| 1 | /* |
| 2 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
| 3 | * All Rights Reserved. |
| 4 | * |
| 5 | * This program is free software; you can redistribute it and/or |
| 6 | * modify it under the terms of the GNU General Public License as |
| 7 | * published by the Free Software Foundation. |
| 8 | * |
| 9 | * This program is distributed in the hope that it would be useful, |
| 10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 12 | * GNU General Public License for more details. |
| 13 | * |
| 14 | * You should have received a copy of the GNU General Public License |
| 15 | * along with this program; if not, write the Free Software Foundation, |
| 16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
| 17 | */ |
| 18 | #include "xfs.h" |
| 19 | #include "xfs_fs.h" |
| 20 | #include "xfs_shared.h" |
| 21 | #include "xfs_format.h" |
| 22 | #include "xfs_log_format.h" |
| 23 | #include "xfs_trans_resv.h" |
| 24 | #include "xfs_mount.h" |
| 25 | #include "xfs_defer.h" |
| 26 | #include "xfs_inode.h" |
| 27 | #include "xfs_error.h" |
| 28 | #include "xfs_cksum.h" |
| 29 | #include "xfs_icache.h" |
| 30 | #include "xfs_trans.h" |
| 31 | #include "xfs_ialloc.h" |
| 32 | #include "xfs_dir2.h" |
| 33 | |
| 34 | /* |
| 35 | * Check that none of the inode's in the buffer have a next |
| 36 | * unlinked field of 0. |
| 37 | */ |
| 38 | #if defined(DEBUG) |
| 39 | void |
| 40 | xfs_inobp_check( |
| 41 | xfs_mount_t *mp, |
| 42 | xfs_buf_t *bp) |
| 43 | { |
| 44 | int i; |
| 45 | int j; |
| 46 | xfs_dinode_t *dip; |
| 47 | |
| 48 | j = mp->m_inode_cluster_size >> mp->m_sb.sb_inodelog; |
| 49 | |
| 50 | for (i = 0; i < j; i++) { |
| 51 | dip = xfs_buf_offset(bp, i * mp->m_sb.sb_inodesize); |
| 52 | if (!dip->di_next_unlinked) { |
| 53 | xfs_alert(mp, |
| 54 | "Detected bogus zero next_unlinked field in inode %d buffer 0x%llx.", |
| 55 | i, (long long)bp->b_bn); |
| 56 | } |
| 57 | } |
| 58 | } |
| 59 | #endif |
| 60 | |
| 61 | bool |
| 62 | xfs_dinode_good_version( |
| 63 | struct xfs_mount *mp, |
| 64 | __u8 version) |
| 65 | { |
| 66 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 67 | return version == 3; |
| 68 | |
| 69 | return version == 1 || version == 2; |
| 70 | } |
| 71 | |
| 72 | /* |
| 73 | * If we are doing readahead on an inode buffer, we might be in log recovery |
| 74 | * reading an inode allocation buffer that hasn't yet been replayed, and hence |
| 75 | * has not had the inode cores stamped into it. Hence for readahead, the buffer |
| 76 | * may be potentially invalid. |
| 77 | * |
| 78 | * If the readahead buffer is invalid, we need to mark it with an error and |
| 79 | * clear the DONE status of the buffer so that a followup read will re-read it |
| 80 | * from disk. We don't report the error otherwise to avoid warnings during log |
| 81 | * recovery and we don't get unnecssary panics on debug kernels. We use EIO here |
| 82 | * because all we want to do is say readahead failed; there is no-one to report |
| 83 | * the error to, so this will distinguish it from a non-ra verifier failure. |
| 84 | * Changes to this readahead error behavour also need to be reflected in |
| 85 | * xfs_dquot_buf_readahead_verify(). |
| 86 | */ |
| 87 | static void |
| 88 | xfs_inode_buf_verify( |
| 89 | struct xfs_buf *bp, |
| 90 | bool readahead) |
| 91 | { |
| 92 | struct xfs_mount *mp = bp->b_target->bt_mount; |
| 93 | int i; |
| 94 | int ni; |
| 95 | |
| 96 | /* |
| 97 | * Validate the magic number and version of every inode in the buffer |
| 98 | */ |
| 99 | ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock; |
| 100 | for (i = 0; i < ni; i++) { |
| 101 | int di_ok; |
| 102 | xfs_dinode_t *dip; |
| 103 | |
| 104 | dip = xfs_buf_offset(bp, (i << mp->m_sb.sb_inodelog)); |
| 105 | di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) && |
| 106 | xfs_dinode_good_version(mp, dip->di_version); |
| 107 | if (unlikely(XFS_TEST_ERROR(!di_ok, mp, |
| 108 | XFS_ERRTAG_ITOBP_INOTOBP, |
| 109 | XFS_RANDOM_ITOBP_INOTOBP))) { |
| 110 | if (readahead) { |
| 111 | bp->b_flags &= ~XBF_DONE; |
| 112 | xfs_buf_ioerror(bp, -EIO); |
| 113 | return; |
| 114 | } |
| 115 | |
| 116 | xfs_buf_ioerror(bp, -EFSCORRUPTED); |
| 117 | xfs_verifier_error(bp); |
| 118 | #ifdef DEBUG |
| 119 | xfs_alert(mp, |
| 120 | "bad inode magic/vsn daddr %lld #%d (magic=%x)", |
| 121 | (unsigned long long)bp->b_bn, i, |
| 122 | be16_to_cpu(dip->di_magic)); |
| 123 | #endif |
| 124 | } |
| 125 | } |
| 126 | xfs_inobp_check(mp, bp); |
| 127 | } |
| 128 | |
| 129 | |
| 130 | static void |
| 131 | xfs_inode_buf_read_verify( |
| 132 | struct xfs_buf *bp) |
| 133 | { |
| 134 | xfs_inode_buf_verify(bp, false); |
| 135 | } |
| 136 | |
| 137 | static void |
| 138 | xfs_inode_buf_readahead_verify( |
| 139 | struct xfs_buf *bp) |
| 140 | { |
| 141 | xfs_inode_buf_verify(bp, true); |
| 142 | } |
| 143 | |
| 144 | static void |
| 145 | xfs_inode_buf_write_verify( |
| 146 | struct xfs_buf *bp) |
| 147 | { |
| 148 | xfs_inode_buf_verify(bp, false); |
| 149 | } |
| 150 | |
| 151 | const struct xfs_buf_ops xfs_inode_buf_ops = { |
| 152 | .name = "xfs_inode", |
| 153 | .verify_read = xfs_inode_buf_read_verify, |
| 154 | .verify_write = xfs_inode_buf_write_verify, |
| 155 | }; |
| 156 | |
| 157 | const struct xfs_buf_ops xfs_inode_buf_ra_ops = { |
| 158 | .name = "xxfs_inode_ra", |
| 159 | .verify_read = xfs_inode_buf_readahead_verify, |
| 160 | .verify_write = xfs_inode_buf_write_verify, |
| 161 | }; |
| 162 | |
| 163 | |
| 164 | /* |
| 165 | * This routine is called to map an inode to the buffer containing the on-disk |
| 166 | * version of the inode. It returns a pointer to the buffer containing the |
| 167 | * on-disk inode in the bpp parameter, and in the dipp parameter it returns a |
| 168 | * pointer to the on-disk inode within that buffer. |
| 169 | * |
| 170 | * If a non-zero error is returned, then the contents of bpp and dipp are |
| 171 | * undefined. |
| 172 | */ |
| 173 | int |
| 174 | xfs_imap_to_bp( |
| 175 | struct xfs_mount *mp, |
| 176 | struct xfs_trans *tp, |
| 177 | struct xfs_imap *imap, |
| 178 | struct xfs_dinode **dipp, |
| 179 | struct xfs_buf **bpp, |
| 180 | uint buf_flags, |
| 181 | uint iget_flags) |
| 182 | { |
| 183 | struct xfs_buf *bp; |
| 184 | int error; |
| 185 | |
| 186 | buf_flags |= XBF_UNMAPPED; |
| 187 | error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno, |
| 188 | (int)imap->im_len, buf_flags, &bp, |
| 189 | &xfs_inode_buf_ops); |
| 190 | if (error) { |
| 191 | if (error == -EAGAIN) { |
| 192 | ASSERT(buf_flags & XBF_TRYLOCK); |
| 193 | return error; |
| 194 | } |
| 195 | |
| 196 | if (error == -EFSCORRUPTED && |
| 197 | (iget_flags & XFS_IGET_UNTRUSTED)) |
| 198 | return -EINVAL; |
| 199 | |
| 200 | xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.", |
| 201 | __func__, error); |
| 202 | return error; |
| 203 | } |
| 204 | |
| 205 | *bpp = bp; |
| 206 | *dipp = xfs_buf_offset(bp, imap->im_boffset); |
| 207 | return 0; |
| 208 | } |
| 209 | |
| 210 | void |
| 211 | xfs_inode_from_disk( |
| 212 | struct xfs_inode *ip, |
| 213 | struct xfs_dinode *from) |
| 214 | { |
| 215 | struct xfs_icdinode *to = &ip->i_d; |
| 216 | struct inode *inode = VFS_I(ip); |
| 217 | |
| 218 | |
| 219 | /* |
| 220 | * Convert v1 inodes immediately to v2 inode format as this is the |
| 221 | * minimum inode version format we support in the rest of the code. |
| 222 | */ |
| 223 | to->di_version = from->di_version; |
| 224 | if (to->di_version == 1) { |
| 225 | set_nlink(inode, be16_to_cpu(from->di_onlink)); |
| 226 | to->di_projid_lo = 0; |
| 227 | to->di_projid_hi = 0; |
| 228 | to->di_version = 2; |
| 229 | } else { |
| 230 | set_nlink(inode, be32_to_cpu(from->di_nlink)); |
| 231 | to->di_projid_lo = be16_to_cpu(from->di_projid_lo); |
| 232 | to->di_projid_hi = be16_to_cpu(from->di_projid_hi); |
| 233 | } |
| 234 | |
| 235 | to->di_format = from->di_format; |
| 236 | to->di_uid = be32_to_cpu(from->di_uid); |
| 237 | to->di_gid = be32_to_cpu(from->di_gid); |
| 238 | to->di_flushiter = be16_to_cpu(from->di_flushiter); |
| 239 | |
| 240 | /* |
| 241 | * Time is signed, so need to convert to signed 32 bit before |
| 242 | * storing in inode timestamp which may be 64 bit. Otherwise |
| 243 | * a time before epoch is converted to a time long after epoch |
| 244 | * on 64 bit systems. |
| 245 | */ |
| 246 | inode->i_atime.tv_sec = (int)be32_to_cpu(from->di_atime.t_sec); |
| 247 | inode->i_atime.tv_nsec = (int)be32_to_cpu(from->di_atime.t_nsec); |
| 248 | inode->i_mtime.tv_sec = (int)be32_to_cpu(from->di_mtime.t_sec); |
| 249 | inode->i_mtime.tv_nsec = (int)be32_to_cpu(from->di_mtime.t_nsec); |
| 250 | inode->i_ctime.tv_sec = (int)be32_to_cpu(from->di_ctime.t_sec); |
| 251 | inode->i_ctime.tv_nsec = (int)be32_to_cpu(from->di_ctime.t_nsec); |
| 252 | inode->i_generation = be32_to_cpu(from->di_gen); |
| 253 | inode->i_mode = be16_to_cpu(from->di_mode); |
| 254 | |
| 255 | to->di_size = be64_to_cpu(from->di_size); |
| 256 | to->di_nblocks = be64_to_cpu(from->di_nblocks); |
| 257 | to->di_extsize = be32_to_cpu(from->di_extsize); |
| 258 | to->di_nextents = be32_to_cpu(from->di_nextents); |
| 259 | to->di_anextents = be16_to_cpu(from->di_anextents); |
| 260 | to->di_forkoff = from->di_forkoff; |
| 261 | to->di_aformat = from->di_aformat; |
| 262 | to->di_dmevmask = be32_to_cpu(from->di_dmevmask); |
| 263 | to->di_dmstate = be16_to_cpu(from->di_dmstate); |
| 264 | to->di_flags = be16_to_cpu(from->di_flags); |
| 265 | |
| 266 | if (to->di_version == 3) { |
| 267 | inode->i_version = be64_to_cpu(from->di_changecount); |
| 268 | to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec); |
| 269 | to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec); |
| 270 | to->di_flags2 = be64_to_cpu(from->di_flags2); |
| 271 | to->di_cowextsize = be32_to_cpu(from->di_cowextsize); |
| 272 | } |
| 273 | } |
| 274 | |
| 275 | void |
| 276 | xfs_inode_to_disk( |
| 277 | struct xfs_inode *ip, |
| 278 | struct xfs_dinode *to, |
| 279 | xfs_lsn_t lsn) |
| 280 | { |
| 281 | struct xfs_icdinode *from = &ip->i_d; |
| 282 | struct inode *inode = VFS_I(ip); |
| 283 | |
| 284 | to->di_magic = cpu_to_be16(XFS_DINODE_MAGIC); |
| 285 | to->di_onlink = 0; |
| 286 | |
| 287 | to->di_version = from->di_version; |
| 288 | to->di_format = from->di_format; |
| 289 | to->di_uid = cpu_to_be32(from->di_uid); |
| 290 | to->di_gid = cpu_to_be32(from->di_gid); |
| 291 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); |
| 292 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); |
| 293 | |
| 294 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
| 295 | to->di_atime.t_sec = cpu_to_be32(inode->i_atime.tv_sec); |
| 296 | to->di_atime.t_nsec = cpu_to_be32(inode->i_atime.tv_nsec); |
| 297 | to->di_mtime.t_sec = cpu_to_be32(inode->i_mtime.tv_sec); |
| 298 | to->di_mtime.t_nsec = cpu_to_be32(inode->i_mtime.tv_nsec); |
| 299 | to->di_ctime.t_sec = cpu_to_be32(inode->i_ctime.tv_sec); |
| 300 | to->di_ctime.t_nsec = cpu_to_be32(inode->i_ctime.tv_nsec); |
| 301 | to->di_nlink = cpu_to_be32(inode->i_nlink); |
| 302 | to->di_gen = cpu_to_be32(inode->i_generation); |
| 303 | to->di_mode = cpu_to_be16(inode->i_mode); |
| 304 | |
| 305 | to->di_size = cpu_to_be64(from->di_size); |
| 306 | to->di_nblocks = cpu_to_be64(from->di_nblocks); |
| 307 | to->di_extsize = cpu_to_be32(from->di_extsize); |
| 308 | to->di_nextents = cpu_to_be32(from->di_nextents); |
| 309 | to->di_anextents = cpu_to_be16(from->di_anextents); |
| 310 | to->di_forkoff = from->di_forkoff; |
| 311 | to->di_aformat = from->di_aformat; |
| 312 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); |
| 313 | to->di_dmstate = cpu_to_be16(from->di_dmstate); |
| 314 | to->di_flags = cpu_to_be16(from->di_flags); |
| 315 | |
| 316 | if (from->di_version == 3) { |
| 317 | to->di_changecount = cpu_to_be64(inode->i_version); |
| 318 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); |
| 319 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); |
| 320 | to->di_flags2 = cpu_to_be64(from->di_flags2); |
| 321 | to->di_cowextsize = cpu_to_be32(from->di_cowextsize); |
| 322 | to->di_ino = cpu_to_be64(ip->i_ino); |
| 323 | to->di_lsn = cpu_to_be64(lsn); |
| 324 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); |
| 325 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); |
| 326 | to->di_flushiter = 0; |
| 327 | } else { |
| 328 | to->di_flushiter = cpu_to_be16(from->di_flushiter); |
| 329 | } |
| 330 | } |
| 331 | |
| 332 | void |
| 333 | xfs_log_dinode_to_disk( |
| 334 | struct xfs_log_dinode *from, |
| 335 | struct xfs_dinode *to) |
| 336 | { |
| 337 | to->di_magic = cpu_to_be16(from->di_magic); |
| 338 | to->di_mode = cpu_to_be16(from->di_mode); |
| 339 | to->di_version = from->di_version; |
| 340 | to->di_format = from->di_format; |
| 341 | to->di_onlink = 0; |
| 342 | to->di_uid = cpu_to_be32(from->di_uid); |
| 343 | to->di_gid = cpu_to_be32(from->di_gid); |
| 344 | to->di_nlink = cpu_to_be32(from->di_nlink); |
| 345 | to->di_projid_lo = cpu_to_be16(from->di_projid_lo); |
| 346 | to->di_projid_hi = cpu_to_be16(from->di_projid_hi); |
| 347 | memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); |
| 348 | |
| 349 | to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec); |
| 350 | to->di_atime.t_nsec = cpu_to_be32(from->di_atime.t_nsec); |
| 351 | to->di_mtime.t_sec = cpu_to_be32(from->di_mtime.t_sec); |
| 352 | to->di_mtime.t_nsec = cpu_to_be32(from->di_mtime.t_nsec); |
| 353 | to->di_ctime.t_sec = cpu_to_be32(from->di_ctime.t_sec); |
| 354 | to->di_ctime.t_nsec = cpu_to_be32(from->di_ctime.t_nsec); |
| 355 | |
| 356 | to->di_size = cpu_to_be64(from->di_size); |
| 357 | to->di_nblocks = cpu_to_be64(from->di_nblocks); |
| 358 | to->di_extsize = cpu_to_be32(from->di_extsize); |
| 359 | to->di_nextents = cpu_to_be32(from->di_nextents); |
| 360 | to->di_anextents = cpu_to_be16(from->di_anextents); |
| 361 | to->di_forkoff = from->di_forkoff; |
| 362 | to->di_aformat = from->di_aformat; |
| 363 | to->di_dmevmask = cpu_to_be32(from->di_dmevmask); |
| 364 | to->di_dmstate = cpu_to_be16(from->di_dmstate); |
| 365 | to->di_flags = cpu_to_be16(from->di_flags); |
| 366 | to->di_gen = cpu_to_be32(from->di_gen); |
| 367 | |
| 368 | if (from->di_version == 3) { |
| 369 | to->di_changecount = cpu_to_be64(from->di_changecount); |
| 370 | to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec); |
| 371 | to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec); |
| 372 | to->di_flags2 = cpu_to_be64(from->di_flags2); |
| 373 | to->di_cowextsize = cpu_to_be32(from->di_cowextsize); |
| 374 | to->di_ino = cpu_to_be64(from->di_ino); |
| 375 | to->di_lsn = cpu_to_be64(from->di_lsn); |
| 376 | memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2)); |
| 377 | uuid_copy(&to->di_uuid, &from->di_uuid); |
| 378 | to->di_flushiter = 0; |
| 379 | } else { |
| 380 | to->di_flushiter = cpu_to_be16(from->di_flushiter); |
| 381 | } |
| 382 | } |
| 383 | |
| 384 | static bool |
| 385 | xfs_dinode_verify( |
| 386 | struct xfs_mount *mp, |
| 387 | xfs_ino_t ino, |
| 388 | struct xfs_dinode *dip) |
| 389 | { |
| 390 | uint16_t mode; |
| 391 | uint16_t flags; |
| 392 | uint64_t flags2; |
| 393 | |
| 394 | if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC)) |
| 395 | return false; |
| 396 | |
| 397 | /* don't allow invalid i_size */ |
| 398 | if (be64_to_cpu(dip->di_size) & (1ULL << 63)) |
| 399 | return false; |
| 400 | |
| 401 | mode = be16_to_cpu(dip->di_mode); |
| 402 | if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN) |
| 403 | return false; |
| 404 | |
| 405 | /* No zero-length symlinks/dirs. */ |
| 406 | if ((S_ISLNK(mode) || S_ISDIR(mode)) && dip->di_size == 0) |
| 407 | return false; |
| 408 | |
| 409 | /* only version 3 or greater inodes are extensively verified here */ |
| 410 | if (dip->di_version < 3) |
| 411 | return true; |
| 412 | |
| 413 | if (!xfs_sb_version_hascrc(&mp->m_sb)) |
| 414 | return false; |
| 415 | if (!xfs_verify_cksum((char *)dip, mp->m_sb.sb_inodesize, |
| 416 | XFS_DINODE_CRC_OFF)) |
| 417 | return false; |
| 418 | if (be64_to_cpu(dip->di_ino) != ino) |
| 419 | return false; |
| 420 | if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid)) |
| 421 | return false; |
| 422 | |
| 423 | flags = be16_to_cpu(dip->di_flags); |
| 424 | flags2 = be64_to_cpu(dip->di_flags2); |
| 425 | |
| 426 | /* don't allow reflink/cowextsize if we don't have reflink */ |
| 427 | if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) && |
| 428 | !xfs_sb_version_hasreflink(&mp->m_sb)) |
| 429 | return false; |
| 430 | |
| 431 | /* don't let reflink and realtime mix */ |
| 432 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME)) |
| 433 | return false; |
| 434 | |
| 435 | /* don't let reflink and dax mix */ |
| 436 | if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX)) |
| 437 | return false; |
| 438 | |
| 439 | return true; |
| 440 | } |
| 441 | |
| 442 | void |
| 443 | xfs_dinode_calc_crc( |
| 444 | struct xfs_mount *mp, |
| 445 | struct xfs_dinode *dip) |
| 446 | { |
| 447 | __uint32_t crc; |
| 448 | |
| 449 | if (dip->di_version < 3) |
| 450 | return; |
| 451 | |
| 452 | ASSERT(xfs_sb_version_hascrc(&mp->m_sb)); |
| 453 | crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize, |
| 454 | XFS_DINODE_CRC_OFF); |
| 455 | dip->di_crc = xfs_end_cksum(crc); |
| 456 | } |
| 457 | |
| 458 | /* |
| 459 | * Read the disk inode attributes into the in-core inode structure. |
| 460 | * |
| 461 | * For version 5 superblocks, if we are initialising a new inode and we are not |
| 462 | * utilising the XFS_MOUNT_IKEEP inode cluster mode, we can simple build the new |
| 463 | * inode core with a random generation number. If we are keeping inodes around, |
| 464 | * we need to read the inode cluster to get the existing generation number off |
| 465 | * disk. Further, if we are using version 4 superblocks (i.e. v1/v2 inode |
| 466 | * format) then log recovery is dependent on the di_flushiter field being |
| 467 | * initialised from the current on-disk value and hence we must also read the |
| 468 | * inode off disk. |
| 469 | */ |
| 470 | int |
| 471 | xfs_iread( |
| 472 | xfs_mount_t *mp, |
| 473 | xfs_trans_t *tp, |
| 474 | xfs_inode_t *ip, |
| 475 | uint iget_flags) |
| 476 | { |
| 477 | xfs_buf_t *bp; |
| 478 | xfs_dinode_t *dip; |
| 479 | int error; |
| 480 | |
| 481 | /* |
| 482 | * Fill in the location information in the in-core inode. |
| 483 | */ |
| 484 | error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, iget_flags); |
| 485 | if (error) |
| 486 | return error; |
| 487 | |
| 488 | /* shortcut IO on inode allocation if possible */ |
| 489 | if ((iget_flags & XFS_IGET_CREATE) && |
| 490 | xfs_sb_version_hascrc(&mp->m_sb) && |
| 491 | !(mp->m_flags & XFS_MOUNT_IKEEP)) { |
| 492 | /* initialise the on-disk inode core */ |
| 493 | memset(&ip->i_d, 0, sizeof(ip->i_d)); |
| 494 | VFS_I(ip)->i_generation = prandom_u32(); |
| 495 | if (xfs_sb_version_hascrc(&mp->m_sb)) |
| 496 | ip->i_d.di_version = 3; |
| 497 | else |
| 498 | ip->i_d.di_version = 2; |
| 499 | return 0; |
| 500 | } |
| 501 | |
| 502 | /* |
| 503 | * Get pointers to the on-disk inode and the buffer containing it. |
| 504 | */ |
| 505 | error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &dip, &bp, 0, iget_flags); |
| 506 | if (error) |
| 507 | return error; |
| 508 | |
| 509 | /* even unallocated inodes are verified */ |
| 510 | if (!xfs_dinode_verify(mp, ip->i_ino, dip)) { |
| 511 | xfs_alert(mp, "%s: validation failed for inode %lld failed", |
| 512 | __func__, ip->i_ino); |
| 513 | |
| 514 | XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, dip); |
| 515 | error = -EFSCORRUPTED; |
| 516 | goto out_brelse; |
| 517 | } |
| 518 | |
| 519 | /* |
| 520 | * If the on-disk inode is already linked to a directory |
| 521 | * entry, copy all of the inode into the in-core inode. |
| 522 | * xfs_iformat_fork() handles copying in the inode format |
| 523 | * specific information. |
| 524 | * Otherwise, just get the truly permanent information. |
| 525 | */ |
| 526 | if (dip->di_mode) { |
| 527 | xfs_inode_from_disk(ip, dip); |
| 528 | error = xfs_iformat_fork(ip, dip); |
| 529 | if (error) { |
| 530 | #ifdef DEBUG |
| 531 | xfs_alert(mp, "%s: xfs_iformat() returned error %d", |
| 532 | __func__, error); |
| 533 | #endif /* DEBUG */ |
| 534 | goto out_brelse; |
| 535 | } |
| 536 | } else { |
| 537 | /* |
| 538 | * Partial initialisation of the in-core inode. Just the bits |
| 539 | * that xfs_ialloc won't overwrite or relies on being correct. |
| 540 | */ |
| 541 | ip->i_d.di_version = dip->di_version; |
| 542 | VFS_I(ip)->i_generation = be32_to_cpu(dip->di_gen); |
| 543 | ip->i_d.di_flushiter = be16_to_cpu(dip->di_flushiter); |
| 544 | |
| 545 | /* |
| 546 | * Make sure to pull in the mode here as well in |
| 547 | * case the inode is released without being used. |
| 548 | * This ensures that xfs_inactive() will see that |
| 549 | * the inode is already free and not try to mess |
| 550 | * with the uninitialized part of it. |
| 551 | */ |
| 552 | VFS_I(ip)->i_mode = 0; |
| 553 | } |
| 554 | |
| 555 | ASSERT(ip->i_d.di_version >= 2); |
| 556 | ip->i_delayed_blks = 0; |
| 557 | |
| 558 | /* |
| 559 | * Mark the buffer containing the inode as something to keep |
| 560 | * around for a while. This helps to keep recently accessed |
| 561 | * meta-data in-core longer. |
| 562 | */ |
| 563 | xfs_buf_set_ref(bp, XFS_INO_REF); |
| 564 | |
| 565 | /* |
| 566 | * Use xfs_trans_brelse() to release the buffer containing the on-disk |
| 567 | * inode, because it was acquired with xfs_trans_read_buf() in |
| 568 | * xfs_imap_to_bp() above. If tp is NULL, this is just a normal |
| 569 | * brelse(). If we're within a transaction, then xfs_trans_brelse() |
| 570 | * will only release the buffer if it is not dirty within the |
| 571 | * transaction. It will be OK to release the buffer in this case, |
| 572 | * because inodes on disk are never destroyed and we will be locking the |
| 573 | * new in-core inode before putting it in the cache where other |
| 574 | * processes can find it. Thus we don't have to worry about the inode |
| 575 | * being changed just because we released the buffer. |
| 576 | */ |
| 577 | out_brelse: |
| 578 | xfs_trans_brelse(tp, bp); |
| 579 | return error; |
| 580 | } |