sys/vfs/ufs/ufs_bmap.c

   1 /*
   2  * Copyright (c) 1989, 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  * (c) UNIX System Laboratories, Inc.
   5  * All or some portions of this file are derived from material licensed
   6  * to the University of California by American Telephone and Telegraph
   7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
   8  * the permission of UNIX System Laboratories, Inc.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  * 1. Redistributions of source code must retain the above copyright
  14  *    notice, this list of conditions and the following disclaimer.
  15  * 2. Redistributions in binary form must reproduce the above copyright
  16  *    notice, this list of conditions and the following disclaimer in the
  17  *    documentation and/or other materials provided with the distribution.
  18  * 3. All advertising materials mentioning features or use of this software
  19  *    must display the following acknowledgement:
  20  *      This product includes software developed by the University of
  21  *      California, Berkeley and its contributors.
  22  * 4. Neither the name of the University nor the names of its contributors
  23  *    may be used to endorse or promote products derived from this software
  24  *    without specific prior written permission.
  25  *
  26  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  36  * SUCH DAMAGE.
  37  *
  38  *      @(#)ufs_bmap.c  8.7 (Berkeley) 3/21/95
  39  * $FreeBSD: src/sys/ufs/ufs/ufs_bmap.c,v 1.34.2.1 2000/03/17 10:12:14 ps Exp $
  40  * $DragonFly: src/sys/vfs/ufs/ufs_bmap.c,v 1.13 2007/08/13 17:31:57 dillon Exp $
  41  */
  42
  43 #include <sys/param.h>
  44 #include <sys/systm.h>
  45 #include <sys/buf.h>
  46 #include <sys/proc.h>
  47 #include <sys/vnode.h>
  48 #include <sys/mount.h>
  49 #include <sys/resourcevar.h>
  50 #include <sys/conf.h>
  51
  52 #include "quota.h"
  53 #include "inode.h"
  54 #include "ufsmount.h"
  55 #include "ufs_extern.h"
  56 #include "fs.h"
  57
  58 /*
  59  * Bmap converts the logical block number of a file to its physical block
  60  * number on the disk. The conversion is done by using the logical block
  61  * number to index into the array of block pointers described by the dinode.
  62  *
  63  * BMAP must return the contiguous before and after run in bytes, inclusive
  64  * of the returned block.
  65  *
  66  * ufs_bmap(struct vnode *a_vp, off_t a_loffset,
  67  *          off_t *a_doffsetp, int *a_runp, int *a_runb)
  68  */
  69 int
  70 ufs_bmap(struct vop_bmap_args *ap)
  71 {
  72         struct fs *fs;
  73         ufs_daddr_t lbn;
  74         ufs_daddr_t dbn;
  75         int error;
  76
  77         /*
  78          * Check for underlying vnode requests and ensure that logical
  79          * to physical mapping is requested.
  80          */
  81         if (ap->a_doffsetp == NULL)
  82                 return (0);
  83
  84         fs = VTOI(ap->a_vp)->i_fs;
  85         KKASSERT(((int)ap->a_loffset & ((1 << fs->fs_bshift) - 1)) == 0);
  86         lbn = ap->a_loffset >> fs->fs_bshift;
  87
  88         error = ufs_bmaparray(ap->a_vp, lbn, &dbn, NULL, NULL,
  89                               ap->a_runp, ap->a_runb);
  90
  91         if (error || dbn == (ufs_daddr_t)-1) {
  92                 *ap->a_doffsetp = NOOFFSET;
  93         } else {
  94                 *ap->a_doffsetp = dbtodoff(fs, dbn);
  95                 if (ap->a_runp)
  96                         *ap->a_runp = (*ap->a_runp + 1) << fs->fs_bshift;
  97                 if (ap->a_runb)
  98                         *ap->a_runb = *ap->a_runb << fs->fs_bshift;
  99         }
 100         return (error);
 101 }
 102
 103 /*
 104  * Indirect blocks are now on the vnode for the file.  They are given negative
 105  * logical block numbers.  Indirect blocks are addressed by the negative
 106  * address of the first data block to which they point.  Double indirect blocks
 107  * are addressed by one less than the address of the first indirect block to
 108  * which they point.  Triple indirect blocks are addressed by one less than
 109  * the address of the first double indirect block to which they point.
 110  *
 111  * ufs_bmaparray does the bmap conversion, and if requested returns the
 112  * array of logical blocks which must be traversed to get to a block.
 113  * Each entry contains the offset into that block that gets you to the
 114  * next block and the disk address of the block (if it is assigned).
 115  */
 116 int
 117 ufs_bmaparray(struct vnode *vp, ufs_daddr_t bn, ufs_daddr_t *bnp,
 118               struct indir *ap, int *nump, int *runp, int *runb)
 119 {
 120         struct inode *ip;
 121         struct buf *bp;
 122         struct ufsmount *ump;
 123         struct mount *mp;
 124         struct vnode *devvp;
 125         struct fs *fs;
 126         struct indir a[NIADDR+1], *xap;
 127         ufs_daddr_t daddr;
 128         long metalbn;
 129         int error, maxrun, num;
 130
 131         ip = VTOI(vp);
 132         mp = vp->v_mount;
 133         ump = VFSTOUFS(mp);
 134         devvp = ump->um_devvp;
 135         fs = ip->i_fs;
 136 #ifdef DIAGNOSTIC
 137         if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
 138                 panic("ufs_bmaparray: invalid arguments");
 139 #endif
 140
 141         if (runp) {
 142                 *runp = 0;
 143         }
 144
 145         if (runb) {
 146                 *runb = 0;
 147         }
 148
 149         maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1;
 150
 151         xap = ap == NULL ? a : ap;
 152         if (!nump)
 153                 nump = &num;
 154         error = ufs_getlbns(vp, bn, xap, nump);
 155         if (error)
 156                 return (error);
 157
 158         num = *nump;
 159         if (num == 0) {
 160                 *bnp = blkptrtodb(ump, ip->i_db[bn]);
 161                 if (*bnp == 0)
 162                         *bnp = -1;
 163                 else if (runp) {
 164                         daddr_t bnb = bn;
 165                         for (++bn; bn < NDADDR && *runp < maxrun &&
 166                             is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
 167                             ++bn, ++*runp);
 168                         bn = bnb;
 169                         if (runb && (bn > 0)) {
 170                                 for (--bn; (bn >= 0) && (*runb < maxrun) &&
 171                                         is_sequential(ump, ip->i_db[bn],
 172                                                 ip->i_db[bn+1]);
 173                                                 --bn, ++*runb);
 174                         }
 175                 }
 176                 return (0);
 177         }
 178
 179
 180         /* Get disk address out of indirect block array */
 181         daddr = ip->i_ib[xap->in_off];
 182
 183         for (bp = NULL, ++xap; --num; ++xap) {
 184                 /*
 185                  * Exit the loop if there is no disk address assigned yet and
 186                  * the indirect block isn't in the cache, or if we were
 187                  * looking for an indirect block and we've found it.
 188                  */
 189
 190                 metalbn = xap->in_lbn;
 191                 if ((daddr == 0 && !findblk(vp, dbtodoff(fs, metalbn))) || metalbn == bn)
 192                         break;
 193                 /*
 194                  * If we get here, we've either got the block in the cache
 195                  * or we have a disk address for it, go fetch it.
 196                  */
 197                 if (bp)
 198                         bqrelse(bp);
 199
 200                 xap->in_exists = 1;
 201                 bp = getblk(vp, lblktodoff(fs, metalbn),
 202                             mp->mnt_stat.f_iosize, 0, 0);
 203                 if ((bp->b_flags & B_CACHE) == 0) {
 204 #ifdef DIAGNOSTIC
 205                         if (!daddr)
 206                                 panic("ufs_bmaparray: indirect block not in cache");
 207 #endif
 208                         bp->b_bio2.bio_offset = fsbtodoff(fs, daddr);
 209                         bp->b_flags &= ~(B_INVAL|B_ERROR);
 210                         bp->b_cmd = BUF_CMD_READ;
 211                         vfs_busy_pages(bp->b_vp, bp);
 212                         vn_strategy(bp->b_vp, &bp->b_bio1);
 213                         error = biowait(bp);
 214                         if (error) {
 215                                 brelse(bp);
 216                                 return (error);
 217                         }
 218                 }
 219
 220                 daddr = ((ufs_daddr_t *)bp->b_data)[xap->in_off];
 221                 if (num == 1 && daddr && runp) {
 222                         for (bn = xap->in_off + 1;
 223                             bn < MNINDIR(ump) && *runp < maxrun &&
 224                             is_sequential(ump,
 225                             ((ufs_daddr_t *)bp->b_data)[bn - 1],
 226                             ((ufs_daddr_t *)bp->b_data)[bn]);
 227                             ++bn, ++*runp);
 228                         bn = xap->in_off;
 229                         if (runb && bn) {
 230                                 for(--bn; bn >= 0 && *runb < maxrun &&
 231                                         is_sequential(ump, ((daddr_t *)bp->b_data)[bn],
 232                                             ((daddr_t *)bp->b_data)[bn+1]);
 233                                         --bn, ++*runb);
 234                         }
 235                 }
 236         }
 237         if (bp)
 238                 bqrelse(bp);
 239
 240         daddr = blkptrtodb(ump, daddr);
 241         *bnp = daddr == 0 ? -1 : daddr;
 242         return (0);
 243 }
 244
 245 /*
 246  * Create an array of logical block number/offset pairs which represent the
 247  * path of indirect blocks required to access a data block.  The first "pair"
 248  * contains the logical block number of the appropriate single, double or
 249  * triple indirect block and the offset into the inode indirect block array.
 250  * Note, the logical block number of the inode single/double/triple indirect
 251  * block appears twice in the array, once with the offset into the i_ib and
 252  * once with the offset into the page itself.
 253  */
 254 int
 255 ufs_getlbns(struct vnode *vp, ufs_daddr_t bn, struct indir *ap, int *nump)
 256 {
 257         long blockcnt, metalbn, realbn;
 258         struct ufsmount *ump;
 259         int i, numlevels, off;
 260         int64_t qblockcnt;
 261
 262         ump = VFSTOUFS(vp->v_mount);
 263         if (nump)
 264                 *nump = 0;
 265         numlevels = 0;
 266         realbn = bn;
 267         if ((long)bn < 0)
 268                 bn = -(long)bn;
 269
 270         /* The first NDADDR blocks are direct blocks. */
 271         if (bn < NDADDR)
 272                 return (0);
 273
 274         /*
 275          * Determine the number of levels of indirection.  After this loop
 276          * is done, blockcnt indicates the number of data blocks possible
 277          * at the previous level of indirection, and NIADDR - i is the number
 278          * of levels of indirection needed to locate the requested block.
 279          */
 280         for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
 281                 if (i == 0)
 282                         return (EFBIG);
 283                 /*
 284                  * Use int64_t's here to avoid overflow for triple indirect
 285                  * blocks when longs have 32 bits and the block size is more
 286                  * than 4K.
 287                  */
 288                 qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
 289                 if (bn < qblockcnt)
 290                         break;
 291                 blockcnt = qblockcnt;
 292         }
 293
 294         /* Calculate the address of the first meta-block. */
 295         if (realbn >= 0)
 296                 metalbn = -(realbn - bn + NIADDR - i);
 297         else
 298                 metalbn = -(-realbn - bn + NIADDR - i);
 299
 300         /*
 301          * At each iteration, off is the offset into the bap array which is
 302          * an array of disk addresses at the current level of indirection.
 303          * The logical block number and the offset in that block are stored
 304          * into the argument array.
 305          */
 306         ap->in_lbn = metalbn;
 307         ap->in_off = off = NIADDR - i;
 308         ap->in_exists = 0;
 309         ap++;
 310         for (++numlevels; i <= NIADDR; i++) {
 311                 /* If searching for a meta-data block, quit when found. */
 312                 if (metalbn == realbn)
 313                         break;
 314
 315                 off = (bn / blockcnt) % MNINDIR(ump);
 316
 317                 ++numlevels;
 318                 ap->in_lbn = metalbn;
 319                 ap->in_off = off;
 320                 ap->in_exists = 0;
 321                 ++ap;
 322
 323                 metalbn -= -1 + off * blockcnt;
 324                 blockcnt /= MNINDIR(ump);
 325         }
 326         if (nump)
 327                 *nump = numlevels;
 328         return (0);
 329 }