1 /*- 2 * Copyright (c) 1992 Keith Muller. 3 * Copyright (c) 1992, 1993 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Keith Muller of the University of California, San Diego. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 3. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 #ifndef lint 35 #if 0 36 static char sccsid[] = "@(#)ar_subs.c 8.2 (Berkeley) 4/18/94"; 37 #endif 38 #endif /* not lint */ 39 #include <sys/cdefs.h> 40 __FBSDID("$FreeBSD$"); 41 42 #include <sys/types.h> 43 #include <sys/time.h> 44 #include <sys/stat.h> 45 #include <signal.h> 46 #include <string.h> 47 #include <stdio.h> 48 #include <fcntl.h> 49 #include <errno.h> 50 #include <unistd.h> 51 #include "pax.h" 52 #include "extern.h" 53 54 static void wr_archive(ARCHD *, int is_app); 55 static int get_arc(void); 56 static int next_head(ARCHD *); 57 58 /* 59 * Routines which control the overall operation modes of pax as specified by 60 * the user: list, append, read ... 61 */ 62 63 static char hdbuf[BLKMULT]; /* space for archive header on read */ 64 u_long flcnt; /* number of files processed */ 65 66 /* 67 * list() 68 * list the contents of an archive which match user supplied pattern(s) 69 * (no pattern matches all). 70 */ 71 72 void 73 list(void) 74 { 75 ARCHD *arcn; 76 int res; 77 ARCHD archd; 78 time_t now; 79 80 arcn = &archd; 81 /* 82 * figure out archive type; pass any format specific options to the 83 * archive option processing routine; call the format init routine. We 84 * also save current time for ls_list() so we do not make a system 85 * call for each file we need to print. If verbose (vflag) start up 86 * the name and group caches. 87 */ 88 if ((get_arc() < 0) || ((*frmt->options)() < 0) || 89 ((*frmt->st_rd)() < 0)) 90 return; 91 92 if (vflag && ((uidtb_start() < 0) || (gidtb_start() < 0))) 93 return; 94 95 now = time(NULL); 96 97 /* 98 * step through the archive until the format says it is done 99 */ 100 while (next_head(arcn) == 0) { 101 /* 102 * check for pattern, and user specified options match. 103 * When all patterns are matched we are done. 104 */ 105 if ((res = pat_match(arcn)) < 0) 106 break; 107 108 if ((res == 0) && (sel_chk(arcn) == 0)) { 109 /* 110 * pattern resulted in a selected file 111 */ 112 if (pat_sel(arcn) < 0) 113 break; 114 115 /* 116 * modify the name as requested by the user if name 117 * survives modification, do a listing of the file 118 */ 119 if ((res = mod_name(arcn)) < 0) 120 break; 121 if (res == 0) 122 ls_list(arcn, now, stdout); 123 } 124 125 /* 126 * skip to next archive format header using values calculated 127 * by the format header read routine 128 */ 129 if (rd_skip(arcn->skip + arcn->pad) == 1) 130 break; 131 } 132 133 /* 134 * all done, let format have a chance to cleanup, and make sure that 135 * the patterns supplied by the user were all matched 136 */ 137 (void)(*frmt->end_rd)(); 138 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 139 ar_close(); 140 pat_chk(); 141 } 142 143 /* 144 * extract() 145 * extract the member(s) of an archive as specified by user supplied 146 * pattern(s) (no patterns extracts all members) 147 */ 148 149 void 150 extract(void) 151 { 152 ARCHD *arcn; 153 int res; 154 off_t cnt; 155 ARCHD archd; 156 struct stat sb; 157 int fd; 158 time_t now; 159 160 arcn = &archd; 161 /* 162 * figure out archive type; pass any format specific options to the 163 * archive option processing routine; call the format init routine; 164 * start up the directory modification time and access mode database 165 */ 166 if ((get_arc() < 0) || ((*frmt->options)() < 0) || 167 ((*frmt->st_rd)() < 0) || (dir_start() < 0)) 168 return; 169 170 /* 171 * When we are doing interactive rename, we store the mapping of names 172 * so we can fix up hard links files later in the archive. 173 */ 174 if (iflag && (name_start() < 0)) 175 return; 176 177 now = time(NULL); 178 179 /* 180 * step through each entry on the archive until the format read routine 181 * says it is done 182 */ 183 while (next_head(arcn) == 0) { 184 185 /* 186 * check for pattern, and user specified options match. When 187 * all the patterns are matched we are done 188 */ 189 if ((res = pat_match(arcn)) < 0) 190 break; 191 192 if ((res > 0) || (sel_chk(arcn) != 0)) { 193 /* 194 * file is not selected. skip past any file data and 195 * padding and go back for the next archive member 196 */ 197 (void)rd_skip(arcn->skip + arcn->pad); 198 continue; 199 } 200 201 /* 202 * with -u or -D only extract when the archive member is newer 203 * than the file with the same name in the file system (nos 204 * test of being the same type is required). 205 * NOTE: this test is done BEFORE name modifications as 206 * specified by pax. this operation can be confusing to the 207 * user who might expect the test to be done on an existing 208 * file AFTER the name mod. In honesty the pax spec is probably 209 * flawed in this respect. 210 */ 211 if ((uflag || Dflag) && ((lstat(arcn->name, &sb) == 0))) { 212 if (uflag && Dflag) { 213 if ((arcn->sb.st_mtime <= sb.st_mtime) && 214 (arcn->sb.st_ctime <= sb.st_ctime)) { 215 (void)rd_skip(arcn->skip + arcn->pad); 216 continue; 217 } 218 } else if (Dflag) { 219 if (arcn->sb.st_ctime <= sb.st_ctime) { 220 (void)rd_skip(arcn->skip + arcn->pad); 221 continue; 222 } 223 } else if (arcn->sb.st_mtime <= sb.st_mtime) { 224 (void)rd_skip(arcn->skip + arcn->pad); 225 continue; 226 } 227 } 228 229 /* 230 * this archive member is now been selected. modify the name. 231 */ 232 if ((pat_sel(arcn) < 0) || ((res = mod_name(arcn)) < 0)) 233 break; 234 if (res > 0) { 235 /* 236 * a bad name mod, skip and purge name from link table 237 */ 238 purg_lnk(arcn); 239 (void)rd_skip(arcn->skip + arcn->pad); 240 continue; 241 } 242 243 /* 244 * Non standard -Y and -Z flag. When the existing file is 245 * same age or newer skip 246 */ 247 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) { 248 if (Yflag && Zflag) { 249 if ((arcn->sb.st_mtime <= sb.st_mtime) && 250 (arcn->sb.st_ctime <= sb.st_ctime)) { 251 (void)rd_skip(arcn->skip + arcn->pad); 252 continue; 253 } 254 } else if (Yflag) { 255 if (arcn->sb.st_ctime <= sb.st_ctime) { 256 (void)rd_skip(arcn->skip + arcn->pad); 257 continue; 258 } 259 } else if (arcn->sb.st_mtime <= sb.st_mtime) { 260 (void)rd_skip(arcn->skip + arcn->pad); 261 continue; 262 } 263 } 264 265 if (vflag) { 266 if (vflag > 1) 267 ls_list(arcn, now, listf); 268 else { 269 (void)fputs(arcn->name, listf); 270 vfpart = 1; 271 } 272 } 273 274 /* 275 * if required, chdir around. 276 */ 277 if ((arcn->pat != NULL) && (arcn->pat->chdname != NULL)) 278 if (chdir(arcn->pat->chdname) != 0) 279 syswarn(1, errno, "Cannot chdir to %s", 280 arcn->pat->chdname); 281 282 /* 283 * all ok, extract this member based on type 284 */ 285 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) { 286 /* 287 * process archive members that are not regular files. 288 * throw out padding and any data that might follow the 289 * header (as determined by the format). 290 */ 291 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) 292 res = lnk_creat(arcn); 293 else 294 res = node_creat(arcn); 295 296 (void)rd_skip(arcn->skip + arcn->pad); 297 if (res < 0) 298 purg_lnk(arcn); 299 300 if (vflag && vfpart) { 301 (void)putc('\n', listf); 302 vfpart = 0; 303 } 304 continue; 305 } 306 /* 307 * we have a file with data here. If we can not create it, skip 308 * over the data and purge the name from hard link table 309 */ 310 if ((fd = file_creat(arcn)) < 0) { 311 (void)rd_skip(arcn->skip + arcn->pad); 312 purg_lnk(arcn); 313 continue; 314 } 315 /* 316 * extract the file from the archive and skip over padding and 317 * any unprocessed data 318 */ 319 res = (*frmt->rd_data)(arcn, fd, &cnt); 320 file_close(arcn, fd); 321 if (vflag && vfpart) { 322 (void)putc('\n', listf); 323 vfpart = 0; 324 } 325 if (!res) 326 (void)rd_skip(cnt + arcn->pad); 327 328 /* 329 * if required, chdir around. 330 */ 331 if ((arcn->pat != NULL) && (arcn->pat->chdname != NULL)) 332 if (fchdir(cwdfd) != 0) 333 syswarn(1, errno, 334 "Can't fchdir to starting directory"); 335 } 336 337 /* 338 * all done, restore directory modes and times as required; make sure 339 * all patterns supplied by the user were matched; block off signals 340 * to avoid chance for multiple entry into the cleanup code. 341 */ 342 (void)(*frmt->end_rd)(); 343 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 344 ar_close(); 345 proc_dir(); 346 pat_chk(); 347 } 348 349 /* 350 * wr_archive() 351 * Write an archive. used in both creating a new archive and appends on 352 * previously written archive. 353 */ 354 355 static void 356 wr_archive(ARCHD *arcn, int is_app) 357 { 358 int res; 359 int hlk; 360 int wr_one; 361 off_t cnt; 362 int (*wrf)(ARCHD *); 363 int fd = -1; 364 time_t now; 365 366 /* 367 * if this format supports hard link storage, start up the database 368 * that detects them. 369 */ 370 if (((hlk = frmt->hlk) == 1) && (lnk_start() < 0)) 371 return; 372 373 /* 374 * start up the file traversal code and format specific write 375 */ 376 if ((ftree_start() < 0) || ((*frmt->st_wr)() < 0)) 377 return; 378 wrf = frmt->wr; 379 380 /* 381 * When we are doing interactive rename, we store the mapping of names 382 * so we can fix up hard links files later in the archive. 383 */ 384 if (iflag && (name_start() < 0)) 385 return; 386 387 /* 388 * if this not append, and there are no files, we do no write a trailer 389 */ 390 wr_one = is_app; 391 392 now = time(NULL); 393 394 /* 395 * while there are files to archive, process them one at at time 396 */ 397 while (next_file(arcn) == 0) { 398 /* 399 * check if this file meets user specified options match. 400 */ 401 if (sel_chk(arcn) != 0) { 402 ftree_notsel(); 403 continue; 404 } 405 fd = -1; 406 if (uflag) { 407 /* 408 * only archive if this file is newer than a file with 409 * the same name that is already stored on the archive 410 */ 411 if ((res = chk_ftime(arcn)) < 0) 412 break; 413 if (res > 0) 414 continue; 415 } 416 417 /* 418 * this file is considered selected now. see if this is a hard 419 * link to a file already stored 420 */ 421 ftree_sel(arcn); 422 if (hlk && (chk_lnk(arcn) < 0)) 423 break; 424 425 if ((arcn->type == PAX_REG) || (arcn->type == PAX_HRG) || 426 (arcn->type == PAX_CTG)) { 427 /* 428 * we will have to read this file. by opening it now we 429 * can avoid writing a header to the archive for a file 430 * we were later unable to read (we also purge it from 431 * the link table). 432 */ 433 if ((fd = open(arcn->org_name, O_RDONLY, 0)) < 0) { 434 syswarn(1,errno, "Unable to open %s to read", 435 arcn->org_name); 436 purg_lnk(arcn); 437 continue; 438 } 439 } 440 441 /* 442 * Now modify the name as requested by the user 443 */ 444 if ((res = mod_name(arcn)) < 0) { 445 /* 446 * name modification says to skip this file, close the 447 * file and purge link table entry 448 */ 449 rdfile_close(arcn, &fd); 450 purg_lnk(arcn); 451 break; 452 } 453 454 if ((res > 0) || (docrc && (set_crc(arcn, fd) < 0))) { 455 /* 456 * unable to obtain the crc we need, close the file, 457 * purge link table entry 458 */ 459 rdfile_close(arcn, &fd); 460 purg_lnk(arcn); 461 continue; 462 } 463 464 if (vflag) { 465 if (vflag > 1) 466 ls_list(arcn, now, listf); 467 else { 468 (void)fputs(arcn->name, listf); 469 vfpart = 1; 470 } 471 } 472 ++flcnt; 473 474 /* 475 * looks safe to store the file, have the format specific 476 * routine write routine store the file header on the archive 477 */ 478 if ((res = (*wrf)(arcn)) < 0) { 479 rdfile_close(arcn, &fd); 480 break; 481 } 482 wr_one = 1; 483 if (res > 0) { 484 /* 485 * format write says no file data needs to be stored 486 * so we are done messing with this file 487 */ 488 if (vflag && vfpart) { 489 (void)putc('\n', listf); 490 vfpart = 0; 491 } 492 rdfile_close(arcn, &fd); 493 continue; 494 } 495 496 /* 497 * Add file data to the archive, quit on write error. if we 498 * cannot write the entire file contents to the archive we 499 * must pad the archive to replace the missing file data 500 * (otherwise during an extract the file header for the file 501 * which FOLLOWS this one will not be where we expect it to 502 * be). 503 */ 504 res = (*frmt->wr_data)(arcn, fd, &cnt); 505 rdfile_close(arcn, &fd); 506 if (vflag && vfpart) { 507 (void)putc('\n', listf); 508 vfpart = 0; 509 } 510 if (res < 0) 511 break; 512 513 /* 514 * pad as required, cnt is number of bytes not written 515 */ 516 if (((cnt > 0) && (wr_skip(cnt) < 0)) || 517 ((arcn->pad > 0) && (wr_skip(arcn->pad) < 0))) 518 break; 519 } 520 521 /* 522 * tell format to write trailer; pad to block boundary; reset directory 523 * mode/access times, and check if all patterns supplied by the user 524 * were matched. block off signals to avoid chance for multiple entry 525 * into the cleanup code 526 */ 527 if (wr_one) { 528 (*frmt->end_wr)(); 529 wr_fin(); 530 } 531 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 532 ar_close(); 533 if (tflag) 534 proc_dir(); 535 ftree_chk(); 536 } 537 538 /* 539 * append() 540 * Add file to previously written archive. Archive format specified by the 541 * user must agree with archive. The archive is read first to collect 542 * modification times (if -u) and locate the archive trailer. The archive 543 * is positioned in front of the record with the trailer and wr_archive() 544 * is called to add the new members. 545 * PAX IMPLEMENTATION DETAIL NOTE: 546 * -u is implemented by adding the new members to the end of the archive. 547 * Care is taken so that these do not end up as links to the older 548 * version of the same file already stored in the archive. It is expected 549 * when extraction occurs these newer versions will over-write the older 550 * ones stored "earlier" in the archive (this may be a bad assumption as 551 * it depends on the implementation of the program doing the extraction). 552 * It is really difficult to splice in members without either re-writing 553 * the entire archive (from the point were the old version was), or having 554 * assistance of the format specification in terms of a special update 555 * header that invalidates a previous archive record. The POSIX spec left 556 * the method used to implement -u unspecified. This pax is able to 557 * over write existing files that it creates. 558 */ 559 560 void 561 append(void) 562 { 563 ARCHD *arcn; 564 int res; 565 ARCHD archd; 566 FSUB *orgfrmt; 567 int udev; 568 off_t tlen; 569 570 arcn = &archd; 571 orgfrmt = frmt; 572 573 /* 574 * Do not allow an append operation if the actual archive is of a 575 * different format than the user specified format. 576 */ 577 if (get_arc() < 0) 578 return; 579 if ((orgfrmt != NULL) && (orgfrmt != frmt)) { 580 paxwarn(1, "Cannot mix current archive format %s with %s", 581 frmt->name, orgfrmt->name); 582 return; 583 } 584 585 /* 586 * pass the format any options and start up format 587 */ 588 if (((*frmt->options)() < 0) || ((*frmt->st_rd)() < 0)) 589 return; 590 591 /* 592 * if we only are adding members that are newer, we need to save the 593 * mod times for all files we see. 594 */ 595 if (uflag && (ftime_start() < 0)) 596 return; 597 598 /* 599 * some archive formats encode hard links by recording the device and 600 * file serial number (inode) but copy the file anyway (multiple times) 601 * to the archive. When we append, we run the risk that newly added 602 * files may have the same device and inode numbers as those recorded 603 * on the archive but during a previous run. If this happens, when the 604 * archive is extracted we get INCORRECT hard links. We avoid this by 605 * remapping the device numbers so that newly added files will never 606 * use the same device number as one found on the archive. remapping 607 * allows new members to safely have links among themselves. remapping 608 * also avoids problems with file inode (serial number) truncations 609 * when the inode number is larger than storage space in the archive 610 * header. See the remap routines for more details. 611 */ 612 if ((udev = frmt->udev) && (dev_start() < 0)) 613 return; 614 615 /* 616 * reading the archive may take a long time. If verbose tell the user 617 */ 618 if (vflag) { 619 (void)fprintf(listf, 620 "%s: Reading archive to position at the end...", argv0); 621 vfpart = 1; 622 } 623 624 /* 625 * step through the archive until the format says it is done 626 */ 627 while (next_head(arcn) == 0) { 628 /* 629 * check if this file meets user specified options. 630 */ 631 if (sel_chk(arcn) != 0) { 632 if (rd_skip(arcn->skip + arcn->pad) == 1) 633 break; 634 continue; 635 } 636 637 if (uflag) { 638 /* 639 * see if this is the newest version of this file has 640 * already been seen, if so skip. 641 */ 642 if ((res = chk_ftime(arcn)) < 0) 643 break; 644 if (res > 0) { 645 if (rd_skip(arcn->skip + arcn->pad) == 1) 646 break; 647 continue; 648 } 649 } 650 651 /* 652 * Store this device number. Device numbers seen during the 653 * read phase of append will cause newly appended files with a 654 * device number seen in the old part of the archive to be 655 * remapped to an unused device number. 656 */ 657 if ((udev && (add_dev(arcn) < 0)) || 658 (rd_skip(arcn->skip + arcn->pad) == 1)) 659 break; 660 } 661 662 /* 663 * done, finish up read and get the number of bytes to back up so we 664 * can add new members. The format might have used the hard link table, 665 * purge it. 666 */ 667 tlen = (*frmt->end_rd)(); 668 lnk_end(); 669 670 /* 671 * try to position for write, if this fails quit. if any error occurs, 672 * we will refuse to write 673 */ 674 if (appnd_start(tlen) < 0) 675 return; 676 677 /* 678 * tell the user we are done reading. 679 */ 680 if (vflag && vfpart) { 681 (void)fputs("done.\n", listf); 682 vfpart = 0; 683 } 684 685 /* 686 * go to the writing phase to add the new members 687 */ 688 wr_archive(arcn, 1); 689 } 690 691 /* 692 * archive() 693 * write a new archive 694 */ 695 696 void 697 archive(void) 698 { 699 ARCHD archd; 700 701 /* 702 * if we only are adding members that are newer, we need to save the 703 * mod times for all files; set up for writing; pass the format any 704 * options write the archive 705 */ 706 if ((uflag && (ftime_start() < 0)) || (wr_start() < 0)) 707 return; 708 if ((*frmt->options)() < 0) 709 return; 710 711 wr_archive(&archd, 0); 712 } 713 714 /* 715 * copy() 716 * copy files from one part of the file system to another. this does not 717 * use any archive storage. The EFFECT OF THE COPY IS THE SAME as if an 718 * archive was written and then extracted in the destination directory 719 * (except the files are forced to be under the destination directory). 720 */ 721 722 void 723 copy(void) 724 { 725 ARCHD *arcn; 726 int res; 727 int fddest; 728 char *dest_pt; 729 int dlen; 730 int drem; 731 int fdsrc = -1; 732 struct stat sb; 733 ARCHD archd; 734 char dirbuf[PAXPATHLEN+1]; 735 736 arcn = &archd; 737 /* 738 * set up the destination dir path and make sure it is a directory. We 739 * make sure we have a trailing / on the destination 740 */ 741 dlen = l_strncpy(dirbuf, dirptr, sizeof(dirbuf) - 1); 742 dest_pt = dirbuf + dlen; 743 if (*(dest_pt-1) != '/') { 744 *dest_pt++ = '/'; 745 ++dlen; 746 } 747 *dest_pt = '\0'; 748 drem = PAXPATHLEN - dlen; 749 750 if (stat(dirptr, &sb) < 0) { 751 syswarn(1, errno, "Cannot access destination directory %s", 752 dirptr); 753 return; 754 } 755 if (!S_ISDIR(sb.st_mode)) { 756 paxwarn(1, "Destination is not a directory %s", dirptr); 757 return; 758 } 759 760 /* 761 * start up the hard link table; file traversal routines and the 762 * modification time and access mode database 763 */ 764 if ((lnk_start() < 0) || (ftree_start() < 0) || (dir_start() < 0)) 765 return; 766 767 /* 768 * When we are doing interactive rename, we store the mapping of names 769 * so we can fix up hard links files later in the archive. 770 */ 771 if (iflag && (name_start() < 0)) 772 return; 773 774 /* 775 * set up to cp file trees 776 */ 777 cp_start(); 778 779 /* 780 * while there are files to archive, process them 781 */ 782 while (next_file(arcn) == 0) { 783 fdsrc = -1; 784 785 /* 786 * check if this file meets user specified options 787 */ 788 if (sel_chk(arcn) != 0) { 789 ftree_notsel(); 790 continue; 791 } 792 793 /* 794 * if there is already a file in the destination directory with 795 * the same name and it is newer, skip the one stored on the 796 * archive. 797 * NOTE: this test is done BEFORE name modifications as 798 * specified by pax. this can be confusing to the user who 799 * might expect the test to be done on an existing file AFTER 800 * the name mod. In honesty the pax spec is probably flawed in 801 * this respect 802 */ 803 if (uflag || Dflag) { 804 /* 805 * create the destination name 806 */ 807 if (*(arcn->name) == '/') 808 res = 1; 809 else 810 res = 0; 811 if ((arcn->nlen - res) > drem) { 812 paxwarn(1, "Destination pathname too long %s", 813 arcn->name); 814 continue; 815 } 816 (void)strncpy(dest_pt, arcn->name + res, drem); 817 dirbuf[PAXPATHLEN] = '\0'; 818 819 /* 820 * if existing file is same age or newer skip 821 */ 822 res = lstat(dirbuf, &sb); 823 *dest_pt = '\0'; 824 825 if (res == 0) { 826 if (uflag && Dflag) { 827 if ((arcn->sb.st_mtime<=sb.st_mtime) && 828 (arcn->sb.st_ctime<=sb.st_ctime)) 829 continue; 830 } else if (Dflag) { 831 if (arcn->sb.st_ctime <= sb.st_ctime) 832 continue; 833 } else if (arcn->sb.st_mtime <= sb.st_mtime) 834 continue; 835 } 836 } 837 838 /* 839 * this file is considered selected. See if this is a hard link 840 * to a previous file; modify the name as requested by the 841 * user; set the final destination. 842 */ 843 ftree_sel(arcn); 844 if ((chk_lnk(arcn) < 0) || ((res = mod_name(arcn)) < 0)) 845 break; 846 if ((res > 0) || (set_dest(arcn, dirbuf, dlen) < 0)) { 847 /* 848 * skip file, purge from link table 849 */ 850 purg_lnk(arcn); 851 continue; 852 } 853 854 /* 855 * Non standard -Y and -Z flag. When the existing file is 856 * same age or newer skip 857 */ 858 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) { 859 if (Yflag && Zflag) { 860 if ((arcn->sb.st_mtime <= sb.st_mtime) && 861 (arcn->sb.st_ctime <= sb.st_ctime)) 862 continue; 863 } else if (Yflag) { 864 if (arcn->sb.st_ctime <= sb.st_ctime) 865 continue; 866 } else if (arcn->sb.st_mtime <= sb.st_mtime) 867 continue; 868 } 869 870 if (vflag) { 871 (void)fputs(arcn->name, listf); 872 vfpart = 1; 873 } 874 ++flcnt; 875 876 /* 877 * try to create a hard link to the src file if requested 878 * but make sure we are not trying to overwrite ourselves. 879 */ 880 if (lflag) 881 res = cross_lnk(arcn); 882 else 883 res = chk_same(arcn); 884 if (res <= 0) { 885 if (vflag && vfpart) { 886 (void)putc('\n', listf); 887 vfpart = 0; 888 } 889 continue; 890 } 891 892 /* 893 * have to create a new file 894 */ 895 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) { 896 /* 897 * create a link or special file 898 */ 899 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) 900 res = lnk_creat(arcn); 901 else 902 res = node_creat(arcn); 903 if (res < 0) 904 purg_lnk(arcn); 905 if (vflag && vfpart) { 906 (void)putc('\n', listf); 907 vfpart = 0; 908 } 909 continue; 910 } 911 912 /* 913 * have to copy a regular file to the destination directory. 914 * first open source file and then create the destination file 915 */ 916 if ((fdsrc = open(arcn->org_name, O_RDONLY, 0)) < 0) { 917 syswarn(1, errno, "Unable to open %s to read", 918 arcn->org_name); 919 purg_lnk(arcn); 920 continue; 921 } 922 if ((fddest = file_creat(arcn)) < 0) { 923 rdfile_close(arcn, &fdsrc); 924 purg_lnk(arcn); 925 continue; 926 } 927 928 /* 929 * copy source file data to the destination file 930 */ 931 cp_file(arcn, fdsrc, fddest); 932 file_close(arcn, fddest); 933 rdfile_close(arcn, &fdsrc); 934 935 if (vflag && vfpart) { 936 (void)putc('\n', listf); 937 vfpart = 0; 938 } 939 } 940 941 /* 942 * restore directory modes and times as required; make sure all 943 * patterns were selected block off signals to avoid chance for 944 * multiple entry into the cleanup code. 945 */ 946 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 947 ar_close(); 948 proc_dir(); 949 ftree_chk(); 950 } 951 952 /* 953 * next_head() 954 * try to find a valid header in the archive. Uses format specific 955 * routines to extract the header and id the trailer. Trailers may be 956 * located within a valid header or in an invalid header (the location 957 * is format specific. The inhead field from the option table tells us 958 * where to look for the trailer). 959 * We keep reading (and resyncing) until we get enough contiguous data 960 * to check for a header. If we cannot find one, we shift by a byte 961 * add a new byte from the archive to the end of the buffer and try again. 962 * If we get a read error, we throw out what we have (as we must have 963 * contiguous data) and start over again. 964 * ASSUMED: headers fit within a BLKMULT header. 965 * Return: 966 * 0 if we got a header, -1 if we are unable to ever find another one 967 * (we reached the end of input, or we reached the limit on retries. see 968 * the specs for rd_wrbuf() for more details) 969 */ 970 971 static int 972 next_head(ARCHD *arcn) 973 { 974 int ret; 975 char *hdend; 976 int res; 977 int shftsz; 978 int hsz; 979 int in_resync = 0; /* set when we are in resync mode */ 980 int cnt = 0; /* counter for trailer function */ 981 int first = 1; /* on 1st read, EOF isn't premature. */ 982 983 /* 984 * set up initial conditions, we want a whole frmt->hsz block as we 985 * have no data yet. 986 */ 987 res = hsz = frmt->hsz; 988 hdend = hdbuf; 989 shftsz = hsz - 1; 990 for(;;) { 991 /* 992 * keep looping until we get a contiguous FULL buffer 993 * (frmt->hsz is the proper size) 994 */ 995 for (;;) { 996 if ((ret = rd_wrbuf(hdend, res)) == res) 997 break; 998 999 /* 1000 * If we read 0 bytes (EOF) from an archive when we 1001 * expect to find a header, we have stepped upon 1002 * an archive without the customary block of zeroes 1003 * end marker. It's just stupid to error out on 1004 * them, so exit gracefully. 1005 */ 1006 if (first && ret == 0) 1007 return(-1); 1008 first = 0; 1009 1010 /* 1011 * some kind of archive read problem, try to resync the 1012 * storage device, better give the user the bad news. 1013 */ 1014 if ((ret == 0) || (rd_sync() < 0)) { 1015 paxwarn(1,"Premature end of file on archive read"); 1016 return(-1); 1017 } 1018 if (!in_resync) { 1019 if (act == APPND) { 1020 paxwarn(1, 1021 "Archive I/O error, cannot continue"); 1022 return(-1); 1023 } 1024 paxwarn(1,"Archive I/O error. Trying to recover."); 1025 ++in_resync; 1026 } 1027 1028 /* 1029 * oh well, throw it all out and start over 1030 */ 1031 res = hsz; 1032 hdend = hdbuf; 1033 } 1034 1035 /* 1036 * ok we have a contiguous buffer of the right size. Call the 1037 * format read routine. If this was not a valid header and this 1038 * format stores trailers outside of the header, call the 1039 * format specific trailer routine to check for a trailer. We 1040 * have to watch out that we do not mis-identify file data or 1041 * block padding as a header or trailer. Format specific 1042 * trailer functions must NOT check for the trailer while we 1043 * are running in resync mode. Some trailer functions may tell 1044 * us that this block cannot contain a valid header either, so 1045 * we then throw out the entire block and start over. 1046 */ 1047 if ((*frmt->rd)(arcn, hdbuf) == 0) 1048 break; 1049 1050 if (!frmt->inhead) { 1051 /* 1052 * this format has trailers outside of valid headers 1053 */ 1054 if ((ret = (*frmt->trail_tar)(hdbuf,in_resync,&cnt)) == 0){ 1055 /* 1056 * valid trailer found, drain input as required 1057 */ 1058 ar_drain(); 1059 return(-1); 1060 } 1061 1062 if (ret == 1) { 1063 /* 1064 * we are in resync and we were told to throw 1065 * the whole block out because none of the 1066 * bytes in this block can be used to form a 1067 * valid header 1068 */ 1069 res = hsz; 1070 hdend = hdbuf; 1071 continue; 1072 } 1073 } 1074 1075 /* 1076 * Brute force section. 1077 * not a valid header. We may be able to find a header yet. So 1078 * we shift over by one byte, and set up to read one byte at a 1079 * time from the archive and place it at the end of the buffer. 1080 * We will keep moving byte at a time until we find a header or 1081 * get a read error and have to start over. 1082 */ 1083 if (!in_resync) { 1084 if (act == APPND) { 1085 paxwarn(1,"Unable to append, archive header flaw"); 1086 return(-1); 1087 } 1088 paxwarn(1,"Invalid header, starting valid header search."); 1089 ++in_resync; 1090 } 1091 memmove(hdbuf, hdbuf+1, shftsz); 1092 res = 1; 1093 hdend = hdbuf + shftsz; 1094 } 1095 1096 /* 1097 * ok got a valid header, check for trailer if format encodes it in 1098 * the header. 1099 */ 1100 if (frmt->inhead && ((*frmt->trail_cpio)(arcn) == 0)) { 1101 /* 1102 * valid trailer found, drain input as required 1103 */ 1104 ar_drain(); 1105 return(-1); 1106 } 1107 1108 ++flcnt; 1109 return(0); 1110 } 1111 1112 /* 1113 * get_arc() 1114 * Figure out what format an archive is. Handles archive with flaws by 1115 * brute force searches for a legal header in any supported format. The 1116 * format id routines have to be careful to NOT mis-identify a format. 1117 * ASSUMED: headers fit within a BLKMULT header. 1118 * Return: 1119 * 0 if archive found -1 otherwise 1120 */ 1121 1122 static int 1123 get_arc(void) 1124 { 1125 int i; 1126 int hdsz = 0; 1127 int res; 1128 int minhd = BLKMULT; 1129 char *hdend; 1130 int notice = 0; 1131 1132 /* 1133 * find the smallest header size in all archive formats and then set up 1134 * to read the archive. 1135 */ 1136 for (i = 0; ford[i] >= 0; ++i) { 1137 if (fsub[ford[i]].hsz < minhd) 1138 minhd = fsub[ford[i]].hsz; 1139 } 1140 if (rd_start() < 0) 1141 return(-1); 1142 res = BLKMULT; 1143 hdsz = 0; 1144 hdend = hdbuf; 1145 for(;;) { 1146 for (;;) { 1147 /* 1148 * fill the buffer with at least the smallest header 1149 */ 1150 i = rd_wrbuf(hdend, res); 1151 if (i > 0) 1152 hdsz += i; 1153 if (hdsz >= minhd) 1154 break; 1155 1156 /* 1157 * if we cannot recover from a read error quit 1158 */ 1159 if ((i == 0) || (rd_sync() < 0)) 1160 goto out; 1161 1162 /* 1163 * when we get an error none of the data we already 1164 * have can be used to create a legal header (we just 1165 * got an error in the middle), so we throw it all out 1166 * and refill the buffer with fresh data. 1167 */ 1168 res = BLKMULT; 1169 hdsz = 0; 1170 hdend = hdbuf; 1171 if (!notice) { 1172 if (act == APPND) 1173 return(-1); 1174 paxwarn(1,"Cannot identify format. Searching..."); 1175 ++notice; 1176 } 1177 } 1178 1179 /* 1180 * we have at least the size of the smallest header in any 1181 * archive format. Look to see if we have a match. The array 1182 * ford[] is used to specify the header id order to reduce the 1183 * chance of incorrectly id'ing a valid header (some formats 1184 * may be subsets of each other and the order would then be 1185 * important). 1186 */ 1187 for (i = 0; ford[i] >= 0; ++i) { 1188 if ((*fsub[ford[i]].id)(hdbuf, hdsz) < 0) 1189 continue; 1190 frmt = &(fsub[ford[i]]); 1191 /* 1192 * yuck, to avoid slow special case code in the extract 1193 * routines, just push this header back as if it was 1194 * not seen. We have left extra space at start of the 1195 * buffer for this purpose. This is a bit ugly, but 1196 * adding all the special case code is far worse. 1197 */ 1198 pback(hdbuf, hdsz); 1199 return(0); 1200 } 1201 1202 /* 1203 * We have a flawed archive, no match. we start searching, but 1204 * we never allow additions to flawed archives 1205 */ 1206 if (!notice) { 1207 if (act == APPND) 1208 return(-1); 1209 paxwarn(1, "Cannot identify format. Searching..."); 1210 ++notice; 1211 } 1212 1213 /* 1214 * brute force search for a header that we can id. 1215 * we shift through byte at a time. this is slow, but we cannot 1216 * determine the nature of the flaw in the archive in a 1217 * portable manner 1218 */ 1219 if (--hdsz > 0) { 1220 memmove(hdbuf, hdbuf+1, hdsz); 1221 res = BLKMULT - hdsz; 1222 hdend = hdbuf + hdsz; 1223 } else { 1224 res = BLKMULT; 1225 hdend = hdbuf; 1226 hdsz = 0; 1227 } 1228 } 1229 1230 out: 1231 /* 1232 * we cannot find a header, bow, apologize and quit 1233 */ 1234 paxwarn(1, "Sorry, unable to determine archive format."); 1235 return(-1); 1236 } 1237