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 1.8 2006/03/05 18:38:39 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
  57 /*
  58  * Bmap converts a the logical block number of a file to its physical block
  59  * number on the disk. The conversion is done by using the logical block
  60  * number to index into the array of block pointers described by the dinode.
  61  *
  62  * ufs_bmap(struct vnode *a_vp, ufs_daddr_t a_bn, struct vnode **a_vpp,
  63  *          ufs_daddr_t *a_bnp, int *a_runp, int *a_runb)
  64  */
  65 int
  66 ufs_bmap(struct vop_bmap_args *ap)
  67 {
  68         /*
  69          * Check for underlying vnode requests and ensure that logical
  70          * to physical mapping is requested.
  71          */
  72         if (ap->a_vpp != NULL)
  73                 *ap->a_vpp = VTOI(ap->a_vp)->i_devvp;
  74         if (ap->a_bnp == NULL)
  75                 return (0);
  76
  77         return (ufs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL,
  78             ap->a_runp, ap->a_runb));
  79 }
  80
  81 /*
  82  * Indirect blocks are now on the vnode for the file.  They are given negative
  83  * logical block numbers.  Indirect blocks are addressed by the negative
  84  * address of the first data block to which they point.  Double indirect blocks
  85  * are addressed by one less than the address of the first indirect block to
  86  * which they point.  Triple indirect blocks are addressed by one less than
  87  * the address of the first double indirect block to which they point.
  88  *
  89  * ufs_bmaparray does the bmap conversion, and if requested returns the
  90  * array of logical blocks which must be traversed to get to a block.
  91  * Each entry contains the offset into that block that gets you to the
  92  * next block and the disk address of the block (if it is assigned).
  93  */
  94
  95 int
  96 ufs_bmaparray(struct vnode *vp, ufs_daddr_t bn, ufs_daddr_t *bnp,
  97               struct indir *ap, int *nump, int *runp, int *runb)
  98 {
  99         struct inode *ip;
 100         struct buf *bp;
 101         struct ufsmount *ump;
 102         struct mount *mp;
 103         struct vnode *devvp;
 104         struct indir a[NIADDR+1], *xap;
 105         ufs_daddr_t daddr;
 106         long metalbn;
 107         int error, maxrun, num;
 108
 109         ip = VTOI(vp);
 110         mp = vp->v_mount;
 111         ump = VFSTOUFS(mp);
 112         devvp = ump->um_devvp;
 113 #ifdef DIAGNOSTIC
 114         if ((ap != NULL && nump == NULL) || (ap == NULL && nump != NULL))
 115                 panic("ufs_bmaparray: invalid arguments");
 116 #endif
 117
 118         if (runp) {
 119                 *runp = 0;
 120         }
 121
 122         if (runb) {
 123                 *runb = 0;
 124         }
 125
 126         maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1;
 127
 128         xap = ap == NULL ? a : ap;
 129         if (!nump)
 130                 nump = &num;
 131         error = ufs_getlbns(vp, bn, xap, nump);
 132         if (error)
 133                 return (error);
 134
 135         num = *nump;
 136         if (num == 0) {
 137                 *bnp = blkptrtodb(ump, ip->i_db[bn]);
 138                 if (*bnp == 0)
 139                         *bnp = -1;
 140                 else if (runp) {
 141                         daddr_t bnb = bn;
 142                         for (++bn; bn < NDADDR && *runp < maxrun &&
 143                             is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
 144                             ++bn, ++*runp);
 145                         bn = bnb;
 146                         if (runb && (bn > 0)) {
 147                                 for (--bn; (bn >= 0) && (*runb < maxrun) &&
 148                                         is_sequential(ump, ip->i_db[bn],
 149                                                 ip->i_db[bn+1]);
 150                                                 --bn, ++*runb);
 151                         }
 152                 }
 153                 return (0);
 154         }
 155
 156
 157         /* Get disk address out of indirect block array */
 158         daddr = ip->i_ib[xap->in_off];
 159
 160         for (bp = NULL, ++xap; --num; ++xap) {
 161                 /*
 162                  * Exit the loop if there is no disk address assigned yet and
 163                  * the indirect block isn't in the cache, or if we were
 164                  * looking for an indirect block and we've found it.
 165                  */
 166
 167                 metalbn = xap->in_lbn;
 168                 if ((daddr == 0 && !findblk(vp, metalbn)) || metalbn == bn)
 169                         break;
 170                 /*
 171                  * If we get here, we've either got the block in the cache
 172                  * or we have a disk address for it, go fetch it.
 173                  */
 174                 if (bp)
 175                         bqrelse(bp);
 176
 177                 xap->in_exists = 1;
 178                 bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
 179                 if ((bp->b_flags & B_CACHE) == 0) {
 180 #ifdef DIAGNOSTIC
 181                         if (!daddr)
 182                                 panic("ufs_bmaparray: indirect block not in cache");
 183 #endif
 184                         bp->b_bio2.bio_blkno = blkptrtodb(ump, daddr);
 185                         bp->b_flags |= B_READ;
 186                         bp->b_flags &= ~(B_INVAL|B_ERROR);
 187                         vfs_busy_pages(bp, 0);
 188                         vn_strategy(bp->b_vp, &bp->b_bio1);
 189                         error = biowait(bp);
 190                         if (error) {
 191                                 brelse(bp);
 192                                 return (error);
 193                         }
 194                 }
 195
 196                 daddr = ((ufs_daddr_t *)bp->b_data)[xap->in_off];
 197                 if (num == 1 && daddr && runp) {
 198                         for (bn = xap->in_off + 1;
 199                             bn < MNINDIR(ump) && *runp < maxrun &&
 200                             is_sequential(ump,
 201                             ((ufs_daddr_t *)bp->b_data)[bn - 1],
 202                             ((ufs_daddr_t *)bp->b_data)[bn]);
 203                             ++bn, ++*runp);
 204                         bn = xap->in_off;
 205                         if (runb && bn) {
 206                                 for(--bn; bn >= 0 && *runb < maxrun &&
 207                                         is_sequential(ump, ((daddr_t *)bp->b_data)[bn],
 208                                             ((daddr_t *)bp->b_data)[bn+1]);
 209                                         --bn, ++*runb);
 210                         }
 211                 }
 212         }
 213         if (bp)
 214                 bqrelse(bp);
 215
 216         daddr = blkptrtodb(ump, daddr);
 217         *bnp = daddr == 0 ? -1 : daddr;
 218         return (0);
 219 }
 220
 221 /*
 222  * Create an array of logical block number/offset pairs which represent the
 223  * path of indirect blocks required to access a data block.  The first "pair"
 224  * contains the logical block number of the appropriate single, double or
 225  * triple indirect block and the offset into the inode indirect block array.
 226  * Note, the logical block number of the inode single/double/triple indirect
 227  * block appears twice in the array, once with the offset into the i_ib and
 228  * once with the offset into the page itself.
 229  */
 230 int
 231 ufs_getlbns(struct vnode *vp, ufs_daddr_t bn, struct indir *ap, int *nump)
 232 {
 233         long blockcnt, metalbn, realbn;
 234         struct ufsmount *ump;
 235         int i, numlevels, off;
 236         int64_t qblockcnt;
 237
 238         ump = VFSTOUFS(vp->v_mount);
 239         if (nump)
 240                 *nump = 0;
 241         numlevels = 0;
 242         realbn = bn;
 243         if ((long)bn < 0)
 244                 bn = -(long)bn;
 245
 246         /* The first NDADDR blocks are direct blocks. */
 247         if (bn < NDADDR)
 248                 return (0);
 249
 250         /*
 251          * Determine the number of levels of indirection.  After this loop
 252          * is done, blockcnt indicates the number of data blocks possible
 253          * at the previous level of indirection, and NIADDR - i is the number
 254          * of levels of indirection needed to locate the requested block.
 255          */
 256         for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
 257                 if (i == 0)
 258                         return (EFBIG);
 259                 /*
 260                  * Use int64_t's here to avoid overflow for triple indirect
 261                  * blocks when longs have 32 bits and the block size is more
 262                  * than 4K.
 263                  */
 264                 qblockcnt = (int64_t)blockcnt * MNINDIR(ump);
 265                 if (bn < qblockcnt)
 266                         break;
 267                 blockcnt = qblockcnt;
 268         }
 269
 270         /* Calculate the address of the first meta-block. */
 271         if (realbn >= 0)
 272                 metalbn = -(realbn - bn + NIADDR - i);
 273         else
 274                 metalbn = -(-realbn - bn + NIADDR - i);
 275
 276         /*
 277          * At each iteration, off is the offset into the bap array which is
 278          * an array of disk addresses at the current level of indirection.
 279          * The logical block number and the offset in that block are stored
 280          * into the argument array.
 281          */
 282         ap->in_lbn = metalbn;
 283         ap->in_off = off = NIADDR - i;
 284         ap->in_exists = 0;
 285         ap++;
 286         for (++numlevels; i <= NIADDR; i++) {
 287                 /* If searching for a meta-data block, quit when found. */
 288                 if (metalbn == realbn)
 289                         break;
 290
 291                 off = (bn / blockcnt) % MNINDIR(ump);
 292
 293                 ++numlevels;
 294                 ap->in_lbn = metalbn;
 295                 ap->in_off = off;
 296                 ap->in_exists = 0;
 297                 ++ap;
 298
 299                 metalbn -= -1 + off * blockcnt;
 300                 blockcnt /= MNINDIR(ump);
 301         }
 302         if (nump)
 303                 *nump = numlevels;
 304         return (0);
 305 }