fs/xfs/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 "xfs.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_inode.h"
  27 #include "xfs_trans.h"
  28 #include "xfs_inode_item.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_error.h"
  34 #include "xfs_quota.h"
  35 #include "xfs_trace.h"
  36 #include "xfs_cksum.h"
  37
  38 /*
  39  * Determine the extent state.
  40  */
  41 /* ARGSUSED */
  42 STATIC xfs_exntst_t
  43 xfs_extent_state(
  44         xfs_filblks_t           blks,
  45         int                     extent_flag)
  46 {
  47         if (extent_flag) {
  48                 ASSERT(blks != 0);      /* saved for DMIG */
  49                 return XFS_EXT_UNWRITTEN;
  50         }
  51         return XFS_EXT_NORM;
  52 }
  53
  54 /*
  55  * Convert on-disk form of btree root to in-memory form.
  56  */
  57 void
  58 xfs_bmdr_to_bmbt(
  59         struct xfs_inode        *ip,
  60         xfs_bmdr_block_t        *dblock,
  61         int                     dblocklen,
  62         struct xfs_btree_block  *rblock,
  63         int                     rblocklen)
  64 {
  65         struct xfs_mount        *mp = ip->i_mount;
  66         int                     dmxr;
  67         xfs_bmbt_key_t          *fkp;
  68         __be64                  *fpp;
  69         xfs_bmbt_key_t          *tkp;
  70         __be64                  *tpp;
  71
  72         if (xfs_sb_version_hascrc(&mp->m_sb))
  73                 xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
  74                                  XFS_BMAP_CRC_MAGIC, 0, 0, ip->i_ino,
  75                                  XFS_BTREE_LONG_PTRS | XFS_BTREE_CRC_BLOCKS);
  76         else
  77                 xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
  78                                  XFS_BMAP_MAGIC, 0, 0, ip->i_ino,
  79                                  XFS_BTREE_LONG_PTRS);
  80
  81         rblock->bb_level = dblock->bb_level;
  82         ASSERT(be16_to_cpu(rblock->bb_level) > 0);
  83         rblock->bb_numrecs = dblock->bb_numrecs;
  84         dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
  85         fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
  86         tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
  87         fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
  88         tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
  89         dmxr = be16_to_cpu(dblock->bb_numrecs);
  90         memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
  91         memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
  92 }
  93
  94 /*
  95  * Convert a compressed bmap extent record to an uncompressed form.
  96  * This code must be in sync with the routines xfs_bmbt_get_startoff,
  97  * xfs_bmbt_get_startblock, xfs_bmbt_get_blockcount and xfs_bmbt_get_state.
  98  */
  99 STATIC void
 100 __xfs_bmbt_get_all(
 101                 __uint64_t l0,
 102                 __uint64_t l1,
 103                 xfs_bmbt_irec_t *s)
 104 {
 105         int     ext_flag;
 106         xfs_exntst_t st;
 107
 108         ext_flag = (int)(l0 >> (64 - BMBT_EXNTFLAG_BITLEN));
 109         s->br_startoff = ((xfs_fileoff_t)l0 &
 110                            xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
 111         s->br_startblock = (((xfs_fsblock_t)l0 & xfs_mask64lo(9)) << 43) |
 112                            (((xfs_fsblock_t)l1) >> 21);
 113         s->br_blockcount = (xfs_filblks_t)(l1 & xfs_mask64lo(21));
 114         /* This is xfs_extent_state() in-line */
 115         if (ext_flag) {
 116                 ASSERT(s->br_blockcount != 0);  /* saved for DMIG */
 117                 st = XFS_EXT_UNWRITTEN;
 118         } else
 119                 st = XFS_EXT_NORM;
 120         s->br_state = st;
 121 }
 122
 123 void
 124 xfs_bmbt_get_all(
 125         xfs_bmbt_rec_host_t *r,
 126         xfs_bmbt_irec_t *s)
 127 {
 128         __xfs_bmbt_get_all(r->l0, r->l1, s);
 129 }
 130
 131 /*
 132  * Extract the blockcount field from an in memory bmap extent record.
 133  */
 134 xfs_filblks_t
 135 xfs_bmbt_get_blockcount(
 136         xfs_bmbt_rec_host_t     *r)
 137 {
 138         return (xfs_filblks_t)(r->l1 & xfs_mask64lo(21));
 139 }
 140
 141 /*
 142  * Extract the startblock field from an in memory bmap extent record.
 143  */
 144 xfs_fsblock_t
 145 xfs_bmbt_get_startblock(
 146         xfs_bmbt_rec_host_t     *r)
 147 {
 148         return (((xfs_fsblock_t)r->l0 & xfs_mask64lo(9)) << 43) |
 149                (((xfs_fsblock_t)r->l1) >> 21);
 150 }
 151
 152 /*
 153  * Extract the startoff field from an in memory bmap extent record.
 154  */
 155 xfs_fileoff_t
 156 xfs_bmbt_get_startoff(
 157         xfs_bmbt_rec_host_t     *r)
 158 {
 159         return ((xfs_fileoff_t)r->l0 &
 160                  xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
 161 }
 162
 163 xfs_exntst_t
 164 xfs_bmbt_get_state(
 165         xfs_bmbt_rec_host_t     *r)
 166 {
 167         int     ext_flag;
 168
 169         ext_flag = (int)((r->l0) >> (64 - BMBT_EXNTFLAG_BITLEN));
 170         return xfs_extent_state(xfs_bmbt_get_blockcount(r),
 171                                 ext_flag);
 172 }
 173
 174 /*
 175  * Extract the blockcount field from an on disk bmap extent record.
 176  */
 177 xfs_filblks_t
 178 xfs_bmbt_disk_get_blockcount(
 179         xfs_bmbt_rec_t  *r)
 180 {
 181         return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
 182 }
 183
 184 /*
 185  * Extract the startoff field from a disk format bmap extent record.
 186  */
 187 xfs_fileoff_t
 188 xfs_bmbt_disk_get_startoff(
 189         xfs_bmbt_rec_t  *r)
 190 {
 191         return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
 192                  xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
 193 }
 194
 195
 196 /*
 197  * Set all the fields in a bmap extent record from the arguments.
 198  */
 199 void
 200 xfs_bmbt_set_allf(
 201         xfs_bmbt_rec_host_t     *r,
 202         xfs_fileoff_t           startoff,
 203         xfs_fsblock_t           startblock,
 204         xfs_filblks_t           blockcount,
 205         xfs_exntst_t            state)
 206 {
 207         int             extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
 208
 209         ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
 210         ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
 211         ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
 212
 213         ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
 214
 215         r->l0 = ((xfs_bmbt_rec_base_t)extent_flag << 63) |
 216                 ((xfs_bmbt_rec_base_t)startoff << 9) |
 217                 ((xfs_bmbt_rec_base_t)startblock >> 43);
 218         r->l1 = ((xfs_bmbt_rec_base_t)startblock << 21) |
 219                 ((xfs_bmbt_rec_base_t)blockcount &
 220                 (xfs_bmbt_rec_base_t)xfs_mask64lo(21));
 221 }
 222
 223 /*
 224  * Set all the fields in a bmap extent record from the uncompressed form.
 225  */
 226 void
 227 xfs_bmbt_set_all(
 228         xfs_bmbt_rec_host_t *r,
 229         xfs_bmbt_irec_t *s)
 230 {
 231         xfs_bmbt_set_allf(r, s->br_startoff, s->br_startblock,
 232                              s->br_blockcount, s->br_state);
 233 }
 234
 235
 236 /*
 237  * Set all the fields in a disk format bmap extent record from the arguments.
 238  */
 239 void
 240 xfs_bmbt_disk_set_allf(
 241         xfs_bmbt_rec_t          *r,
 242         xfs_fileoff_t           startoff,
 243         xfs_fsblock_t           startblock,
 244         xfs_filblks_t           blockcount,
 245         xfs_exntst_t            state)
 246 {
 247         int                     extent_flag = (state == XFS_EXT_NORM) ? 0 : 1;
 248
 249         ASSERT(state == XFS_EXT_NORM || state == XFS_EXT_UNWRITTEN);
 250         ASSERT((startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)) == 0);
 251         ASSERT((blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)) == 0);
 252         ASSERT((startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)) == 0);
 253
 254         r->l0 = cpu_to_be64(
 255                 ((xfs_bmbt_rec_base_t)extent_flag << 63) |
 256                  ((xfs_bmbt_rec_base_t)startoff << 9) |
 257                  ((xfs_bmbt_rec_base_t)startblock >> 43));
 258         r->l1 = cpu_to_be64(
 259                 ((xfs_bmbt_rec_base_t)startblock << 21) |
 260                  ((xfs_bmbt_rec_base_t)blockcount &
 261                   (xfs_bmbt_rec_base_t)xfs_mask64lo(21)));
 262 }
 263
 264 /*
 265  * Set all the fields in a bmap extent record from the uncompressed form.
 266  */
 267 STATIC void
 268 xfs_bmbt_disk_set_all(
 269         xfs_bmbt_rec_t  *r,
 270         xfs_bmbt_irec_t *s)
 271 {
 272         xfs_bmbt_disk_set_allf(r, s->br_startoff, s->br_startblock,
 273                                   s->br_blockcount, s->br_state);
 274 }
 275
 276 /*
 277  * Set the blockcount field in a bmap extent record.
 278  */
 279 void
 280 xfs_bmbt_set_blockcount(
 281         xfs_bmbt_rec_host_t *r,
 282         xfs_filblks_t   v)
 283 {
 284         ASSERT((v & xfs_mask64hi(43)) == 0);
 285         r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64hi(43)) |
 286                   (xfs_bmbt_rec_base_t)(v & xfs_mask64lo(21));
 287 }
 288
 289 /*
 290  * Set the startblock field in a bmap extent record.
 291  */
 292 void
 293 xfs_bmbt_set_startblock(
 294         xfs_bmbt_rec_host_t *r,
 295         xfs_fsblock_t   v)
 296 {
 297         ASSERT((v & xfs_mask64hi(12)) == 0);
 298         r->l0 = (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64hi(55)) |
 299                   (xfs_bmbt_rec_base_t)(v >> 43);
 300         r->l1 = (r->l1 & (xfs_bmbt_rec_base_t)xfs_mask64lo(21)) |
 301                   (xfs_bmbt_rec_base_t)(v << 21);
 302 }
 303
 304 /*
 305  * Set the startoff field in a bmap extent record.
 306  */
 307 void
 308 xfs_bmbt_set_startoff(
 309         xfs_bmbt_rec_host_t *r,
 310         xfs_fileoff_t   v)
 311 {
 312         ASSERT((v & xfs_mask64hi(9)) == 0);
 313         r->l0 = (r->l0 & (xfs_bmbt_rec_base_t) xfs_mask64hi(1)) |
 314                 ((xfs_bmbt_rec_base_t)v << 9) |
 315                   (r->l0 & (xfs_bmbt_rec_base_t)xfs_mask64lo(9));
 316 }
 317
 318 /*
 319  * Set the extent state field in a bmap extent record.
 320  */
 321 void
 322 xfs_bmbt_set_state(
 323         xfs_bmbt_rec_host_t *r,
 324         xfs_exntst_t    v)
 325 {
 326         ASSERT(v == XFS_EXT_NORM || v == XFS_EXT_UNWRITTEN);
 327         if (v == XFS_EXT_NORM)
 328                 r->l0 &= xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN);
 329         else
 330                 r->l0 |= xfs_mask64hi(BMBT_EXNTFLAG_BITLEN);
 331 }
 332
 333 /*
 334  * Convert in-memory form of btree root to on-disk form.
 335  */
 336 void
 337 xfs_bmbt_to_bmdr(
 338         struct xfs_mount        *mp,
 339         struct xfs_btree_block  *rblock,
 340         int                     rblocklen,
 341         xfs_bmdr_block_t        *dblock,
 342         int                     dblocklen)
 343 {
 344         int                     dmxr;
 345         xfs_bmbt_key_t          *fkp;
 346         __be64                  *fpp;
 347         xfs_bmbt_key_t          *tkp;
 348         __be64                  *tpp;
 349
 350         if (xfs_sb_version_hascrc(&mp->m_sb)) {
 351                 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
 352                 ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
 353                        &mp->m_sb.sb_meta_uuid));
 354                 ASSERT(rblock->bb_u.l.bb_blkno ==
 355                        cpu_to_be64(XFS_BUF_DADDR_NULL));
 356         } else
 357                 ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
 358         ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
 359         ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
 360         ASSERT(rblock->bb_level != 0);
 361         dblock->bb_level = rblock->bb_level;
 362         dblock->bb_numrecs = rblock->bb_numrecs;
 363         dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
 364         fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
 365         tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
 366         fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
 367         tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
 368         dmxr = be16_to_cpu(dblock->bb_numrecs);
 369         memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
 370         memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
 371 }
 372
 373 /*
 374  * Check extent records, which have just been read, for
 375  * any bit in the extent flag field. ASSERT on debug
 376  * kernels, as this condition should not occur.
 377  * Return an error condition (1) if any flags found,
 378  * otherwise return 0.
 379  */
 380
 381 int
 382 xfs_check_nostate_extents(
 383         xfs_ifork_t             *ifp,
 384         xfs_extnum_t            idx,
 385         xfs_extnum_t            num)
 386 {
 387         for (; num > 0; num--, idx++) {
 388                 xfs_bmbt_rec_host_t *ep = xfs_iext_get_ext(ifp, idx);
 389                 if ((ep->l0 >>
 390                      (64 - BMBT_EXNTFLAG_BITLEN)) != 0) {
 391                         ASSERT(0);
 392                         return 1;
 393                 }
 394         }
 395         return 0;
 396 }
 397
 398
 399 STATIC struct xfs_btree_cur *
 400 xfs_bmbt_dup_cursor(
 401         struct xfs_btree_cur    *cur)
 402 {
 403         struct xfs_btree_cur    *new;
 404
 405         new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
 406                         cur->bc_private.b.ip, cur->bc_private.b.whichfork);
 407
 408         /*
 409          * Copy the firstblock, flist, and flags values,
 410          * since init cursor doesn't get them.
 411          */
 412         new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
 413         new->bc_private.b.flist = cur->bc_private.b.flist;
 414         new->bc_private.b.flags = cur->bc_private.b.flags;
 415
 416         return new;
 417 }
 418
 419 STATIC void
 420 xfs_bmbt_update_cursor(
 421         struct xfs_btree_cur    *src,
 422         struct xfs_btree_cur    *dst)
 423 {
 424         ASSERT((dst->bc_private.b.firstblock != NULLFSBLOCK) ||
 425                (dst->bc_private.b.ip->i_d.di_flags & XFS_DIFLAG_REALTIME));
 426         ASSERT(dst->bc_private.b.flist == src->bc_private.b.flist);
 427
 428         dst->bc_private.b.allocated += src->bc_private.b.allocated;
 429         dst->bc_private.b.firstblock = src->bc_private.b.firstblock;
 430
 431         src->bc_private.b.allocated = 0;
 432 }
 433
 434 STATIC int
 435 xfs_bmbt_alloc_block(
 436         struct xfs_btree_cur    *cur,
 437         union xfs_btree_ptr     *start,
 438         union xfs_btree_ptr     *new,
 439         int                     *stat)
 440 {
 441         xfs_alloc_arg_t         args;           /* block allocation args */
 442         int                     error;          /* error return value */
 443
 444         memset(&args, 0, sizeof(args));
 445         args.tp = cur->bc_tp;
 446         args.mp = cur->bc_mp;
 447         args.fsbno = cur->bc_private.b.firstblock;
 448         args.firstblock = args.fsbno;
 449
 450         if (args.fsbno == NULLFSBLOCK) {
 451                 args.fsbno = be64_to_cpu(start->l);
 452                 args.type = XFS_ALLOCTYPE_START_BNO;
 453                 /*
 454                  * Make sure there is sufficient room left in the AG to
 455                  * complete a full tree split for an extent insert.  If
 456                  * we are converting the middle part of an extent then
 457                  * we may need space for two tree splits.
 458                  *
 459                  * We are relying on the caller to make the correct block
 460                  * reservation for this operation to succeed.  If the
 461                  * reservation amount is insufficient then we may fail a
 462                  * block allocation here and corrupt the filesystem.
 463                  */
 464                 args.minleft = xfs_trans_get_block_res(args.tp);
 465         } else if (cur->bc_private.b.flist->xbf_low) {
 466                 args.type = XFS_ALLOCTYPE_START_BNO;
 467         } else {
 468                 args.type = XFS_ALLOCTYPE_NEAR_BNO;
 469         }
 470
 471         args.minlen = args.maxlen = args.prod = 1;
 472         args.wasdel = cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL;
 473         if (!args.wasdel && xfs_trans_get_block_res(args.tp) == 0) {
 474                 error = -ENOSPC;
 475                 goto error0;
 476         }
 477         error = xfs_alloc_vextent(&args);
 478         if (error)
 479                 goto error0;
 480
 481         if (args.fsbno == NULLFSBLOCK && args.minleft) {
 482                 /*
 483                  * Could not find an AG with enough free space to satisfy
 484                  * a full btree split.  Try again without minleft and if
 485                  * successful activate the lowspace algorithm.
 486                  */
 487                 args.fsbno = 0;
 488                 args.type = XFS_ALLOCTYPE_FIRST_AG;
 489                 args.minleft = 0;
 490                 error = xfs_alloc_vextent(&args);
 491                 if (error)
 492                         goto error0;
 493                 cur->bc_private.b.flist->xbf_low = 1;
 494         }
 495         if (args.fsbno == NULLFSBLOCK) {
 496                 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 497                 *stat = 0;
 498                 return 0;
 499         }
 500         ASSERT(args.len == 1);
 501         cur->bc_private.b.firstblock = args.fsbno;
 502         cur->bc_private.b.allocated++;
 503         cur->bc_private.b.ip->i_d.di_nblocks++;
 504         xfs_trans_log_inode(args.tp, cur->bc_private.b.ip, XFS_ILOG_CORE);
 505         xfs_trans_mod_dquot_byino(args.tp, cur->bc_private.b.ip,
 506                         XFS_TRANS_DQ_BCOUNT, 1L);
 507
 508         new->l = cpu_to_be64(args.fsbno);
 509
 510         XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 511         *stat = 1;
 512         return 0;
 513
 514  error0:
 515         XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 516         return error;
 517 }
 518
 519 STATIC int
 520 xfs_bmbt_free_block(
 521         struct xfs_btree_cur    *cur,
 522         struct xfs_buf          *bp)
 523 {
 524         struct xfs_mount        *mp = cur->bc_mp;
 525         struct xfs_inode        *ip = cur->bc_private.b.ip;
 526         struct xfs_trans        *tp = cur->bc_tp;
 527         xfs_fsblock_t           fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
 528
 529         xfs_bmap_add_free(fsbno, 1, cur->bc_private.b.flist, mp);
 530         ip->i_d.di_nblocks--;
 531
 532         xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 533         xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
 534         xfs_trans_binval(tp, bp);
 535         return 0;
 536 }
 537
 538 STATIC int
 539 xfs_bmbt_get_minrecs(
 540         struct xfs_btree_cur    *cur,
 541         int                     level)
 542 {
 543         if (level == cur->bc_nlevels - 1) {
 544                 struct xfs_ifork        *ifp;
 545
 546                 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
 547                                     cur->bc_private.b.whichfork);
 548
 549                 return xfs_bmbt_maxrecs(cur->bc_mp,
 550                                         ifp->if_broot_bytes, level == 0) / 2;
 551         }
 552
 553         return cur->bc_mp->m_bmap_dmnr[level != 0];
 554 }
 555
 556 int
 557 xfs_bmbt_get_maxrecs(
 558         struct xfs_btree_cur    *cur,
 559         int                     level)
 560 {
 561         if (level == cur->bc_nlevels - 1) {
 562                 struct xfs_ifork        *ifp;
 563
 564                 ifp = XFS_IFORK_PTR(cur->bc_private.b.ip,
 565                                     cur->bc_private.b.whichfork);
 566
 567                 return xfs_bmbt_maxrecs(cur->bc_mp,
 568                                         ifp->if_broot_bytes, level == 0);
 569         }
 570
 571         return cur->bc_mp->m_bmap_dmxr[level != 0];
 572
 573 }
 574
 575 /*
 576  * Get the maximum records we could store in the on-disk format.
 577  *
 578  * For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
 579  * for the root node this checks the available space in the dinode fork
 580  * so that we can resize the in-memory buffer to match it.  After a
 581  * resize to the maximum size this function returns the same value
 582  * as xfs_bmbt_get_maxrecs for the root node, too.
 583  */
 584 STATIC int
 585 xfs_bmbt_get_dmaxrecs(
 586         struct xfs_btree_cur    *cur,
 587         int                     level)
 588 {
 589         if (level != cur->bc_nlevels - 1)
 590                 return cur->bc_mp->m_bmap_dmxr[level != 0];
 591         return xfs_bmdr_maxrecs(cur->bc_private.b.forksize, level == 0);
 592 }
 593
 594 STATIC void
 595 xfs_bmbt_init_key_from_rec(
 596         union xfs_btree_key     *key,
 597         union xfs_btree_rec     *rec)
 598 {
 599         key->bmbt.br_startoff =
 600                 cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
 601 }
 602
 603 STATIC void
 604 xfs_bmbt_init_rec_from_key(
 605         union xfs_btree_key     *key,
 606         union xfs_btree_rec     *rec)
 607 {
 608         ASSERT(key->bmbt.br_startoff != 0);
 609
 610         xfs_bmbt_disk_set_allf(&rec->bmbt, be64_to_cpu(key->bmbt.br_startoff),
 611                                0, 0, XFS_EXT_NORM);
 612 }
 613
 614 STATIC void
 615 xfs_bmbt_init_rec_from_cur(
 616         struct xfs_btree_cur    *cur,
 617         union xfs_btree_rec     *rec)
 618 {
 619         xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
 620 }
 621
 622 STATIC void
 623 xfs_bmbt_init_ptr_from_cur(
 624         struct xfs_btree_cur    *cur,
 625         union xfs_btree_ptr     *ptr)
 626 {
 627         ptr->l = 0;
 628 }
 629
 630 STATIC __int64_t
 631 xfs_bmbt_key_diff(
 632         struct xfs_btree_cur    *cur,
 633         union xfs_btree_key     *key)
 634 {
 635         return (__int64_t)be64_to_cpu(key->bmbt.br_startoff) -
 636                                       cur->bc_rec.b.br_startoff;
 637 }
 638
 639 static bool
 640 xfs_bmbt_verify(
 641         struct xfs_buf          *bp)
 642 {
 643         struct xfs_mount        *mp = bp->b_target->bt_mount;
 644         struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 645         unsigned int            level;
 646
 647         switch (block->bb_magic) {
 648         case cpu_to_be32(XFS_BMAP_CRC_MAGIC):
 649                 if (!xfs_sb_version_hascrc(&mp->m_sb))
 650                         return false;
 651                 if (!uuid_equal(&block->bb_u.l.bb_uuid, &mp->m_sb.sb_meta_uuid))
 652                         return false;
 653                 if (be64_to_cpu(block->bb_u.l.bb_blkno) != bp->b_bn)
 654                         return false;
 655                 /*
 656                  * XXX: need a better way of verifying the owner here. Right now
 657                  * just make sure there has been one set.
 658                  */
 659                 if (be64_to_cpu(block->bb_u.l.bb_owner) == 0)
 660                         return false;
 661                 /* fall through */
 662         case cpu_to_be32(XFS_BMAP_MAGIC):
 663                 break;
 664         default:
 665                 return false;
 666         }
 667
 668         /*
 669          * numrecs and level verification.
 670          *
 671          * We don't know what fork we belong to, so just verify that the level
 672          * is less than the maximum of the two. Later checks will be more
 673          * precise.
 674          */
 675         level = be16_to_cpu(block->bb_level);
 676         if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
 677                 return false;
 678         if (be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
 679                 return false;
 680
 681         /* sibling pointer verification */
 682         if (!block->bb_u.l.bb_leftsib ||
 683             (block->bb_u.l.bb_leftsib != cpu_to_be64(NULLFSBLOCK) &&
 684              !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_leftsib))))
 685                 return false;
 686         if (!block->bb_u.l.bb_rightsib ||
 687             (block->bb_u.l.bb_rightsib != cpu_to_be64(NULLFSBLOCK) &&
 688              !XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_u.l.bb_rightsib))))
 689                 return false;
 690
 691         return true;
 692 }
 693
 694 static void
 695 xfs_bmbt_read_verify(
 696         struct xfs_buf  *bp)
 697 {
 698         if (!xfs_btree_lblock_verify_crc(bp))
 699                 xfs_buf_ioerror(bp, -EFSBADCRC);
 700         else if (!xfs_bmbt_verify(bp))
 701                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 702
 703         if (bp->b_error) {
 704                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 705                 xfs_verifier_error(bp);
 706         }
 707 }
 708
 709 static void
 710 xfs_bmbt_write_verify(
 711         struct xfs_buf  *bp)
 712 {
 713         if (!xfs_bmbt_verify(bp)) {
 714                 trace_xfs_btree_corrupt(bp, _RET_IP_);
 715                 xfs_buf_ioerror(bp, -EFSCORRUPTED);
 716                 xfs_verifier_error(bp);
 717                 return;
 718         }
 719         xfs_btree_lblock_calc_crc(bp);
 720 }
 721
 722 const struct xfs_buf_ops xfs_bmbt_buf_ops = {
 723         .name = "xfs_bmbt",
 724         .verify_read = xfs_bmbt_read_verify,
 725         .verify_write = xfs_bmbt_write_verify,
 726 };
 727
 728
 729 #if defined(DEBUG) || defined(XFS_WARN)
 730 STATIC int
 731 xfs_bmbt_keys_inorder(
 732         struct xfs_btree_cur    *cur,
 733         union xfs_btree_key     *k1,
 734         union xfs_btree_key     *k2)
 735 {
 736         return be64_to_cpu(k1->bmbt.br_startoff) <
 737                 be64_to_cpu(k2->bmbt.br_startoff);
 738 }
 739
 740 STATIC int
 741 xfs_bmbt_recs_inorder(
 742         struct xfs_btree_cur    *cur,
 743         union xfs_btree_rec     *r1,
 744         union xfs_btree_rec     *r2)
 745 {
 746         return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
 747                 xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
 748                 xfs_bmbt_disk_get_startoff(&r2->bmbt);
 749 }
 750 #endif  /* DEBUG */
 751
 752 static const struct xfs_btree_ops xfs_bmbt_ops = {
 753         .rec_len                = sizeof(xfs_bmbt_rec_t),
 754         .key_len                = sizeof(xfs_bmbt_key_t),
 755
 756         .dup_cursor             = xfs_bmbt_dup_cursor,
 757         .update_cursor          = xfs_bmbt_update_cursor,
 758         .alloc_block            = xfs_bmbt_alloc_block,
 759         .free_block             = xfs_bmbt_free_block,
 760         .get_maxrecs            = xfs_bmbt_get_maxrecs,
 761         .get_minrecs            = xfs_bmbt_get_minrecs,
 762         .get_dmaxrecs           = xfs_bmbt_get_dmaxrecs,
 763         .init_key_from_rec      = xfs_bmbt_init_key_from_rec,
 764         .init_rec_from_key      = xfs_bmbt_init_rec_from_key,
 765         .init_rec_from_cur      = xfs_bmbt_init_rec_from_cur,
 766         .init_ptr_from_cur      = xfs_bmbt_init_ptr_from_cur,
 767         .key_diff               = xfs_bmbt_key_diff,
 768         .buf_ops                = &xfs_bmbt_buf_ops,
 769 #if defined(DEBUG) || defined(XFS_WARN)
 770         .keys_inorder           = xfs_bmbt_keys_inorder,
 771         .recs_inorder           = xfs_bmbt_recs_inorder,
 772 #endif
 773 };
 774
 775 /*
 776  * Allocate a new bmap btree cursor.
 777  */
 778 struct xfs_btree_cur *                          /* new bmap btree cursor */
 779 xfs_bmbt_init_cursor(
 780         struct xfs_mount        *mp,            /* file system mount point */
 781         struct xfs_trans        *tp,            /* transaction pointer */
 782         struct xfs_inode        *ip,            /* inode owning the btree */
 783         int                     whichfork)      /* data or attr fork */
 784 {
 785         struct xfs_ifork        *ifp = XFS_IFORK_PTR(ip, whichfork);
 786         struct xfs_btree_cur    *cur;
 787
 788         cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
 789
 790         cur->bc_tp = tp;
 791         cur->bc_mp = mp;
 792         cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
 793         cur->bc_btnum = XFS_BTNUM_BMAP;
 794         cur->bc_blocklog = mp->m_sb.sb_blocklog;
 795
 796         cur->bc_ops = &xfs_bmbt_ops;
 797         cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
 798         if (xfs_sb_version_hascrc(&mp->m_sb))
 799                 cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 800
 801         cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
 802         cur->bc_private.b.ip = ip;
 803         cur->bc_private.b.firstblock = NULLFSBLOCK;
 804         cur->bc_private.b.flist = NULL;
 805         cur->bc_private.b.allocated = 0;
 806         cur->bc_private.b.flags = 0;
 807         cur->bc_private.b.whichfork = whichfork;
 808
 809         return cur;
 810 }
 811
 812 /*
 813  * Calculate number of records in a bmap btree block.
 814  */
 815 int
 816 xfs_bmbt_maxrecs(
 817         struct xfs_mount        *mp,
 818         int                     blocklen,
 819         int                     leaf)
 820 {
 821         blocklen -= XFS_BMBT_BLOCK_LEN(mp);
 822
 823         if (leaf)
 824                 return blocklen / sizeof(xfs_bmbt_rec_t);
 825         return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
 826 }
 827
 828 /*
 829  * Calculate number of records in a bmap btree inode root.
 830  */
 831 int
 832 xfs_bmdr_maxrecs(
 833         int                     blocklen,
 834         int                     leaf)
 835 {
 836         blocklen -= sizeof(xfs_bmdr_block_t);
 837
 838         if (leaf)
 839                 return blocklen / sizeof(xfs_bmdr_rec_t);
 840         return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
 841 }
 842
 843 /*
 844  * Change the owner of a btree format fork fo the inode passed in. Change it to
 845  * the owner of that is passed in so that we can change owners before or after
 846  * we switch forks between inodes. The operation that the caller is doing will
 847  * determine whether is needs to change owner before or after the switch.
 848  *
 849  * For demand paged transactional modification, the fork switch should be done
 850  * after reading in all the blocks, modifying them and pinning them in the
 851  * transaction. For modification when the buffers are already pinned in memory,
 852  * the fork switch can be done before changing the owner as we won't need to
 853  * validate the owner until the btree buffers are unpinned and writes can occur
 854  * again.
 855  *
 856  * For recovery based ownership change, there is no transactional context and
 857  * so a buffer list must be supplied so that we can record the buffers that we
 858  * modified for the caller to issue IO on.
 859  */
 860 int
 861 xfs_bmbt_change_owner(
 862         struct xfs_trans        *tp,
 863         struct xfs_inode        *ip,
 864         int                     whichfork,
 865         xfs_ino_t               new_owner,
 866         struct list_head        *buffer_list)
 867 {
 868         struct xfs_btree_cur    *cur;
 869         int                     error;
 870
 871         ASSERT(tp || buffer_list);
 872         ASSERT(!(tp && buffer_list));
 873         if (whichfork == XFS_DATA_FORK)
 874                 ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_BTREE);
 875         else
 876                 ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_BTREE);
 877
 878         cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
 879         if (!cur)
 880                 return -ENOMEM;
 881
 882         error = xfs_btree_change_owner(cur, new_owner, buffer_list);
 883         xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 884         return error;
 885 }