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