1 /* $NetBSD: traverse.c,v 1.19 1997/09/16 06:41:23 lukem Exp $ */ 2 3 /*- 4 * Copyright (c) 1980, 1988, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36 #include <sys/cdefs.h> 37 #ifndef lint 38 #if 0 39 static char sccsid[] = "@(#)traverse.c 8.7 (Berkeley) 6/15/95"; 40 #else 41 __RCSID("$NetBSD: traverse.c,v 1.19 1997/09/16 06:41:23 lukem Exp $"); 42 #endif 43 #endif /* not lint */ 44 45 #include <sys/param.h> 46 #include <sys/time.h> 47 #include <sys/stat.h> 48 #ifdef sunos 49 #include <sys/vnode.h> 50 51 #include <ufs/fs.h> 52 #include <ufs/fsdir.h> 53 #include <ufs/inode.h> 54 #else 55 #include <ufs/ufs/dir.h> 56 #include <ufs/ufs/dinode.h> 57 #include <ufs/ffs/fs.h> 58 #endif 59 60 #include <protocols/dumprestore.h> 61 62 #include <ctype.h> 63 #include <errno.h> 64 #include <fts.h> 65 #include <stdio.h> 66 #ifdef __STDC__ 67 #include <string.h> 68 #include <unistd.h> 69 #endif 70 71 #include "dump.h" 72 73 #define HASDUMPEDFILE 0x1 74 #define HASSUBDIRS 0x2 75 76 #ifdef FS_44INODEFMT 77 typedef quad_t fsizeT; 78 #else 79 typedef int32_t fsizeT; 80 #endif 81 82 static int dirindir __P((ino_t ino, daddr_t blkno, int level, long *size)); 83 static void dmpindir __P((ino_t ino, daddr_t blk, int level, fsizeT *size)); 84 static int searchdir __P((ino_t ino, daddr_t blkno, long size, long filesize)); 85 86 /* 87 * This is an estimation of the number of TP_BSIZE blocks in the file. 88 * It estimates the number of blocks in files with holes by assuming 89 * that all of the blocks accounted for by di_blocks are data blocks 90 * (when some of the blocks are usually used for indirect pointers); 91 * hence the estimate may be high. 92 */ 93 long 94 blockest(dp) 95 struct dinode *dp; 96 { 97 long blkest, sizeest; 98 99 /* 100 * dp->di_size is the size of the file in bytes. 101 * dp->di_blocks stores the number of sectors actually in the file. 102 * If there are more sectors than the size would indicate, this just 103 * means that there are indirect blocks in the file or unused 104 * sectors in the last file block; we can safely ignore these 105 * (blkest = sizeest below). 106 * If the file is bigger than the number of sectors would indicate, 107 * then the file has holes in it. In this case we must use the 108 * block count to estimate the number of data blocks used, but 109 * we use the actual size for estimating the number of indirect 110 * dump blocks (sizeest vs. blkest in the indirect block 111 * calculation). 112 */ 113 blkest = howmany(dbtob(dp->di_blocks), TP_BSIZE); 114 sizeest = howmany(dp->di_size, TP_BSIZE); 115 if (blkest > sizeest) 116 blkest = sizeest; 117 if (dp->di_size > sblock->fs_bsize * NDADDR) { 118 /* calculate the number of indirect blocks on the dump tape */ 119 blkest += 120 howmany(sizeest - NDADDR * sblock->fs_bsize / TP_BSIZE, 121 TP_NINDIR); 122 } 123 return (blkest + 1); 124 } 125 126 /* Auxiliary macro to pick up files changed since previous dump. */ 127 #define CHANGEDSINCE(dp, t) \ 128 ((dp)->di_mtime >= (t) || (dp)->di_ctime >= (t)) 129 130 /* The WANTTODUMP macro decides whether a file should be dumped. */ 131 #ifdef UF_NODUMP 132 #define WANTTODUMP(dp) \ 133 (CHANGEDSINCE(dp, spcl.c_ddate) && \ 134 (nonodump || ((dp)->di_flags & UF_NODUMP) != UF_NODUMP)) 135 #else 136 #define WANTTODUMP(dp) CHANGEDSINCE(dp, spcl.c_ddate) 137 #endif 138 139 /* 140 * Determine if given inode should be dumped 141 */ 142 void 143 mapfileino(ino, tapesize, dirskipped) 144 ino_t ino; 145 long *tapesize; 146 int *dirskipped; 147 { 148 int mode; 149 struct dinode *dp; 150 151 dp = getino(ino); 152 if ((mode = (dp->di_mode & IFMT)) == 0) 153 return; 154 SETINO(ino, usedinomap); 155 if (mode == IFDIR) 156 SETINO(ino, dumpdirmap); 157 if (WANTTODUMP(dp)) { 158 SETINO(ino, dumpinomap); 159 if (mode != IFREG && mode != IFDIR && mode != IFLNK) 160 *tapesize += 1; 161 else 162 *tapesize += blockest(dp); 163 return; 164 } 165 if (mode == IFDIR) 166 *dirskipped = 1; 167 } 168 169 /* 170 * Dump pass 1. 171 * 172 * Walk the inode list for a filesystem to find all allocated inodes 173 * that have been modified since the previous dump time. Also, find all 174 * the directories in the filesystem. 175 */ 176 int 177 mapfiles(maxino, tapesize, disk, dirv) 178 ino_t maxino; 179 long *tapesize; 180 char *disk; 181 char * const *dirv; 182 { 183 int anydirskipped = 0; 184 185 if (dirv != NULL) { 186 char curdir[MAXPATHLEN]; 187 FTS *dirh; 188 FTSENT *entry; 189 int d; 190 191 if (getcwd(curdir, sizeof(curdir)) == NULL) { 192 msg("Can't determine cwd: %s\n", strerror(errno)); 193 dumpabort(0); 194 } 195 if ((dirh = fts_open(dirv, FTS_PHYSICAL|FTS_SEEDOT|FTS_XDEV, 196 NULL)) == NULL) { 197 msg("fts_open failed: %s\n", strerror(errno)); 198 dumpabort(0); 199 } 200 while ((entry = fts_read(dirh)) != NULL) { 201 switch (entry->fts_info) { 202 case FTS_DNR: /* an error */ 203 case FTS_ERR: 204 case FTS_NS: 205 msg("Can't fts_read %s: %s\n", entry->fts_path, 206 strerror(errno)); 207 case FTS_DP: /* already seen dir */ 208 continue; 209 } 210 mapfileino(entry->fts_statp->st_ino, tapesize, 211 &anydirskipped); 212 } 213 (void)fts_close(dirh); 214 215 /* 216 * Add any parent directories 217 */ 218 for (d = 0 ; dirv[d] != NULL ; d++) { 219 char path[MAXPATHLEN]; 220 221 if (dirv[d][0] != '/') 222 (void)snprintf(path, sizeof(path), "%s/%s", 223 curdir, dirv[d]); 224 else 225 (void)snprintf(path, sizeof(path), "%s", 226 dirv[d]); 227 while (strcmp(path, disk) != 0) { 228 char *p; 229 struct stat sb; 230 231 if (*path == '\0') 232 break; 233 if ((p = strrchr(path, '/')) == NULL) 234 break; 235 if (p == path) 236 break; 237 *p = '\0'; 238 if (stat(path, &sb) == -1) { 239 msg("Can't stat %s: %s\n", path, 240 strerror(errno)); 241 break; 242 } 243 mapfileino(sb.st_ino, tapesize, &anydirskipped); 244 } 245 } 246 247 /* 248 * Ensure that the root inode actually appears in the 249 * file list for a subdir 250 */ 251 mapfileino(ROOTINO, tapesize, &anydirskipped); 252 } else { 253 ino_t ino; 254 255 for (ino = ROOTINO; ino < maxino; ino++) { 256 mapfileino(ino, tapesize, &anydirskipped); 257 } 258 } 259 /* 260 * Restore gets very upset if the root is not dumped, 261 * so ensure that it always is dumped. 262 */ 263 SETINO(ROOTINO, dumpinomap); 264 return (anydirskipped); 265 } 266 267 /* 268 * Dump pass 2. 269 * 270 * Scan each directory on the filesystem to see if it has any modified 271 * files in it. If it does, and has not already been added to the dump 272 * list (because it was itself modified), then add it. If a directory 273 * has not been modified itself, contains no modified files and has no 274 * subdirectories, then it can be deleted from the dump list and from 275 * the list of directories. By deleting it from the list of directories, 276 * its parent may now qualify for the same treatment on this or a later 277 * pass using this algorithm. 278 */ 279 int 280 mapdirs(maxino, tapesize) 281 ino_t maxino; 282 long *tapesize; 283 { 284 struct dinode *dp; 285 int i, isdir; 286 char *map; 287 ino_t ino; 288 long filesize; 289 int ret, change = 0; 290 291 isdir = 0; /* XXX just to get gcc to shut up */ 292 for (map = dumpdirmap, ino = 1; ino < maxino; ino++) { 293 if (((ino - 1) % NBBY) == 0) /* map is offset by 1 */ 294 isdir = *map++; 295 else 296 isdir >>= 1; 297 if ((isdir & 1) == 0 || TSTINO(ino, dumpinomap)) 298 continue; 299 dp = getino(ino); 300 filesize = dp->di_size; 301 for (ret = 0, i = 0; filesize > 0 && i < NDADDR; i++) { 302 if (dp->di_db[i] != 0) 303 ret |= searchdir(ino, dp->di_db[i], 304 (long)dblksize(sblock, dp, i), 305 filesize); 306 if (ret & HASDUMPEDFILE) 307 filesize = 0; 308 else 309 filesize -= sblock->fs_bsize; 310 } 311 for (i = 0; filesize > 0 && i < NIADDR; i++) { 312 if (dp->di_ib[i] == 0) 313 continue; 314 ret |= dirindir(ino, dp->di_ib[i], i, &filesize); 315 } 316 if (ret & HASDUMPEDFILE) { 317 SETINO(ino, dumpinomap); 318 *tapesize += blockest(dp); 319 change = 1; 320 continue; 321 } 322 if ((ret & HASSUBDIRS) == 0) { 323 if (!TSTINO(ino, dumpinomap)) { 324 CLRINO(ino, dumpdirmap); 325 change = 1; 326 } 327 } 328 } 329 return (change); 330 } 331 332 /* 333 * Read indirect blocks, and pass the data blocks to be searched 334 * as directories. Quit as soon as any entry is found that will 335 * require the directory to be dumped. 336 */ 337 static int 338 dirindir(ino, blkno, ind_level, filesize) 339 ino_t ino; 340 daddr_t blkno; 341 int ind_level; 342 long *filesize; 343 { 344 int ret = 0; 345 int i; 346 daddr_t idblk[MAXNINDIR]; 347 348 bread(fsbtodb(sblock, blkno), (char *)idblk, (int)sblock->fs_bsize); 349 if (ind_level <= 0) { 350 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) { 351 blkno = idblk[i]; 352 if (blkno != 0) 353 ret |= searchdir(ino, blkno, sblock->fs_bsize, 354 *filesize); 355 if (ret & HASDUMPEDFILE) 356 *filesize = 0; 357 else 358 *filesize -= sblock->fs_bsize; 359 } 360 return (ret); 361 } 362 ind_level--; 363 for (i = 0; *filesize > 0 && i < NINDIR(sblock); i++) { 364 blkno = idblk[i]; 365 if (blkno != 0) 366 ret |= dirindir(ino, blkno, ind_level, filesize); 367 } 368 return (ret); 369 } 370 371 /* 372 * Scan a disk block containing directory information looking to see if 373 * any of the entries are on the dump list and to see if the directory 374 * contains any subdirectories. 375 */ 376 static int 377 searchdir(ino, blkno, size, filesize) 378 ino_t ino; 379 daddr_t blkno; 380 long size; 381 long filesize; 382 { 383 struct direct *dp; 384 long loc, ret = 0; 385 char dblk[MAXBSIZE]; 386 387 bread(fsbtodb(sblock, blkno), dblk, (int)size); 388 if (filesize < size) 389 size = filesize; 390 for (loc = 0; loc < size; ) { 391 dp = (struct direct *)(dblk + loc); 392 if (dp->d_reclen == 0) { 393 msg("corrupted directory, inumber %d\n", ino); 394 break; 395 } 396 loc += dp->d_reclen; 397 if (dp->d_ino == 0) 398 continue; 399 if (dp->d_name[0] == '.') { 400 if (dp->d_name[1] == '\0') 401 continue; 402 if (dp->d_name[1] == '.' && dp->d_name[2] == '\0') 403 continue; 404 } 405 if (TSTINO(dp->d_ino, dumpinomap)) { 406 ret |= HASDUMPEDFILE; 407 if (ret & HASSUBDIRS) 408 break; 409 } 410 if (TSTINO(dp->d_ino, dumpdirmap)) { 411 ret |= HASSUBDIRS; 412 if (ret & HASDUMPEDFILE) 413 break; 414 } 415 } 416 return (ret); 417 } 418 419 /* 420 * Dump passes 3 and 4. 421 * 422 * Dump the contents of an inode to tape. 423 */ 424 void 425 dumpino(dp, ino) 426 struct dinode *dp; 427 ino_t ino; 428 { 429 int ind_level, cnt; 430 fsizeT size; 431 char buf[TP_BSIZE]; 432 433 if (newtape) { 434 newtape = 0; 435 dumpmap(dumpinomap, TS_BITS, ino); 436 } 437 CLRINO(ino, dumpinomap); 438 spcl.c_dinode = *dp; 439 spcl.c_type = TS_INODE; 440 spcl.c_count = 0; 441 switch (dp->di_mode & IFMT) { 442 443 case 0: 444 /* 445 * Freed inode. 446 */ 447 return; 448 449 case IFLNK: 450 /* 451 * Check for short symbolic link. 452 */ 453 if (dp->di_size > 0 && 454 #ifdef FS_44INODEFMT 455 (dp->di_size < sblock->fs_maxsymlinklen || 456 (sblock->fs_maxsymlinklen == 0 && dp->di_blocks == 0))) { 457 #else 458 dp->di_blocks == 0) { 459 #endif 460 spcl.c_addr[0] = 1; 461 spcl.c_count = 1; 462 writeheader(ino); 463 memmove(buf, dp->di_shortlink, (u_long)dp->di_size); 464 buf[dp->di_size] = '\0'; 465 writerec(buf, 0); 466 return; 467 } 468 /* fall through */ 469 470 case IFDIR: 471 case IFREG: 472 if (dp->di_size > 0) 473 break; 474 /* fall through */ 475 476 case IFIFO: 477 case IFSOCK: 478 case IFCHR: 479 case IFBLK: 480 writeheader(ino); 481 return; 482 483 default: 484 msg("Warning: undefined file type 0%o\n", dp->di_mode & IFMT); 485 return; 486 } 487 if (dp->di_size > NDADDR * sblock->fs_bsize) 488 cnt = NDADDR * sblock->fs_frag; 489 else 490 cnt = howmany(dp->di_size, sblock->fs_fsize); 491 blksout(&dp->di_db[0], cnt, ino); 492 if ((size = dp->di_size - NDADDR * sblock->fs_bsize) <= 0) 493 return; 494 for (ind_level = 0; ind_level < NIADDR; ind_level++) { 495 dmpindir(ino, dp->di_ib[ind_level], ind_level, &size); 496 if (size <= 0) 497 return; 498 } 499 } 500 501 /* 502 * Read indirect blocks, and pass the data blocks to be dumped. 503 */ 504 static void 505 dmpindir(ino, blk, ind_level, size) 506 ino_t ino; 507 daddr_t blk; 508 int ind_level; 509 fsizeT *size; 510 { 511 int i, cnt; 512 daddr_t idblk[MAXNINDIR]; 513 514 if (blk != 0) 515 bread(fsbtodb(sblock, blk), (char *)idblk, (int) sblock->fs_bsize); 516 else 517 memset(idblk, 0, (int)sblock->fs_bsize); 518 if (ind_level <= 0) { 519 if (*size < NINDIR(sblock) * sblock->fs_bsize) 520 cnt = howmany(*size, sblock->fs_fsize); 521 else 522 cnt = NINDIR(sblock) * sblock->fs_frag; 523 *size -= NINDIR(sblock) * sblock->fs_bsize; 524 blksout(&idblk[0], cnt, ino); 525 return; 526 } 527 ind_level--; 528 for (i = 0; i < NINDIR(sblock); i++) { 529 dmpindir(ino, idblk[i], ind_level, size); 530 if (*size <= 0) 531 return; 532 } 533 } 534 535 /* 536 * Collect up the data into tape record sized buffers and output them. 537 */ 538 void 539 blksout(blkp, frags, ino) 540 daddr_t *blkp; 541 int frags; 542 ino_t ino; 543 { 544 daddr_t *bp; 545 int i, j, count, blks, tbperdb; 546 547 blks = howmany(frags * sblock->fs_fsize, TP_BSIZE); 548 tbperdb = sblock->fs_bsize >> tp_bshift; 549 for (i = 0; i < blks; i += TP_NINDIR) { 550 if (i + TP_NINDIR > blks) 551 count = blks; 552 else 553 count = i + TP_NINDIR; 554 for (j = i; j < count; j++) 555 if (blkp[j / tbperdb] != 0) 556 spcl.c_addr[j - i] = 1; 557 else 558 spcl.c_addr[j - i] = 0; 559 spcl.c_count = count - i; 560 writeheader(ino); 561 bp = &blkp[i / tbperdb]; 562 for (j = i; j < count; j += tbperdb, bp++) 563 if (*bp != 0) 564 if (j + tbperdb <= count) 565 dumpblock(*bp, (int)sblock->fs_bsize); 566 else 567 dumpblock(*bp, (count - j) * TP_BSIZE); 568 spcl.c_type = TS_ADDR; 569 } 570 } 571 572 /* 573 * Dump a map to the tape. 574 */ 575 void 576 dumpmap(map, type, ino) 577 char *map; 578 int type; 579 ino_t ino; 580 { 581 int i; 582 char *cp; 583 584 spcl.c_type = type; 585 spcl.c_count = howmany(mapsize * sizeof(char), TP_BSIZE); 586 writeheader(ino); 587 for (i = 0, cp = map; i < spcl.c_count; i++, cp += TP_BSIZE) 588 writerec(cp, 0); 589 } 590 591 /* 592 * Write a header record to the dump tape. 593 */ 594 void 595 writeheader(ino) 596 ino_t ino; 597 { 598 int32_t sum, cnt, *lp; 599 600 spcl.c_inumber = ino; 601 spcl.c_magic = NFS_MAGIC; 602 spcl.c_checksum = 0; 603 lp = (int32_t *)&spcl; 604 sum = 0; 605 cnt = sizeof(union u_spcl) / (4 * sizeof(int32_t)); 606 while (--cnt >= 0) { 607 sum += *lp++; 608 sum += *lp++; 609 sum += *lp++; 610 sum += *lp++; 611 } 612 spcl.c_checksum = CHECKSUM - sum; 613 writerec((char *)&spcl, 1); 614 } 615 616 struct dinode * 617 getino(inum) 618 ino_t inum; 619 { 620 static daddr_t minino, maxino; 621 static struct dinode inoblock[MAXINOPB]; 622 623 curino = inum; 624 if (inum >= minino && inum < maxino) 625 return (&inoblock[inum - minino]); 626 bread(fsbtodb(sblock, ino_to_fsba(sblock, inum)), (char *)inoblock, 627 (int)sblock->fs_bsize); 628 minino = inum - (inum % INOPB(sblock)); 629 maxino = minino + INOPB(sblock); 630 return (&inoblock[inum - minino]); 631 } 632 633 /* 634 * Read a chunk of data from the disk. 635 * Try to recover from hard errors by reading in sector sized pieces. 636 * Error recovery is attempted at most BREADEMAX times before seeking 637 * consent from the operator to continue. 638 */ 639 int breaderrors = 0; 640 #define BREADEMAX 32 641 642 void 643 bread(blkno, buf, size) 644 daddr_t blkno; 645 char *buf; 646 int size; 647 { 648 int cnt, i; 649 extern int errno; 650 651 loop: 652 if (lseek(diskfd, ((off_t)blkno << dev_bshift), 0) < 0) 653 msg("bread: lseek fails\n"); 654 if ((cnt = read(diskfd, buf, size)) == size) 655 return; 656 if (blkno + (size / dev_bsize) > fsbtodb(sblock, sblock->fs_size)) { 657 /* 658 * Trying to read the final fragment. 659 * 660 * NB - dump only works in TP_BSIZE blocks, hence 661 * rounds `dev_bsize' fragments up to TP_BSIZE pieces. 662 * It should be smarter about not actually trying to 663 * read more than it can get, but for the time being 664 * we punt and scale back the read only when it gets 665 * us into trouble. (mkm 9/25/83) 666 */ 667 size -= dev_bsize; 668 goto loop; 669 } 670 if (cnt == -1) 671 msg("read error from %s: %s: [block %d]: count=%d\n", 672 disk, strerror(errno), blkno, size); 673 else 674 msg("short read error from %s: [block %d]: count=%d, got=%d\n", 675 disk, blkno, size, cnt); 676 if (++breaderrors > BREADEMAX) { 677 msg("More than %d block read errors from %d\n", 678 BREADEMAX, disk); 679 broadcast("DUMP IS AILING!\n"); 680 msg("This is an unrecoverable error.\n"); 681 if (!query("Do you want to attempt to continue?")){ 682 dumpabort(0); 683 /*NOTREACHED*/ 684 } else 685 breaderrors = 0; 686 } 687 /* 688 * Zero buffer, then try to read each sector of buffer separately. 689 */ 690 memset(buf, 0, size); 691 for (i = 0; i < size; i += dev_bsize, buf += dev_bsize, blkno++) { 692 if (lseek(diskfd, ((off_t)blkno << dev_bshift), 0) < 0) 693 msg("bread: lseek2 fails!\n"); 694 if ((cnt = read(diskfd, buf, (int)dev_bsize)) == dev_bsize) 695 continue; 696 if (cnt == -1) { 697 msg("read error from %s: %s: [sector %d]: count=%d\n", 698 disk, strerror(errno), blkno, dev_bsize); 699 continue; 700 } 701 msg("short read error from %s: [sector %d]: count=%d, got=%d\n", 702 disk, blkno, dev_bsize, cnt); 703 } 704 } 705