tar.c revision 1.13
1/* $NetBSD: tar.c,v 1.13 1999/01/20 14:45:09 mrg 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 1.13 1999/01/20 14:45:09 mrg 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 default: 518 /* 519 * If we have a trailing / this is a directory and NOT a file. 520 */ 521 arcn->ln_name[0] = '\0'; 522 arcn->ln_nlen = 0; 523 if (*pt == '/') { 524 /* 525 * it is a directory, set the mode for -v printing 526 */ 527 arcn->type = PAX_DIR; 528 arcn->sb.st_mode |= S_IFDIR; 529 arcn->sb.st_nlink = 2; 530 } else { 531 /* 532 * have a file that will be followed by data. Set the 533 * skip value to the size field and caluculate the size 534 * of the padding. 535 */ 536 arcn->type = PAX_REG; 537 arcn->sb.st_mode |= S_IFREG; 538 arcn->pad = TAR_PAD(arcn->sb.st_size); 539 arcn->skip = arcn->sb.st_size; 540 } 541 break; 542 } 543 544 /* 545 * strip off any trailing slash. 546 */ 547 if (*pt == '/') { 548 *pt = '\0'; 549 --arcn->nlen; 550 } 551 return(0); 552} 553 554/* 555 * tar_wr() 556 * write a tar header for the file specified in the ARCHD to the archive. 557 * Have to check for file types that cannot be stored and file names that 558 * are too long. Be careful of the term (last arg) to ul_oct, each field 559 * of tar has it own spec for the termination character(s). 560 * ASSUMED: space after header in header block is zero filled 561 * Return: 562 * 0 if file has data to be written after the header, 1 if file has NO 563 * data to write after the header, -1 if archive write failed 564 */ 565 566#if __STDC__ 567int 568tar_wr(ARCHD *arcn) 569#else 570int 571tar_wr(arcn) 572 ARCHD *arcn; 573#endif 574{ 575 HD_TAR *hd; 576 int len; 577 char hdblk[sizeof(HD_TAR)]; 578 579 /* 580 * check for those file system types which tar cannot store 581 */ 582 switch(arcn->type) { 583 case PAX_DIR: 584 /* 585 * user asked that dirs not be written to the archive 586 */ 587 if (tar_nodir) 588 return(1); 589 break; 590 case PAX_CHR: 591 tty_warn(1, "Tar cannot archive a character device %s", 592 arcn->org_name); 593 return(1); 594 case PAX_BLK: 595 tty_warn(1, 596 "Tar cannot archive a block device %s", arcn->org_name); 597 return(1); 598 case PAX_SCK: 599 tty_warn(1, "Tar cannot archive a socket %s", arcn->org_name); 600 return(1); 601 case PAX_FIF: 602 tty_warn(1, "Tar cannot archive a fifo %s", arcn->org_name); 603 return(1); 604 case PAX_SLK: 605 case PAX_HLK: 606 case PAX_HRG: 607 if (arcn->ln_nlen > sizeof(hd->linkname)) { 608 tty_warn(1,"Link name too long for tar %s", 609 arcn->ln_name); 610 return(1); 611 } 612 break; 613 case PAX_REG: 614 case PAX_CTG: 615 default: 616 break; 617 } 618 619 /* 620 * check file name len, remember extra char for dirs (the / at the end) 621 */ 622 len = arcn->nlen; 623 if (arcn->type == PAX_DIR) 624 ++len; 625 if (len > sizeof(hd->name)) { 626 tty_warn(1, "File name too long for tar %s", arcn->name); 627 return(1); 628 } 629 630 /* 631 * copy the data out of the ARCHD into the tar header based on the type 632 * of the file. Remember many tar readers want the unused fields to be 633 * padded with zero. We set the linkflag field (type), the linkname 634 * (or zero if not used),the size, and set the padding (if any) to be 635 * added after the file data (0 for all other types, as they only have 636 * a header) 637 */ 638 hd = (HD_TAR *)hdblk; 639 zf_strncpy(hd->name, arcn->name, sizeof(hd->name)); 640 arcn->pad = 0; 641 642 if (arcn->type == PAX_DIR) { 643 /* 644 * directories are the same as files, except have a filename 645 * that ends with a /, we add the slash here. No data follows, 646 * dirs, so no pad. 647 */ 648 hd->linkflag = AREGTYPE; 649 memset(hd->linkname, 0, sizeof(hd->linkname)); 650 hd->name[len-1] = '/'; 651 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 652 goto out; 653 } else if (arcn->type == PAX_SLK) { 654 /* 655 * no data follows this file, so no pad 656 */ 657 hd->linkflag = SYMTYPE; 658 zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname)); 659 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 660 goto out; 661 } else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) { 662 /* 663 * no data follows this file, so no pad 664 */ 665 hd->linkflag = LNKTYPE; 666 zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname)); 667 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 668 goto out; 669 } else { 670 /* 671 * data follows this file, so set the pad 672 */ 673 hd->linkflag = AREGTYPE; 674 memset(hd->linkname, 0, sizeof(hd->linkname)); 675# ifdef NET2_STAT 676 if (ul_oct((u_long)arcn->sb.st_size, hd->size, 677 sizeof(hd->size), 1)) { 678# else 679 if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size, 680 sizeof(hd->size), 1)) { 681# endif 682 tty_warn(1,"File is too large for tar %s", 683 arcn->org_name); 684 return(1); 685 } 686 arcn->pad = TAR_PAD(arcn->sb.st_size); 687 } 688 689 /* 690 * copy those fields that are independent of the type 691 */ 692 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) || 693 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) || 694 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) || 695 ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1)) 696 goto out; 697 698 /* 699 * calculate and add the checksum, then write the header. A return of 700 * 0 tells the caller to now write the file data, 1 says no data needs 701 * to be written 702 */ 703 if (ul_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum, 704 sizeof(hd->chksum), 2)) 705 goto out; 706 if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0) 707 return(-1); 708 if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0) 709 return(-1); 710 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 711 return(0); 712 return(1); 713 714 out: 715 /* 716 * header field is out of range 717 */ 718 tty_warn(1, "Tar header field is too small for %s", arcn->org_name); 719 return(1); 720} 721 722/* 723 * Routines for POSIX ustar 724 */ 725 726/* 727 * ustar_strd() 728 * initialization for ustar read 729 * Return: 730 * 0 if ok, -1 otherwise 731 */ 732 733#if __STDC__ 734int 735ustar_strd(void) 736#else 737int 738ustar_strd() 739#endif 740{ 741 return(0); 742} 743 744/* 745 * ustar_stwr() 746 * initialization for ustar write 747 * Return: 748 * 0 if ok, -1 otherwise 749 */ 750 751#if __STDC__ 752int 753ustar_stwr(void) 754#else 755int 756ustar_stwr() 757#endif 758{ 759 return(0); 760} 761 762/* 763 * ustar_id() 764 * determine if a block given to us is a valid ustar header. We have to 765 * be on the lookout for those pesky blocks of all zero's 766 * Return: 767 * 0 if a ustar header, -1 otherwise 768 */ 769 770#if __STDC__ 771int 772ustar_id(char *blk, int size) 773#else 774int 775ustar_id(blk, size) 776 char *blk; 777 int size; 778#endif 779{ 780 HD_USTAR *hd; 781 782 if (size < BLKMULT) 783 return(-1); 784 hd = (HD_USTAR *)blk; 785 786 /* 787 * check for block of zero's first, a simple and fast test then check 788 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive 789 * programs are fouled up and create archives missing the \0. Last we 790 * check the checksum. If ok we have to assume it is a valid header. 791 */ 792 if (hd->name[0] == '\0') 793 return(-1); 794 if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0) 795 return(-1); 796 if (!strncmp(hd->magic, "ustar ", 8)) 797 is_oldgnutar = 1; 798 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 799 return(-1); 800 return(0); 801} 802 803/* 804 * ustar_rd() 805 * extract the values out of block already determined to be a ustar header. 806 * store the values in the ARCHD parameter. 807 * Return: 808 * 0 809 */ 810 811#if __STDC__ 812int 813ustar_rd(ARCHD *arcn, char *buf) 814#else 815int 816ustar_rd(arcn, buf) 817 ARCHD *arcn; 818 char *buf; 819#endif 820{ 821 HD_USTAR *hd; 822 char *dest; 823 int cnt; 824 dev_t devmajor; 825 dev_t devminor; 826 827 /* 828 * we only get proper sized buffers 829 */ 830 if (ustar_id(buf, BLKMULT) < 0) 831 return(-1); 832 arcn->org_name = arcn->name; 833 arcn->sb.st_nlink = 1; 834 arcn->pat = NULL; 835 hd = (HD_USTAR *)buf; 836 837 /* 838 * see if the filename is split into two parts. if, so joint the parts. 839 * we copy the prefix first and add a / between the prefix and name. 840 */ 841 dest = arcn->name; 842 if (*(hd->prefix) != '\0') { 843 cnt = l_strncpy(arcn->name, hd->prefix, sizeof(hd->prefix)); 844 dest += cnt; 845 *dest++ = '/'; 846 } 847 cnt = l_strncpy(dest, hd->name, sizeof(hd->name)); 848 dest += cnt; 849 *dest = '\0'; 850 arcn->nlen = dest - arcn->name; 851 852 /* 853 * follow the spec to the letter. we should only have mode bits, strip 854 * off all other crud we may be passed. 855 */ 856 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) & 857 0xfff); 858 arcn->sb.st_size = (size_t)asc_ul(hd->size, sizeof(hd->size), OCT); 859 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 860 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 861 862 /* 863 * If we can find the ascii names for gname and uname in the password 864 * and group files we will use the uid's and gid they bind. Otherwise 865 * we use the uid and gid values stored in the header. (This is what 866 * the posix spec wants). 867 */ 868 hd->gname[sizeof(hd->gname) - 1] = '\0'; 869 if (gid_from_group(hd->gname, &(arcn->sb.st_gid)) < 0) 870 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 871 hd->uname[sizeof(hd->uname) - 1] = '\0'; 872 if (uid_from_user(hd->uname, &(arcn->sb.st_uid)) < 0) 873 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 874 875 /* 876 * set the defaults, these may be changed depending on the file type 877 */ 878 arcn->ln_name[0] = '\0'; 879 arcn->ln_nlen = 0; 880 arcn->pad = 0; 881 arcn->skip = 0; 882 arcn->sb.st_rdev = (dev_t)0; 883 884 /* 885 * set the mode and PAX type according to the typeflag in the header 886 */ 887 switch(hd->typeflag) { 888 case FIFOTYPE: 889 arcn->type = PAX_FIF; 890 arcn->sb.st_mode |= S_IFIFO; 891 break; 892 case DIRTYPE: 893 arcn->type = PAX_DIR; 894 arcn->sb.st_mode |= S_IFDIR; 895 arcn->sb.st_nlink = 2; 896 897 /* 898 * Some programs that create ustar archives append a '/' 899 * to the pathname for directories. This clearly violates 900 * ustar specs, but we will silently strip it off anyway. 901 */ 902 if (arcn->name[arcn->nlen - 1] == '/') 903 arcn->name[--arcn->nlen] = '\0'; 904 break; 905 case BLKTYPE: 906 case CHRTYPE: 907 /* 908 * this type requires the rdev field to be set. 909 */ 910 if (hd->typeflag == BLKTYPE) { 911 arcn->type = PAX_BLK; 912 arcn->sb.st_mode |= S_IFBLK; 913 } else { 914 arcn->type = PAX_CHR; 915 arcn->sb.st_mode |= S_IFCHR; 916 } 917 devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT); 918 devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT); 919 arcn->sb.st_rdev = TODEV(devmajor, devminor); 920 break; 921 case SYMTYPE: 922 case LNKTYPE: 923 if (hd->typeflag == SYMTYPE) { 924 arcn->type = PAX_SLK; 925 arcn->sb.st_mode |= S_IFLNK; 926 } else { 927 arcn->type = PAX_HLK; 928 /* 929 * so printing looks better 930 */ 931 arcn->sb.st_mode |= S_IFREG; 932 arcn->sb.st_nlink = 2; 933 } 934 /* 935 * copy the link name 936 */ 937 arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname, 938 sizeof(hd->linkname)); 939 arcn->ln_name[arcn->ln_nlen] = '\0'; 940 break; 941 case CONTTYPE: 942 case AREGTYPE: 943 case REGTYPE: 944 default: 945 /* 946 * these types have file data that follows. Set the skip and 947 * pad fields. 948 */ 949 arcn->type = PAX_REG; 950 arcn->pad = TAR_PAD(arcn->sb.st_size); 951 arcn->skip = arcn->sb.st_size; 952 arcn->sb.st_mode |= S_IFREG; 953 break; 954 } 955 return(0); 956} 957 958/* 959 * ustar_wr() 960 * write a ustar header for the file specified in the ARCHD to the archive 961 * Have to check for file types that cannot be stored and file names that 962 * are too long. Be careful of the term (last arg) to ul_oct, we only use 963 * '\0' for the termination character (this is different than picky tar) 964 * ASSUMED: space after header in header block is zero filled 965 * Return: 966 * 0 if file has data to be written after the header, 1 if file has NO 967 * data to write after the header, -1 if archive write failed 968 */ 969 970#if __STDC__ 971int 972ustar_wr(ARCHD *arcn) 973#else 974int 975ustar_wr(arcn) 976 ARCHD *arcn; 977#endif 978{ 979 HD_USTAR *hd; 980 char *pt; 981 char hdblk[sizeof(HD_USTAR)]; 982 const char *user, *group; 983 984 /* 985 * check for those file system types ustar cannot store 986 */ 987 if (arcn->type == PAX_SCK) { 988 tty_warn(1, "Ustar cannot archive a socket %s", arcn->org_name); 989 return(1); 990 } 991 992 /* 993 * check the length of the linkname 994 */ 995 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) || 996 (arcn->type == PAX_HRG)) && (arcn->ln_nlen > sizeof(hd->linkname))){ 997 tty_warn(1, "Link name too long for ustar %s", arcn->ln_name); 998 return(1); 999 } 1000 1001 /* 1002 * split the path name into prefix and name fields (if needed). if 1003 * pt != arcn->name, the name has to be split 1004 */ 1005 if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) { 1006 tty_warn(1, "File name too long for ustar %s", arcn->name); 1007 return(1); 1008 } 1009 hd = (HD_USTAR *)hdblk; 1010 arcn->pad = 0L; 1011 1012 /* 1013 * split the name, or zero out the prefix 1014 */ 1015 if (pt != arcn->name) { 1016 /* 1017 * name was split, pt points at the / where the split is to 1018 * occur, we remove the / and copy the first part to the prefix 1019 */ 1020 *pt = '\0'; 1021 zf_strncpy(hd->prefix, arcn->name, sizeof(hd->prefix)); 1022 *pt++ = '/'; 1023 } else 1024 memset(hd->prefix, 0, sizeof(hd->prefix)); 1025 1026 /* 1027 * copy the name part. this may be the whole path or the part after 1028 * the prefix 1029 */ 1030 zf_strncpy(hd->name, pt, sizeof(hd->name)); 1031 1032 /* 1033 * set the fields in the header that are type dependent 1034 */ 1035 switch(arcn->type) { 1036 case PAX_DIR: 1037 hd->typeflag = DIRTYPE; 1038 memset(hd->linkname, 0, sizeof(hd->linkname)); 1039 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1040 memset(hd->devminor, 0, sizeof(hd->devminor)); 1041 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1042 goto out; 1043 break; 1044 case PAX_CHR: 1045 case PAX_BLK: 1046 if (arcn->type == PAX_CHR) 1047 hd->typeflag = CHRTYPE; 1048 else 1049 hd->typeflag = BLKTYPE; 1050 memset(hd->linkname, 0, sizeof(hd->linkname)); 1051 if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor, 1052 sizeof(hd->devmajor), 3) || 1053 ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor, 1054 sizeof(hd->devminor), 3) || 1055 ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1056 goto out; 1057 break; 1058 case PAX_FIF: 1059 hd->typeflag = FIFOTYPE; 1060 memset(hd->linkname, 0, sizeof(hd->linkname)); 1061 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1062 memset(hd->devminor, 0, sizeof(hd->devminor)); 1063 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1064 goto out; 1065 break; 1066 case PAX_SLK: 1067 case PAX_HLK: 1068 case PAX_HRG: 1069 if (arcn->type == PAX_SLK) 1070 hd->typeflag = SYMTYPE; 1071 else 1072 hd->typeflag = LNKTYPE; 1073 zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname)); 1074 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1075 memset(hd->devminor, 0, sizeof(hd->devminor)); 1076 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1077 goto out; 1078 break; 1079 case PAX_REG: 1080 case PAX_CTG: 1081 default: 1082 /* 1083 * file data with this type, set the padding 1084 */ 1085 if (arcn->type == PAX_CTG) 1086 hd->typeflag = CONTTYPE; 1087 else 1088 hd->typeflag = REGTYPE; 1089 memset(hd->linkname, 0, sizeof(hd->linkname)); 1090 memset(hd->devmajor, 0, sizeof(hd->devmajor)); 1091 memset(hd->devminor, 0, sizeof(hd->devminor)); 1092 arcn->pad = TAR_PAD(arcn->sb.st_size); 1093# ifdef NET2_STAT 1094 if (ul_oct((u_long)arcn->sb.st_size, hd->size, 1095 sizeof(hd->size), 3)) { 1096# else 1097 if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size, 1098 sizeof(hd->size), 3)) { 1099# endif 1100 tty_warn(1,"File is too long for ustar %s", 1101 arcn->org_name); 1102 return(1); 1103 } 1104 break; 1105 } 1106 1107 zf_strncpy(hd->magic, TMAGIC, TMAGLEN); 1108 zf_strncpy(hd->version, TVERSION, TVERSLEN); 1109 1110 /* 1111 * set the remaining fields. Some versions want all 16 bits of mode 1112 * we better humor them (they really do not meet spec though).... 1113 */ 1114 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) || 1115 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3) || 1116 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) || 1117 ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3)) 1118 goto out; 1119 user = user_from_uid(arcn->sb.st_uid, 1); 1120 group = group_from_gid(arcn->sb.st_gid, 1); 1121 zf_strncpy(hd->uname, user ? user : "", sizeof(hd->uname)); 1122 zf_strncpy(hd->gname, group ? group : "", sizeof(hd->gname)); 1123 1124 /* 1125 * calculate and store the checksum write the header to the archive 1126 * return 0 tells the caller to now write the file data, 1 says no data 1127 * needs to be written 1128 */ 1129 if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum, 1130 sizeof(hd->chksum), 3)) 1131 goto out; 1132 if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0) 1133 return(-1); 1134 if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0) 1135 return(-1); 1136 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 1137 return(0); 1138 return(1); 1139 1140 out: 1141 /* 1142 * header field is out of range 1143 */ 1144 tty_warn(1, "Ustar header field is too small for %s", arcn->org_name); 1145 return(1); 1146} 1147 1148/* 1149 * name_split() 1150 * see if the name has to be split for storage in a ustar header. We try 1151 * to fit the entire name in the name field without splitting if we can. 1152 * The split point is always at a / 1153 * Return 1154 * character pointer to split point (always the / that is to be removed 1155 * if the split is not needed, the points is set to the start of the file 1156 * name (it would violate the spec to split there). A NULL is returned if 1157 * the file name is too long 1158 */ 1159 1160#if __STDC__ 1161static char * 1162name_split(char *name, int len) 1163#else 1164static char * 1165name_split(name, len) 1166 char *name; 1167 int len; 1168#endif 1169{ 1170 char *start; 1171 1172 /* 1173 * check to see if the file name is small enough to fit in the name 1174 * field. if so just return a pointer to the name. 1175 */ 1176 if (len <= TNMSZ) 1177 return(name); 1178 if (len > (TPFSZ + TNMSZ + 1)) 1179 return(NULL); 1180 1181 /* 1182 * we start looking at the biggest sized piece that fits in the name 1183 * field. We walk foward looking for a slash to split at. The idea is 1184 * to find the biggest piece to fit in the name field (or the smallest 1185 * prefix we can find) (the -1 is correct the biggest piece would 1186 * include the slash between the two parts that gets thrown away) 1187 */ 1188 start = name + len - TNMSZ - 1; 1189 while ((*start != '\0') && (*start != '/')) 1190 ++start; 1191 1192 /* 1193 * if we hit the end of the string, this name cannot be split, so we 1194 * cannot store this file. 1195 */ 1196 if (*start == '\0') 1197 return(NULL); 1198 len = start - name; 1199 1200 /* 1201 * NOTE: /str where the length of str == TNMSZ can not be stored under 1202 * the p1003.1-1990 spec for ustar. We could force a prefix of / and 1203 * the file would then expand on extract to //str. The len == 0 below 1204 * makes this special case follow the spec to the letter. 1205 */ 1206 if ((len > TPFSZ) || (len == 0)) 1207 return(NULL); 1208 1209 /* 1210 * ok have a split point, return it to the caller 1211 */ 1212 return(start); 1213} 1214 1215/* 1216 * deal with GNU tar -X switch. basically, we go through each line of 1217 * the file, building a string from the "glob" lines in the file into 1218 * RE lines, of the form `/^RE$//', which we pass to rep_add(), which 1219 * will add a empty replacement (exclusion), for the named files. 1220 */ 1221int 1222tar_gnutar_X_compat(path) 1223 const char *path; 1224{ 1225 char *line, sbuf[MAXPATHLEN * 2 + 1 + 5]; 1226 FILE *fp; 1227 int lineno = 0, len, i, j; 1228 1229 fp = fopen(path, "r"); 1230 if (fp == NULL) { 1231 tty_warn(1, "can not open %s: %s", path, 1232 strerror(errno)); 1233 return(-1); 1234 } 1235 1236 while ((line = fgetln(fp, &len))) { 1237 lineno++; 1238 if (len > MAXPATHLEN) { 1239 tty_warn(0, "pathname too long, line %d of %s", 1240 lineno, path); 1241 } 1242 if (line[len - 1] == '\n') 1243 len--; 1244 for (i = 0, j = 2; i < len; i++) { 1245 /* 1246 * convert glob to regexp, escaping everything 1247 */ 1248 if (line[i] == '*') 1249 sbuf[j++] = '.'; 1250 else if (line[i] == '?') 1251 line[i] = '.'; 1252 else if (!isalnum(line[i]) && !isblank(line[i])) 1253 sbuf[j++] = '\\'; 1254 sbuf[j++] = line[i]; 1255 } 1256 sbuf[0] = sbuf[j + 1] = sbuf[j + 2] = '/'; 1257 sbuf[1] = '^'; 1258 sbuf[j] = '$'; 1259 sbuf[j + 3] = '\0'; 1260 if (rep_add(sbuf) < 0) 1261 return (-1); 1262 } 1263 return (0); 1264} 1265