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 * $FreeBSD$ 36 */ 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/bio.h> 41 #include <sys/buf.h> 42 #include <sys/proc.h> 43 #include <sys/vnode.h> 44 #include <sys/mount.h> 45 #include <sys/resourcevar.h> 46 #include <sys/stat.h> 47 48 #include <fs/ext2fs/inode.h> 49 #include <fs/ext2fs/fs.h> 50 #include <fs/ext2fs/ext2fs.h> 51 #include <fs/ext2fs/ext2_dinode.h> 52 #include <fs/ext2fs/ext2_extern.h> 53 #include <fs/ext2fs/ext2_mount.h> 54 55 static int ext4_bmapext(struct vnode *, int32_t, int64_t *, int *, int *); 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 */ 62 int 63 ext2_bmap(struct vop_bmap_args *ap) 64 { 65 daddr_t blkno; 66 int error; 67 68 /* 69 * Check for underlying vnode requests and ensure that logical 70 * to physical mapping is requested. 71 */ 72 if (ap->a_bop != NULL) 73 *ap->a_bop = &VTOI(ap->a_vp)->i_devvp->v_bufobj; 74 if (ap->a_bnp == NULL) 75 return (0); 76 77 if (VTOI(ap->a_vp)->i_flag & IN_E4EXTENTS) 78 error = ext4_bmapext(ap->a_vp, ap->a_bn, &blkno, 79 ap->a_runp, ap->a_runb); 80 else 81 error = ext2_bmaparray(ap->a_vp, ap->a_bn, &blkno, 82 ap->a_runp, ap->a_runb); 83 *ap->a_bnp = blkno; 84 return (error); 85 } 86 87 /* 88 * This function converts the logical block number of a file to 89 * its physical block number on the disk within ext4 extents. 90 */ 91 static int 92 ext4_bmapext(struct vnode *vp, int32_t bn, int64_t *bnp, int *runp, int *runb) 93 { 94 struct inode *ip; 95 struct m_ext2fs *fs; 96 struct ext4_extent *ep; 97 struct ext4_extent_path path = { .ep_bp = NULL }; 98 daddr_t lbn; 99 100 ip = VTOI(vp); 101 fs = ip->i_e2fs; 102 lbn = bn; 103 104 /* 105 * TODO: need to implement read ahead to improve the performance. 106 */ 107 if (runp != NULL) 108 *runp = 0; 109 110 if (runb != NULL) 111 *runb = 0; 112 113 ext4_ext_find_extent(fs, ip, lbn, &path); 114 ep = path.ep_ext; 115 if (ep == NULL) 116 return (EIO); 117 118 *bnp = fsbtodb(fs, lbn - ep->e_blk + 119 (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32)); 120 121 if (*bnp == 0) 122 *bnp = -1; 123 124 return (0); 125 } 126 127 /* 128 * Indirect blocks are now on the vnode for the file. They are given negative 129 * logical block numbers. Indirect blocks are addressed by the negative 130 * address of the first data block to which they point. Double indirect blocks 131 * are addressed by one less than the address of the first indirect block to 132 * which they point. Triple indirect blocks are addressed by one less than 133 * the address of the first double indirect block to which they point. 134 * 135 * ext2_bmaparray does the bmap conversion, and if requested returns the 136 * array of logical blocks which must be traversed to get to a block. 137 * Each entry contains the offset into that block that gets you to the 138 * next block and the disk address of the block (if it is assigned). 139 */ 140 141 int 142 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb) 143 { 144 struct inode *ip; 145 struct buf *bp; 146 struct ext2mount *ump; 147 struct mount *mp; 148 struct indir a[NIADDR+1], *ap; 149 daddr_t daddr; 150 e2fs_lbn_t metalbn; 151 int error, num, maxrun = 0, bsize; 152 int *nump; 153 154 ap = NULL; 155 ip = VTOI(vp); 156 mp = vp->v_mount; 157 ump = VFSTOEXT2(mp); 158 159 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs); 160 161 if (runp) { 162 maxrun = mp->mnt_iosize_max / bsize - 1; 163 *runp = 0; 164 } 165 166 if (runb) { 167 *runb = 0; 168 } 169 170 171 ap = a; 172 nump = # 173 error = ext2_getlbns(vp, bn, ap, nump); 174 if (error) 175 return (error); 176 177 num = *nump; 178 if (num == 0) { 179 *bnp = blkptrtodb(ump, ip->i_db[bn]); 180 if (*bnp == 0) { 181 *bnp = -1; 182 } else if (runp) { 183 daddr_t bnb = bn; 184 for (++bn; bn < NDADDR && *runp < maxrun && 185 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]); 186 ++bn, ++*runp); 187 bn = bnb; 188 if (runb && (bn > 0)) { 189 for (--bn; (bn >= 0) && (*runb < maxrun) && 190 is_sequential(ump, ip->i_db[bn], 191 ip->i_db[bn + 1]); 192 --bn, ++*runb); 193 } 194 } 195 return (0); 196 } 197 198 199 /* Get disk address out of indirect block array */ 200 daddr = ip->i_ib[ap->in_off]; 201 202 for (bp = NULL, ++ap; --num; ++ap) { 203 /* 204 * Exit the loop if there is no disk address assigned yet and 205 * the indirect block isn't in the cache, or if we were 206 * looking for an indirect block and we've found it. 207 */ 208 209 metalbn = ap->in_lbn; 210 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 211 break; 212 /* 213 * If we get here, we've either got the block in the cache 214 * or we have a disk address for it, go fetch it. 215 */ 216 if (bp) 217 bqrelse(bp); 218 219 bp = getblk(vp, metalbn, bsize, 0, 0, 0); 220 if ((bp->b_flags & B_CACHE) == 0) { 221 #ifdef INVARIANTS 222 if (!daddr) 223 panic("ext2_bmaparray: indirect block not in cache"); 224 #endif 225 bp->b_blkno = blkptrtodb(ump, daddr); 226 bp->b_iocmd = BIO_READ; 227 bp->b_flags &= ~B_INVAL; 228 bp->b_ioflags &= ~BIO_ERROR; 229 vfs_busy_pages(bp, 0); 230 bp->b_iooffset = dbtob(bp->b_blkno); 231 bstrategy(bp); 232 curthread->td_ru.ru_inblock++; 233 error = bufwait(bp); 234 if (error) { 235 brelse(bp); 236 return (error); 237 } 238 } 239 240 daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off]; 241 if (num == 1 && daddr && runp) { 242 for (bn = ap->in_off + 1; 243 bn < MNINDIR(ump) && *runp < maxrun && 244 is_sequential(ump, 245 ((e2fs_daddr_t *)bp->b_data)[bn - 1], 246 ((e2fs_daddr_t *)bp->b_data)[bn]); 247 ++bn, ++*runp); 248 bn = ap->in_off; 249 if (runb && bn) { 250 for (--bn; bn >= 0 && *runb < maxrun && 251 is_sequential(ump, 252 ((e2fs_daddr_t *)bp->b_data)[bn], 253 ((e2fs_daddr_t *)bp->b_data)[bn + 1]); 254 --bn, ++*runb); 255 } 256 } 257 } 258 if (bp) 259 bqrelse(bp); 260 261 /* 262 * Since this is FFS independent code, we are out of scope for the 263 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 264 * will fall in the range 1..um_seqinc, so we use that test and 265 * return a request for a zeroed out buffer if attempts are made 266 * to read a BLK_NOCOPY or BLK_SNAP block. 267 */ 268 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){ 269 *bnp = -1; 270 return (0); 271 } 272 *bnp = blkptrtodb(ump, daddr); 273 if (*bnp == 0) { 274 *bnp = -1; 275 } 276 return (0); 277 } 278 279 /* 280 * Create an array of logical block number/offset pairs which represent the 281 * path of indirect blocks required to access a data block. The first "pair" 282 * contains the logical block number of the appropriate single, double or 283 * triple indirect block and the offset into the inode indirect block array. 284 * Note, the logical block number of the inode single/double/triple indirect 285 * block appears twice in the array, once with the offset into the i_ib and 286 * once with the offset into the page itself. 287 */ 288 int 289 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump) 290 { 291 long blockcnt; 292 e2fs_lbn_t metalbn, realbn; 293 struct ext2mount *ump; 294 int i, numlevels, off; 295 int64_t qblockcnt; 296 297 ump = VFSTOEXT2(vp->v_mount); 298 if (nump) 299 *nump = 0; 300 numlevels = 0; 301 realbn = bn; 302 if ((long)bn < 0) 303 bn = -(long)bn; 304 305 /* The first NDADDR blocks are direct blocks. */ 306 if (bn < NDADDR) 307 return (0); 308 309 /* 310 * Determine the number of levels of indirection. After this loop 311 * is done, blockcnt indicates the number of data blocks possible 312 * at the previous level of indirection, and NIADDR - i is the number 313 * of levels of indirection needed to locate the requested block. 314 */ 315 for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) { 316 if (i == 0) 317 return (EFBIG); 318 /* 319 * Use int64_t's here to avoid overflow for triple indirect 320 * blocks when longs have 32 bits and the block size is more 321 * than 4K. 322 */ 323 qblockcnt = (int64_t)blockcnt * MNINDIR(ump); 324 if (bn < qblockcnt) 325 break; 326 blockcnt = qblockcnt; 327 } 328 329 /* Calculate the address of the first meta-block. */ 330 if (realbn >= 0) 331 metalbn = -(realbn - bn + NIADDR - i); 332 else 333 metalbn = -(-realbn - bn + NIADDR - i); 334 335 /* 336 * At each iteration, off is the offset into the bap array which is 337 * an array of disk addresses at the current level of indirection. 338 * The logical block number and the offset in that block are stored 339 * into the argument array. 340 */ 341 ap->in_lbn = metalbn; 342 ap->in_off = off = NIADDR - i; 343 ap++; 344 for (++numlevels; i <= NIADDR; i++) { 345 /* If searching for a meta-data block, quit when found. */ 346 if (metalbn == realbn) 347 break; 348 349 off = (bn / blockcnt) % MNINDIR(ump); 350 351 ++numlevels; 352 ap->in_lbn = metalbn; 353 ap->in_off = off; 354 ++ap; 355 356 metalbn -= -1 + off * blockcnt; 357 blockcnt /= MNINDIR(ump); 358 } 359 if (nump) 360 *nump = numlevels; 361 return (0); 362 } 363