libxfs/xfs_bmap_btree.c

   1 /*
   2  * Copyright (c) 2000-2003,2005 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 "libxfs_priv.h"
  19 #include "xfs_fs.h"
  20 #include "xfs_shared.h"
  21 #include "xfs_format.h"
  22 #include "xfs_log_format.h"
  23 #include "xfs_trans_resv.h"
  24 #include "xfs_bit.h"
  25 #include "xfs_mount.h"
  26 #include "xfs_defer.h"
  27 #include "xfs_inode.h"
  28 #include "xfs_trans.h"
  29 #include "xfs_alloc.h"
  30 #include "xfs_btree.h"
  31 #include "xfs_bmap_btree.h"
  32 #include "xfs_bmap.h"
  33 #include "xfs_trace.h"
  34 #include "xfs_cksum.h"
  35 #include "xfs_rmap.h"
  36
  37 /*
  38  * Convert on-disk form of btree root to in-memory form.
  39  */
  40 void
  41 xfs_bmdr_to_bmbt(
  42         struct xfs_inode        *ip,
  43         xfs_bmdr_block_t        *dblock,
  44         int                     dblocklen,
  45         struct xfs_btree_block  *rblock,
  46         int                     rblocklen)
  47 {
  48         struct xfs_mount        *mp = ip->i_mount;
  49         int                     dmxr;
  50         xfs_bmbt_key_t          *fkp;
  51         __be64                  *fpp;
  52         xfs_bmbt_key_t          *tkp;
  53         __be64                  *tpp;
  54
  55         xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
  56                                  XFS_BTNUM_BMAP, 0, 0, ip->i_ino,
  57                                  XFS_BTREE_LONG_PTRS);
  58         rblock->bb_level = dblock->bb_level;
  59         ASSERT(be16_to_cpu(rblock->bb_level) > 0);
  60         rblock->bb_numrecs = dblock->bb_numrecs;
  61         dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
  62         fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
  63         tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
  64         fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
  65         tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
  66         dmxr = be16_to_cpu(dblock->bb_numrecs);
  67         memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
  68         memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
  69 }
  70
  71 void
  72 xfs_bmbt_disk_get_all(
  73         struct xfs_bmbt_rec     *rec,
  74         struct xfs_bmbt_irec    *irec)
  75 {
  76         uint64_t                l0 = get_unaligned_be64(&rec->l0);
  77         uint64_t                l1 = get_unaligned_be64(&rec->l1);
  78
  79         irec->br_startoff = (l0 & xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
  80         irec->br_startblock = ((l0 & xfs_mask64lo(9)) << 43) | (l1 >> 21);
  81         irec->br_blockcount = l1 & xfs_mask64lo(21);
  82         if (l0 >> (64 - BMBT_EXNTFLAG_BITLEN))
  83                 irec->br_state = XFS_EXT_UNWRITTEN;
  84         else
  85                 irec->br_state = XFS_EXT_NORM;
  86 }
  87
  88 /*
  89  * Extract the blockcount field from an on disk bmap extent record.
  90  */
  91 xfs_filblks_t
  92 xfs_bmbt_disk_get_blockcount(
  93         xfs_bmbt_rec_t  *r)
  94 {
  95         return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
  96 }
  97
  98 /*
  99  * Extract the startoff field from a disk format bmap extent record.
 100  */
 101 xfs_fileoff_t
 102 xfs_bmbt_disk_get_startoff(
 103         xfs_bmbt_rec_t  *r)
 104 {
 105         return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
 106                  xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
 107 }
 108
 109 /*
 110  * Set all the fields in a bmap extent record from the uncompressed form.
 111  */
 112 void
 113 xfs_bmbt_disk_set_all(
 114         struct xfs_bmbt_rec     *r,
 115         struct xfs_bmbt_irec    *s)
 116 {
 117         int                     extent_flag = (s->br_state != XFS_EXT_NORM);
 118
 119         ASSERT(s->br_state == XFS_EXT_NORM || s->br_state == XFS_EXT_UNWRITTEN);
 120         ASSERT(!(s->br_startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)));
 121         ASSERT(!(s->br_blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)));
 122         ASSERT(!(s->br_startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)));
 123
 124         put_unaligned_be64(
 125                 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
 126                  ((xfs_bmbt_rec_base_t)s->br_startoff << 9) |
 127                  ((xfs_bmbt_rec_base_t)s->br_startblock >> 43), &r->l0);
 128         put_unaligned_be64(
 129                 ((xfs_bmbt_rec_base_t)s->br_startblock << 21) |
 130                  ((xfs_bmbt_rec_base_t)s->br_blockcount &
 131                   (xfs_bmbt_rec_base_t)xfs_mask64lo(21)), &r->l1);
 132 }
 133
 134 /*
 135  * Convert in-memory form of btree root to on-disk form.
 136  */
 137 void
 138 xfs_bmbt_to_bmdr(
 139         struct xfs_mount        *mp,
 140         struct xfs_btree_block  *rblock,
 141         int                     rblocklen,
 142         xfs_bmdr_block_t        *dblock,
 143         int                     dblocklen)
 144 {
 145         int                     dmxr;
 146         xfs_bmbt_key_t          *fkp;
 147         __be64                  *fpp;
 148         xfs_bmbt_key_t          *tkp;
 149         __be64                  *tpp;
 150
 151         if (xfs_sb_version_hascrc(&mp->m_sb)) {
 152                 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
 153                 ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
 154                        &mp->m_sb.sb_meta_uuid));
 155                 ASSERT(rblock->bb_u.l.bb_blkno ==
 156                        cpu_to_be64(XFS_BUF_DADDR_NULL));
 157         } else
 158                 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
 159         ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
 160         ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
 161         ASSERT(rblock->bb_level != 0);
 162         dblock->bb_level = rblock->bb_level;
 163         dblock->bb_numrecs = rblock->bb_numrecs;
 164         dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
 165         fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
 166         tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
 167         fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
 168         tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
 169         dmxr = be16_to_cpu(dblock->bb_numrecs);
 170         memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
 171         memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
 172 }
 173
 174 STATIC struct xfs_btree_cur *
 175 xfs_bmbt_dup_cursor(
 176         struct xfs_btree_cur    *cur)
 177 {
 178         struct xfs_btree_cur    *new;
 179
 180         new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
 181                         cur->bc_private.b.ip, cur->bc_private.b.whichfork);
 182
 183         /*
 184          * Copy the firstblock, dfops, and flags values,
 185          * since init cursor doesn't get them.
 186          */
 187         new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
 188         new->bc_private.b.dfops = cur->bc_private.b.dfops;
 189         new->bc_private.b.flags = cur->bc_private.b.flags;
 190
 191         return new;
 192 }
 193
 194 STATIC void
 195 xfs_bmbt_update_cursor(
 196         struct xfs_btree_cur    *src,
 197         struct xfs_btree_cur    *dst)
 198 {
 199         ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
 200                (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
 201         ASSERT(dst->bc_private.b.dfops == src->bc_private.b.dfops);
 202
 203         dst->bc_private.b.allocated += src->bc_private.b.allocated;
 204         dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
 205
 206         src->bc_private.b.allocated = 0;
 207 }
 208
 209 STATIC int
 210 xfs_bmbt_alloc_block(
 211         struct xfs_btree_cur    *cur,
 212         union xfs_btree_ptr     *start,
 213         union xfs_btree_ptr     *new,
 214         int                     *stat)
 215 {
 216         xfs_alloc_arg_t         args;           /* block allocation args */
 217         int                     error;          /* error return value */
 218
 219         memset(&args, 0, sizeof(args));
 220         args.tp = cur->bc_tp;
 221         args.mp = cur->bc_mp;
 222         args.fsbno = cur->bc_private.b.firstblock;
 223         args.firstblock = args.fsbno;
 224         xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_private.b.ip->i_ino,
 225                         cur->bc_private.b.whichfork);
 226
 227         if (args.fsbno == NULLFSBLOCK) {
 228                 args.fsbno = be64_to_cpu(start->l);
 229                 args.type = XFS_ALLOCTYPE_START_BNO;
 230                 /*
 231                  * Make sure there is sufficient room left in the AG to
 232                  * complete a full tree split for an extent insert.  If
 233                  * we are converting the middle part of an extent then
 234                  * we may need space for two tree splits.
 235                  *
 236                  * We are relying on the caller to make the correct block
 237                  * reservation for this operation to succeed.  If the
 238                  * reservation amount is insufficient then we may fail a
 239                  * block allocation here and corrupt the filesystem.
 240                  */
 241                 args.minleft = args.tp->t_blk_res;
 242         } else if (cur->bc_private.b.dfops->dop_low) {
 243                 args.type = XFS_ALLOCTYPE_START_BNO;
 244         } else {
 245                 args.type = XFS_ALLOCTYPE_NEAR_BNO;
 246         }
 247
 248         args.minlen = args.maxlen = args.prod = 1;
 249         args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
 250         if (!args.wasdel && args.tp->t_blk_res == 0) {
 251                 error = -ENOSPC;
 252                 goto error0;
 253         }
 254         error = xfs_alloc_vextent(&args);
 255         if (error)
 256                 goto error0;
 257
 258         if (args.fsbno == NULLFSBLOCK && args.minleft) {
 259                 /*
 260                  * Could not find an AG with enough free space to satisfy
 261                  * a full btree split.  Try again and if
 262                  * successful activate the lowspace algorithm.
 263                  */
 264                 args.fsbno = 0;
 265                 args.type = XFS_ALLOCTYPE_FIRST_AG;
 266                 error = xfs_alloc_vextent(&args);
 267                 if (error)
 268                         goto error0;
 269                 cur->bc_private.b.dfops->dop_low = true;
 270         }
 271         if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
 272                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 273                 *stat = 0;
 274                 return 0;
 275         }
 276         ASSERT(args.len == 1);
 277         cur->bc_private.b.firstblock = args.fsbno;
 278         cur->bc_private.b.allocated++;
 279         cur->bc_private.b.ip->i_d.di_nblocks++;
 280         xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
 281         xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
 282                         XFS_TRANS_DQ_BCOUNT, 1L);
 283
 284         new->l = cpu_to_be64(args.fsbno);
 285
 286         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 287         *stat = 1;
 288         return 0;
 289
 290  error0:
 291         XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 292         return error;
 293 }
 294
 295 STATIC int
 296 xfs_bmbt_free_block(
 297         struct xfs_btree_cur    *cur,
 298         struct xfs_buf          *bp)
 299 {
 300         struct xfs_mount        *mp = cur->bc_mp;
 301         struct xfs_inode        *ip = cur->bc_private.b.ip;
 302         struct xfs_trans        *tp = cur->bc_tp;
 303         xfs_fsblock_t           fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
 304         struct xfs_owner_info   oinfo;
 305
 306         xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, cur->bc_private.b.whichfork);
 307         xfs_bmap_add_free(mp, cur->bc_private.b.dfops, fsbno, 1, &oinfo);
 308         ip->i_d.di_nblocks--;
 309
 310         xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 311         xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
 312         return 0;
 313 }
 314
 315 STATIC int
 316 xfs_bmbt_get_minrecs(
 317         struct xfs_btree_cur    *cur,
 318         int                     level)
 319 {
 320         if (level == cur->bc_nlevels - 1) {
 321                 struct xfs_ifork        *ifp;
 322
 323                 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
 324                                     cur->bc_private.b.whichfork);
 325
 326                 return xfs_bmbt_maxrecs(cur->bc_mp,
 327                                         ifp->if_broot_bytes, level == 0) / 2;
 328         }
 329
 330         return cur->bc_mp->m_bmap_dmnr[level != 0];
 331 }
 332
 333 int
 334 xfs_bmbt_get_maxrecs(
 335         struct xfs_btree_cur    *cur,
 336         int                     level)
 337 {
 338         if (level == cur->bc_nlevels - 1) {
 339                 struct xfs_ifork        *ifp;
 340
 341                 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
 342                                     cur->bc_private.b.whichfork);
 343
 344                 return xfs_bmbt_maxrecs(cur->bc_mp,
 345                                         ifp->if_broot_bytes, level == 0);
 346         }
 347
 348         return cur->bc_mp->m_bmap_dmxr[level != 0];
 349
 350 }
 351
 352 /*
 353  * Get the maximum records we could store in the on-disk format.
 354  *
 355  * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
 356  * for the root node this checks the available space in the dinode fork
 357  * so that we can resize the in-memory buffer to match it.  After a
 358  * resize to the maximum size this function returns the same value
 359  * as xfs_bmbt_get_maxrecs for the root node, too.
 360  */
 361 STATIC int
 362 xfs_bmbt_get_dmaxrecs(
 363         struct xfs_btree_cur    *cur,
 364         int                     level)
 365 {
 366         if (level != cur->bc_nlevels - 1)
 367                 return cur->bc_mp->m_bmap_dmxr[level != 0];
 368         return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
 369 }
 370
 371 STATIC void
 372 xfs_bmbt_init_key_from_rec(
 373         union xfs_btree_key     *key,
 374         union xfs_btree_rec     *rec)
 375 {
 376         key->bmbt.br_startoff =
 377                 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
 378 }
 379
 380 STATIC void
 381 xfs_bmbt_init_high_key_from_rec(
 382         union xfs_btree_key     *key,
 383         union xfs_btree_rec     *rec)
 384 {
 385         key->bmbt.br_startoff = cpu_to_be64(
 386                         xfs_bmbt_disk_get_startoff(&rec->bmbt) +
 387                         xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
 388 }
 389
 390 STATIC void
 391 xfs_bmbt_init_rec_from_cur(
 392         struct xfs_btree_cur    *cur,
 393         union xfs_btree_rec     *rec)
 394 {
 395         xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
 396 }
 397
 398 STATIC void
 399 xfs_bmbt_init_ptr_from_cur(
 400         struct xfs_btree_cur    *cur,
 401         union xfs_btree_ptr     *ptr)
 402 {
 403         ptr->l = 0;
 404 }
 405
 406 STATIC int64_t
 407 xfs_bmbt_key_diff(
 408         struct xfs_btree_cur    *cur,
 409         union xfs_btree_key     *key)
 410 {
 411         return (int64_t)be64_to_cpu(key->bmbt.br_startoff) -
 412                                       cur->bc_rec.b.br_startoff;
 413 }
 414
 415 STATIC int64_t
 416 xfs_bmbt_diff_two_keys(
 417         struct xfs_btree_cur    *cur,
 418         union xfs_btree_key     *k1,
 419         union xfs_btree_key     *k2)
 420 {
 421         return (int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
 422                           be64_to_cpu(k2->bmbt.br_startoff);
 423 }
 424
 425 static xfs_failaddr_t
 426 xfs_bmbt_verify(
 427         struct xfs_buf          *bp)
 428 {
 429         struct xfs_mount        *mp = bp->b_target->bt_mount;
 430         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 431         xfs_failaddr_t          fa;
 432         unsigned int            level;
 433
 434         switch (block->bb_magic) {
 435         case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
 436                 /*
 437                  * XXX: need a better way of verifying the owner here. Right now
 438                  * just make sure there has been one set.
 439                  */
 440                 fa = xfs_btree_lblock_v5hdr_verify(bp, XFS_RMAP_OWN_UNKNOWN);
 441                 if (fa)
 442                         return fa;
 443                 /* fall through */
 444         case cpu_to_be32(XFS_BMAP_MAGIC):
 445                 break;
 446         default:
 447                 return __this_address;
 448         }
 449
 450         /*
 451          * numrecs and level verification.
 452          *
 453          * We don't know what fork we belong to, so just verify that the level
 454          * is less than the maximum of the two. Later checks will be more
 455          * precise.
 456          */
 457         level = be16_to_cpu(block->bb_level);
 458         if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
 459                 return __this_address;
 460
 461         return xfs_btree_lblock_verify(bp, mp->m_bmap_dmxr[level != 0]);
 462 }
 463
 464 static void
 465 xfs_bmbt_read_verify(
 466         struct xfs_buf  *bp)
 467 {
 468         xfs_failaddr_t  fa;
 469
 470         if (!xfs_btree_lblock_verify_crc(bp))
 471                 xfs_verifier_error(bp, -EFSBADCRC, __this_address);
 472         else {
 473                 fa = xfs_bmbt_verify(bp);
 474                 if (fa)
 475                         xfs_verifier_error(bp, -EFSCORRUPTED, fa);
 476         }
 477
 478         if (bp->b_error)
 479                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 480 }
 481
 482 static void
 483 xfs_bmbt_write_verify(
 484         struct xfs_buf  *bp)
 485 {
 486         xfs_failaddr_t  fa;
 487
 488         fa = xfs_bmbt_verify(bp);
 489         if (fa) {
 490                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 491                 xfs_verifier_error(bp, -EFSCORRUPTED, fa);
 492                 return;
 493         }
 494         xfs_btree_lblock_calc_crc(bp);
 495 }
 496
 497 const struct xfs_buf_ops xfs_bmbt_buf_ops = {
 498         .name = "xfs_bmbt",
 499         .verify_read = xfs_bmbt_read_verify,
 500         .verify_write = xfs_bmbt_write_verify,
 501         .verify_struct = xfs_bmbt_verify,
 502 };
 503
 504
 505 STATIC int
 506 xfs_bmbt_keys_inorder(
 507         struct xfs_btree_cur    *cur,
 508         union xfs_btree_key     *k1,
 509         union xfs_btree_key     *k2)
 510 {
 511         return be64_to_cpu(k1->bmbt.br_startoff) <
 512                 be64_to_cpu(k2->bmbt.br_startoff);
 513 }
 514
 515 STATIC int
 516 xfs_bmbt_recs_inorder(
 517         struct xfs_btree_cur    *cur,
 518         union xfs_btree_rec     *r1,
 519         union xfs_btree_rec     *r2)
 520 {
 521         return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
 522                 xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
 523                 xfs_bmbt_disk_get_startoff(&r2->bmbt);
 524 }
 525
 526 static const struct xfs_btree_ops xfs_bmbt_ops = {
 527         .rec_len                = sizeof(xfs_bmbt_rec_t),
 528         .key_len                = sizeof(xfs_bmbt_key_t),
 529
 530         .dup_cursor             = xfs_bmbt_dup_cursor,
 531         .update_cursor          = xfs_bmbt_update_cursor,
 532         .alloc_block            = xfs_bmbt_alloc_block,
 533         .free_block             = xfs_bmbt_free_block,
 534         .get_maxrecs            = xfs_bmbt_get_maxrecs,
 535         .get_minrecs            = xfs_bmbt_get_minrecs,
 536         .get_dmaxrecs           = xfs_bmbt_get_dmaxrecs,
 537         .init_key_from_rec      = xfs_bmbt_init_key_from_rec,
 538         .init_high_key_from_rec = xfs_bmbt_init_high_key_from_rec,
 539         .init_rec_from_cur      = xfs_bmbt_init_rec_from_cur,
 540         .init_ptr_from_cur      = xfs_bmbt_init_ptr_from_cur,
 541         .key_diff               = xfs_bmbt_key_diff,
 542         .diff_two_keys          = xfs_bmbt_diff_two_keys,
 543         .buf_ops                = &xfs_bmbt_buf_ops,
 544         .keys_inorder           = xfs_bmbt_keys_inorder,
 545         .recs_inorder           = xfs_bmbt_recs_inorder,
 546 };
 547
 548 /*
 549  * Allocate a new bmap btree cursor.
 550  */
 551 struct xfs_btree_cur *                          /* new bmap btree cursor */
 552 xfs_bmbt_init_cursor(
 553         struct xfs_mount        *mp,            /* file system mount point */
 554         struct xfs_trans        *tp,            /* transaction pointer */
 555         struct xfs_inode        *ip,            /* inode owning the btree */
 556         int                     whichfork)      /* data or attr fork */
 557 {
 558         struct xfs_ifork        *ifp = XFS_IFORK_PTR(ip, whichfork);
 559         struct xfs_btree_cur    *cur;
 560         ASSERT(whichfork != XFS_COW_FORK);
 561
 562         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
 563
 564         cur->bc_tp = tp;
 565         cur->bc_mp = mp;
 566         cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
 567         cur->bc_btnum = XFS_BTNUM_BMAP;
 568         cur->bc_blocklog = mp->m_sb.sb_blocklog;
 569         cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
 570
 571         cur->bc_ops = &xfs_bmbt_ops;
 572         cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
 573         if (xfs_sb_version_hascrc(&mp->m_sb))
 574                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 575
 576         cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
 577         cur->bc_private.b.ip = ip;
 578         cur->bc_private.b.firstblock = NULLFSBLOCK;
 579         cur->bc_private.b.dfops = NULL;
 580         cur->bc_private.b.allocated = 0;
 581         cur->bc_private.b.flags = 0;
 582         cur->bc_private.b.whichfork = whichfork;
 583
 584         return cur;
 585 }
 586
 587 /*
 588  * Calculate number of records in a bmap btree block.
 589  */
 590 int
 591 xfs_bmbt_maxrecs(
 592         struct xfs_mount        *mp,
 593         int                     blocklen,
 594         int                     leaf)
 595 {
 596         blocklen -= XFS_BMBT_BLOCK_LEN(mp);
 597
 598         if (leaf)
 599                 return blocklen / sizeof(xfs_bmbt_rec_t);
 600         return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
 601 }
 602
 603 /*
 604  * Calculate number of records in a bmap btree inode root.
 605  */
 606 int
 607 xfs_bmdr_maxrecs(
 608         int                     blocklen,
 609         int                     leaf)
 610 {
 611         blocklen -= sizeof(xfs_bmdr_block_t);
 612
 613         if (leaf)
 614                 return blocklen / sizeof(xfs_bmdr_rec_t);
 615         return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
 616 }
 617
 618 /*
 619  * Change the owner of a btree format fork fo the inode passed in. Change it to
 620  * the owner of that is passed in so that we can change owners before or after
 621  * we switch forks between inodes. The operation that the caller is doing will
 622  * determine whether is needs to change owner before or after the switch.
 623  *
 624  * For demand paged transactional modification, the fork switch should be done
 625  * after reading in all the blocks, modifying them and pinning them in the
 626  * transaction. For modification when the buffers are already pinned in memory,
 627  * the fork switch can be done before changing the owner as we won't need to
 628  * validate the owner until the btree buffers are unpinned and writes can occur
 629  * again.
 630  *
 631  * For recovery based ownership change, there is no transactional context and
 632  * so a buffer list must be supplied so that we can record the buffers that we
 633  * modified for the caller to issue IO on.
 634  */
 635 int
 636 xfs_bmbt_change_owner(
 637         struct xfs_trans        *tp,
 638         struct xfs_inode        *ip,
 639         int                     whichfork,
 640         xfs_ino_t               new_owner,
 641         struct list_head        *buffer_list)
 642 {
 643         struct xfs_btree_cur    *cur;
 644         int                     error;
 645
 646         ASSERT(tp || buffer_list);
 647         ASSERT(!(tp && buffer_list));
 648         if (whichfork == XFS_DATA_FORK)
 649                 ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
 650         else
 651                 ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
 652
 653         cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
 654         if (!cur)
 655                 return -ENOMEM;
 656         cur->bc_private.b.flags |= XFS_BTCUR_BPRV_INVALID_OWNER;
 657
 658         error = xfs_btree_change_owner(cur, new_owner, buffer_list);
 659         xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 660         return error;
 661 }