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 int ret = 0; 100 101 ip = VTOI(vp); 102 fs = ip->i_e2fs; 103 lbn = bn; 104 105 /* 106 * TODO: need to implement read ahead to improve the performance. 107 */ 108 if (runp != NULL) 109 *runp = 0; 110 111 if (runb != NULL) 112 *runb = 0; 113 114 ext4_ext_find_extent(fs, ip, lbn, &path); 115 ep = path.ep_ext; 116 if (ep == NULL) 117 ret = EIO; 118 else { 119 *bnp = fsbtodb(fs, lbn - ep->e_blk + 120 (ep->e_start_lo | (daddr_t)ep->e_start_hi << 32)); 121 122 if (*bnp == 0) 123 *bnp = -1; 124 } 125 126 if (path.ep_bp != NULL) { 127 brelse(path.ep_bp); 128 path.ep_bp = NULL; 129 } 130 131 return (ret); 132 } 133 134 /* 135 * Indirect blocks are now on the vnode for the file. They are given negative 136 * logical block numbers. Indirect blocks are addressed by the negative 137 * address of the first data block to which they point. Double indirect blocks 138 * are addressed by one less than the address of the first indirect block to 139 * which they point. Triple indirect blocks are addressed by one less than 140 * the address of the first double indirect block to which they point. 141 * 142 * ext2_bmaparray does the bmap conversion, and if requested returns the 143 * array of logical blocks which must be traversed to get to a block. 144 * Each entry contains the offset into that block that gets you to the 145 * next block and the disk address of the block (if it is assigned). 146 */ 147 148 int 149 ext2_bmaparray(struct vnode *vp, daddr_t bn, daddr_t *bnp, int *runp, int *runb) 150 { 151 struct inode *ip; 152 struct buf *bp; 153 struct ext2mount *ump; 154 struct mount *mp; 155 struct indir a[NIADDR+1], *ap; 156 daddr_t daddr; 157 e2fs_lbn_t metalbn; 158 int error, num, maxrun = 0, bsize; 159 int *nump; 160 161 ap = NULL; 162 ip = VTOI(vp); 163 mp = vp->v_mount; 164 ump = VFSTOEXT2(mp); 165 166 bsize = EXT2_BLOCK_SIZE(ump->um_e2fs); 167 168 if (runp) { 169 maxrun = mp->mnt_iosize_max / bsize - 1; 170 *runp = 0; 171 } 172 173 if (runb) { 174 *runb = 0; 175 } 176 177 178 ap = a; 179 nump = # 180 error = ext2_getlbns(vp, bn, ap, nump); 181 if (error) 182 return (error); 183 184 num = *nump; 185 if (num == 0) { 186 *bnp = blkptrtodb(ump, ip->i_db[bn]); 187 if (*bnp == 0) { 188 *bnp = -1; 189 } else if (runp) { 190 daddr_t bnb = bn; 191 for (++bn; bn < NDADDR && *runp < maxrun && 192 is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]); 193 ++bn, ++*runp); 194 bn = bnb; 195 if (runb && (bn > 0)) { 196 for (--bn; (bn >= 0) && (*runb < maxrun) && 197 is_sequential(ump, ip->i_db[bn], 198 ip->i_db[bn + 1]); 199 --bn, ++*runb); 200 } 201 } 202 return (0); 203 } 204 205 206 /* Get disk address out of indirect block array */ 207 daddr = ip->i_ib[ap->in_off]; 208 209 for (bp = NULL, ++ap; --num; ++ap) { 210 /* 211 * Exit the loop if there is no disk address assigned yet and 212 * the indirect block isn't in the cache, or if we were 213 * looking for an indirect block and we've found it. 214 */ 215 216 metalbn = ap->in_lbn; 217 if ((daddr == 0 && !incore(&vp->v_bufobj, metalbn)) || metalbn == bn) 218 break; 219 /* 220 * If we get here, we've either got the block in the cache 221 * or we have a disk address for it, go fetch it. 222 */ 223 if (bp) 224 bqrelse(bp); 225 226 bp = getblk(vp, metalbn, bsize, 0, 0, 0); 227 if ((bp->b_flags & B_CACHE) == 0) { 228 #ifdef INVARIANTS 229 if (!daddr) 230 panic("ext2_bmaparray: indirect block not in cache"); 231 #endif 232 bp->b_blkno = blkptrtodb(ump, daddr); 233 bp->b_iocmd = BIO_READ; 234 bp->b_flags &= ~B_INVAL; 235 bp->b_ioflags &= ~BIO_ERROR; 236 vfs_busy_pages(bp, 0); 237 bp->b_iooffset = dbtob(bp->b_blkno); 238 bstrategy(bp); 239 curthread->td_ru.ru_inblock++; 240 error = bufwait(bp); 241 if (error) { 242 brelse(bp); 243 return (error); 244 } 245 } 246 247 daddr = ((e2fs_daddr_t *)bp->b_data)[ap->in_off]; 248 if (num == 1 && daddr && runp) { 249 for (bn = ap->in_off + 1; 250 bn < MNINDIR(ump) && *runp < maxrun && 251 is_sequential(ump, 252 ((e2fs_daddr_t *)bp->b_data)[bn - 1], 253 ((e2fs_daddr_t *)bp->b_data)[bn]); 254 ++bn, ++*runp); 255 bn = ap->in_off; 256 if (runb && bn) { 257 for (--bn; bn >= 0 && *runb < maxrun && 258 is_sequential(ump, 259 ((e2fs_daddr_t *)bp->b_data)[bn], 260 ((e2fs_daddr_t *)bp->b_data)[bn + 1]); 261 --bn, ++*runb); 262 } 263 } 264 } 265 if (bp) 266 bqrelse(bp); 267 268 /* 269 * Since this is FFS independent code, we are out of scope for the 270 * definitions of BLK_NOCOPY and BLK_SNAP, but we do know that they 271 * will fall in the range 1..um_seqinc, so we use that test and 272 * return a request for a zeroed out buffer if attempts are made 273 * to read a BLK_NOCOPY or BLK_SNAP block. 274 */ 275 if ((ip->i_flags & SF_SNAPSHOT) && daddr > 0 && daddr < ump->um_seqinc){ 276 *bnp = -1; 277 return (0); 278 } 279 *bnp = blkptrtodb(ump, daddr); 280 if (*bnp == 0) { 281 *bnp = -1; 282 } 283 return (0); 284 } 285 286 /* 287 * Create an array of logical block number/offset pairs which represent the 288 * path of indirect blocks required to access a data block. The first "pair" 289 * contains the logical block number of the appropriate single, double or 290 * triple indirect block and the offset into the inode indirect block array. 291 * Note, the logical block number of the inode single/double/triple indirect 292 * block appears twice in the array, once with the offset into the i_ib and 293 * once with the offset into the page itself. 294 */ 295 int 296 ext2_getlbns(struct vnode *vp, daddr_t bn, struct indir *ap, int *nump) 297 { 298 long blockcnt; 299 e2fs_lbn_t metalbn, realbn; 300 struct ext2mount *ump; 301 int i, numlevels, off; 302 int64_t qblockcnt; 303 304 ump = VFSTOEXT2(vp->v_mount); 305 if (nump) 306 *nump = 0; 307 numlevels = 0; 308 realbn = bn; 309 if ((long)bn < 0) 310 bn = -(long)bn; 311 312 /* The first NDADDR blocks are direct blocks. */ 313 if (bn < NDADDR) 314 return (0); 315 316 /* 317 * Determine the number of levels of indirection. After this loop 318 * is done, blockcnt indicates the number of data blocks possible 319 * at the previous level of indirection, and NIADDR - i is the number 320 * of levels of indirection needed to locate the requested block. 321 */ 322 for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) { 323 if (i == 0) 324 return (EFBIG); 325 /* 326 * Use int64_t's here to avoid overflow for triple indirect 327 * blocks when longs have 32 bits and the block size is more 328 * than 4K. 329 */ 330 qblockcnt = (int64_t)blockcnt * MNINDIR(ump); 331 if (bn < qblockcnt) 332 break; 333 blockcnt = qblockcnt; 334 } 335 336 /* Calculate the address of the first meta-block. */ 337 if (realbn >= 0) 338 metalbn = -(realbn - bn + NIADDR - i); 339 else 340 metalbn = -(-realbn - bn + NIADDR - i); 341 342 /* 343 * At each iteration, off is the offset into the bap array which is 344 * an array of disk addresses at the current level of indirection. 345 * The logical block number and the offset in that block are stored 346 * into the argument array. 347 */ 348 ap->in_lbn = metalbn; 349 ap->in_off = off = NIADDR - i; 350 ap++; 351 for (++numlevels; i <= NIADDR; i++) { 352 /* If searching for a meta-data block, quit when found. */ 353 if (metalbn == realbn) 354 break; 355 356 off = (bn / blockcnt) % MNINDIR(ump); 357 358 ++numlevels; 359 ap->in_lbn = metalbn; 360 ap->in_off = off; 361 ++ap; 362 363 metalbn -= -1 + off * blockcnt; 364 blockcnt /= MNINDIR(ump); 365 } 366 if (nump) 367 *nump = numlevels; 368 return (0); 369 } 370