1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 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 #ifndef lint 37 #if 0 38 static char sccsid[] = "@(#)ar_subs.c 8.2 (Berkeley) 4/18/94"; 39 #endif 40 #endif /* not lint */ 41 #include <sys/cdefs.h> 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 (no 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 is not append, and there are no files, we do not write a 389 * trailer 390 */ 391 wr_one = is_app; 392 393 now = time(NULL); 394 395 /* 396 * while there are files to archive, process them one at at time 397 */ 398 while (next_file(arcn) == 0) { 399 /* 400 * check if this file meets user specified options match. 401 */ 402 if (sel_chk(arcn) != 0) { 403 ftree_notsel(); 404 continue; 405 } 406 fd = -1; 407 if (uflag) { 408 /* 409 * only archive if this file is newer than a file with 410 * the same name that is already stored on the archive 411 */ 412 if ((res = chk_ftime(arcn)) < 0) 413 break; 414 if (res > 0) 415 continue; 416 } 417 418 /* 419 * this file is considered selected now. see if this is a hard 420 * link to a file already stored 421 */ 422 ftree_sel(arcn); 423 if (hlk && (chk_lnk(arcn) < 0)) 424 break; 425 426 if ((arcn->type == PAX_REG) || (arcn->type == PAX_HRG) || 427 (arcn->type == PAX_CTG)) { 428 /* 429 * we will have to read this file. by opening it now we 430 * can avoid writing a header to the archive for a file 431 * we were later unable to read (we also purge it from 432 * the link table). 433 */ 434 if ((fd = open(arcn->org_name, O_RDONLY, 0)) < 0) { 435 syswarn(1,errno, "Unable to open %s to read", 436 arcn->org_name); 437 purg_lnk(arcn); 438 continue; 439 } 440 } 441 442 /* 443 * Now modify the name as requested by the user 444 */ 445 if ((res = mod_name(arcn)) < 0) { 446 /* 447 * name modification says to skip this file, close the 448 * file and purge link table entry 449 */ 450 rdfile_close(arcn, &fd); 451 purg_lnk(arcn); 452 break; 453 } 454 455 if ((res > 0) || (docrc && (set_crc(arcn, fd) < 0))) { 456 /* 457 * unable to obtain the crc we need, close the file, 458 * purge link table entry 459 */ 460 rdfile_close(arcn, &fd); 461 purg_lnk(arcn); 462 continue; 463 } 464 465 if (vflag) { 466 if (vflag > 1) 467 ls_list(arcn, now, listf); 468 else { 469 (void)fputs(arcn->name, listf); 470 vfpart = 1; 471 } 472 } 473 ++flcnt; 474 475 /* 476 * looks safe to store the file, have the format specific 477 * routine write routine store the file header on the archive 478 */ 479 if ((res = (*wrf)(arcn)) < 0) { 480 rdfile_close(arcn, &fd); 481 break; 482 } 483 wr_one = 1; 484 if (res > 0) { 485 /* 486 * format write says no file data needs to be stored 487 * so we are done messing with this file 488 */ 489 if (vflag && vfpart) { 490 (void)putc('\n', listf); 491 vfpart = 0; 492 } 493 rdfile_close(arcn, &fd); 494 continue; 495 } 496 497 /* 498 * Add file data to the archive, quit on write error. if we 499 * cannot write the entire file contents to the archive we 500 * must pad the archive to replace the missing file data 501 * (otherwise during an extract the file header for the file 502 * which FOLLOWS this one will not be where we expect it to 503 * be). 504 */ 505 res = (*frmt->wr_data)(arcn, fd, &cnt); 506 rdfile_close(arcn, &fd); 507 if (vflag && vfpart) { 508 (void)putc('\n', listf); 509 vfpart = 0; 510 } 511 if (res < 0) 512 break; 513 514 /* 515 * pad as required, cnt is number of bytes not written 516 */ 517 if (((cnt > 0) && (wr_skip(cnt) < 0)) || 518 ((arcn->pad > 0) && (wr_skip(arcn->pad) < 0))) 519 break; 520 } 521 522 /* 523 * tell format to write trailer; pad to block boundary; reset directory 524 * mode/access times, and check if all patterns supplied by the user 525 * were matched. block off signals to avoid chance for multiple entry 526 * into the cleanup code 527 */ 528 if (wr_one) { 529 (*frmt->end_wr)(); 530 wr_fin(); 531 } 532 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 533 ar_close(); 534 if (tflag) 535 proc_dir(); 536 ftree_chk(); 537 } 538 539 /* 540 * append() 541 * Add file to previously written archive. Archive format specified by the 542 * user must agree with archive. The archive is read first to collect 543 * modification times (if -u) and locate the archive trailer. The archive 544 * is positioned in front of the record with the trailer and wr_archive() 545 * is called to add the new members. 546 * PAX IMPLEMENTATION DETAIL NOTE: 547 * -u is implemented by adding the new members to the end of the archive. 548 * Care is taken so that these do not end up as links to the older 549 * version of the same file already stored in the archive. It is expected 550 * when extraction occurs these newer versions will over-write the older 551 * ones stored "earlier" in the archive (this may be a bad assumption as 552 * it depends on the implementation of the program doing the extraction). 553 * It is really difficult to splice in members without either re-writing 554 * the entire archive (from the point were the old version was), or having 555 * assistance of the format specification in terms of a special update 556 * header that invalidates a previous archive record. The POSIX spec left 557 * the method used to implement -u unspecified. This pax is able to 558 * over write existing files that it creates. 559 */ 560 561 void 562 append(void) 563 { 564 ARCHD *arcn; 565 int res; 566 ARCHD archd; 567 FSUB *orgfrmt; 568 int udev; 569 off_t tlen; 570 571 arcn = &archd; 572 orgfrmt = frmt; 573 574 /* 575 * Do not allow an append operation if the actual archive is of a 576 * different format than the user specified format. 577 */ 578 if (get_arc() < 0) 579 return; 580 if ((orgfrmt != NULL) && (orgfrmt != frmt)) { 581 paxwarn(1, "Cannot mix current archive format %s with %s", 582 frmt->name, orgfrmt->name); 583 return; 584 } 585 586 /* 587 * pass the format any options and start up format 588 */ 589 if (((*frmt->options)() < 0) || ((*frmt->st_rd)() < 0)) 590 return; 591 592 /* 593 * if we only are adding members that are newer, we need to save the 594 * mod times for all files we see. 595 */ 596 if (uflag && (ftime_start() < 0)) 597 return; 598 599 /* 600 * some archive formats encode hard links by recording the device and 601 * file serial number (inode) but copy the file anyway (multiple times) 602 * to the archive. When we append, we run the risk that newly added 603 * files may have the same device and inode numbers as those recorded 604 * on the archive but during a previous run. If this happens, when the 605 * archive is extracted we get INCORRECT hard links. We avoid this by 606 * remapping the device numbers so that newly added files will never 607 * use the same device number as one found on the archive. remapping 608 * allows new members to safely have links among themselves. remapping 609 * also avoids problems with file inode (serial number) truncations 610 * when the inode number is larger than storage space in the archive 611 * header. See the remap routines for more details. 612 */ 613 if ((udev = frmt->udev) && (dev_start() < 0)) 614 return; 615 616 /* 617 * reading the archive may take a long time. If verbose tell the user 618 */ 619 if (vflag) { 620 (void)fprintf(listf, 621 "%s: Reading archive to position at the end...", argv0); 622 vfpart = 1; 623 } 624 625 /* 626 * step through the archive until the format says it is done 627 */ 628 while (next_head(arcn) == 0) { 629 /* 630 * check if this file meets user specified options. 631 */ 632 if (sel_chk(arcn) != 0) { 633 if (rd_skip(arcn->skip + arcn->pad) == 1) 634 break; 635 continue; 636 } 637 638 if (uflag) { 639 /* 640 * see if this is the newest version of this file has 641 * already been seen, if so skip. 642 */ 643 if ((res = chk_ftime(arcn)) < 0) 644 break; 645 if (res > 0) { 646 if (rd_skip(arcn->skip + arcn->pad) == 1) 647 break; 648 continue; 649 } 650 } 651 652 /* 653 * Store this device number. Device numbers seen during the 654 * read phase of append will cause newly appended files with a 655 * device number seen in the old part of the archive to be 656 * remapped to an unused device number. 657 */ 658 if ((udev && (add_dev(arcn) < 0)) || 659 (rd_skip(arcn->skip + arcn->pad) == 1)) 660 break; 661 } 662 663 /* 664 * done, finish up read and get the number of bytes to back up so we 665 * can add new members. The format might have used the hard link table, 666 * purge it. 667 */ 668 tlen = (*frmt->end_rd)(); 669 lnk_end(); 670 671 /* 672 * try to position for write, if this fails quit. if any error occurs, 673 * we will refuse to write 674 */ 675 if (appnd_start(tlen) < 0) 676 return; 677 678 /* 679 * tell the user we are done reading. 680 */ 681 if (vflag && vfpart) { 682 (void)fputs("done.\n", listf); 683 vfpart = 0; 684 } 685 686 /* 687 * go to the writing phase to add the new members 688 */ 689 wr_archive(arcn, 1); 690 } 691 692 /* 693 * archive() 694 * write a new archive 695 */ 696 697 void 698 archive(void) 699 { 700 ARCHD archd; 701 702 /* 703 * if we only are adding members that are newer, we need to save the 704 * mod times for all files; set up for writing; pass the format any 705 * options write the archive 706 */ 707 if ((uflag && (ftime_start() < 0)) || (wr_start() < 0)) 708 return; 709 if ((*frmt->options)() < 0) 710 return; 711 712 wr_archive(&archd, 0); 713 } 714 715 /* 716 * copy() 717 * copy files from one part of the file system to another. this does not 718 * use any archive storage. The EFFECT OF THE COPY IS THE SAME as if an 719 * archive was written and then extracted in the destination directory 720 * (except the files are forced to be under the destination directory). 721 */ 722 723 void 724 copy(void) 725 { 726 ARCHD *arcn; 727 int res; 728 int fddest; 729 char *dest_pt; 730 int dlen; 731 int drem; 732 int fdsrc = -1; 733 struct stat sb; 734 ARCHD archd; 735 char dirbuf[PAXPATHLEN+1]; 736 737 arcn = &archd; 738 /* 739 * set up the destination dir path and make sure it is a directory. We 740 * make sure we have a trailing / on the destination 741 */ 742 dlen = l_strncpy(dirbuf, dirptr, sizeof(dirbuf) - 1); 743 dest_pt = dirbuf + dlen; 744 if (*(dest_pt-1) != '/') { 745 *dest_pt++ = '/'; 746 ++dlen; 747 } 748 *dest_pt = '\0'; 749 drem = PAXPATHLEN - dlen; 750 751 if (stat(dirptr, &sb) < 0) { 752 syswarn(1, errno, "Cannot access destination directory %s", 753 dirptr); 754 return; 755 } 756 if (!S_ISDIR(sb.st_mode)) { 757 paxwarn(1, "Destination is not a directory %s", dirptr); 758 return; 759 } 760 761 /* 762 * start up the hard link table; file traversal routines and the 763 * modification time and access mode database 764 */ 765 if ((lnk_start() < 0) || (ftree_start() < 0) || (dir_start() < 0)) 766 return; 767 768 /* 769 * When we are doing interactive rename, we store the mapping of names 770 * so we can fix up hard links files later in the archive. 771 */ 772 if (iflag && (name_start() < 0)) 773 return; 774 775 /* 776 * set up to cp file trees 777 */ 778 cp_start(); 779 780 /* 781 * while there are files to archive, process them 782 */ 783 while (next_file(arcn) == 0) { 784 fdsrc = -1; 785 786 /* 787 * check if this file meets user specified options 788 */ 789 if (sel_chk(arcn) != 0) { 790 ftree_notsel(); 791 continue; 792 } 793 794 /* 795 * if there is already a file in the destination directory with 796 * the same name and it is newer, skip the one stored on the 797 * archive. 798 * NOTE: this test is done BEFORE name modifications as 799 * specified by pax. this can be confusing to the user who 800 * might expect the test to be done on an existing file AFTER 801 * the name mod. In honesty the pax spec is probably flawed in 802 * this respect 803 */ 804 if (uflag || Dflag) { 805 /* 806 * create the destination name 807 */ 808 if (*(arcn->name) == '/') 809 res = 1; 810 else 811 res = 0; 812 if ((arcn->nlen - res) > drem) { 813 paxwarn(1, "Destination pathname too long %s", 814 arcn->name); 815 continue; 816 } 817 (void)strncpy(dest_pt, arcn->name + res, drem); 818 dirbuf[PAXPATHLEN] = '\0'; 819 820 /* 821 * if existing file is same age or newer skip 822 */ 823 res = lstat(dirbuf, &sb); 824 *dest_pt = '\0'; 825 826 if (res == 0) { 827 if (uflag && Dflag) { 828 if ((arcn->sb.st_mtime<=sb.st_mtime) && 829 (arcn->sb.st_ctime<=sb.st_ctime)) 830 continue; 831 } else if (Dflag) { 832 if (arcn->sb.st_ctime <= sb.st_ctime) 833 continue; 834 } else if (arcn->sb.st_mtime <= sb.st_mtime) 835 continue; 836 } 837 } 838 839 /* 840 * this file is considered selected. See if this is a hard link 841 * to a previous file; modify the name as requested by the 842 * user; set the final destination. 843 */ 844 ftree_sel(arcn); 845 if ((chk_lnk(arcn) < 0) || ((res = mod_name(arcn)) < 0)) 846 break; 847 if ((res > 0) || (set_dest(arcn, dirbuf, dlen) < 0)) { 848 /* 849 * skip file, purge from link table 850 */ 851 purg_lnk(arcn); 852 continue; 853 } 854 855 /* 856 * Non standard -Y and -Z flag. When the existing file is 857 * same age or newer skip 858 */ 859 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) { 860 if (Yflag && Zflag) { 861 if ((arcn->sb.st_mtime <= sb.st_mtime) && 862 (arcn->sb.st_ctime <= sb.st_ctime)) 863 continue; 864 } else if (Yflag) { 865 if (arcn->sb.st_ctime <= sb.st_ctime) 866 continue; 867 } else if (arcn->sb.st_mtime <= sb.st_mtime) 868 continue; 869 } 870 871 if (vflag) { 872 (void)fputs(arcn->name, listf); 873 vfpart = 1; 874 } 875 ++flcnt; 876 877 /* 878 * try to create a hard link to the src file if requested 879 * but make sure we are not trying to overwrite ourselves. 880 */ 881 if (lflag) 882 res = cross_lnk(arcn); 883 else 884 res = chk_same(arcn); 885 if (res <= 0) { 886 if (vflag && vfpart) { 887 (void)putc('\n', listf); 888 vfpart = 0; 889 } 890 continue; 891 } 892 893 /* 894 * have to create a new file 895 */ 896 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) { 897 /* 898 * create a link or special file 899 */ 900 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) 901 res = lnk_creat(arcn); 902 else 903 res = node_creat(arcn); 904 if (res < 0) 905 purg_lnk(arcn); 906 if (vflag && vfpart) { 907 (void)putc('\n', listf); 908 vfpart = 0; 909 } 910 continue; 911 } 912 913 /* 914 * have to copy a regular file to the destination directory. 915 * first open source file and then create the destination file 916 */ 917 if ((fdsrc = open(arcn->org_name, O_RDONLY, 0)) < 0) { 918 syswarn(1, errno, "Unable to open %s to read", 919 arcn->org_name); 920 purg_lnk(arcn); 921 continue; 922 } 923 if ((fddest = file_creat(arcn)) < 0) { 924 rdfile_close(arcn, &fdsrc); 925 purg_lnk(arcn); 926 continue; 927 } 928 929 /* 930 * copy source file data to the destination file 931 */ 932 cp_file(arcn, fdsrc, fddest); 933 file_close(arcn, fddest); 934 rdfile_close(arcn, &fdsrc); 935 936 if (vflag && vfpart) { 937 (void)putc('\n', listf); 938 vfpart = 0; 939 } 940 } 941 942 /* 943 * restore directory modes and times as required; make sure all 944 * patterns were selected block off signals to avoid chance for 945 * multiple entry into the cleanup code. 946 */ 947 (void)sigprocmask(SIG_BLOCK, &s_mask, NULL); 948 ar_close(); 949 proc_dir(); 950 ftree_chk(); 951 } 952 953 /* 954 * next_head() 955 * try to find a valid header in the archive. Uses format specific 956 * routines to extract the header and id the trailer. Trailers may be 957 * located within a valid header or in an invalid header (the location 958 * is format specific. The inhead field from the option table tells us 959 * where to look for the trailer). 960 * We keep reading (and resyncing) until we get enough contiguous data 961 * to check for a header. If we cannot find one, we shift by a byte 962 * add a new byte from the archive to the end of the buffer and try again. 963 * If we get a read error, we throw out what we have (as we must have 964 * contiguous data) and start over again. 965 * ASSUMED: headers fit within a BLKMULT header. 966 * Return: 967 * 0 if we got a header, -1 if we are unable to ever find another one 968 * (we reached the end of input, or we reached the limit on retries. see 969 * the specs for rd_wrbuf() for more details) 970 */ 971 972 static int 973 next_head(ARCHD *arcn) 974 { 975 int ret; 976 char *hdend; 977 int res; 978 int shftsz; 979 int hsz; 980 int in_resync = 0; /* set when we are in resync mode */ 981 int cnt = 0; /* counter for trailer function */ 982 int first = 1; /* on 1st read, EOF isn't premature. */ 983 984 /* 985 * set up initial conditions, we want a whole frmt->hsz block as we 986 * have no data yet. 987 */ 988 res = hsz = frmt->hsz; 989 hdend = hdbuf; 990 shftsz = hsz - 1; 991 for(;;) { 992 /* 993 * keep looping until we get a contiguous FULL buffer 994 * (frmt->hsz is the proper size) 995 */ 996 for (;;) { 997 if ((ret = rd_wrbuf(hdend, res)) == res) 998 break; 999 1000 /* 1001 * If we read 0 bytes (EOF) from an archive when we 1002 * expect to find a header, we have stepped upon 1003 * an archive without the customary block of zeroes 1004 * end marker. It's just stupid to error out on 1005 * them, so exit gracefully. 1006 */ 1007 if (first && ret == 0) 1008 return(-1); 1009 first = 0; 1010 1011 /* 1012 * some kind of archive read problem, try to resync the 1013 * storage device, better give the user the bad news. 1014 */ 1015 if ((ret == 0) || (rd_sync() < 0)) { 1016 paxwarn(1,"Premature end of file on archive read"); 1017 return(-1); 1018 } 1019 if (!in_resync) { 1020 if (act == APPND) { 1021 paxwarn(1, 1022 "Archive I/O error, cannot continue"); 1023 return(-1); 1024 } 1025 paxwarn(1,"Archive I/O error. Trying to recover."); 1026 ++in_resync; 1027 } 1028 1029 /* 1030 * oh well, throw it all out and start over 1031 */ 1032 res = hsz; 1033 hdend = hdbuf; 1034 } 1035 1036 /* 1037 * ok we have a contiguous buffer of the right size. Call the 1038 * format read routine. If this was not a valid header and this 1039 * format stores trailers outside of the header, call the 1040 * format specific trailer routine to check for a trailer. We 1041 * have to watch out that we do not mis-identify file data or 1042 * block padding as a header or trailer. Format specific 1043 * trailer functions must NOT check for the trailer while we 1044 * are running in resync mode. Some trailer functions may tell 1045 * us that this block cannot contain a valid header either, so 1046 * we then throw out the entire block and start over. 1047 */ 1048 if ((*frmt->rd)(arcn, hdbuf) == 0) 1049 break; 1050 1051 if (!frmt->inhead) { 1052 /* 1053 * this format has trailers outside of valid headers 1054 */ 1055 if ((ret = (*frmt->trail_tar)(hdbuf,in_resync,&cnt)) == 0){ 1056 /* 1057 * valid trailer found, drain input as required 1058 */ 1059 ar_drain(); 1060 return(-1); 1061 } 1062 1063 if (ret == 1) { 1064 /* 1065 * we are in resync and we were told to throw 1066 * the whole block out because none of the 1067 * bytes in this block can be used to form a 1068 * valid header 1069 */ 1070 res = hsz; 1071 hdend = hdbuf; 1072 continue; 1073 } 1074 } 1075 1076 /* 1077 * Brute force section. 1078 * not a valid header. We may be able to find a header yet. So 1079 * we shift over by one byte, and set up to read one byte at a 1080 * time from the archive and place it at the end of the buffer. 1081 * We will keep moving byte at a time until we find a header or 1082 * get a read error and have to start over. 1083 */ 1084 if (!in_resync) { 1085 if (act == APPND) { 1086 paxwarn(1,"Unable to append, archive header flaw"); 1087 return(-1); 1088 } 1089 paxwarn(1,"Invalid header, starting valid header search."); 1090 ++in_resync; 1091 } 1092 memmove(hdbuf, hdbuf+1, shftsz); 1093 res = 1; 1094 hdend = hdbuf + shftsz; 1095 } 1096 1097 /* 1098 * ok got a valid header, check for trailer if format encodes it in 1099 * the header. 1100 */ 1101 if (frmt->inhead && ((*frmt->trail_cpio)(arcn) == 0)) { 1102 /* 1103 * valid trailer found, drain input as required 1104 */ 1105 ar_drain(); 1106 return(-1); 1107 } 1108 1109 ++flcnt; 1110 return(0); 1111 } 1112 1113 /* 1114 * get_arc() 1115 * Figure out what format an archive is. Handles archive with flaws by 1116 * brute force searches for a legal header in any supported format. The 1117 * format id routines have to be careful to NOT mis-identify a format. 1118 * ASSUMED: headers fit within a BLKMULT header. 1119 * Return: 1120 * 0 if archive found -1 otherwise 1121 */ 1122 1123 static int 1124 get_arc(void) 1125 { 1126 int i; 1127 int hdsz = 0; 1128 int res; 1129 int minhd = BLKMULT; 1130 char *hdend; 1131 int notice = 0; 1132 1133 /* 1134 * find the smallest header size in all archive formats and then set up 1135 * to read the archive. 1136 */ 1137 for (i = 0; ford[i] >= 0; ++i) { 1138 if (fsub[ford[i]].hsz < minhd) 1139 minhd = fsub[ford[i]].hsz; 1140 } 1141 if (rd_start() < 0) 1142 return(-1); 1143 res = BLKMULT; 1144 hdsz = 0; 1145 hdend = hdbuf; 1146 for(;;) { 1147 for (;;) { 1148 /* 1149 * fill the buffer with at least the smallest header 1150 */ 1151 i = rd_wrbuf(hdend, res); 1152 if (i > 0) 1153 hdsz += i; 1154 if (hdsz >= minhd) 1155 break; 1156 1157 /* 1158 * if we cannot recover from a read error quit 1159 */ 1160 if ((i == 0) || (rd_sync() < 0)) 1161 goto out; 1162 1163 /* 1164 * when we get an error none of the data we already 1165 * have can be used to create a legal header (we just 1166 * got an error in the middle), so we throw it all out 1167 * and refill the buffer with fresh data. 1168 */ 1169 res = BLKMULT; 1170 hdsz = 0; 1171 hdend = hdbuf; 1172 if (!notice) { 1173 if (act == APPND) 1174 return(-1); 1175 paxwarn(1,"Cannot identify format. Searching..."); 1176 ++notice; 1177 } 1178 } 1179 1180 /* 1181 * we have at least the size of the smallest header in any 1182 * archive format. Look to see if we have a match. The array 1183 * ford[] is used to specify the header id order to reduce the 1184 * chance of incorrectly id'ing a valid header (some formats 1185 * may be subsets of each other and the order would then be 1186 * important). 1187 */ 1188 for (i = 0; ford[i] >= 0; ++i) { 1189 if ((*fsub[ford[i]].id)(hdbuf, hdsz) < 0) 1190 continue; 1191 frmt = &(fsub[ford[i]]); 1192 /* 1193 * yuck, to avoid slow special case code in the extract 1194 * routines, just push this header back as if it was 1195 * not seen. We have left extra space at start of the 1196 * buffer for this purpose. This is a bit ugly, but 1197 * adding all the special case code is far worse. 1198 */ 1199 pback(hdbuf, hdsz); 1200 return(0); 1201 } 1202 1203 /* 1204 * We have a flawed archive, no match. we start searching, but 1205 * we never allow additions to flawed archives 1206 */ 1207 if (!notice) { 1208 if (act == APPND) 1209 return(-1); 1210 paxwarn(1, "Cannot identify format. Searching..."); 1211 ++notice; 1212 } 1213 1214 /* 1215 * brute force search for a header that we can id. 1216 * we shift through byte at a time. this is slow, but we cannot 1217 * determine the nature of the flaw in the archive in a 1218 * portable manner 1219 */ 1220 if (--hdsz > 0) { 1221 memmove(hdbuf, hdbuf+1, hdsz); 1222 res = BLKMULT - hdsz; 1223 hdend = hdbuf + hdsz; 1224 } else { 1225 res = BLKMULT; 1226 hdend = hdbuf; 1227 hdsz = 0; 1228 } 1229 } 1230 1231 out: 1232 /* 1233 * we cannot find a header, bow, apologize and quit 1234 */ 1235 paxwarn(1, "Sorry, unable to determine archive format."); 1236 return(-1); 1237 } 1238