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 /*
  72  * Convert a compressed bmap extent record to an uncompressed form.
  73  * This code must be in sync with the routines xfs_bmbt_get_startoff,
  74  * xfs_bmbt_get_startblock and xfs_bmbt_get_blockcount.
  75  */
  76 STATIC void
  77 __xfs_bmbt_get_all(
  78                 uint64_t l0,
  79                 uint64_t l1,
  80                 xfs_bmbt_irec_t *s)
  81 {
  82         int     ext_flag;
  83         xfs_exntst_t st;
  84
  85         ext_flag = (int)(l0 >> (64 - BMBT_EXNTFLAG_BITLEN));
  86         s->br_startoff = ((xfs_fileoff_t)l0 &
  87                            xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
  88         s->br_startblock = (((xfs_fsblock_t)l0 & xfs_mask64lo(9)) << 43) |
  89                            (((xfs_fsblock_t)l1) >> 21);
  90         s->br_blockcount = (xfs_filblks_t)(l1 & xfs_mask64lo(21));
  91         /* This is xfs_extent_state() in-line */
  92         if (ext_flag) {
  93                 ASSERT(s->br_blockcount != 0);  /* saved for DMIG */
  94                 st = XFS_EXT_UNWRITTEN;
  95         } else
  96                 st = XFS_EXT_NORM;
  97         s->br_state = st;
  98 }
  99
 100 void
 101 xfs_bmbt_get_all(
 102         xfs_bmbt_rec_host_t *r,
 103         xfs_bmbt_irec_t *s)
 104 {
 105         __xfs_bmbt_get_all(r->l0, r->l1, s);
 106 }
 107
 108 /*
 109  * Extract the blockcount field from an in memory bmap extent record.
 110  */
 111 xfs_filblks_t
 112 xfs_bmbt_get_blockcount(
 113         xfs_bmbt_rec_host_t     *r)
 114 {
 115         return (xfs_filblks_t)(r->l1 & xfs_mask64lo(21));
 116 }
 117
 118 /*
 119  * Extract the startblock field from an in memory bmap extent record.
 120  */
 121 xfs_fsblock_t
 122 xfs_bmbt_get_startblock(
 123         xfs_bmbt_rec_host_t     *r)
 124 {
 125         return (((xfs_fsblock_t)r->l0 & xfs_mask64lo(9)) << 43) |
 126                (((xfs_fsblock_t)r->l1) >> 21);
 127 }
 128
 129 /*
 130  * Extract the startoff field from an in memory bmap extent record.
 131  */
 132 xfs_fileoff_t
 133 xfs_bmbt_get_startoff(
 134         xfs_bmbt_rec_host_t     *r)
 135 {
 136         return ((xfs_fileoff_t)r->l0 &
 137                  xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
 138 }
 139
 140 /*
 141  * Extract the blockcount field from an on disk bmap extent record.
 142  */
 143 xfs_filblks_t
 144 xfs_bmbt_disk_get_blockcount(
 145         xfs_bmbt_rec_t  *r)
 146 {
 147         return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
 148 }
 149
 150 /*
 151  * Extract the startoff field from a disk format bmap extent record.
 152  */
 153 xfs_fileoff_t
 154 xfs_bmbt_disk_get_startoff(
 155         xfs_bmbt_rec_t  *r)
 156 {
 157         return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
 158                  xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
 159 }
 160
 161 /*
 162  * Set all the fields in a bmap extent record from the uncompressed form.
 163  */
 164 void
 165 xfs_bmbt_set_all(
 166         struct xfs_bmbt_rec_host *r,
 167         struct xfs_bmbt_irec    *s)
 168 {
 169         int                     extent_flag = (s->br_state != XFS_EXT_NORM);
 170
 171         ASSERT(s->br_state == XFS_EXT_NORM || s->br_state == XFS_EXT_UNWRITTEN);
 172         ASSERT(!(s->br_startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)));
 173         ASSERT(!(s->br_blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)));
 174         ASSERT(!(s->br_startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)));
 175
 176         r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
 177                  ((xfs_bmbt_rec_base_t)s->br_startoff << 9) |
 178                  ((xfs_bmbt_rec_base_t)s->br_startblock >> 43);
 179         r->l1 = ((xfs_bmbt_rec_base_t)s->br_startblock << 21) |
 180                  ((xfs_bmbt_rec_base_t)s->br_blockcount &
 181                   (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
 182 }
 183
 184 /*
 185  * Set all the fields in a bmap extent record from the uncompressed form.
 186  */
 187 void
 188 xfs_bmbt_disk_set_all(
 189         struct xfs_bmbt_rec     *r,
 190         struct xfs_bmbt_irec    *s)
 191 {
 192         int                     extent_flag = (s->br_state != XFS_EXT_NORM);
 193
 194         ASSERT(s->br_state == XFS_EXT_NORM || s->br_state == XFS_EXT_UNWRITTEN);
 195         ASSERT(!(s->br_startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)));
 196         ASSERT(!(s->br_blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)));
 197         ASSERT(!(s->br_startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)));
 198
 199         put_unaligned_be64(
 200                 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
 201                  ((xfs_bmbt_rec_base_t)s->br_startoff << 9) |
 202                  ((xfs_bmbt_rec_base_t)s->br_startblock >> 43), &r->l0);
 203         put_unaligned_be64(
 204                 ((xfs_bmbt_rec_base_t)s->br_startblock << 21) |
 205                  ((xfs_bmbt_rec_base_t)s->br_blockcount &
 206                   (xfs_bmbt_rec_base_t)xfs_mask64lo(21)), &r->l1);
 207 }
 208
 209 /*
 210  * Convert in-memory form of btree root to on-disk form.
 211  */
 212 void
 213 xfs_bmbt_to_bmdr(
 214         struct xfs_mount        *mp,
 215         struct xfs_btree_block  *rblock,
 216         int                     rblocklen,
 217         xfs_bmdr_block_t        *dblock,
 218         int                     dblocklen)
 219 {
 220         int                     dmxr;
 221         xfs_bmbt_key_t          *fkp;
 222         __be64                  *fpp;
 223         xfs_bmbt_key_t          *tkp;
 224         __be64                  *tpp;
 225
 226         if (xfs_sb_version_hascrc(&mp->m_sb)) {
 227                 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
 228                 ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
 229                        &mp->m_sb.sb_meta_uuid));
 230                 ASSERT(rblock->bb_u.l.bb_blkno ==
 231                        cpu_to_be64(XFS_BUF_DADDR_NULL));
 232         } else
 233                 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
 234         ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
 235         ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
 236         ASSERT(rblock->bb_level != 0);
 237         dblock->bb_level = rblock->bb_level;
 238         dblock->bb_numrecs = rblock->bb_numrecs;
 239         dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
 240         fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
 241         tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
 242         fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
 243         tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
 244         dmxr = be16_to_cpu(dblock->bb_numrecs);
 245         memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
 246         memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
 247 }
 248
 249 STATIC struct xfs_btree_cur *
 250 xfs_bmbt_dup_cursor(
 251         struct xfs_btree_cur    *cur)
 252 {
 253         struct xfs_btree_cur    *new;
 254
 255         new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
 256                         cur->bc_private.b.ip, cur->bc_private.b.whichfork);
 257
 258         /*
 259          * Copy the firstblock, dfops, and flags values,
 260          * since init cursor doesn't get them.
 261          */
 262         new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
 263         new->bc_private.b.dfops = cur->bc_private.b.dfops;
 264         new->bc_private.b.flags = cur->bc_private.b.flags;
 265
 266         return new;
 267 }
 268
 269 STATIC void
 270 xfs_bmbt_update_cursor(
 271         struct xfs_btree_cur    *src,
 272         struct xfs_btree_cur    *dst)
 273 {
 274         ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
 275                (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
 276         ASSERT(dst->bc_private.b.dfops == src->bc_private.b.dfops);
 277
 278         dst->bc_private.b.allocated += src->bc_private.b.allocated;
 279         dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
 280
 281         src->bc_private.b.allocated = 0;
 282 }
 283
 284 STATIC int
 285 xfs_bmbt_alloc_block(
 286         struct xfs_btree_cur    *cur,
 287         union xfs_btree_ptr     *start,
 288         union xfs_btree_ptr     *new,
 289         int                     *stat)
 290 {
 291         xfs_alloc_arg_t         args;           /* block allocation args */
 292         int                     error;          /* error return value */
 293
 294         memset(&args, 0, sizeof(args));
 295         args.tp = cur->bc_tp;
 296         args.mp = cur->bc_mp;
 297         args.fsbno = cur->bc_private.b.firstblock;
 298         args.firstblock = args.fsbno;
 299         xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_private.b.ip->i_ino,
 300                         cur->bc_private.b.whichfork);
 301
 302         if (args.fsbno == NULLFSBLOCK) {
 303                 args.fsbno = be64_to_cpu(start->l);
 304                 args.type = XFS_ALLOCTYPE_START_BNO;
 305                 /*
 306                  * Make sure there is sufficient room left in the AG to
 307                  * complete a full tree split for an extent insert.  If
 308                  * we are converting the middle part of an extent then
 309                  * we may need space for two tree splits.
 310                  *
 311                  * We are relying on the caller to make the correct block
 312                  * reservation for this operation to succeed.  If the
 313                  * reservation amount is insufficient then we may fail a
 314                  * block allocation here and corrupt the filesystem.
 315                  */
 316                 args.minleft = args.tp->t_blk_res;
 317         } else if (cur->bc_private.b.dfops->dop_low) {
 318                 args.type = XFS_ALLOCTYPE_START_BNO;
 319         } else {
 320                 args.type = XFS_ALLOCTYPE_NEAR_BNO;
 321         }
 322
 323         args.minlen = args.maxlen = args.prod = 1;
 324         args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
 325         if (!args.wasdel && args.tp->t_blk_res == 0) {
 326                 error = -ENOSPC;
 327                 goto error0;
 328         }
 329         error = xfs_alloc_vextent(&args);
 330         if (error)
 331                 goto error0;
 332
 333         if (args.fsbno == NULLFSBLOCK && args.minleft) {
 334                 /*
 335                  * Could not find an AG with enough free space to satisfy
 336                  * a full btree split.  Try again and if
 337                  * successful activate the lowspace algorithm.
 338                  */
 339                 args.fsbno = 0;
 340                 args.type = XFS_ALLOCTYPE_FIRST_AG;
 341                 error = xfs_alloc_vextent(&args);
 342                 if (error)
 343                         goto error0;
 344                 cur->bc_private.b.dfops->dop_low = true;
 345         }
 346         if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
 347                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 348                 *stat = 0;
 349                 return 0;
 350         }
 351         ASSERT(args.len == 1);
 352         cur->bc_private.b.firstblock = args.fsbno;
 353         cur->bc_private.b.allocated++;
 354         cur->bc_private.b.ip->i_d.di_nblocks++;
 355         xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
 356         xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
 357                         XFS_TRANS_DQ_BCOUNT, 1L);
 358
 359         new->l = cpu_to_be64(args.fsbno);
 360
 361         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 362         *stat = 1;
 363         return 0;
 364
 365  error0:
 366         XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 367         return error;
 368 }
 369
 370 STATIC int
 371 xfs_bmbt_free_block(
 372         struct xfs_btree_cur    *cur,
 373         struct xfs_buf          *bp)
 374 {
 375         struct xfs_mount        *mp = cur->bc_mp;
 376         struct xfs_inode        *ip = cur->bc_private.b.ip;
 377         struct xfs_trans        *tp = cur->bc_tp;
 378         xfs_fsblock_t           fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
 379         struct xfs_owner_info   oinfo;
 380
 381         xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, cur->bc_private.b.whichfork);
 382         xfs_bmap_add_free(mp, cur->bc_private.b.dfops, fsbno, 1, &oinfo);
 383         ip->i_d.di_nblocks--;
 384
 385         xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 386         xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
 387         return 0;
 388 }
 389
 390 STATIC int
 391 xfs_bmbt_get_minrecs(
 392         struct xfs_btree_cur    *cur,
 393         int                     level)
 394 {
 395         if (level == cur->bc_nlevels - 1) {
 396                 struct xfs_ifork        *ifp;
 397
 398                 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
 399                                     cur->bc_private.b.whichfork);
 400
 401                 return xfs_bmbt_maxrecs(cur->bc_mp,
 402                                         ifp->if_broot_bytes, level == 0) / 2;
 403         }
 404
 405         return cur->bc_mp->m_bmap_dmnr[level != 0];
 406 }
 407
 408 int
 409 xfs_bmbt_get_maxrecs(
 410         struct xfs_btree_cur    *cur,
 411         int                     level)
 412 {
 413         if (level == cur->bc_nlevels - 1) {
 414                 struct xfs_ifork        *ifp;
 415
 416                 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
 417                                     cur->bc_private.b.whichfork);
 418
 419                 return xfs_bmbt_maxrecs(cur->bc_mp,
 420                                         ifp->if_broot_bytes, level == 0);
 421         }
 422
 423         return cur->bc_mp->m_bmap_dmxr[level != 0];
 424
 425 }
 426
 427 /*
 428  * Get the maximum records we could store in the on-disk format.
 429  *
 430  * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
 431  * for the root node this checks the available space in the dinode fork
 432  * so that we can resize the in-memory buffer to match it.  After a
 433  * resize to the maximum size this function returns the same value
 434  * as xfs_bmbt_get_maxrecs for the root node, too.
 435  */
 436 STATIC int
 437 xfs_bmbt_get_dmaxrecs(
 438         struct xfs_btree_cur    *cur,
 439         int                     level)
 440 {
 441         if (level != cur->bc_nlevels - 1)
 442                 return cur->bc_mp->m_bmap_dmxr[level != 0];
 443         return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
 444 }
 445
 446 STATIC void
 447 xfs_bmbt_init_key_from_rec(
 448         union xfs_btree_key     *key,
 449         union xfs_btree_rec     *rec)
 450 {
 451         key->bmbt.br_startoff =
 452                 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
 453 }
 454
 455 STATIC void
 456 xfs_bmbt_init_high_key_from_rec(
 457         union xfs_btree_key     *key,
 458         union xfs_btree_rec     *rec)
 459 {
 460         key->bmbt.br_startoff = cpu_to_be64(
 461                         xfs_bmbt_disk_get_startoff(&rec->bmbt) +
 462                         xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
 463 }
 464
 465 STATIC void
 466 xfs_bmbt_init_rec_from_cur(
 467         struct xfs_btree_cur    *cur,
 468         union xfs_btree_rec     *rec)
 469 {
 470         xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
 471 }
 472
 473 STATIC void
 474 xfs_bmbt_init_ptr_from_cur(
 475         struct xfs_btree_cur    *cur,
 476         union xfs_btree_ptr     *ptr)
 477 {
 478         ptr->l = 0;
 479 }
 480
 481 STATIC int64_t
 482 xfs_bmbt_key_diff(
 483         struct xfs_btree_cur    *cur,
 484         union xfs_btree_key     *key)
 485 {
 486         return (int64_t)be64_to_cpu(key->bmbt.br_startoff) -
 487                                       cur->bc_rec.b.br_startoff;
 488 }
 489
 490 STATIC int64_t
 491 xfs_bmbt_diff_two_keys(
 492         struct xfs_btree_cur    *cur,
 493         union xfs_btree_key     *k1,
 494         union xfs_btree_key     *k2)
 495 {
 496         return (int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
 497                           be64_to_cpu(k2->bmbt.br_startoff);
 498 }
 499
 500 static bool
 501 xfs_bmbt_verify(
 502         struct xfs_buf          *bp)
 503 {
 504         struct xfs_mount        *mp = bp->b_target->bt_mount;
 505         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 506         unsigned int            level;
 507
 508         switch (block->bb_magic) {
 509         case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
 510                 if (!xfs_sb_version_hascrc(&mp->m_sb))
 511                         return false;
 512                 if (!uuid_equal(&block->bb_u.l.bb_uuid, &mp->m_sb.sb_meta_uuid))
 513                         return false;
 514                 if (be64_to_cpu(block->bb_u.l.bb_blkno) != bp->b_bn)
 515                         return false;
 516                 /*
 517                  * XXX: need a better way of verifying the owner here. Right now
 518                  * just make sure there has been one set.
 519                  */
 520                 if (be64_to_cpu(block->bb_u.l.bb_owner) == 0)
 521                         return false;
 522                 /* fall through */
 523         case cpu_to_be32(XFS_BMAP_MAGIC):
 524                 break;
 525         default:
 526                 return false;
 527         }
 528
 529         /*
 530          * numrecs and level verification.
 531          *
 532          * We don't know what fork we belong to, so just verify that the level
 533          * is less than the maximum of the two. Later checks will be more
 534          * precise.
 535          */
 536         level = be16_to_cpu(block->bb_level);
 537         if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
 538                 return false;
 539         if (be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
 540                 return false;
 541
 542         /* sibling pointer verification */
 543         if (!block->bb_u.l.bb_leftsib ||
 544             (block->bb_u.l.bb_leftsib != cpu_to_be64(NULLFSBLOCK) &&
 545              !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_leftsib))))
 546                 return false;
 547         if (!block->bb_u.l.bb_rightsib ||
 548             (block->bb_u.l.bb_rightsib != cpu_to_be64(NULLFSBLOCK) &&
 549              !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_rightsib))))
 550                 return false;
 551
 552         return true;
 553 }
 554
 555 static void
 556 xfs_bmbt_read_verify(
 557         struct xfs_buf  *bp)
 558 {
 559         if (!xfs_btree_lblock_verify_crc(bp))
 560                 xfs_buf_ioerror(bp, -EFSBADCRC);
 561         else if (!xfs_bmbt_verify(bp))
 562                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 563
 564         if (bp->b_error) {
 565                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 566                 xfs_verifier_error(bp);
 567         }
 568 }
 569
 570 static void
 571 xfs_bmbt_write_verify(
 572         struct xfs_buf  *bp)
 573 {
 574         if (!xfs_bmbt_verify(bp)) {
 575                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 576                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 577                 xfs_verifier_error(bp);
 578                 return;
 579         }
 580         xfs_btree_lblock_calc_crc(bp);
 581 }
 582
 583 const struct xfs_buf_ops xfs_bmbt_buf_ops = {
 584         .name = "xfs_bmbt",
 585         .verify_read = xfs_bmbt_read_verify,
 586         .verify_write = xfs_bmbt_write_verify,
 587 };
 588
 589
 590 STATIC int
 591 xfs_bmbt_keys_inorder(
 592         struct xfs_btree_cur    *cur,
 593         union xfs_btree_key     *k1,
 594         union xfs_btree_key     *k2)
 595 {
 596         return be64_to_cpu(k1->bmbt.br_startoff) <
 597                 be64_to_cpu(k2->bmbt.br_startoff);
 598 }
 599
 600 STATIC int
 601 xfs_bmbt_recs_inorder(
 602         struct xfs_btree_cur    *cur,
 603         union xfs_btree_rec     *r1,
 604         union xfs_btree_rec     *r2)
 605 {
 606         return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
 607                 xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
 608                 xfs_bmbt_disk_get_startoff(&r2->bmbt);
 609 }
 610
 611 static const struct xfs_btree_ops xfs_bmbt_ops = {
 612         .rec_len                = sizeof(xfs_bmbt_rec_t),
 613         .key_len                = sizeof(xfs_bmbt_key_t),
 614
 615         .dup_cursor             = xfs_bmbt_dup_cursor,
 616         .update_cursor          = xfs_bmbt_update_cursor,
 617         .alloc_block            = xfs_bmbt_alloc_block,
 618         .free_block             = xfs_bmbt_free_block,
 619         .get_maxrecs            = xfs_bmbt_get_maxrecs,
 620         .get_minrecs            = xfs_bmbt_get_minrecs,
 621         .get_dmaxrecs           = xfs_bmbt_get_dmaxrecs,
 622         .init_key_from_rec      = xfs_bmbt_init_key_from_rec,
 623         .init_high_key_from_rec = xfs_bmbt_init_high_key_from_rec,
 624         .init_rec_from_cur      = xfs_bmbt_init_rec_from_cur,
 625         .init_ptr_from_cur      = xfs_bmbt_init_ptr_from_cur,
 626         .key_diff               = xfs_bmbt_key_diff,
 627         .diff_two_keys          = xfs_bmbt_diff_two_keys,
 628         .buf_ops                = &xfs_bmbt_buf_ops,
 629         .keys_inorder           = xfs_bmbt_keys_inorder,
 630         .recs_inorder           = xfs_bmbt_recs_inorder,
 631 };
 632
 633 /*
 634  * Allocate a new bmap btree cursor.
 635  */
 636 struct xfs_btree_cur *                          /* new bmap btree cursor */
 637 xfs_bmbt_init_cursor(
 638         struct xfs_mount        *mp,            /* file system mount point */
 639         struct xfs_trans        *tp,            /* transaction pointer */
 640         struct xfs_inode        *ip,            /* inode owning the btree */
 641         int                     whichfork)      /* data or attr fork */
 642 {
 643         struct xfs_ifork        *ifp = XFS_IFORK_PTR(ip, whichfork);
 644         struct xfs_btree_cur    *cur;
 645         ASSERT(whichfork != XFS_COW_FORK);
 646
 647         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
 648
 649         cur->bc_tp = tp;
 650         cur->bc_mp = mp;
 651         cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
 652         cur->bc_btnum = XFS_BTNUM_BMAP;
 653         cur->bc_blocklog = mp->m_sb.sb_blocklog;
 654         cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
 655
 656         cur->bc_ops = &xfs_bmbt_ops;
 657         cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
 658         if (xfs_sb_version_hascrc(&mp->m_sb))
 659                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 660
 661         cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
 662         cur->bc_private.b.ip = ip;
 663         cur->bc_private.b.firstblock = NULLFSBLOCK;
 664         cur->bc_private.b.dfops = NULL;
 665         cur->bc_private.b.allocated = 0;
 666         cur->bc_private.b.flags = 0;
 667         cur->bc_private.b.whichfork = whichfork;
 668
 669         return cur;
 670 }
 671
 672 /*
 673  * Calculate number of records in a bmap btree block.
 674  */
 675 int
 676 xfs_bmbt_maxrecs(
 677         struct xfs_mount        *mp,
 678         int                     blocklen,
 679         int                     leaf)
 680 {
 681         blocklen -= XFS_BMBT_BLOCK_LEN(mp);
 682
 683         if (leaf)
 684                 return blocklen / sizeof(xfs_bmbt_rec_t);
 685         return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
 686 }
 687
 688 /*
 689  * Calculate number of records in a bmap btree inode root.
 690  */
 691 int
 692 xfs_bmdr_maxrecs(
 693         int                     blocklen,
 694         int                     leaf)
 695 {
 696         blocklen -= sizeof(xfs_bmdr_block_t);
 697
 698         if (leaf)
 699                 return blocklen / sizeof(xfs_bmdr_rec_t);
 700         return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
 701 }
 702
 703 /*
 704  * Change the owner of a btree format fork fo the inode passed in. Change it to
 705  * the owner of that is passed in so that we can change owners before or after
 706  * we switch forks between inodes. The operation that the caller is doing will
 707  * determine whether is needs to change owner before or after the switch.
 708  *
 709  * For demand paged transactional modification, the fork switch should be done
 710  * after reading in all the blocks, modifying them and pinning them in the
 711  * transaction. For modification when the buffers are already pinned in memory,
 712  * the fork switch can be done before changing the owner as we won't need to
 713  * validate the owner until the btree buffers are unpinned and writes can occur
 714  * again.
 715  *
 716  * For recovery based ownership change, there is no transactional context and
 717  * so a buffer list must be supplied so that we can record the buffers that we
 718  * modified for the caller to issue IO on.
 719  */
 720 int
 721 xfs_bmbt_change_owner(
 722         struct xfs_trans        *tp,
 723         struct xfs_inode        *ip,
 724         int                     whichfork,
 725         xfs_ino_t               new_owner,
 726         struct list_head        *buffer_list)
 727 {
 728         struct xfs_btree_cur    *cur;
 729         int                     error;
 730
 731         ASSERT(tp || buffer_list);
 732         ASSERT(!(tp && buffer_list));
 733         if (whichfork == XFS_DATA_FORK)
 734                 ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
 735         else
 736                 ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
 737
 738         cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
 739         if (!cur)
 740                 return -ENOMEM;
 741         cur->bc_private.b.flags |= XFS_BTCUR_BPRV_INVALID_OWNER;
 742
 743         error = xfs_btree_change_owner(cur, new_owner, buffer_list);
 744         xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 745         return error;
 746 }