1 /* $NetBSD: ar_subs.c,v 1.16 2002/01/31 19:27:53 tv Exp $ */ 2 3 /*- 4 * Copyright (c) 1992 Keith Muller. 5 * Copyright (c) 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * Keith Muller of the University of California, San Diego. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 */ 39 40 #include <sys/cdefs.h> 41 #if defined(__RCSID) && !defined(lint) 42 #if 0 43 static char sccsid[] = "@(#)ar_subs.c 8.2 (Berkeley) 4/18/94"; 44 #else 45 __RCSID("$NetBSD: ar_subs.c,v 1.16 2002/01/31 19:27:53 tv Exp $"); 46 #endif 47 #endif /* not lint */ 48 49 #include <sys/types.h> 50 #include <sys/time.h> 51 #include <sys/stat.h> 52 #include <sys/param.h> 53 #include <signal.h> 54 #include <string.h> 55 #include <stdio.h> 56 #include <ctype.h> 57 #include <fcntl.h> 58 #include <errno.h> 59 #include <time.h> 60 #include <unistd.h> 61 #include <stdlib.h> 62 #include "pax.h" 63 #include "extern.h" 64 65 static void wr_archive(ARCHD *, int is_app); 66 static int get_arc(void); 67 static int next_head(ARCHD *); 68 extern sigset_t s_mask; 69 70 /* 71 * Routines which control the overall operation modes of pax as specified by 72 * the user: list, append, read ... 73 */ 74 75 static char hdbuf[BLKMULT]; /* space for archive header on read */ 76 u_long flcnt; /* number of files processed */ 77 ARCHD archd; 78 79 /* 80 * list() 81 * list the contents of an archive which match user supplied pattern(s) 82 * (no pattern matches all). 83 */ 84 85 void 86 list(void) 87 { 88 ARCHD *arcn; 89 int res; 90 time_t now; 91 92 arcn = &archd; 93 /* 94 * figure out archive type; pass any format specific options to the 95 * archive option processing routine; call the format init routine. We 96 * also save current time for ls_list() so we do not make a system 97 * call for each file we need to print. If verbose (vflag) start up 98 * the name and group caches. 99 */ 100 if ((get_arc() < 0) || ((*frmt->options)() < 0) || 101 ((*frmt->st_rd)() < 0)) 102 return; 103 104 now = time((time_t *)NULL); 105 106 /* 107 * step through the archive until the format says it is done 108 */ 109 while (next_head(arcn) == 0) { 110 if (arcn->name[0] == '/' && !check_Aflag()) { 111 memmove(arcn->name, arcn->name + 1, strlen(arcn->name)); 112 } 113 /* 114 * check for pattern, and user specified options match. 115 * When all patterns are matched we are done. 116 */ 117 if ((res = pat_match(arcn)) < 0) 118 break; 119 120 if ((res == 0) && (sel_chk(arcn) == 0)) { 121 /* 122 * pattern resulted in a selected file 123 */ 124 if (pat_sel(arcn) < 0) 125 break; 126 127 /* 128 * modify the name as requested by the user if name 129 * survives modification, do a listing of the file 130 */ 131 if ((res = mod_name(arcn)) < 0) 132 break; 133 if (res == 0) 134 ls_list(arcn, now); 135 } 136 137 /* 138 * skip to next archive format header using values calculated 139 * by the format header read routine 140 */ 141 if (rd_skip(arcn->skip + arcn->pad) == 1) 142 break; 143 } 144 145 /* 146 * all done, let format have a chance to cleanup, and make sure that 147 * the patterns supplied by the user were all matched 148 */ 149 (void)(*frmt->end_rd)(); 150 (void)sigprocmask(SIG_BLOCK, &s_mask, (sigset_t *)NULL); 151 ar_close(); 152 pat_chk(); 153 } 154 155 /* 156 * extract() 157 * extract the member(s) of an archive as specified by user supplied 158 * pattern(s) (no patterns extracts all members) 159 */ 160 161 void 162 extract(void) 163 { 164 ARCHD *arcn; 165 int res; 166 off_t cnt; 167 struct stat sb; 168 int fd; 169 170 arcn = &archd; 171 /* 172 * figure out archive type; pass any format specific options to the 173 * archive option processing routine; call the format init routine; 174 * start up the directory modification time and access mode database 175 */ 176 if ((get_arc() < 0) || ((*frmt->options)() < 0) || 177 ((*frmt->st_rd)() < 0) || (dir_start() < 0)) 178 return; 179 180 /* 181 * When we are doing interactive rename, we store the mapping of names 182 * so we can fix up hard links files later in the archive. 183 */ 184 if (iflag && (name_start() < 0)) 185 return; 186 187 /* 188 * step through each entry on the archive until the format read routine 189 * says it is done 190 */ 191 while (next_head(arcn) == 0) { 192 int gnu_longlink_hack = 193 (arcn->type == PAX_GLL || arcn->type == PAX_GLF); 194 195 if (arcn->name[0] == '/' && !check_Aflag()) { 196 memmove(arcn->name, arcn->name + 1, strlen(arcn->name)); 197 } 198 /* 199 * check for pattern, and user specified options match. When 200 * all the patterns are matched we are done 201 */ 202 if (!gnu_longlink_hack) { 203 if ((res = pat_match(arcn)) < 0) 204 break; 205 206 if ((res > 0) || (sel_chk(arcn) != 0)) { 207 /* 208 * file is not selected. skip past any file 209 * data and padding and go back for the next 210 * archive member 211 */ 212 (void)rd_skip(arcn->skip + arcn->pad); 213 continue; 214 } 215 } 216 217 /* 218 * with -u or -D only extract when the archive member is newer 219 * than the file with the same name in the file system (nos 220 * test of being the same type is required). 221 * NOTE: this test is done BEFORE name modifications as 222 * specified by pax. this operation can be confusing to the 223 * user who might expect the test to be done on an existing 224 * file AFTER the name mod. In honesty the pax spec is probably 225 * flawed in this respect. ignore this for GNU long links. 226 */ 227 if ((uflag || Dflag) && ((lstat(arcn->name, &sb) == 0)) && 228 !gnu_longlink_hack) { 229 if (uflag && Dflag) { 230 if ((arcn->sb.st_mtime <= sb.st_mtime) && 231 (arcn->sb.st_ctime <= sb.st_ctime)) { 232 (void)rd_skip(arcn->skip + arcn->pad); 233 continue; 234 } 235 } else if (Dflag) { 236 if (arcn->sb.st_ctime <= sb.st_ctime) { 237 (void)rd_skip(arcn->skip + arcn->pad); 238 continue; 239 } 240 } else if (arcn->sb.st_mtime <= sb.st_mtime) { 241 (void)rd_skip(arcn->skip + arcn->pad); 242 continue; 243 } 244 } 245 246 /* 247 * this archive member is now been selected. modify the name. 248 */ 249 if ((pat_sel(arcn) < 0) || ((res = mod_name(arcn)) < 0)) 250 break; 251 if (res > 0) { 252 /* 253 * a bad name mod, skip and purge name from link table 254 */ 255 purg_lnk(arcn); 256 (void)rd_skip(arcn->skip + arcn->pad); 257 continue; 258 } 259 260 /* 261 * Non standard -Y and -Z flag. When the existing file is 262 * same age or newer skip; ignore this for GNU long links. 263 */ 264 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0)) && 265 !gnu_longlink_hack) { 266 if (Yflag && Zflag) { 267 if ((arcn->sb.st_mtime <= sb.st_mtime) && 268 (arcn->sb.st_ctime <= sb.st_ctime)) { 269 (void)rd_skip(arcn->skip + arcn->pad); 270 continue; 271 } 272 } else if (Yflag) { 273 if (arcn->sb.st_ctime <= sb.st_ctime) { 274 (void)rd_skip(arcn->skip + arcn->pad); 275 continue; 276 } 277 } else if (arcn->sb.st_mtime <= sb.st_mtime) { 278 (void)rd_skip(arcn->skip + arcn->pad); 279 continue; 280 } 281 } 282 283 if (vflag) { 284 (void)fputs(arcn->name, stderr); 285 vfpart = 1; 286 } 287 288 /* 289 * all ok, extract this member based on type 290 */ 291 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG) && 292 !gnu_longlink_hack) { 293 /* 294 * process archive members that are not regular files. 295 * throw out padding and any data that might follow the 296 * header (as determined by the format). 297 */ 298 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) 299 res = lnk_creat(arcn); 300 else 301 res = node_creat(arcn); 302 303 (void)rd_skip(arcn->skip + arcn->pad); 304 if (res < 0) 305 purg_lnk(arcn); 306 307 if (vflag && vfpart) { 308 (void)putc('\n', stderr); 309 vfpart = 0; 310 } 311 continue; 312 } 313 /* 314 * we have a file with data here. If we can not create it, skip 315 * over the data and purge the name from hard link table 316 */ 317 if (gnu_longlink_hack) 318 fd = -1; /* this tells the pax internals to DTRT */ 319 else if ((fd = file_creat(arcn)) < 0) { 320 (void)rd_skip(arcn->skip + arcn->pad); 321 purg_lnk(arcn); 322 continue; 323 } 324 /* 325 * extract the file from the archive and skip over padding and 326 * any unprocessed data 327 */ 328 res = (*frmt->rd_data)(arcn, fd, &cnt); 329 if (!gnu_longlink_hack) 330 file_close(arcn, fd); 331 if (vflag && vfpart) { 332 (void)putc('\n', stderr); 333 vfpart = 0; 334 } 335 if (!res) 336 (void)rd_skip(cnt + arcn->pad); 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, (sigset_t *)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; 366 367 /* 368 * if this format supports hard link storage, start up the database 369 * that detects them. 370 */ 371 if (((hlk = frmt->hlk) == 1) && (lnk_start() < 0)) 372 return; 373 374 /* 375 * start up the file traversal code and format specific write 376 */ 377 if ((ftree_start() < 0) || ((*frmt->st_wr)() < 0)) 378 return; 379 wrf = frmt->wr; 380 381 /* 382 * When we are doing interactive rename, we store the mapping of names 383 * so we can fix up hard links files later in the archive. 384 */ 385 if (iflag && (name_start() < 0)) 386 return; 387 388 /* 389 * if this not append, and there are no files, we do no write a trailer 390 */ 391 wr_one = is_app; 392 393 /* 394 * while there are files to archive, process them one at at time 395 */ 396 while (next_file(arcn) == 0) { 397 /* 398 * check if this file meets user specified options match. 399 */ 400 if (sel_chk(arcn) != 0) 401 continue; 402 fd = -1; 403 if (uflag) { 404 /* 405 * only archive if this file is newer than a file with 406 * the same name that is already stored on the archive 407 */ 408 if ((res = chk_ftime(arcn)) < 0) 409 break; 410 if (res > 0) 411 continue; 412 } 413 414 /* 415 * this file is considered selected now. see if this is a hard 416 * link to a file already stored 417 */ 418 ftree_sel(arcn); 419 if (hlk && (chk_lnk(arcn) < 0)) 420 break; 421 422 if ((arcn->type == PAX_REG) || (arcn->type == PAX_HRG) || 423 (arcn->type == PAX_CTG)) { 424 /* 425 * we will have to read this file. by opening it now we 426 * can avoid writing a header to the archive for a file 427 * we were later unable to read (we also purge it from 428 * the link table). 429 */ 430 if ((fd = open(arcn->org_name, O_RDONLY, 0)) < 0) { 431 syswarn(1,errno, "Unable to open %s to read", 432 arcn->org_name); 433 purg_lnk(arcn); 434 continue; 435 } 436 } 437 438 if (arcn->name[0] == '/' && !check_Aflag()) { 439 memmove(arcn->name, arcn->name + 1, strlen(arcn->name)); 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 (void)fputs(arcn->name, stderr); 466 vfpart = 1; 467 } 468 ++flcnt; 469 470 /* 471 * looks safe to store the file, have the format specific 472 * routine write routine store the file header on the archive 473 */ 474 if ((res = (*wrf)(arcn)) < 0) { 475 rdfile_close(arcn, &fd); 476 break; 477 } 478 wr_one = 1; 479 if (res > 0) { 480 /* 481 * format write says no file data needs to be stored 482 * so we are done messing with this file 483 */ 484 if (vflag && vfpart) { 485 (void)putc('\n', stderr); 486 vfpart = 0; 487 } 488 rdfile_close(arcn, &fd); 489 continue; 490 } 491 492 /* 493 * Add file data to the archive, quit on write error. if we 494 * cannot write the entire file contents to the archive we 495 * must pad the archive to replace the missing file data 496 * (otherwise during an extract the file header for the file 497 * which FOLLOWS this one will not be where we expect it to 498 * be). 499 */ 500 res = (*frmt->wr_data)(arcn, fd, &cnt); 501 rdfile_close(arcn, &fd); 502 if (vflag && vfpart) { 503 (void)putc('\n', stderr); 504 vfpart = 0; 505 } 506 if (res < 0) 507 break; 508 509 /* 510 * pad as required, cnt is number of bytes not written 511 */ 512 if (((cnt > 0) && (wr_skip(cnt) < 0)) || 513 ((arcn->pad > 0) && (wr_skip(arcn->pad) < 0))) 514 break; 515 } 516 517 /* 518 * tell format to write trailer; pad to block boundary; reset directory 519 * mode/access times, and check if all patterns supplied by the user 520 * were matched. block off signals to avoid chance for multiple entry 521 * into the cleanup code 522 */ 523 if (wr_one) { 524 (*frmt->end_wr)(); 525 wr_fin(); 526 } 527 (void)sigprocmask(SIG_BLOCK, &s_mask, (sigset_t *)NULL); 528 ar_close(); 529 if (tflag) 530 proc_dir(); 531 ftree_chk(); 532 } 533 534 /* 535 * append() 536 * Add file to previously written archive. Archive format specified by the 537 * user must agree with archive. The archive is read first to collect 538 * modification times (if -u) and locate the archive trailer. The archive 539 * is positioned in front of the record with the trailer and wr_archive() 540 * is called to add the new members. 541 * PAX IMPLEMENTATION DETAIL NOTE: 542 * -u is implemented by adding the new members to the end of the archive. 543 * Care is taken so that these do not end up as links to the older 544 * version of the same file already stored in the archive. It is expected 545 * when extraction occurs these newer versions will over-write the older 546 * ones stored "earlier" in the archive (this may be a bad assumption as 547 * it depends on the implementation of the program doing the extraction). 548 * It is really difficult to splice in members without either re-writing 549 * the entire archive (from the point were the old version was), or having 550 * assistance of the format specification in terms of a special update 551 * header that invalidates a previous archive record. The posix spec left 552 * the method used to implement -u unspecified. This pax is able to 553 * over write existing files that it creates. 554 */ 555 556 void 557 append(void) 558 { 559 ARCHD *arcn; 560 int res; 561 FSUB *orgfrmt; 562 int udev; 563 off_t tlen; 564 565 arcn = &archd; 566 orgfrmt = frmt; 567 568 /* 569 * Do not allow an append operation if the actual archive is of a 570 * different format than the user specified format. 571 */ 572 if (get_arc() < 0) 573 return; 574 if ((orgfrmt != NULL) && (orgfrmt != frmt)) { 575 tty_warn(1, "Cannot mix current archive format %s with %s", 576 frmt->name, orgfrmt->name); 577 return; 578 } 579 580 /* 581 * pass the format any options and start up format 582 */ 583 if (((*frmt->options)() < 0) || ((*frmt->st_rd)() < 0)) 584 return; 585 586 /* 587 * if we only are adding members that are newer, we need to save the 588 * mod times for all files we see. 589 */ 590 if (uflag && (ftime_start() < 0)) 591 return; 592 593 /* 594 * some archive formats encode hard links by recording the device and 595 * file serial number (inode) but copy the file anyway (multiple times) 596 * to the archive. When we append, we run the risk that newly added 597 * files may have the same device and inode numbers as those recorded 598 * on the archive but during a previous run. If this happens, when the 599 * archive is extracted we get INCORRECT hard links. We avoid this by 600 * remapping the device numbers so that newly added files will never 601 * use the same device number as one found on the archive. remapping 602 * allows new members to safely have links among themselves. remapping 603 * also avoids problems with file inode (serial number) truncations 604 * when the inode number is larger than storage space in the archive 605 * header. See the remap routines for more details. 606 */ 607 if ((udev = frmt->udev) && (dev_start() < 0)) 608 return; 609 610 /* 611 * reading the archive may take a long time. If verbose tell the user 612 */ 613 if (vflag) { 614 (void)fprintf(stderr, 615 "%s: Reading archive to position at the end...", argv0); 616 vfpart = 1; 617 } 618 619 /* 620 * step through the archive until the format says it is done 621 */ 622 while (next_head(arcn) == 0) { 623 /* 624 * check if this file meets user specified options. 625 */ 626 if (sel_chk(arcn) != 0) { 627 if (rd_skip(arcn->skip + arcn->pad) == 1) 628 break; 629 continue; 630 } 631 632 if (uflag) { 633 /* 634 * see if this is the newest version of this file has 635 * already been seen, if so skip. 636 */ 637 if ((res = chk_ftime(arcn)) < 0) 638 break; 639 if (res > 0) { 640 if (rd_skip(arcn->skip + arcn->pad) == 1) 641 break; 642 continue; 643 } 644 } 645 646 /* 647 * Store this device number. Device numbers seen during the 648 * read phase of append will cause newly appended files with a 649 * device number seen in the old part of the archive to be 650 * remapped to an unused device number. 651 */ 652 if ((udev && (add_dev(arcn) < 0)) || 653 (rd_skip(arcn->skip + arcn->pad) == 1)) 654 break; 655 } 656 657 /* 658 * done, finish up read and get the number of bytes to back up so we 659 * can add new members. The format might have used the hard link table, 660 * purge it. 661 */ 662 tlen = (*frmt->end_rd)(); 663 lnk_end(); 664 665 /* 666 * try to position for write, if this fails quit. if any error occurs, 667 * we will refuse to write 668 */ 669 if (appnd_start(tlen) < 0) 670 return; 671 672 /* 673 * tell the user we are done reading. 674 */ 675 if (vflag && vfpart) { 676 (void)fputs("done.\n", stderr); 677 vfpart = 0; 678 } 679 680 /* 681 * go to the writing phase to add the new members 682 */ 683 wr_archive(arcn, 1); 684 } 685 686 /* 687 * archive() 688 * write a new archive 689 */ 690 691 void 692 archive(void) 693 { 694 695 /* 696 * if we only are adding members that are newer, we need to save the 697 * mod times for all files; set up for writing; pass the format any 698 * options write the archive 699 */ 700 if ((uflag && (ftime_start() < 0)) || (wr_start() < 0)) 701 return; 702 if ((*frmt->options)() < 0) 703 return; 704 705 wr_archive(&archd, 0); 706 } 707 708 /* 709 * copy() 710 * copy files from one part of the file system to another. this does not 711 * use any archive storage. The EFFECT OF THE COPY IS THE SAME as if an 712 * archive was written and then extracted in the destination directory 713 * (except the files are forced to be under the destination directory). 714 */ 715 716 void 717 copy(void) 718 { 719 ARCHD *arcn; 720 int res; 721 int fddest; 722 char *dest_pt; 723 int dlen; 724 int drem; 725 int fdsrc = -1; 726 struct stat sb; 727 char dirbuf[PAXPATHLEN+1]; 728 729 arcn = &archd; 730 /* 731 * set up the destination dir path and make sure it is a directory. We 732 * make sure we have a trailing / on the destination 733 */ 734 dlen = l_strncpy(dirbuf, dirptr, PAXPATHLEN); 735 dest_pt = dirbuf + dlen; 736 if (*(dest_pt-1) != '/') { 737 *dest_pt++ = '/'; 738 ++dlen; 739 } 740 *dest_pt = '\0'; 741 drem = PAXPATHLEN - dlen; 742 743 if (stat(dirptr, &sb) < 0) { 744 syswarn(1, errno, "Cannot access destination directory %s", 745 dirptr); 746 return; 747 } 748 if (!S_ISDIR(sb.st_mode)) { 749 tty_warn(1, "Destination is not a directory %s", dirptr); 750 return; 751 } 752 753 /* 754 * start up the hard link table; file traversal routines and the 755 * modification time and access mode database 756 */ 757 if ((lnk_start() < 0) || (ftree_start() < 0) || (dir_start() < 0)) 758 return; 759 760 /* 761 * When we are doing interactive rename, we store the mapping of names 762 * so we can fix up hard links files later in the archive. 763 */ 764 if (iflag && (name_start() < 0)) 765 return; 766 767 /* 768 * set up to cp file trees 769 */ 770 cp_start(); 771 772 /* 773 * while there are files to archive, process them 774 */ 775 while (next_file(arcn) == 0) { 776 fdsrc = -1; 777 778 /* 779 * check if this file meets user specified options 780 */ 781 if (sel_chk(arcn) != 0) 782 continue; 783 784 /* 785 * if there is already a file in the destination directory with 786 * the same name and it is newer, skip the one stored on the 787 * archive. 788 * NOTE: this test is done BEFORE name modifications as 789 * specified by pax. this can be confusing to the user who 790 * might expect the test to be done on an existing file AFTER 791 * the name mod. In honesty the pax spec is probably flawed in 792 * this respect 793 */ 794 if (uflag || Dflag) { 795 /* 796 * create the destination name 797 */ 798 if (*(arcn->name) == '/') 799 res = 1; 800 else 801 res = 0; 802 if ((arcn->nlen - res) > drem) { 803 tty_warn(1, "Destination pathname too long %s", 804 arcn->name); 805 continue; 806 } 807 (void)strncpy(dest_pt, arcn->name + res, drem); 808 dirbuf[PAXPATHLEN] = '\0'; 809 810 /* 811 * if existing file is same age or newer skip 812 */ 813 res = lstat(dirbuf, &sb); 814 *dest_pt = '\0'; 815 816 if (res == 0) { 817 if (uflag && Dflag) { 818 if ((arcn->sb.st_mtime<=sb.st_mtime) && 819 (arcn->sb.st_ctime<=sb.st_ctime)) 820 continue; 821 } else if (Dflag) { 822 if (arcn->sb.st_ctime <= sb.st_ctime) 823 continue; 824 } else if (arcn->sb.st_mtime <= sb.st_mtime) 825 continue; 826 } 827 } 828 829 /* 830 * this file is considered selected. See if this is a hard link 831 * to a previous file; modify the name as requested by the 832 * user; set the final destination. 833 */ 834 ftree_sel(arcn); 835 if ((chk_lnk(arcn) < 0) || ((res = mod_name(arcn)) < 0)) 836 break; 837 if ((res > 0) || (set_dest(arcn, dirbuf, dlen) < 0)) { 838 /* 839 * skip file, purge from link table 840 */ 841 purg_lnk(arcn); 842 continue; 843 } 844 845 /* 846 * Non standard -Y and -Z flag. When the exisiting file is 847 * same age or newer skip 848 */ 849 if ((Yflag || Zflag) && ((lstat(arcn->name, &sb) == 0))) { 850 if (Yflag && Zflag) { 851 if ((arcn->sb.st_mtime <= sb.st_mtime) && 852 (arcn->sb.st_ctime <= sb.st_ctime)) 853 continue; 854 } else if (Yflag) { 855 if (arcn->sb.st_ctime <= sb.st_ctime) 856 continue; 857 } else if (arcn->sb.st_mtime <= sb.st_mtime) 858 continue; 859 } 860 861 if (vflag) { 862 (void)fputs(arcn->name, stderr); 863 vfpart = 1; 864 } 865 ++flcnt; 866 867 /* 868 * try to create a hard link to the src file if requested 869 * but make sure we are not trying to overwrite ourselves. 870 */ 871 if (lflag) 872 res = cross_lnk(arcn); 873 else 874 res = chk_same(arcn); 875 if (res <= 0) { 876 if (vflag && vfpart) { 877 (void)putc('\n', stderr); 878 vfpart = 0; 879 } 880 continue; 881 } 882 883 /* 884 * have to create a new file 885 */ 886 if ((arcn->type != PAX_REG) && (arcn->type != PAX_CTG)) { 887 /* 888 * create a link or special file 889 */ 890 if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) 891 res = lnk_creat(arcn); 892 else 893 res = node_creat(arcn); 894 if (res < 0) 895 purg_lnk(arcn); 896 if (vflag && vfpart) { 897 (void)putc('\n', stderr); 898 vfpart = 0; 899 } 900 continue; 901 } 902 903 /* 904 * have to copy a regular file to the destination directory. 905 * first open source file and then create the destination file 906 */ 907 if ((fdsrc = open(arcn->org_name, O_RDONLY, 0)) < 0) { 908 syswarn(1, errno, "Unable to open %s to read", 909 arcn->org_name); 910 purg_lnk(arcn); 911 continue; 912 } 913 if ((fddest = file_creat(arcn)) < 0) { 914 rdfile_close(arcn, &fdsrc); 915 purg_lnk(arcn); 916 continue; 917 } 918 919 /* 920 * copy source file data to the destination file 921 */ 922 cp_file(arcn, fdsrc, fddest); 923 file_close(arcn, fddest); 924 rdfile_close(arcn, &fdsrc); 925 926 if (vflag && vfpart) { 927 (void)putc('\n', stderr); 928 vfpart = 0; 929 } 930 } 931 932 /* 933 * restore directory modes and times as required; make sure all 934 * patterns were selected block off signals to avoid chance for 935 * multiple entry into the cleanup code. 936 */ 937 (void)sigprocmask(SIG_BLOCK, &s_mask, (sigset_t *)NULL); 938 ar_close(); 939 proc_dir(); 940 ftree_chk(); 941 } 942 943 /* 944 * next_head() 945 * try to find a valid header in the archive. Uses format specific 946 * routines to extract the header and id the trailer. Trailers may be 947 * located within a valid header or in an invalid header (the location 948 * is format specific. The inhead field from the option table tells us 949 * where to look for the trailer). 950 * We keep reading (and resyncing) until we get enough contiguous data 951 * to check for a header. If we cannot find one, we shift by a byte 952 * add a new byte from the archive to the end of the buffer and try again. 953 * If we get a read error, we throw out what we have (as we must have 954 * contiguous data) and start over again. 955 * ASSUMED: headers fit within a BLKMULT header. 956 * Return: 957 * 0 if we got a header, -1 if we are unable to ever find another one 958 * (we reached the end of input, or we reached the limit on retries. see 959 * the specs for rd_wrbuf() for more details) 960 */ 961 962 static int 963 next_head(ARCHD *arcn) 964 { 965 int ret; 966 char *hdend; 967 int res; 968 int shftsz; 969 int hsz; 970 int in_resync = 0; /* set when we are in resync mode */ 971 int cnt = 0; /* counter for trailer function */ 972 973 /* 974 * set up initial conditions, we want a whole frmt->hsz block as we 975 * have no data yet. 976 */ 977 res = hsz = frmt->hsz; 978 hdend = hdbuf; 979 shftsz = hsz - 1; 980 for(;;) { 981 /* 982 * keep looping until we get a contiguous FULL buffer 983 * (frmt->hsz is the proper size) 984 */ 985 for (;;) { 986 if ((ret = rd_wrbuf(hdend, res)) == res) 987 break; 988 989 /* 990 * some kind of archive read problem, try to resync the 991 * storage device, better give the user the bad news. 992 */ 993 if ((ret == 0) || (rd_sync() < 0)) { 994 if (!is_oldgnutar) 995 tty_warn(1, 996 "Premature end of file on archive read"); 997 return(-1); 998 } 999 if (!in_resync) { 1000 if (act == APPND) { 1001 tty_warn(1, 1002 "Archive I/O error, cannot continue"); 1003 return(-1); 1004 } 1005 tty_warn(1, 1006 "Archive I/O error. Trying to recover."); 1007 ++in_resync; 1008 } 1009 1010 /* 1011 * oh well, throw it all out and start over 1012 */ 1013 res = hsz; 1014 hdend = hdbuf; 1015 } 1016 1017 /* 1018 * ok we have a contiguous buffer of the right size. Call the 1019 * format read routine. If this was not a valid header and this 1020 * format stores trailers outside of the header, call the 1021 * format specific trailer routine to check for a trailer. We 1022 * have to watch out that we do not mis-identify file data or 1023 * block padding as a header or trailer. Format specific 1024 * trailer functions must NOT check for the trailer while we 1025 * are running in resync mode. Some trailer functions may tell 1026 * us that this block cannot contain a valid header either, so 1027 * we then throw out the entire block and start over. 1028 */ 1029 if ((*frmt->rd)(arcn, hdbuf) == 0) 1030 break; 1031 1032 if (!frmt->inhead) { 1033 /* 1034 * this format has trailers outside of valid headers 1035 */ 1036 if ((ret = (*frmt->trail)(hdbuf,in_resync,&cnt)) == 0){ 1037 /* 1038 * valid trailer found, drain input as required 1039 */ 1040 ar_drain(); 1041 return(-1); 1042 } 1043 1044 if (ret == 1) { 1045 /* 1046 * we are in resync and we were told to throw 1047 * the whole block out because none of the 1048 * bytes in this block can be used to form a 1049 * valid header 1050 */ 1051 res = hsz; 1052 hdend = hdbuf; 1053 continue; 1054 } 1055 } 1056 1057 /* 1058 * Brute force section. 1059 * not a valid header. We may be able to find a header yet. So 1060 * we shift over by one byte, and set up to read one byte at a 1061 * time from the archive and place it at the end of the buffer. 1062 * We will keep moving byte at a time until we find a header or 1063 * get a read error and have to start over. 1064 */ 1065 if (!in_resync) { 1066 if (act == APPND) { 1067 tty_warn(1, 1068 "Unable to append, archive header flaw"); 1069 return(-1); 1070 } 1071 tty_warn(1, 1072 "Invalid header, starting valid header search."); 1073 ++in_resync; 1074 } 1075 memmove(hdbuf, hdbuf+1, shftsz); 1076 res = 1; 1077 hdend = hdbuf + shftsz; 1078 } 1079 1080 /* 1081 * ok got a valid header, check for trailer if format encodes it in the 1082 * the header. NOTE: the parameters are different than trailer routines 1083 * which encode trailers outside of the header! 1084 */ 1085 if (frmt->inhead && ((*frmt->subtrail)(arcn) == 0)) { 1086 /* 1087 * valid trailer found, drain input as required 1088 */ 1089 ar_drain(); 1090 return(-1); 1091 } 1092 1093 ++flcnt; 1094 return(0); 1095 } 1096 1097 /* 1098 * get_arc() 1099 * Figure out what format an archive is. Handles archive with flaws by 1100 * brute force searches for a legal header in any supported format. The 1101 * format id routines have to be careful to NOT mis-identify a format. 1102 * ASSUMED: headers fit within a BLKMULT header. 1103 * Return: 1104 * 0 if archive found -1 otherwise 1105 */ 1106 1107 static int 1108 get_arc(void) 1109 { 1110 int i; 1111 int hdsz = 0; 1112 int res; 1113 int minhd = BLKMULT; 1114 char *hdend; 1115 int notice = 0; 1116 1117 /* 1118 * find the smallest header size in all archive formats and then set up 1119 * to read the archive. 1120 */ 1121 for (i = 0; ford[i] >= 0; ++i) { 1122 if (fsub[ford[i]].hsz < minhd) 1123 minhd = fsub[ford[i]].hsz; 1124 } 1125 if (rd_start() < 0) 1126 return(-1); 1127 res = BLKMULT; 1128 hdsz = 0; 1129 hdend = hdbuf; 1130 for(;;) { 1131 for (;;) { 1132 /* 1133 * fill the buffer with at least the smallest header 1134 */ 1135 i = rd_wrbuf(hdend, res); 1136 if (i > 0) 1137 hdsz += i; 1138 if (hdsz >= minhd) 1139 break; 1140 1141 /* 1142 * if we cannot recover from a read error quit 1143 */ 1144 if ((i == 0) || (rd_sync() < 0)) 1145 goto out; 1146 1147 /* 1148 * when we get an error none of the data we already 1149 * have can be used to create a legal header (we just 1150 * got an error in the middle), so we throw it all out 1151 * and refill the buffer with fresh data. 1152 */ 1153 res = BLKMULT; 1154 hdsz = 0; 1155 hdend = hdbuf; 1156 if (!notice) { 1157 if (act == APPND) 1158 return(-1); 1159 tty_warn(1, 1160 "Cannot identify format. Searching..."); 1161 ++notice; 1162 } 1163 } 1164 1165 /* 1166 * we have at least the size of the smallest header in any 1167 * archive format. Look to see if we have a match. The array 1168 * ford[] is used to specify the header id order to reduce the 1169 * chance of incorrectly id'ing a valid header (some formats 1170 * may be subsets of each other and the order would then be 1171 * important). 1172 */ 1173 for (i = 0; ford[i] >= 0; ++i) { 1174 if ((*fsub[ford[i]].id)(hdbuf, hdsz) < 0) 1175 continue; 1176 frmt = &(fsub[ford[i]]); 1177 /* 1178 * yuck, to avoid slow special case code in the extract 1179 * routines, just push this header back as if it was 1180 * not seen. We have left extra space at start of the 1181 * buffer for this purpose. This is a bit ugly, but 1182 * adding all the special case code is far worse. 1183 */ 1184 pback(hdbuf, hdsz); 1185 return(0); 1186 } 1187 1188 /* 1189 * We have a flawed archive, no match. we start searching, but 1190 * we never allow additions to flawed archives 1191 */ 1192 if (!notice) { 1193 if (act == APPND) 1194 return(-1); 1195 tty_warn(1, "Cannot identify format. Searching..."); 1196 ++notice; 1197 } 1198 1199 /* 1200 * brute force search for a header that we can id. 1201 * we shift through byte at a time. this is slow, but we cannot 1202 * determine the nature of the flaw in the archive in a 1203 * portable manner 1204 */ 1205 if (--hdsz > 0) { 1206 memmove(hdbuf, hdbuf+1, hdsz); 1207 res = BLKMULT - hdsz; 1208 hdend = hdbuf + hdsz; 1209 } else { 1210 res = BLKMULT; 1211 hdend = hdbuf; 1212 hdsz = 0; 1213 } 1214 } 1215 1216 out: 1217 /* 1218 * we cannot find a header, bow, apologize and quit 1219 */ 1220 tty_warn(1, "Sorry, unable to determine archive format."); 1221 return(-1); 1222 } 1223