libxfs/xfs_inode_buf.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (c) 2000-2006 Silicon Graphics, Inc.
   4  * All Rights Reserved.
   5  */
   6 #include "libxfs_priv.h"
   7 #include "xfs_fs.h"
   8 #include "xfs_shared.h"
   9 #include "xfs_format.h"
  10 #include "xfs_log_format.h"
  11 #include "xfs_trans_resv.h"
  12 #include "xfs_mount.h"
  13 #include "xfs_inode.h"
  14 #include "xfs_trans.h"
  15 #include "xfs_ialloc.h"
  16 #include "xfs_dir2.h"
  17
  18
  19 /*
  20  * Check that none of the inode's in the buffer have a next
  21  * unlinked field of 0.
  22  */
  23 #if defined(DEBUG)
  24 void
  25 xfs_inobp_check(
  26         xfs_mount_t     *mp,
  27         xfs_buf_t       *bp)
  28 {
  29         int             i;
  30         xfs_dinode_t    *dip;
  31
  32         for (i = 0; i < M_IGEO(mp)->inodes_per_cluster; i++) {
  33                 dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize);
  34                 if (!dip->di_next_unlinked)  {
  35                         xfs_alert(mp,
  36         "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
  37                                 i, (long long)bp->b_bn);
  38                 }
  39         }
  40 }
  41 #endif
  42
  43 bool
  44 xfs_dinode_good_version(
  45         struct xfs_mount *mp,
  46         __u8            version)
  47 {
  48         if (xfs_sb_version_hascrc(&mp->m_sb))
  49                 return version == 3;
  50
  51         return version == 1 || version == 2;
  52 }
  53
  54 /*
  55  * If we are doing readahead on an inode buffer, we might be in log recovery
  56  * reading an inode allocation buffer that hasn't yet been replayed, and hence
  57  * has not had the inode cores stamped into it. Hence for readahead, the buffer
  58  * may be potentially invalid.
  59  *
  60  * If the readahead buffer is invalid, we need to mark it with an error and
  61  * clear the DONE status of the buffer so that a followup read will re-read it
  62  * from disk. We don't report the error otherwise to avoid warnings during log
  63  * recovery and we don't get unnecssary panics on debug kernels. We use EIO here
  64  * because all we want to do is say readahead failed; there is no-one to report
  65  * the error to, so this will distinguish it from a non-ra verifier failure.
  66  * Changes to this readahead error behavour also need to be reflected in
  67  * xfs_dquot_buf_readahead_verify().
  68  */
  69 static void
  70 xfs_inode_buf_verify(
  71         struct xfs_buf  *bp,
  72         bool            readahead)
  73 {
  74         struct xfs_mount *mp = bp->b_mount;
  75         xfs_agnumber_t  agno;
  76         int             i;
  77         int             ni;
  78
  79         /*
  80          * Validate the magic number and version of every inode in the buffer
  81          */
  82         agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp));
  83         ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
  84         for (i = 0; i < ni; i++) {
  85                 int             di_ok;
  86                 xfs_dinode_t    *dip;
  87                 xfs_agino_t     unlinked_ino;
  88
  89                 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
  90                 unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
  91                 di_ok = xfs_verify_magic16(bp, dip->di_magic) &&
  92                         xfs_dinode_good_version(mp, dip->di_version) &&
  93                         xfs_verify_agino_or_null(mp, agno, unlinked_ino);
  94                 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
  95                                                 XFS_ERRTAG_ITOBP_INOTOBP))) {
  96                         if (readahead) {
  97                                 bp->b_flags &= ~XBF_DONE;
  98                                 xfs_buf_ioerror(bp, -EIO);
  99                                 return;
 100                         }
 101
 102 #ifdef DEBUG
 103                         xfs_alert(mp,
 104                                 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
 105                                 (unsigned long long)bp->b_bn, i,
 106                                 be16_to_cpu(dip->di_magic));
 107 #endif
 108                         xfs_buf_verifier_error(bp, -EFSCORRUPTED,
 109                                         __func__, dip, sizeof(*dip),
 110                                         NULL);
 111                         return;
 112                 }
 113         }
 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         .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
 141                      cpu_to_be16(XFS_DINODE_MAGIC) },
 142         .verify_read = xfs_inode_buf_read_verify,
 143         .verify_write = xfs_inode_buf_write_verify,
 144 };
 145
 146 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
 147         .name = "xfs_inode_ra",
 148         .magic16 = { cpu_to_be16(XFS_DINODE_MAGIC),
 149                      cpu_to_be16(XFS_DINODE_MAGIC) },
 150         .verify_read = xfs_inode_buf_readahead_verify,
 151         .verify_write = xfs_inode_buf_write_verify,
 152 };
 153
 154
 155 /*
 156  * This routine is called to map an inode to the buffer containing the on-disk
 157  * version of the inode.  It returns a pointer to the buffer containing the
 158  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
 159  * pointer to the on-disk inode within that buffer.
 160  *
 161  * If a non-zero error is returned, then the contents of bpp and dipp are
 162  * undefined.
 163  */
 164 int
 165 xfs_imap_to_bp(
 166         struct xfs_mount        *mp,
 167         struct xfs_trans        *tp,
 168         struct xfs_imap         *imap,
 169         struct xfs_dinode       **dipp,
 170         struct xfs_buf          **bpp,
 171         uint                    buf_flags,
 172         uint                    iget_flags)
 173 {
 174         struct xfs_buf          *bp;
 175         int                     error;
 176
 177         buf_flags |= XBF_UNMAPPED;
 178         error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
 179                                    (int)imap->im_len, buf_flags, &bp,
 180                                    &xfs_inode_buf_ops);
 181         if (error) {
 182                 if (error == -EAGAIN) {
 183                         ASSERT(buf_flags & XBF_TRYLOCK);
 184                         return error;
 185                 }
 186                 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
 187                         __func__, error);
 188                 return error;
 189         }
 190
 191         *bpp = bp;
 192         *dipp = xfs_buf_offset(bp, imap->im_boffset);
 193         return 0;
 194 }
 195
 196 void
 197 xfs_inode_from_disk(
 198         struct xfs_inode        *ip,
 199         struct xfs_dinode       *from)
 200 {
 201         struct xfs_icdinode     *to = &ip->i_d;
 202         struct inode            *inode = VFS_I(ip);
 203
 204
 205         /*
 206          * Convert v1 inodes immediately to v2 inode format as this is the
 207          * minimum inode version format we support in the rest of the code.
 208          */
 209         to->di_version = from->di_version;
 210         if (to->di_version == 1) {
 211                 set_nlink(inode, be16_to_cpu(from->di_onlink));
 212                 to->di_projid_lo = 0;
 213                 to->di_projid_hi = 0;
 214                 to->di_version = 2;
 215         } else {
 216                 set_nlink(inode, be32_to_cpu(from->di_nlink));
 217                 to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
 218                 to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
 219         }
 220
 221         to->di_format = from->di_format;
 222         to->di_uid = be32_to_cpu(from->di_uid);
 223         to->di_gid = be32_to_cpu(from->di_gid);
 224         to->di_flushiter = be16_to_cpu(from->di_flushiter);
 225
 226         /*
 227          * Time is signed, so need to convert to signed 32 bit before
 228          * storing in inode timestamp which may be 64 bit. Otherwise
 229          * a time before epoch is converted to a time long after epoch
 230          * on 64 bit systems.
 231          */
 232         inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
 233         inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
 234         inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
 235         inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
 236         inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
 237         inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
 238         inode->i_generation = be32_to_cpu(from->di_gen);
 239         inode->i_mode = be16_to_cpu(from->di_mode);
 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
 252         if (to->di_version == 3) {
 253                 inode_set_iversion_queried(inode,
 254                                            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                 to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
 259         }
 260 }
 261
 262 void
 263 xfs_inode_to_disk(
 264         struct xfs_inode        *ip,
 265         struct xfs_dinode       *to,
 266         xfs_lsn_t               lsn)
 267 {
 268         struct xfs_icdinode     *from = &ip->i_d;
 269         struct inode            *inode = VFS_I(ip);
 270
 271         to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
 272         to->di_onlink = 0;
 273
 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         to->di_gen = cpu_to_be32(inode->i_generation);
 290         to->di_mode = cpu_to_be16(inode->i_mode);
 291
 292         to->di_size = cpu_to_be64(from->di_size);
 293         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 294         to->di_extsize = cpu_to_be32(from->di_extsize);
 295         to->di_nextents = cpu_to_be32(from->di_nextents);
 296         to->di_anextents = cpu_to_be16(from->di_anextents);
 297         to->di_forkoff = from->di_forkoff;
 298         to->di_aformat = from->di_aformat;
 299         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 300         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 301         to->di_flags = cpu_to_be16(from->di_flags);
 302
 303         if (from->di_version == 3) {
 304                 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
 305                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 306                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 307                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 308                 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 309                 to->di_ino = cpu_to_be64(ip->i_ino);
 310                 to->di_lsn = cpu_to_be64(lsn);
 311                 memset(to->di_pad2, 0, sizeof(to->di_pad2));
 312                 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
 313                 to->di_flushiter = 0;
 314         } else {
 315                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 316         }
 317 }
 318
 319 void
 320 xfs_log_dinode_to_disk(
 321         struct xfs_log_dinode   *from,
 322         struct xfs_dinode       *to)
 323 {
 324         to->di_magic = cpu_to_be16(from->di_magic);
 325         to->di_mode = cpu_to_be16(from->di_mode);
 326         to->di_version = from->di_version;
 327         to->di_format = from->di_format;
 328         to->di_onlink = 0;
 329         to->di_uid = cpu_to_be32(from->di_uid);
 330         to->di_gid = cpu_to_be32(from->di_gid);
 331         to->di_nlink = cpu_to_be32(from->di_nlink);
 332         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 333         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 334         memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
 335
 336         to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
 337         to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
 338         to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
 339         to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
 340         to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
 341         to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
 342
 343         to->di_size = cpu_to_be64(from->di_size);
 344         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 345         to->di_extsize = cpu_to_be32(from->di_extsize);
 346         to->di_nextents = cpu_to_be32(from->di_nextents);
 347         to->di_anextents = cpu_to_be16(from->di_anextents);
 348         to->di_forkoff = from->di_forkoff;
 349         to->di_aformat = from->di_aformat;
 350         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 351         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 352         to->di_flags = cpu_to_be16(from->di_flags);
 353         to->di_gen = cpu_to_be32(from->di_gen);
 354
 355         if (from->di_version == 3) {
 356                 to->di_changecount = cpu_to_be64(from->di_changecount);
 357                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 358                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 359                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 360                 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 361                 to->di_ino = cpu_to_be64(from->di_ino);
 362                 to->di_lsn = cpu_to_be64(from->di_lsn);
 363                 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 364                 uuid_copy(&to->di_uuid, &from->di_uuid);
 365                 to->di_flushiter = 0;
 366         } else {
 367                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 368         }
 369 }
 370
 371 static xfs_failaddr_t
 372 xfs_dinode_verify_fork(
 373         struct xfs_dinode       *dip,
 374         struct xfs_mount        *mp,
 375         int                     whichfork)
 376 {
 377         uint32_t                di_nextents = XFS_DFORK_NEXTENTS(dip, whichfork);
 378
 379         switch (XFS_DFORK_FORMAT(dip, whichfork)) {
 380         case XFS_DINODE_FMT_LOCAL:
 381                 /*
 382                  * no local regular files yet
 383                  */
 384                 if (whichfork == XFS_DATA_FORK) {
 385                         if (S_ISREG(be16_to_cpu(dip->di_mode)))
 386                                 return __this_address;
 387                         if (be64_to_cpu(dip->di_size) >
 388                                         XFS_DFORK_SIZE(dip, mp, whichfork))
 389                                 return __this_address;
 390                 }
 391                 if (di_nextents)
 392                         return __this_address;
 393                 break;
 394         case XFS_DINODE_FMT_EXTENTS:
 395                 if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
 396                         return __this_address;
 397                 break;
 398         case XFS_DINODE_FMT_BTREE:
 399                 if (whichfork == XFS_ATTR_FORK) {
 400                         if (di_nextents > MAXAEXTNUM)
 401                                 return __this_address;
 402                 } else if (di_nextents > MAXEXTNUM) {
 403                         return __this_address;
 404                 }
 405                 break;
 406         default:
 407                 return __this_address;
 408         }
 409         return NULL;
 410 }
 411
 412 static xfs_failaddr_t
 413 xfs_dinode_verify_forkoff(
 414         struct xfs_dinode       *dip,
 415         struct xfs_mount        *mp)
 416 {
 417         if (!XFS_DFORK_Q(dip))
 418                 return NULL;
 419
 420         switch (dip->di_format)  {
 421         case XFS_DINODE_FMT_DEV:
 422                 if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
 423                         return __this_address;
 424                 break;
 425         case XFS_DINODE_FMT_LOCAL:      /* fall through ... */
 426         case XFS_DINODE_FMT_EXTENTS:    /* fall through ... */
 427         case XFS_DINODE_FMT_BTREE:
 428                 if (dip->di_forkoff >= (XFS_LITINO(mp, dip->di_version) >> 3))
 429                         return __this_address;
 430                 break;
 431         default:
 432                 return __this_address;
 433         }
 434         return NULL;
 435 }
 436
 437 xfs_failaddr_t
 438 xfs_dinode_verify(
 439         struct xfs_mount        *mp,
 440         xfs_ino_t               ino,
 441         struct xfs_dinode       *dip)
 442 {
 443         xfs_failaddr_t          fa;
 444         uint16_t                mode;
 445         uint16_t                flags;
 446         uint64_t                flags2;
 447         uint64_t                di_size;
 448
 449         if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
 450                 return __this_address;
 451
 452         /* Verify v3 integrity information first */
 453         if (dip->di_version >= 3) {
 454                 if (!xfs_sb_version_hascrc(&mp->m_sb))
 455                         return __this_address;
 456                 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
 457                                       XFS_DINODE_CRC_OFF))
 458                         return __this_address;
 459                 if (be64_to_cpu(dip->di_ino) != ino)
 460                         return __this_address;
 461                 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
 462                         return __this_address;
 463         }
 464
 465         /* don't allow invalid i_size */
 466         di_size = be64_to_cpu(dip->di_size);
 467         if (di_size & (1ULL << 63))
 468                 return __this_address;
 469
 470         mode = be16_to_cpu(dip->di_mode);
 471         if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
 472                 return __this_address;
 473
 474         /* No zero-length symlinks/dirs. */
 475         if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
 476                 return __this_address;
 477
 478         /* Fork checks carried over from xfs_iformat_fork */
 479         if (mode &&
 480             be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
 481                         be64_to_cpu(dip->di_nblocks))
 482                 return __this_address;
 483
 484         if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
 485                 return __this_address;
 486
 487         flags = be16_to_cpu(dip->di_flags);
 488
 489         if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
 490                 return __this_address;
 491
 492         /* check for illegal values of forkoff */
 493         fa = xfs_dinode_verify_forkoff(dip, mp);
 494         if (fa)
 495                 return fa;
 496
 497         /* Do we have appropriate data fork formats for the mode? */
 498         switch (mode & S_IFMT) {
 499         case S_IFIFO:
 500         case S_IFCHR:
 501         case S_IFBLK:
 502         case S_IFSOCK:
 503                 if (dip->di_format != XFS_DINODE_FMT_DEV)
 504                         return __this_address;
 505                 break;
 506         case S_IFREG:
 507         case S_IFLNK:
 508         case S_IFDIR:
 509                 fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
 510                 if (fa)
 511                         return fa;
 512                 break;
 513         case 0:
 514                 /* Uninitialized inode ok. */
 515                 break;
 516         default:
 517                 return __this_address;
 518         }
 519
 520         if (XFS_DFORK_Q(dip)) {
 521                 fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
 522                 if (fa)
 523                         return fa;
 524         } else {
 525                 /*
 526                  * If there is no fork offset, this may be a freshly-made inode
 527                  * in a new disk cluster, in which case di_aformat is zeroed.
 528                  * Otherwise, such an inode must be in EXTENTS format; this goes
 529                  * for freed inodes as well.
 530                  */
 531                 switch (dip->di_aformat) {
 532                 case 0:
 533                 case XFS_DINODE_FMT_EXTENTS:
 534                         break;
 535                 default:
 536                         return __this_address;
 537                 }
 538                 if (dip->di_anextents)
 539                         return __this_address;
 540         }
 541
 542         /* extent size hint validation */
 543         fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
 544                         mode, flags);
 545         if (fa)
 546                 return fa;
 547
 548         /* only version 3 or greater inodes are extensively verified here */
 549         if (dip->di_version < 3)
 550                 return NULL;
 551
 552         flags2 = be64_to_cpu(dip->di_flags2);
 553
 554         /* don't allow reflink/cowextsize if we don't have reflink */
 555         if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
 556              !xfs_sb_version_hasreflink(&mp->m_sb))
 557                 return __this_address;
 558
 559         /* only regular files get reflink */
 560         if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
 561                 return __this_address;
 562
 563         /* don't let reflink and realtime mix */
 564         if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
 565                 return __this_address;
 566
 567         /* don't let reflink and dax mix */
 568         if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
 569                 return __this_address;
 570
 571         /* COW extent size hint validation */
 572         fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
 573                         mode, flags, flags2);
 574         if (fa)
 575                 return fa;
 576
 577         return NULL;
 578 }
 579
 580 void
 581 xfs_dinode_calc_crc(
 582         struct xfs_mount        *mp,
 583         struct xfs_dinode       *dip)
 584 {
 585         uint32_t                crc;
 586
 587         if (dip->di_version < 3)
 588                 return;
 589
 590         ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
 591         crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
 592                               XFS_DINODE_CRC_OFF);
 593         dip->di_crc = xfs_end_cksum(crc);
 594 }
 595
 596 /*
 597  * Read the disk inode attributes into the in-core inode structure.
 598  *
 599  * For version 5 superblocks, if we are initialising a new inode and we are not
 600  * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
 601  * inode core with a random generation number. If we are keeping inodes around,
 602  * we need to read the inode cluster to get the existing generation number off
 603  * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
 604  * format) then log recovery is dependent on the di_flushiter field being
 605  * initialised from the current on-disk value and hence we must also read the
 606  * inode off disk.
 607  */
 608 int
 609 xfs_iread(
 610         xfs_mount_t     *mp,
 611         xfs_trans_t     *tp,
 612         xfs_inode_t     *ip,
 613         uint            iget_flags)
 614 {
 615         xfs_buf_t       *bp;
 616         xfs_dinode_t    *dip;
 617         xfs_failaddr_t  fa;
 618         int             error;
 619
 620         /*
 621          * Fill in the location information in the in-core inode.
 622          */
 623         error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
 624         if (error)
 625                 return error;
 626
 627         /* shortcut IO on inode allocation if possible */
 628         if ((iget_flags & XFS_IGET_CREATE) &&
 629             xfs_sb_version_hascrc(&mp->m_sb) &&
 630             !(mp->m_flags & XFS_MOUNT_IKEEP)) {
 631                 /* initialise the on-disk inode core */
 632                 memset(&ip->i_d, 0, sizeof(ip->i_d));
 633                 VFS_I(ip)->i_generation = prandom_u32();
 634                 ip->i_d.di_version = 3;
 635                 return 0;
 636         }
 637
 638         /*
 639          * Get pointers to the on-disk inode and the buffer containing it.
 640          */
 641         error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
 642         if (error)
 643                 return error;
 644
 645         /* even unallocated inodes are verified */
 646         fa = xfs_dinode_verify(mp, ip->i_ino, dip);
 647         if (fa) {
 648                 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
 649                                 sizeof(*dip), fa);
 650                 error = -EFSCORRUPTED;
 651                 goto out_brelse;
 652         }
 653
 654         /*
 655          * If the on-disk inode is already linked to a directory
 656          * entry, copy all of the inode into the in-core inode.
 657          * xfs_iformat_fork() handles copying in the inode format
 658          * specific information.
 659          * Otherwise, just get the truly permanent information.
 660          */
 661         if (dip->di_mode) {
 662                 xfs_inode_from_disk(ip, dip);
 663                 error = xfs_iformat_fork(ip, dip);
 664                 if (error)  {
 665 #ifdef DEBUG
 666                         xfs_alert(mp, "%s: xfs_iformat() returned error %d",
 667                                 __func__, error);
 668 #endif /* DEBUG */
 669                         goto out_brelse;
 670                 }
 671         } else {
 672                 /*
 673                  * Partial initialisation of the in-core inode. Just the bits
 674                  * that xfs_ialloc won't overwrite or relies on being correct.
 675                  */
 676                 ip->i_d.di_version = dip->di_version;
 677                 VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
 678                 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
 679
 680                 /*
 681                  * Make sure to pull in the mode here as well in
 682                  * case the inode is released without being used.
 683                  * This ensures that xfs_inactive() will see that
 684                  * the inode is already free and not try to mess
 685                  * with the uninitialized part of it.
 686                  */
 687                 VFS_I(ip)->i_mode = 0;
 688         }
 689
 690         ASSERT(ip->i_d.di_version >= 2);
 691         ip->i_delayed_blks = 0;
 692
 693         /*
 694          * Mark the buffer containing the inode as something to keep
 695          * around for a while.  This helps to keep recently accessed
 696          * meta-data in-core longer.
 697          */
 698         xfs_buf_set_ref(bp, XFS_INO_REF);
 699
 700         /*
 701          * Use xfs_trans_brelse() to release the buffer containing the on-disk
 702          * inode, because it was acquired with xfs_trans_read_buf() in
 703          * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
 704          * brelse().  If we're within a transaction, then xfs_trans_brelse()
 705          * will only release the buffer if it is not dirty within the
 706          * transaction.  It will be OK to release the buffer in this case,
 707          * because inodes on disk are never destroyed and we will be locking the
 708          * new in-core inode before putting it in the cache where other
 709          * processes can find it.  Thus we don't have to worry about the inode
 710          * being changed just because we released the buffer.
 711          */
 712  out_brelse:
 713         xfs_trans_brelse(tp, bp);
 714         return error;
 715 }
 716
 717 /*
 718  * Validate di_extsize hint.
 719  *
 720  * The rules are documented at xfs_ioctl_setattr_check_extsize().
 721  * These functions must be kept in sync with each other.
 722  */
 723 xfs_failaddr_t
 724 xfs_inode_validate_extsize(
 725         struct xfs_mount                *mp,
 726         uint32_t                        extsize,
 727         uint16_t                        mode,
 728         uint16_t                        flags)
 729 {
 730         bool                            rt_flag;
 731         bool                            hint_flag;
 732         bool                            inherit_flag;
 733         uint32_t                        extsize_bytes;
 734         uint32_t                        blocksize_bytes;
 735
 736         rt_flag = (flags & XFS_DIFLAG_REALTIME);
 737         hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
 738         inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
 739         extsize_bytes = XFS_FSB_TO_B(mp, extsize);
 740
 741         if (rt_flag)
 742                 blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
 743         else
 744                 blocksize_bytes = mp->m_sb.sb_blocksize;
 745
 746         if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
 747                 return __this_address;
 748
 749         if (hint_flag && !S_ISREG(mode))
 750                 return __this_address;
 751
 752         if (inherit_flag && !S_ISDIR(mode))
 753                 return __this_address;
 754
 755         if ((hint_flag || inherit_flag) && extsize == 0)
 756                 return __this_address;
 757
 758         /* free inodes get flags set to zero but extsize remains */
 759         if (mode && !(hint_flag || inherit_flag) && extsize != 0)
 760                 return __this_address;
 761
 762         if (extsize_bytes % blocksize_bytes)
 763                 return __this_address;
 764
 765         if (extsize > MAXEXTLEN)
 766                 return __this_address;
 767
 768         if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
 769                 return __this_address;
 770
 771         return NULL;
 772 }
 773
 774 /*
 775  * Validate di_cowextsize hint.
 776  *
 777  * The rules are documented at xfs_ioctl_setattr_check_cowextsize().
 778  * These functions must be kept in sync with each other.
 779  */
 780 xfs_failaddr_t
 781 xfs_inode_validate_cowextsize(
 782         struct xfs_mount                *mp,
 783         uint32_t                        cowextsize,
 784         uint16_t                        mode,
 785         uint16_t                        flags,
 786         uint64_t                        flags2)
 787 {
 788         bool                            rt_flag;
 789         bool                            hint_flag;
 790         uint32_t                        cowextsize_bytes;
 791
 792         rt_flag = (flags & XFS_DIFLAG_REALTIME);
 793         hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
 794         cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
 795
 796         if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
 797                 return __this_address;
 798
 799         if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
 800                 return __this_address;
 801
 802         if (hint_flag && cowextsize == 0)
 803                 return __this_address;
 804
 805         /* free inodes get flags set to zero but cowextsize remains */
 806         if (mode && !hint_flag && cowextsize != 0)
 807                 return __this_address;
 808
 809         if (hint_flag && rt_flag)
 810                 return __this_address;
 811
 812         if (cowextsize_bytes % mp->m_sb.sb_blocksize)
 813                 return __this_address;
 814
 815         if (cowextsize > MAXEXTLEN)
 816                 return __this_address;
 817
 818         if (cowextsize > mp->m_sb.sb_agblocks / 2)
 819                 return __this_address;
 820
 821         return NULL;
 822 }