tar.c revision 1.15
1/* $NetBSD: tar.c,v 1.15 1999/08/18 17:46:28 kleink 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#ifndef lint 42#if 0 43static char sccsid[] = "@(#)tar.c 8.2 (Berkeley) 4/18/94"; 44#else 45__RCSID("$NetBSD: tar.c,v 1.15 1999/08/18 17:46:28 kleink 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 54#include <ctype.h> 55#include <errno.h> 56#include <grp.h> 57#include <pwd.h> 58#include <stdio.h> 59#include <stdlib.h> 60#include <string.h> 61#include <unistd.h> 62 63#include "pax.h" 64#include "extern.h" 65#include "tar.h" 66 67/* 68 * Routines for reading, writing and header identify of various versions of tar 69 */ 70 71static u_long tar_chksm __P((char *, int)); 72static char *name_split __P((char *, int)); 73static int ul_oct __P((u_long, char *, int, int)); 74#ifndef NET2_STAT 75static int uqd_oct __P((u_quad_t, char *, int, int)); 76#endif 77 78/* 79 * Routines common to all versions of tar 80 */ 81 82static int tar_nodir; /* do not write dirs under old tar */ 83int is_oldgnutar; /* skip end-ofvolume checks */ 84 85/* 86 * tar_endwr() 87 * add the tar trailer of two null blocks 88 * Return: 89 * 0 if ok, -1 otherwise (what wr_skip returns) 90 */ 91 92#if __STDC__ 93int 94tar_endwr(void) 95#else 96int 97tar_endwr() 98#endif 99{ 100 return(wr_skip((off_t)(NULLCNT*BLKMULT))); 101} 102 103/* 104 * tar_endrd() 105 * no cleanup needed here, just return size of trailer (for append) 106 * Return: 107 * size of trailer (2 * BLKMULT) 108 */ 109 110#if __STDC__ 111off_t 112tar_endrd(void) 113#else 114off_t 115tar_endrd() 116#endif 117{ 118 return((off_t)(NULLCNT*BLKMULT)); 119} 120 121/* 122 * tar_trail() 123 * Called to determine if a header block is a valid trailer. We are passed 124 * the block, the in_sync flag (which tells us we are in resync mode; 125 * looking for a valid header), and cnt (which starts at zero) which is 126 * used to count the number of empty blocks we have seen so far. 127 * Return: 128 * 0 if a valid trailer, -1 if not a valid trailer, or 1 if the block 129 * could never contain a header. 130 */ 131 132#if __STDC__ 133int 134tar_trail(char *buf, int in_resync, int *cnt) 135#else 136int 137tar_trail(buf, in_resync, cnt) 138 char *buf; 139 int in_resync; 140 int *cnt; 141#endif 142{ 143 int i; 144 145 /* 146 * look for all zero, trailer is two consecutive blocks of zero 147 */ 148 for (i = 0; i < BLKMULT; ++i) { 149 if (buf[i] != '\0') 150 break; 151 } 152 153 /* 154 * if not all zero it is not a trailer, but MIGHT be a header. 155 */ 156 if (i != BLKMULT) 157 return(-1); 158 159 /* 160 * When given a zero block, we must be careful! 161 * If we are not in resync mode, check for the trailer. Have to watch 162 * out that we do not mis-identify file data as the trailer, so we do 163 * NOT try to id a trailer during resync mode. During resync mode we 164 * might as well throw this block out since a valid header can NEVER be 165 * a block of all 0 (we must have a valid file name). 166 */ 167 if (!in_resync && (++*cnt >= NULLCNT)) 168 return(0); 169 return(1); 170} 171 172/* 173 * ul_oct() 174 * convert an unsigned long to an octal string. many oddball field 175 * termination characters are used by the various versions of tar in the 176 * different fields. term selects which kind to use. str is '0' padded 177 * at the front to len. we are unable to use only one format as many old 178 * tar readers are very cranky about this. 179 * Return: 180 * 0 if the number fit into the string, -1 otherwise 181 */ 182 183#if __STDC__ 184static int 185ul_oct(u_long val, char *str, int len, int term) 186#else 187static int 188ul_oct(val, str, len, term) 189 u_long val; 190 char *str; 191 int len; 192 int term; 193#endif 194{ 195 char *pt; 196 197 /* 198 * term selects the appropriate character(s) for the end of the string 199 */ 200 pt = str + len - 1; 201 switch(term) { 202 case 3: 203 *pt-- = '\0'; 204 break; 205 case 2: 206 *pt-- = ' '; 207 *pt-- = '\0'; 208 break; 209 case 1: 210 *pt-- = ' '; 211 break; 212 case 0: 213 default: 214 *pt-- = '\0'; 215 *pt-- = ' '; 216 break; 217 } 218 219 /* 220 * convert and blank pad if there is space 221 */ 222 while (pt >= str) { 223 *pt-- = '0' + (char)(val & 0x7); 224 if ((val = val >> 3) == (u_long)0) 225 break; 226 } 227 228 while (pt >= str) 229 *pt-- = '0'; 230 if (val != (u_long)0) 231 return(-1); 232 return(0); 233} 234 235#ifndef NET2_STAT 236/* 237 * uqd_oct() 238 * convert an u_quad_t to an octal string. one of many oddball field 239 * termination characters are used by the various versions of tar in the 240 * different fields. term selects which kind to use. str is '0' padded 241 * at the front to len. we are unable to use only one format as many old 242 * tar readers are very cranky about this. 243 * Return: 244 * 0 if the number fit into the string, -1 otherwise 245 */ 246 247#if __STDC__ 248static int 249uqd_oct(u_quad_t val, char *str, int len, int term) 250#else 251static int 252uqd_oct(val, str, len, term) 253 u_quad_t val; 254 char *str; 255 int len; 256 int term; 257#endif 258{ 259 char *pt; 260 261 /* 262 * term selects the appropriate character(s) for the end of the string 263 */ 264 pt = str + len - 1; 265 switch(term) { 266 case 3: 267 *pt-- = '\0'; 268 break; 269 case 2: 270 *pt-- = ' '; 271 *pt-- = '\0'; 272 break; 273 case 1: 274 *pt-- = ' '; 275 break; 276 case 0: 277 default: 278 *pt-- = '\0'; 279 *pt-- = ' '; 280 break; 281 } 282 283 /* 284 * convert and blank pad if there is space 285 */ 286 while (pt >= str) { 287 *pt-- = '0' + (char)(val & 0x7); 288 if ((val = val >> 3) == 0) 289 break; 290 } 291 292 while (pt >= str) 293 *pt-- = '0'; 294 if (val != (u_quad_t)0) 295 return(-1); 296 return(0); 297} 298#endif 299 300/* 301 * tar_chksm() 302 * calculate the checksum for a tar block counting the checksum field as 303 * all blanks (BLNKSUM is that value pre-calculated, the sume of 8 blanks). 304 * NOTE: we use len to short circuit summing 0's on write since we ALWAYS 305 * pad headers with 0. 306 * Return: 307 * unsigned long checksum 308 */ 309 310#if __STDC__ 311static u_long 312tar_chksm(char *blk, int len) 313#else 314static u_long 315tar_chksm(blk, len) 316 char *blk; 317 int len; 318#endif 319{ 320 char *stop; 321 char *pt; 322 u_long chksm = BLNKSUM; /* inital value is checksum field sum */ 323 324 /* 325 * add the part of the block before the checksum field 326 */ 327 pt = blk; 328 stop = blk + CHK_OFFSET; 329 while (pt < stop) 330 chksm += (u_long)(*pt++ & 0xff); 331 /* 332 * move past the checksum field and keep going, spec counts the 333 * checksum field as the sum of 8 blanks (which is pre-computed as 334 * BLNKSUM). 335 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding 336 * starts, no point in summing zero's) 337 */ 338 pt += CHK_LEN; 339 stop = blk + len; 340 while (pt < stop) 341 chksm += (u_long)(*pt++ & 0xff); 342 return(chksm); 343} 344 345/* 346 * Routines for old BSD style tar (also made portable to sysV tar) 347 */ 348 349/* 350 * tar_id() 351 * determine if a block given to us is a valid tar header (and not a USTAR 352 * header). We have to be on the lookout for those pesky blocks of all 353 * zero's. 354 * Return: 355 * 0 if a tar header, -1 otherwise 356 */ 357 358#if __STDC__ 359int 360tar_id(char *blk, int size) 361#else 362int 363tar_id(blk, size) 364 char *blk; 365 int size; 366#endif 367{ 368 HD_TAR *hd; 369 HD_USTAR *uhd; 370 371 if (size < BLKMULT) 372 return(-1); 373 hd = (HD_TAR *)blk; 374 uhd = (HD_USTAR *)blk; 375 376 /* 377 * check for block of zero's first, a simple and fast test, then make 378 * sure this is not a ustar header by looking for the ustar magic 379 * cookie. We should use TMAGLEN, but some USTAR archive programs are 380 * wrong and create archives missing the \0. Last we check the 381 * checksum. If this is ok we have to assume it is a valid header. 382 */ 383 if (hd->name[0] == '\0') 384 return(-1); 385 if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0) 386 return(-1); 387 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 388 return(-1); 389 return(0); 390} 391 392/* 393 * tar_opt() 394 * handle tar format specific -o options 395 * Return: 396 * 0 if ok -1 otherwise 397 */ 398 399#if __STDC__ 400int 401tar_opt(void) 402#else 403int 404tar_opt() 405#endif 406{ 407 OPLIST *opt; 408 409 while ((opt = opt_next()) != NULL) { 410 if (strcmp(opt->name, TAR_OPTION) || 411 strcmp(opt->value, TAR_NODIR)) { 412 tty_warn(1, 413 "Unknown tar format -o option/value pair %s=%s", 414 opt->name, opt->value); 415 tty_warn(1, 416 "%s=%s is the only supported tar format option", 417 TAR_OPTION, TAR_NODIR); 418 return(-1); 419 } 420 421 /* 422 * we only support one option, and only when writing 423 */ 424 if ((act != APPND) && (act != ARCHIVE)) { 425 tty_warn(1, "%s=%s is only supported when writing.", 426 opt->name, opt->value); 427 return(-1); 428 } 429 tar_nodir = 1; 430 } 431 return(0); 432} 433 434 435/* 436 * tar_rd() 437 * extract the values out of block already determined to be a tar header. 438 * store the values in the ARCHD parameter. 439 * Return: 440 * 0 441 */ 442 443#if __STDC__ 444int 445tar_rd(ARCHD *arcn, char *buf) 446#else 447int 448tar_rd(arcn, buf) 449 ARCHD *arcn; 450 char *buf; 451#endif 452{ 453 HD_TAR *hd; 454 char *pt; 455 456 /* 457 * we only get proper sized buffers passed to us 458 */ 459 if (tar_id(buf, BLKMULT) < 0) 460 return(-1); 461 arcn->org_name = arcn->name; 462 arcn->sb.st_nlink = 1; 463 arcn->pat = NULL; 464 465 /* 466 * copy out the name and values in the stat buffer 467 */ 468 hd = (HD_TAR *)buf; 469 arcn->nlen = l_strncpy(arcn->name, hd->name, sizeof(hd->name)); 470 arcn->name[arcn->nlen] = '\0'; 471 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) & 472 0xfff); 473 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 474 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 475 arcn->sb.st_size = (size_t)asc_ul(hd->size, sizeof(hd->size), OCT); 476 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 477 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 478 479 /* 480 * have to look at the last character, it may be a '/' and that is used 481 * to encode this as a directory 482 */ 483 pt = &(arcn->name[arcn->nlen - 1]); 484 arcn->pad = 0; 485 arcn->skip = 0; 486 switch(hd->linkflag) { 487 case SYMTYPE: 488 /* 489 * symbolic link, need to get the link name and set the type in 490 * the st_mode so -v printing will look correct. 491 */ 492 arcn->type = PAX_SLK; 493 arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname, 494 sizeof(hd->linkname)); 495 arcn->ln_name[arcn->ln_nlen] = '\0'; 496 arcn->sb.st_mode |= S_IFLNK; 497 break; 498 case LNKTYPE: 499 /* 500 * hard link, need to get the link name, set the type in the 501 * st_mode and st_nlink so -v printing will look better. 502 */ 503 arcn->type = PAX_HLK; 504 arcn->sb.st_nlink = 2; 505 arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname, 506 sizeof(hd->linkname)); 507 arcn->ln_name[arcn->ln_nlen] = '\0'; 508 509 /* 510 * no idea of what type this thing really points at, but 511 * we set something for printing only. 512 */ 513 arcn->sb.st_mode |= S_IFREG; 514 break; 515 case AREGTYPE: 516 case REGTYPE: 517 case DIRTYPE: /* see below */ 518 default: 519 /* 520 * If we have a trailing / this is a directory and NOT a file. 521 * Note: V7 tar doesn't actually have DIRTYPE, but it was 522 * reported that V7 archives using USTAR directories do exist. 523 */ 524 arcn->ln_name[0] = '\0'; 525 arcn->ln_nlen = 0; 526 if (*pt == '/' || hd->linkflag == DIRTYPE) { 527 /* 528 * it is a directory, set the mode for -v printing 529 */ 530 arcn->type = PAX_DIR; 531 arcn->sb.st_mode |= S_IFDIR; 532 arcn->sb.st_nlink = 2; 533 } else { 534 /* 535 * have a file that will be followed by data. Set the 536 * skip value to the size field and caluculate the size 537 * of the padding. 538 */ 539 arcn->type = PAX_REG; 540 arcn->sb.st_mode |= S_IFREG; 541 arcn->pad = TAR_PAD(arcn->sb.st_size); 542 arcn->skip = arcn->sb.st_size; 543 } 544 break; 545 } 546 547 /* 548 * strip off any trailing slash. 549 */ 550 if (*pt == '/') { 551 *pt = '\0'; 552 --arcn->nlen; 553 } 554 return(0); 555} 556 557/* 558 * tar_wr() 559 * write a tar header for the file specified in the ARCHD to the archive. 560 * Have to check for file types that cannot be stored and file names that 561 * are too long. Be careful of the term (last arg) to ul_oct, each field 562 * of tar has it own spec for the termination character(s). 563 * ASSUMED: space after header in header block is zero filled 564 * Return: 565 * 0 if file has data to be written after the header, 1 if file has NO 566 * data to write after the header, -1 if archive write failed 567 */ 568 569#if __STDC__ 570int 571tar_wr(ARCHD *arcn) 572#else 573int 574tar_wr(arcn) 575 ARCHD *arcn; 576#endif 577{ 578 HD_TAR *hd; 579 int len; 580 char hdblk[sizeof(HD_TAR)]; 581 582 /* 583 * check for those file system types which tar cannot store 584 */ 585 switch(arcn->type) { 586 case PAX_DIR: 587 /* 588 * user asked that dirs not be written to the archive 589 */ 590 if (tar_nodir) 591 return(1); 592 break; 593 case PAX_CHR: 594 tty_warn(1, "Tar cannot archive a character device %s", 595 arcn->org_name); 596 return(1); 597 case PAX_BLK: 598 tty_warn(1, 599 "Tar cannot archive a block device %s", arcn->org_name); 600 return(1); 601 case PAX_SCK: 602 tty_warn(1, "Tar cannot archive a socket %s", arcn->org_name); 603 return(1); 604 case PAX_FIF: 605 tty_warn(1, "Tar cannot archive a fifo %s", arcn->org_name); 606 return(1); 607 case PAX_SLK: 608 case PAX_HLK: 609 case PAX_HRG: 610 if (arcn->ln_nlen > sizeof(hd->linkname)) { 611 tty_warn(1,"Link name too long for tar %s", 612 arcn->ln_name); 613 return(1); 614 } 615 break; 616 case PAX_REG: 617 case PAX_CTG: 618 default: 619 break; 620 } 621 622 /* 623 * check file name len, remember extra char for dirs (the / at the end) 624 */ 625 len = arcn->nlen; 626 if (arcn->type == PAX_DIR) 627 ++len; 628 if (len > sizeof(hd->name)) { 629 tty_warn(1, "File name too long for tar %s", arcn->name); 630 return(1); 631 } 632 633 /* 634 * copy the data out of the ARCHD into the tar header based on the type 635 * of the file. Remember many tar readers want the unused fields to be 636 * padded with zero. We set the linkflag field (type), the linkname 637 * (or zero if not used),the size, and set the padding (if any) to be 638 * added after the file data (0 for all other types, as they only have 639 * a header) 640 */ 641 hd = (HD_TAR *)hdblk; 642 zf_strncpy(hd->name, arcn->name, sizeof(hd->name)); 643 arcn->pad = 0; 644 645 if (arcn->type == PAX_DIR) { 646 /* 647 * directories are the same as files, except have a filename 648 * that ends with a /, we add the slash here. No data follows, 649 * dirs, so no pad. 650 */ 651 hd->linkflag = AREGTYPE; 652 memset(hd->linkname, 0, sizeof(hd->linkname)); 653 hd->name[len-1] = '/'; 654 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 655 goto out; 656 } else if (arcn->type == PAX_SLK) { 657 /* 658 * no data follows this file, so no pad 659 */ 660 hd->linkflag = SYMTYPE; 661 zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname)); 662 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 663 goto out; 664 } else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) { 665 /* 666 * no data follows this file, so no pad 667 */ 668 hd->linkflag = LNKTYPE; 669 zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname)); 670 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 671 goto out; 672 } else { 673 /* 674 * data follows this file, so set the pad 675 */ 676 hd->linkflag = AREGTYPE; 677 memset(hd->linkname, 0, sizeof(hd->linkname)); 678# ifdef NET2_STAT 679 if (ul_oct((u_long)arcn->sb.st_size, hd->size, 680 sizeof(hd->size), 1)) { 681# else 682 if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size, 683 sizeof(hd->size), 1)) { 684# endif 685 tty_warn(1,"File is too large for tar %s", 686 arcn->org_name); 687 return(1); 688 } 689 arcn->pad = TAR_PAD(arcn->sb.st_size); 690 } 691 692 /* 693 * copy those fields that are independent of the type 694 */ 695 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) || 696 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) || 697 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) || 698 ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1)) 699 goto out; 700 701 /* 702 * calculate and add the checksum, then write the header. A return of 703 * 0 tells the caller to now write the file data, 1 says no data needs 704 * to be written 705 */ 706 if (ul_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum, 707 sizeof(hd->chksum), 2)) 708 goto out; 709 if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0) 710 return(-1); 711 if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0) 712 return(-1); 713 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 714 return(0); 715 return(1); 716 717 out: 718 /* 719 * header field is out of range 720 */ 721 tty_warn(1, "Tar header field is too small for %s", arcn->org_name); 722 return(1); 723} 724 725/* 726 * Routines for POSIX ustar 727 */ 728 729/* 730 * ustar_strd() 731 * initialization for ustar read 732 * Return: 733 * 0 if ok, -1 otherwise 734 */ 735 736#if __STDC__ 737int 738ustar_strd(void) 739#else 740int 741ustar_strd() 742#endif 743{ 744 return(0); 745} 746 747/* 748 * ustar_stwr() 749 * initialization for ustar write 750 * Return: 751 * 0 if ok, -1 otherwise 752 */ 753 754#if __STDC__ 755int 756ustar_stwr(void) 757#else 758int 759ustar_stwr() 760#endif 761{ 762 return(0); 763} 764 765/* 766 * ustar_id() 767 * determine if a block given to us is a valid ustar header. We have to 768 * be on the lookout for those pesky blocks of all zero's 769 * Return: 770 * 0 if a ustar header, -1 otherwise 771 */ 772 773#if __STDC__ 774int 775ustar_id(char *blk, int size) 776#else 777int 778ustar_id(blk, size) 779 char *blk; 780 int size; 781#endif 782{ 783 HD_USTAR *hd; 784 785 if (size < BLKMULT) 786 return(-1); 787 hd = (HD_USTAR *)blk; 788 789 /* 790 * check for block of zero's first, a simple and fast test then check 791 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive 792 * programs are fouled up and create archives missing the \0. Last we 793 * check the checksum. If ok we have to assume it is a valid header. 794 */ 795 if (hd->name[0] == '\0') 796 return(-1); 797 if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0) 798 return(-1); 799 if (!strncmp(hd->magic, "ustar ", 8)) 800 is_oldgnutar = 1; 801 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 802 return(-1); 803 return(0); 804} 805 806/* 807 * ustar_rd() 808 * extract the values out of block already determined to be a ustar header. 809 * store the values in the ARCHD parameter. 810 * Return: 811 * 0 812 */ 813 814#if __STDC__ 815int 816ustar_rd(ARCHD *arcn, char *buf) 817#else 818int 819ustar_rd(arcn, buf) 820 ARCHD *arcn; 821 char *buf; 822#endif 823{ 824 HD_USTAR *hd; 825 char *dest; 826 int cnt; 827 dev_t devmajor; 828 dev_t devminor; 829 830 /* 831 * we only get proper sized buffers 832 */ 833 if (ustar_id(buf, BLKMULT) < 0) 834 return(-1); 835 arcn->org_name = arcn->name; 836 arcn->sb.st_nlink = 1; 837 arcn->pat = NULL; 838 hd = (HD_USTAR *)buf; 839 840 /* 841 * see if the filename is split into two parts. if, so joint the parts. 842 * we copy the prefix first and add a / between the prefix and name. 843 */ 844 dest = arcn->name; 845 if (*(hd->prefix) != '\0') { 846 cnt = l_strncpy(arcn->name, hd->prefix, sizeof(hd->prefix)); 847 dest += cnt; 848 *dest++ = '/'; 849 } 850 cnt = l_strncpy(dest, hd->name, sizeof(hd->name)); 851 dest += cnt; 852 *dest = '\0'; 853 arcn->nlen = dest - arcn->name; 854 855 /* 856 * follow the spec to the letter. we should only have mode bits, strip 857 * off all other crud we may be passed. 858 */ 859 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) & 860 0xfff); 861 arcn->sb.st_size = (size_t)asc_ul(hd->size, sizeof(hd->size), OCT); 862 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 863 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 864 865 /* 866 * If we can find the ascii names for gname and uname in the password 867 * and group files we will use the uid's and gid they bind. Otherwise 868 * we use the uid and gid values stored in the header. (This is what 869 * the posix spec wants). 870 */ 871 hd->gname[sizeof(hd->gname) - 1] = '\0'; 872 if (gid_from_group(hd->gname, &(arcn->sb.st_gid)) < 0) 873 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 874 hd->uname[sizeof(hd->uname) - 1] = '\0'; 875 if (uid_from_user(hd->uname, &(arcn->sb.st_uid)) < 0) 876 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 877 878 /* 879 * set the defaults, these may be changed depending on the file type 880 */ 881 arcn->ln_name[0] = '\0'; 882 arcn->ln_nlen = 0; 883 arcn->pad = 0; 884 arcn->skip = 0; 885 arcn->sb.st_rdev = (dev_t)0; 886 887 /* 888 * set the mode and PAX type according to the typeflag in the header 889 */ 890 switch(hd->typeflag) { 891 case FIFOTYPE: 892 arcn->type = PAX_FIF; 893 arcn->sb.st_mode |= S_IFIFO; 894 break; 895 case DIRTYPE: 896 arcn->type = PAX_DIR; 897 arcn->sb.st_mode |= S_IFDIR; 898 arcn->sb.st_nlink = 2; 899 900 /* 901 * Some programs that create ustar archives append a '/' 902 * to the pathname for directories. This clearly violates 903 * ustar specs, but we will silently strip it off anyway. 904 */ 905 if (arcn->name[arcn->nlen - 1] == '/') 906 arcn->name[--arcn->nlen] = '\0'; 907 break; 908 case BLKTYPE: 909 case CHRTYPE: 910 /* 911 * this type requires the rdev field to be set. 912 */ 913 if (hd->typeflag == BLKTYPE) { 914 arcn->type = PAX_BLK; 915 arcn->sb.st_mode |= S_IFBLK; 916 } else { 917 arcn->type = PAX_CHR; 918 arcn->sb.st_mode |= S_IFCHR; 919 } 920 devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT); 921 devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT); 922 arcn->sb.st_rdev = TODEV(devmajor, devminor); 923 break; 924 case SYMTYPE: 925 case LNKTYPE: 926 if (hd->typeflag == SYMTYPE) { 927 arcn->type = PAX_SLK; 928 arcn->sb.st_mode |= S_IFLNK; 929 } else { 930 arcn->type = PAX_HLK; 931 /* 932 * so printing looks better 933 */ 934 arcn->sb.st_mode |= S_IFREG; 935 arcn->sb.st_nlink = 2; 936 } 937 /* 938 * copy the link name 939 */ 940 arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname, 941 sizeof(hd->linkname)); 942 arcn->ln_name[arcn->ln_nlen] = '\0'; 943 break; 944 case CONTTYPE: 945 case AREGTYPE: 946 case REGTYPE: 947 default: 948 /* 949 * these types have file data that follows. Set the skip and 950 * pad fields. 951 */ 952 arcn->type = PAX_REG; 953 arcn->pad = TAR_PAD(arcn->sb.st_size); 954 arcn->skip = arcn->sb.st_size; 955 arcn->sb.st_mode |= S_IFREG; 956 break; 957 } 958 return(0); 959} 960 961/* 962 * ustar_wr() 963 * write a ustar header for the file specified in the ARCHD to the archive 964 * Have to check for file types that cannot be stored and file names that 965 * are too long. Be careful of the term (last arg) to ul_oct, we only use 966 * '\0' for the termination character (this is different than picky tar) 967 * ASSUMED: space after header in header block is zero filled 968 * Return: 969 * 0 if file has data to be written after the header, 1 if file has NO 970 * data to write after the header, -1 if archive write failed 971 */ 972 973#if __STDC__ 974int 975ustar_wr(ARCHD *arcn) 976#else 977int 978ustar_wr(arcn) 979 ARCHD *arcn; 980#endif 981{ 982 HD_USTAR *hd; 983 char *pt; 984 char hdblk[sizeof(HD_USTAR)]; 985 const char *user, *group; 986 987 /* 988 * check for those file system types ustar cannot store 989 */ 990 if (arcn->type == PAX_SCK) { 991 tty_warn(1, "Ustar cannot archive a socket %s", arcn->org_name); 992 return(1); 993 } 994 995 /* 996 * check the length of the linkname 997 */ 998 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) || 999 (arcn->type == PAX_HRG)) && (arcn->ln_nlen > sizeof(hd->linkname))){ 1000 tty_warn(1, "Link name too long for ustar %s", arcn->ln_name); 1001 return(1); 1002 } 1003 1004 /* 1005 * split the path name into prefix and name fields (if needed). if 1006 * pt != arcn->name, the name has to be split 1007 */ 1008 if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) { 1009 tty_warn(1, "File name too long for ustar %s", arcn->name); 1010 return(1); 1011 } 1012 hd = (HD_USTAR *)hdblk; 1013 arcn->pad = 0L; 1014 1015 /* 1016 * split the name, or zero out the prefix 1017 */ 1018 if (pt != arcn->name) { 1019 /* 1020 * name was split, pt points at the / where the split is to 1021 * occur, we remove the / and copy the first part to the prefix 1022 */ 1023 *pt = '\0'; 1024 zf_strncpy(hd->prefix, arcn->name, sizeof(hd->prefix)); 1025 *pt++ = '/'; 1026 } else 1027 memset(hd->prefix, 0, sizeof(hd->prefix)); 1028 1029 /* 1030 * copy the name part. this may be the whole path or the part after 1031 * the prefix 1032 */ 1033 zf_strncpy(hd->name, pt, sizeof(hd->name)); 1034 1035 /* 1036 * set the fields in the header that are type dependent 1037 */ 1038 switch(arcn->type) { 1039 case PAX_DIR: 1040 hd->typeflag = DIRTYPE; 1041 memset(hd->linkname, 0, sizeof(hd->linkname)); 1042 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1043 memset(hd->devminor, 0, sizeof(hd->devminor)); 1044 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1045 goto out; 1046 break; 1047 case PAX_CHR: 1048 case PAX_BLK: 1049 if (arcn->type == PAX_CHR) 1050 hd->typeflag = CHRTYPE; 1051 else 1052 hd->typeflag = BLKTYPE; 1053 memset(hd->linkname, 0, sizeof(hd->linkname)); 1054 if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor, 1055 sizeof(hd->devmajor), 3) || 1056 ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor, 1057 sizeof(hd->devminor), 3) || 1058 ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1059 goto out; 1060 break; 1061 case PAX_FIF: 1062 hd->typeflag = FIFOTYPE; 1063 memset(hd->linkname, 0, sizeof(hd->linkname)); 1064 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1065 memset(hd->devminor, 0, sizeof(hd->devminor)); 1066 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1067 goto out; 1068 break; 1069 case PAX_SLK: 1070 case PAX_HLK: 1071 case PAX_HRG: 1072 if (arcn->type == PAX_SLK) 1073 hd->typeflag = SYMTYPE; 1074 else 1075 hd->typeflag = LNKTYPE; 1076 zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname)); 1077 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1078 memset(hd->devminor, 0, sizeof(hd->devminor)); 1079 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1080 goto out; 1081 break; 1082 case PAX_REG: 1083 case PAX_CTG: 1084 default: 1085 /* 1086 * file data with this type, set the padding 1087 */ 1088 if (arcn->type == PAX_CTG) 1089 hd->typeflag = CONTTYPE; 1090 else 1091 hd->typeflag = REGTYPE; 1092 memset(hd->linkname, 0, sizeof(hd->linkname)); 1093 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1094 memset(hd->devminor, 0, sizeof(hd->devminor)); 1095 arcn->pad = TAR_PAD(arcn->sb.st_size); 1096# ifdef NET2_STAT 1097 if (ul_oct((u_long)arcn->sb.st_size, hd->size, 1098 sizeof(hd->size), 3)) { 1099# else 1100 if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size, 1101 sizeof(hd->size), 3)) { 1102# endif 1103 tty_warn(1,"File is too long for ustar %s", 1104 arcn->org_name); 1105 return(1); 1106 } 1107 break; 1108 } 1109 1110 zf_strncpy(hd->magic, TMAGIC, TMAGLEN); 1111 zf_strncpy(hd->version, TVERSION, TVERSLEN); 1112 1113 /* 1114 * set the remaining fields. Some versions want all 16 bits of mode 1115 * we better humor them (they really do not meet spec though).... 1116 */ 1117 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) || 1118 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3) || 1119 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) || 1120 ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3)) 1121 goto out; 1122 user = user_from_uid(arcn->sb.st_uid, 1); 1123 group = group_from_gid(arcn->sb.st_gid, 1); 1124 zf_strncpy(hd->uname, user ? user : "", sizeof(hd->uname)); 1125 zf_strncpy(hd->gname, group ? group : "", sizeof(hd->gname)); 1126 1127 /* 1128 * calculate and store the checksum write the header to the archive 1129 * return 0 tells the caller to now write the file data, 1 says no data 1130 * needs to be written 1131 */ 1132 if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum, 1133 sizeof(hd->chksum), 3)) 1134 goto out; 1135 if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0) 1136 return(-1); 1137 if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0) 1138 return(-1); 1139 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 1140 return(0); 1141 return(1); 1142 1143 out: 1144 /* 1145 * header field is out of range 1146 */ 1147 tty_warn(1, "Ustar header field is too small for %s", arcn->org_name); 1148 return(1); 1149} 1150 1151/* 1152 * name_split() 1153 * see if the name has to be split for storage in a ustar header. We try 1154 * to fit the entire name in the name field without splitting if we can. 1155 * The split point is always at a / 1156 * Return 1157 * character pointer to split point (always the / that is to be removed 1158 * if the split is not needed, the points is set to the start of the file 1159 * name (it would violate the spec to split there). A NULL is returned if 1160 * the file name is too long 1161 */ 1162 1163#if __STDC__ 1164static char * 1165name_split(char *name, int len) 1166#else 1167static char * 1168name_split(name, len) 1169 char *name; 1170 int len; 1171#endif 1172{ 1173 char *start; 1174 1175 /* 1176 * check to see if the file name is small enough to fit in the name 1177 * field. if so just return a pointer to the name. 1178 */ 1179 if (len <= TNMSZ) 1180 return(name); 1181 if (len > (TPFSZ + TNMSZ + 1)) 1182 return(NULL); 1183 1184 /* 1185 * we start looking at the biggest sized piece that fits in the name 1186 * field. We walk foward looking for a slash to split at. The idea is 1187 * to find the biggest piece to fit in the name field (or the smallest 1188 * prefix we can find) (the -1 is correct the biggest piece would 1189 * include the slash between the two parts that gets thrown away) 1190 */ 1191 start = name + len - TNMSZ - 1; 1192 while ((*start != '\0') && (*start != '/')) 1193 ++start; 1194 1195 /* 1196 * if we hit the end of the string, this name cannot be split, so we 1197 * cannot store this file. 1198 */ 1199 if (*start == '\0') 1200 return(NULL); 1201 len = start - name; 1202 1203 /* 1204 * NOTE: /str where the length of str == TNMSZ can not be stored under 1205 * the p1003.1-1990 spec for ustar. We could force a prefix of / and 1206 * the file would then expand on extract to //str. The len == 0 below 1207 * makes this special case follow the spec to the letter. 1208 */ 1209 if ((len > TPFSZ) || (len == 0)) 1210 return(NULL); 1211 1212 /* 1213 * ok have a split point, return it to the caller 1214 */ 1215 return(start); 1216} 1217 1218/* 1219 * deal with GNU tar -X switch. basically, we go through each line of 1220 * the file, building a string from the "glob" lines in the file into 1221 * RE lines, of the form `/^RE$//', which we pass to rep_add(), which 1222 * will add a empty replacement (exclusion), for the named files. 1223 */ 1224int 1225tar_gnutar_X_compat(path) 1226 const char *path; 1227{ 1228 char *line, sbuf[MAXPATHLEN * 2 + 1 + 5]; 1229 FILE *fp; 1230 int lineno = 0, i, j; 1231 size_t len; 1232 1233 fp = fopen(path, "r"); 1234 if (fp == NULL) { 1235 tty_warn(1, "can not open %s: %s", path, 1236 strerror(errno)); 1237 return(-1); 1238 } 1239 1240 while ((line = fgetln(fp, &len))) { 1241 lineno++; 1242 if (len > MAXPATHLEN) { 1243 tty_warn(0, "pathname too long, line %d of %s", 1244 lineno, path); 1245 } 1246 if (line[len - 1] == '\n') 1247 len--; 1248 for (i = 0, j = 2; i < len; i++) { 1249 /* 1250 * convert glob to regexp, escaping everything 1251 */ 1252 if (line[i] == '*') 1253 sbuf[j++] = '.'; 1254 else if (line[i] == '?') 1255 line[i] = '.'; 1256 else if (!isalnum(line[i]) && !isblank(line[i])) 1257 sbuf[j++] = '\\'; 1258 sbuf[j++] = line[i]; 1259 } 1260 sbuf[0] = sbuf[j + 1] = sbuf[j + 2] = '/'; 1261 sbuf[1] = '^'; 1262 sbuf[j] = '$'; 1263 sbuf[j + 3] = '\0'; 1264 if (rep_add(sbuf) < 0) 1265 return (-1); 1266 } 1267 return (0); 1268} 1269