1/* $NetBSD: ar_subs.c,v 1.58 2023/05/28 21:42:40 lukem Exp $ */ 2 3/*- 4 * Copyright (c) 1992 Keith Muller. 5 * Copyright (c) 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * Keith Muller of the University of California, San Diego. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. 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#if HAVE_NBTOOL_CONFIG_H 37#include "nbtool_config.h" 38#endif 39 40#include <sys/cdefs.h> 41#if !defined(lint) 42#if 0 43static char sccsid[] = "@(#)ar_subs.c 8.2 (Berkeley) 4/18/94"; 44#else 45__RCSID("$NetBSD: ar_subs.c,v 1.58 2023/05/28 21:42:40 lukem Exp $"); 46#endif 47#endif /* not lint */ 48 49#include <sys/types.h> 50#include <sys/time.h> 51#include <sys/stat.h> 52#include <sys/param.h> 53#include <signal.h> 54#include <string.h> 55#include <stdio.h> 56#include <ctype.h> 57#include <fcntl.h> 58#include <errno.h> 59#include <time.h> 60#include <unistd.h> 61#include <stdlib.h> 62#include "pax.h" 63#include "pat_rep.h" 64#include "extern.h" 65 66static int path_check(ARCHD *, int); 67static int wr_archive(ARCHD *, int is_app); 68static int get_arc(void); 69static int next_head(ARCHD *); 70#if !HAVE_NBTOOL_CONFIG_H 71static int fdochroot(int); 72#endif 73extern sigset_t s_mask; 74 75/* 76 * Routines which control the overall operation modes of pax as specified by 77 * the user: list, append, read ... 78 */ 79 80static char hdbuf[BLKMULT]; /* space for archive header on read */ 81u_long flcnt; /* number of files processed */ 82ARCHD archd; 83 84static char cwdpath[MAXPATHLEN]; /* current working directory path */ 85static size_t cwdpathlen; /* current working directory path len */ 86 87int 88updatepath(void) 89{ 90 if (getcwd(cwdpath, sizeof(cwdpath)) == NULL) { 91 syswarn(1, errno, "Cannot get working directory"); 92 return -1; 93 } 94 cwdpathlen = strlen(cwdpath); 95 return 0; 96} 97 98int 99fdochdir(int fcwd) 100{ 101 if (fchdir(fcwd) == -1) { 102 syswarn(1, errno, "Cannot chdir to `.'"); 103 return -1; 104 } 105 return updatepath(); 106} 107 108int 109dochdir(const char *name) 110{ 111 if (chdir(name) == -1) 112 syswarn(1, errno, "Cannot chdir to `%s'", name); 113 return updatepath(); 114} 115 116#if !HAVE_NBTOOL_CONFIG_H 117static int 118fdochroot(int fcwd) 119{ 120 if (fchroot(fcwd) != 0) { 121 syswarn(1, errno, "Can't fchroot to \".\""); 122 return -1; 123 } 124 return updatepath(); 125} 126#endif 127 128/* 129 * mkdir(), but if we failed, check if someone else made it for us 130 * already and don't error out. 131 */ 132int 133domkdir(const char *fname, mode_t mode) 134{ 135 int error; 136 struct stat sb; 137 138 if ((error = mkdir(fname, mode)) != -1) 139 return error; 140 141 switch (errno) { 142 case EISDIR: 143 return 0; 144 case EEXIST: 145 case EACCES: 146 case ENOSYS: /* Grr Solaris */ 147 case EROFS: 148 error = errno; 149 if (stat(fname, &sb) != -1 && S_ISDIR(sb.st_mode)) 150 return 0; 151 errno = error; 152 /*FALLTHROUGH*/ 153 default: 154 return -1; 155 } 156} 157 158static int 159path_check(ARCHD *arcn, int level) 160{ 161 char buf[MAXPATHLEN]; 162 char *p; 163 164 if ((p = strrchr(arcn->name, '/')) == NULL) 165 return 0; 166 *p = '\0'; 167 168 if (realpath(arcn->name, buf) == NULL) { 169 int error; 170 error = path_check(arcn, level + 1); 171 *p = '/'; 172 if (error == 0) 173 return 0; 174 if (level == 0) 175 syswarn(1, 0, "Cannot resolve `%s'", arcn->name); 176 return -1; 177 } 178 if (strncmp(buf, cwdpath, cwdpathlen) != 0) { 179 *p = '/'; 180 syswarn(1, 0, "Attempt to write file `%s' that resolves into " 181 "`%s/%s' outside current working directory `%s' ignored", 182 arcn->name, buf, p + 1, cwdpath); 183 return -1; 184 } 185 *p = '/'; 186 return 0; 187} 188 189/* 190 * list() 191 * list the contents of an archive which match user supplied pattern(s) 192 * (if no pattern is supplied, list entire contents). 193 */ 194 195int 196list(void) 197{ 198 ARCHD *arcn; 199 int res; 200 time_t now; 201 202 arcn = &archd; 203 /* 204 * figure out archive type; pass any format specific options to the 205 * archive option processing routine; call the format init routine. We 206 * also save current time for ls_list() so we do not make a system 207 * call for each file we need to print. If verbose (vflag) start up 208 * the name and group caches. 209 */ 210 if ((get_arc() < 0) || ((*frmt->options)() < 0) || 211 ((*frmt->st_rd)() < 0)) 212 return 1; 213 214 now = time(NULL); 215 216 /* 217 * step through the archive until the format says it is done 218 */ 219 while (next_head(arcn) == 0) { 220 if (arcn->type == PAX_GLL || arcn->type == PAX_GLF) { 221 /* 222 * we need to read, to get the real filename 223 */ 224 off_t cnt; 225 if (!(*frmt->rd_data)(arcn, -arcn->type, &cnt)) 226 (void)rd_skip(cnt + arcn->pad); 227 continue; 228 } 229 230 /* 231 * check for pattern, and user specified options match. 232 * When all patterns are matched we are done. 233 */ 234 if ((res = pat_match(arcn)) < 0) 235 break; 236 237 if ((res == 0) && (sel_chk(arcn) == 0)) { 238 /* 239 * pattern resulted in a selected file 240 */ 241 if (pat_sel(arcn) < 0) 242 break; 243 244 /* 245 * modify the name as requested by the user if name 246 * survives modification, do a listing of the file 247 */ 248 if ((res = mod_name(arcn, RENM)) < 0) 249 break; 250 if (res == 0) { 251 if (arcn->name[0] == '/' && !check_Aflag()) { 252 memmove(arcn->name, arcn->name + 1, 253 strlen(arcn->name)); 254 } 255 ls_list(arcn, now, stdout); 256 } 257 /* 258 * if there's an error writing to stdout then we must 259 * stop now -- we're probably writing to a pipe that 260 * has been closed by the reader. 261 */ 262 if (ferror(stdout)) { 263 syswarn(1, errno, "Listing incomplete."); 264 break; 265 } 266 } 267 /* 268 * skip to next archive format header using values calculated 269 * by the format header read routine 270 */ 271 if (rd_skip(arcn->skip + arcn->pad) == 1) 272 break; 273 } 274 275 /* 276 * all done, let format have a chance to cleanup, and make sure that 277 * the patterns supplied by the user were all matched 278 */ 279 (void)(*frmt->end_rd)(); 280 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 281 ar_close(); 282 pat_chk(); 283 284 return 0; 285} 286 287/* 288 * extract() 289 * extract the member(s) of an archive as specified by user supplied 290 * pattern(s) (no patterns extracts all members) 291 */ 292 293int 294extract(void) 295{ 296 ARCHD *arcn; 297 int res; 298 off_t cnt; 299 struct stat sb; 300 int fd; 301 time_t now; 302 303 arcn = &archd; 304 /* 305 * figure out archive type; pass any format specific options to the 306 * archive option processing routine; call the format init routine; 307 * start up the directory modification time and access mode database 308 */ 309 if ((get_arc() < 0) || ((*frmt->options)() < 0) || 310 ((*frmt->st_rd)() < 0) || (dir_start() < 0)) 311 return 1; 312 313 now = time(NULL); 314#if !HAVE_NBTOOL_CONFIG_H 315 if (do_chroot) 316 (void)fdochroot(cwdfd); 317#endif 318 319 /* 320 * When we are doing interactive rename, we store the mapping of names 321 * so we can fix up hard links files later in the archive. 322 */ 323 if (iflag && (name_start() < 0)) 324 return 1; 325 326 /* 327 * step through each entry on the archive until the format read routine 328 * says it is done 329 */ 330 while (next_head(arcn) == 0) { 331 int write_to_hard_link = 0; 332 333 if (arcn->type == PAX_GLL || arcn->type == PAX_GLF) { 334 /* 335 * we need to read, to get the real filename 336 */ 337 if (!(*frmt->rd_data)(arcn, -arcn->type, &cnt)) 338 (void)rd_skip(cnt + arcn->pad); 339 continue; 340 } 341 342 /* 343 * check for pattern, and user specified options match. When 344 * all the patterns are matched we are done 345 */ 346 if ((res = pat_match(arcn)) < 0) 347 break; 348 349 if ((res > 0) || (sel_chk(arcn) != 0)) { 350 /* 351 * file is not selected. skip past any file 352 * data and padding and go back for the next 353 * archive member 354 */ 355 (void)rd_skip(arcn->skip + arcn->pad); 356 continue; 357 } 358 359 if (kflag && (lstat(arcn->name, &sb) == 0)) { 360 (void)rd_skip(arcn->skip + arcn->pad); 361 continue; 362 } 363 364 /* 365 * with -u or -D only extract when the archive member is newer 366 * than the file with the same name in the file system (no 367 * test of being the same type is required). 368 * NOTE: this test is done BEFORE name modifications as 369 * specified by pax. this operation can be confusing to the 370 * user who might expect the test to be done on an existing 371 * file AFTER the name mod. In honesty the pax spec is probably 372 * flawed in this respect. ignore this for GNU long links. 373 */ 374 if ((uflag || Dflag) && ((lstat(arcn->name, &sb) == 0))) { 375 if (uflag && Dflag) { 376 if ((arcn->sb.st_mtime <= sb.st_mtime) && 377 (arcn->sb.st_ctime <= sb.st_ctime)) { 378 (void)rd_skip(arcn->skip + arcn->pad); 379 continue; 380 } 381 } else if (Dflag) { 382 if (arcn->sb.st_ctime <= sb.st_ctime) { 383 (void)rd_skip(arcn->skip + arcn->pad); 384 continue; 385 } 386 } else if (arcn->sb.st_mtime <= sb.st_mtime) { 387 (void)rd_skip(arcn->skip + arcn->pad); 388 continue; 389 } 390 } 391 392 /* 393 * this archive member is now been selected. modify the name. 394 */ 395 if ((pat_sel(arcn) < 0) || ((res = mod_name(arcn, RENM)) < 0)) 396 break; 397 if (res > 0) { 398 /* 399 * a bad name mod, skip and purge name from link table 400 */ 401 purg_lnk(arcn); 402 (void)rd_skip(arcn->skip + arcn->pad); 403 continue; 404 } 405 406 if (arcn->name[0] == '/' && !check_Aflag()) { 407 memmove(arcn->name, arcn->name + 1, strlen(arcn->name)); 408 } 409 /* 410 * Non standard -Y and -Z flag. When the existing file is 411 * same age or newer skip; ignore this for GNU long links. 412 */ 413 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) { 414 if (Yflag && Zflag) { 415 if ((arcn->sb.st_mtime <= sb.st_mtime) && 416 (arcn->sb.st_ctime <= sb.st_ctime)) { 417 (void)rd_skip(arcn->skip + arcn->pad); 418 continue; 419 } 420 } else if (Yflag) { 421 if (arcn->sb.st_ctime <= sb.st_ctime) { 422 (void)rd_skip(arcn->skip + arcn->pad); 423 continue; 424 } 425 } else if (arcn->sb.st_mtime <= sb.st_mtime) { 426 (void)rd_skip(arcn->skip + arcn->pad); 427 continue; 428 } 429 } 430 431 if (vflag) { 432 if (vflag > 1) 433 ls_list(arcn, now, listf); 434 else { 435 (void)safe_print(arcn->name, listf); 436 vfpart = 1; 437 } 438 } 439 440 /* 441 * if required, chdir around. 442 */ 443 if ((arcn->pat != NULL) && (arcn->pat->chdname != NULL) && 444 !to_stdout) 445 dochdir(arcn->pat->chdname); 446 447 if (secure && path_check(arcn, 0) != 0) { 448 (void)rd_skip(arcn->skip + arcn->pad); 449 continue; 450 } 451 452 453 /* 454 * all ok, extract this member based on type 455 */ 456 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) { 457 /* 458 * process archive members that are not regular files. 459 * throw out padding and any data that might follow the 460 * header (as determined by the format). 461 */ 462 if ((arcn->type == PAX_HLK) || 463 (arcn->type == PAX_HRG)) 464 res = lnk_creat(arcn, &write_to_hard_link); 465 else 466 res = node_creat(arcn); 467 468 if (!write_to_hard_link) { 469 (void)rd_skip(arcn->skip + arcn->pad); 470 if (res < 0) 471 purg_lnk(arcn); 472 473 if (vflag && vfpart) { 474 (void)putc('\n', listf); 475 vfpart = 0; 476 } 477 continue; 478 } 479 } 480 if (to_stdout) 481 fd = STDOUT_FILENO; 482 else { 483 /* 484 * We have a file with data here. If we cannot create 485 * it, skip over the data and purge the name from hard 486 * link table. 487 */ 488 if ((fd = file_creat(arcn, write_to_hard_link)) < 0) { 489 (void)fflush(listf); 490 (void)rd_skip(arcn->skip + arcn->pad); 491 purg_lnk(arcn); 492 continue; 493 } 494 } 495 /* 496 * extract the file from the archive and skip over padding and 497 * any unprocessed data 498 */ 499 res = (*frmt->rd_data)(arcn, fd, &cnt); 500 if (!to_stdout) 501 file_close(arcn, fd); 502 if (vflag && vfpart) { 503 (void)putc('\n', listf); 504 vfpart = 0; 505 } 506 if (!res) 507 (void)rd_skip(cnt + arcn->pad); 508 509 /* 510 * if required, chdir around. 511 */ 512 if ((arcn->pat != NULL) && (arcn->pat->chdname != NULL)) 513 fdochdir(cwdfd); 514 } 515 516 /* 517 * all done, restore directory modes and times as required; make sure 518 * all patterns supplied by the user were matched; block off signals 519 * to avoid chance for multiple entry into the cleanup code. 520 */ 521 (void)(*frmt->end_rd)(); 522 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 523 ar_close(); 524 proc_dir(); 525 pat_chk(); 526 527 return 0; 528} 529 530/* 531 * wr_archive() 532 * Write an archive. used in both creating a new archive and appends on 533 * previously written archive. 534 */ 535 536static int 537wr_archive(ARCHD *arcn, int is_app) 538{ 539 int res; 540 int hlk; 541 int wr_one; 542 off_t cnt; 543 int (*wrf)(ARCHD *); 544 int fd = -1; 545 time_t now; 546 547 /* 548 * if this format supports hard link storage, start up the database 549 * that detects them. 550 */ 551 if (((hlk = frmt->hlk) == 1) && (lnk_start() < 0)) 552 return 1; 553 554 /* 555 * start up the file traversal code and format specific write 556 */ 557 if ((ftree_start() < 0) || ((*frmt->st_wr)() < 0)) 558 return 1; 559 wrf = frmt->wr; 560 561 now = time(NULL); 562 563 /* 564 * When we are doing interactive rename, we store the mapping of names 565 * so we can fix up hard links files later in the archive. 566 */ 567 if (iflag && (name_start() < 0)) 568 return 1; 569 570 /* 571 * if this is not append, and there are no files, we do no write a trailer 572 */ 573 wr_one = is_app; 574 575 /* 576 * while there are files to archive, process them one at at time 577 */ 578 while (next_file(arcn) == 0) { 579 /* 580 * check if this file meets user specified options match. 581 */ 582 if (sel_chk(arcn) != 0) 583 continue; 584 /* 585 * Here we handle the exclusion -X gnu style patterns which 586 * are implemented like a pattern list. We don't modify the 587 * name as this will be done below again, and we don't want 588 * to double modify it. 589 */ 590 if ((res = mod_name(arcn, 0)) < 0) 591 break; 592 if (res == 1) 593 continue; 594 fd = -1; 595 if (uflag) { 596 /* 597 * only archive if this file is newer than a file with 598 * the same name that is already stored on the archive 599 */ 600 if ((res = chk_ftime(arcn)) < 0) 601 break; 602 if (res > 0) 603 continue; 604 } 605 606 /* 607 * this file is considered selected now. see if this is a hard 608 * link to a file already stored 609 */ 610 ftree_sel(arcn); 611 if (hlk && (chk_lnk(arcn) < 0)) 612 break; 613 614 if ((arcn->type == PAX_REG) || (arcn->type == PAX_HRG) || 615 (arcn->type == PAX_CTG)) { 616 /* 617 * we will have to read this file. by opening it now we 618 * can avoid writing a header to the archive for a file 619 * we were later unable to read (we also purge it from 620 * the link table). 621 */ 622 if ((fd = open(arcn->org_name, O_RDONLY, 0)) < 0) { 623 syswarn(1, errno, "Unable to open %s to read", 624 arcn->org_name); 625 purg_lnk(arcn); 626 continue; 627 } 628 } 629 630 /* 631 * Now modify the name as requested by the user 632 */ 633 if ((res = mod_name(arcn, RENM)) < 0) { 634 /* 635 * name modification says to skip this file, close the 636 * file and purge link table entry 637 */ 638 rdfile_close(arcn, &fd); 639 purg_lnk(arcn); 640 break; 641 } 642 643 if (arcn->name[0] == '/' && !check_Aflag()) { 644 memmove(arcn->name, arcn->name + 1, strlen(arcn->name)); 645 } 646 647 if ((res > 0) || (docrc && (set_crc(arcn, fd) < 0))) { 648 /* 649 * unable to obtain the crc we need, close the file, 650 * purge link table entry 651 */ 652 rdfile_close(arcn, &fd); 653 purg_lnk(arcn); 654 continue; 655 } 656 657 if (vflag) { 658 if (vflag > 1) 659 ls_list(arcn, now, listf); 660 else { 661 (void)safe_print(arcn->name, listf); 662 vfpart = 1; 663 } 664 } 665 ++flcnt; 666 667 /* 668 * looks safe to store the file, have the format specific 669 * routine write routine store the file header on the archive 670 */ 671 if ((res = (*wrf)(arcn)) < 0) { 672 rdfile_close(arcn, &fd); 673 break; 674 } 675 wr_one = 1; 676 if (res > 0) { 677 /* 678 * format write says no file data needs to be stored 679 * so we are done messing with this file 680 */ 681 if (vflag && vfpart) { 682 (void)putc('\n', listf); 683 vfpart = 0; 684 } 685 rdfile_close(arcn, &fd); 686 continue; 687 } 688 689 /* 690 * Add file data to the archive, quit on write error. if we 691 * cannot write the entire file contents to the archive we 692 * must pad the archive to replace the missing file data 693 * (otherwise during an extract the file header for the file 694 * which FOLLOWS this one will not be where we expect it to 695 * be). 696 */ 697 res = (*frmt->wr_data)(arcn, fd, &cnt); 698 rdfile_close(arcn, &fd); 699 if (vflag && vfpart) { 700 (void)putc('\n', listf); 701 vfpart = 0; 702 } 703 if (res < 0) 704 break; 705 706 /* 707 * pad as required, cnt is number of bytes not written 708 */ 709 if (((cnt > 0) && (wr_skip(cnt) < 0)) || 710 ((arcn->pad > 0) && (wr_skip(arcn->pad) < 0))) 711 break; 712 } 713 714 /* 715 * tell format to write trailer; pad to block boundary; reset directory 716 * mode/access times, and check if all patterns supplied by the user 717 * were matched. block off signals to avoid chance for multiple entry 718 * into the cleanup code 719 */ 720 if (wr_one) { 721 (*frmt->end_wr)(); 722 wr_fin(); 723 } 724 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 725 ar_close(); 726 if (tflag) 727 proc_dir(); 728 ftree_chk(); 729 730 return 0; 731} 732 733/* 734 * append() 735 * Add file to previously written archive. Archive format specified by the 736 * user must agree with archive. The archive is read first to collect 737 * modification times (if -u) and locate the archive trailer. The archive 738 * is positioned in front of the record with the trailer and wr_archive() 739 * is called to add the new members. 740 * PAX IMPLEMENTATION DETAIL NOTE: 741 * -u is implemented by adding the new members to the end of the archive. 742 * Care is taken so that these do not end up as links to the older 743 * version of the same file already stored in the archive. It is expected 744 * when extraction occurs these newer versions will over-write the older 745 * ones stored "earlier" in the archive (this may be a bad assumption as 746 * it depends on the implementation of the program doing the extraction). 747 * It is really difficult to splice in members without either re-writing 748 * the entire archive (from the point were the old version was), or having 749 * assistance of the format specification in terms of a special update 750 * header that invalidates a previous archive record. The posix spec left 751 * the method used to implement -u unspecified. This pax is able to 752 * over write existing files that it creates. 753 */ 754 755int 756append(void) 757{ 758 ARCHD *arcn; 759 int res; 760 FSUB *orgfrmt; 761 int udev; 762 off_t tlen; 763 764 arcn = &archd; 765 orgfrmt = frmt; 766 767 /* 768 * Do not allow an append operation if the actual archive is of a 769 * different format than the user specified format. 770 */ 771 if (get_arc() < 0) 772 return 1; 773 if ((orgfrmt != NULL) && (orgfrmt != frmt)) { 774 tty_warn(1, "Cannot mix current archive format %s with %s", 775 frmt->name, orgfrmt->name); 776 return 1; 777 } 778 779 /* 780 * pass the format any options and start up format 781 */ 782 if (((*frmt->options)() < 0) || ((*frmt->st_rd)() < 0)) 783 return 1; 784 785 /* 786 * if we only are adding members that are newer, we need to save the 787 * mod times for all files we see. 788 */ 789 if (uflag && (ftime_start() < 0)) 790 return 1; 791 792 /* 793 * some archive formats encode hard links by recording the device and 794 * file serial number (inode) but copy the file anyway (multiple times) 795 * to the archive. When we append, we run the risk that newly added 796 * files may have the same device and inode numbers as those recorded 797 * on the archive but during a previous run. If this happens, when the 798 * archive is extracted we get INCORRECT hard links. We avoid this by 799 * remapping the device numbers so that newly added files will never 800 * use the same device number as one found on the archive. remapping 801 * allows new members to safely have links among themselves. remapping 802 * also avoids problems with file inode (serial number) truncations 803 * when the inode number is larger than storage space in the archive 804 * header. See the remap routines for more details. 805 */ 806 if ((udev = frmt->udev) && (dev_start() < 0)) 807 return 1; 808 809 /* 810 * reading the archive may take a long time. If verbose tell the user 811 */ 812 if (vflag || Vflag) { 813 (void)fprintf(listf, 814 "%s: Reading archive to position at the end...", argv0); 815 vfpart = 1; 816 } 817 818 /* 819 * step through the archive until the format says it is done 820 */ 821 while (next_head(arcn) == 0) { 822 /* 823 * check if this file meets user specified options. 824 */ 825 if (sel_chk(arcn) != 0) { 826 if (rd_skip(arcn->skip + arcn->pad) == 1) 827 break; 828 continue; 829 } 830 831 if (uflag) { 832 /* 833 * see if this is the newest version of this file has 834 * already been seen, if so skip. 835 */ 836 if ((res = chk_ftime(arcn)) < 0) 837 break; 838 if (res > 0) { 839 if (rd_skip(arcn->skip + arcn->pad) == 1) 840 break; 841 continue; 842 } 843 } 844 845 /* 846 * Store this device number. Device numbers seen during the 847 * read phase of append will cause newly appended files with a 848 * device number seen in the old part of the archive to be 849 * remapped to an unused device number. 850 */ 851 if ((udev && (add_dev(arcn) < 0)) || 852 (rd_skip(arcn->skip + arcn->pad) == 1)) 853 break; 854 } 855 856 /* 857 * done, finish up read and get the number of bytes to back up so we 858 * can add new members. The format might have used the hard link table, 859 * purge it. 860 */ 861 tlen = (*frmt->end_rd)(); 862 lnk_end(); 863 864 /* 865 * try to position for write, if this fails quit. if any error occurs, 866 * we will refuse to write 867 */ 868 if (appnd_start(tlen) < 0) 869 return 1; 870 871 /* 872 * tell the user we are done reading. 873 */ 874 if ((vflag || Vflag) && vfpart) { 875 (void)safe_print("done.\n", listf); 876 vfpart = 0; 877 } 878 879 /* 880 * go to the writing phase to add the new members 881 */ 882 res = wr_archive(arcn, 1); 883 if (res == 1) { 884 /* 885 * wr_archive failed in some way, but before any files were 886 * added. These are the only steps needed to cleanup (and 887 * not truncate the archive). 888 */ 889 wr_fin(); 890 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 891 ar_close(); 892 } 893 return res; 894} 895 896/* 897 * archive() 898 * write a new archive 899 */ 900 901int 902archive(void) 903{ 904 905 /* 906 * if we only are adding members that are newer, we need to save the 907 * mod times for all files; set up for writing; pass the format any 908 * options write the archive 909 */ 910 if ((uflag && (ftime_start() < 0)) || (wr_start() < 0)) 911 return 1; 912 if ((*frmt->options)() < 0) 913 return 1; 914 915 return wr_archive(&archd, 0); 916} 917 918/* 919 * copy() 920 * copy files from one part of the file system to another. this does not 921 * use any archive storage. The EFFECT OF THE COPY IS THE SAME as if an 922 * archive was written and then extracted in the destination directory 923 * (except the files are forced to be under the destination directory). 924 */ 925 926int 927copy(void) 928{ 929 ARCHD *arcn; 930 int res; 931 int fddest; 932 char *dest_pt; 933 size_t dlen; 934 size_t drem; 935 int fdsrc = -1; 936 struct stat sb; 937 char dirbuf[PAXPATHLEN+1]; 938 939 arcn = &archd; 940 /* 941 * set up the destination dir path and make sure it is a directory. We 942 * make sure we have a trailing / on the destination 943 */ 944 dlen = strlcpy(dirbuf, dirptr, sizeof(dirbuf)); 945 if (dlen >= sizeof(dirbuf) || 946 (dlen == sizeof(dirbuf) - 1 && dirbuf[dlen - 1] != '/')) { 947 tty_warn(1, "directory name is too long %s", dirptr); 948 return 1; 949 } 950 dest_pt = dirbuf + dlen; 951 if (*(dest_pt-1) != '/') { 952 *dest_pt++ = '/'; 953 ++dlen; 954 } 955 *dest_pt = '\0'; 956 drem = PAXPATHLEN - dlen; 957 958 if (stat(dirptr, &sb) < 0) { 959 syswarn(1, errno, "Cannot access destination directory %s", 960 dirptr); 961 return 1; 962 } 963 if (!S_ISDIR(sb.st_mode)) { 964 tty_warn(1, "Destination is not a directory %s", dirptr); 965 return 1; 966 } 967 968 /* 969 * start up the hard link table; file traversal routines and the 970 * modification time and access mode database 971 */ 972 if ((lnk_start() < 0) || (ftree_start() < 0) || (dir_start() < 0)) 973 return 1; 974 975 /* 976 * When we are doing interactive rename, we store the mapping of names 977 * so we can fix up hard links files later in the archive. 978 */ 979 if (iflag && (name_start() < 0)) 980 return 1; 981 982 /* 983 * set up to cp file trees 984 */ 985 cp_start(); 986 987 /* 988 * while there are files to archive, process them 989 */ 990 while (next_file(arcn) == 0) { 991 fdsrc = -1; 992 993 /* 994 * check if this file meets user specified options 995 */ 996 if (sel_chk(arcn) != 0) 997 continue; 998 999 /* 1000 * if there is already a file in the destination directory with 1001 * the same name and it is newer, skip the one stored on the 1002 * archive. 1003 * NOTE: this test is done BEFORE name modifications as 1004 * specified by pax. this can be confusing to the user who 1005 * might expect the test to be done on an existing file AFTER 1006 * the name mod. In honesty the pax spec is probably flawed in 1007 * this respect 1008 */ 1009 if (uflag || Dflag) { 1010 /* 1011 * create the destination name 1012 */ 1013 if (strlcpy(dest_pt, arcn->name + (*arcn->name == '/'), 1014 drem + 1) > drem) { 1015 tty_warn(1, "Destination pathname too long %s", 1016 arcn->name); 1017 continue; 1018 } 1019 1020 /* 1021 * if existing file is same age or newer skip 1022 */ 1023 res = lstat(dirbuf, &sb); 1024 *dest_pt = '\0'; 1025 1026 if (res == 0) { 1027 if (uflag && Dflag) { 1028 if ((arcn->sb.st_mtime<=sb.st_mtime) && 1029 (arcn->sb.st_ctime<=sb.st_ctime)) 1030 continue; 1031 } else if (Dflag) { 1032 if (arcn->sb.st_ctime <= sb.st_ctime) 1033 continue; 1034 } else if (arcn->sb.st_mtime <= sb.st_mtime) 1035 continue; 1036 } 1037 } 1038 1039 /* 1040 * this file is considered selected. See if this is a hard link 1041 * to a previous file; modify the name as requested by the 1042 * user; set the final destination. 1043 */ 1044 ftree_sel(arcn); 1045 if ((chk_lnk(arcn) < 0) || ((res = mod_name(arcn, RENM)) < 0)) 1046 break; 1047 if ((res > 0) || (set_dest(arcn, dirbuf, dlen) < 0)) { 1048 /* 1049 * skip file, purge from link table 1050 */ 1051 purg_lnk(arcn); 1052 continue; 1053 } 1054 1055 /* 1056 * Non standard -Y and -Z flag. When the existing file is 1057 * same age or newer skip 1058 */ 1059 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) { 1060 if (Yflag && Zflag) { 1061 if ((arcn->sb.st_mtime <= sb.st_mtime) && 1062 (arcn->sb.st_ctime <= sb.st_ctime)) 1063 continue; 1064 } else if (Yflag) { 1065 if (arcn->sb.st_ctime <= sb.st_ctime) 1066 continue; 1067 } else if (arcn->sb.st_mtime <= sb.st_mtime) 1068 continue; 1069 } 1070 1071 if (vflag) { 1072 (void)safe_print(arcn->name, listf); 1073 vfpart = 1; 1074 } 1075 ++flcnt; 1076 1077 /* 1078 * try to create a hard link to the src file if requested 1079 * but make sure we are not trying to overwrite ourselves. 1080 */ 1081 if (lflag) 1082 res = cross_lnk(arcn); 1083 else 1084 res = chk_same(arcn); 1085 if (res <= 0) { 1086 if (vflag && vfpart) { 1087 (void)putc('\n', listf); 1088 vfpart = 0; 1089 } 1090 continue; 1091 } 1092 1093 /* 1094 * have to create a new file 1095 */ 1096 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) { 1097 /* 1098 * create a link or special file 1099 */ 1100 if ((arcn->type == PAX_HLK) || 1101 (arcn->type == PAX_HRG)) { 1102 int payload; 1103 1104 res = lnk_creat(arcn, &payload); 1105 } else { 1106 res = node_creat(arcn); 1107 } 1108 if (res < 0) 1109 purg_lnk(arcn); 1110 if (vflag && vfpart) { 1111 (void)putc('\n', listf); 1112 vfpart = 0; 1113 } 1114 continue; 1115 } 1116 1117 /* 1118 * have to copy a regular file to the destination directory. 1119 * first open source file and then create the destination file 1120 */ 1121 if ((fdsrc = open(arcn->org_name, O_RDONLY, 0)) < 0) { 1122 syswarn(1, errno, "Unable to open %s to read", 1123 arcn->org_name); 1124 purg_lnk(arcn); 1125 continue; 1126 } 1127 if ((fddest = file_creat(arcn, 0)) < 0) { 1128 rdfile_close(arcn, &fdsrc); 1129 purg_lnk(arcn); 1130 continue; 1131 } 1132 1133 /* 1134 * copy source file data to the destination file. 1135 * if there was a failure, remove the temporary file 1136 * and leave any existing destination file unmodified. 1137 */ 1138 if (cp_file(arcn, fdsrc, fddest) < 0) 1139 file_cleanup(arcn, fddest); 1140 else 1141 file_close(arcn, fddest); 1142 rdfile_close(arcn, &fdsrc); 1143 1144 if (vflag && vfpart) { 1145 (void)putc('\n', listf); 1146 vfpart = 0; 1147 } 1148 } 1149 1150 /* 1151 * restore directory modes and times as required; make sure all 1152 * patterns were selected block off signals to avoid chance for 1153 * multiple entry into the cleanup code. 1154 */ 1155 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 1156 ar_close(); 1157 proc_dir(); 1158 ftree_chk(); 1159 1160 return 0; 1161} 1162 1163/* 1164 * next_head() 1165 * try to find a valid header in the archive. Uses format specific 1166 * routines to extract the header and id the trailer. Trailers may be 1167 * located within a valid header or in an invalid header (the location 1168 * is format specific. The inhead field from the option table tells us 1169 * where to look for the trailer). 1170 * We keep reading (and resyncing) until we get enough contiguous data 1171 * to check for a header. If we cannot find one, we shift by a byte 1172 * add a new byte from the archive to the end of the buffer and try again. 1173 * If we get a read error, we throw out what we have (as we must have 1174 * contiguous data) and start over again. 1175 * ASSUMED: headers fit within a BLKMULT header. 1176 * Return: 1177 * 0 if we got a header, -1 if we are unable to ever find another one 1178 * (we reached the end of input, or we reached the limit on retries. see 1179 * the specs for rd_wrbuf() for more details) 1180 */ 1181 1182static int 1183next_head(ARCHD *arcn) 1184{ 1185 int ret; 1186 char *hdend; 1187 int res; 1188 int shftsz; 1189 int hsz; 1190 int in_resync = 0; /* set when we are in resync mode */ 1191 int cnt = 0; /* counter for trailer function */ 1192 int first = 1; /* on 1st read, EOF isn't premature. */ 1193 1194 /* 1195 * set up initial conditions, we want a whole frmt->hsz block as we 1196 * have no data yet. 1197 */ 1198 res = hsz = frmt->hsz; 1199 hdend = hdbuf; 1200 shftsz = hsz - 1; 1201 for(;;) { 1202 /* 1203 * keep looping until we get a contiguous FULL buffer 1204 * (frmt->hsz is the proper size) 1205 */ 1206 for (;;) { 1207 if ((ret = rd_wrbuf(hdend, res)) == res) 1208 break; 1209 1210 /* 1211 * If we read 0 bytes (EOF) from an archive when we 1212 * expect to find a header, we have stepped upon 1213 * an archive without the customary block of zeroes 1214 * end marker. It's just stupid to error out on 1215 * them, so exit gracefully. 1216 */ 1217 if (first && ret == 0) 1218 return -1; 1219 first = 0; 1220 1221 /* 1222 * some kind of archive read problem, try to resync the 1223 * storage device, better give the user the bad news. 1224 */ 1225 if ((ret == 0) || (rd_sync() < 0)) { 1226 tty_warn(1, 1227 "Premature end of file on archive read"); 1228 return -1; 1229 } 1230 if (!in_resync) { 1231 if (act == APPND) { 1232 tty_warn(1, 1233 "Archive I/O error, cannot continue"); 1234 return -1; 1235 } 1236 tty_warn(1, 1237 "Archive I/O error. Trying to recover."); 1238 ++in_resync; 1239 } 1240 1241 /* 1242 * oh well, throw it all out and start over 1243 */ 1244 res = hsz; 1245 hdend = hdbuf; 1246 } 1247 1248 /* 1249 * ok we have a contiguous buffer of the right size. Call the 1250 * format read routine. If this was not a valid header and this 1251 * format stores trailers outside of the header, call the 1252 * format specific trailer routine to check for a trailer. We 1253 * have to watch out that we do not mis-identify file data or 1254 * block padding as a header or trailer. Format specific 1255 * trailer functions must NOT check for the trailer while we 1256 * are running in resync mode. Some trailer functions may tell 1257 * us that this block cannot contain a valid header either, so 1258 * we then throw out the entire block and start over. 1259 */ 1260 if ((*frmt->rd)(arcn, hdbuf) == 0) 1261 break; 1262 1263 if (!frmt->inhead) { 1264 /* 1265 * this format has trailers outside of valid headers 1266 */ 1267 if ((ret = (*frmt->trail)(hdbuf,in_resync,&cnt)) == 0){ 1268 /* 1269 * valid trailer found, drain input as required 1270 */ 1271 ar_drain(); 1272 return -1; 1273 } 1274 1275 if (ret == 1) { 1276 /* 1277 * we are in resync and we were told to throw 1278 * the whole block out because none of the 1279 * bytes in this block can be used to form a 1280 * valid header 1281 */ 1282 res = hsz; 1283 hdend = hdbuf; 1284 continue; 1285 } 1286 } 1287 1288 /* 1289 * Brute force section. 1290 * not a valid header. We may be able to find a header yet. So 1291 * we shift over by one byte, and set up to read one byte at a 1292 * time from the archive and place it at the end of the buffer. 1293 * We will keep moving byte at a time until we find a header or 1294 * get a read error and have to start over. 1295 */ 1296 if (!in_resync) { 1297 if (act == APPND) { 1298 tty_warn(1, 1299 "Unable to append, archive header flaw"); 1300 return -1; 1301 } 1302 tty_warn(1, 1303 "Invalid header, starting valid header search."); 1304 ++in_resync; 1305 } 1306 memmove(hdbuf, hdbuf+1, shftsz); 1307 res = 1; 1308 hdend = hdbuf + shftsz; 1309 } 1310 1311 /* 1312 * ok got a valid header, check for trailer if format encodes it in the 1313 * the header. NOTE: the parameters are different than trailer routines 1314 * which encode trailers outside of the header! 1315 */ 1316 if (frmt->inhead && ((*frmt->subtrail)(arcn) == 0)) { 1317 /* 1318 * valid trailer found, drain input as required 1319 */ 1320 ar_drain(); 1321 return -1; 1322 } 1323 1324 ++flcnt; 1325 return 0; 1326} 1327 1328/* 1329 * get_arc() 1330 * Figure out what format an archive is. Handles archive with flaws by 1331 * brute force searches for a legal header in any supported format. The 1332 * format id routines have to be careful to NOT mis-identify a format. 1333 * ASSUMED: headers fit within a BLKMULT header. 1334 * Return: 1335 * 0 if archive found -1 otherwise 1336 */ 1337 1338static int 1339get_arc(void) 1340{ 1341 int i; 1342 int hdsz = 0; 1343 int res; 1344 int minhd = BLKMULT; 1345 char *hdend; 1346 int notice = 0; 1347 1348 /* 1349 * find the smallest header size in all archive formats and then set up 1350 * to read the archive. 1351 */ 1352 for (i = 0; ford[i] >= 0; ++i) { 1353 if (fsub[ford[i]].hsz < minhd) 1354 minhd = fsub[ford[i]].hsz; 1355 } 1356 if (rd_start() < 0) 1357 return -1; 1358 res = BLKMULT; 1359 hdsz = 0; 1360 hdend = hdbuf; 1361 for(;;) { 1362 for (;;) { 1363 /* 1364 * fill the buffer with at least the smallest header 1365 */ 1366 i = rd_wrbuf(hdend, res); 1367 if (i > 0) 1368 hdsz += i; 1369 if (hdsz >= minhd) 1370 break; 1371 1372 /* 1373 * if we cannot recover from a read error quit 1374 */ 1375 if ((i == 0) || (rd_sync() < 0)) 1376 goto out; 1377 1378 /* 1379 * when we get an error none of the data we already 1380 * have can be used to create a legal header (we just 1381 * got an error in the middle), so we throw it all out 1382 * and refill the buffer with fresh data. 1383 */ 1384 res = BLKMULT; 1385 hdsz = 0; 1386 hdend = hdbuf; 1387 if (!notice) { 1388 if (act == APPND) 1389 return -1; 1390 tty_warn(1, 1391 "Cannot identify format. Searching..."); 1392 ++notice; 1393 } 1394 } 1395 1396 /* 1397 * we have at least the size of the smallest header in any 1398 * archive format. Look to see if we have a match. The array 1399 * ford[] is used to specify the header id order to reduce the 1400 * chance of incorrectly id'ing a valid header (some formats 1401 * may be subsets of each other and the order would then be 1402 * important). 1403 */ 1404 for (i = 0; ford[i] >= 0; ++i) { 1405 if ((*fsub[ford[i]].id)(hdbuf, hdsz) < 0) 1406 continue; 1407 frmt = &(fsub[ford[i]]); 1408 /* 1409 * yuck, to avoid slow special case code in the extract 1410 * routines, just push this header back as if it was 1411 * not seen. We have left extra space at start of the 1412 * buffer for this purpose. This is a bit ugly, but 1413 * adding all the special case code is far worse. 1414 */ 1415 pback(hdbuf, hdsz); 1416 return 0; 1417 } 1418 1419 /* 1420 * We have a flawed archive, no match. we start searching, but 1421 * we never allow additions to flawed archives 1422 */ 1423 if (!notice) { 1424 if (act == APPND) 1425 return -1; 1426 tty_warn(1, "Cannot identify format. Searching..."); 1427 ++notice; 1428 } 1429 1430 /* 1431 * brute force search for a header that we can id. 1432 * we shift through byte at a time. this is slow, but we cannot 1433 * determine the nature of the flaw in the archive in a 1434 * portable manner 1435 */ 1436 if (--hdsz > 0) { 1437 memmove(hdbuf, hdbuf+1, hdsz); 1438 res = BLKMULT - hdsz; 1439 hdend = hdbuf + hdsz; 1440 } else { 1441 res = BLKMULT; 1442 hdend = hdbuf; 1443 hdsz = 0; 1444 } 1445 } 1446 1447 out: 1448 /* 1449 * we cannot find a header, bow, apologize and quit 1450 */ 1451 tty_warn(1, "Sorry, unable to determine archive format."); 1452 return -1; 1453} 1454