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 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)ufs_bmap.c 8.7 (Berkeley) 3/21/95 35 */ 36 37#include <sys/cdefs.h> 38__FBSDID("$FreeBSD: stable/11/sys/ufs/ufs/ufs_bmap.c 306553 2016-10-01 09:19:43Z kib $"); 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/bio.h> 43#include <sys/buf.h> 44#include <sys/proc.h> 45#include <sys/vnode.h> 46#include <sys/mount.h> 47#include <sys/racct.h> 48#include <sys/resourcevar.h> 49#include <sys/stat.h> 50 51#include <ufs/ufs/extattr.h> 52#include <ufs/ufs/quota.h> 53#include <ufs/ufs/inode.h> 54#include <ufs/ufs/ufsmount.h> 55#include <ufs/ufs/ufs_extern.h> 56 57/* 58 * Bmap converts 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 */ 62int 63ufs_bmap(ap) 64 struct vop_bmap_args /* { 65 struct vnode *a_vp; 66 daddr_t a_bn; 67 struct bufobj **a_bop; 68 daddr_t *a_bnp; 69 int *a_runp; 70 int *a_runb; 71 } */ *ap; 72{ 73 ufs2_daddr_t blkno; 74 int error; 75 76 /* 77 * Check for underlying vnode requests and ensure that logical 78 * to physical mapping is requested. 79 */ 80 if (ap->a_bop != NULL) 81 *ap->a_bop = &VFSTOUFS(ap->a_vp->v_mount)->um_devvp->v_bufobj; 82 if (ap->a_bnp == NULL) 83 return (0); 84 85 error = ufs_bmaparray(ap->a_vp, ap->a_bn, &blkno, NULL, 86 ap->a_runp, ap->a_runb); 87 *ap->a_bnp = blkno; 88 return (error); 89} 90 91/* 92 * Indirect blocks are now on the vnode for the file. They are given negative 93 * logical block numbers. Indirect blocks are addressed by the negative 94 * address of the first data block to which they point. Double indirect blocks 95 * are addressed by one less than the address of the first indirect block to 96 * which they point. Triple indirect blocks are addressed by one less than 97 * the address of the first double indirect block to which they point. 98 * 99 * ufs_bmaparray does the bmap conversion, and if requested returns the 100 * array of logical blocks which must be traversed to get to a block. 101 * Each entry contains the offset into that block that gets you to the 102 * next block and the disk address of the block (if it is assigned). 103 */ 104 105int 106ufs_bmaparray(vp, bn, bnp, nbp, runp, runb) 107 struct vnode *vp; 108 ufs2_daddr_t bn; 109 ufs2_daddr_t *bnp; 110 struct buf *nbp; 111 int *runp; 112 int *runb; 113{ 114 struct inode *ip; 115 struct buf *bp; 116 struct ufsmount *ump; 117 struct mount *mp; 118 struct indir a[NIADDR+1], *ap; 119 ufs2_daddr_t daddr; 120 ufs_lbn_t metalbn; 121 int error, num, maxrun = 0; 122 int *nump; 123 124 ap = NULL; 125 ip = VTOI(vp); 126 mp = vp->v_mount; 127 ump = VFSTOUFS(mp); 128 129 if (runp) { 130 maxrun = mp->mnt_iosize_max / mp->mnt_stat.f_iosize - 1; 131 *runp = 0; 132 } 133 134 if (runb) { 135 *runb = 0; 136 } 137 138 139 ap = a; 140 nump = # 141 error = ufs_getlbns(vp, bn, ap, nump); 142 if (error) 143 return (error); 144 145 num = *nump; 146 if (num == 0) { 147 if (bn >= 0 && bn < NDADDR) { 148 *bnp = blkptrtodb(ump, DIP(ip, i_db[bn])); 149 } else if (bn < 0 && bn >= -NXADDR) { 150 *bnp = blkptrtodb(ump, ip->i_din2->di_extb[-1 - bn]); 151 if (*bnp == 0) 152 *bnp = -1; 153 if (nbp == NULL) 154 panic("ufs_bmaparray: mapping ext data"); 155 nbp->b_xflags |= BX_ALTDATA; 156 return (0); 157 } else { 158 panic("ufs_bmaparray: blkno out of range"); 159 } 160 /* 161 * Since this is FFS independent code, we are out of 162 * scope for the definitions of BLK_NOCOPY and 163 * BLK_SNAP, but we do know that they will fall in 164 * the range 1..um_seqinc, so we use that test and 165 * return a request for a zeroed out buffer if attempts 166 * are made to read a BLK_NOCOPY or BLK_SNAP block. 167 */ 168 if ((ip->i_flags & SF_SNAPSHOT) && DIP(ip, i_db[bn]) > 0 && 169 DIP(ip, i_db[bn]) < ump->um_seqinc) { 170 *bnp = -1; 171 } else if (*bnp == 0) { 172 if (ip->i_flags & SF_SNAPSHOT) 173 *bnp = blkptrtodb(ump, bn * ump->um_seqinc); 174 else 175 *bnp = -1; 176 } else if (runp) { 177 ufs2_daddr_t bnb = bn; 178 for (++bn; bn < NDADDR && *runp < maxrun && 179 is_sequential(ump, DIP(ip, i_db[bn - 1]), 180 DIP(ip, i_db[bn])); 181 ++bn, ++*runp); 182 bn = bnb; 183 if (runb && (bn > 0)) { 184 for (--bn; (bn >= 0) && (*runb < maxrun) && 185 is_sequential(ump, DIP(ip, i_db[bn]), 186 DIP(ip, i_db[bn+1])); 187 --bn, ++*runb); 188 } 189 } 190 return (0); 191 } 192 193 194 /* Get disk address out of indirect block array */ 195 daddr = DIP(ip, i_ib[ap->in_off]); 196 197 for (bp = NULL, ++ap; --num; ++ap) { 198 /* 199 * Exit the loop if there is no disk address assigned yet and 200 * the indirect block isn't in the cache, or if we were 201 * looking for an indirect block and we've found it. 202 */ 203 204 metalbn = ap->in_lbn; 205 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 206 break; 207 /* 208 * If we get here, we've either got the block in the cache 209 * or we have a disk address for it, go fetch it. 210 */ 211 if (bp) 212 bqrelse(bp); 213 214 bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0, 0); 215 if ((bp->b_flags & B_CACHE) == 0) { 216#ifdef INVARIANTS 217 if (!daddr) 218 panic("ufs_bmaparray: indirect block not in cache"); 219#endif 220 bp->b_blkno = blkptrtodb(ump, daddr); 221 bp->b_iocmd = BIO_READ; 222 bp->b_flags &= ~B_INVAL; 223 bp->b_ioflags &= ~BIO_ERROR; 224 vfs_busy_pages(bp, 0); 225 bp->b_iooffset = dbtob(bp->b_blkno); 226 bstrategy(bp); 227#ifdef RACCT 228 if (racct_enable) { 229 PROC_LOCK(curproc); 230 racct_add_buf(curproc, bp, 0); 231 PROC_UNLOCK(curproc); 232 } 233#endif /* RACCT */ 234 curthread->td_ru.ru_inblock++; 235 error = bufwait(bp); 236 if (error) { 237 brelse(bp); 238 return (error); 239 } 240 } 241 242 if (I_IS_UFS1(ip)) { 243 daddr = ((ufs1_daddr_t *)bp->b_data)[ap->in_off]; 244 if (num == 1 && daddr && runp) { 245 for (bn = ap->in_off + 1; 246 bn < MNINDIR(ump) && *runp < maxrun && 247 is_sequential(ump, 248 ((ufs1_daddr_t *)bp->b_data)[bn - 1], 249 ((ufs1_daddr_t *)bp->b_data)[bn]); 250 ++bn, ++*runp); 251 bn = ap->in_off; 252 if (runb && bn) { 253 for (--bn; bn >= 0 && *runb < maxrun && 254 is_sequential(ump, 255 ((ufs1_daddr_t *)bp->b_data)[bn], 256 ((ufs1_daddr_t *)bp->b_data)[bn+1]); 257 --bn, ++*runb); 258 } 259 } 260 continue; 261 } 262 daddr = ((ufs2_daddr_t *)bp->b_data)[ap->in_off]; 263 if (num == 1 && daddr && runp) { 264 for (bn = ap->in_off + 1; 265 bn < MNINDIR(ump) && *runp < maxrun && 266 is_sequential(ump, 267 ((ufs2_daddr_t *)bp->b_data)[bn - 1], 268 ((ufs2_daddr_t *)bp->b_data)[bn]); 269 ++bn, ++*runp); 270 bn = ap->in_off; 271 if (runb && bn) { 272 for (--bn; bn >= 0 && *runb < maxrun && 273 is_sequential(ump, 274 ((ufs2_daddr_t *)bp->b_data)[bn], 275 ((ufs2_daddr_t *)bp->b_data)[bn + 1]); 276 --bn, ++*runb); 277 } 278 } 279 } 280 if (bp) 281 bqrelse(bp); 282 283 /* 284 * Since this is FFS independent code, we are out of scope for the 285 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 286 * will fall in the range 1..um_seqinc, so we use that test and 287 * return a request for a zeroed out buffer if attempts are made 288 * to read a BLK_NOCOPY or BLK_SNAP block. 289 */ 290 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){ 291 *bnp = -1; 292 return (0); 293 } 294 *bnp = blkptrtodb(ump, daddr); 295 if (*bnp == 0) { 296 if (ip->i_flags & SF_SNAPSHOT) 297 *bnp = blkptrtodb(ump, bn * ump->um_seqinc); 298 else 299 *bnp = -1; 300 } 301 return (0); 302} 303 304/* 305 * Create an array of logical block number/offset pairs which represent the 306 * path of indirect blocks required to access a data block. The first "pair" 307 * contains the logical block number of the appropriate single, double or 308 * triple indirect block and the offset into the inode indirect block array. 309 * Note, the logical block number of the inode single/double/triple indirect 310 * block appears twice in the array, once with the offset into the i_ib and 311 * once with the offset into the page itself. 312 */ 313int 314ufs_getlbns(vp, bn, ap, nump) 315 struct vnode *vp; 316 ufs2_daddr_t bn; 317 struct indir *ap; 318 int *nump; 319{ 320 ufs2_daddr_t blockcnt; 321 ufs_lbn_t metalbn, realbn; 322 struct ufsmount *ump; 323 int i, numlevels, off; 324 325 ump = VFSTOUFS(vp->v_mount); 326 if (nump) 327 *nump = 0; 328 numlevels = 0; 329 realbn = bn; 330 if (bn < 0) 331 bn = -bn; 332 333 /* The first NDADDR blocks are direct blocks. */ 334 if (bn < NDADDR) 335 return (0); 336 337 /* 338 * Determine the number of levels of indirection. After this loop 339 * is done, blockcnt indicates the number of data blocks possible 340 * at the previous level of indirection, and NIADDR - i is the number 341 * of levels of indirection needed to locate the requested block. 342 */ 343 for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) { 344 if (i == 0) 345 return (EFBIG); 346 blockcnt *= MNINDIR(ump); 347 if (bn < blockcnt) 348 break; 349 } 350 351 /* Calculate the address of the first meta-block. */ 352 if (realbn >= 0) 353 metalbn = -(realbn - bn + NIADDR - i); 354 else 355 metalbn = -(-realbn - bn + NIADDR - i); 356 357 /* 358 * At each iteration, off is the offset into the bap array which is 359 * an array of disk addresses at the current level of indirection. 360 * The logical block number and the offset in that block are stored 361 * into the argument array. 362 */ 363 ap->in_lbn = metalbn; 364 ap->in_off = off = NIADDR - i; 365 ap++; 366 for (++numlevels; i <= NIADDR; i++) { 367 /* If searching for a meta-data block, quit when found. */ 368 if (metalbn == realbn) 369 break; 370 371 blockcnt /= MNINDIR(ump); 372 off = (bn / blockcnt) % MNINDIR(ump); 373 374 ++numlevels; 375 ap->in_lbn = metalbn; 376 ap->in_off = off; 377 ++ap; 378 379 metalbn -= -1 + off * blockcnt; 380 } 381 if (nump) 382 *nump = numlevels; 383 return (0); 384} 385