fs/xfs/xfs_ialloc.c

   1 /*
   2  * Copyright (c) 2000-2002,2005 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_log.h"
  23 #include "xfs_inum.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_btree.h"
  34 #include "xfs_ialloc.h"
  35 #include "xfs_alloc.h"
  36 #include "xfs_rtalloc.h"
  37 #include "xfs_error.h"
  38 #include "xfs_bmap.h"
  39
  40
  41 /*
  42  * Allocation group level functions.
  43  */
  44 static inline int
  45 xfs_ialloc_cluster_alignment(
  46         xfs_alloc_arg_t *args)
  47 {
  48         if (xfs_sb_version_hasalign(&args->mp->m_sb) &&
  49             args->mp->m_sb.sb_inoalignmt >=
  50              XFS_B_TO_FSBT(args->mp, XFS_INODE_CLUSTER_SIZE(args->mp)))
  51                 return args->mp->m_sb.sb_inoalignmt;
  52         return 1;
  53 }
  54
  55 /*
  56  * Lookup a record by ino in the btree given by cur.
  57  */
  58 int                                     /* error */
  59 xfs_inobt_lookup(
  60         struct xfs_btree_cur    *cur,   /* btree cursor */
  61         xfs_agino_t             ino,    /* starting inode of chunk */
  62         xfs_lookup_t            dir,    /* <=, >=, == */
  63         int                     *stat)  /* success/failure */
  64 {
  65         cur->bc_rec.i.ir_startino = ino;
  66         cur->bc_rec.i.ir_freecount = 0;
  67         cur->bc_rec.i.ir_free = 0;
  68         return xfs_btree_lookup(cur, dir, stat);
  69 }
  70
  71 /*
  72  * Update the record referred to by cur to the value given.
  73  * This either works (return 0) or gets an EFSCORRUPTED error.
  74  */
  75 STATIC int                              /* error */
  76 xfs_inobt_update(
  77         struct xfs_btree_cur    *cur,   /* btree cursor */
  78         xfs_inobt_rec_incore_t  *irec)  /* btree record */
  79 {
  80         union xfs_btree_rec     rec;
  81
  82         rec.inobt.ir_startino = cpu_to_be32(irec->ir_startino);
  83         rec.inobt.ir_freecount = cpu_to_be32(irec->ir_freecount);
  84         rec.inobt.ir_free = cpu_to_be64(irec->ir_free);
  85         return xfs_btree_update(cur, &rec);
  86 }
  87
  88 /*
  89  * Get the data from the pointed-to record.
  90  */
  91 int                                     /* error */
  92 xfs_inobt_get_rec(
  93         struct xfs_btree_cur    *cur,   /* btree cursor */
  94         xfs_inobt_rec_incore_t  *irec,  /* btree record */
  95         int                     *stat)  /* output: success/failure */
  96 {
  97         union xfs_btree_rec     *rec;
  98         int                     error;
  99
 100         error = xfs_btree_get_rec(cur, &rec, stat);
 101         if (!error && *stat == 1) {
 102                 irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino);
 103                 irec->ir_freecount = be32_to_cpu(rec->inobt.ir_freecount);
 104                 irec->ir_free = be64_to_cpu(rec->inobt.ir_free);
 105         }
 106         return error;
 107 }
 108
 109 /*
 110  * Verify that the number of free inodes in the AGI is correct.
 111  */
 112 #ifdef DEBUG
 113 STATIC int
 114 xfs_check_agi_freecount(
 115         struct xfs_btree_cur    *cur,
 116         struct xfs_agi          *agi)
 117 {
 118         if (cur->bc_nlevels == 1) {
 119                 xfs_inobt_rec_incore_t rec;
 120                 int             freecount = 0;
 121                 int             error;
 122                 int             i;
 123
 124                 error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i);
 125                 if (error)
 126                         return error;
 127
 128                 do {
 129                         error = xfs_inobt_get_rec(cur, &rec, &i);
 130                         if (error)
 131                                 return error;
 132
 133                         if (i) {
 134                                 freecount += rec.ir_freecount;
 135                                 error = xfs_btree_increment(cur, 0, &i);
 136                                 if (error)
 137                                         return error;
 138                         }
 139                 } while (i == 1);
 140
 141                 if (!XFS_FORCED_SHUTDOWN(cur->bc_mp))
 142                         ASSERT(freecount == be32_to_cpu(agi->agi_freecount));
 143         }
 144         return 0;
 145 }
 146 #else
 147 #define xfs_check_agi_freecount(cur, agi)       0
 148 #endif
 149
 150 /*
 151  * Initialise a new set of inodes.
 152  */
 153 STATIC int
 154 xfs_ialloc_inode_init(
 155         struct xfs_mount        *mp,
 156         struct xfs_trans        *tp,
 157         xfs_agnumber_t          agno,
 158         xfs_agblock_t           agbno,
 159         xfs_agblock_t           length,
 160         unsigned int            gen)
 161 {
 162         struct xfs_buf          *fbuf;
 163         struct xfs_dinode       *free;
 164         int                     blks_per_cluster, nbufs, ninodes;
 165         int                     version;
 166         int                     i, j;
 167         xfs_daddr_t             d;
 168
 169         /*
 170          * Loop over the new block(s), filling in the inodes.
 171          * For small block sizes, manipulate the inodes in buffers
 172          * which are multiples of the blocks size.
 173          */
 174         if (mp->m_sb.sb_blocksize >= XFS_INODE_CLUSTER_SIZE(mp)) {
 175                 blks_per_cluster = 1;
 176                 nbufs = length;
 177                 ninodes = mp->m_sb.sb_inopblock;
 178         } else {
 179                 blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) /
 180                                    mp->m_sb.sb_blocksize;
 181                 nbufs = length / blks_per_cluster;
 182                 ninodes = blks_per_cluster * mp->m_sb.sb_inopblock;
 183         }
 184
 185         /*
 186          * Figure out what version number to use in the inodes we create.
 187          * If the superblock version has caught up to the one that supports
 188          * the new inode format, then use the new inode version.  Otherwise
 189          * use the old version so that old kernels will continue to be
 190          * able to use the file system.
 191          */
 192         if (xfs_sb_version_hasnlink(&mp->m_sb))
 193                 version = 2;
 194         else
 195                 version = 1;
 196
 197         for (j = 0; j < nbufs; j++) {
 198                 /*
 199                  * Get the block.
 200                  */
 201                 d = XFS_AGB_TO_DADDR(mp, agno, agbno + (j * blks_per_cluster));
 202                 fbuf = xfs_trans_get_buf(tp, mp->m_ddev_targp, d,
 203                                          mp->m_bsize * blks_per_cluster,
 204                                          XBF_LOCK);
 205                 if (!fbuf)
 206                         return ENOMEM;
 207                 /*
 208                  * Initialize all inodes in this buffer and then log them.
 209                  *
 210                  * XXX: It would be much better if we had just one transaction
 211                  *      to log a whole cluster of inodes instead of all the
 212                  *      individual transactions causing a lot of log traffic.
 213                  */
 214                 xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog);
 215                 for (i = 0; i < ninodes; i++) {
 216                         int     ioffset = i << mp->m_sb.sb_inodelog;
 217                         uint    isize = sizeof(struct xfs_dinode);
 218
 219                         free = xfs_make_iptr(mp, fbuf, i);
 220                         free->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
 221                         free->di_version = version;
 222                         free->di_gen = cpu_to_be32(gen);
 223                         free->di_next_unlinked = cpu_to_be32(NULLAGINO);
 224                         xfs_trans_log_buf(tp, fbuf, ioffset, ioffset + isize - 1);
 225                 }
 226                 xfs_trans_inode_alloc_buf(tp, fbuf);
 227         }
 228         return 0;
 229 }
 230
 231 /*
 232  * Allocate new inodes in the allocation group specified by agbp.
 233  * Return 0 for success, else error code.
 234  */
 235 STATIC int                              /* error code or 0 */
 236 xfs_ialloc_ag_alloc(
 237         xfs_trans_t     *tp,            /* transaction pointer */
 238         xfs_buf_t       *agbp,          /* alloc group buffer */
 239         int             *alloc)
 240 {
 241         xfs_agi_t       *agi;           /* allocation group header */
 242         xfs_alloc_arg_t args;           /* allocation argument structure */
 243         xfs_btree_cur_t *cur;           /* inode btree cursor */
 244         xfs_agnumber_t  agno;
 245         int             error;
 246         int             i;
 247         xfs_agino_t     newino;         /* new first inode's number */
 248         xfs_agino_t     newlen;         /* new number of inodes */
 249         xfs_agino_t     thisino;        /* current inode number, for loop */
 250         int             isaligned = 0;  /* inode allocation at stripe unit */
 251                                         /* boundary */
 252         struct xfs_perag *pag;
 253
 254         args.tp = tp;
 255         args.mp = tp->t_mountp;
 256
 257         /*
 258          * Locking will ensure that we don't have two callers in here
 259          * at one time.
 260          */
 261         newlen = XFS_IALLOC_INODES(args.mp);
 262         if (args.mp->m_maxicount &&
 263             args.mp->m_sb.sb_icount + newlen > args.mp->m_maxicount)
 264                 return XFS_ERROR(ENOSPC);
 265         args.minlen = args.maxlen = XFS_IALLOC_BLOCKS(args.mp);
 266         /*
 267          * First try to allocate inodes contiguous with the last-allocated
 268          * chunk of inodes.  If the filesystem is striped, this will fill
 269          * an entire stripe unit with inodes.
 270          */
 271         agi = XFS_BUF_TO_AGI(agbp);
 272         newino = be32_to_cpu(agi->agi_newino);
 273         agno = be32_to_cpu(agi->agi_seqno);
 274         args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) +
 275                         XFS_IALLOC_BLOCKS(args.mp);
 276         if (likely(newino != NULLAGINO &&
 277                   (args.agbno < be32_to_cpu(agi->agi_length)))) {
 278                 args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
 279                 args.type = XFS_ALLOCTYPE_THIS_BNO;
 280                 args.mod = args.total = args.wasdel = args.isfl =
 281                         args.userdata = args.minalignslop = 0;
 282                 args.prod = 1;
 283
 284                 /*
 285                  * We need to take into account alignment here to ensure that
 286                  * we don't modify the free list if we fail to have an exact
 287                  * block. If we don't have an exact match, and every oher
 288                  * attempt allocation attempt fails, we'll end up cancelling
 289                  * a dirty transaction and shutting down.
 290                  *
 291                  * For an exact allocation, alignment must be 1,
 292                  * however we need to take cluster alignment into account when
 293                  * fixing up the freelist. Use the minalignslop field to
 294                  * indicate that extra blocks might be required for alignment,
 295                  * but not to use them in the actual exact allocation.
 296                  */
 297                 args.alignment = 1;
 298                 args.minalignslop = xfs_ialloc_cluster_alignment(&args) - 1;
 299
 300                 /* Allow space for the inode btree to split. */
 301                 args.minleft = args.mp->m_in_maxlevels - 1;
 302                 if ((error = xfs_alloc_vextent(&args)))
 303                         return error;
 304         } else
 305                 args.fsbno = NULLFSBLOCK;
 306
 307         if (unlikely(args.fsbno == NULLFSBLOCK)) {
 308                 /*
 309                  * Set the alignment for the allocation.
 310                  * If stripe alignment is turned on then align at stripe unit
 311                  * boundary.
 312                  * If the cluster size is smaller than a filesystem block
 313                  * then we're doing I/O for inodes in filesystem block size
 314                  * pieces, so don't need alignment anyway.
 315                  */
 316                 isaligned = 0;
 317                 if (args.mp->m_sinoalign) {
 318                         ASSERT(!(args.mp->m_flags & XFS_MOUNT_NOALIGN));
 319                         args.alignment = args.mp->m_dalign;
 320                         isaligned = 1;
 321                 } else
 322                         args.alignment = xfs_ialloc_cluster_alignment(&args);
 323                 /*
 324                  * Need to figure out where to allocate the inode blocks.
 325                  * Ideally they should be spaced out through the a.g.
 326                  * For now, just allocate blocks up front.
 327                  */
 328                 args.agbno = be32_to_cpu(agi->agi_root);
 329                 args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
 330                 /*
 331                  * Allocate a fixed-size extent of inodes.
 332                  */
 333                 args.type = XFS_ALLOCTYPE_NEAR_BNO;
 334                 args.mod = args.total = args.wasdel = args.isfl =
 335                         args.userdata = args.minalignslop = 0;
 336                 args.prod = 1;
 337                 /*
 338                  * Allow space for the inode btree to split.
 339                  */
 340                 args.minleft = args.mp->m_in_maxlevels - 1;
 341                 if ((error = xfs_alloc_vextent(&args)))
 342                         return error;
 343         }
 344
 345         /*
 346          * If stripe alignment is turned on, then try again with cluster
 347          * alignment.
 348          */
 349         if (isaligned && args.fsbno == NULLFSBLOCK) {
 350                 args.type = XFS_ALLOCTYPE_NEAR_BNO;
 351                 args.agbno = be32_to_cpu(agi->agi_root);
 352                 args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
 353                 args.alignment = xfs_ialloc_cluster_alignment(&args);
 354                 if ((error = xfs_alloc_vextent(&args)))
 355                         return error;
 356         }
 357
 358         if (args.fsbno == NULLFSBLOCK) {
 359                 *alloc = 0;
 360                 return 0;
 361         }
 362         ASSERT(args.len == args.minlen);
 363
 364         /*
 365          * Stamp and write the inode buffers.
 366          *
 367          * Seed the new inode cluster with a random generation number. This
 368          * prevents short-term reuse of generation numbers if a chunk is
 369          * freed and then immediately reallocated. We use random numbers
 370          * rather than a linear progression to prevent the next generation
 371          * number from being easily guessable.
 372          */
 373         error = xfs_ialloc_inode_init(args.mp, tp, agno, args.agbno,
 374                         args.len, random32());
 375
 376         if (error)
 377                 return error;
 378         /*
 379          * Convert the results.
 380          */
 381         newino = XFS_OFFBNO_TO_AGINO(args.mp, args.agbno, 0);
 382         be32_add_cpu(&agi->agi_count, newlen);
 383         be32_add_cpu(&agi->agi_freecount, newlen);
 384         pag = xfs_perag_get(args.mp, agno);
 385         pag->pagi_freecount += newlen;
 386         xfs_perag_put(pag);
 387         agi->agi_newino = cpu_to_be32(newino);
 388
 389         /*
 390          * Insert records describing the new inode chunk into the btree.
 391          */
 392         cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno);
 393         for (thisino = newino;
 394              thisino < newino + newlen;
 395              thisino += XFS_INODES_PER_CHUNK) {
 396                 cur->bc_rec.i.ir_startino = thisino;
 397                 cur->bc_rec.i.ir_freecount = XFS_INODES_PER_CHUNK;
 398                 cur->bc_rec.i.ir_free = XFS_INOBT_ALL_FREE;
 399                 error = xfs_btree_lookup(cur, XFS_LOOKUP_EQ, &i);
 400                 if (error) {
 401                         xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 402                         return error;
 403                 }
 404                 ASSERT(i == 0);
 405                 error = xfs_btree_insert(cur, &i);
 406                 if (error) {
 407                         xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 408                         return error;
 409                 }
 410                 ASSERT(i == 1);
 411         }
 412         xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
 413         /*
 414          * Log allocation group header fields
 415          */
 416         xfs_ialloc_log_agi(tp, agbp,
 417                 XFS_AGI_COUNT | XFS_AGI_FREECOUNT | XFS_AGI_NEWINO);
 418         /*
 419          * Modify/log superblock values for inode count and inode free count.
 420          */
 421         xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, (long)newlen);
 422         xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, (long)newlen);
 423         *alloc = 1;
 424         return 0;
 425 }
 426
 427 STATIC xfs_agnumber_t
 428 xfs_ialloc_next_ag(
 429         xfs_mount_t     *mp)
 430 {
 431         xfs_agnumber_t  agno;
 432
 433         spin_lock(&mp->m_agirotor_lock);
 434         agno = mp->m_agirotor;
 435         if (++mp->m_agirotor == mp->m_maxagi)
 436                 mp->m_agirotor = 0;
 437         spin_unlock(&mp->m_agirotor_lock);
 438
 439         return agno;
 440 }
 441
 442 /*
 443  * Select an allocation group to look for a free inode in, based on the parent
 444  * inode and then mode.  Return the allocation group buffer.
 445  */
 446 STATIC xfs_buf_t *                      /* allocation group buffer */
 447 xfs_ialloc_ag_select(
 448         xfs_trans_t     *tp,            /* transaction pointer */
 449         xfs_ino_t       parent,         /* parent directory inode number */
 450         mode_t          mode,           /* bits set to indicate file type */
 451         int             okalloc)        /* ok to allocate more space */
 452 {
 453         xfs_buf_t       *agbp;          /* allocation group header buffer */
 454         xfs_agnumber_t  agcount;        /* number of ag's in the filesystem */
 455         xfs_agnumber_t  agno;           /* current ag number */
 456         int             flags;          /* alloc buffer locking flags */
 457         xfs_extlen_t    ineed;          /* blocks needed for inode allocation */
 458         xfs_extlen_t    longest = 0;    /* longest extent available */
 459         xfs_mount_t     *mp;            /* mount point structure */
 460         int             needspace;      /* file mode implies space allocated */
 461         xfs_perag_t     *pag;           /* per allocation group data */
 462         xfs_agnumber_t  pagno;          /* parent (starting) ag number */
 463
 464         /*
 465          * Files of these types need at least one block if length > 0
 466          * (and they won't fit in the inode, but that's hard to figure out).
 467          */
 468         needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode);
 469         mp = tp->t_mountp;
 470         agcount = mp->m_maxagi;
 471         if (S_ISDIR(mode))
 472                 pagno = xfs_ialloc_next_ag(mp);
 473         else {
 474                 pagno = XFS_INO_TO_AGNO(mp, parent);
 475                 if (pagno >= agcount)
 476                         pagno = 0;
 477         }
 478         ASSERT(pagno < agcount);
 479         /*
 480          * Loop through allocation groups, looking for one with a little
 481          * free space in it.  Note we don't look for free inodes, exactly.
 482          * Instead, we include whether there is a need to allocate inodes
 483          * to mean that blocks must be allocated for them,
 484          * if none are currently free.
 485          */
 486         agno = pagno;
 487         flags = XFS_ALLOC_FLAG_TRYLOCK;
 488         for (;;) {
 489                 pag = xfs_perag_get(mp, agno);
 490                 if (!pag->pagi_init) {
 491                         if (xfs_ialloc_read_agi(mp, tp, agno, &agbp)) {
 492                                 agbp = NULL;
 493                                 goto nextag;
 494                         }
 495                 } else
 496                         agbp = NULL;
 497
 498                 if (!pag->pagi_inodeok) {
 499                         xfs_ialloc_next_ag(mp);
 500                         goto unlock_nextag;
 501                 }
 502
 503                 /*
 504                  * Is there enough free space for the file plus a block
 505                  * of inodes (if we need to allocate some)?
 506                  */
 507                 ineed = pag->pagi_freecount ? 0 : XFS_IALLOC_BLOCKS(mp);
 508                 if (ineed && !pag->pagf_init) {
 509                         if (agbp == NULL &&
 510                             xfs_ialloc_read_agi(mp, tp, agno, &agbp)) {
 511                                 agbp = NULL;
 512                                 goto nextag;
 513                         }
 514                         (void)xfs_alloc_pagf_init(mp, tp, agno, flags);
 515                 }
 516                 if (!ineed || pag->pagf_init) {
 517                         if (ineed && !(longest = pag->pagf_longest))
 518                                 longest = pag->pagf_flcount > 0;
 519                         if (!ineed ||
 520                             (pag->pagf_freeblks >= needspace + ineed &&
 521                              longest >= ineed &&
 522                              okalloc)) {
 523                                 if (agbp == NULL &&
 524                                     xfs_ialloc_read_agi(mp, tp, agno, &agbp)) {
 525                                         agbp = NULL;
 526                                         goto nextag;
 527                                 }
 528                                 xfs_perag_put(pag);
 529                                 return agbp;
 530                         }
 531                 }
 532 unlock_nextag:
 533                 if (agbp)
 534                         xfs_trans_brelse(tp, agbp);
 535 nextag:
 536                 xfs_perag_put(pag);
 537                 /*
 538                  * No point in iterating over the rest, if we're shutting
 539                  * down.
 540                  */
 541                 if (XFS_FORCED_SHUTDOWN(mp))
 542                         return NULL;
 543                 agno++;
 544                 if (agno >= agcount)
 545                         agno = 0;
 546                 if (agno == pagno) {
 547                         if (flags == 0)
 548                                 return NULL;
 549                         flags = 0;
 550                 }
 551         }
 552 }
 553
 554 /*
 555  * Try to retrieve the next record to the left/right from the current one.
 556  */
 557 STATIC int
 558 xfs_ialloc_next_rec(
 559         struct xfs_btree_cur    *cur,
 560         xfs_inobt_rec_incore_t  *rec,
 561         int                     *done,
 562         int                     left)
 563 {
 564         int                     error;
 565         int                     i;
 566
 567         if (left)
 568                 error = xfs_btree_decrement(cur, 0, &i);
 569         else
 570                 error = xfs_btree_increment(cur, 0, &i);
 571
 572         if (error)
 573                 return error;
 574         *done = !i;
 575         if (i) {
 576                 error = xfs_inobt_get_rec(cur, rec, &i);
 577                 if (error)
 578                         return error;
 579                 XFS_WANT_CORRUPTED_RETURN(i == 1);
 580         }
 581
 582         return 0;
 583 }
 584
 585 STATIC int
 586 xfs_ialloc_get_rec(
 587         struct xfs_btree_cur    *cur,
 588         xfs_agino_t             agino,
 589         xfs_inobt_rec_incore_t  *rec,
 590         int                     *done,
 591         int                     left)
 592 {
 593         int                     error;
 594         int                     i;
 595
 596         error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_EQ, &i);
 597         if (error)
 598                 return error;
 599         *done = !i;
 600         if (i) {
 601                 error = xfs_inobt_get_rec(cur, rec, &i);
 602                 if (error)
 603                         return error;
 604                 XFS_WANT_CORRUPTED_RETURN(i == 1);
 605         }
 606
 607         return 0;
 608 }
 609
 610 /*
 611  * Visible inode allocation functions.
 612  */
 613
 614 /*
 615  * Allocate an inode on disk.
 616  * Mode is used to tell whether the new inode will need space, and whether
 617  * it is a directory.
 618  *
 619  * The arguments IO_agbp and alloc_done are defined to work within
 620  * the constraint of one allocation per transaction.
 621  * xfs_dialloc() is designed to be called twice if it has to do an
 622  * allocation to make more free inodes.  On the first call,
 623  * IO_agbp should be set to NULL. If an inode is available,
 624  * i.e., xfs_dialloc() did not need to do an allocation, an inode
 625  * number is returned.  In this case, IO_agbp would be set to the
 626  * current ag_buf and alloc_done set to false.
 627  * If an allocation needed to be done, xfs_dialloc would return
 628  * the current ag_buf in IO_agbp and set alloc_done to true.
 629  * The caller should then commit the current transaction, allocate a new
 630  * transaction, and call xfs_dialloc() again, passing in the previous
 631  * value of IO_agbp.  IO_agbp should be held across the transactions.
 632  * Since the agbp is locked across the two calls, the second call is
 633  * guaranteed to have a free inode available.
 634  *
 635  * Once we successfully pick an inode its number is returned and the
 636  * on-disk data structures are updated.  The inode itself is not read
 637  * in, since doing so would break ordering constraints with xfs_reclaim.
 638  */
 639 int
 640 xfs_dialloc(
 641         xfs_trans_t     *tp,            /* transaction pointer */
 642         xfs_ino_t       parent,         /* parent inode (directory) */
 643         mode_t          mode,           /* mode bits for new inode */
 644         int             okalloc,        /* ok to allocate more space */
 645         xfs_buf_t       **IO_agbp,      /* in/out ag header's buffer */
 646         boolean_t       *alloc_done,    /* true if we needed to replenish
 647                                            inode freelist */
 648         xfs_ino_t       *inop)          /* inode number allocated */
 649 {
 650         xfs_agnumber_t  agcount;        /* number of allocation groups */
 651         xfs_buf_t       *agbp;          /* allocation group header's buffer */
 652         xfs_agnumber_t  agno;           /* allocation group number */
 653         xfs_agi_t       *agi;           /* allocation group header structure */
 654         xfs_btree_cur_t *cur;           /* inode allocation btree cursor */
 655         int             error;          /* error return value */
 656         int             i;              /* result code */
 657         int             ialloced;       /* inode allocation status */
 658         int             noroom = 0;     /* no space for inode blk allocation */
 659         xfs_ino_t       ino;            /* fs-relative inode to be returned */
 660         /* REFERENCED */
 661         int             j;              /* result code */
 662         xfs_mount_t     *mp;            /* file system mount structure */
 663         int             offset;         /* index of inode in chunk */
 664         xfs_agino_t     pagino;         /* parent's AG relative inode # */
 665         xfs_agnumber_t  pagno;          /* parent's AG number */
 666         xfs_inobt_rec_incore_t rec;     /* inode allocation record */
 667         xfs_agnumber_t  tagno;          /* testing allocation group number */
 668         xfs_btree_cur_t *tcur;          /* temp cursor */
 669         xfs_inobt_rec_incore_t trec;    /* temp inode allocation record */
 670         struct xfs_perag *pag;
 671
 672
 673         if (*IO_agbp == NULL) {
 674                 /*
 675                  * We do not have an agbp, so select an initial allocation
 676                  * group for inode allocation.
 677                  */
 678                 agbp = xfs_ialloc_ag_select(tp, parent, mode, okalloc);
 679                 /*
 680                  * Couldn't find an allocation group satisfying the
 681                  * criteria, give up.
 682                  */
 683                 if (!agbp) {
 684                         *inop = NULLFSINO;
 685                         return 0;
 686                 }
 687                 agi = XFS_BUF_TO_AGI(agbp);
 688                 ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC));
 689         } else {
 690                 /*
 691                  * Continue where we left off before.  In this case, we
 692                  * know that the allocation group has free inodes.
 693                  */
 694                 agbp = *IO_agbp;
 695                 agi = XFS_BUF_TO_AGI(agbp);
 696                 ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC));
 697                 ASSERT(be32_to_cpu(agi->agi_freecount) > 0);
 698         }
 699         mp = tp->t_mountp;
 700         agcount = mp->m_sb.sb_agcount;
 701         agno = be32_to_cpu(agi->agi_seqno);
 702         tagno = agno;
 703         pagno = XFS_INO_TO_AGNO(mp, parent);
 704         pagino = XFS_INO_TO_AGINO(mp, parent);
 705
 706         /*
 707          * If we have already hit the ceiling of inode blocks then clear
 708          * okalloc so we scan all available agi structures for a free
 709          * inode.
 710          */
 711
 712         if (mp->m_maxicount &&
 713             mp->m_sb.sb_icount + XFS_IALLOC_INODES(mp) > mp->m_maxicount) {
 714                 noroom = 1;
 715                 okalloc = 0;
 716         }
 717
 718         /*
 719          * Loop until we find an allocation group that either has free inodes
 720          * or in which we can allocate some inodes.  Iterate through the
 721          * allocation groups upward, wrapping at the end.
 722          */
 723         *alloc_done = B_FALSE;
 724         while (!agi->agi_freecount) {
 725                 /*
 726                  * Don't do anything if we're not supposed to allocate
 727                  * any blocks, just go on to the next ag.
 728                  */
 729                 if (okalloc) {
 730                         /*
 731                          * Try to allocate some new inodes in the allocation
 732                          * group.
 733                          */
 734                         if ((error = xfs_ialloc_ag_alloc(tp, agbp, &ialloced))) {
 735                                 xfs_trans_brelse(tp, agbp);
 736                                 if (error == ENOSPC) {
 737                                         *inop = NULLFSINO;
 738                                         return 0;
 739                                 } else
 740                                         return error;
 741                         }
 742                         if (ialloced) {
 743                                 /*
 744                                  * We successfully allocated some inodes, return
 745                                  * the current context to the caller so that it
 746                                  * can commit the current transaction and call
 747                                  * us again where we left off.
 748                                  */
 749                                 ASSERT(be32_to_cpu(agi->agi_freecount) > 0);
 750                                 *alloc_done = B_TRUE;
 751                                 *IO_agbp = agbp;
 752                                 *inop = NULLFSINO;
 753                                 return 0;
 754                         }
 755                 }
 756                 /*
 757                  * If it failed, give up on this ag.
 758                  */
 759                 xfs_trans_brelse(tp, agbp);
 760                 /*
 761                  * Go on to the next ag: get its ag header.
 762                  */
 763 nextag:
 764                 if (++tagno == agcount)
 765                         tagno = 0;
 766                 if (tagno == agno) {
 767                         *inop = NULLFSINO;
 768                         return noroom ? ENOSPC : 0;
 769                 }
 770                 pag = xfs_perag_get(mp, tagno);
 771                 if (pag->pagi_inodeok == 0) {
 772                         xfs_perag_put(pag);
 773                         goto nextag;
 774                 }
 775                 error = xfs_ialloc_read_agi(mp, tp, tagno, &agbp);
 776                 xfs_perag_put(pag);
 777                 if (error)
 778                         goto nextag;
 779                 agi = XFS_BUF_TO_AGI(agbp);
 780                 ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC));
 781         }
 782         /*
 783          * Here with an allocation group that has a free inode.
 784          * Reset agno since we may have chosen a new ag in the
 785          * loop above.
 786          */
 787         agno = tagno;
 788         *IO_agbp = NULL;
 789         pag = xfs_perag_get(mp, agno);
 790
 791  restart_pagno:
 792         cur = xfs_inobt_init_cursor(mp, tp, agbp, be32_to_cpu(agi->agi_seqno));
 793         /*
 794          * If pagino is 0 (this is the root inode allocation) use newino.
 795          * This must work because we've just allocated some.
 796          */
 797         if (!pagino)
 798                 pagino = be32_to_cpu(agi->agi_newino);
 799
 800         error = xfs_check_agi_freecount(cur, agi);
 801         if (error)
 802                 goto error0;
 803
 804         /*
 805          * If in the same AG as the parent, try to get near the parent.
 806          */
 807         if (pagno == agno) {
 808                 int             doneleft;       /* done, to the left */
 809                 int             doneright;      /* done, to the right */
 810                 int             searchdistance = 10;
 811
 812                 error = xfs_inobt_lookup(cur, pagino, XFS_LOOKUP_LE, &i);
 813                 if (error)
 814                         goto error0;
 815                 XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 816
 817                 error = xfs_inobt_get_rec(cur, &rec, &j);
 818                 if (error)
 819                         goto error0;
 820                 XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 821
 822                 if (rec.ir_freecount > 0) {
 823                         /*
 824                          * Found a free inode in the same chunk
 825                          * as the parent, done.
 826                          */
 827                         goto alloc_inode;
 828                 }
 829
 830
 831                 /*
 832                  * In the same AG as parent, but parent's chunk is full.
 833                  */
 834
 835                 /* duplicate the cursor, search left & right simultaneously */
 836                 error = xfs_btree_dup_cursor(cur, &tcur);
 837                 if (error)
 838                         goto error0;
 839
 840                 /*
 841                  * Skip to last blocks looked up if same parent inode.
 842                  */
 843                 if (pagino != NULLAGINO &&
 844                     pag->pagl_pagino == pagino &&
 845                     pag->pagl_leftrec != NULLAGINO &&
 846                     pag->pagl_rightrec != NULLAGINO) {
 847                         error = xfs_ialloc_get_rec(tcur, pag->pagl_leftrec,
 848                                                    &trec, &doneleft, 1);
 849                         if (error)
 850                                 goto error1;
 851
 852                         error = xfs_ialloc_get_rec(cur, pag->pagl_rightrec,
 853                                                    &rec, &doneright, 0);
 854                         if (error)
 855                                 goto error1;
 856                 } else {
 857                         /* search left with tcur, back up 1 record */
 858                         error = xfs_ialloc_next_rec(tcur, &trec, &doneleft, 1);
 859                         if (error)
 860                                 goto error1;
 861
 862                         /* search right with cur, go forward 1 record. */
 863                         error = xfs_ialloc_next_rec(cur, &rec, &doneright, 0);
 864                         if (error)
 865                                 goto error1;
 866                 }
 867
 868                 /*
 869                  * Loop until we find an inode chunk with a free inode.
 870                  */
 871                 while (!doneleft || !doneright) {
 872                         int     useleft;  /* using left inode chunk this time */
 873
 874                         if (!--searchdistance) {
 875                                 /*
 876                                  * Not in range - save last search
 877                                  * location and allocate a new inode
 878                                  */
 879                                 xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
 880                                 pag->pagl_leftrec = trec.ir_startino;
 881                                 pag->pagl_rightrec = rec.ir_startino;
 882                                 pag->pagl_pagino = pagino;
 883                                 goto newino;
 884                         }
 885
 886                         /* figure out the closer block if both are valid. */
 887                         if (!doneleft && !doneright) {
 888                                 useleft = pagino -
 889                                  (trec.ir_startino + XFS_INODES_PER_CHUNK - 1) <
 890                                   rec.ir_startino - pagino;
 891                         } else {
 892                                 useleft = !doneleft;
 893                         }
 894
 895                         /* free inodes to the left? */
 896                         if (useleft && trec.ir_freecount) {
 897                                 rec = trec;
 898                                 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
 899                                 cur = tcur;
 900
 901                                 pag->pagl_leftrec = trec.ir_startino;
 902                                 pag->pagl_rightrec = rec.ir_startino;
 903                                 pag->pagl_pagino = pagino;
 904                                 goto alloc_inode;
 905                         }
 906
 907                         /* free inodes to the right? */
 908                         if (!useleft && rec.ir_freecount) {
 909                                 xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
 910
 911                                 pag->pagl_leftrec = trec.ir_startino;
 912                                 pag->pagl_rightrec = rec.ir_startino;
 913                                 pag->pagl_pagino = pagino;
 914                                 goto alloc_inode;
 915                         }
 916
 917                         /* get next record to check */
 918                         if (useleft) {
 919                                 error = xfs_ialloc_next_rec(tcur, &trec,
 920                                                                  &doneleft, 1);
 921                         } else {
 922                                 error = xfs_ialloc_next_rec(cur, &rec,
 923                                                                  &doneright, 0);
 924                         }
 925                         if (error)
 926                                 goto error1;
 927                 }
 928
 929                 /*
 930                  * We've reached the end of the btree. because
 931                  * we are only searching a small chunk of the
 932                  * btree each search, there is obviously free
 933                  * inodes closer to the parent inode than we
 934                  * are now. restart the search again.
 935                  */
 936                 pag->pagl_pagino = NULLAGINO;
 937                 pag->pagl_leftrec = NULLAGINO;
 938                 pag->pagl_rightrec = NULLAGINO;
 939                 xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
 940                 xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
 941                 goto restart_pagno;
 942         }
 943
 944         /*
 945          * In a different AG from the parent.
 946          * See if the most recently allocated block has any free.
 947          */
 948 newino:
 949         if (agi->agi_newino != cpu_to_be32(NULLAGINO)) {
 950                 error = xfs_inobt_lookup(cur, be32_to_cpu(agi->agi_newino),
 951                                          XFS_LOOKUP_EQ, &i);
 952                 if (error)
 953                         goto error0;
 954
 955                 if (i == 1) {
 956                         error = xfs_inobt_get_rec(cur, &rec, &j);
 957                         if (error)
 958                                 goto error0;
 959
 960                         if (j == 1 && rec.ir_freecount > 0) {
 961                                 /*
 962                                  * The last chunk allocated in the group
 963                                  * still has a free inode.
 964                                  */
 965                                 goto alloc_inode;
 966                         }
 967                 }
 968         }
 969
 970         /*
 971          * None left in the last group, search the whole AG
 972          */
 973         error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i);
 974         if (error)
 975                 goto error0;
 976         XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 977
 978         for (;;) {
 979                 error = xfs_inobt_get_rec(cur, &rec, &i);
 980                 if (error)
 981                         goto error0;
 982                 XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 983                 if (rec.ir_freecount > 0)
 984                         break;
 985                 error = xfs_btree_increment(cur, 0, &i);
 986                 if (error)
 987                         goto error0;
 988                 XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 989         }
 990
 991 alloc_inode:
 992         offset = xfs_ialloc_find_free(&rec.ir_free);
 993         ASSERT(offset >= 0);
 994         ASSERT(offset < XFS_INODES_PER_CHUNK);
 995         ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) %
 996                                    XFS_INODES_PER_CHUNK) == 0);
 997         ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset);
 998         rec.ir_free &= ~XFS_INOBT_MASK(offset);
 999         rec.ir_freecount--;
1000         error = xfs_inobt_update(cur, &rec);
1001         if (error)
1002                 goto error0;
1003         be32_add_cpu(&agi->agi_freecount, -1);
1004         xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT);
1005         pag->pagi_freecount--;
1006
1007         error = xfs_check_agi_freecount(cur, agi);
1008         if (error)
1009                 goto error0;
1010
1011         xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
1012         xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1);
1013         xfs_perag_put(pag);
1014         *inop = ino;
1015         return 0;
1016 error1:
1017         xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
1018 error0:
1019         xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
1020         xfs_perag_put(pag);
1021         return error;
1022 }
1023
1024 /*
1025  * Free disk inode.  Carefully avoids touching the incore inode, all
1026  * manipulations incore are the caller's responsibility.
1027  * The on-disk inode is not changed by this operation, only the
1028  * btree (free inode mask) is changed.
1029  */
1030 int
1031 xfs_difree(
1032         xfs_trans_t     *tp,            /* transaction pointer */
1033         xfs_ino_t       inode,          /* inode to be freed */
1034         xfs_bmap_free_t *flist,         /* extents to free */
1035         int             *delete,        /* set if inode cluster was deleted */
1036         xfs_ino_t       *first_ino)     /* first inode in deleted cluster */
1037 {
1038         /* REFERENCED */
1039         xfs_agblock_t   agbno;  /* block number containing inode */
1040         xfs_buf_t       *agbp;  /* buffer containing allocation group header */
1041         xfs_agino_t     agino;  /* inode number relative to allocation group */
1042         xfs_agnumber_t  agno;   /* allocation group number */
1043         xfs_agi_t       *agi;   /* allocation group header */
1044         xfs_btree_cur_t *cur;   /* inode btree cursor */
1045         int             error;  /* error return value */
1046         int             i;      /* result code */
1047         int             ilen;   /* inodes in an inode cluster */
1048         xfs_mount_t     *mp;    /* mount structure for filesystem */
1049         int             off;    /* offset of inode in inode chunk */
1050         xfs_inobt_rec_incore_t rec;     /* btree record */
1051         struct xfs_perag *pag;
1052
1053         mp = tp->t_mountp;
1054
1055         /*
1056          * Break up inode number into its components.
1057          */
1058         agno = XFS_INO_TO_AGNO(mp, inode);
1059         if (agno >= mp->m_sb.sb_agcount)  {
1060                 xfs_warn(mp, "%s: agno >= mp->m_sb.sb_agcount (%d >= %d).",
1061                         __func__, agno, mp->m_sb.sb_agcount);
1062                 ASSERT(0);
1063                 return XFS_ERROR(EINVAL);
1064         }
1065         agino = XFS_INO_TO_AGINO(mp, inode);
1066         if (inode != XFS_AGINO_TO_INO(mp, agno, agino))  {
1067                 xfs_warn(mp, "%s: inode != XFS_AGINO_TO_INO() (%llu != %llu).",
1068                         __func__, (unsigned long long)inode,
1069                         (unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino));
1070                 ASSERT(0);
1071                 return XFS_ERROR(EINVAL);
1072         }
1073         agbno = XFS_AGINO_TO_AGBNO(mp, agino);
1074         if (agbno >= mp->m_sb.sb_agblocks)  {
1075                 xfs_warn(mp, "%s: agbno >= mp->m_sb.sb_agblocks (%d >= %d).",
1076                         __func__, agbno, mp->m_sb.sb_agblocks);
1077                 ASSERT(0);
1078                 return XFS_ERROR(EINVAL);
1079         }
1080         /*
1081          * Get the allocation group header.
1082          */
1083         error = xfs_ialloc_read_agi(mp, tp, agno, &agbp);
1084         if (error) {
1085                 xfs_warn(mp, "%s: xfs_ialloc_read_agi() returned error %d.",
1086                         __func__, error);
1087                 return error;
1088         }
1089         agi = XFS_BUF_TO_AGI(agbp);
1090         ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC));
1091         ASSERT(agbno < be32_to_cpu(agi->agi_length));
1092         /*
1093          * Initialize the cursor.
1094          */
1095         cur = xfs_inobt_init_cursor(mp, tp, agbp, agno);
1096
1097         error = xfs_check_agi_freecount(cur, agi);
1098         if (error)
1099                 goto error0;
1100
1101         /*
1102          * Look for the entry describing this inode.
1103          */
1104         if ((error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i))) {
1105                 xfs_warn(mp, "%s: xfs_inobt_lookup() returned error %d.",
1106                         __func__, error);
1107                 goto error0;
1108         }
1109         XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
1110         error = xfs_inobt_get_rec(cur, &rec, &i);
1111         if (error) {
1112                 xfs_warn(mp, "%s: xfs_inobt_get_rec() returned error %d.",
1113                         __func__, error);
1114                 goto error0;
1115         }
1116         XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
1117         /*
1118          * Get the offset in the inode chunk.
1119          */
1120         off = agino - rec.ir_startino;
1121         ASSERT(off >= 0 && off < XFS_INODES_PER_CHUNK);
1122         ASSERT(!(rec.ir_free & XFS_INOBT_MASK(off)));
1123         /*
1124          * Mark the inode free & increment the count.
1125          */
1126         rec.ir_free |= XFS_INOBT_MASK(off);
1127         rec.ir_freecount++;
1128
1129         /*
1130          * When an inode cluster is free, it becomes eligible for removal
1131          */
1132         if (!(mp->m_flags & XFS_MOUNT_IKEEP) &&
1133             (rec.ir_freecount == XFS_IALLOC_INODES(mp))) {
1134
1135                 *delete = 1;
1136                 *first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino);
1137
1138                 /*
1139                  * Remove the inode cluster from the AGI B+Tree, adjust the
1140                  * AGI and Superblock inode counts, and mark the disk space
1141                  * to be freed when the transaction is committed.
1142                  */
1143                 ilen = XFS_IALLOC_INODES(mp);
1144                 be32_add_cpu(&agi->agi_count, -ilen);
1145                 be32_add_cpu(&agi->agi_freecount, -(ilen - 1));
1146                 xfs_ialloc_log_agi(tp, agbp, XFS_AGI_COUNT | XFS_AGI_FREECOUNT);
1147                 pag = xfs_perag_get(mp, agno);
1148                 pag->pagi_freecount -= ilen - 1;
1149                 xfs_perag_put(pag);
1150                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen);
1151                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1));
1152
1153                 if ((error = xfs_btree_delete(cur, &i))) {
1154                         xfs_warn(mp, "%s: xfs_btree_delete returned error %d.",
1155                                 __func__, error);
1156                         goto error0;
1157                 }
1158
1159                 xfs_bmap_add_free(XFS_AGB_TO_FSB(mp,
1160                                 agno, XFS_INO_TO_AGBNO(mp,rec.ir_startino)),
1161                                 XFS_IALLOC_BLOCKS(mp), flist, mp);
1162         } else {
1163                 *delete = 0;
1164
1165                 error = xfs_inobt_update(cur, &rec);
1166                 if (error) {
1167                         xfs_warn(mp, "%s: xfs_inobt_update returned error %d.",
1168                                 __func__, error);
1169                         goto error0;
1170                 }
1171
1172                 /*
1173                  * Change the inode free counts and log the ag/sb changes.
1174                  */
1175                 be32_add_cpu(&agi->agi_freecount, 1);
1176                 xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT);
1177                 pag = xfs_perag_get(mp, agno);
1178                 pag->pagi_freecount++;
1179                 xfs_perag_put(pag);
1180                 xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1);
1181         }
1182
1183         error = xfs_check_agi_freecount(cur, agi);
1184         if (error)
1185                 goto error0;
1186
1187         xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
1188         return 0;
1189
1190 error0:
1191         xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
1192         return error;
1193 }
1194
1195 STATIC int
1196 xfs_imap_lookup(
1197         struct xfs_mount        *mp,
1198         struct xfs_trans        *tp,
1199         xfs_agnumber_t          agno,
1200         xfs_agino_t             agino,
1201         xfs_agblock_t           agbno,
1202         xfs_agblock_t           *chunk_agbno,
1203         xfs_agblock_t           *offset_agbno,
1204         int                     flags)
1205 {
1206         struct xfs_inobt_rec_incore rec;
1207         struct xfs_btree_cur    *cur;
1208         struct xfs_buf          *agbp;
1209         int                     error;
1210         int                     i;
1211
1212         error = xfs_ialloc_read_agi(mp, tp, agno, &agbp);
1213         if (error) {
1214                 xfs_alert(mp,
1215                         "%s: xfs_ialloc_read_agi() returned error %d, agno %d",
1216                         __func__, error, agno);
1217                 return error;
1218         }
1219
1220         /*
1221          * Lookup the inode record for the given agino. If the record cannot be
1222          * found, then it's an invalid inode number and we should abort. Once
1223          * we have a record, we need to ensure it contains the inode number
1224          * we are looking up.
1225          */
1226         cur = xfs_inobt_init_cursor(mp, tp, agbp, agno);
1227         error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i);
1228         if (!error) {
1229                 if (i)
1230                         error = xfs_inobt_get_rec(cur, &rec, &i);
1231                 if (!error && i == 0)
1232                         error = EINVAL;
1233         }
1234
1235         xfs_trans_brelse(tp, agbp);
1236         xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
1237         if (error)
1238                 return error;
1239
1240         /* check that the returned record contains the required inode */
1241         if (rec.ir_startino > agino ||
1242             rec.ir_startino + XFS_IALLOC_INODES(mp) <= agino)
1243                 return EINVAL;
1244
1245         /* for untrusted inodes check it is allocated first */
1246         if ((flags & XFS_IGET_UNTRUSTED) &&
1247             (rec.ir_free & XFS_INOBT_MASK(agino - rec.ir_startino)))
1248                 return EINVAL;
1249
1250         *chunk_agbno = XFS_AGINO_TO_AGBNO(mp, rec.ir_startino);
1251         *offset_agbno = agbno - *chunk_agbno;
1252         return 0;
1253 }
1254
1255 /*
1256  * Return the location of the inode in imap, for mapping it into a buffer.
1257  */
1258 int
1259 xfs_imap(
1260         xfs_mount_t      *mp,   /* file system mount structure */
1261         xfs_trans_t      *tp,   /* transaction pointer */
1262         xfs_ino_t       ino,    /* inode to locate */
1263         struct xfs_imap *imap,  /* location map structure */
1264         uint            flags)  /* flags for inode btree lookup */
1265 {
1266         xfs_agblock_t   agbno;  /* block number of inode in the alloc group */
1267         xfs_agino_t     agino;  /* inode number within alloc group */
1268         xfs_agnumber_t  agno;   /* allocation group number */
1269         int             blks_per_cluster; /* num blocks per inode cluster */
1270         xfs_agblock_t   chunk_agbno;    /* first block in inode chunk */
1271         xfs_agblock_t   cluster_agbno;  /* first block in inode cluster */
1272         int             error;  /* error code */
1273         int             offset; /* index of inode in its buffer */
1274         int             offset_agbno;   /* blks from chunk start to inode */
1275
1276         ASSERT(ino != NULLFSINO);
1277
1278         /*
1279          * Split up the inode number into its parts.
1280          */
1281         agno = XFS_INO_TO_AGNO(mp, ino);
1282         agino = XFS_INO_TO_AGINO(mp, ino);
1283         agbno = XFS_AGINO_TO_AGBNO(mp, agino);
1284         if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks ||
1285             ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
1286 #ifdef DEBUG
1287                 /*
1288                  * Don't output diagnostic information for untrusted inodes
1289                  * as they can be invalid without implying corruption.
1290                  */
1291                 if (flags & XFS_IGET_UNTRUSTED)
1292                         return XFS_ERROR(EINVAL);
1293                 if (agno >= mp->m_sb.sb_agcount) {
1294                         xfs_alert(mp,
1295                                 "%s: agno (%d) >= mp->m_sb.sb_agcount (%d)",
1296                                 __func__, agno, mp->m_sb.sb_agcount);
1297                 }
1298                 if (agbno >= mp->m_sb.sb_agblocks) {
1299                         xfs_alert(mp,
1300                 "%s: agbno (0x%llx) >= mp->m_sb.sb_agblocks (0x%lx)",
1301                                 __func__, (unsigned long long)agbno,
1302                                 (unsigned long)mp->m_sb.sb_agblocks);
1303                 }
1304                 if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
1305                         xfs_alert(mp,
1306                 "%s: ino (0x%llx) != XFS_AGINO_TO_INO() (0x%llx)",
1307                                 __func__, ino,
1308                                 XFS_AGINO_TO_INO(mp, agno, agino));
1309                 }
1310                 xfs_stack_trace();
1311 #endif /* DEBUG */
1312                 return XFS_ERROR(EINVAL);
1313         }
1314
1315         blks_per_cluster = XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_blocklog;
1316
1317         /*
1318          * For bulkstat and handle lookups, we have an untrusted inode number
1319          * that we have to verify is valid. We cannot do this just by reading
1320          * the inode buffer as it may have been unlinked and removed leaving
1321          * inodes in stale state on disk. Hence we have to do a btree lookup
1322          * in all cases where an untrusted inode number is passed.
1323          */
1324         if (flags & XFS_IGET_UNTRUSTED) {
1325                 error = xfs_imap_lookup(mp, tp, agno, agino, agbno,
1326                                         &chunk_agbno, &offset_agbno, flags);
1327                 if (error)
1328                         return error;
1329                 goto out_map;
1330         }
1331
1332         /*
1333          * If the inode cluster size is the same as the blocksize or
1334          * smaller we get to the buffer by simple arithmetics.
1335          */
1336         if (XFS_INODE_CLUSTER_SIZE(mp) <= mp->m_sb.sb_blocksize) {
1337                 offset = XFS_INO_TO_OFFSET(mp, ino);
1338                 ASSERT(offset < mp->m_sb.sb_inopblock);
1339
1340                 imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno);
1341                 imap->im_len = XFS_FSB_TO_BB(mp, 1);
1342                 imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog);
1343                 return 0;
1344         }
1345
1346         /*
1347          * If the inode chunks are aligned then use simple maths to
1348          * find the location. Otherwise we have to do a btree
1349          * lookup to find the location.
1350          */
1351         if (mp->m_inoalign_mask) {
1352                 offset_agbno = agbno & mp->m_inoalign_mask;
1353                 chunk_agbno = agbno - offset_agbno;
1354         } else {
1355                 error = xfs_imap_lookup(mp, tp, agno, agino, agbno,
1356                                         &chunk_agbno, &offset_agbno, flags);
1357                 if (error)
1358                         return error;
1359         }
1360
1361 out_map:
1362         ASSERT(agbno >= chunk_agbno);
1363         cluster_agbno = chunk_agbno +
1364                 ((offset_agbno / blks_per_cluster) * blks_per_cluster);
1365         offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) +
1366                 XFS_INO_TO_OFFSET(mp, ino);
1367
1368         imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno);
1369         imap->im_len = XFS_FSB_TO_BB(mp, blks_per_cluster);
1370         imap->im_boffset = (ushort)(offset << mp->m_sb.sb_inodelog);
1371
1372         /*
1373          * If the inode number maps to a block outside the bounds
1374          * of the file system then return NULL rather than calling
1375          * read_buf and panicing when we get an error from the
1376          * driver.
1377          */
1378         if ((imap->im_blkno + imap->im_len) >
1379             XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)) {
1380                 xfs_alert(mp,
1381         "%s: (im_blkno (0x%llx) + im_len (0x%llx)) > sb_dblocks (0x%llx)",
1382                         __func__, (unsigned long long) imap->im_blkno,
1383                         (unsigned long long) imap->im_len,
1384                         XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks));
1385                 return XFS_ERROR(EINVAL);
1386         }
1387         return 0;
1388 }
1389
1390 /*
1391  * Compute and fill in value of m_in_maxlevels.
1392  */
1393 void
1394 xfs_ialloc_compute_maxlevels(
1395         xfs_mount_t     *mp)            /* file system mount structure */
1396 {
1397         int             level;
1398         uint            maxblocks;
1399         uint            maxleafents;
1400         int             minleafrecs;
1401         int             minnoderecs;
1402
1403         maxleafents = (1LL << XFS_INO_AGINO_BITS(mp)) >>
1404                 XFS_INODES_PER_CHUNK_LOG;
1405         minleafrecs = mp->m_alloc_mnr[0];
1406         minnoderecs = mp->m_alloc_mnr[1];
1407         maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs;
1408         for (level = 1; maxblocks > 1; level++)
1409                 maxblocks = (maxblocks + minnoderecs - 1) / minnoderecs;
1410         mp->m_in_maxlevels = level;
1411 }
1412
1413 /*
1414  * Log specified fields for the ag hdr (inode section)
1415  */
1416 void
1417 xfs_ialloc_log_agi(
1418         xfs_trans_t     *tp,            /* transaction pointer */
1419         xfs_buf_t       *bp,            /* allocation group header buffer */
1420         int             fields)         /* bitmask of fields to log */
1421 {
1422         int                     first;          /* first byte number */
1423         int                     last;           /* last byte number */
1424         static const short      offsets[] = {   /* field starting offsets */
1425                                         /* keep in sync with bit definitions */
1426                 offsetof(xfs_agi_t, agi_magicnum),
1427                 offsetof(xfs_agi_t, agi_versionnum),
1428                 offsetof(xfs_agi_t, agi_seqno),
1429                 offsetof(xfs_agi_t, agi_length),
1430                 offsetof(xfs_agi_t, agi_count),
1431                 offsetof(xfs_agi_t, agi_root),
1432                 offsetof(xfs_agi_t, agi_level),
1433                 offsetof(xfs_agi_t, agi_freecount),
1434                 offsetof(xfs_agi_t, agi_newino),
1435                 offsetof(xfs_agi_t, agi_dirino),
1436                 offsetof(xfs_agi_t, agi_unlinked),
1437                 sizeof(xfs_agi_t)
1438         };
1439 #ifdef DEBUG
1440         xfs_agi_t               *agi;   /* allocation group header */
1441
1442         agi = XFS_BUF_TO_AGI(bp);
1443         ASSERT(agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC));
1444 #endif
1445         /*
1446          * Compute byte offsets for the first and last fields.
1447          */
1448         xfs_btree_offsets(fields, offsets, XFS_AGI_NUM_BITS, &first, &last);
1449         /*
1450          * Log the allocation group inode header buffer.
1451          */
1452         xfs_trans_log_buf(tp, bp, first, last);
1453 }
1454
1455 #ifdef DEBUG
1456 STATIC void
1457 xfs_check_agi_unlinked(
1458         struct xfs_agi          *agi)
1459 {
1460         int                     i;
1461
1462         for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++)
1463                 ASSERT(agi->agi_unlinked[i]);
1464 }
1465 #else
1466 #define xfs_check_agi_unlinked(agi)
1467 #endif
1468
1469 /*
1470  * Read in the allocation group header (inode allocation section)
1471  */
1472 int
1473 xfs_read_agi(
1474         struct xfs_mount        *mp,    /* file system mount structure */
1475         struct xfs_trans        *tp,    /* transaction pointer */
1476         xfs_agnumber_t          agno,   /* allocation group number */
1477         struct xfs_buf          **bpp)  /* allocation group hdr buf */
1478 {
1479         struct xfs_agi          *agi;   /* allocation group header */
1480         int                     agi_ok; /* agi is consistent */
1481         int                     error;
1482
1483         ASSERT(agno != NULLAGNUMBER);
1484
1485         error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
1486                         XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
1487                         XFS_FSS_TO_BB(mp, 1), 0, bpp);
1488         if (error)
1489                 return error;
1490
1491         ASSERT(!xfs_buf_geterror(*bpp));
1492         agi = XFS_BUF_TO_AGI(*bpp);
1493
1494         /*
1495          * Validate the magic number of the agi block.
1496          */
1497         agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
1498                 XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) &&
1499                 be32_to_cpu(agi->agi_seqno) == agno;
1500         if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
1501                         XFS_RANDOM_IALLOC_READ_AGI))) {
1502                 XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW,
1503                                      mp, agi);
1504                 xfs_trans_brelse(tp, *bpp);
1505                 return XFS_ERROR(EFSCORRUPTED);
1506         }
1507
1508         xfs_buf_set_ref(*bpp, XFS_AGI_REF);
1509
1510         xfs_check_agi_unlinked(agi);
1511         return 0;
1512 }
1513
1514 int
1515 xfs_ialloc_read_agi(
1516         struct xfs_mount        *mp,    /* file system mount structure */
1517         struct xfs_trans        *tp,    /* transaction pointer */
1518         xfs_agnumber_t          agno,   /* allocation group number */
1519         struct xfs_buf          **bpp)  /* allocation group hdr buf */
1520 {
1521         struct xfs_agi          *agi;   /* allocation group header */
1522         struct xfs_perag        *pag;   /* per allocation group data */
1523         int                     error;
1524
1525         error = xfs_read_agi(mp, tp, agno, bpp);
1526         if (error)
1527                 return error;
1528
1529         agi = XFS_BUF_TO_AGI(*bpp);
1530         pag = xfs_perag_get(mp, agno);
1531         if (!pag->pagi_init) {
1532                 pag->pagi_freecount = be32_to_cpu(agi->agi_freecount);
1533                 pag->pagi_count = be32_to_cpu(agi->agi_count);
1534                 pag->pagi_init = 1;
1535         }
1536
1537         /*
1538          * It's possible for these to be out of sync if
1539          * we are in the middle of a forced shutdown.
1540          */
1541         ASSERT(pag->pagi_freecount == be32_to_cpu(agi->agi_freecount) ||
1542                 XFS_FORCED_SHUTDOWN(mp));
1543         xfs_perag_put(pag);
1544         return 0;
1545 }
1546
1547 /*
1548  * Read in the agi to initialise the per-ag data in the mount structure
1549  */
1550 int
1551 xfs_ialloc_pagi_init(
1552         xfs_mount_t     *mp,            /* file system mount structure */
1553         xfs_trans_t     *tp,            /* transaction pointer */
1554         xfs_agnumber_t  agno)           /* allocation group number */
1555 {
1556         xfs_buf_t       *bp = NULL;
1557         int             error;
1558
1559         error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
1560         if (error)
1561                 return error;
1562         if (bp)
1563                 xfs_trans_brelse(tp, bp);
1564         return 0;
1565 }