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 "libxfs_priv.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_defer.h"
  26 #include "xfs_inode.h"
  27 #include "xfs_errortag.h"
  28 #include "xfs_cksum.h"
  29 #include "xfs_trans.h"
  30 #include "xfs_ialloc.h"
  31 #include "xfs_dir2.h"
  32
  33 /*
  34  * Check that none of the inode's in the buffer have a next
  35  * unlinked field of 0.
  36  */
  37 #if defined(DEBUG)
  38 void
  39 xfs_inobp_check(
  40         xfs_mount_t     *mp,
  41         xfs_buf_t       *bp)
  42 {
  43         int             i;
  44         int             j;
  45         xfs_dinode_t    *dip;
  46
  47         j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog;
  48
  49         for (i = 0; i < j; i++) {
  50                 dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize);
  51                 if (!dip->di_next_unlinked)  {
  52                         xfs_alert(mp,
  53         "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.",
  54                                 i, (long long)bp->b_bn);
  55                 }
  56         }
  57 }
  58 #endif
  59
  60 bool
  61 xfs_dinode_good_version(
  62         struct xfs_mount *mp,
  63         __u8            version)
  64 {
  65         if (xfs_sb_version_hascrc(&mp->m_sb))
  66                 return version == 3;
  67
  68         return version == 1 || version == 2;
  69 }
  70
  71 /*
  72  * If we are doing readahead on an inode buffer, we might be in log recovery
  73  * reading an inode allocation buffer that hasn't yet been replayed, and hence
  74  * has not had the inode cores stamped into it. Hence for readahead, the buffer
  75  * may be potentially invalid.
  76  *
  77  * If the readahead buffer is invalid, we need to mark it with an error and
  78  * clear the DONE status of the buffer so that a followup read will re-read it
  79  * from disk. We don't report the error otherwise to avoid warnings during log
  80  * recovery and we don't get unnecssary panics on debug kernels. We use EIO here
  81  * because all we want to do is say readahead failed; there is no-one to report
  82  * the error to, so this will distinguish it from a non-ra verifier failure.
  83  * Changes to this readahead error behavour also need to be reflected in
  84  * xfs_dquot_buf_readahead_verify().
  85  */
  86 static void
  87 xfs_inode_buf_verify(
  88         struct xfs_buf  *bp,
  89         bool            readahead)
  90 {
  91         struct xfs_mount *mp = bp->b_target->bt_mount;
  92         xfs_agnumber_t  agno;
  93         int             i;
  94         int             ni;
  95
  96         /*
  97          * Validate the magic number and version of every inode in the buffer
  98          */
  99         agno = xfs_daddr_to_agno(mp, XFS_BUF_ADDR(bp));
 100         ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
 101         for (i = 0; i < ni; i++) {
 102                 int             di_ok;
 103                 xfs_dinode_t    *dip;
 104                 xfs_agino_t     unlinked_ino;
 105
 106                 dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog));
 107                 unlinked_ino = be32_to_cpu(dip->di_next_unlinked);
 108                 di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
 109                         xfs_dinode_good_version(mp, dip->di_version) &&
 110                         (unlinked_ino == NULLAGINO ||
 111                          xfs_verify_agino(mp, agno, unlinked_ino));
 112                 if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
 113                                                 XFS_ERRTAG_ITOBP_INOTOBP))) {
 114                         if (readahead) {
 115                                 bp->b_flags &= ~XBF_DONE;
 116                                 xfs_buf_ioerror(bp, -EIO);
 117                                 return;
 118                         }
 119
 120 #ifdef DEBUG
 121                         xfs_alert(mp,
 122                                 "bad inode magic/vsn daddr %lld #%d (magic=%x)",
 123                                 (unsigned long long)bp->b_bn, i,
 124                                 be16_to_cpu(dip->di_magic));
 125 #endif
 126                         xfs_buf_verifier_error(bp, -EFSCORRUPTED,
 127                                         __func__, dip, sizeof(*dip),
 128                                         NULL);
 129                         return;
 130                 }
 131         }
 132 }
 133
 134
 135 static void
 136 xfs_inode_buf_read_verify(
 137         struct xfs_buf  *bp)
 138 {
 139         xfs_inode_buf_verify(bp, false);
 140 }
 141
 142 static void
 143 xfs_inode_buf_readahead_verify(
 144         struct xfs_buf  *bp)
 145 {
 146         xfs_inode_buf_verify(bp, true);
 147 }
 148
 149 static void
 150 xfs_inode_buf_write_verify(
 151         struct xfs_buf  *bp)
 152 {
 153         xfs_inode_buf_verify(bp, false);
 154 }
 155
 156 const struct xfs_buf_ops xfs_inode_buf_ops = {
 157         .name = "xfs_inode",
 158         .verify_read = xfs_inode_buf_read_verify,
 159         .verify_write = xfs_inode_buf_write_verify,
 160 };
 161
 162 const struct xfs_buf_ops xfs_inode_buf_ra_ops = {
 163         .name = "xxfs_inode_ra",
 164         .verify_read = xfs_inode_buf_readahead_verify,
 165         .verify_write = xfs_inode_buf_write_verify,
 166 };
 167
 168
 169 /*
 170  * This routine is called to map an inode to the buffer containing the on-disk
 171  * version of the inode.  It returns a pointer to the buffer containing the
 172  * on-disk inode in the bpp parameter, and in the dipp parameter it returns a
 173  * pointer to the on-disk inode within that buffer.
 174  *
 175  * If a non-zero error is returned, then the contents of bpp and dipp are
 176  * undefined.
 177  */
 178 int
 179 xfs_imap_to_bp(
 180         struct xfs_mount        *mp,
 181         struct xfs_trans        *tp,
 182         struct xfs_imap         *imap,
 183         struct xfs_dinode       **dipp,
 184         struct xfs_buf          **bpp,
 185         uint                    buf_flags,
 186         uint                    iget_flags)
 187 {
 188         struct xfs_buf          *bp;
 189         int                     error;
 190
 191         buf_flags |= XBF_UNMAPPED;
 192         error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
 193                                    (int)imap->im_len, buf_flags, &bp,
 194                                    &xfs_inode_buf_ops);
 195         if (error) {
 196                 if (error == -EAGAIN) {
 197                         ASSERT(buf_flags & XBF_TRYLOCK);
 198                         return error;
 199                 }
 200
 201                 if (error == -EFSCORRUPTED &&
 202                     (iget_flags & XFS_IGET_UNTRUSTED))
 203                         return -EINVAL;
 204
 205                 xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
 206                         __func__, error);
 207                 return error;
 208         }
 209
 210         *bpp = bp;
 211         *dipp = xfs_buf_offset(bp, imap->im_boffset);
 212         return 0;
 213 }
 214
 215 void
 216 xfs_inode_from_disk(
 217         struct xfs_inode        *ip,
 218         struct xfs_dinode       *from)
 219 {
 220         struct xfs_icdinode     *to = &ip->i_d;
 221         struct inode            *inode = VFS_I(ip);
 222
 223
 224         /*
 225          * Convert v1 inodes immediately to v2 inode format as this is the
 226          * minimum inode version format we support in the rest of the code.
 227          */
 228         to->di_version = from->di_version;
 229         if (to->di_version == 1) {
 230                 set_nlink(inode, be16_to_cpu(from->di_onlink));
 231                 to->di_projid_lo = 0;
 232                 to->di_projid_hi = 0;
 233                 to->di_version = 2;
 234         } else {
 235                 set_nlink(inode, be32_to_cpu(from->di_nlink));
 236                 to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
 237                 to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
 238         }
 239
 240         to->di_format = from->di_format;
 241         to->di_uid = be32_to_cpu(from->di_uid);
 242         to->di_gid = be32_to_cpu(from->di_gid);
 243         to->di_flushiter = be16_to_cpu(from->di_flushiter);
 244
 245         /*
 246          * Time is signed, so need to convert to signed 32 bit before
 247          * storing in inode timestamp which may be 64 bit. Otherwise
 248          * a time before epoch is converted to a time long after epoch
 249          * on 64 bit systems.
 250          */
 251         inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec);
 252         inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec);
 253         inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec);
 254         inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec);
 255         inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec);
 256         inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec);
 257         inode->i_generation = be32_to_cpu(from->di_gen);
 258         inode->i_mode = be16_to_cpu(from->di_mode);
 259
 260         to->di_size = be64_to_cpu(from->di_size);
 261         to->di_nblocks = be64_to_cpu(from->di_nblocks);
 262         to->di_extsize = be32_to_cpu(from->di_extsize);
 263         to->di_nextents = be32_to_cpu(from->di_nextents);
 264         to->di_anextents = be16_to_cpu(from->di_anextents);
 265         to->di_forkoff = from->di_forkoff;
 266         to->di_aformat  = from->di_aformat;
 267         to->di_dmevmask = be32_to_cpu(from->di_dmevmask);
 268         to->di_dmstate  = be16_to_cpu(from->di_dmstate);
 269         to->di_flags    = be16_to_cpu(from->di_flags);
 270
 271         if (to->di_version == 3) {
 272                 inode_set_iversion_queried(inode,
 273                                            be64_to_cpu(from->di_changecount));
 274                 to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
 275                 to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
 276                 to->di_flags2 = be64_to_cpu(from->di_flags2);
 277                 to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
 278         }
 279 }
 280
 281 void
 282 xfs_inode_to_disk(
 283         struct xfs_inode        *ip,
 284         struct xfs_dinode       *to,
 285         xfs_lsn_t               lsn)
 286 {
 287         struct xfs_icdinode     *from = &ip->i_d;
 288         struct inode            *inode = VFS_I(ip);
 289
 290         to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
 291         to->di_onlink = 0;
 292
 293         to->di_version = from->di_version;
 294         to->di_format = from->di_format;
 295         to->di_uid = cpu_to_be32(from->di_uid);
 296         to->di_gid = cpu_to_be32(from->di_gid);
 297         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 298         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 299
 300         memset(to->di_pad, 0, sizeof(to->di_pad));
 301         to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec);
 302         to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec);
 303         to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec);
 304         to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec);
 305         to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec);
 306         to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec);
 307         to->di_nlink = cpu_to_be32(inode->i_nlink);
 308         to->di_gen = cpu_to_be32(inode->i_generation);
 309         to->di_mode = cpu_to_be16(inode->i_mode);
 310
 311         to->di_size = cpu_to_be64(from->di_size);
 312         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 313         to->di_extsize = cpu_to_be32(from->di_extsize);
 314         to->di_nextents = cpu_to_be32(from->di_nextents);
 315         to->di_anextents = cpu_to_be16(from->di_anextents);
 316         to->di_forkoff = from->di_forkoff;
 317         to->di_aformat = from->di_aformat;
 318         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 319         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 320         to->di_flags = cpu_to_be16(from->di_flags);
 321
 322         if (from->di_version == 3) {
 323                 to->di_changecount = cpu_to_be64(inode_peek_iversion(inode));
 324                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 325                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 326                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 327                 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 328                 to->di_ino = cpu_to_be64(ip->i_ino);
 329                 to->di_lsn = cpu_to_be64(lsn);
 330                 memset(to->di_pad2, 0, sizeof(to->di_pad2));
 331                 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
 332                 to->di_flushiter = 0;
 333         } else {
 334                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 335         }
 336 }
 337
 338 void
 339 xfs_log_dinode_to_disk(
 340         struct xfs_log_dinode   *from,
 341         struct xfs_dinode       *to)
 342 {
 343         to->di_magic = cpu_to_be16(from->di_magic);
 344         to->di_mode = cpu_to_be16(from->di_mode);
 345         to->di_version = from->di_version;
 346         to->di_format = from->di_format;
 347         to->di_onlink = 0;
 348         to->di_uid = cpu_to_be32(from->di_uid);
 349         to->di_gid = cpu_to_be32(from->di_gid);
 350         to->di_nlink = cpu_to_be32(from->di_nlink);
 351         to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
 352         to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
 353         memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
 354
 355         to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
 356         to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec);
 357         to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec);
 358         to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec);
 359         to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec);
 360         to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec);
 361
 362         to->di_size = cpu_to_be64(from->di_size);
 363         to->di_nblocks = cpu_to_be64(from->di_nblocks);
 364         to->di_extsize = cpu_to_be32(from->di_extsize);
 365         to->di_nextents = cpu_to_be32(from->di_nextents);
 366         to->di_anextents = cpu_to_be16(from->di_anextents);
 367         to->di_forkoff = from->di_forkoff;
 368         to->di_aformat = from->di_aformat;
 369         to->di_dmevmask = cpu_to_be32(from->di_dmevmask);
 370         to->di_dmstate = cpu_to_be16(from->di_dmstate);
 371         to->di_flags = cpu_to_be16(from->di_flags);
 372         to->di_gen = cpu_to_be32(from->di_gen);
 373
 374         if (from->di_version == 3) {
 375                 to->di_changecount = cpu_to_be64(from->di_changecount);
 376                 to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 377                 to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 378                 to->di_flags2 = cpu_to_be64(from->di_flags2);
 379                 to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 380                 to->di_ino = cpu_to_be64(from->di_ino);
 381                 to->di_lsn = cpu_to_be64(from->di_lsn);
 382                 memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
 383                 uuid_copy(&to->di_uuid, &from->di_uuid);
 384                 to->di_flushiter = 0;
 385         } else {
 386                 to->di_flushiter = cpu_to_be16(from->di_flushiter);
 387         }
 388 }
 389
 390 xfs_failaddr_t
 391 xfs_dinode_verify(
 392         struct xfs_mount        *mp,
 393         xfs_ino_t               ino,
 394         struct xfs_dinode       *dip)
 395 {
 396         xfs_failaddr_t          fa;
 397         uint16_t                mode;
 398         uint16_t                flags;
 399         uint64_t                flags2;
 400         uint64_t                di_size;
 401
 402         if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
 403                 return __this_address;
 404
 405         /* Verify v3 integrity information first */
 406         if (dip->di_version >= 3) {
 407                 if (!xfs_sb_version_hascrc(&mp->m_sb))
 408                         return __this_address;
 409                 if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize,
 410                                       XFS_DINODE_CRC_OFF))
 411                         return __this_address;
 412                 if (be64_to_cpu(dip->di_ino) != ino)
 413                         return __this_address;
 414                 if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
 415                         return __this_address;
 416         }
 417
 418         /* don't allow invalid i_size */
 419         di_size = be64_to_cpu(dip->di_size);
 420         if (di_size & (1ULL << 63))
 421                 return __this_address;
 422
 423         mode = be16_to_cpu(dip->di_mode);
 424         if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
 425                 return __this_address;
 426
 427         /* No zero-length symlinks/dirs. */
 428         if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
 429                 return __this_address;
 430
 431         /* Fork checks carried over from xfs_iformat_fork */
 432         if (mode &&
 433             be32_to_cpu(dip->di_nextents) + be16_to_cpu(dip->di_anextents) >
 434                         be64_to_cpu(dip->di_nblocks))
 435                 return __this_address;
 436
 437         if (mode && XFS_DFORK_BOFF(dip) > mp->m_sb.sb_inodesize)
 438                 return __this_address;
 439
 440         flags = be16_to_cpu(dip->di_flags);
 441
 442         if (mode && (flags & XFS_DIFLAG_REALTIME) && !mp->m_rtdev_targp)
 443                 return __this_address;
 444
 445         /* Do we have appropriate data fork formats for the mode? */
 446         switch (mode & S_IFMT) {
 447         case S_IFIFO:
 448         case S_IFCHR:
 449         case S_IFBLK:
 450         case S_IFSOCK:
 451                 if (dip->di_format != XFS_DINODE_FMT_DEV)
 452                         return __this_address;
 453                 break;
 454         case S_IFREG:
 455         case S_IFLNK:
 456         case S_IFDIR:
 457                 switch (dip->di_format) {
 458                 case XFS_DINODE_FMT_LOCAL:
 459                         /*
 460                          * no local regular files yet
 461                          */
 462                         if (S_ISREG(mode))
 463                                 return __this_address;
 464                         if (di_size > XFS_DFORK_DSIZE(dip, mp))
 465                                 return __this_address;
 466                         if (dip->di_nextents)
 467                                 return __this_address;
 468                         /* fall through */
 469                 case XFS_DINODE_FMT_EXTENTS:
 470                 case XFS_DINODE_FMT_BTREE:
 471                         break;
 472                 default:
 473                         return __this_address;
 474                 }
 475                 break;
 476         case 0:
 477                 /* Uninitialized inode ok. */
 478                 break;
 479         default:
 480                 return __this_address;
 481         }
 482
 483         if (XFS_DFORK_Q(dip)) {
 484                 switch (dip->di_aformat) {
 485                 case XFS_DINODE_FMT_LOCAL:
 486                         if (dip->di_anextents)
 487                                 return __this_address;
 488                 /* fall through */
 489                 case XFS_DINODE_FMT_EXTENTS:
 490                 case XFS_DINODE_FMT_BTREE:
 491                         break;
 492                 default:
 493                         return __this_address;
 494                 }
 495         } else {
 496                 /*
 497                  * If there is no fork offset, this may be a freshly-made inode
 498                  * in a new disk cluster, in which case di_aformat is zeroed.
 499                  * Otherwise, such an inode must be in EXTENTS format; this goes
 500                  * for freed inodes as well.
 501                  */
 502                 switch (dip->di_aformat) {
 503                 case 0:
 504                 case XFS_DINODE_FMT_EXTENTS:
 505                         break;
 506                 default:
 507                         return __this_address;
 508                 }
 509                 if (dip->di_anextents)
 510                         return __this_address;
 511         }
 512
 513         /* extent size hint validation */
 514         fa = xfs_inode_validate_extsize(mp, be32_to_cpu(dip->di_extsize),
 515                         mode, flags);
 516         if (fa)
 517                 return fa;
 518
 519         /* only version 3 or greater inodes are extensively verified here */
 520         if (dip->di_version < 3)
 521                 return NULL;
 522
 523         flags2 = be64_to_cpu(dip->di_flags2);
 524
 525         /* don't allow reflink/cowextsize if we don't have reflink */
 526         if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
 527              !xfs_sb_version_hasreflink(&mp->m_sb))
 528                 return __this_address;
 529
 530         /* only regular files get reflink */
 531         if ((flags2 & XFS_DIFLAG2_REFLINK) && (mode & S_IFMT) != S_IFREG)
 532                 return __this_address;
 533
 534         /* don't let reflink and realtime mix */
 535         if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
 536                 return __this_address;
 537
 538         /* don't let reflink and dax mix */
 539         if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
 540                 return __this_address;
 541
 542         return NULL;
 543 }
 544
 545 void
 546 xfs_dinode_calc_crc(
 547         struct xfs_mount        *mp,
 548         struct xfs_dinode       *dip)
 549 {
 550         uint32_t                crc;
 551
 552         if (dip->di_version < 3)
 553                 return;
 554
 555         ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
 556         crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
 557                               XFS_DINODE_CRC_OFF);
 558         dip->di_crc = xfs_end_cksum(crc);
 559 }
 560
 561 /*
 562  * Read the disk inode attributes into the in-core inode structure.
 563  *
 564  * For version 5 superblocks, if we are initialising a new inode and we are not
 565  * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new
 566  * inode core with a random generation number. If we are keeping inodes around,
 567  * we need to read the inode cluster to get the existing generation number off
 568  * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode
 569  * format) then log recovery is dependent on the di_flushiter field being
 570  * initialised from the current on-disk value and hence we must also read the
 571  * inode off disk.
 572  */
 573 int
 574 xfs_iread(
 575         xfs_mount_t     *mp,
 576         xfs_trans_t     *tp,
 577         xfs_inode_t     *ip,
 578         uint            iget_flags)
 579 {
 580         xfs_buf_t       *bp;
 581         xfs_dinode_t    *dip;
 582         xfs_failaddr_t  fa;
 583         int             error;
 584
 585         /*
 586          * Fill in the location information in the in-core inode.
 587          */
 588         error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags);
 589         if (error)
 590                 return error;
 591
 592         /* shortcut IO on inode allocation if possible */
 593         if ((iget_flags & XFS_IGET_CREATE) &&
 594             xfs_sb_version_hascrc(&mp->m_sb) &&
 595             !(mp->m_flags & XFS_MOUNT_IKEEP)) {
 596                 /* initialise the on-disk inode core */
 597                 memset(&ip->i_d, 0, sizeof(ip->i_d));
 598                 VFS_I(ip)->i_generation = prandom_u32();
 599                 ip->i_d.di_version = 3;
 600                 return 0;
 601         }
 602
 603         /*
 604          * Get pointers to the on-disk inode and the buffer containing it.
 605          */
 606         error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags);
 607         if (error)
 608                 return error;
 609
 610         /* even unallocated inodes are verified */
 611         fa = xfs_dinode_verify(mp, ip->i_ino, dip);
 612         if (fa) {
 613                 xfs_inode_verifier_error(ip, -EFSCORRUPTED, "dinode", dip,
 614                                 sizeof(*dip), fa);
 615                 error = -EFSCORRUPTED;
 616                 goto out_brelse;
 617         }
 618
 619         /*
 620          * If the on-disk inode is already linked to a directory
 621          * entry, copy all of the inode into the in-core inode.
 622          * xfs_iformat_fork() handles copying in the inode format
 623          * specific information.
 624          * Otherwise, just get the truly permanent information.
 625          */
 626         if (dip->di_mode) {
 627                 xfs_inode_from_disk(ip, dip);
 628                 error = xfs_iformat_fork(ip, dip);
 629                 if (error)  {
 630 #ifdef DEBUG
 631                         xfs_alert(mp, "%s: xfs_iformat() returned error %d",
 632                                 __func__, error);
 633 #endif /* DEBUG */
 634                         goto out_brelse;
 635                 }
 636         } else {
 637                 /*
 638                  * Partial initialisation of the in-core inode. Just the bits
 639                  * that xfs_ialloc won't overwrite or relies on being correct.
 640                  */
 641                 ip->i_d.di_version = dip->di_version;
 642                 VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen);
 643                 ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter);
 644
 645                 /*
 646                  * Make sure to pull in the mode here as well in
 647                  * case the inode is released without being used.
 648                  * This ensures that xfs_inactive() will see that
 649                  * the inode is already free and not try to mess
 650                  * with the uninitialized part of it.
 651                  */
 652                 VFS_I(ip)->i_mode = 0;
 653         }
 654
 655         ASSERT(ip->i_d.di_version >= 2);
 656         ip->i_delayed_blks = 0;
 657
 658         /*
 659          * Mark the buffer containing the inode as something to keep
 660          * around for a while.  This helps to keep recently accessed
 661          * meta-data in-core longer.
 662          */
 663         xfs_buf_set_ref(bp, XFS_INO_REF);
 664
 665         /*
 666          * Use xfs_trans_brelse() to release the buffer containing the on-disk
 667          * inode, because it was acquired with xfs_trans_read_buf() in
 668          * xfs_imap_to_bp() above.  If tp is NULL, this is just a normal
 669          * brelse().  If we're within a transaction, then xfs_trans_brelse()
 670          * will only release the buffer if it is not dirty within the
 671          * transaction.  It will be OK to release the buffer in this case,
 672          * because inodes on disk are never destroyed and we will be locking the
 673          * new in-core inode before putting it in the cache where other
 674          * processes can find it.  Thus we don't have to worry about the inode
 675          * being changed just because we released the buffer.
 676          */
 677  out_brelse:
 678         xfs_trans_brelse(tp, bp);
 679         return error;
 680 }
 681
 682 /*
 683  * Validate di_extsize hint.
 684  *
 685  * The rules are documented at xfs_ioctl_setattr_check_extsize().
 686  * These functions must be kept in sync with each other.
 687  */
 688 xfs_failaddr_t
 689 xfs_inode_validate_extsize(
 690         struct xfs_mount                *mp,
 691         uint32_t                        extsize,
 692         uint16_t                        mode,
 693         uint16_t                        flags)
 694 {
 695         bool                            rt_flag;
 696         bool                            hint_flag;
 697         bool                            inherit_flag;
 698         uint32_t                        extsize_bytes;
 699         uint32_t                        blocksize_bytes;
 700
 701         rt_flag = (flags & XFS_DIFLAG_REALTIME);
 702         hint_flag = (flags & XFS_DIFLAG_EXTSIZE);
 703         inherit_flag = (flags & XFS_DIFLAG_EXTSZINHERIT);
 704         extsize_bytes = XFS_FSB_TO_B(mp, extsize);
 705
 706         if (rt_flag)
 707                 blocksize_bytes = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
 708         else
 709                 blocksize_bytes = mp->m_sb.sb_blocksize;
 710
 711         if ((hint_flag || inherit_flag) && !(S_ISDIR(mode) || S_ISREG(mode)))
 712                 return __this_address;
 713
 714         if (hint_flag && !S_ISREG(mode))
 715                 return __this_address;
 716
 717         if (inherit_flag && !S_ISDIR(mode))
 718                 return __this_address;
 719
 720         if ((hint_flag || inherit_flag) && extsize == 0)
 721                 return __this_address;
 722
 723         if (!(hint_flag || inherit_flag) && extsize != 0)
 724                 return __this_address;
 725
 726         if (extsize_bytes % blocksize_bytes)
 727                 return __this_address;
 728
 729         if (extsize > MAXEXTLEN)
 730                 return __this_address;
 731
 732         if (!rt_flag && extsize > mp->m_sb.sb_agblocks / 2)
 733                 return __this_address;
 734
 735         return NULL;
 736 }
 737
 738 /*
 739  * Validate di_cowextsize hint.
 740  *
 741  * The rules are documented at xfs_ioctl_setattr_check_cowextsize().
 742  * These functions must be kept in sync with each other.
 743  */
 744 xfs_failaddr_t
 745 xfs_inode_validate_cowextsize(
 746         struct xfs_mount                *mp,
 747         uint32_t                        cowextsize,
 748         uint16_t                        mode,
 749         uint16_t                        flags,
 750         uint64_t                        flags2)
 751 {
 752         bool                            rt_flag;
 753         bool                            hint_flag;
 754         uint32_t                        cowextsize_bytes;
 755
 756         rt_flag = (flags & XFS_DIFLAG_REALTIME);
 757         hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
 758         cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
 759
 760         if (hint_flag && !xfs_sb_version_hasreflink(&mp->m_sb))
 761                 return __this_address;
 762
 763         if (hint_flag && !(S_ISDIR(mode) || S_ISREG(mode)))
 764                 return __this_address;
 765
 766         if (hint_flag && cowextsize == 0)
 767                 return __this_address;
 768
 769         if (!hint_flag && cowextsize != 0)
 770                 return __this_address;
 771
 772         if (hint_flag && rt_flag)
 773                 return __this_address;
 774
 775         if (cowextsize_bytes % mp->m_sb.sb_blocksize)
 776                 return __this_address;
 777
 778         if (cowextsize > MAXEXTLEN)
 779                 return __this_address;
 780
 781         if (cowextsize > mp->m_sb.sb_agblocks / 2)
 782                 return __this_address;
 783
 784         return NULL;
 785 }