs_dinode       *dip)
 398 {
 399         __uint32_t              crc;
 400
 401         if (dip->di_version < 3)
 402                 return;
 403
 404         ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
 405         crc = xfs_start_cksum((char *)dip, mp->m_sb.sb_inodesize,
 406                               XFS_DINODE_CRC_OFF);
 407         dip->di_crc = xfs_end_cksum(crc);
 408 }
 409
 410 /*
 411  * Read the disk inode attributes into the in-core inode structure.
 412  *
 413  * For version 5 superblocks, if we are initialising a new inode and we are not
 414  * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
 415  * inode core with a random generation number. If we are keeping inodes around,
 416  * we need to read the inode cluster to get the existing generation number off
 417  * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
 418  * format) then log recovery is dependent on the di_flushiter field being
 419  * initialised from the current on-disk value and hence we must also read the
 420  * inode off disk.
 421  */
 422 int
 423 xfs_iread(
 424         xfs_mount_t     *mp,
 425         xfs_trans_t     *tp,
 426         xfs_inode_t     *ip,
 427         uint            iget_flags)
 428 {
 429         xfs_buf_t       *bp;
 430         xfs_dinode_t    *dip;
 431         int             error;
 432
 433         /*
 434          * Fill in the location information in the in-core inode.
 435          */
 436         error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
 437         if (error)
 438                 return error;
 439
 440         /* shortcut IO on inode allocation if possible */
 441         if ((iget_flags & XFS_IGET_CREATE) &&
 442             xfs_sb_version_hascrc(&mp->m_sb) &&
 443             !(mp->m_flags & XFS_MOUNT_IKEEP)) {
 444                 /* initialise the on-disk inode core */
 445                 memset(&ip->i_d, 0, sizeof(ip->i_d));
 446                 ip->i_d.di_gen = prandom_u32();
 447                 if (xfs_sb_version_hascrc(&mp->m_sb))
 448                         ip->i_d.di_version = 3;
 449                 else
 450                         ip->i_d.di_version = 2;
 451                 return 0;
 452         }
 453
 454         /*
 455          * Get pointers to the on-disk inode and the buffer containing it.
 456          */
 457         error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
 458         if (error)
 459                 return error;
 460
 461         /* even unallocated inodes are verified */
 462         if (!xfs_dinode_verify(mp, ip, dip)) {
 463                 xfs_alert(mp, "%s: validation failed for inode %lld failed",
 464                                 __func__, ip->i_ino);
 465
 466                 XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, dip);
 467                 error = -EFSCORRUPTED;
 468                 goto out_brelse;
 469         }
 470
 471         /*
 472          * If the on-disk inode is already linked to a directory
 473          * entry, copy all of the inode into the in-core inode.
 474          * xfs_iformat_fork() handles copying in the inode format
 475          * specific information.
 476          * Otherwise, just get the truly permanent information.
 477          */
 478         if (dip->di_mode) {
 479                 xfs_inode_from_disk(ip, dip);
 480                 error = xfs_iformat_fork(ip, dip);
 481                 if (error)  {
 482 #ifdef DEBUG
 483                         xfs_alert(mp, "%s: xfs_iformat() returned error %d",
 484                                 __func__, error);
 485 #endif /* DEBUG */
 486                         goto out_brelse;
 487                 }
 488         } else {
 489                 /*
 490                  * Partial initialisation of the in-core inode. Just the bits
 491                  * that xfs_ialloc won't overwrite or relies on being correct.
 492                  */
 493                 ip->i_d.di_version = dip->di_version;
 494                 ip->i_d.di_gen = be32_to_cpu(dip->di_gen);
 495                 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
 496
 497                 /*
 498                  * Make sure to pull in the mode here as well in
 499                  * case the inode is released without being used.
 500                  * This ensures that xfs_inactive() will see that
 501                  * the inode is already free and not try to mess
 502                  * with the uninitialized part of it.
 503                  */
 504                 ip->i_d.di_mode = 0;
 505         }
 506
 507         ASSERT(ip->i_d.di_version >= 2);
 508         ip->i_delayed_blks = 0;
 509
 510         /*
 511          * Mark the buffer containing the inode as something to keep
 512          * around for a while.  This helps to keep recently accessed
 513          * meta-data in-core longer.
 514          */
 515         xfs_buf_set_ref(bp, XFS_INO_REF);
 516
 517         /*
 518          * Use xfs_trans_brelse() to release the buffer containing the on-disk
 519          * inode, because it was acquired with xfs_trans_read_buf() in
 520          * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
 521          * brelse().  If we're within a transaction, then xfs_trans_brelse()
 522          * will only release the buffer if it is not dirty within the
 523          * transaction.  It will be OK to release the buffer in this case,
 524          * because inodes on disk are never destroyed and we will be locking the
 525          * new in-core inode before putting it in the cache where other
 526          * processes can find it.  Thus we don't have to worry about the inode
 527          * being changed just because we released the buffer.
 528          */
 529  out_brelse:
 530         xfs_trans_brelse(tp, bp);
 531         return error;
 532 }