1 /* 2 * Copyright (c) 2000-2005 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_bit.h" 25 #include "xfs_sb.h" 26 #include "xfs_mount.h" 27 #include "xfs_inode.h" 28 #include "xfs_ialloc.h" 29 #include "xfs_alloc.h" 30 #include "xfs_error.h" 31 #include "xfs_trace.h" 32 #include "xfs_cksum.h" 33 #include "xfs_trans.h" 34 #include "xfs_buf_item.h" 35 #include "xfs_bmap_btree.h" 36 #include "xfs_alloc_btree.h" 37 #include "xfs_ialloc_btree.h" 38 39 /* 40 * Physical superblock buffer manipulations. Shared with libxfs in userspace. 41 */ 42 43 /* 44 * Reference counting access wrappers to the perag structures. 45 * Because we never free per-ag structures, the only thing we 46 * have to protect against changes is the tree structure itself. 47 */ 48 struct xfs_perag * 49 xfs_perag_get( 50 struct xfs_mount *mp, 51 xfs_agnumber_t agno) 52 { 53 struct xfs_perag *pag; 54 int ref = 0; 55 56 rcu_read_lock(); 57 pag = radix_tree_lookup(&mp->m_perag_tree, agno); 58 if (pag) { 59 ASSERT(atomic_read(&pag->pag_ref) >= 0); 60 ref = atomic_inc_return(&pag->pag_ref); 61 } 62 rcu_read_unlock(); 63 trace_xfs_perag_get(mp, agno, ref, _RET_IP_); 64 return pag; 65 } 66 67 /* 68 * search from @first to find the next perag with the given tag set. 69 */ 70 struct xfs_perag * 71 xfs_perag_get_tag( 72 struct xfs_mount *mp, 73 xfs_agnumber_t first, 74 int tag) 75 { 76 struct xfs_perag *pag; 77 int found; 78 int ref; 79 80 rcu_read_lock(); 81 found = radix_tree_gang_lookup_tag(&mp->m_perag_tree, 82 (void **)&pag, first, 1, tag); 83 if (found <= 0) { 84 rcu_read_unlock(); 85 return NULL; 86 } 87 ref = atomic_inc_return(&pag->pag_ref); 88 rcu_read_unlock(); 89 trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_); 90 return pag; 91 } 92 93 void 94 xfs_perag_put( 95 struct xfs_perag *pag) 96 { 97 int ref; 98 99 ASSERT(atomic_read(&pag->pag_ref) > 0); 100 ref = atomic_dec_return(&pag->pag_ref); 101 trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_); 102 } 103 104 /* 105 * Check the validity of the SB found. 106 */ 107 STATIC int 108 xfs_mount_validate_sb( 109 xfs_mount_t *mp, 110 xfs_sb_t *sbp, 111 bool check_inprogress, 112 bool check_version) 113 { 114 if (sbp->sb_magicnum != XFS_SB_MAGIC) { 115 xfs_warn(mp, "bad magic number"); 116 return -EWRONGFS; 117 } 118 119 120 if (!xfs_sb_good_version(sbp)) { 121 xfs_warn(mp, "bad version"); 122 return -EWRONGFS; 123 } 124 125 /* 126 * Version 5 superblock feature mask validation. Reject combinations the 127 * kernel cannot support up front before checking anything else. For 128 * write validation, we don't need to check feature masks. 129 */ 130 if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) { 131 if (xfs_sb_has_compat_feature(sbp, 132 XFS_SB_FEAT_COMPAT_UNKNOWN)) { 133 xfs_warn(mp, 134 "Superblock has unknown compatible features (0x%x) enabled.\n" 135 "Using a more recent kernel is recommended.", 136 (sbp->sb_features_compat & 137 XFS_SB_FEAT_COMPAT_UNKNOWN)); 138 } 139 140 if (xfs_sb_has_ro_compat_feature(sbp, 141 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { 142 xfs_alert(mp, 143 "Superblock has unknown read-only compatible features (0x%x) enabled.", 144 (sbp->sb_features_ro_compat & 145 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); 146 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { 147 xfs_warn(mp, 148 "Attempted to mount read-only compatible filesystem read-write.\n" 149 "Filesystem can only be safely mounted read only."); 150 return -EINVAL; 151 } 152 } 153 if (xfs_sb_has_incompat_feature(sbp, 154 XFS_SB_FEAT_INCOMPAT_UNKNOWN)) { 155 xfs_warn(mp, 156 "Superblock has unknown incompatible features (0x%x) enabled.\n" 157 "Filesystem can not be safely mounted by this kernel.", 158 (sbp->sb_features_incompat & 159 XFS_SB_FEAT_INCOMPAT_UNKNOWN)); 160 return -EINVAL; 161 } 162 } 163 164 if (xfs_sb_version_has_pquotino(sbp)) { 165 if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) { 166 xfs_notice(mp, 167 "Version 5 of Super block has XFS_OQUOTA bits."); 168 return -EFSCORRUPTED; 169 } 170 } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD | 171 XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) { 172 xfs_notice(mp, 173 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits."); 174 return -EFSCORRUPTED; 175 } 176 177 /* 178 * Full inode chunks must be aligned to inode chunk size when 179 * sparse inodes are enabled to support the sparse chunk 180 * allocation algorithm and prevent overlapping inode records. 181 */ 182 if (xfs_sb_version_hassparseinodes(sbp)) { 183 uint32_t align; 184 185 xfs_alert(mp, 186 "EXPERIMENTAL sparse inode feature enabled. Use at your own risk!"); 187 188 align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize 189 >> sbp->sb_blocklog; 190 if (sbp->sb_inoalignmt != align) { 191 xfs_warn(mp, 192 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.", 193 sbp->sb_inoalignmt, align); 194 return -EINVAL; 195 } 196 } 197 198 if (unlikely( 199 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) { 200 xfs_warn(mp, 201 "filesystem is marked as having an external log; " 202 "specify logdev on the mount command line."); 203 return -EINVAL; 204 } 205 206 if (unlikely( 207 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) { 208 xfs_warn(mp, 209 "filesystem is marked as having an internal log; " 210 "do not specify logdev on the mount command line."); 211 return -EINVAL; 212 } 213 214 /* 215 * More sanity checking. Most of these were stolen directly from 216 * xfs_repair. 217 */ 218 if (unlikely( 219 sbp->sb_agcount <= 0 || 220 sbp->sb_sectsize < XFS_MIN_SECTORSIZE || 221 sbp->sb_sectsize > XFS_MAX_SECTORSIZE || 222 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG || 223 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG || 224 sbp->sb_sectsize != (1 << sbp->sb_sectlog) || 225 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE || 226 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE || 227 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG || 228 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || 229 sbp->sb_blocksize != (1 << sbp->sb_blocklog) || 230 sbp->sb_dirblklog > XFS_MAX_BLOCKSIZE_LOG || 231 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE || 232 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE || 233 sbp->sb_inodelog < XFS_DINODE_MIN_LOG || 234 sbp->sb_inodelog > XFS_DINODE_MAX_LOG || 235 sbp->sb_inodesize != (1 << sbp->sb_inodelog) || 236 sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE || 237 sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) || 238 (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) || 239 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) || 240 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) || 241 (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) || 242 sbp->sb_dblocks == 0 || 243 sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) || 244 sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) || 245 sbp->sb_shared_vn != 0)) { 246 xfs_notice(mp, "SB sanity check failed"); 247 return -EFSCORRUPTED; 248 } 249 250 /* 251 * Until this is fixed only page-sized or smaller data blocks work. 252 */ 253 if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) { 254 xfs_warn(mp, 255 "File system with blocksize %d bytes. " 256 "Only pagesize (%ld) or less will currently work.", 257 sbp->sb_blocksize, PAGE_SIZE); 258 return -ENOSYS; 259 } 260 261 /* 262 * Currently only very few inode sizes are supported. 263 */ 264 switch (sbp->sb_inodesize) { 265 case 256: 266 case 512: 267 case 1024: 268 case 2048: 269 break; 270 default: 271 xfs_warn(mp, "inode size of %d bytes not supported", 272 sbp->sb_inodesize); 273 return -ENOSYS; 274 } 275 276 if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) || 277 xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) { 278 xfs_warn(mp, 279 "file system too large to be mounted on this system."); 280 return -EFBIG; 281 } 282 283 if (check_inprogress && sbp->sb_inprogress) { 284 xfs_warn(mp, "Offline file system operation in progress!"); 285 return -EFSCORRUPTED; 286 } 287 return 0; 288 } 289 290 void 291 xfs_sb_quota_from_disk(struct xfs_sb *sbp) 292 { 293 /* 294 * older mkfs doesn't initialize quota inodes to NULLFSINO. This 295 * leads to in-core values having two different values for a quota 296 * inode to be invalid: 0 and NULLFSINO. Change it to a single value 297 * NULLFSINO. 298 * 299 * Note that this change affect only the in-core values. These 300 * values are not written back to disk unless any quota information 301 * is written to the disk. Even in that case, sb_pquotino field is 302 * not written to disk unless the superblock supports pquotino. 303 */ 304 if (sbp->sb_uquotino == 0) 305 sbp->sb_uquotino = NULLFSINO; 306 if (sbp->sb_gquotino == 0) 307 sbp->sb_gquotino = NULLFSINO; 308 if (sbp->sb_pquotino == 0) 309 sbp->sb_pquotino = NULLFSINO; 310 311 /* 312 * We need to do these manipilations only if we are working 313 * with an older version of on-disk superblock. 314 */ 315 if (xfs_sb_version_has_pquotino(sbp)) 316 return; 317 318 if (sbp->sb_qflags & XFS_OQUOTA_ENFD) 319 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? 320 XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD; 321 if (sbp->sb_qflags & XFS_OQUOTA_CHKD) 322 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? 323 XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD; 324 sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD); 325 326 if (sbp->sb_qflags & XFS_PQUOTA_ACCT) { 327 /* 328 * In older version of superblock, on-disk superblock only 329 * has sb_gquotino, and in-core superblock has both sb_gquotino 330 * and sb_pquotino. But, only one of them is supported at any 331 * point of time. So, if PQUOTA is set in disk superblock, 332 * copy over sb_gquotino to sb_pquotino. 333 */ 334 sbp->sb_pquotino = sbp->sb_gquotino; 335 sbp->sb_gquotino = NULLFSINO; 336 } 337 } 338 339 static void 340 __xfs_sb_from_disk( 341 struct xfs_sb *to, 342 xfs_dsb_t *from, 343 bool convert_xquota) 344 { 345 to->sb_magicnum = be32_to_cpu(from->sb_magicnum); 346 to->sb_blocksize = be32_to_cpu(from->sb_blocksize); 347 to->sb_dblocks = be64_to_cpu(from->sb_dblocks); 348 to->sb_rblocks = be64_to_cpu(from->sb_rblocks); 349 to->sb_rextents = be64_to_cpu(from->sb_rextents); 350 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); 351 to->sb_logstart = be64_to_cpu(from->sb_logstart); 352 to->sb_rootino = be64_to_cpu(from->sb_rootino); 353 to->sb_rbmino = be64_to_cpu(from->sb_rbmino); 354 to->sb_rsumino = be64_to_cpu(from->sb_rsumino); 355 to->sb_rextsize = be32_to_cpu(from->sb_rextsize); 356 to->sb_agblocks = be32_to_cpu(from->sb_agblocks); 357 to->sb_agcount = be32_to_cpu(from->sb_agcount); 358 to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks); 359 to->sb_logblocks = be32_to_cpu(from->sb_logblocks); 360 to->sb_versionnum = be16_to_cpu(from->sb_versionnum); 361 to->sb_sectsize = be16_to_cpu(from->sb_sectsize); 362 to->sb_inodesize = be16_to_cpu(from->sb_inodesize); 363 to->sb_inopblock = be16_to_cpu(from->sb_inopblock); 364 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); 365 to->sb_blocklog = from->sb_blocklog; 366 to->sb_sectlog = from->sb_sectlog; 367 to->sb_inodelog = from->sb_inodelog; 368 to->sb_inopblog = from->sb_inopblog; 369 to->sb_agblklog = from->sb_agblklog; 370 to->sb_rextslog = from->sb_rextslog; 371 to->sb_inprogress = from->sb_inprogress; 372 to->sb_imax_pct = from->sb_imax_pct; 373 to->sb_icount = be64_to_cpu(from->sb_icount); 374 to->sb_ifree = be64_to_cpu(from->sb_ifree); 375 to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks); 376 to->sb_frextents = be64_to_cpu(from->sb_frextents); 377 to->sb_uquotino = be64_to_cpu(from->sb_uquotino); 378 to->sb_gquotino = be64_to_cpu(from->sb_gquotino); 379 to->sb_qflags = be16_to_cpu(from->sb_qflags); 380 to->sb_flags = from->sb_flags; 381 to->sb_shared_vn = from->sb_shared_vn; 382 to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt); 383 to->sb_unit = be32_to_cpu(from->sb_unit); 384 to->sb_width = be32_to_cpu(from->sb_width); 385 to->sb_dirblklog = from->sb_dirblklog; 386 to->sb_logsectlog = from->sb_logsectlog; 387 to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize); 388 to->sb_logsunit = be32_to_cpu(from->sb_logsunit); 389 to->sb_features2 = be32_to_cpu(from->sb_features2); 390 to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2); 391 to->sb_features_compat = be32_to_cpu(from->sb_features_compat); 392 to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat); 393 to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat); 394 to->sb_features_log_incompat = 395 be32_to_cpu(from->sb_features_log_incompat); 396 /* crc is only used on disk, not in memory; just init to 0 here. */ 397 to->sb_crc = 0; 398 to->sb_spino_align = be32_to_cpu(from->sb_spino_align); 399 to->sb_pquotino = be64_to_cpu(from->sb_pquotino); 400 to->sb_lsn = be64_to_cpu(from->sb_lsn); 401 /* Convert on-disk flags to in-memory flags? */ 402 if (convert_xquota) 403 xfs_sb_quota_from_disk(to); 404 } 405 406 void 407 xfs_sb_from_disk( 408 struct xfs_sb *to, 409 xfs_dsb_t *from) 410 { 411 __xfs_sb_from_disk(to, from, true); 412 } 413 414 static void 415 xfs_sb_quota_to_disk( 416 struct xfs_dsb *to, 417 struct xfs_sb *from) 418 { 419 __uint16_t qflags = from->sb_qflags; 420 421 to->sb_uquotino = cpu_to_be64(from->sb_uquotino); 422 if (xfs_sb_version_has_pquotino(from)) { 423 to->sb_qflags = cpu_to_be16(from->sb_qflags); 424 to->sb_gquotino = cpu_to_be64(from->sb_gquotino); 425 to->sb_pquotino = cpu_to_be64(from->sb_pquotino); 426 return; 427 } 428 429 /* 430 * The in-core version of sb_qflags do not have XFS_OQUOTA_* 431 * flags, whereas the on-disk version does. So, convert incore 432 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags. 433 */ 434 qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD | 435 XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD); 436 437 if (from->sb_qflags & 438 (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD)) 439 qflags |= XFS_OQUOTA_ENFD; 440 if (from->sb_qflags & 441 (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) 442 qflags |= XFS_OQUOTA_CHKD; 443 to->sb_qflags = cpu_to_be16(qflags); 444 445 /* 446 * GQUOTINO and PQUOTINO cannot be used together in versions 447 * of superblock that do not have pquotino. from->sb_flags 448 * tells us which quota is active and should be copied to 449 * disk. If neither are active, we should NULL the inode. 450 * 451 * In all cases, the separate pquotino must remain 0 because it 452 * it beyond the "end" of the valid non-pquotino superblock. 453 */ 454 if (from->sb_qflags & XFS_GQUOTA_ACCT) 455 to->sb_gquotino = cpu_to_be64(from->sb_gquotino); 456 else if (from->sb_qflags & XFS_PQUOTA_ACCT) 457 to->sb_gquotino = cpu_to_be64(from->sb_pquotino); 458 else { 459 /* 460 * We can't rely on just the fields being logged to tell us 461 * that it is safe to write NULLFSINO - we should only do that 462 * if quotas are not actually enabled. Hence only write 463 * NULLFSINO if both in-core quota inodes are NULL. 464 */ 465 if (from->sb_gquotino == NULLFSINO && 466 from->sb_pquotino == NULLFSINO) 467 to->sb_gquotino = cpu_to_be64(NULLFSINO); 468 } 469 470 to->sb_pquotino = 0; 471 } 472 473 void 474 xfs_sb_to_disk( 475 struct xfs_dsb *to, 476 struct xfs_sb *from) 477 { 478 xfs_sb_quota_to_disk(to, from); 479 480 to->sb_magicnum = cpu_to_be32(from->sb_magicnum); 481 to->sb_blocksize = cpu_to_be32(from->sb_blocksize); 482 to->sb_dblocks = cpu_to_be64(from->sb_dblocks); 483 to->sb_rblocks = cpu_to_be64(from->sb_rblocks); 484 to->sb_rextents = cpu_to_be64(from->sb_rextents); 485 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); 486 to->sb_logstart = cpu_to_be64(from->sb_logstart); 487 to->sb_rootino = cpu_to_be64(from->sb_rootino); 488 to->sb_rbmino = cpu_to_be64(from->sb_rbmino); 489 to->sb_rsumino = cpu_to_be64(from->sb_rsumino); 490 to->sb_rextsize = cpu_to_be32(from->sb_rextsize); 491 to->sb_agblocks = cpu_to_be32(from->sb_agblocks); 492 to->sb_agcount = cpu_to_be32(from->sb_agcount); 493 to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks); 494 to->sb_logblocks = cpu_to_be32(from->sb_logblocks); 495 to->sb_versionnum = cpu_to_be16(from->sb_versionnum); 496 to->sb_sectsize = cpu_to_be16(from->sb_sectsize); 497 to->sb_inodesize = cpu_to_be16(from->sb_inodesize); 498 to->sb_inopblock = cpu_to_be16(from->sb_inopblock); 499 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); 500 to->sb_blocklog = from->sb_blocklog; 501 to->sb_sectlog = from->sb_sectlog; 502 to->sb_inodelog = from->sb_inodelog; 503 to->sb_inopblog = from->sb_inopblog; 504 to->sb_agblklog = from->sb_agblklog; 505 to->sb_rextslog = from->sb_rextslog; 506 to->sb_inprogress = from->sb_inprogress; 507 to->sb_imax_pct = from->sb_imax_pct; 508 to->sb_icount = cpu_to_be64(from->sb_icount); 509 to->sb_ifree = cpu_to_be64(from->sb_ifree); 510 to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks); 511 to->sb_frextents = cpu_to_be64(from->sb_frextents); 512 513 to->sb_flags = from->sb_flags; 514 to->sb_shared_vn = from->sb_shared_vn; 515 to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt); 516 to->sb_unit = cpu_to_be32(from->sb_unit); 517 to->sb_width = cpu_to_be32(from->sb_width); 518 to->sb_dirblklog = from->sb_dirblklog; 519 to->sb_logsectlog = from->sb_logsectlog; 520 to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize); 521 to->sb_logsunit = cpu_to_be32(from->sb_logsunit); 522 523 /* 524 * We need to ensure that bad_features2 always matches features2. 525 * Hence we enforce that here rather than having to remember to do it 526 * everywhere else that updates features2. 527 */ 528 from->sb_bad_features2 = from->sb_features2; 529 to->sb_features2 = cpu_to_be32(from->sb_features2); 530 to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2); 531 532 if (xfs_sb_version_hascrc(from)) { 533 to->sb_features_compat = cpu_to_be32(from->sb_features_compat); 534 to->sb_features_ro_compat = 535 cpu_to_be32(from->sb_features_ro_compat); 536 to->sb_features_incompat = 537 cpu_to_be32(from->sb_features_incompat); 538 to->sb_features_log_incompat = 539 cpu_to_be32(from->sb_features_log_incompat); 540 to->sb_spino_align = cpu_to_be32(from->sb_spino_align); 541 to->sb_lsn = cpu_to_be64(from->sb_lsn); 542 } 543 } 544 545 static int 546 xfs_sb_verify( 547 struct xfs_buf *bp, 548 bool check_version) 549 { 550 struct xfs_mount *mp = bp->b_target->bt_mount; 551 struct xfs_sb sb; 552 553 /* 554 * Use call variant which doesn't convert quota flags from disk 555 * format, because xfs_mount_validate_sb checks the on-disk flags. 556 */ 557 __xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false); 558 559 /* 560 * Only check the in progress field for the primary superblock as 561 * mkfs.xfs doesn't clear it from secondary superblocks. 562 */ 563 return xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR, 564 check_version); 565 } 566 567 /* 568 * If the superblock has the CRC feature bit set or the CRC field is non-null, 569 * check that the CRC is valid. We check the CRC field is non-null because a 570 * single bit error could clear the feature bit and unused parts of the 571 * superblock are supposed to be zero. Hence a non-null crc field indicates that 572 * we've potentially lost a feature bit and we should check it anyway. 573 * 574 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the 575 * last field in V4 secondary superblocks. So for secondary superblocks, 576 * we are more forgiving, and ignore CRC failures if the primary doesn't 577 * indicate that the fs version is V5. 578 */ 579 static void 580 xfs_sb_read_verify( 581 struct xfs_buf *bp) 582 { 583 struct xfs_mount *mp = bp->b_target->bt_mount; 584 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); 585 int error; 586 587 /* 588 * open code the version check to avoid needing to convert the entire 589 * superblock from disk order just to check the version number 590 */ 591 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) && 592 (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) == 593 XFS_SB_VERSION_5) || 594 dsb->sb_crc != 0)) { 595 596 if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) { 597 /* Only fail bad secondaries on a known V5 filesystem */ 598 if (bp->b_bn == XFS_SB_DADDR || 599 xfs_sb_version_hascrc(&mp->m_sb)) { 600 error = -EFSBADCRC; 601 goto out_error; 602 } 603 } 604 } 605 error = xfs_sb_verify(bp, true); 606 607 out_error: 608 if (error) { 609 xfs_buf_ioerror(bp, error); 610 if (error == -EFSCORRUPTED || error == -EFSBADCRC) 611 xfs_verifier_error(bp); 612 } 613 } 614 615 /* 616 * We may be probed for a filesystem match, so we may not want to emit 617 * messages when the superblock buffer is not actually an XFS superblock. 618 * If we find an XFS superblock, then run a normal, noisy mount because we are 619 * really going to mount it and want to know about errors. 620 */ 621 static void 622 xfs_sb_quiet_read_verify( 623 struct xfs_buf *bp) 624 { 625 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); 626 627 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) { 628 /* XFS filesystem, verify noisily! */ 629 xfs_sb_read_verify(bp); 630 return; 631 } 632 /* quietly fail */ 633 xfs_buf_ioerror(bp, -EWRONGFS); 634 } 635 636 static void 637 xfs_sb_write_verify( 638 struct xfs_buf *bp) 639 { 640 struct xfs_mount *mp = bp->b_target->bt_mount; 641 struct xfs_buf_log_item *bip = bp->b_fspriv; 642 int error; 643 644 error = xfs_sb_verify(bp, false); 645 if (error) { 646 xfs_buf_ioerror(bp, error); 647 xfs_verifier_error(bp); 648 return; 649 } 650 651 if (!xfs_sb_version_hascrc(&mp->m_sb)) 652 return; 653 654 if (bip) 655 XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn); 656 657 xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF); 658 } 659 660 const struct xfs_buf_ops xfs_sb_buf_ops = { 661 .verify_read = xfs_sb_read_verify, 662 .verify_write = xfs_sb_write_verify, 663 }; 664 665 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = { 666 .verify_read = xfs_sb_quiet_read_verify, 667 .verify_write = xfs_sb_write_verify, 668 }; 669 670 /* 671 * xfs_mount_common 672 * 673 * Mount initialization code establishing various mount 674 * fields from the superblock associated with the given 675 * mount structure 676 */ 677 void 678 xfs_sb_mount_common( 679 struct xfs_mount *mp, 680 struct xfs_sb *sbp) 681 { 682 mp->m_agfrotor = mp->m_agirotor = 0; 683 spin_lock_init(&mp->m_agirotor_lock); 684 mp->m_maxagi = mp->m_sb.sb_agcount; 685 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG; 686 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT; 687 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT; 688 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1; 689 mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog; 690 mp->m_blockmask = sbp->sb_blocksize - 1; 691 mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG; 692 mp->m_blockwmask = mp->m_blockwsize - 1; 693 694 mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1); 695 mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0); 696 mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2; 697 mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2; 698 699 mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1); 700 mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0); 701 mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2; 702 mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2; 703 704 mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1); 705 mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0); 706 mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2; 707 mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2; 708 709 mp->m_bsize = XFS_FSB_TO_BB(mp, 1); 710 mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK, 711 sbp->sb_inopblock); 712 mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog; 713 714 if (sbp->sb_spino_align) 715 mp->m_ialloc_min_blks = sbp->sb_spino_align; 716 else 717 mp->m_ialloc_min_blks = mp->m_ialloc_blks; 718 } 719 720 /* 721 * xfs_initialize_perag_data 722 * 723 * Read in each per-ag structure so we can count up the number of 724 * allocated inodes, free inodes and used filesystem blocks as this 725 * information is no longer persistent in the superblock. Once we have 726 * this information, write it into the in-core superblock structure. 727 */ 728 int 729 xfs_initialize_perag_data( 730 struct xfs_mount *mp, 731 xfs_agnumber_t agcount) 732 { 733 xfs_agnumber_t index; 734 xfs_perag_t *pag; 735 xfs_sb_t *sbp = &mp->m_sb; 736 uint64_t ifree = 0; 737 uint64_t ialloc = 0; 738 uint64_t bfree = 0; 739 uint64_t bfreelst = 0; 740 uint64_t btree = 0; 741 int error; 742 743 for (index = 0; index < agcount; index++) { 744 /* 745 * read the agf, then the agi. This gets us 746 * all the information we need and populates the 747 * per-ag structures for us. 748 */ 749 error = xfs_alloc_pagf_init(mp, NULL, index, 0); 750 if (error) 751 return error; 752 753 error = xfs_ialloc_pagi_init(mp, NULL, index); 754 if (error) 755 return error; 756 pag = xfs_perag_get(mp, index); 757 ifree += pag->pagi_freecount; 758 ialloc += pag->pagi_count; 759 bfree += pag->pagf_freeblks; 760 bfreelst += pag->pagf_flcount; 761 btree += pag->pagf_btreeblks; 762 xfs_perag_put(pag); 763 } 764 765 /* Overwrite incore superblock counters with just-read data */ 766 spin_lock(&mp->m_sb_lock); 767 sbp->sb_ifree = ifree; 768 sbp->sb_icount = ialloc; 769 sbp->sb_fdblocks = bfree + bfreelst + btree; 770 spin_unlock(&mp->m_sb_lock); 771 772 xfs_reinit_percpu_counters(mp); 773 774 return 0; 775 } 776 777 /* 778 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock 779 * into the superblock buffer to be logged. It does not provide the higher 780 * level of locking that is needed to protect the in-core superblock from 781 * concurrent access. 782 */ 783 void 784 xfs_log_sb( 785 struct xfs_trans *tp) 786 { 787 struct xfs_mount *mp = tp->t_mountp; 788 struct xfs_buf *bp = xfs_trans_getsb(tp, mp, 0); 789 790 mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); 791 mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree); 792 mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks); 793 794 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); 795 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); 796 xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb)); 797 } 798 799 /* 800 * xfs_sync_sb 801 * 802 * Sync the superblock to disk. 803 * 804 * Note that the caller is responsible for checking the frozen state of the 805 * filesystem. This procedure uses the non-blocking transaction allocator and 806 * thus will allow modifications to a frozen fs. This is required because this 807 * code can be called during the process of freezing where use of the high-level 808 * allocator would deadlock. 809 */ 810 int 811 xfs_sync_sb( 812 struct xfs_mount *mp, 813 bool wait) 814 { 815 struct xfs_trans *tp; 816 int error; 817 818 tp = _xfs_trans_alloc(mp, XFS_TRANS_SB_CHANGE, KM_SLEEP); 819 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0); 820 if (error) { 821 xfs_trans_cancel(tp); 822 return error; 823 } 824 825 xfs_log_sb(tp); 826 if (wait) 827 xfs_trans_set_sync(tp); 828 return xfs_trans_commit(tp); 829 } 830