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_defer.h" 28 #include "xfs_inode.h" 29 #include "xfs_ialloc.h" 30 #include "xfs_alloc.h" 31 #include "xfs_error.h" 32 #include "xfs_trace.h" 33 #include "xfs_cksum.h" 34 #include "xfs_trans.h" 35 #include "xfs_buf_item.h" 36 #include "xfs_bmap_btree.h" 37 #include "xfs_alloc_btree.h" 38 #include "xfs_ialloc_btree.h" 39 #include "xfs_log.h" 40 #include "xfs_rmap_btree.h" 41 #include "xfs_bmap.h" 42 #include "xfs_refcount_btree.h" 43 44 /* 45 * Physical superblock buffer manipulations. Shared with libxfs in userspace. 46 */ 47 48 /* 49 * Reference counting access wrappers to the perag structures. 50 * Because we never free per-ag structures, the only thing we 51 * have to protect against changes is the tree structure itself. 52 */ 53 struct xfs_perag * 54 xfs_perag_get( 55 struct xfs_mount *mp, 56 xfs_agnumber_t agno) 57 { 58 struct xfs_perag *pag; 59 int ref = 0; 60 61 rcu_read_lock(); 62 pag = radix_tree_lookup(&mp->m_perag_tree, agno); 63 if (pag) { 64 ASSERT(atomic_read(&pag->pag_ref) >= 0); 65 ref = atomic_inc_return(&pag->pag_ref); 66 } 67 rcu_read_unlock(); 68 trace_xfs_perag_get(mp, agno, ref, _RET_IP_); 69 return pag; 70 } 71 72 /* 73 * search from @first to find the next perag with the given tag set. 74 */ 75 struct xfs_perag * 76 xfs_perag_get_tag( 77 struct xfs_mount *mp, 78 xfs_agnumber_t first, 79 int tag) 80 { 81 struct xfs_perag *pag; 82 int found; 83 int ref; 84 85 rcu_read_lock(); 86 found = radix_tree_gang_lookup_tag(&mp->m_perag_tree, 87 (void **)&pag, first, 1, tag); 88 if (found <= 0) { 89 rcu_read_unlock(); 90 return NULL; 91 } 92 ref = atomic_inc_return(&pag->pag_ref); 93 rcu_read_unlock(); 94 trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_); 95 return pag; 96 } 97 98 void 99 xfs_perag_put( 100 struct xfs_perag *pag) 101 { 102 int ref; 103 104 ASSERT(atomic_read(&pag->pag_ref) > 0); 105 ref = atomic_dec_return(&pag->pag_ref); 106 trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_); 107 } 108 109 /* 110 * Check the validity of the SB found. 111 */ 112 STATIC int 113 xfs_mount_validate_sb( 114 xfs_mount_t *mp, 115 xfs_sb_t *sbp, 116 bool check_inprogress, 117 bool check_version) 118 { 119 if (sbp->sb_magicnum != XFS_SB_MAGIC) { 120 xfs_warn(mp, "bad magic number"); 121 return -EWRONGFS; 122 } 123 124 125 if (!xfs_sb_good_version(sbp)) { 126 xfs_warn(mp, "bad version"); 127 return -EWRONGFS; 128 } 129 130 /* 131 * Version 5 superblock feature mask validation. Reject combinations the 132 * kernel cannot support up front before checking anything else. For 133 * write validation, we don't need to check feature masks. 134 */ 135 if (check_version && XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) { 136 if (xfs_sb_has_compat_feature(sbp, 137 XFS_SB_FEAT_COMPAT_UNKNOWN)) { 138 xfs_warn(mp, 139 "Superblock has unknown compatible features (0x%x) enabled.", 140 (sbp->sb_features_compat & 141 XFS_SB_FEAT_COMPAT_UNKNOWN)); 142 xfs_warn(mp, 143 "Using a more recent kernel is recommended."); 144 } 145 146 if (xfs_sb_has_ro_compat_feature(sbp, 147 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { 148 xfs_alert(mp, 149 "Superblock has unknown read-only compatible features (0x%x) enabled.", 150 (sbp->sb_features_ro_compat & 151 XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); 152 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { 153 xfs_warn(mp, 154 "Attempted to mount read-only compatible filesystem read-write."); 155 xfs_warn(mp, 156 "Filesystem can only be safely mounted read only."); 157 158 return -EINVAL; 159 } 160 } 161 if (xfs_sb_has_incompat_feature(sbp, 162 XFS_SB_FEAT_INCOMPAT_UNKNOWN)) { 163 xfs_warn(mp, 164 "Superblock has unknown incompatible features (0x%x) enabled.", 165 (sbp->sb_features_incompat & 166 XFS_SB_FEAT_INCOMPAT_UNKNOWN)); 167 xfs_warn(mp, 168 "Filesystem can not be safely mounted by this kernel."); 169 return -EINVAL; 170 } 171 } else if (xfs_sb_version_hascrc(sbp)) { 172 /* 173 * We can't read verify the sb LSN because the read verifier is 174 * called before the log is allocated and processed. We know the 175 * log is set up before write verifier (!check_version) calls, 176 * so just check it here. 177 */ 178 if (!xfs_log_check_lsn(mp, sbp->sb_lsn)) 179 return -EFSCORRUPTED; 180 } 181 182 if (xfs_sb_version_has_pquotino(sbp)) { 183 if (sbp->sb_qflags & (XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD)) { 184 xfs_notice(mp, 185 "Version 5 of Super block has XFS_OQUOTA bits."); 186 return -EFSCORRUPTED; 187 } 188 } else if (sbp->sb_qflags & (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD | 189 XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) { 190 xfs_notice(mp, 191 "Superblock earlier than Version 5 has XFS_[PQ]UOTA_{ENFD|CHKD} bits."); 192 return -EFSCORRUPTED; 193 } 194 195 /* 196 * Full inode chunks must be aligned to inode chunk size when 197 * sparse inodes are enabled to support the sparse chunk 198 * allocation algorithm and prevent overlapping inode records. 199 */ 200 if (xfs_sb_version_hassparseinodes(sbp)) { 201 uint32_t align; 202 203 align = XFS_INODES_PER_CHUNK * sbp->sb_inodesize 204 >> sbp->sb_blocklog; 205 if (sbp->sb_inoalignmt != align) { 206 xfs_warn(mp, 207 "Inode block alignment (%u) must match chunk size (%u) for sparse inodes.", 208 sbp->sb_inoalignmt, align); 209 return -EINVAL; 210 } 211 } 212 213 if (unlikely( 214 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) { 215 xfs_warn(mp, 216 "filesystem is marked as having an external log; " 217 "specify logdev on the mount command line."); 218 return -EINVAL; 219 } 220 221 if (unlikely( 222 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) { 223 xfs_warn(mp, 224 "filesystem is marked as having an internal log; " 225 "do not specify logdev on the mount command line."); 226 return -EINVAL; 227 } 228 229 /* 230 * More sanity checking. Most of these were stolen directly from 231 * xfs_repair. 232 */ 233 if (unlikely( 234 sbp->sb_agcount <= 0 || 235 sbp->sb_sectsize < XFS_MIN_SECTORSIZE || 236 sbp->sb_sectsize > XFS_MAX_SECTORSIZE || 237 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG || 238 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG || 239 sbp->sb_sectsize != (1 << sbp->sb_sectlog) || 240 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE || 241 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE || 242 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG || 243 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || 244 sbp->sb_blocksize != (1 << sbp->sb_blocklog) || 245 sbp->sb_dirblklog + sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG || 246 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE || 247 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE || 248 sbp->sb_inodelog < XFS_DINODE_MIN_LOG || 249 sbp->sb_inodelog > XFS_DINODE_MAX_LOG || 250 sbp->sb_inodesize != (1 << sbp->sb_inodelog) || 251 sbp->sb_logsunit > XLOG_MAX_RECORD_BSIZE || 252 sbp->sb_inopblock != howmany(sbp->sb_blocksize,sbp->sb_inodesize) || 253 (sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) || 254 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) || 255 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) || 256 (sbp->sb_imax_pct > 100 /* zero sb_imax_pct is valid */) || 257 sbp->sb_dblocks == 0 || 258 sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) || 259 sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp) || 260 sbp->sb_shared_vn != 0)) { 261 xfs_notice(mp, "SB sanity check failed"); 262 return -EFSCORRUPTED; 263 } 264 265 if (xfs_sb_version_hascrc(&mp->m_sb) && 266 sbp->sb_blocksize < XFS_MIN_CRC_BLOCKSIZE) { 267 xfs_notice(mp, "v5 SB sanity check failed"); 268 return -EFSCORRUPTED; 269 } 270 271 /* 272 * Until this is fixed only page-sized or smaller data blocks work. 273 */ 274 if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) { 275 xfs_warn(mp, 276 "File system with blocksize %d bytes. " 277 "Only pagesize (%ld) or less will currently work.", 278 sbp->sb_blocksize, PAGE_SIZE); 279 return -ENOSYS; 280 } 281 282 /* 283 * Currently only very few inode sizes are supported. 284 */ 285 switch (sbp->sb_inodesize) { 286 case 256: 287 case 512: 288 case 1024: 289 case 2048: 290 break; 291 default: 292 xfs_warn(mp, "inode size of %d bytes not supported", 293 sbp->sb_inodesize); 294 return -ENOSYS; 295 } 296 297 if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) || 298 xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) { 299 xfs_warn(mp, 300 "file system too large to be mounted on this system."); 301 return -EFBIG; 302 } 303 304 if (check_inprogress && sbp->sb_inprogress) { 305 xfs_warn(mp, "Offline file system operation in progress!"); 306 return -EFSCORRUPTED; 307 } 308 return 0; 309 } 310 311 void 312 xfs_sb_quota_from_disk(struct xfs_sb *sbp) 313 { 314 /* 315 * older mkfs doesn't initialize quota inodes to NULLFSINO. This 316 * leads to in-core values having two different values for a quota 317 * inode to be invalid: 0 and NULLFSINO. Change it to a single value 318 * NULLFSINO. 319 * 320 * Note that this change affect only the in-core values. These 321 * values are not written back to disk unless any quota information 322 * is written to the disk. Even in that case, sb_pquotino field is 323 * not written to disk unless the superblock supports pquotino. 324 */ 325 if (sbp->sb_uquotino == 0) 326 sbp->sb_uquotino = NULLFSINO; 327 if (sbp->sb_gquotino == 0) 328 sbp->sb_gquotino = NULLFSINO; 329 if (sbp->sb_pquotino == 0) 330 sbp->sb_pquotino = NULLFSINO; 331 332 /* 333 * We need to do these manipilations only if we are working 334 * with an older version of on-disk superblock. 335 */ 336 if (xfs_sb_version_has_pquotino(sbp)) 337 return; 338 339 if (sbp->sb_qflags & XFS_OQUOTA_ENFD) 340 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? 341 XFS_PQUOTA_ENFD : XFS_GQUOTA_ENFD; 342 if (sbp->sb_qflags & XFS_OQUOTA_CHKD) 343 sbp->sb_qflags |= (sbp->sb_qflags & XFS_PQUOTA_ACCT) ? 344 XFS_PQUOTA_CHKD : XFS_GQUOTA_CHKD; 345 sbp->sb_qflags &= ~(XFS_OQUOTA_ENFD | XFS_OQUOTA_CHKD); 346 347 if (sbp->sb_qflags & XFS_PQUOTA_ACCT && 348 sbp->sb_gquotino != NULLFSINO) { 349 /* 350 * In older version of superblock, on-disk superblock only 351 * has sb_gquotino, and in-core superblock has both sb_gquotino 352 * and sb_pquotino. But, only one of them is supported at any 353 * point of time. So, if PQUOTA is set in disk superblock, 354 * copy over sb_gquotino to sb_pquotino. The NULLFSINO test 355 * above is to make sure we don't do this twice and wipe them 356 * both out! 357 */ 358 sbp->sb_pquotino = sbp->sb_gquotino; 359 sbp->sb_gquotino = NULLFSINO; 360 } 361 } 362 363 static void 364 __xfs_sb_from_disk( 365 struct xfs_sb *to, 366 xfs_dsb_t *from, 367 bool convert_xquota) 368 { 369 to->sb_magicnum = be32_to_cpu(from->sb_magicnum); 370 to->sb_blocksize = be32_to_cpu(from->sb_blocksize); 371 to->sb_dblocks = be64_to_cpu(from->sb_dblocks); 372 to->sb_rblocks = be64_to_cpu(from->sb_rblocks); 373 to->sb_rextents = be64_to_cpu(from->sb_rextents); 374 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); 375 to->sb_logstart = be64_to_cpu(from->sb_logstart); 376 to->sb_rootino = be64_to_cpu(from->sb_rootino); 377 to->sb_rbmino = be64_to_cpu(from->sb_rbmino); 378 to->sb_rsumino = be64_to_cpu(from->sb_rsumino); 379 to->sb_rextsize = be32_to_cpu(from->sb_rextsize); 380 to->sb_agblocks = be32_to_cpu(from->sb_agblocks); 381 to->sb_agcount = be32_to_cpu(from->sb_agcount); 382 to->sb_rbmblocks = be32_to_cpu(from->sb_rbmblocks); 383 to->sb_logblocks = be32_to_cpu(from->sb_logblocks); 384 to->sb_versionnum = be16_to_cpu(from->sb_versionnum); 385 to->sb_sectsize = be16_to_cpu(from->sb_sectsize); 386 to->sb_inodesize = be16_to_cpu(from->sb_inodesize); 387 to->sb_inopblock = be16_to_cpu(from->sb_inopblock); 388 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); 389 to->sb_blocklog = from->sb_blocklog; 390 to->sb_sectlog = from->sb_sectlog; 391 to->sb_inodelog = from->sb_inodelog; 392 to->sb_inopblog = from->sb_inopblog; 393 to->sb_agblklog = from->sb_agblklog; 394 to->sb_rextslog = from->sb_rextslog; 395 to->sb_inprogress = from->sb_inprogress; 396 to->sb_imax_pct = from->sb_imax_pct; 397 to->sb_icount = be64_to_cpu(from->sb_icount); 398 to->sb_ifree = be64_to_cpu(from->sb_ifree); 399 to->sb_fdblocks = be64_to_cpu(from->sb_fdblocks); 400 to->sb_frextents = be64_to_cpu(from->sb_frextents); 401 to->sb_uquotino = be64_to_cpu(from->sb_uquotino); 402 to->sb_gquotino = be64_to_cpu(from->sb_gquotino); 403 to->sb_qflags = be16_to_cpu(from->sb_qflags); 404 to->sb_flags = from->sb_flags; 405 to->sb_shared_vn = from->sb_shared_vn; 406 to->sb_inoalignmt = be32_to_cpu(from->sb_inoalignmt); 407 to->sb_unit = be32_to_cpu(from->sb_unit); 408 to->sb_width = be32_to_cpu(from->sb_width); 409 to->sb_dirblklog = from->sb_dirblklog; 410 to->sb_logsectlog = from->sb_logsectlog; 411 to->sb_logsectsize = be16_to_cpu(from->sb_logsectsize); 412 to->sb_logsunit = be32_to_cpu(from->sb_logsunit); 413 to->sb_features2 = be32_to_cpu(from->sb_features2); 414 to->sb_bad_features2 = be32_to_cpu(from->sb_bad_features2); 415 to->sb_features_compat = be32_to_cpu(from->sb_features_compat); 416 to->sb_features_ro_compat = be32_to_cpu(from->sb_features_ro_compat); 417 to->sb_features_incompat = be32_to_cpu(from->sb_features_incompat); 418 to->sb_features_log_incompat = 419 be32_to_cpu(from->sb_features_log_incompat); 420 /* crc is only used on disk, not in memory; just init to 0 here. */ 421 to->sb_crc = 0; 422 to->sb_spino_align = be32_to_cpu(from->sb_spino_align); 423 to->sb_pquotino = be64_to_cpu(from->sb_pquotino); 424 to->sb_lsn = be64_to_cpu(from->sb_lsn); 425 /* 426 * sb_meta_uuid is only on disk if it differs from sb_uuid and the 427 * feature flag is set; if not set we keep it only in memory. 428 */ 429 if (xfs_sb_version_hasmetauuid(to)) 430 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid); 431 else 432 uuid_copy(&to->sb_meta_uuid, &from->sb_uuid); 433 /* Convert on-disk flags to in-memory flags? */ 434 if (convert_xquota) 435 xfs_sb_quota_from_disk(to); 436 } 437 438 void 439 xfs_sb_from_disk( 440 struct xfs_sb *to, 441 xfs_dsb_t *from) 442 { 443 __xfs_sb_from_disk(to, from, true); 444 } 445 446 static void 447 xfs_sb_quota_to_disk( 448 struct xfs_dsb *to, 449 struct xfs_sb *from) 450 { 451 __uint16_t qflags = from->sb_qflags; 452 453 to->sb_uquotino = cpu_to_be64(from->sb_uquotino); 454 if (xfs_sb_version_has_pquotino(from)) { 455 to->sb_qflags = cpu_to_be16(from->sb_qflags); 456 to->sb_gquotino = cpu_to_be64(from->sb_gquotino); 457 to->sb_pquotino = cpu_to_be64(from->sb_pquotino); 458 return; 459 } 460 461 /* 462 * The in-core version of sb_qflags do not have XFS_OQUOTA_* 463 * flags, whereas the on-disk version does. So, convert incore 464 * XFS_{PG}QUOTA_* flags to on-disk XFS_OQUOTA_* flags. 465 */ 466 qflags &= ~(XFS_PQUOTA_ENFD | XFS_PQUOTA_CHKD | 467 XFS_GQUOTA_ENFD | XFS_GQUOTA_CHKD); 468 469 if (from->sb_qflags & 470 (XFS_PQUOTA_ENFD | XFS_GQUOTA_ENFD)) 471 qflags |= XFS_OQUOTA_ENFD; 472 if (from->sb_qflags & 473 (XFS_PQUOTA_CHKD | XFS_GQUOTA_CHKD)) 474 qflags |= XFS_OQUOTA_CHKD; 475 to->sb_qflags = cpu_to_be16(qflags); 476 477 /* 478 * GQUOTINO and PQUOTINO cannot be used together in versions 479 * of superblock that do not have pquotino. from->sb_flags 480 * tells us which quota is active and should be copied to 481 * disk. If neither are active, we should NULL the inode. 482 * 483 * In all cases, the separate pquotino must remain 0 because it 484 * it beyond the "end" of the valid non-pquotino superblock. 485 */ 486 if (from->sb_qflags & XFS_GQUOTA_ACCT) 487 to->sb_gquotino = cpu_to_be64(from->sb_gquotino); 488 else if (from->sb_qflags & XFS_PQUOTA_ACCT) 489 to->sb_gquotino = cpu_to_be64(from->sb_pquotino); 490 else { 491 /* 492 * We can't rely on just the fields being logged to tell us 493 * that it is safe to write NULLFSINO - we should only do that 494 * if quotas are not actually enabled. Hence only write 495 * NULLFSINO if both in-core quota inodes are NULL. 496 */ 497 if (from->sb_gquotino == NULLFSINO && 498 from->sb_pquotino == NULLFSINO) 499 to->sb_gquotino = cpu_to_be64(NULLFSINO); 500 } 501 502 to->sb_pquotino = 0; 503 } 504 505 void 506 xfs_sb_to_disk( 507 struct xfs_dsb *to, 508 struct xfs_sb *from) 509 { 510 xfs_sb_quota_to_disk(to, from); 511 512 to->sb_magicnum = cpu_to_be32(from->sb_magicnum); 513 to->sb_blocksize = cpu_to_be32(from->sb_blocksize); 514 to->sb_dblocks = cpu_to_be64(from->sb_dblocks); 515 to->sb_rblocks = cpu_to_be64(from->sb_rblocks); 516 to->sb_rextents = cpu_to_be64(from->sb_rextents); 517 memcpy(&to->sb_uuid, &from->sb_uuid, sizeof(to->sb_uuid)); 518 to->sb_logstart = cpu_to_be64(from->sb_logstart); 519 to->sb_rootino = cpu_to_be64(from->sb_rootino); 520 to->sb_rbmino = cpu_to_be64(from->sb_rbmino); 521 to->sb_rsumino = cpu_to_be64(from->sb_rsumino); 522 to->sb_rextsize = cpu_to_be32(from->sb_rextsize); 523 to->sb_agblocks = cpu_to_be32(from->sb_agblocks); 524 to->sb_agcount = cpu_to_be32(from->sb_agcount); 525 to->sb_rbmblocks = cpu_to_be32(from->sb_rbmblocks); 526 to->sb_logblocks = cpu_to_be32(from->sb_logblocks); 527 to->sb_versionnum = cpu_to_be16(from->sb_versionnum); 528 to->sb_sectsize = cpu_to_be16(from->sb_sectsize); 529 to->sb_inodesize = cpu_to_be16(from->sb_inodesize); 530 to->sb_inopblock = cpu_to_be16(from->sb_inopblock); 531 memcpy(&to->sb_fname, &from->sb_fname, sizeof(to->sb_fname)); 532 to->sb_blocklog = from->sb_blocklog; 533 to->sb_sectlog = from->sb_sectlog; 534 to->sb_inodelog = from->sb_inodelog; 535 to->sb_inopblog = from->sb_inopblog; 536 to->sb_agblklog = from->sb_agblklog; 537 to->sb_rextslog = from->sb_rextslog; 538 to->sb_inprogress = from->sb_inprogress; 539 to->sb_imax_pct = from->sb_imax_pct; 540 to->sb_icount = cpu_to_be64(from->sb_icount); 541 to->sb_ifree = cpu_to_be64(from->sb_ifree); 542 to->sb_fdblocks = cpu_to_be64(from->sb_fdblocks); 543 to->sb_frextents = cpu_to_be64(from->sb_frextents); 544 545 to->sb_flags = from->sb_flags; 546 to->sb_shared_vn = from->sb_shared_vn; 547 to->sb_inoalignmt = cpu_to_be32(from->sb_inoalignmt); 548 to->sb_unit = cpu_to_be32(from->sb_unit); 549 to->sb_width = cpu_to_be32(from->sb_width); 550 to->sb_dirblklog = from->sb_dirblklog; 551 to->sb_logsectlog = from->sb_logsectlog; 552 to->sb_logsectsize = cpu_to_be16(from->sb_logsectsize); 553 to->sb_logsunit = cpu_to_be32(from->sb_logsunit); 554 555 /* 556 * We need to ensure that bad_features2 always matches features2. 557 * Hence we enforce that here rather than having to remember to do it 558 * everywhere else that updates features2. 559 */ 560 from->sb_bad_features2 = from->sb_features2; 561 to->sb_features2 = cpu_to_be32(from->sb_features2); 562 to->sb_bad_features2 = cpu_to_be32(from->sb_bad_features2); 563 564 if (xfs_sb_version_hascrc(from)) { 565 to->sb_features_compat = cpu_to_be32(from->sb_features_compat); 566 to->sb_features_ro_compat = 567 cpu_to_be32(from->sb_features_ro_compat); 568 to->sb_features_incompat = 569 cpu_to_be32(from->sb_features_incompat); 570 to->sb_features_log_incompat = 571 cpu_to_be32(from->sb_features_log_incompat); 572 to->sb_spino_align = cpu_to_be32(from->sb_spino_align); 573 to->sb_lsn = cpu_to_be64(from->sb_lsn); 574 if (xfs_sb_version_hasmetauuid(from)) 575 uuid_copy(&to->sb_meta_uuid, &from->sb_meta_uuid); 576 } 577 } 578 579 static int 580 xfs_sb_verify( 581 struct xfs_buf *bp, 582 bool check_version) 583 { 584 struct xfs_mount *mp = bp->b_target->bt_mount; 585 struct xfs_sb sb; 586 587 /* 588 * Use call variant which doesn't convert quota flags from disk 589 * format, because xfs_mount_validate_sb checks the on-disk flags. 590 */ 591 __xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp), false); 592 593 /* 594 * Only check the in progress field for the primary superblock as 595 * mkfs.xfs doesn't clear it from secondary superblocks. 596 */ 597 return xfs_mount_validate_sb(mp, &sb, 598 bp->b_maps[0].bm_bn == XFS_SB_DADDR, 599 check_version); 600 } 601 602 /* 603 * If the superblock has the CRC feature bit set or the CRC field is non-null, 604 * check that the CRC is valid. We check the CRC field is non-null because a 605 * single bit error could clear the feature bit and unused parts of the 606 * superblock are supposed to be zero. Hence a non-null crc field indicates that 607 * we've potentially lost a feature bit and we should check it anyway. 608 * 609 * However, past bugs (i.e. in growfs) left non-zeroed regions beyond the 610 * last field in V4 secondary superblocks. So for secondary superblocks, 611 * we are more forgiving, and ignore CRC failures if the primary doesn't 612 * indicate that the fs version is V5. 613 */ 614 static void 615 xfs_sb_read_verify( 616 struct xfs_buf *bp) 617 { 618 struct xfs_mount *mp = bp->b_target->bt_mount; 619 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); 620 int error; 621 622 /* 623 * open code the version check to avoid needing to convert the entire 624 * superblock from disk order just to check the version number 625 */ 626 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC) && 627 (((be16_to_cpu(dsb->sb_versionnum) & XFS_SB_VERSION_NUMBITS) == 628 XFS_SB_VERSION_5) || 629 dsb->sb_crc != 0)) { 630 631 if (!xfs_buf_verify_cksum(bp, XFS_SB_CRC_OFF)) { 632 /* Only fail bad secondaries on a known V5 filesystem */ 633 if (bp->b_bn == XFS_SB_DADDR || 634 xfs_sb_version_hascrc(&mp->m_sb)) { 635 error = -EFSBADCRC; 636 goto out_error; 637 } 638 } 639 } 640 error = xfs_sb_verify(bp, true); 641 642 out_error: 643 if (error) { 644 xfs_buf_ioerror(bp, error); 645 if (error == -EFSCORRUPTED || error == -EFSBADCRC) 646 xfs_verifier_error(bp); 647 } 648 } 649 650 /* 651 * We may be probed for a filesystem match, so we may not want to emit 652 * messages when the superblock buffer is not actually an XFS superblock. 653 * If we find an XFS superblock, then run a normal, noisy mount because we are 654 * really going to mount it and want to know about errors. 655 */ 656 static void 657 xfs_sb_quiet_read_verify( 658 struct xfs_buf *bp) 659 { 660 struct xfs_dsb *dsb = XFS_BUF_TO_SBP(bp); 661 662 if (dsb->sb_magicnum == cpu_to_be32(XFS_SB_MAGIC)) { 663 /* XFS filesystem, verify noisily! */ 664 xfs_sb_read_verify(bp); 665 return; 666 } 667 /* quietly fail */ 668 xfs_buf_ioerror(bp, -EWRONGFS); 669 } 670 671 static void 672 xfs_sb_write_verify( 673 struct xfs_buf *bp) 674 { 675 struct xfs_mount *mp = bp->b_target->bt_mount; 676 struct xfs_buf_log_item *bip = bp->b_fspriv; 677 int error; 678 679 error = xfs_sb_verify(bp, false); 680 if (error) { 681 xfs_buf_ioerror(bp, error); 682 xfs_verifier_error(bp); 683 return; 684 } 685 686 if (!xfs_sb_version_hascrc(&mp->m_sb)) 687 return; 688 689 if (bip) 690 XFS_BUF_TO_SBP(bp)->sb_lsn = cpu_to_be64(bip->bli_item.li_lsn); 691 692 xfs_buf_update_cksum(bp, XFS_SB_CRC_OFF); 693 } 694 695 const struct xfs_buf_ops xfs_sb_buf_ops = { 696 .name = "xfs_sb", 697 .verify_read = xfs_sb_read_verify, 698 .verify_write = xfs_sb_write_verify, 699 }; 700 701 const struct xfs_buf_ops xfs_sb_quiet_buf_ops = { 702 .name = "xfs_sb_quiet", 703 .verify_read = xfs_sb_quiet_read_verify, 704 .verify_write = xfs_sb_write_verify, 705 }; 706 707 /* 708 * xfs_mount_common 709 * 710 * Mount initialization code establishing various mount 711 * fields from the superblock associated with the given 712 * mount structure 713 */ 714 void 715 xfs_sb_mount_common( 716 struct xfs_mount *mp, 717 struct xfs_sb *sbp) 718 { 719 mp->m_agfrotor = mp->m_agirotor = 0; 720 spin_lock_init(&mp->m_agirotor_lock); 721 mp->m_maxagi = mp->m_sb.sb_agcount; 722 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG; 723 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT; 724 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT; 725 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1; 726 mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog; 727 mp->m_blockmask = sbp->sb_blocksize - 1; 728 mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG; 729 mp->m_blockwmask = mp->m_blockwsize - 1; 730 731 mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1); 732 mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0); 733 mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2; 734 mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2; 735 736 mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1); 737 mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0); 738 mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2; 739 mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2; 740 741 mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1); 742 mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0); 743 mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2; 744 mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2; 745 746 mp->m_rmap_mxr[0] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 1); 747 mp->m_rmap_mxr[1] = xfs_rmapbt_maxrecs(mp, sbp->sb_blocksize, 0); 748 mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2; 749 mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2; 750 751 mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, 752 true); 753 mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize, 754 false); 755 mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2; 756 mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2; 757 758 mp->m_bsize = XFS_FSB_TO_BB(mp, 1); 759 mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK, 760 sbp->sb_inopblock); 761 mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog; 762 763 if (sbp->sb_spino_align) 764 mp->m_ialloc_min_blks = sbp->sb_spino_align; 765 else 766 mp->m_ialloc_min_blks = mp->m_ialloc_blks; 767 mp->m_alloc_set_aside = xfs_alloc_set_aside(mp); 768 mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp); 769 } 770 771 /* 772 * xfs_initialize_perag_data 773 * 774 * Read in each per-ag structure so we can count up the number of 775 * allocated inodes, free inodes and used filesystem blocks as this 776 * information is no longer persistent in the superblock. Once we have 777 * this information, write it into the in-core superblock structure. 778 */ 779 int 780 xfs_initialize_perag_data( 781 struct xfs_mount *mp, 782 xfs_agnumber_t agcount) 783 { 784 xfs_agnumber_t index; 785 xfs_perag_t *pag; 786 xfs_sb_t *sbp = &mp->m_sb; 787 uint64_t ifree = 0; 788 uint64_t ialloc = 0; 789 uint64_t bfree = 0; 790 uint64_t bfreelst = 0; 791 uint64_t btree = 0; 792 int error; 793 794 for (index = 0; index < agcount; index++) { 795 /* 796 * read the agf, then the agi. This gets us 797 * all the information we need and populates the 798 * per-ag structures for us. 799 */ 800 error = xfs_alloc_pagf_init(mp, NULL, index, 0); 801 if (error) 802 return error; 803 804 error = xfs_ialloc_pagi_init(mp, NULL, index); 805 if (error) 806 return error; 807 pag = xfs_perag_get(mp, index); 808 ifree += pag->pagi_freecount; 809 ialloc += pag->pagi_count; 810 bfree += pag->pagf_freeblks; 811 bfreelst += pag->pagf_flcount; 812 btree += pag->pagf_btreeblks; 813 xfs_perag_put(pag); 814 } 815 816 /* Overwrite incore superblock counters with just-read data */ 817 spin_lock(&mp->m_sb_lock); 818 sbp->sb_ifree = ifree; 819 sbp->sb_icount = ialloc; 820 sbp->sb_fdblocks = bfree + bfreelst + btree; 821 spin_unlock(&mp->m_sb_lock); 822 823 xfs_reinit_percpu_counters(mp); 824 825 return 0; 826 } 827 828 /* 829 * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock 830 * into the superblock buffer to be logged. It does not provide the higher 831 * level of locking that is needed to protect the in-core superblock from 832 * concurrent access. 833 */ 834 void 835 xfs_log_sb( 836 struct xfs_trans *tp) 837 { 838 struct xfs_mount *mp = tp->t_mountp; 839 struct xfs_buf *bp = xfs_trans_getsb(tp, mp, 0); 840 841 mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); 842 mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree); 843 mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks); 844 845 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); 846 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); 847 xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb)); 848 } 849 850 /* 851 * xfs_sync_sb 852 * 853 * Sync the superblock to disk. 854 * 855 * Note that the caller is responsible for checking the frozen state of the 856 * filesystem. This procedure uses the non-blocking transaction allocator and 857 * thus will allow modifications to a frozen fs. This is required because this 858 * code can be called during the process of freezing where use of the high-level 859 * allocator would deadlock. 860 */ 861 int 862 xfs_sync_sb( 863 struct xfs_mount *mp, 864 bool wait) 865 { 866 struct xfs_trans *tp; 867 int error; 868 869 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_sb, 0, 0, 870 XFS_TRANS_NO_WRITECOUNT, &tp); 871 if (error) 872 return error; 873 874 xfs_log_sb(tp); 875 if (wait) 876 xfs_trans_set_sync(tp); 877 return xfs_trans_commit(tp); 878 } 879