fs/xfs/xfs_fsops.c

   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_bit.h"
  22 #include "xfs_inum.h"
  23 #include "xfs_log.h"
  24 #include "xfs_trans.h"
  25 #include "xfs_sb.h"
  26 #include "xfs_ag.h"
  27 #include "xfs_mount.h"
  28 #include "xfs_bmap_btree.h"
  29 #include "xfs_alloc_btree.h"
  30 #include "xfs_ialloc_btree.h"
  31 #include "xfs_dinode.h"
  32 #include "xfs_inode.h"
  33 #include "xfs_inode_item.h"
  34 #include "xfs_btree.h"
  35 #include "xfs_error.h"
  36 #include "xfs_alloc.h"
  37 #include "xfs_ialloc.h"
  38 #include "xfs_fsops.h"
  39 #include "xfs_itable.h"
  40 #include "xfs_trans_space.h"
  41 #include "xfs_rtalloc.h"
  42 #include "xfs_rw.h"
  43 #include "xfs_filestream.h"
  44 #include "xfs_trace.h"
  45
  46 /*
  47  * File system operations
  48  */
  49
  50 int
  51 xfs_fs_geometry(
  52         xfs_mount_t             *mp,
  53         xfs_fsop_geom_t         *geo,
  54         int                     new_version)
  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                 geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
 102                                 mp->m_sb.sb_logsectsize : BBSIZE;
 103                 geo->rtsectsize = mp->m_sb.sb_blocksize;
 104                 geo->dirblocksize = mp->m_dirblksize;
 105         }
 106         if (new_version >= 4) {
 107                 geo->flags |=
 108                         (xfs_sb_version_haslogv2(&mp->m_sb) ?
 109                                 XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
 110                 geo->logsunit = mp->m_sb.sb_logsunit;
 111         }
 112         return 0;
 113 }
 114
 115 static int
 116 xfs_growfs_data_private(
 117         xfs_mount_t             *mp,            /* mount point for filesystem */
 118         xfs_growfs_data_t       *in)            /* growfs data input struct */
 119 {
 120         xfs_agf_t               *agf;
 121         xfs_agi_t               *agi;
 122         xfs_agnumber_t          agno;
 123         xfs_extlen_t            agsize;
 124         xfs_extlen_t            tmpsize;
 125         xfs_alloc_rec_t         *arec;
 126         struct xfs_btree_block  *block;
 127         xfs_buf_t               *bp;
 128         int                     bucket;
 129         int                     dpct;
 130         int                     error;
 131         xfs_agnumber_t          nagcount;
 132         xfs_agnumber_t          nagimax = 0;
 133         xfs_rfsblock_t          nb, nb_mod;
 134         xfs_rfsblock_t          new;
 135         xfs_rfsblock_t          nfree;
 136         xfs_agnumber_t          oagcount;
 137         int                     pct;
 138         xfs_trans_t             *tp;
 139
 140         nb = in->newblocks;
 141         pct = in->imaxpct;
 142         if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
 143                 return XFS_ERROR(EINVAL);
 144         if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
 145                 return error;
 146         dpct = pct - mp->m_sb.sb_imax_pct;
 147         error = xfs_read_buf(mp, mp->m_ddev_targp,
 148                         XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
 149                         XFS_FSS_TO_BB(mp, 1), 0, &bp);
 150         if (error)
 151                 return error;
 152         ASSERT(bp);
 153         xfs_buf_relse(bp);
 154
 155         new = nb;       /* use new as a temporary here */
 156         nb_mod = do_div(new, mp->m_sb.sb_agblocks);
 157         nagcount = new + (nb_mod != 0);
 158         if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
 159                 nagcount--;
 160                 nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
 161                 if (nb < mp->m_sb.sb_dblocks)
 162                         return XFS_ERROR(EINVAL);
 163         }
 164         new = nb - mp->m_sb.sb_dblocks;
 165         oagcount = mp->m_sb.sb_agcount;
 166
 167         /* allocate the new per-ag structures */
 168         if (nagcount > oagcount) {
 169                 error = xfs_initialize_perag(mp, nagcount, &nagimax);
 170                 if (error)
 171                         return error;
 172         }
 173
 174         tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
 175         tp->t_flags |= XFS_TRANS_RESERVE;
 176         if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
 177                         XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
 178                 xfs_trans_cancel(tp, 0);
 179                 return error;
 180         }
 181
 182         /*
 183          * Write new AG headers to disk. Non-transactional, but written
 184          * synchronously so they are completed prior to the growfs transaction
 185          * being logged.
 186          */
 187         nfree = 0;
 188         for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
 189                 /*
 190                  * AG freelist header block
 191                  */
 192                 bp = xfs_buf_get(mp->m_ddev_targp,
 193                                  XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
 194                                  XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
 195                 agf = XFS_BUF_TO_AGF(bp);
 196                 memset(agf, 0, mp->m_sb.sb_sectsize);
 197                 agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
 198                 agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
 199                 agf->agf_seqno = cpu_to_be32(agno);
 200                 if (agno == nagcount - 1)
 201                         agsize =
 202                                 nb -
 203                                 (agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
 204                 else
 205                         agsize = mp->m_sb.sb_agblocks;
 206                 agf->agf_length = cpu_to_be32(agsize);
 207                 agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
 208                 agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
 209                 agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
 210                 agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
 211                 agf->agf_flfirst = 0;
 212                 agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
 213                 agf->agf_flcount = 0;
 214                 tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
 215                 agf->agf_freeblks = cpu_to_be32(tmpsize);
 216                 agf->agf_longest = cpu_to_be32(tmpsize);
 217                 error = xfs_bwrite(mp, bp);
 218                 if (error) {
 219                         goto error0;
 220                 }
 221                 /*
 222                  * AG inode header block
 223                  */
 224                 bp = xfs_buf_get(mp->m_ddev_targp,
 225                                  XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
 226                                  XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
 227                 agi = XFS_BUF_TO_AGI(bp);
 228                 memset(agi, 0, mp->m_sb.sb_sectsize);
 229                 agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
 230                 agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
 231                 agi->agi_seqno = cpu_to_be32(agno);
 232                 agi->agi_length = cpu_to_be32(agsize);
 233                 agi->agi_count = 0;
 234                 agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
 235                 agi->agi_level = cpu_to_be32(1);
 236                 agi->agi_freecount = 0;
 237                 agi->agi_newino = cpu_to_be32(NULLAGINO);
 238                 agi->agi_dirino = cpu_to_be32(NULLAGINO);
 239                 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
 240                         agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
 241                 error = xfs_bwrite(mp, bp);
 242                 if (error) {
 243                         goto error0;
 244                 }
 245                 /*
 246                  * BNO btree root block
 247                  */
 248                 bp = xfs_buf_get(mp->m_ddev_targp,
 249                                  XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
 250                                  BTOBB(mp->m_sb.sb_blocksize),
 251                                  XBF_LOCK | XBF_MAPPED);
 252                 block = XFS_BUF_TO_BLOCK(bp);
 253                 memset(block, 0, mp->m_sb.sb_blocksize);
 254                 block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
 255                 block->bb_level = 0;
 256                 block->bb_numrecs = cpu_to_be16(1);
 257                 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
 258                 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
 259                 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
 260                 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
 261                 arec->ar_blockcount = cpu_to_be32(
 262                         agsize - be32_to_cpu(arec->ar_startblock));
 263                 error = xfs_bwrite(mp, bp);
 264                 if (error) {
 265                         goto error0;
 266                 }
 267                 /*
 268                  * CNT btree root block
 269                  */
 270                 bp = xfs_buf_get(mp->m_ddev_targp,
 271                                  XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
 272                                  BTOBB(mp->m_sb.sb_blocksize),
 273                                  XBF_LOCK | XBF_MAPPED);
 274                 block = XFS_BUF_TO_BLOCK(bp);
 275                 memset(block, 0, mp->m_sb.sb_blocksize);
 276                 block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
 277                 block->bb_level = 0;
 278                 block->bb_numrecs = cpu_to_be16(1);
 279                 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
 280                 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
 281                 arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
 282                 arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
 283                 arec->ar_blockcount = cpu_to_be32(
 284                         agsize - be32_to_cpu(arec->ar_startblock));
 285                 nfree += be32_to_cpu(arec->ar_blockcount);
 286                 error = xfs_bwrite(mp, bp);
 287                 if (error) {
 288                         goto error0;
 289                 }
 290                 /*
 291                  * INO btree root block
 292                  */
 293                 bp = xfs_buf_get(mp->m_ddev_targp,
 294                                  XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
 295                                  BTOBB(mp->m_sb.sb_blocksize),
 296                                  XBF_LOCK | XBF_MAPPED);
 297                 block = XFS_BUF_TO_BLOCK(bp);
 298                 memset(block, 0, mp->m_sb.sb_blocksize);
 299                 block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
 300                 block->bb_level = 0;
 301                 block->bb_numrecs = 0;
 302                 block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
 303                 block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
 304                 error = xfs_bwrite(mp, bp);
 305                 if (error) {
 306                         goto error0;
 307                 }
 308         }
 309         xfs_trans_agblocks_delta(tp, nfree);
 310         /*
 311          * There are new blocks in the old last a.g.
 312          */
 313         if (new) {
 314                 /*
 315                  * Change the agi length.
 316                  */
 317                 error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
 318                 if (error) {
 319                         goto error0;
 320                 }
 321                 ASSERT(bp);
 322                 agi = XFS_BUF_TO_AGI(bp);
 323                 be32_add_cpu(&agi->agi_length, new);
 324                 ASSERT(nagcount == oagcount ||
 325                        be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
 326                 xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
 327                 /*
 328                  * Change agf length.
 329                  */
 330                 error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
 331                 if (error) {
 332                         goto error0;
 333                 }
 334                 ASSERT(bp);
 335                 agf = XFS_BUF_TO_AGF(bp);
 336                 be32_add_cpu(&agf->agf_length, new);
 337                 ASSERT(be32_to_cpu(agf->agf_length) ==
 338                        be32_to_cpu(agi->agi_length));
 339
 340                 xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
 341                 /*
 342                  * Free the new space.
 343                  */
 344                 error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
 345                         be32_to_cpu(agf->agf_length) - new), new);
 346                 if (error) {
 347                         goto error0;
 348                 }
 349         }
 350
 351         /*
 352          * Update changed superblock fields transactionally. These are not
 353          * seen by the rest of the world until the transaction commit applies
 354          * them atomically to the superblock.
 355          */
 356         if (nagcount > oagcount)
 357                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
 358         if (nb > mp->m_sb.sb_dblocks)
 359                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
 360                                  nb - mp->m_sb.sb_dblocks);
 361         if (nfree)
 362                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
 363         if (dpct)
 364                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
 365         error = xfs_trans_commit(tp, 0);
 366         if (error)
 367                 return error;
 368
 369         /* New allocation groups fully initialized, so update mount struct */
 370         if (nagimax)
 371                 mp->m_maxagi = nagimax;
 372         if (mp->m_sb.sb_imax_pct) {
 373                 __uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
 374                 do_div(icount, 100);
 375                 mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
 376         } else
 377                 mp->m_maxicount = 0;
 378
 379         /* update secondary superblocks. */
 380         for (agno = 1; agno < nagcount; agno++) {
 381                 error = xfs_read_buf(mp, mp->m_ddev_targp,
 382                                   XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
 383                                   XFS_FSS_TO_BB(mp, 1), 0, &bp);
 384                 if (error) {
 385                         xfs_fs_cmn_err(CE_WARN, mp,
 386                         "error %d reading secondary superblock for ag %d",
 387                                 error, agno);
 388                         break;
 389                 }
 390                 xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
 391                 /*
 392                  * If we get an error writing out the alternate superblocks,
 393                  * just issue a warning and continue.  The real work is
 394                  * already done and committed.
 395                  */
 396                 if (!(error = xfs_bwrite(mp, bp))) {
 397                         continue;
 398                 } else {
 399                         xfs_fs_cmn_err(CE_WARN, mp,
 400                 "write error %d updating secondary superblock for ag %d",
 401                                 error, agno);
 402                         break; /* no point in continuing */
 403                 }
 404         }
 405         return 0;
 406
 407  error0:
 408         xfs_trans_cancel(tp, XFS_TRANS_ABORT);
 409         return error;
 410 }
 411
 412 static int
 413 xfs_growfs_log_private(
 414         xfs_mount_t             *mp,    /* mount point for filesystem */
 415         xfs_growfs_log_t        *in)    /* growfs log input struct */
 416 {
 417         xfs_extlen_t            nb;
 418
 419         nb = in->newblocks;
 420         if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
 421                 return XFS_ERROR(EINVAL);
 422         if (nb == mp->m_sb.sb_logblocks &&
 423             in->isint == (mp->m_sb.sb_logstart != 0))
 424                 return XFS_ERROR(EINVAL);
 425         /*
 426          * Moving the log is hard, need new interfaces to sync
 427          * the log first, hold off all activity while moving it.
 428          * Can have shorter or longer log in the same space,
 429          * or transform internal to external log or vice versa.
 430          */
 431         return XFS_ERROR(ENOSYS);
 432 }
 433
 434 /*
 435  * protected versions of growfs function acquire and release locks on the mount
 436  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
 437  * XFS_IOC_FSGROWFSRT
 438  */
 439
 440
 441 int
 442 xfs_growfs_data(
 443         xfs_mount_t             *mp,
 444         xfs_growfs_data_t       *in)
 445 {
 446         int error;
 447
 448         if (!capable(CAP_SYS_ADMIN))
 449                 return XFS_ERROR(EPERM);
 450         if (!mutex_trylock(&mp->m_growlock))
 451                 return XFS_ERROR(EWOULDBLOCK);
 452         error = xfs_growfs_data_private(mp, in);
 453         mutex_unlock(&mp->m_growlock);
 454         return error;
 455 }
 456
 457 int
 458 xfs_growfs_log(
 459         xfs_mount_t             *mp,
 460         xfs_growfs_log_t        *in)
 461 {
 462         int error;
 463
 464         if (!capable(CAP_SYS_ADMIN))
 465                 return XFS_ERROR(EPERM);
 466         if (!mutex_trylock(&mp->m_growlock))
 467                 return XFS_ERROR(EWOULDBLOCK);
 468         error = xfs_growfs_log_private(mp, in);
 469         mutex_unlock(&mp->m_growlock);
 470         return error;
 471 }
 472
 473 /*
 474  * exported through ioctl XFS_IOC_FSCOUNTS
 475  */
 476
 477 int
 478 xfs_fs_counts(
 479         xfs_mount_t             *mp,
 480         xfs_fsop_counts_t       *cnt)
 481 {
 482         xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
 483         spin_lock(&mp->m_sb_lock);
 484         cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
 485         cnt->freertx = mp->m_sb.sb_frextents;
 486         cnt->freeino = mp->m_sb.sb_ifree;
 487         cnt->allocino = mp->m_sb.sb_icount;
 488         spin_unlock(&mp->m_sb_lock);
 489         return 0;
 490 }
 491
 492 /*
 493  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
 494  *
 495  * xfs_reserve_blocks is called to set m_resblks
 496  * in the in-core mount table. The number of unused reserved blocks
 497  * is kept in m_resblks_avail.
 498  *
 499  * Reserve the requested number of blocks if available. Otherwise return
 500  * as many as possible to satisfy the request. The actual number
 501  * reserved are returned in outval
 502  *
 503  * A null inval pointer indicates that only the current reserved blocks
 504  * available  should  be returned no settings are changed.
 505  */
 506
 507 int
 508 xfs_reserve_blocks(
 509         xfs_mount_t             *mp,
 510         __uint64_t              *inval,
 511         xfs_fsop_resblks_t      *outval)
 512 {
 513         __int64_t               lcounter, delta, fdblks_delta;
 514         __uint64_t              request;
 515
 516         /* If inval is null, report current values and return */
 517         if (inval == (__uint64_t *)NULL) {
 518                 if (!outval)
 519                         return EINVAL;
 520                 outval->resblks = mp->m_resblks;
 521                 outval->resblks_avail = mp->m_resblks_avail;
 522                 return 0;
 523         }
 524
 525         request = *inval;
 526
 527         /*
 528          * With per-cpu counters, this becomes an interesting
 529          * problem. we needto work out if we are freeing or allocation
 530          * blocks first, then we can do the modification as necessary.
 531          *
 532          * We do this under the m_sb_lock so that if we are near
 533          * ENOSPC, we will hold out any changes while we work out
 534          * what to do. This means that the amount of free space can
 535          * change while we do this, so we need to retry if we end up
 536          * trying to reserve more space than is available.
 537          *
 538          * We also use the xfs_mod_incore_sb() interface so that we
 539          * don't have to care about whether per cpu counter are
 540          * enabled, disabled or even compiled in....
 541          */
 542 retry:
 543         spin_lock(&mp->m_sb_lock);
 544         xfs_icsb_sync_counters_locked(mp, 0);
 545
 546         /*
 547          * If our previous reservation was larger than the current value,
 548          * then move any unused blocks back to the free pool.
 549          */
 550         fdblks_delta = 0;
 551         if (mp->m_resblks > request) {
 552                 lcounter = mp->m_resblks_avail - request;
 553                 if (lcounter  > 0) {            /* release unused blocks */
 554                         fdblks_delta = lcounter;
 555                         mp->m_resblks_avail -= lcounter;
 556                 }
 557                 mp->m_resblks = request;
 558         } else {
 559                 __int64_t       free;
 560
 561                 free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
 562                 if (!free)
 563                         goto out; /* ENOSPC and fdblks_delta = 0 */
 564
 565                 delta = request - mp->m_resblks;
 566                 lcounter = free - delta;
 567                 if (lcounter < 0) {
 568                         /* We can't satisfy the request, just get what we can */
 569                         mp->m_resblks += free;
 570                         mp->m_resblks_avail += free;
 571                         fdblks_delta = -free;
 572                 } else {
 573                         fdblks_delta = -delta;
 574                         mp->m_resblks = request;
 575                         mp->m_resblks_avail += delta;
 576                 }
 577         }
 578 out:
 579         if (outval) {
 580                 outval->resblks = mp->m_resblks;
 581                 outval->resblks_avail = mp->m_resblks_avail;
 582         }
 583         spin_unlock(&mp->m_sb_lock);
 584
 585         if (fdblks_delta) {
 586                 /*
 587                  * If we are putting blocks back here, m_resblks_avail is
 588                  * already at its max so this will put it in the free pool.
 589                  *
 590                  * If we need space, we'll either succeed in getting it
 591                  * from the free block count or we'll get an enospc. If
 592                  * we get a ENOSPC, it means things changed while we were
 593                  * calculating fdblks_delta and so we should try again to
 594                  * see if there is anything left to reserve.
 595                  *
 596                  * Don't set the reserved flag here - we don't want to reserve
 597                  * the extra reserve blocks from the reserve.....
 598                  */
 599                 int error;
 600                 error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0);
 601                 if (error == ENOSPC)
 602                         goto retry;
 603         }
 604         return 0;
 605 }
 606
 607 /*
 608  * Dump a transaction into the log that contains no real change. This is needed
 609  * to be able to make the log dirty or stamp the current tail LSN into the log
 610  * during the covering operation.
 611  *
 612  * We cannot use an inode here for this - that will push dirty state back up
 613  * into the VFS and then periodic inode flushing will prevent log covering from
 614  * making progress. Hence we log a field in the superblock instead.
 615  */
 616 int
 617 xfs_fs_log_dummy(
 618         xfs_mount_t     *mp,
 619         int             flags)
 620 {
 621         xfs_trans_t     *tp;
 622         int             error;
 623
 624         tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
 625         error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
 626                                         XFS_DEFAULT_LOG_COUNT);
 627         if (error) {
 628                 xfs_trans_cancel(tp, 0);
 629                 return error;
 630         }
 631
 632         /* log the UUID because it is an unchanging field */
 633         xfs_mod_sb(tp, XFS_SB_UUID);
 634         if (flags & SYNC_WAIT)
 635                 xfs_trans_set_sync(tp);
 636         return xfs_trans_commit(tp, 0);
 637 }
 638
 639 int
 640 xfs_fs_goingdown(
 641         xfs_mount_t     *mp,
 642         __uint32_t      inflags)
 643 {
 644         switch (inflags) {
 645         case XFS_FSOP_GOING_FLAGS_DEFAULT: {
 646                 struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
 647
 648                 if (sb && !IS_ERR(sb)) {
 649                         xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
 650                         thaw_bdev(sb->s_bdev, sb);
 651                 }
 652
 653                 break;
 654         }
 655         case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
 656                 xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
 657                 break;
 658         case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
 659                 xfs_force_shutdown(mp,
 660                                 SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
 661                 break;
 662         default:
 663                 return XFS_ERROR(EINVAL);
 664         }
 665
 666         return 0;
 667 }