1 /*- 2 * Copyright (c) 1980, 1988, 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)traverse.c 8.7 (Berkeley) 6/15/95 30 * $FreeBSD: src/sbin/dump/traverse.c,v 1.10.2.6 2003/04/14 20:10:35 johan Exp $ 31 */ 32 33 #include <sys/param.h> 34 #include <sys/stat.h> 35 #include <vfs/ufs/dir.h> 36 #include <vfs/ufs/dinode.h> 37 #include <vfs/ufs/fs.h> 38 39 #include <protocols/dumprestore.h> 40 41 #include <ctype.h> 42 #include <stdio.h> 43 #include <errno.h> 44 #include <string.h> 45 #include <unistd.h> 46 47 #include "dump.h" 48 49 #define HASDUMPEDFILE 0x1 50 #define HASSUBDIRS 0x2 51 52 #ifdef FS_44INODEFMT 53 typedef quad_t fsizeT; 54 #else 55 typedef long fsizeT; 56 #endif 57 58 static int dirindir(ufs1_ino_t, daddr_t, int, long *, long *, int); 59 static void dmpindir(ufs1_ino_t, daddr_t, int, fsizeT *); 60 static int searchdir(ufs1_ino_t, daddr_t, long, long, long *, int); 61 62 /* 63 * This is an estimation of the number of TP_BSIZE blocks in the file. 64 * It estimates the number of blocks in files with holes by assuming 65 * that all of the blocks accounted for by di_blocks are data blocks 66 * (when some of the blocks are usually used for indirect pointers); 67 * hence the estimate may be high. 68 */ 69 long 70 blockest(struct ufs1_dinode *dp) 71 { 72 long blkest, sizeest; 73 74 /* 75 * dp->di_size is the size of the file in bytes. 76 * dp->di_blocks stores the number of sectors actually in the file. 77 * If there are more sectors than the size would indicate, this just 78 * means that there are indirect blocks in the file or unused 79 * sectors in the last file block; we can safely ignore these 80 * (blkest = sizeest below). 81 * If the file is bigger than the number of sectors would indicate, 82 * then the file has holes in it. In this case we must use the 83 * block count to estimate the number of data blocks used, but 84 * we use the actual size for estimating the number of indirect 85 * dump blocks (sizeest vs. blkest in the indirect block 86 * calculation). 87 */ 88 blkest = howmany(dbtob(dp->di_blocks), TP_BSIZE); 89 sizeest = howmany(dp->di_size, TP_BSIZE); 90 if (blkest > sizeest) 91 blkest = sizeest; 92 if (dp->di_size > (unsigned)sblock->fs_bsize * NDADDR) { 93 /* calculate the number of indirect blocks on the dump tape */ 94 blkest += 95 howmany(sizeest - NDADDR * sblock->fs_bsize / TP_BSIZE, 96 TP_NINDIR); 97 } 98 return (blkest + 1); 99 } 100 101 /* Auxiliary macro to pick up files changed since previous dump. */ 102 #define CHANGEDSINCE(dp, t) \ 103 ((dp)->di_mtime >= (t) || (dp)->di_ctime >= (t)) 104 105 /* The WANTTODUMP macro decides whether a file should be dumped. */ 106 #ifdef UF_NODUMP 107 #define WANTTODUMP(dp) \ 108 (CHANGEDSINCE(dp, spcl.c_ddate) && \ 109 (nonodump || ((dp)->di_flags & UF_NODUMP) != UF_NODUMP)) 110 #else 111 #define WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate) 112 #endif 113 114 /* 115 * Dump pass 1. 116 * 117 * Walk the inode list for a filesystem to find all allocated inodes 118 * that have been modified since the previous dump time. Also, find all 119 * the directories in the filesystem. 120 */ 121 int 122 mapfiles(ufs1_ino_t maxino, long *tape_size) 123 { 124 int mode; 125 ufs1_ino_t ino; 126 struct ufs1_dinode *dp; 127 int anydirskipped = 0; 128 129 for (ino = ROOTINO; ino < maxino; ino++) { 130 dp = getino(ino); 131 if ((mode = (dp->di_mode & IFMT)) == 0) 132 continue; 133 /* 134 * Everything must go in usedinomap so that a check 135 * for "in dumpdirmap but not in usedinomap" to detect 136 * dirs with nodump set has a chance of succeeding 137 * (this is used in mapdirs()). 138 */ 139 SETINO(ino, usedinomap); 140 if (mode == IFDIR) 141 SETINO(ino, dumpdirmap); 142 if (WANTTODUMP(dp)) { 143 SETINO(ino, dumpinomap); 144 if (mode != IFREG && mode != IFDIR && mode != IFLNK) 145 *tape_size += 1; 146 else 147 *tape_size += blockest(dp); 148 continue; 149 } 150 if (mode == IFDIR) { 151 if (!nonodump && (dp->di_flags & UF_NODUMP)) 152 CLRINO(ino, usedinomap); 153 anydirskipped = 1; 154 } 155 } 156 /* 157 * Restore gets very upset if the root is not dumped, 158 * so ensure that it always is dumped. 159 */ 160 SETINO(ROOTINO, dumpinomap); 161 return (anydirskipped); 162 } 163 164 /* 165 * Dump pass 2. 166 * 167 * Scan each directory on the filesystem to see if it has any modified 168 * files in it. If it does, and has not already been added to the dump 169 * list (because it was itself modified), then add it. If a directory 170 * has not been modified itself, contains no modified files and has no 171 * subdirectories, then it can be deleted from the dump list and from 172 * the list of directories. By deleting it from the list of directories, 173 * its parent may now qualify for the same treatment on this or a later 174 * pass using this algorithm. 175 */ 176 int 177 mapdirs(ufs1_ino_t maxino, long *tape_size) 178 { 179 struct ufs1_dinode *dp; 180 int isdir, nodump; 181 unsigned int i; 182 char *map; 183 ufs1_ino_t ino; 184 struct ufs1_dinode di; 185 long filesize; 186 int ret, change = 0; 187 188 isdir = 0; /* XXX just to get gcc to shut up */ 189 for (map = dumpdirmap, ino = 1; ino < maxino; ino++) { 190 if (((ino - 1) % NBBY) == 0) /* map is offset by 1 */ 191 isdir = *map++; 192 else 193 isdir >>= 1; 194 /* 195 * If a directory has been removed from usedinomap, it 196 * either has the nodump flag set, or has inherited 197 * it. Although a directory can't be in dumpinomap if 198 * it isn't in usedinomap, we have to go through it to 199 * propagate the nodump flag. 200 */ 201 nodump = !nonodump && (TSTINO(ino, usedinomap) == 0); 202 if ((isdir & 1) == 0 || (TSTINO(ino, dumpinomap) && !nodump)) 203 continue; 204 dp = getino(ino); 205 di = *dp; /* inode buf may change in searchdir(). */ 206 filesize = di.di_size; 207 for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) { 208 if (di.di_db[i] != 0) { 209 ret |= searchdir(ino, di.di_db[i], 210 (long)dblksize(sblock, &di, i), 211 filesize, tape_size, nodump); 212 } 213 if (ret & HASDUMPEDFILE) 214 filesize = 0; 215 else 216 filesize -= sblock->fs_bsize; 217 } 218 for (i = 0; filesize > 0 && i < NIADDR; i++) { 219 if (di.di_ib[i] == 0) 220 continue; 221 ret |= dirindir(ino, di.di_ib[i], i, &filesize, 222 tape_size, nodump); 223 } 224 if (ret & HASDUMPEDFILE) { 225 SETINO(ino, dumpinomap); 226 *tape_size += blockest(&di); 227 change = 1; 228 continue; 229 } 230 if (nodump) { 231 if (ret & HASSUBDIRS) 232 change = 1; /* subdirs inherit nodump */ 233 CLRINO(ino, dumpdirmap); 234 } else if ((ret & HASSUBDIRS) == 0) 235 if (!TSTINO(ino, dumpinomap)) { 236 CLRINO(ino, dumpdirmap); 237 change = 1; 238 } 239 } 240 return (change); 241 } 242 243 /* 244 * Read indirect blocks, and pass the data blocks to be searched 245 * as directories. Quit as soon as any entry is found that will 246 * require the directory to be dumped. 247 */ 248 static int 249 dirindir(ufs1_ino_t ino, daddr_t blkno, int ind_level, long *filesize, 250 long *tape_size, int nodump) 251 { 252 int ret = 0; 253 int i; 254 daddr_t idblk[MAXNINDIR]; 255 256 bread(fsbtodb(sblock, blkno), (char *)idblk, (int)sblock->fs_bsize); 257 if (ind_level <= 0) { 258 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) { 259 blkno = idblk[i]; 260 if (blkno != 0) { 261 ret |= searchdir(ino, blkno, sblock->fs_bsize, 262 *filesize, tape_size, nodump); 263 } 264 if (ret & HASDUMPEDFILE) 265 *filesize = 0; 266 else 267 *filesize -= sblock->fs_bsize; 268 } 269 return (ret); 270 } 271 ind_level--; 272 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) { 273 blkno = idblk[i]; 274 if (blkno != 0) { 275 ret |= dirindir(ino, blkno, ind_level, filesize, 276 tape_size, nodump); 277 } 278 } 279 return (ret); 280 } 281 282 /* 283 * Scan a disk block containing directory information looking to see if 284 * any of the entries are on the dump list and to see if the directory 285 * contains any subdirectories. 286 */ 287 static int 288 searchdir(ufs1_ino_t ino, daddr_t blkno, long size, long filesize, 289 long *tape_size, int nodump) 290 { 291 struct direct *dp; 292 struct ufs1_dinode *ip; 293 long loc, ret = 0; 294 char dblk[MAXBSIZE]; 295 296 bread(fsbtodb(sblock, blkno), dblk, (int)size); 297 if (filesize < size) 298 size = filesize; 299 for (loc = 0; loc < size; ) { 300 dp = (struct direct *)(dblk + loc); 301 if (dp->d_reclen == 0) { 302 msg("corrupted directory, inumber %d\n", ino); 303 break; 304 } 305 loc += dp->d_reclen; 306 if (dp->d_ino == 0) 307 continue; 308 if (dp->d_name[0] == '.') { 309 if (dp->d_name[1] == '\0') 310 continue; 311 if (dp->d_name[1] == '.' && dp->d_name[2] == '\0') 312 continue; 313 } 314 if (nodump) { 315 ip = getino(dp->d_ino); 316 if (TSTINO(dp->d_ino, dumpinomap)) { 317 CLRINO(dp->d_ino, dumpinomap); 318 *tape_size -= blockest(ip); 319 } 320 /* 321 * Add back to dumpdirmap and remove from usedinomap 322 * to propagate nodump. 323 */ 324 if ((ip->di_mode & IFMT) == IFDIR) { 325 SETINO(dp->d_ino, dumpdirmap); 326 CLRINO(dp->d_ino, usedinomap); 327 ret |= HASSUBDIRS; 328 } 329 } else { 330 if (TSTINO(dp->d_ino, dumpinomap)) { 331 ret |= HASDUMPEDFILE; 332 if (ret & HASSUBDIRS) 333 break; 334 } 335 if (TSTINO(dp->d_ino, dumpdirmap)) { 336 ret |= HASSUBDIRS; 337 if (ret & HASDUMPEDFILE) 338 break; 339 } 340 } 341 } 342 return (ret); 343 } 344 345 /* 346 * Dump passes 3 and 4. 347 * 348 * Dump the contents of an inode to tape. 349 */ 350 void 351 dumpino(struct ufs1_dinode *dp, ufs1_ino_t ino) 352 { 353 int ind_level, cnt; 354 fsizeT size; 355 char buf[TP_BSIZE]; 356 357 if (newtape) { 358 newtape = 0; 359 dumpmap(dumpinomap, TS_BITS, ino); 360 } 361 CLRINO(ino, dumpinomap); 362 spcl.c_dinode = *dp; 363 spcl.c_type = TS_INODE; 364 spcl.c_count = 0; 365 switch (dp->di_mode & S_IFMT) { 366 367 case 0: 368 /* 369 * Freed inode. 370 */ 371 return; 372 373 case S_IFLNK: 374 /* 375 * Check for short symbolic link. 376 */ 377 #ifdef FS_44INODEFMT 378 if (dp->di_size > 0 && 379 dp->di_size < (unsigned)sblock->fs_maxsymlinklen) { 380 spcl.c_addr[0] = 1; 381 spcl.c_count = 1; 382 writeheader(ino); 383 memmove(buf, dp->di_shortlink, (u_long)dp->di_size); 384 buf[dp->di_size] = '\0'; 385 writerec(buf, 0); 386 return; 387 } 388 #endif 389 /* fall through */ 390 391 case S_IFDIR: 392 case S_IFREG: 393 if (dp->di_size > 0) 394 break; 395 /* fall through */ 396 397 case S_IFIFO: 398 case S_IFSOCK: 399 case S_IFCHR: 400 case S_IFBLK: 401 writeheader(ino); 402 return; 403 404 default: 405 msg("Warning: undefined file type 0%o\n", dp->di_mode & IFMT); 406 return; 407 } 408 if (dp->di_size > NDADDR * (unsigned)sblock->fs_bsize) 409 cnt = NDADDR * sblock->fs_frag; 410 else 411 cnt = howmany(dp->di_size, sblock->fs_fsize); 412 blksout(&dp->di_db[0], cnt, ino); 413 if ((size = dp->di_size - NDADDR * sblock->fs_bsize) <= 0) 414 return; 415 for (ind_level = 0; ind_level < NIADDR; ind_level++) { 416 dmpindir(ino, dp->di_ib[ind_level], ind_level, &size); 417 if (size <= 0) 418 return; 419 } 420 } 421 422 /* 423 * Read indirect blocks, and pass the data blocks to be dumped. 424 */ 425 static void 426 dmpindir(ufs1_ino_t ino, daddr_t blk, int ind_level, fsizeT *size) 427 { 428 int i, cnt; 429 daddr_t idblk[MAXNINDIR]; 430 431 if (blk != 0) 432 bread(fsbtodb(sblock, blk), (char *)idblk, (int) sblock->fs_bsize); 433 else 434 memset(idblk, 0, (int)sblock->fs_bsize); 435 if (ind_level <= 0) { 436 if (*size < NINDIR(sblock) * sblock->fs_bsize) 437 cnt = howmany(*size, sblock->fs_fsize); 438 else 439 cnt = NINDIR(sblock) * sblock->fs_frag; 440 *size -= NINDIR(sblock) * sblock->fs_bsize; 441 blksout(&idblk[0], cnt, ino); 442 return; 443 } 444 ind_level--; 445 for (i = 0; i < NINDIR(sblock); i++) { 446 dmpindir(ino, idblk[i], ind_level, size); 447 if (*size <= 0) 448 return; 449 } 450 } 451 452 /* 453 * Collect up the data into tape record sized buffers and output them. 454 */ 455 void 456 blksout(daddr_t *blkp, int frags, ufs1_ino_t ino) 457 { 458 daddr_t *bp; 459 int i, j, count, blks, tbperdb; 460 461 blks = howmany(frags * sblock->fs_fsize, TP_BSIZE); 462 tbperdb = sblock->fs_bsize >> tp_bshift; 463 for (i = 0; i < blks; i += TP_NINDIR) { 464 if (i + TP_NINDIR > blks) 465 count = blks; 466 else 467 count = i + TP_NINDIR; 468 for (j = i; j < count; j++) 469 if (blkp[j / tbperdb] != 0) 470 spcl.c_addr[j - i] = 1; 471 else 472 spcl.c_addr[j - i] = 0; 473 spcl.c_count = count - i; 474 writeheader(ino); 475 bp = &blkp[i / tbperdb]; 476 for (j = i; j < count; j += tbperdb, bp++) 477 if (*bp != 0) { 478 if (j + tbperdb <= count) 479 dumpblock(*bp, (int)sblock->fs_bsize); 480 else 481 dumpblock(*bp, (count - j) * TP_BSIZE); 482 } 483 spcl.c_type = TS_ADDR; 484 } 485 } 486 487 /* 488 * Dump a map to the tape. 489 */ 490 void 491 dumpmap(const char *map, int type, ufs1_ino_t ino) 492 { 493 int i; 494 const char *cp; 495 496 spcl.c_type = type; 497 spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE); 498 writeheader(ino); 499 for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE) 500 writerec(cp, 0); 501 } 502 503 /* 504 * Write a header record to the dump tape. 505 */ 506 void 507 writeheader(ufs1_ino_t ino) 508 { 509 int32_t sum, cnt, *lp; 510 511 spcl.c_inumber = ino; 512 spcl.c_magic = NFS_MAGIC; 513 spcl.c_checksum = 0; 514 lp = (int32_t *)&spcl; 515 sum = 0; 516 cnt = sizeof(union u_spcl) / (4 * sizeof(int32_t)); 517 while (--cnt >= 0) { 518 sum += *lp++; 519 sum += *lp++; 520 sum += *lp++; 521 sum += *lp++; 522 } 523 spcl.c_checksum = CHECKSUM - sum; 524 writerec(&spcl, 1); 525 } 526 527 struct ufs1_dinode * 528 getino(ufs1_ino_t inum) 529 { 530 static daddr_t minino, maxino; 531 static struct ufs1_dinode inoblock[MAXINOPB]; 532 533 curino = inum; 534 if (inum >= (unsigned)minino && inum < (unsigned)maxino) 535 return (&inoblock[inum - minino]); 536 bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), (char *)inoblock, 537 (int)sblock->fs_bsize); 538 minino = inum - (inum % INOPB(sblock)); 539 maxino = minino + INOPB(sblock); 540 return (&inoblock[inum - minino]); 541 } 542 543 /* 544 * Read a chunk of data from the disk. 545 * Try to recover from hard errors by reading in sector sized pieces. 546 * Error recovery is attempted at most BREADEMAX times before seeking 547 * consent from the operator to continue. 548 */ 549 int breaderrors = 0; 550 #define BREADEMAX 32 551 552 void 553 bread(daddr_t blkno, char *buf, int size) 554 { 555 int cnt, i; 556 557 loop: 558 cnt = cread(diskfd, buf, size, ((off_t)blkno << dev_bshift)); 559 if (cnt == size) 560 return; 561 if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) { 562 /* 563 * Trying to read the final fragment. 564 * 565 * NB - dump only works in TP_BSIZE blocks, hence 566 * rounds `dev_bsize' fragments up to TP_BSIZE pieces. 567 * It should be smarter about not actually trying to 568 * read more than it can get, but for the time being 569 * we punt and scale back the read only when it gets 570 * us into trouble. (mkm 9/25/83) 571 */ 572 size -= dev_bsize; 573 goto loop; 574 } 575 if (cnt == -1) 576 msg("read error from %s: %s: [block %d]: count=%d\n", 577 disk, strerror(errno), blkno, size); 578 else 579 msg("short read error from %s: [block %d]: count=%d, got=%d\n", 580 disk, blkno, size, cnt); 581 if (++breaderrors > BREADEMAX) { 582 msg("More than %d block read errors from %s\n", 583 BREADEMAX, disk); 584 broadcast("DUMP IS AILING!\n"); 585 msg("This is an unrecoverable error.\n"); 586 if (!query("Do you want to attempt to continue?")){ 587 dumpabort(0); 588 /*NOTREACHED*/ 589 } else 590 breaderrors = 0; 591 } 592 /* 593 * Zero buffer, then try to read each sector of buffer separately, 594 * and bypass the cache. 595 */ 596 memset(buf, 0, size); 597 for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) { 598 if ((cnt = pread(diskfd, buf, (int)dev_bsize, 599 ((off_t)blkno << dev_bshift))) == dev_bsize) 600 continue; 601 if (cnt == -1) { 602 msg("read error from %s: %s: [sector %d]: count=%ld\n", 603 disk, strerror(errno), blkno, dev_bsize); 604 continue; 605 } 606 msg("short read error from %s: [sector %d]: count=%ld, got=%d\n", 607 disk, blkno, dev_bsize, cnt); 608 } 609 } 610