sys/ufs/lfs/ulfs_bmap.c

   1 /*      $NetBSD: ulfs_bmap.c,v 1.5 2013/07/28 01:10:49 dholland Exp $   */
   2 /*  from NetBSD: ufs_bmap.c,v 1.50 2013/01/22 09:39:18 dholland Exp  */
   3
   4 /*
   5  * Copyright (c) 1989, 1991, 1993
   6  *      The Regents of the University of California.  All rights reserved.
   7  * (c) UNIX System Laboratories, Inc.
   8  * All or some portions of this file are derived from material licensed
   9  * to the University of California by American Telephone and Telegraph
  10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  11  * the permission of UNIX System Laboratories, Inc.
  12  *
  13  * Redistribution and use in source and binary forms, with or without
  14  * modification, are permitted provided that the following conditions
  15  * are met:
  16  * 1. Redistributions of source code must retain the above copyright
  17  *    notice, this list of conditions and the following disclaimer.
  18  * 2. Redistributions in binary form must reproduce the above copyright
  19  *    notice, this list of conditions and the following disclaimer in the
  20  *    documentation and/or other materials provided with the distribution.
  21  * 3. Neither the name of the University nor the names of its contributors
  22  *    may be used to endorse or promote products derived from this software
  23  *    without specific prior written permission.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  35  * SUCH DAMAGE.
  36  *
  37  *      @(#)ufs_bmap.c  8.8 (Berkeley) 8/11/95
  38  */
  39
  40 #include <sys/cdefs.h>
  41 __KERNEL_RCSID(0, "$NetBSD: ulfs_bmap.c,v 1.5 2013/07/28 01:10:49 dholland Exp $");
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/stat.h>
  46 #include <sys/buf.h>
  47 #include <sys/proc.h>
  48 #include <sys/vnode.h>
  49 #include <sys/mount.h>
  50 #include <sys/resourcevar.h>
  51 #include <sys/trace.h>
  52 #include <sys/fstrans.h>
  53
  54 #include <miscfs/specfs/specdev.h>
  55
  56 #include <ufs/lfs/ulfs_inode.h>
  57 #include <ufs/lfs/ulfsmount.h>
  58 #include <ufs/lfs/ulfs_extern.h>
  59 #include <ufs/lfs/ulfs_bswap.h>
  60
  61 static bool
  62 ulfs_issequential(const struct lfs *fs, daddr_t daddr0, daddr_t daddr1)
  63 {
  64
  65         /* for ulfs, blocks in a hole is not 'contiguous'. */
  66         if (daddr0 == 0)
  67                 return false;
  68
  69         return (daddr0 + fs->um_seqinc == daddr1);
  70 }
  71
  72 /*
  73  * Bmap converts the logical block number of a file to its physical block
  74  * number on the disk. The conversion is done by using the logical block
  75  * number to index into the array of block pointers described by the dinode.
  76  */
  77 int
  78 ulfs_bmap(void *v)
  79 {
  80         struct vop_bmap_args /* {
  81                 struct vnode *a_vp;
  82                 daddr_t  a_bn;
  83                 struct vnode **a_vpp;
  84                 daddr_t *a_bnp;
  85                 int *a_runp;
  86         } */ *ap = v;
  87         int error;
  88
  89         /*
  90          * Check for underlying vnode requests and ensure that logical
  91          * to physical mapping is requested.
  92          */
  93         if (ap->a_vpp != NULL)
  94                 *ap->a_vpp = VTOI(ap->a_vp)->i_devvp;
  95         if (ap->a_bnp == NULL)
  96                 return (0);
  97
  98         fstrans_start(ap->a_vp->v_mount, FSTRANS_SHARED);
  99         error = ulfs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL,
 100             ap->a_runp, ulfs_issequential);
 101         fstrans_done(ap->a_vp->v_mount);
 102         return error;
 103 }
 104
 105 /*
 106  * Indirect blocks are now on the vnode for the file.  They are given negative
 107  * logical block numbers.  Indirect blocks are addressed by the negative
 108  * address of the first data block to which they point.  Double indirect blocks
 109  * are addressed by one less than the address of the first indirect block to
 110  * which they point.  Triple indirect blocks are addressed by one less than
 111  * the address of the first double indirect block to which they point.
 112  *
 113  * ulfs_bmaparray does the bmap conversion, and if requested returns the
 114  * array of logical blocks which must be traversed to get to a block.
 115  * Each entry contains the offset into that block that gets you to the
 116  * next block and the disk address of the block (if it is assigned).
 117  */
 118
 119 int
 120 ulfs_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, struct indir *ap,
 121     int *nump, int *runp, ulfs_issequential_callback_t is_sequential)
 122 {
 123         struct inode *ip;
 124         struct buf *bp, *cbp;
 125         struct ulfsmount *ump;
 126         struct lfs *fs;
 127         struct mount *mp;
 128         struct indir a[ULFS_NIADDR + 1], *xap;
 129         daddr_t daddr;
 130         daddr_t metalbn;
 131         int error, maxrun = 0, num;
 132
 133         ip = VTOI(vp);
 134         mp = vp->v_mount;
 135         ump = ip->i_ump;
 136         fs = ip->i_lfs;
 137 #ifdef DIAGNOSTIC
 138         if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
 139                 panic("ulfs_bmaparray: invalid arguments");
 140 #endif
 141
 142         if (runp) {
 143                 /*
 144                  * XXX
 145                  * If MAXBSIZE is the largest transfer the disks can handle,
 146                  * we probably want maxrun to be 1 block less so that we
 147                  * don't create a block larger than the device can handle.
 148                  */
 149                 *runp = 0;
 150                 maxrun = MAXPHYS / mp->mnt_stat.f_iosize - 1;
 151         }
 152
 153         if (bn >= 0 && bn < ULFS_NDADDR) {
 154                 if (nump != NULL)
 155                         *nump = 0;
 156                 if (ump->um_fstype == ULFS1)
 157                         daddr = ulfs_rw32(ip->i_ffs1_db[bn],
 158                             ULFS_MPNEEDSWAP(fs));
 159                 else
 160                         daddr = ulfs_rw64(ip->i_ffs2_db[bn],
 161                             ULFS_MPNEEDSWAP(fs));
 162                 *bnp = blkptrtodb(fs, daddr);
 163                 /*
 164                  * Since this is FFS independent code, we are out of
 165                  * scope for the definitions of BLK_NOCOPY and
 166                  * BLK_SNAP, but we do know that they will fall in
 167                  * the range 1..um_seqinc, so we use that test and
 168                  * return a request for a zeroed out buffer if attempts
 169                  * are made to read a BLK_NOCOPY or BLK_SNAP block.
 170                  */
 171                 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT
 172                     && daddr > 0 &&
 173                     daddr < fs->um_seqinc) {
 174                         *bnp = -1;
 175                 } else if (*bnp == 0) {
 176                         if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))
 177                             == SF_SNAPSHOT) {
 178                                 *bnp = blkptrtodb(fs, bn * fs->um_seqinc);
 179                         } else {
 180                                 *bnp = -1;
 181                         }
 182                 } else if (runp) {
 183                         if (ump->um_fstype == ULFS1) {
 184                                 for (++bn; bn < ULFS_NDADDR && *runp < maxrun &&
 185                                     is_sequential(fs,
 186                                         ulfs_rw32(ip->i_ffs1_db[bn - 1],
 187                                             ULFS_MPNEEDSWAP(fs)),
 188                                         ulfs_rw32(ip->i_ffs1_db[bn],
 189                                             ULFS_MPNEEDSWAP(fs)));
 190                                     ++bn, ++*runp);
 191                         } else {
 192                                 for (++bn; bn < ULFS_NDADDR && *runp < maxrun &&
 193                                     is_sequential(fs,
 194                                         ulfs_rw64(ip->i_ffs2_db[bn - 1],
 195                                             ULFS_MPNEEDSWAP(fs)),
 196                                         ulfs_rw64(ip->i_ffs2_db[bn],
 197                                             ULFS_MPNEEDSWAP(fs)));
 198                                     ++bn, ++*runp);
 199                         }
 200                 }
 201                 return (0);
 202         }
 203
 204         xap = ap == NULL ? a : ap;
 205         if (!nump)
 206                 nump = &num;
 207         if ((error = ulfs_getlbns(vp, bn, xap, nump)) != 0)
 208                 return (error);
 209
 210         num = *nump;
 211
 212         /* Get disk address out of indirect block array */
 213         if (ump->um_fstype == ULFS1)
 214                 daddr = ulfs_rw32(ip->i_ffs1_ib[xap->in_off],
 215                     ULFS_MPNEEDSWAP(fs));
 216         else
 217                 daddr = ulfs_rw64(ip->i_ffs2_ib[xap->in_off],
 218                     ULFS_MPNEEDSWAP(fs));
 219
 220         for (bp = NULL, ++xap; --num; ++xap) {
 221                 /*
 222                  * Exit the loop if there is no disk address assigned yet and
 223                  * the indirect block isn't in the cache, or if we were
 224                  * looking for an indirect block and we've found it.
 225                  */
 226
 227                 metalbn = xap->in_lbn;
 228                 if (metalbn == bn)
 229                         break;
 230                 if (daddr == 0) {
 231                         mutex_enter(&bufcache_lock);
 232                         cbp = incore(vp, metalbn);
 233                         mutex_exit(&bufcache_lock);
 234                         if (cbp == NULL)
 235                                 break;
 236                 }
 237
 238                 /*
 239                  * If we get here, we've either got the block in the cache
 240                  * or we have a disk address for it, go fetch it.
 241                  */
 242                 if (bp)
 243                         brelse(bp, 0);
 244
 245                 xap->in_exists = 1;
 246                 bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
 247                 if (bp == NULL) {
 248
 249                         /*
 250                          * getblk() above returns NULL only iff we are
 251                          * pagedaemon.  See the implementation of getblk
 252                          * for detail.
 253                          */
 254
 255                         return (ENOMEM);
 256                 }
 257                 if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
 258                         trace(TR_BREADHIT, pack(vp, size), metalbn);
 259                 }
 260 #ifdef DIAGNOSTIC
 261                 else if (!daddr)
 262                         panic("ulfs_bmaparray: indirect block not in cache");
 263 #endif
 264                 else {
 265                         trace(TR_BREADMISS, pack(vp, size), metalbn);
 266                         bp->b_blkno = blkptrtodb(fs, daddr);
 267                         bp->b_flags |= B_READ;
 268                         BIO_SETPRIO(bp, BPRIO_TIMECRITICAL);
 269                         VOP_STRATEGY(vp, bp);
 270                         curlwp->l_ru.ru_inblock++;      /* XXX */
 271                         if ((error = biowait(bp)) != 0) {
 272                                 brelse(bp, 0);
 273                                 return (error);
 274                         }
 275                 }
 276                 if (ump->um_fstype == ULFS1) {
 277                         daddr = ulfs_rw32(((u_int32_t *)bp->b_data)[xap->in_off],
 278                             ULFS_MPNEEDSWAP(fs));
 279                         if (num == 1 && daddr && runp) {
 280                                 for (bn = xap->in_off + 1;
 281                                     bn < MNINDIR(fs) && *runp < maxrun &&
 282                                     is_sequential(fs,
 283                                         ulfs_rw32(((int32_t *)bp->b_data)[bn-1],
 284                                             ULFS_MPNEEDSWAP(fs)),
 285                                         ulfs_rw32(((int32_t *)bp->b_data)[bn],
 286                                             ULFS_MPNEEDSWAP(fs)));
 287                                     ++bn, ++*runp);
 288                         }
 289                 } else {
 290                         daddr = ulfs_rw64(((u_int64_t *)bp->b_data)[xap->in_off],
 291                             ULFS_MPNEEDSWAP(fs));
 292                         if (num == 1 && daddr && runp) {
 293                                 for (bn = xap->in_off + 1;
 294                                     bn < MNINDIR(fs) && *runp < maxrun &&
 295                                     is_sequential(fs,
 296                                         ulfs_rw64(((int64_t *)bp->b_data)[bn-1],
 297                                             ULFS_MPNEEDSWAP(fs)),
 298                                         ulfs_rw64(((int64_t *)bp->b_data)[bn],
 299                                             ULFS_MPNEEDSWAP(fs)));
 300                                     ++bn, ++*runp);
 301                         }
 302                 }
 303         }
 304         if (bp)
 305                 brelse(bp, 0);
 306
 307         /*
 308          * Since this is FFS independent code, we are out of scope for the
 309          * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they
 310          * will fall in the range 1..um_seqinc, so we use that test and
 311          * return a request for a zeroed out buffer if attempts are made
 312          * to read a BLK_NOCOPY or BLK_SNAP block.
 313          */
 314         if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT
 315             && daddr > 0 && daddr < fs->um_seqinc) {
 316                 *bnp = -1;
 317                 return (0);
 318         }
 319         *bnp = blkptrtodb(fs, daddr);
 320         if (*bnp == 0) {
 321                 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL))
 322                     == SF_SNAPSHOT) {
 323                         *bnp = blkptrtodb(fs, bn * fs->um_seqinc);
 324                 } else {
 325                         *bnp = -1;
 326                 }
 327         }
 328         return (0);
 329 }
 330
 331 /*
 332  * Create an array of logical block number/offset pairs which represent the
 333  * path of indirect blocks required to access a data block.  The first "pair"
 334  * contains the logical block number of the appropriate single, double or
 335  * triple indirect block and the offset into the inode indirect block array.
 336  * Note, the logical block number of the inode single/double/triple indirect
 337  * block appears twice in the array, once with the offset into the i_ffs1_ib and
 338  * once with the offset into the page itself.
 339  */
 340 int
 341 ulfs_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump)
 342 {
 343         daddr_t metalbn, realbn;
 344         struct ulfsmount *ump;
 345         struct lfs *fs;
 346         int64_t blockcnt;
 347         int lbc;
 348         int i, numlevels, off;
 349
 350         ump = VFSTOULFS(vp->v_mount);
 351         fs = ump->um_lfs;
 352         if (nump)
 353                 *nump = 0;
 354         numlevels = 0;
 355         realbn = bn;
 356         if (bn < 0)
 357                 bn = -bn;
 358         KASSERT(bn >= ULFS_NDADDR);
 359
 360         /*
 361          * Determine the number of levels of indirection.  After this loop
 362          * is done, blockcnt indicates the number of data blocks possible
 363          * at the given level of indirection, and ULFS_NIADDR - i is the number
 364          * of levels of indirection needed to locate the requested block.
 365          */
 366
 367         bn -= ULFS_NDADDR;
 368         for (lbc = 0, i = ULFS_NIADDR;; i--, bn -= blockcnt) {
 369                 if (i == 0)
 370                         return (EFBIG);
 371
 372                 lbc += fs->um_lognindir;
 373                 blockcnt = (int64_t)1 << lbc;
 374
 375                 if (bn < blockcnt)
 376                         break;
 377         }
 378
 379         /* Calculate the address of the first meta-block. */
 380         metalbn = -((realbn >= 0 ? realbn : -realbn) - bn + ULFS_NIADDR - i);
 381
 382         /*
 383          * At each iteration, off is the offset into the bap array which is
 384          * an array of disk addresses at the current level of indirection.
 385          * The logical block number and the offset in that block are stored
 386          * into the argument array.
 387          */
 388         ap->in_lbn = metalbn;
 389         ap->in_off = off = ULFS_NIADDR - i;
 390         ap->in_exists = 0;
 391         ap++;
 392         for (++numlevels; i <= ULFS_NIADDR; i++) {
 393                 /* If searching for a meta-data block, quit when found. */
 394                 if (metalbn == realbn)
 395                         break;
 396
 397                 lbc -= fs->um_lognindir;
 398                 off = (bn >> lbc) & (MNINDIR(fs) - 1);
 399
 400                 ++numlevels;
 401                 ap->in_lbn = metalbn;
 402                 ap->in_off = off;
 403                 ap->in_exists = 0;
 404                 ++ap;
 405
 406                 metalbn -= -1 + ((int64_t)off << lbc);
 407         }
 408         if (nump)
 409                 *nump = numlevels;
 410         return (0);
 411 }