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_defer.h"
  14 #include "xfs_inode.h"
  15 #include "xfs_errortag.h"
  16 #include "xfs_cksum.h"
  17 #include "xfs_trans.h"
  18 #include "xfs_ialloc.h"
  19 #include "xfs_dir2.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_buf_offset(bp, i * mp->m_sb.sb_inodesize);
  39                 if (!dip->di_next_unlinked)  {
  40                         xfs_alert(mp,
  41         "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
  42                                 i, (long long)bp->b_bn);
  43                 }
  44         }
  45 }
  46 #endif
  47
  48 bool
  49 xfs_dinode_good_version(
  50         struct xfs_mount *mp,
  51         __u8            version)
  52 {
  53         if (xfs_sb_version_hascrc(&mp->m_sb))
  54                 return version == 3;
  55
  56         return version == 1 || version == 2;
  57 }
  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         xfs_agnumber_t  agno;
  81         int             i;
  82         int             ni;
  83
  84         /*
  85          * Validate the magic number and version of every inode in the buffer
  86          */
  87         agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp));
  88         ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
  89         for (i = 0; i < ni; i++) {
  90                 int             di_ok;
  91                 xfs_dinode_t    *dip;
  92                 xfs_agino_t     unlinked_ino;
  93
  94                 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
  95                 unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
  96                 di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
  97                         xfs_dinode_good_version(mp, dip->di_version) &&
  98                         (unlinked_ino == NULLAGINO ||
  99                          xfs_verify_agino(mp, agno, unlinked_ino));
 100                 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
 101                                                 XFS_ERRTAG_ITOBP_INOTOBP))) {
 102                         if (readahead) {
 103                                 bp->b_flags &= ~XBF_DONE;
 104                                 xfs_buf_ioerror(bp, -EIO);
 105                                 return;
 106                         }
 107
 108 #ifdef DEBUG
 109                         xfs_alert(mp,
 110                                 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
 111                                 (unsigned long long)bp->b_bn, i,
 112                                 be16_to_cpu(dip->di_magic));
 113 #endif
 114                         xfs_buf_verifier_error(bp, -EFSCORRUPTED,
 115                                         __func__, dip, sizeof(*dip),
 116                                         NULL);
 117                         return;
 118                 }
 119         }
 120 }
 121
 122
 123 static void
 124 xfs_inode_buf_read_verify(
 125         struct xfs_buf  *bp)
 126 {
 127         xfs_inode_buf_verify(bp, false);
 128 }
 129
 130 static void
 131 xfs_inode_buf_readahead_verify(
 132         struct xfs_buf  *bp)
 133 {
 134         xfs_inode_buf_verify(bp, true);
 135 }
 136
 137 static void
 138 xfs_inode_buf_write_verify(
 139         struct xfs_buf  *bp)
 140 {
 141         xfs_inode_buf_verify(bp, false);
 142 }
 143
 144 const struct xfs_buf_ops xfs_inode_buf_ops = {
 145         .name = "xfs_inode",
 146         .verify_read = xfs_inode_buf_read_verify,
 147         .verify_write = xfs_inode_buf_write_verify,
 148 };
 149
 150 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
 151         .name = "xxfs_inode_ra",
 152         .verify_read = xfs_inode_buf_readahead_verify,
 153         .verify_write = xfs_inode_buf_write_verify,
 154 };
 155
 156
 157 /*
 158  * This routine is called to map an inode to the buffer containing the on-disk
 159  * version of the inode.  It returns a pointer to the buffer containing the
 160  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
 161  * pointer to the on-disk inode within that buffer.
 162  *
 163  * If a non-zero error is returned, then the contents of bpp and dipp are
 164  * undefined.
 165  */
 166 int
 167 xfs_imap_to_bp(
 168         struct xfs_mount        *mp,
 169         struct xfs_trans        *tp,
 170         struct xfs_imap         *imap,
 171         struct xfs_dinode       **dipp,
 172         struct xfs_buf          **bpp,
 173         uint                    buf_flags,
 174         uint                    iget_flags)
 175 {
 176         struct xfs_buf          *bp;
 177         int                     error;
 178
 179         buf_flags |= XBF_UNMAPPED;
 180         error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
 181                                    (int)imap->im_len, buf_flags, &bp,
 182                                    &xfs_inode_buf_ops);
 183         if (error) {
 184                 if (error == -EAGAIN) {
 185                         ASSERT(buf_flags & XBF_TRYLOCK);
 186                         return error;
 187                 }
 188                 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
 189                         __func__, error);
 190                 return error;
 191         }
 192
 193         *bpp = bp;
 194         *dipp = xfs_buf_offset(bp, imap->im_boffset);
 195         return 0;
 196 }
 197
 198 void
 199 xfs_inode_from_disk(
 200         struct xfs_inode        *ip,
 201         struct xfs_dinode       *from)
 202 {
 203         struct xfs_icdinode     *to = &ip->i_d;
 204         struct inode            *inode = VFS_I(ip);
 205
 206
 207         /*
 208          * Convert v1 inodes immediately to v2 inode format as this is the
 209          * minimum inode version format we support in the rest of the code.
 210          */
 211         to->di_version = from->di_version;
 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         inode->i_generation = be32_to_cpu(from->di_gen);
 241         inode->i_mode = be16_to_cpu(from->di_mode);
 242
 243         to->di_size = be64_to_cpu(from->di_size);
 244         to->di_nblocks = be64_to_cpu(from->di_nblocks);
 245         to->di_extsize = be32_to_cpu(from->di_extsize);
 246         to->di_nextents = be32_to_cpu(from->di_nextents);
 247         to->di_anextents = be16_to_cpu(from->di_anextents);
 248         to->di_forkoff = from->di_forkoff;
 249         to->di_aformat  = from->di_aformat;
 250         to->di_dmevmask = be32_to_cpu(from->di_dmevmask);
 251         to->di_dmstate  = be16_to_cpu(from->di_dmstate);
 252         to->di_flags    = be16_to_cpu(from->di_flags);
 253
 254         if (to->di_version == 3) {
 255                 inode_set_iversion_queried(inode,
 256                                            be64_to_cpu(from->di_changecount));
 257                 to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
 258                 to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
 259                 to->di_flags2 = be64_to_cpu(from->di_flags2);
 260                 to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
 261         }
 262 }
 263
 264 void
 265 xfs_inode_to_disk(
 266         struct xfs_inode        *ip,
 267         struct xfs_dinode       *to,
 268         xfs_lsn_t               lsn)
 269 {
 270         struct xfs_icdinode     *from = &ip->i_d;
 271         struct inode            *inode = VFS_I(ip);
 272
 273         to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
 274         to->di_onlink = 0;
 275
 276         to->di_version = from->di_version;
 277         to->di_format = from->di_format;
 278         to->di_uid = cpu_to_be32(from->di_uid);
 279         to->di_gid = cpu_to_be32(from->di_gid);
 280         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 281         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 282
 283         memset(to->di_pad, 0, sizeof(to->di_pad));
 284         to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
 285         to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
 286         to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
 287         to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
 288         to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
 289         to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
 290         to->di_nlink = cpu_to_be32(inode->i_nlink);
 291         to->di_gen = cpu_to_be32(inode->i_generation);
 292         to->di_mode = cpu_to_be16(inode->i_mode);
 293
 294         to->di_size = cpu_to_be64(from->di_size);
 295         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 296         to->di_extsize = cpu_to_be32(from->di_extsize);
 297         to->di_nextents = cpu_to_be32(from->di_nextents);
 298         to->di_anextents = cpu_to_be16(from->di_anextents);
 299         to->di_forkoff = from->di_forkoff;
 300         to->di_aformat = from->di_aformat;
 301         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 302         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 303         to->di_flags = cpu_to_be16(from->di_flags);
 304
 305         if (from->di_version == 3) {
 306                 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
 307                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 308                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 309                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 310                 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 311                 to->di_ino = cpu_to_be64(ip->i_ino);
 312                 to->di_lsn = cpu_to_be64(lsn);
 313                 memset(to->di_pad2, 0, sizeof(to->di_pad2));
 314                 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
 315                 to->di_flushiter = 0;
 316         } else {
 317                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 318         }
 319 }
 320
 321 void
 322 xfs_log_dinode_to_disk(
 323         struct xfs_log_dinode   *from,
 324         struct xfs_dinode       *to)
 325 {
 326         to->di_magic = cpu_to_be16(from->di_magic);
 327         to->di_mode = cpu_to_be16(from->di_mode);
 328         to->di_version = from->di_version;
 329         to->di_format = from->di_format;
 330         to->di_onlink = 0;
 331         to->di_uid = cpu_to_be32(from->di_uid);
 332         to->di_gid = cpu_to_be32(from->di_gid);
 333         to->di_nlink = cpu_to_be32(from->di_nlink);
 334         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 335         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 336         memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
 337
 338         to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
 339         to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
 340         to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
 341         to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
 342         to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
 343         to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
 344
 345         to->di_size = cpu_to_be64(from->di_size);
 346         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 347         to->di_extsize = cpu_to_be32(from->di_extsize);
 348         to->di_nextents = cpu_to_be32(from->di_nextents);
 349         to->di_anextents = cpu_to_be16(from->di_anextents);
 350         to->di_forkoff = from->di_forkoff;
 351         to->di_aformat = from->di_aformat;
 352         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 353         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 354         to->di_flags = cpu_to_be16(from->di_flags);
 355         to->di_gen = cpu_to_be32(from->di_gen);
 356
 357         if (from->di_version == 3) {
 358                 to->di_changecount = cpu_to_be64(from->di_changecount);
 359                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 360                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 361                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 362                 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 363                 to->di_ino = cpu_to_be64(from->di_ino);
 364                 to->di_lsn = cpu_to_be64(from->di_lsn);
 365                 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 366                 uuid_copy(&to->di_uuid, &from->di_uuid);
 367                 to->di_flushiter = 0;
 368         } else {
 369                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 370         }
 371 }
 372
 373 static xfs_failaddr_t
 374 xfs_dinode_verify_fork(
 375         struct xfs_dinode       *dip,
 376         struct xfs_mount        *mp,
 377         int                     whichfork)
 378 {
 379         uint32_t                di_nextents = XFS_DFORK_NEXTENTS(dip, whichfork);
 380
 381         switch (XFS_DFORK_FORMAT(dip, whichfork)) {
 382         case XFS_DINODE_FMT_LOCAL:
 383                 /*
 384                  * no local regular files yet
 385                  */
 386                 if (whichfork == XFS_DATA_FORK) {
 387                         if (S_ISREG(be16_to_cpu(dip->di_mode)))
 388                                 return __this_address;
 389                         if (be64_to_cpu(dip->di_size) >
 390                                         XFS_DFORK_SIZE(dip, mp, whichfork))
 391                                 return __this_address;
 392                 }
 393                 if (di_nextents)
 394                         return __this_address;
 395                 break;
 396         case XFS_DINODE_FMT_EXTENTS:
 397                 if (di_nextents > XFS_DFORK_MAXEXT(dip, mp, whichfork))
 398                         return __this_address;
 399                 break;
 400         case XFS_DINODE_FMT_BTREE:
 401                 if (whichfork == XFS_ATTR_FORK) {
 402                         if (di_nextents > MAXAEXTNUM)
 403                                 return __this_address;
 404                 } else if (di_nextents > MAXEXTNUM) {
 405                         return __this_address;
 406                 }
 407                 break;
 408         default:
 409                 return __this_address;
 410         }
 411         return NULL;
 412 }
 413
 414 static xfs_failaddr_t
 415 xfs_dinode_verify_forkoff(
 416         struct xfs_dinode       *dip,
 417         struct xfs_mount        *mp)
 418 {
 419         if (!XFS_DFORK_Q(dip))
 420                 return NULL;
 421
 422         switch (dip->di_format)  {
 423         case XFS_DINODE_FMT_DEV:
 424                 if (dip->di_forkoff != (roundup(sizeof(xfs_dev_t), 8) >> 3))
 425                         return __this_address;
 426                 break;
 427         case XFS_DINODE_FMT_LOCAL:      /* fall through ... */
 428         case XFS_DINODE_FMT_EXTENTS:    /* fall through ... */
 429         case XFS_DINODE_FMT_BTREE:
 430                 if (dip->di_forkoff >= (XFS_LITINO(mp, dip->di_version) >> 3))
 431                         return __this_address;
 432                 break;
 433         default:
 434                 return __this_address;
 435         }
 436         return NULL;
 437 }
 438
 439 xfs_failaddr_t
 440 xfs_dinode_verify(
 441         struct xfs_mount        *mp,
 442         xfs_ino_t               ino,
 443         struct xfs_dinode       *dip)
 444 {
 445         xfs_failaddr_t          fa;
 446         uint16_t                mode;
 447         uint16_t                flags;
 448         uint64_t                flags2;
 449         uint64_t                di_size;
 450
 451         if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
 452                 return __this_address;
 453
 454         /* Verify v3 integrity information first */
 455         if (dip->di_version >= 3) {
 456                 if (!xfs_sb_version_hascrc(&mp->m_sb))
 457                         return __this_address;
 458                 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
 459                                       XFS_DINODE_CRC_OFF))
 460                         return __this_address;
 461                 if (be64_to_cpu(dip->di_ino) != ino)
 462                         return __this_address;
 463                 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
 464                         return __this_address;
 465         }
 466
 467         /* don't allow invalid i_size */
 468         di_size = be64_to_cpu(dip->di_size);
 469         if (di_size & (1ULL << 63))
 470                 return __this_address;
 471
 472         mode = be16_to_cpu(dip->di_mode);
 473         if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
 474                 return __this_address;
 475
 476         /* No zero-length symlinks/dirs. */
 477         if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
 478                 return __this_address;
 479
 480         /* Fork checks carried over from xfs_iformat_fork */
 481         if (mode &&
 482             be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
 483                         be64_to_cpu(dip->di_nblocks))
 484                 return __this_address;
 485
 486         if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
 487                 return __this_address;
 488
 489         flags = be16_to_cpu(dip->di_flags);
 490
 491         if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
 492                 return __this_address;
 493
 494         /* check for illegal values of forkoff */
 495         fa = xfs_dinode_verify_forkoff(dip, mp);
 496         if (fa)
 497                 return fa;
 498
 499         /* Do we have appropriate data fork formats for the mode? */
 500         switch (mode & S_IFMT) {
 501         case S_IFIFO:
 502         case S_IFCHR:
 503         case S_IFBLK:
 504         case S_IFSOCK:
 505                 if (dip->di_format != XFS_DINODE_FMT_DEV)
 506                         return __this_address;
 507                 break;
 508         case S_IFREG:
 509         case S_IFLNK:
 510         case S_IFDIR:
 511                 fa = xfs_dinode_verify_fork(dip, mp, XFS_DATA_FORK);
 512                 if (fa)
 513                         return fa;
 514                 break;
 515         case 0:
 516                 /* Uninitialized inode ok. */
 517                 break;
 518         default:
 519                 return __this_address;
 520         }
 521
 522         if (XFS_DFORK_Q(dip)) {
 523                 fa = xfs_dinode_verify_fork(dip, mp, XFS_ATTR_FORK);
 524                 if (fa)
 525                         return fa;
 526         } else {
 527                 /*
 528                  * If there is no fork offset, this may be a freshly-made inode
 529                  * in a new disk cluster, in which case di_aformat is zeroed.
 530                  * Otherwise, such an inode must be in EXTENTS format; this goes
 531                  * for freed inodes as well.
 532                  */
 533                 switch (dip->di_aformat) {
 534                 case 0:
 535                 case XFS_DINODE_FMT_EXTENTS:
 536                         break;
 537                 default:
 538                         return __this_address;
 539                 }
 540                 if (dip->di_anextents)
 541                         return __this_address;
 542         }
 543
 544         /* extent size hint validation */
 545         fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
 546                         mode, flags);
 547         if (fa)
 548                 return fa;
 549
 550         /* only version 3 or greater inodes are extensively verified here */
 551         if (dip->di_version < 3)
 552                 return NULL;
 553
 554         flags2 = be64_to_cpu(dip->di_flags2);
 555
 556         /* don't allow reflink/cowextsize if we don't have reflink */
 557         if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
 558              !xfs_sb_version_hasreflink(&mp->m_sb))
 559                 return __this_address;
 560
 561         /* only regular files get reflink */
 562         if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
 563                 return __this_address;
 564
 565         /* don't let reflink and realtime mix */
 566         if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
 567                 return __this_address;
 568
 569         /* don't let reflink and dax mix */
 570         if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
 571                 return __this_address;
 572
 573         /* COW extent size hint validation */
 574         fa = xfs_inode_validate_cowextsize(mp, be32_to_cpu(dip->di_cowextsize),
 575                         mode, flags, flags2);
 576         if (fa)
 577                 return fa;
 578
 579         return NULL;
 580 }
 581
 582 void
 583 xfs_dinode_calc_crc(
 584         struct xfs_mount        *mp,
 585         struct xfs_dinode       *dip)
 586 {
 587         uint32_t                crc;
 588
 589         if (dip->di_version < 3)
 590                 return;
 591
 592         ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
 593         crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
 594                               XFS_DINODE_CRC_OFF);
 595         dip->di_crc = xfs_end_cksum(crc);
 596 }
 597
 598 /*
 599  * Read the disk inode attributes into the in-core inode structure.
 600  *
 601  * For version 5 superblocks, if we are initialising a new inode and we are not
 602  * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
 603  * inode core with a random generation number. If we are keeping inodes around,
 604  * we need to read the inode cluster to get the existing generation number off
 605  * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
 606  * format) then log recovery is dependent on the di_flushiter field being
 607  * initialised from the current on-disk value and hence we must also read the
 608  * inode off disk.
 609  */
 610 int
 611 xfs_iread(
 612         xfs_mount_t     *mp,
 613         xfs_trans_t     *tp,
 614         xfs_inode_t     *ip,
 615         uint            iget_flags)
 616 {
 617         xfs_buf_t       *bp;
 618         xfs_dinode_t    *dip;
 619         xfs_failaddr_t  fa;
 620         int             error;
 621
 622         /*
 623          * Fill in the location information in the in-core inode.
 624          */
 625         error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
 626         if (error)
 627                 return error;
 628
 629         /* shortcut IO on inode allocation if possible */
 630         if ((iget_flags & XFS_IGET_CREATE) &&
 631             xfs_sb_version_hascrc(&mp->m_sb) &&
 632             !(mp->m_flags & XFS_MOUNT_IKEEP)) {
 633                 /* initialise the on-disk inode core */
 634                 memset(&ip->i_d, 0, sizeof(ip->i_d));
 635                 VFS_I(ip)->i_generation = prandom_u32();
 636                 ip->i_d.di_version = 3;
 637                 return 0;
 638         }
 639
 640         /*
 641          * Get pointers to the on-disk inode and the buffer containing it.
 642          */
 643         error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
 644         if (error)
 645                 return error;
 646
 647         /* even unallocated inodes are verified */
 648         fa = xfs_dinode_verify(mp, ip->i_ino, dip);
 649         if (fa) {
 650                 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
 651                                 sizeof(*dip), fa);
 652                 error = -EFSCORRUPTED;
 653                 goto out_brelse;
 654         }
 655
 656         /*
 657          * If the on-disk inode is already linked to a directory
 658          * entry, copy all of the inode into the in-core inode.
 659          * xfs_iformat_fork() handles copying in the inode format
 660          * specific information.
 661          * Otherwise, just get the truly permanent information.
 662          */
 663         if (dip->di_mode) {
 664                 xfs_inode_from_disk(ip, dip);
 665                 error = xfs_iformat_fork(ip, dip);
 666                 if (error)  {
 667 #ifdef DEBUG
 668                         xfs_alert(mp, "%s: xfs_iformat() returned error %d",
 669                                 __func__, error);
 670 #endif /* DEBUG */
 671                         goto out_brelse;
 672                 }
 673         } else {
 674                 /*
 675                  * Partial initialisation of the in-core inode. Just the bits
 676                  * that xfs_ialloc won't overwrite or relies on being correct.
 677                  */
 678                 ip->i_d.di_version = dip->di_version;
 679                 VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
 680                 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
 681
 682                 /*
 683                  * Make sure to pull in the mode here as well in
 684                  * case the inode is released without being used.
 685                  * This ensures that xfs_inactive() will see that
 686                  * the inode is already free and not try to mess
 687                  * with the uninitialized part of it.
 688                  */
 689                 VFS_I(ip)->i_mode = 0;
 690         }
 691
 692         ASSERT(ip->i_d.di_version >= 2);
 693         ip->i_delayed_blks = 0;
 694
 695         /*
 696          * Mark the buffer containing the inode as something to keep
 697          * around for a while.  This helps to keep recently accessed
 698          * meta-data in-core longer.
 699          */
 700         xfs_buf_set_ref(bp, XFS_INO_REF);
 701
 702         /*
 703          * Use xfs_trans_brelse() to release the buffer containing the on-disk
 704          * inode, because it was acquired with xfs_trans_read_buf() in
 705          * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
 706          * brelse().  If we're within a transaction, then xfs_trans_brelse()
 707          * will only release the buffer if it is not dirty within the
 708          * transaction.  It will be OK to release the buffer in this case,
 709          * because inodes on disk are never destroyed and we will be locking the
 710          * new in-core inode before putting it in the cache where other
 711          * processes can find it.  Thus we don't have to worry about the inode
 712          * being changed just because we released the buffer.
 713          */
 714  out_brelse:
 715         xfs_trans_brelse(tp, bp);
 716         return error;
 717 }
 718
 719 /*
 720  * Validate di_extsize hint.
 721  *
 722  * The rules are documented at xfs_ioctl_setattr_check_extsize().
 723  * These functions must be kept in sync with each other.
 724  */
 725 xfs_failaddr_t
 726 xfs_inode_validate_extsize(
 727         struct xfs_mount                *mp,
 728         uint32_t                        extsize,
 729         uint16_t                        mode,
 730         uint16_t                        flags)
 731 {
 732         bool                            rt_flag;
 733         bool                            hint_flag;
 734         bool                            inherit_flag;
 735         uint32_t                        extsize_bytes;
 736         uint32_t                        blocksize_bytes;
 737
 738         rt_flag = (flags & XFS_DIFLAG_REALTIME);
 739         hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
 740         inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
 741         extsize_bytes = XFS_FSB_TO_B(mp, extsize);
 742
 743         if (rt_flag)
 744                 blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
 745         else
 746                 blocksize_bytes = mp->m_sb.sb_blocksize;
 747
 748         if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
 749                 return __this_address;
 750
 751         if (hint_flag && !S_ISREG(mode))
 752                 return __this_address;
 753
 754         if (inherit_flag && !S_ISDIR(mode))
 755                 return __this_address;
 756
 757         if ((hint_flag || inherit_flag) && extsize == 0)
 758                 return __this_address;
 759
 760         /* free inodes get flags set to zero but extsize remains */
 761         if (mode && !(hint_flag || inherit_flag) && extsize != 0)
 762                 return __this_address;
 763
 764         if (extsize_bytes % blocksize_bytes)
 765                 return __this_address;
 766
 767         if (extsize > MAXEXTLEN)
 768                 return __this_address;
 769
 770         if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
 771                 return __this_address;
 772
 773         return NULL;
 774 }
 775
 776 /*
 777  * Validate di_cowextsize hint.
 778  *
 779  * The rules are documented at xfs_ioctl_setattr_check_cowextsize().
 780  * These functions must be kept in sync with each other.
 781  */
 782 xfs_failaddr_t
 783 xfs_inode_validate_cowextsize(
 784         struct xfs_mount                *mp,
 785         uint32_t                        cowextsize,
 786         uint16_t                        mode,
 787         uint16_t                        flags,
 788         uint64_t                        flags2)
 789 {
 790         bool                            rt_flag;
 791         bool                            hint_flag;
 792         uint32_t                        cowextsize_bytes;
 793
 794         rt_flag = (flags & XFS_DIFLAG_REALTIME);
 795         hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
 796         cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
 797
 798         if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
 799                 return __this_address;
 800
 801         if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
 802                 return __this_address;
 803
 804         if (hint_flag && cowextsize == 0)
 805                 return __this_address;
 806
 807         /* free inodes get flags set to zero but cowextsize remains */
 808         if (mode && !hint_flag && cowextsize != 0)
 809                 return __this_address;
 810
 811         if (hint_flag && rt_flag)
 812                 return __this_address;
 813
 814         if (cowextsize_bytes % mp->m_sb.sb_blocksize)
 815                 return __this_address;
 816
 817         if (cowextsize > MAXEXTLEN)
 818                 return __this_address;
 819
 820         if (cowextsize > mp->m_sb.sb_agblocks / 2)
 821                 return __this_address;
 822
 823         return NULL;
 824 }