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_types.h" 21 #include "xfs_log.h" 22 #include "xfs_trans.h" 23 #include "xfs_sb.h" 24 #include "xfs_ag.h" 25 #include "xfs_mount.h" 26 #include "xfs_bmap_btree.h" 27 #include "xfs_alloc_btree.h" 28 #include "xfs_ialloc_btree.h" 29 #include "xfs_dinode.h" 30 #include "xfs_inode.h" 31 #include "xfs_inode_item.h" 32 #include "xfs_btree.h" 33 #include "xfs_error.h" 34 #include "xfs_alloc.h" 35 #include "xfs_ialloc.h" 36 #include "xfs_fsops.h" 37 #include "xfs_itable.h" 38 #include "xfs_trans_space.h" 39 #include "xfs_rtalloc.h" 40 #include "xfs_filestream.h" 41 #include "xfs_trace.h" 42 43 /* 44 * File system operations 45 */ 46 47 int 48 xfs_fs_geometry( 49 xfs_mount_t *mp, 50 xfs_fsop_geom_t *geo, 51 int new_version) 52 { 53 54 memset(geo, 0, sizeof(*geo)); 55 56 geo->blocksize = mp->m_sb.sb_blocksize; 57 geo->rtextsize = mp->m_sb.sb_rextsize; 58 geo->agblocks = mp->m_sb.sb_agblocks; 59 geo->agcount = mp->m_sb.sb_agcount; 60 geo->logblocks = mp->m_sb.sb_logblocks; 61 geo->sectsize = mp->m_sb.sb_sectsize; 62 geo->inodesize = mp->m_sb.sb_inodesize; 63 geo->imaxpct = mp->m_sb.sb_imax_pct; 64 geo->datablocks = mp->m_sb.sb_dblocks; 65 geo->rtblocks = mp->m_sb.sb_rblocks; 66 geo->rtextents = mp->m_sb.sb_rextents; 67 geo->logstart = mp->m_sb.sb_logstart; 68 ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid)); 69 memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid)); 70 if (new_version >= 2) { 71 geo->sunit = mp->m_sb.sb_unit; 72 geo->swidth = mp->m_sb.sb_width; 73 } 74 if (new_version >= 3) { 75 geo->version = XFS_FSOP_GEOM_VERSION; 76 geo->flags = 77 (xfs_sb_version_hasattr(&mp->m_sb) ? 78 XFS_FSOP_GEOM_FLAGS_ATTR : 0) | 79 (xfs_sb_version_hasnlink(&mp->m_sb) ? 80 XFS_FSOP_GEOM_FLAGS_NLINK : 0) | 81 (xfs_sb_version_hasquota(&mp->m_sb) ? 82 XFS_FSOP_GEOM_FLAGS_QUOTA : 0) | 83 (xfs_sb_version_hasalign(&mp->m_sb) ? 84 XFS_FSOP_GEOM_FLAGS_IALIGN : 0) | 85 (xfs_sb_version_hasdalign(&mp->m_sb) ? 86 XFS_FSOP_GEOM_FLAGS_DALIGN : 0) | 87 (xfs_sb_version_hasshared(&mp->m_sb) ? 88 XFS_FSOP_GEOM_FLAGS_SHARED : 0) | 89 (xfs_sb_version_hasextflgbit(&mp->m_sb) ? 90 XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) | 91 (xfs_sb_version_hasdirv2(&mp->m_sb) ? 92 XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) | 93 (xfs_sb_version_hassector(&mp->m_sb) ? 94 XFS_FSOP_GEOM_FLAGS_SECTOR : 0) | 95 (xfs_sb_version_hasasciici(&mp->m_sb) ? 96 XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) | 97 (xfs_sb_version_haslazysbcount(&mp->m_sb) ? 98 XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) | 99 (xfs_sb_version_hasattr2(&mp->m_sb) ? 100 XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) | 101 (xfs_sb_version_hasprojid32bit(&mp->m_sb) ? 102 XFS_FSOP_GEOM_FLAGS_PROJID32 : 0); 103 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ? 104 mp->m_sb.sb_logsectsize : BBSIZE; 105 geo->rtsectsize = mp->m_sb.sb_blocksize; 106 geo->dirblocksize = mp->m_dirblksize; 107 } 108 if (new_version >= 4) { 109 geo->flags |= 110 (xfs_sb_version_haslogv2(&mp->m_sb) ? 111 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0); 112 geo->logsunit = mp->m_sb.sb_logsunit; 113 } 114 return 0; 115 } 116 117 static struct xfs_buf * 118 xfs_growfs_get_hdr_buf( 119 struct xfs_mount *mp, 120 xfs_daddr_t blkno, 121 size_t numblks, 122 int flags, 123 const struct xfs_buf_ops *ops) 124 { 125 struct xfs_buf *bp; 126 127 bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags); 128 if (!bp) 129 return NULL; 130 131 xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); 132 bp->b_bn = blkno; 133 bp->b_maps[0].bm_bn = blkno; 134 bp->b_ops = ops; 135 136 return bp; 137 } 138 139 static int 140 xfs_growfs_data_private( 141 xfs_mount_t *mp, /* mount point for filesystem */ 142 xfs_growfs_data_t *in) /* growfs data input struct */ 143 { 144 xfs_agf_t *agf; 145 struct xfs_agfl *agfl; 146 xfs_agi_t *agi; 147 xfs_agnumber_t agno; 148 xfs_extlen_t agsize; 149 xfs_extlen_t tmpsize; 150 xfs_alloc_rec_t *arec; 151 xfs_buf_t *bp; 152 int bucket; 153 int dpct; 154 int error; 155 xfs_agnumber_t nagcount; 156 xfs_agnumber_t nagimax = 0; 157 xfs_rfsblock_t nb, nb_mod; 158 xfs_rfsblock_t new; 159 xfs_rfsblock_t nfree; 160 xfs_agnumber_t oagcount; 161 int pct; 162 xfs_trans_t *tp; 163 164 nb = in->newblocks; 165 pct = in->imaxpct; 166 if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100) 167 return XFS_ERROR(EINVAL); 168 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb))) 169 return error; 170 dpct = pct - mp->m_sb.sb_imax_pct; 171 bp = xfs_buf_read_uncached(mp->m_ddev_targp, 172 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1), 173 XFS_FSS_TO_BB(mp, 1), 0, NULL); 174 if (!bp) 175 return EIO; 176 if (bp->b_error) { 177 int error = bp->b_error; 178 xfs_buf_relse(bp); 179 return error; 180 } 181 xfs_buf_relse(bp); 182 183 new = nb; /* use new as a temporary here */ 184 nb_mod = do_div(new, mp->m_sb.sb_agblocks); 185 nagcount = new + (nb_mod != 0); 186 if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) { 187 nagcount--; 188 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks; 189 if (nb < mp->m_sb.sb_dblocks) 190 return XFS_ERROR(EINVAL); 191 } 192 new = nb - mp->m_sb.sb_dblocks; 193 oagcount = mp->m_sb.sb_agcount; 194 195 /* allocate the new per-ag structures */ 196 if (nagcount > oagcount) { 197 error = xfs_initialize_perag(mp, nagcount, &nagimax); 198 if (error) 199 return error; 200 } 201 202 tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS); 203 tp->t_flags |= XFS_TRANS_RESERVE; 204 if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp), 205 XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) { 206 xfs_trans_cancel(tp, 0); 207 return error; 208 } 209 210 /* 211 * Write new AG headers to disk. Non-transactional, but written 212 * synchronously so they are completed prior to the growfs transaction 213 * being logged. 214 */ 215 nfree = 0; 216 for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) { 217 /* 218 * AG freespace header block 219 */ 220 bp = xfs_growfs_get_hdr_buf(mp, 221 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)), 222 XFS_FSS_TO_BB(mp, 1), 0, 223 &xfs_agf_buf_ops); 224 if (!bp) { 225 error = ENOMEM; 226 goto error0; 227 } 228 229 agf = XFS_BUF_TO_AGF(bp); 230 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC); 231 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION); 232 agf->agf_seqno = cpu_to_be32(agno); 233 if (agno == nagcount - 1) 234 agsize = 235 nb - 236 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks); 237 else 238 agsize = mp->m_sb.sb_agblocks; 239 agf->agf_length = cpu_to_be32(agsize); 240 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp)); 241 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp)); 242 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1); 243 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1); 244 agf->agf_flfirst = 0; 245 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1); 246 agf->agf_flcount = 0; 247 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp); 248 agf->agf_freeblks = cpu_to_be32(tmpsize); 249 agf->agf_longest = cpu_to_be32(tmpsize); 250 error = xfs_bwrite(bp); 251 xfs_buf_relse(bp); 252 if (error) 253 goto error0; 254 255 /* 256 * AG freelist header block 257 */ 258 bp = xfs_growfs_get_hdr_buf(mp, 259 XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)), 260 XFS_FSS_TO_BB(mp, 1), 0, 261 &xfs_agfl_buf_ops); 262 if (!bp) { 263 error = ENOMEM; 264 goto error0; 265 } 266 267 agfl = XFS_BUF_TO_AGFL(bp); 268 for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++) 269 agfl->agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK); 270 271 error = xfs_bwrite(bp); 272 xfs_buf_relse(bp); 273 if (error) 274 goto error0; 275 276 /* 277 * AG inode header block 278 */ 279 bp = xfs_growfs_get_hdr_buf(mp, 280 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), 281 XFS_FSS_TO_BB(mp, 1), 0, 282 &xfs_agi_buf_ops); 283 if (!bp) { 284 error = ENOMEM; 285 goto error0; 286 } 287 288 agi = XFS_BUF_TO_AGI(bp); 289 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC); 290 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION); 291 agi->agi_seqno = cpu_to_be32(agno); 292 agi->agi_length = cpu_to_be32(agsize); 293 agi->agi_count = 0; 294 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp)); 295 agi->agi_level = cpu_to_be32(1); 296 agi->agi_freecount = 0; 297 agi->agi_newino = cpu_to_be32(NULLAGINO); 298 agi->agi_dirino = cpu_to_be32(NULLAGINO); 299 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++) 300 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO); 301 error = xfs_bwrite(bp); 302 xfs_buf_relse(bp); 303 if (error) 304 goto error0; 305 306 /* 307 * BNO btree root block 308 */ 309 bp = xfs_growfs_get_hdr_buf(mp, 310 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)), 311 BTOBB(mp->m_sb.sb_blocksize), 0, 312 &xfs_allocbt_buf_ops); 313 314 if (!bp) { 315 error = ENOMEM; 316 goto error0; 317 } 318 319 xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1, 0); 320 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1); 321 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp)); 322 arec->ar_blockcount = cpu_to_be32( 323 agsize - be32_to_cpu(arec->ar_startblock)); 324 325 error = xfs_bwrite(bp); 326 xfs_buf_relse(bp); 327 if (error) 328 goto error0; 329 330 /* 331 * CNT btree root block 332 */ 333 bp = xfs_growfs_get_hdr_buf(mp, 334 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)), 335 BTOBB(mp->m_sb.sb_blocksize), 0, 336 &xfs_allocbt_buf_ops); 337 if (!bp) { 338 error = ENOMEM; 339 goto error0; 340 } 341 342 xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1, 0); 343 arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1); 344 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp)); 345 arec->ar_blockcount = cpu_to_be32( 346 agsize - be32_to_cpu(arec->ar_startblock)); 347 nfree += be32_to_cpu(arec->ar_blockcount); 348 349 error = xfs_bwrite(bp); 350 xfs_buf_relse(bp); 351 if (error) 352 goto error0; 353 354 /* 355 * INO btree root block 356 */ 357 bp = xfs_growfs_get_hdr_buf(mp, 358 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)), 359 BTOBB(mp->m_sb.sb_blocksize), 0, 360 &xfs_inobt_buf_ops); 361 if (!bp) { 362 error = ENOMEM; 363 goto error0; 364 } 365 366 xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0, 0); 367 368 error = xfs_bwrite(bp); 369 xfs_buf_relse(bp); 370 if (error) 371 goto error0; 372 } 373 xfs_trans_agblocks_delta(tp, nfree); 374 /* 375 * There are new blocks in the old last a.g. 376 */ 377 if (new) { 378 /* 379 * Change the agi length. 380 */ 381 error = xfs_ialloc_read_agi(mp, tp, agno, &bp); 382 if (error) { 383 goto error0; 384 } 385 ASSERT(bp); 386 agi = XFS_BUF_TO_AGI(bp); 387 be32_add_cpu(&agi->agi_length, new); 388 ASSERT(nagcount == oagcount || 389 be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks); 390 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH); 391 /* 392 * Change agf length. 393 */ 394 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp); 395 if (error) { 396 goto error0; 397 } 398 ASSERT(bp); 399 agf = XFS_BUF_TO_AGF(bp); 400 be32_add_cpu(&agf->agf_length, new); 401 ASSERT(be32_to_cpu(agf->agf_length) == 402 be32_to_cpu(agi->agi_length)); 403 404 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH); 405 /* 406 * Free the new space. 407 */ 408 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno, 409 be32_to_cpu(agf->agf_length) - new), new); 410 if (error) { 411 goto error0; 412 } 413 } 414 415 /* 416 * Update changed superblock fields transactionally. These are not 417 * seen by the rest of the world until the transaction commit applies 418 * them atomically to the superblock. 419 */ 420 if (nagcount > oagcount) 421 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount); 422 if (nb > mp->m_sb.sb_dblocks) 423 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS, 424 nb - mp->m_sb.sb_dblocks); 425 if (nfree) 426 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree); 427 if (dpct) 428 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct); 429 error = xfs_trans_commit(tp, 0); 430 if (error) 431 return error; 432 433 /* New allocation groups fully initialized, so update mount struct */ 434 if (nagimax) 435 mp->m_maxagi = nagimax; 436 if (mp->m_sb.sb_imax_pct) { 437 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct; 438 do_div(icount, 100); 439 mp->m_maxicount = icount << mp->m_sb.sb_inopblog; 440 } else 441 mp->m_maxicount = 0; 442 xfs_set_low_space_thresholds(mp); 443 444 /* update secondary superblocks. */ 445 for (agno = 1; agno < nagcount; agno++) { 446 error = 0; 447 /* 448 * new secondary superblocks need to be zeroed, not read from 449 * disk as the contents of the new area we are growing into is 450 * completely unknown. 451 */ 452 if (agno < oagcount) { 453 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, 454 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)), 455 XFS_FSS_TO_BB(mp, 1), 0, &bp, 456 &xfs_sb_buf_ops); 457 } else { 458 bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp, 459 XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)), 460 XFS_FSS_TO_BB(mp, 1), 0); 461 if (bp) { 462 bp->b_ops = &xfs_sb_buf_ops; 463 xfs_buf_zero(bp, 0, BBTOB(bp->b_length)); 464 } else 465 error = ENOMEM; 466 } 467 468 if (error) { 469 xfs_warn(mp, 470 "error %d reading secondary superblock for ag %d", 471 error, agno); 472 break; 473 } 474 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS); 475 476 /* 477 * If we get an error writing out the alternate superblocks, 478 * just issue a warning and continue. The real work is 479 * already done and committed. 480 */ 481 error = xfs_bwrite(bp); 482 xfs_buf_relse(bp); 483 if (error) { 484 xfs_warn(mp, 485 "write error %d updating secondary superblock for ag %d", 486 error, agno); 487 break; /* no point in continuing */ 488 } 489 } 490 return error; 491 492 error0: 493 xfs_trans_cancel(tp, XFS_TRANS_ABORT); 494 return error; 495 } 496 497 static int 498 xfs_growfs_log_private( 499 xfs_mount_t *mp, /* mount point for filesystem */ 500 xfs_growfs_log_t *in) /* growfs log input struct */ 501 { 502 xfs_extlen_t nb; 503 504 nb = in->newblocks; 505 if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES)) 506 return XFS_ERROR(EINVAL); 507 if (nb == mp->m_sb.sb_logblocks && 508 in->isint == (mp->m_sb.sb_logstart != 0)) 509 return XFS_ERROR(EINVAL); 510 /* 511 * Moving the log is hard, need new interfaces to sync 512 * the log first, hold off all activity while moving it. 513 * Can have shorter or longer log in the same space, 514 * or transform internal to external log or vice versa. 515 */ 516 return XFS_ERROR(ENOSYS); 517 } 518 519 /* 520 * protected versions of growfs function acquire and release locks on the mount 521 * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG, 522 * XFS_IOC_FSGROWFSRT 523 */ 524 525 526 int 527 xfs_growfs_data( 528 xfs_mount_t *mp, 529 xfs_growfs_data_t *in) 530 { 531 int error; 532 533 if (!capable(CAP_SYS_ADMIN)) 534 return XFS_ERROR(EPERM); 535 if (!mutex_trylock(&mp->m_growlock)) 536 return XFS_ERROR(EWOULDBLOCK); 537 error = xfs_growfs_data_private(mp, in); 538 mutex_unlock(&mp->m_growlock); 539 return error; 540 } 541 542 int 543 xfs_growfs_log( 544 xfs_mount_t *mp, 545 xfs_growfs_log_t *in) 546 { 547 int error; 548 549 if (!capable(CAP_SYS_ADMIN)) 550 return XFS_ERROR(EPERM); 551 if (!mutex_trylock(&mp->m_growlock)) 552 return XFS_ERROR(EWOULDBLOCK); 553 error = xfs_growfs_log_private(mp, in); 554 mutex_unlock(&mp->m_growlock); 555 return error; 556 } 557 558 /* 559 * exported through ioctl XFS_IOC_FSCOUNTS 560 */ 561 562 int 563 xfs_fs_counts( 564 xfs_mount_t *mp, 565 xfs_fsop_counts_t *cnt) 566 { 567 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); 568 spin_lock(&mp->m_sb_lock); 569 cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); 570 cnt->freertx = mp->m_sb.sb_frextents; 571 cnt->freeino = mp->m_sb.sb_ifree; 572 cnt->allocino = mp->m_sb.sb_icount; 573 spin_unlock(&mp->m_sb_lock); 574 return 0; 575 } 576 577 /* 578 * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS 579 * 580 * xfs_reserve_blocks is called to set m_resblks 581 * in the in-core mount table. The number of unused reserved blocks 582 * is kept in m_resblks_avail. 583 * 584 * Reserve the requested number of blocks if available. Otherwise return 585 * as many as possible to satisfy the request. The actual number 586 * reserved are returned in outval 587 * 588 * A null inval pointer indicates that only the current reserved blocks 589 * available should be returned no settings are changed. 590 */ 591 592 int 593 xfs_reserve_blocks( 594 xfs_mount_t *mp, 595 __uint64_t *inval, 596 xfs_fsop_resblks_t *outval) 597 { 598 __int64_t lcounter, delta, fdblks_delta; 599 __uint64_t request; 600 601 /* If inval is null, report current values and return */ 602 if (inval == (__uint64_t *)NULL) { 603 if (!outval) 604 return EINVAL; 605 outval->resblks = mp->m_resblks; 606 outval->resblks_avail = mp->m_resblks_avail; 607 return 0; 608 } 609 610 request = *inval; 611 612 /* 613 * With per-cpu counters, this becomes an interesting 614 * problem. we needto work out if we are freeing or allocation 615 * blocks first, then we can do the modification as necessary. 616 * 617 * We do this under the m_sb_lock so that if we are near 618 * ENOSPC, we will hold out any changes while we work out 619 * what to do. This means that the amount of free space can 620 * change while we do this, so we need to retry if we end up 621 * trying to reserve more space than is available. 622 * 623 * We also use the xfs_mod_incore_sb() interface so that we 624 * don't have to care about whether per cpu counter are 625 * enabled, disabled or even compiled in.... 626 */ 627 retry: 628 spin_lock(&mp->m_sb_lock); 629 xfs_icsb_sync_counters_locked(mp, 0); 630 631 /* 632 * If our previous reservation was larger than the current value, 633 * then move any unused blocks back to the free pool. 634 */ 635 fdblks_delta = 0; 636 if (mp->m_resblks > request) { 637 lcounter = mp->m_resblks_avail - request; 638 if (lcounter > 0) { /* release unused blocks */ 639 fdblks_delta = lcounter; 640 mp->m_resblks_avail -= lcounter; 641 } 642 mp->m_resblks = request; 643 } else { 644 __int64_t free; 645 646 free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); 647 if (!free) 648 goto out; /* ENOSPC and fdblks_delta = 0 */ 649 650 delta = request - mp->m_resblks; 651 lcounter = free - delta; 652 if (lcounter < 0) { 653 /* We can't satisfy the request, just get what we can */ 654 mp->m_resblks += free; 655 mp->m_resblks_avail += free; 656 fdblks_delta = -free; 657 } else { 658 fdblks_delta = -delta; 659 mp->m_resblks = request; 660 mp->m_resblks_avail += delta; 661 } 662 } 663 out: 664 if (outval) { 665 outval->resblks = mp->m_resblks; 666 outval->resblks_avail = mp->m_resblks_avail; 667 } 668 spin_unlock(&mp->m_sb_lock); 669 670 if (fdblks_delta) { 671 /* 672 * If we are putting blocks back here, m_resblks_avail is 673 * already at its max so this will put it in the free pool. 674 * 675 * If we need space, we'll either succeed in getting it 676 * from the free block count or we'll get an enospc. If 677 * we get a ENOSPC, it means things changed while we were 678 * calculating fdblks_delta and so we should try again to 679 * see if there is anything left to reserve. 680 * 681 * Don't set the reserved flag here - we don't want to reserve 682 * the extra reserve blocks from the reserve..... 683 */ 684 int error; 685 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, 686 fdblks_delta, 0); 687 if (error == ENOSPC) 688 goto retry; 689 } 690 return 0; 691 } 692 693 /* 694 * Dump a transaction into the log that contains no real change. This is needed 695 * to be able to make the log dirty or stamp the current tail LSN into the log 696 * during the covering operation. 697 * 698 * We cannot use an inode here for this - that will push dirty state back up 699 * into the VFS and then periodic inode flushing will prevent log covering from 700 * making progress. Hence we log a field in the superblock instead and use a 701 * synchronous transaction to ensure the superblock is immediately unpinned 702 * and can be written back. 703 */ 704 int 705 xfs_fs_log_dummy( 706 xfs_mount_t *mp) 707 { 708 xfs_trans_t *tp; 709 int error; 710 711 tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP); 712 error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0, 713 XFS_DEFAULT_LOG_COUNT); 714 if (error) { 715 xfs_trans_cancel(tp, 0); 716 return error; 717 } 718 719 /* log the UUID because it is an unchanging field */ 720 xfs_mod_sb(tp, XFS_SB_UUID); 721 xfs_trans_set_sync(tp); 722 return xfs_trans_commit(tp, 0); 723 } 724 725 int 726 xfs_fs_goingdown( 727 xfs_mount_t *mp, 728 __uint32_t inflags) 729 { 730 switch (inflags) { 731 case XFS_FSOP_GOING_FLAGS_DEFAULT: { 732 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev); 733 734 if (sb && !IS_ERR(sb)) { 735 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); 736 thaw_bdev(sb->s_bdev, sb); 737 } 738 739 break; 740 } 741 case XFS_FSOP_GOING_FLAGS_LOGFLUSH: 742 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); 743 break; 744 case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH: 745 xfs_force_shutdown(mp, 746 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR); 747 break; 748 default: 749 return XFS_ERROR(EINVAL); 750 } 751 752 return 0; 753 } 754 755 /* 756 * Force a shutdown of the filesystem instantly while keeping the filesystem 757 * consistent. We don't do an unmount here; just shutdown the shop, make sure 758 * that absolutely nothing persistent happens to this filesystem after this 759 * point. 760 */ 761 void 762 xfs_do_force_shutdown( 763 xfs_mount_t *mp, 764 int flags, 765 char *fname, 766 int lnnum) 767 { 768 int logerror; 769 770 logerror = flags & SHUTDOWN_LOG_IO_ERROR; 771 772 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { 773 xfs_notice(mp, 774 "%s(0x%x) called from line %d of file %s. Return address = 0x%p", 775 __func__, flags, lnnum, fname, __return_address); 776 } 777 /* 778 * No need to duplicate efforts. 779 */ 780 if (XFS_FORCED_SHUTDOWN(mp) && !logerror) 781 return; 782 783 /* 784 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't 785 * queue up anybody new on the log reservations, and wakes up 786 * everybody who's sleeping on log reservations to tell them 787 * the bad news. 788 */ 789 if (xfs_log_force_umount(mp, logerror)) 790 return; 791 792 if (flags & SHUTDOWN_CORRUPT_INCORE) { 793 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT, 794 "Corruption of in-memory data detected. Shutting down filesystem"); 795 if (XFS_ERRLEVEL_HIGH <= xfs_error_level) 796 xfs_stack_trace(); 797 } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { 798 if (logerror) { 799 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR, 800 "Log I/O Error Detected. Shutting down filesystem"); 801 } else if (flags & SHUTDOWN_DEVICE_REQ) { 802 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR, 803 "All device paths lost. Shutting down filesystem"); 804 } else if (!(flags & SHUTDOWN_REMOTE_REQ)) { 805 xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR, 806 "I/O Error Detected. Shutting down filesystem"); 807 } 808 } 809 if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { 810 xfs_alert(mp, 811 "Please umount the filesystem and rectify the problem(s)"); 812 } 813 } 814