tar.c revision 1.52
1/* $NetBSD: tar.c,v 1.52 2004/04/25 15:52:30 christos 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. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36#if HAVE_NBTOOL_CONFIG_H 37#include "nbtool_config.h" 38#endif 39 40#include <sys/cdefs.h> 41#if !defined(lint) 42#if 0 43static char sccsid[] = "@(#)tar.c 8.2 (Berkeley) 4/18/94"; 44#else 45__RCSID("$NetBSD: tar.c,v 1.52 2004/04/25 15:52:30 christos 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 int expandname(char *, size_t, char **, const char *, size_t); 72static void longlink(ARCHD *); 73static u_long tar_chksm(char *, int); 74static char *name_split(char *, int); 75static int ul_oct(u_long, char *, int, int); 76#if !defined(NET2_STAT) && !defined(_LP64) 77static int ull_oct(unsigned long long, char *, int, int); 78#endif 79static int tar_gnutar_exclude_one(const char *, size_t); 80static int check_sum(char *, size_t, char *, size_t, int); 81 82/* 83 * Routines common to all versions of tar 84 */ 85 86static int tar_nodir; /* do not write dirs under old tar */ 87int is_gnutar; /* behave like gnu tar; enable gnu 88 * extensions and skip end-ofvolume 89 * checks 90 */ 91static int seen_gnu_warning; /* Have we warned yet? */ 92static char *gnu_hack_string; /* ././@LongLink hackery */ 93static int gnu_hack_len; /* len of gnu_hack_string */ 94char *gnu_name_string; /* ././@LongLink hackery name */ 95char *gnu_link_string; /* ././@LongLink hackery link */ 96static int gnu_short_trailer; /* gnu short trailer */ 97 98static int 99check_sum(char *hd, size_t hdlen, char *bl, size_t bllen, int quiet) 100{ 101 u_long hdck, blck; 102 103 hdck = asc_ul(hd, hdlen, OCT); 104 blck = tar_chksm(bl, bllen); 105 106 if (hdck != blck) { 107 if (!quiet) 108 tty_warn(0, "Header checksum %lo does not match %lo", 109 hdck, blck); 110 return(-1); 111 } 112 return(0); 113} 114 115 116/* 117 * tar_endwr() 118 * add the tar trailer of two null blocks 119 * Return: 120 * 0 if ok, -1 otherwise (what wr_skip returns) 121 */ 122 123int 124tar_endwr(void) 125{ 126 return(wr_skip((off_t)(NULLCNT * BLKMULT))); 127} 128 129/* 130 * tar_endrd() 131 * no cleanup needed here, just return size of trailer (for append) 132 * Return: 133 * size of trailer BLKMULT 134 */ 135 136off_t 137tar_endrd(void) 138{ 139 return((off_t)((gnu_short_trailer ? 1 : NULLCNT) * BLKMULT)); 140} 141 142/* 143 * tar_trail() 144 * Called to determine if a header block is a valid trailer. We are passed 145 * the block, the in_sync flag (which tells us we are in resync mode; 146 * looking for a valid header), and cnt (which starts at zero) which is 147 * used to count the number of empty blocks we have seen so far. 148 * Return: 149 * 0 if a valid trailer, -1 if not a valid trailer, or 1 if the block 150 * could never contain a header. 151 */ 152 153int 154tar_trail(char *buf, int in_resync, int *cnt) 155{ 156 int i; 157 158 gnu_short_trailer = 0; 159 /* 160 * look for all zero, trailer is two consecutive blocks of zero 161 */ 162 for (i = 0; i < BLKMULT; ++i) { 163 if (buf[i] != '\0') 164 break; 165 } 166 167 /* 168 * if not all zero it is not a trailer, but MIGHT be a header. 169 */ 170 if (i != BLKMULT) 171 return(-1); 172 173 /* 174 * When given a zero block, we must be careful! 175 * If we are not in resync mode, check for the trailer. Have to watch 176 * out that we do not mis-identify file data as the trailer, so we do 177 * NOT try to id a trailer during resync mode. During resync mode we 178 * might as well throw this block out since a valid header can NEVER be 179 * a block of all 0 (we must have a valid file name). 180 */ 181 if (!in_resync) { 182 ++*cnt; 183 /* 184 * old GNU tar (up through 1.13) only writes one block of 185 * trailers, so we pretend we got another 186 */ 187 if (is_gnutar) { 188 gnu_short_trailer = 1; 189 ++*cnt; 190 } 191 if (*cnt >= NULLCNT) 192 return(0); 193 } 194 return(1); 195} 196 197/* 198 * ul_oct() 199 * convert an unsigned long to an octal string. many oddball field 200 * termination characters are used by the various versions of tar in the 201 * different fields. term selects which kind to use. str is '0' padded 202 * at the front to len. we are unable to use only one format as many old 203 * tar readers are very cranky about this. 204 * Return: 205 * 0 if the number fit into the string, -1 otherwise 206 */ 207 208static int 209ul_oct(u_long val, char *str, int len, int term) 210{ 211 char *pt; 212 213 /* 214 * term selects the appropriate character(s) for the end of the string 215 */ 216 pt = str + len - 1; 217 switch(term) { 218 case 3: 219 *pt-- = '\0'; 220 break; 221 case 2: 222 *pt-- = ' '; 223 *pt-- = '\0'; 224 break; 225 case 1: 226 *pt-- = ' '; 227 break; 228 case 0: 229 default: 230 *pt-- = '\0'; 231 *pt-- = ' '; 232 break; 233 } 234 235 /* 236 * convert and blank pad if there is space 237 */ 238 while (pt >= str) { 239 *pt-- = '0' + (char)(val & 0x7); 240 if ((val = val >> 3) == (u_long)0) 241 break; 242 } 243 244 while (pt >= str) 245 *pt-- = '0'; 246 if (val != (u_long)0) 247 return(-1); 248 return(0); 249} 250 251#if !defined(NET2_STAT) && !defined(_LP64) 252/* 253 * ull_oct() 254 * convert an unsigned long long to an octal string. one of many oddball 255 * field termination characters are used by the various versions of tar 256 * in the different fields. term selects which kind to use. str is '0' 257 * padded at the front to len. we are unable to use only one format as 258 * many old tar readers are very cranky about this. 259 * Return: 260 * 0 if the number fit into the string, -1 otherwise 261 */ 262 263static int 264ull_oct(unsigned long long val, char *str, int len, int term) 265{ 266 char *pt; 267 268 /* 269 * term selects the appropriate character(s) for the end of the string 270 */ 271 pt = str + len - 1; 272 switch(term) { 273 case 3: 274 *pt-- = '\0'; 275 break; 276 case 2: 277 *pt-- = ' '; 278 *pt-- = '\0'; 279 break; 280 case 1: 281 *pt-- = ' '; 282 break; 283 case 0: 284 default: 285 *pt-- = '\0'; 286 *pt-- = ' '; 287 break; 288 } 289 290 /* 291 * convert and blank pad if there is space 292 */ 293 while (pt >= str) { 294 *pt-- = '0' + (char)(val & 0x7); 295 if ((val = val >> 3) == 0) 296 break; 297 } 298 299 while (pt >= str) 300 *pt-- = '0'; 301 if (val != (unsigned long long)0) 302 return(-1); 303 return(0); 304} 305#endif 306 307/* 308 * tar_chksm() 309 * calculate the checksum for a tar block counting the checksum field as 310 * all blanks (BLNKSUM is that value pre-calculated, the sum of 8 blanks). 311 * NOTE: we use len to short circuit summing 0's on write since we ALWAYS 312 * pad headers with 0. 313 * Return: 314 * unsigned long checksum 315 */ 316 317static u_long 318tar_chksm(char *blk, int len) 319{ 320 char *stop; 321 char *pt; 322 u_long chksm = BLNKSUM; /* initial 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 358int 359tar_id(char *blk, int size) 360{ 361 HD_TAR *hd; 362 HD_USTAR *uhd; 363 364 if (size < BLKMULT) 365 return(-1); 366 hd = (HD_TAR *)blk; 367 uhd = (HD_USTAR *)blk; 368 369 /* 370 * check for block of zero's first, a simple and fast test, then make 371 * sure this is not a ustar header by looking for the ustar magic 372 * cookie. We should use TMAGLEN, but some USTAR archive programs are 373 * wrong and create archives missing the \0. Last we check the 374 * checksum. If this is ok we have to assume it is a valid header. 375 */ 376 if (hd->name[0] == '\0') 377 return(-1); 378 if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0) 379 return(-1); 380 return check_sum(hd->chksum, sizeof(hd->chksum), blk, BLKMULT, 1); 381} 382 383/* 384 * tar_opt() 385 * handle tar format specific -o options 386 * Return: 387 * 0 if ok -1 otherwise 388 */ 389 390int 391tar_opt(void) 392{ 393 OPLIST *opt; 394 395 while ((opt = opt_next()) != NULL) { 396 if (strcmp(opt->name, TAR_OPTION) || 397 strcmp(opt->value, TAR_NODIR)) { 398 tty_warn(1, 399 "Unknown tar format -o option/value pair %s=%s", 400 opt->name, opt->value); 401 tty_warn(1, 402 "%s=%s is the only supported tar format option", 403 TAR_OPTION, TAR_NODIR); 404 return(-1); 405 } 406 407 /* 408 * we only support one option, and only when writing 409 */ 410 if ((act != APPND) && (act != ARCHIVE)) { 411 tty_warn(1, "%s=%s is only supported when writing.", 412 opt->name, opt->value); 413 return(-1); 414 } 415 tar_nodir = 1; 416 } 417 return(0); 418} 419 420 421/* 422 * tar_rd() 423 * extract the values out of block already determined to be a tar header. 424 * store the values in the ARCHD parameter. 425 * Return: 426 * 0 427 */ 428 429int 430tar_rd(ARCHD *arcn, char *buf) 431{ 432 HD_TAR *hd; 433 char *pt; 434 435 /* 436 * we only get proper sized buffers passed to us 437 */ 438 if (tar_id(buf, BLKMULT) < 0) 439 return(-1); 440 memset(arcn, 0, sizeof(*arcn)); 441 arcn->org_name = arcn->name; 442 arcn->pat = NULL; 443 arcn->sb.st_nlink = 1; 444 445 /* 446 * copy out the name and values in the stat buffer 447 */ 448 hd = (HD_TAR *)buf; 449 if (hd->linkflag != LONGLINKTYPE && hd->linkflag != LONGNAMETYPE) { 450 arcn->nlen = expandname(arcn->name, sizeof(arcn->name), 451 &gnu_name_string, hd->name, sizeof(hd->name)); 452 arcn->ln_nlen = expandname(arcn->ln_name, sizeof(arcn->ln_name), 453 &gnu_link_string, hd->linkname, sizeof(hd->linkname)); 454 } 455 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) & 456 0xfff); 457 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 458 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 459 arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT); 460 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 461 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 462 463 /* 464 * have to look at the last character, it may be a '/' and that is used 465 * to encode this as a directory 466 */ 467 pt = &(arcn->name[arcn->nlen - 1]); 468 arcn->pad = 0; 469 arcn->skip = 0; 470 switch(hd->linkflag) { 471 case SYMTYPE: 472 /* 473 * symbolic link, need to get the link name and set the type in 474 * the st_mode so -v printing will look correct. 475 */ 476 arcn->type = PAX_SLK; 477 arcn->sb.st_mode |= S_IFLNK; 478 break; 479 case LNKTYPE: 480 /* 481 * hard link, need to get the link name, set the type in the 482 * st_mode and st_nlink so -v printing will look better. 483 */ 484 arcn->type = PAX_HLK; 485 arcn->sb.st_nlink = 2; 486 487 /* 488 * no idea of what type this thing really points at, but 489 * we set something for printing only. 490 */ 491 arcn->sb.st_mode |= S_IFREG; 492 break; 493 case LONGLINKTYPE: 494 arcn->type = PAX_GLL; 495 /* FALLTHROUGH */ 496 case LONGNAMETYPE: 497 /* 498 * GNU long link/file; we tag these here and let the 499 * pax internals deal with it -- too ugly otherwise. 500 */ 501 if (hd->linkflag != LONGLINKTYPE) 502 arcn->type = PAX_GLF; 503 arcn->pad = TAR_PAD(arcn->sb.st_size); 504 arcn->skip = arcn->sb.st_size; 505 break; 506 case AREGTYPE: 507 case REGTYPE: 508 case DIRTYPE: /* see below */ 509 default: 510 /* 511 * If we have a trailing / this is a directory and NOT a file. 512 * Note: V7 tar doesn't actually have DIRTYPE, but it was 513 * reported that V7 archives using USTAR directories do exist. 514 */ 515 if (*pt == '/' || hd->linkflag == DIRTYPE) { 516 /* 517 * it is a directory, set the mode for -v printing 518 */ 519 arcn->type = PAX_DIR; 520 arcn->sb.st_mode |= S_IFDIR; 521 arcn->sb.st_nlink = 2; 522 } else { 523 /* 524 * have a file that will be followed by data. Set the 525 * skip value to the size field and calculate the size 526 * of the padding. 527 */ 528 arcn->type = PAX_REG; 529 arcn->sb.st_mode |= S_IFREG; 530 arcn->pad = TAR_PAD(arcn->sb.st_size); 531 arcn->skip = arcn->sb.st_size; 532 } 533 break; 534 } 535 536 /* 537 * strip off any trailing slash. 538 */ 539 if (*pt == '/') { 540 *pt = '\0'; 541 --arcn->nlen; 542 } 543 return(0); 544} 545 546/* 547 * tar_wr() 548 * write a tar header for the file specified in the ARCHD to the archive. 549 * Have to check for file types that cannot be stored and file names that 550 * are too long. Be careful of the term (last arg) to ul_oct, each field 551 * of tar has it own spec for the termination character(s). 552 * ASSUMED: space after header in header block is zero filled 553 * Return: 554 * 0 if file has data to be written after the header, 1 if file has NO 555 * data to write after the header, -1 if archive write failed 556 */ 557 558int 559tar_wr(ARCHD *arcn) 560{ 561 HD_TAR *hd; 562 int len; 563 char hdblk[sizeof(HD_TAR)]; 564 565 /* 566 * check for those file system types which tar cannot store 567 */ 568 switch(arcn->type) { 569 case PAX_DIR: 570 /* 571 * user asked that dirs not be written to the archive 572 */ 573 if (tar_nodir) 574 return(1); 575 break; 576 case PAX_CHR: 577 tty_warn(1, "Tar cannot archive a character device %s", 578 arcn->org_name); 579 return(1); 580 case PAX_BLK: 581 tty_warn(1, 582 "Tar cannot archive a block device %s", arcn->org_name); 583 return(1); 584 case PAX_SCK: 585 tty_warn(1, "Tar cannot archive a socket %s", arcn->org_name); 586 return(1); 587 case PAX_FIF: 588 tty_warn(1, "Tar cannot archive a fifo %s", arcn->org_name); 589 return(1); 590 case PAX_SLK: 591 case PAX_HLK: 592 case PAX_HRG: 593 if (arcn->ln_nlen > sizeof(hd->linkname)) { 594 tty_warn(1,"Link name too long for tar %s", 595 arcn->ln_name); 596 return(1); 597 } 598 break; 599 case PAX_REG: 600 case PAX_CTG: 601 default: 602 break; 603 } 604 605 /* 606 * check file name len, remember extra char for dirs (the / at the end) 607 */ 608 len = arcn->nlen; 609 if (arcn->type == PAX_DIR) 610 ++len; 611 if (len >= sizeof(hd->name)) { 612 tty_warn(1, "File name too long for tar %s", arcn->name); 613 return(1); 614 } 615 616 /* 617 * copy the data out of the ARCHD into the tar header based on the type 618 * of the file. Remember many tar readers want the unused fields to be 619 * padded with zero. We set the linkflag field (type), the linkname 620 * (or zero if not used),the size, and set the padding (if any) to be 621 * added after the file data (0 for all other types, as they only have 622 * a header) 623 */ 624 memset(hdblk, 0, sizeof(hdblk)); 625 hd = (HD_TAR *)hdblk; 626 strlcpy(hd->name, arcn->name, sizeof(hd->name)); 627 arcn->pad = 0; 628 629 if (arcn->type == PAX_DIR) { 630 /* 631 * directories are the same as files, except have a filename 632 * that ends with a /, we add the slash here. No data follows, 633 * dirs, so no pad. 634 */ 635 hd->linkflag = AREGTYPE; 636 hd->name[len-1] = '/'; 637 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 638 goto out; 639 } else if (arcn->type == PAX_SLK) { 640 /* 641 * no data follows this file, so no pad 642 */ 643 hd->linkflag = SYMTYPE; 644 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 645 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 646 goto out; 647 } else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) { 648 /* 649 * no data follows this file, so no pad 650 */ 651 hd->linkflag = LNKTYPE; 652 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 653 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 654 goto out; 655 } else { 656 /* 657 * data follows this file, so set the pad 658 */ 659 hd->linkflag = AREGTYPE; 660 if (OFFT_OCT(arcn->sb.st_size, hd->size, sizeof(hd->size), 1)) { 661 tty_warn(1,"File is too large for tar %s", 662 arcn->org_name); 663 return(1); 664 } 665 arcn->pad = TAR_PAD(arcn->sb.st_size); 666 } 667 668 /* 669 * copy those fields that are independent of the type 670 */ 671 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) || 672 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) || 673 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) || 674 ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1)) 675 goto out; 676 677 /* 678 * calculate and add the checksum, then write the header. A return of 679 * 0 tells the caller to now write the file data, 1 says no data needs 680 * to be written 681 */ 682 if (ul_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum, 683 sizeof(hd->chksum), 3)) 684 goto out; /* XXX Something's wrong here 685 * because a zero-byte file can 686 * cause this to be done and 687 * yet the resulting warning 688 * seems incorrect */ 689 690 if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0) 691 return(-1); 692 if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0) 693 return(-1); 694 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 695 return(0); 696 return(1); 697 698 out: 699 /* 700 * header field is out of range 701 */ 702 tty_warn(1, "Tar header field is too small for %s", arcn->org_name); 703 return(1); 704} 705 706/* 707 * Routines for POSIX ustar 708 */ 709 710/* 711 * ustar_strd() 712 * initialization for ustar read 713 * Return: 714 * 0 if ok, -1 otherwise 715 */ 716 717int 718ustar_strd(void) 719{ 720 return(0); 721} 722 723/* 724 * ustar_stwr() 725 * initialization for ustar write 726 * Return: 727 * 0 if ok, -1 otherwise 728 */ 729 730int 731ustar_stwr(void) 732{ 733 return(0); 734} 735 736/* 737 * ustar_id() 738 * determine if a block given to us is a valid ustar header. We have to 739 * be on the lookout for those pesky blocks of all zero's 740 * Return: 741 * 0 if a ustar header, -1 otherwise 742 */ 743 744int 745ustar_id(char *blk, int size) 746{ 747 HD_USTAR *hd; 748 749 if (size < BLKMULT) 750 return(-1); 751 hd = (HD_USTAR *)blk; 752 753 /* 754 * check for block of zero's first, a simple and fast test then check 755 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive 756 * programs are fouled up and create archives missing the \0. Last we 757 * check the checksum. If ok we have to assume it is a valid header. 758 */ 759 if (hd->name[0] == '\0') 760 return(-1); 761 if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0) 762 return(-1); 763 /* This is GNU tar */ 764 if (strncmp(hd->magic, "ustar ", 8) == 0 && !is_gnutar && 765 !seen_gnu_warning) { 766 seen_gnu_warning = 1; 767 tty_warn(0, 768 "Trying to read GNU tar archive with extensions off"); 769 } 770 return check_sum(hd->chksum, sizeof(hd->chksum), blk, BLKMULT, 0); 771} 772 773/* 774 * ustar_rd() 775 * extract the values out of block already determined to be a ustar header. 776 * store the values in the ARCHD parameter. 777 * Return: 778 * 0 779 */ 780 781int 782ustar_rd(ARCHD *arcn, char *buf) 783{ 784 HD_USTAR *hd; 785 char *dest; 786 int cnt; 787 dev_t devmajor; 788 dev_t devminor; 789 790 /* 791 * we only get proper sized buffers 792 */ 793 if (ustar_id(buf, BLKMULT) < 0) 794 return(-1); 795 796 memset(arcn, 0, sizeof(*arcn)); 797 arcn->org_name = arcn->name; 798 arcn->pat = NULL; 799 arcn->sb.st_nlink = 1; 800 hd = (HD_USTAR *)buf; 801 802 /* 803 * see if the filename is split into two parts. if, so joint the parts. 804 * we copy the prefix first and add a / between the prefix and name. 805 */ 806 dest = arcn->name; 807 if (*(hd->prefix) != '\0') { 808 cnt = strlcpy(arcn->name, hd->prefix, sizeof(arcn->name)); 809 dest += cnt; 810 *dest++ = '/'; 811 cnt++; 812 } else { 813 cnt = 0; 814 } 815 816 if (hd->typeflag != LONGLINKTYPE && hd->typeflag != LONGNAMETYPE) { 817 arcn->nlen = expandname(dest, sizeof(arcn->name) - cnt, 818 &gnu_name_string, hd->name, sizeof(hd->name)) + cnt; 819 arcn->ln_nlen = expandname(arcn->ln_name, 820 sizeof(arcn->ln_name), &gnu_link_string, hd->linkname, 821 sizeof(hd->linkname)); 822 } 823 824 /* 825 * follow the spec to the letter. we should only have mode bits, strip 826 * off all other crud we may be passed. 827 */ 828 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) & 829 0xfff); 830 arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT); 831 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 832 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 833 834 /* 835 * If we can find the ascii names for gname and uname in the password 836 * and group files we will use the uid's and gid they bind. Otherwise 837 * we use the uid and gid values stored in the header. (This is what 838 * the posix spec wants). 839 */ 840 hd->gname[sizeof(hd->gname) - 1] = '\0'; 841 if (gid_from_group(hd->gname, &(arcn->sb.st_gid)) < 0) 842 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 843 hd->uname[sizeof(hd->uname) - 1] = '\0'; 844 if (uid_from_user(hd->uname, &(arcn->sb.st_uid)) < 0) 845 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 846 847 /* 848 * set the defaults, these may be changed depending on the file type 849 */ 850 arcn->pad = 0; 851 arcn->skip = 0; 852 arcn->sb.st_rdev = (dev_t)0; 853 854 /* 855 * set the mode and PAX type according to the typeflag in the header 856 */ 857 switch(hd->typeflag) { 858 case FIFOTYPE: 859 arcn->type = PAX_FIF; 860 arcn->sb.st_mode |= S_IFIFO; 861 break; 862 case DIRTYPE: 863 arcn->type = PAX_DIR; 864 arcn->sb.st_mode |= S_IFDIR; 865 arcn->sb.st_nlink = 2; 866 867 /* 868 * Some programs that create ustar archives append a '/' 869 * to the pathname for directories. This clearly violates 870 * ustar specs, but we will silently strip it off anyway. 871 */ 872 if (arcn->name[arcn->nlen - 1] == '/') 873 arcn->name[--arcn->nlen] = '\0'; 874 break; 875 case BLKTYPE: 876 case CHRTYPE: 877 /* 878 * this type requires the rdev field to be set. 879 */ 880 if (hd->typeflag == BLKTYPE) { 881 arcn->type = PAX_BLK; 882 arcn->sb.st_mode |= S_IFBLK; 883 } else { 884 arcn->type = PAX_CHR; 885 arcn->sb.st_mode |= S_IFCHR; 886 } 887 devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT); 888 devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT); 889 arcn->sb.st_rdev = TODEV(devmajor, devminor); 890 break; 891 case SYMTYPE: 892 case LNKTYPE: 893 if (hd->typeflag == SYMTYPE) { 894 arcn->type = PAX_SLK; 895 arcn->sb.st_mode |= S_IFLNK; 896 } else { 897 arcn->type = PAX_HLK; 898 /* 899 * so printing looks better 900 */ 901 arcn->sb.st_mode |= S_IFREG; 902 arcn->sb.st_nlink = 2; 903 } 904 break; 905 case LONGLINKTYPE: 906 if (is_gnutar) 907 arcn->type = PAX_GLL; 908 /* FALLTHROUGH */ 909 case LONGNAMETYPE: 910 if (is_gnutar) { 911 /* 912 * GNU long link/file; we tag these here and let the 913 * pax internals deal with it -- too ugly otherwise. 914 */ 915 if (hd->typeflag != LONGLINKTYPE) 916 arcn->type = PAX_GLF; 917 arcn->pad = TAR_PAD(arcn->sb.st_size); 918 arcn->skip = arcn->sb.st_size; 919 } else { 920 tty_warn(1, "GNU Long %s found in posix ustar archive.", 921 hd->typeflag == LONGLINKTYPE ? "Link" : "File"); 922 } 923 break; 924 case CONTTYPE: 925 case AREGTYPE: 926 case REGTYPE: 927 default: 928 /* 929 * these types have file data that follows. Set the skip and 930 * pad fields. 931 */ 932 arcn->type = PAX_REG; 933 arcn->pad = TAR_PAD(arcn->sb.st_size); 934 arcn->skip = arcn->sb.st_size; 935 arcn->sb.st_mode |= S_IFREG; 936 break; 937 } 938 return(0); 939} 940 941static int 942expandname(char *buf, size_t len, char **gnu_name, const char *name, 943 size_t nlen) 944{ 945 if (*gnu_name) { 946 len = strlcpy(buf, *gnu_name, len); 947 free(*gnu_name); 948 *gnu_name = NULL; 949 } else { 950 if (len > ++nlen) 951 len = nlen; 952 len = strlcpy(buf, name, len); 953 } 954 return len; 955} 956 957static void 958longlink(ARCHD *arcn) 959{ 960 ARCHD larc; 961 962 memset(&larc, 0, sizeof(larc)); 963 964 switch (arcn->type) { 965 case PAX_SLK: 966 case PAX_HRG: 967 case PAX_HLK: 968 larc.type = PAX_GLL; 969 larc.ln_nlen = strlcpy(larc.ln_name, "././@LongLink", 970 sizeof(larc.ln_name)); 971 gnu_hack_string = arcn->ln_name; 972 gnu_hack_len = arcn->ln_nlen + 1; 973 break; 974 default: 975 larc.nlen = strlcpy(larc.name, "././@LongLink", 976 sizeof(larc.name)); 977 gnu_hack_string = arcn->name; 978 gnu_hack_len = arcn->nlen + 1; 979 larc.type = PAX_GLF; 980 } 981 /* 982 * We need a longlink now. 983 */ 984 ustar_wr(&larc); 985} 986 987/* 988 * ustar_wr() 989 * write a ustar header for the file specified in the ARCHD to the archive 990 * Have to check for file types that cannot be stored and file names that 991 * are too long. Be careful of the term (last arg) to ul_oct, we only use 992 * '\0' for the termination character (this is different than picky tar) 993 * ASSUMED: space after header in header block is zero filled 994 * Return: 995 * 0 if file has data to be written after the header, 1 if file has NO 996 * data to write after the header, -1 if archive write failed 997 */ 998 999int 1000ustar_wr(ARCHD *arcn) 1001{ 1002 HD_USTAR *hd; 1003 char *pt; 1004 char hdblk[sizeof(HD_USTAR)]; 1005 const char *user, *group; 1006 1007 /* 1008 * check for those file system types ustar cannot store 1009 */ 1010 if (arcn->type == PAX_SCK) { 1011 if (!is_gnutar) 1012 tty_warn(1, "Ustar cannot archive a socket %s", 1013 arcn->org_name); 1014 return(1); 1015 } 1016 1017 /* 1018 * check the length of the linkname 1019 */ 1020 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) || 1021 (arcn->type == PAX_HRG)) && 1022 (arcn->ln_nlen >= sizeof(hd->linkname))){ 1023 if (is_gnutar) { 1024 longlink(arcn); 1025 } else { 1026 tty_warn(1, "Link name too long for ustar %s", 1027 arcn->ln_name); 1028 return(1); 1029 } 1030 } 1031 1032 /* 1033 * split the path name into prefix and name fields (if needed). if 1034 * pt != arcn->name, the name has to be split 1035 */ 1036 if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) { 1037 if (is_gnutar) { 1038 longlink(arcn); 1039 pt = arcn->name; 1040 } else { 1041 tty_warn(1, "File name too long for ustar %s", 1042 arcn->name); 1043 return(1); 1044 } 1045 } 1046 1047 /* 1048 * zero out the header so we don't have to worry about zero fill below 1049 */ 1050 memset(hdblk, 0, sizeof(hdblk)); 1051 hd = (HD_USTAR *)hdblk; 1052 arcn->pad = 0L; 1053 1054 /* 1055 * split the name, or zero out the prefix 1056 */ 1057 if (pt != arcn->name) { 1058 /* 1059 * name was split, pt points at the / where the split is to 1060 * occur, we remove the / and copy the first part to the prefix 1061 */ 1062 *pt = '\0'; 1063 strlcpy(hd->prefix, arcn->name, sizeof(hd->prefix)); 1064 *pt++ = '/'; 1065 } 1066 1067 /* 1068 * copy the name part. this may be the whole path or the part after 1069 * the prefix 1070 */ 1071 strlcpy(hd->name, pt, sizeof(hd->name)); 1072 1073 /* 1074 * set the fields in the header that are type dependent 1075 */ 1076 switch(arcn->type) { 1077 case PAX_DIR: 1078 hd->typeflag = DIRTYPE; 1079 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1080 goto out; 1081 break; 1082 case PAX_CHR: 1083 case PAX_BLK: 1084 if (arcn->type == PAX_CHR) 1085 hd->typeflag = CHRTYPE; 1086 else 1087 hd->typeflag = BLKTYPE; 1088 if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor, 1089 sizeof(hd->devmajor), 3) || 1090 ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor, 1091 sizeof(hd->devminor), 3) || 1092 ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1093 goto out; 1094 break; 1095 case PAX_FIF: 1096 hd->typeflag = FIFOTYPE; 1097 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1098 goto out; 1099 break; 1100 case PAX_GLL: 1101 case PAX_SLK: 1102 case PAX_HLK: 1103 case PAX_HRG: 1104 if (arcn->type == PAX_SLK) 1105 hd->typeflag = SYMTYPE; 1106 else if (arcn->type == PAX_GLL) 1107 hd->typeflag = LONGLINKTYPE; 1108 else 1109 hd->typeflag = LNKTYPE; 1110 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 1111 if (ul_oct((u_long)gnu_hack_len, hd->size, 1112 sizeof(hd->size), 3)) 1113 goto out; 1114 break; 1115 case PAX_GLF: 1116 case PAX_REG: 1117 case PAX_CTG: 1118 default: 1119 /* 1120 * file data with this type, set the padding 1121 */ 1122 if (arcn->type == PAX_GLF) { 1123 hd->typeflag = LONGNAMETYPE; 1124 arcn->pad = TAR_PAD(gnu_hack_len); 1125 if (OFFT_OCT((u_long)gnu_hack_len, hd->size, 1126 sizeof(hd->size), 3)) { 1127 tty_warn(1,"File is too long for ustar %s", 1128 arcn->org_name); 1129 return(1); 1130 } 1131 } else { 1132 if (arcn->type == PAX_CTG) 1133 hd->typeflag = CONTTYPE; 1134 else 1135 hd->typeflag = REGTYPE; 1136 arcn->pad = TAR_PAD(arcn->sb.st_size); 1137 if (OFFT_OCT(arcn->sb.st_size, hd->size, 1138 sizeof(hd->size), 3)) { 1139 tty_warn(1,"File is too long for ustar %s", 1140 arcn->org_name); 1141 return(1); 1142 } 1143 } 1144 break; 1145 } 1146 1147 strncpy(hd->magic, TMAGIC, TMAGLEN); 1148 if (is_gnutar) 1149 hd->magic[TMAGLEN - 1] = hd->magic[TMAGLEN] = ' '; 1150 else 1151 strncpy(hd->version, TVERSION, TVERSLEN); 1152 1153 /* 1154 * set the remaining fields. Some versions want all 16 bits of mode 1155 * we better humor them (they really do not meet spec though).... 1156 */ 1157 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) || 1158 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3) || 1159 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) || 1160 ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3)) 1161 goto out; 1162 user = user_from_uid(arcn->sb.st_uid, 1); 1163 group = group_from_gid(arcn->sb.st_gid, 1); 1164 strncpy(hd->uname, user ? user : "", sizeof(hd->uname)); 1165 strncpy(hd->gname, group ? group : "", sizeof(hd->gname)); 1166 1167 /* 1168 * calculate and store the checksum write the header to the archive 1169 * return 0 tells the caller to now write the file data, 1 says no data 1170 * needs to be written 1171 */ 1172 if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum, 1173 sizeof(hd->chksum), 3)) 1174 goto out; 1175 if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0) 1176 return(-1); 1177 if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0) 1178 return(-1); 1179 if (gnu_hack_string) { 1180 int res = wr_rdbuf(gnu_hack_string, gnu_hack_len); 1181 int pad = gnu_hack_len; 1182 gnu_hack_string = NULL; 1183 gnu_hack_len = 0; 1184 if (res < 0) 1185 return(-1); 1186 if (wr_skip((off_t)(BLKMULT - (pad % BLKMULT))) < 0) 1187 return(-1); 1188 } 1189 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 1190 return(0); 1191 return(1); 1192 1193 out: 1194 /* 1195 * header field is out of range 1196 */ 1197 tty_warn(1, "Ustar header field is too small for %s", arcn->org_name); 1198 return(1); 1199} 1200 1201/* 1202 * name_split() 1203 * see if the name has to be split for storage in a ustar header. We try 1204 * to fit the entire name in the name field without splitting if we can. 1205 * The split point is always at a / 1206 * Return 1207 * character pointer to split point (always the / that is to be removed 1208 * if the split is not needed, the points is set to the start of the file 1209 * name (it would violate the spec to split there). A NULL is returned if 1210 * the file name is too long 1211 */ 1212 1213static char * 1214name_split(char *name, int len) 1215{ 1216 char *start; 1217 1218 /* 1219 * check to see if the file name is small enough to fit in the name 1220 * field. if so just return a pointer to the name. 1221 */ 1222 if (len < TNMSZ) 1223 return(name); 1224 if (len > (TPFSZ + TNMSZ)) 1225 return(NULL); 1226 1227 /* 1228 * we start looking at the biggest sized piece that fits in the name 1229 * field. We walk forward looking for a slash to split at. The idea is 1230 * to find the biggest piece to fit in the name field (or the smallest 1231 * prefix we can find) (the -1 is correct the biggest piece would 1232 * include the slash between the two parts that gets thrown away) 1233 */ 1234 start = name + len - TNMSZ; 1235 while ((*start != '\0') && (*start != '/')) 1236 ++start; 1237 1238 /* 1239 * if we hit the end of the string, this name cannot be split, so we 1240 * cannot store this file. 1241 */ 1242 if (*start == '\0') 1243 return(NULL); 1244 len = start - name; 1245 1246 /* 1247 * NOTE: /str where the length of str == TNMSZ cannot be stored under 1248 * the p1003.1-1990 spec for ustar. We could force a prefix of / and 1249 * the file would then expand on extract to //str. The len == 0 below 1250 * makes this special case follow the spec to the letter. 1251 */ 1252 if ((len >= TPFSZ) || (len == 0)) 1253 return(NULL); 1254 1255 /* 1256 * ok have a split point, return it to the caller 1257 */ 1258 return(start); 1259} 1260 1261/* 1262 * convert a glob into a RE, and add it to the list. we convert to 1263 * four different RE's (because we're using BRE's and can't use | 1264 * alternation :-() with this padding: 1265 * .*\/ and $ 1266 * .*\/ and \/.* 1267 * ^ and $ 1268 * ^ and \/.* 1269 */ 1270static int 1271tar_gnutar_exclude_one(const char *line, size_t len) 1272{ 1273 /* 2 * buffer len + nul */ 1274 char sbuf[MAXPATHLEN * 2 + 1]; 1275 /* + / + // + .*""/\/ + \/.* */ 1276 char rabuf[MAXPATHLEN * 2 + 1 + 1 + 2 + 4 + 4]; 1277 int i, j; 1278 1279 if (line[len - 1] == '\n') 1280 len--; 1281 strncpy(sbuf, ".*" "\\/", j = 4); 1282 for (i = 0; i < len; i++) { 1283 /* 1284 * convert glob to regexp, escaping everything 1285 */ 1286 if (line[i] == '*') 1287 sbuf[j++] = '.'; 1288 else if (line[i] == '?') { 1289 sbuf[j++] = '.'; 1290 continue; 1291 } else if (!isalnum(line[i]) && !isblank(line[i])) 1292 sbuf[j++] = '\\'; 1293 sbuf[j++] = line[i]; 1294 } 1295 /* don't need the .*\/ ones if we start with /, i guess */ 1296 if (line[0] != '/') { 1297 (void)snprintf(rabuf, sizeof rabuf, "/.*\\/%s$//", sbuf); 1298 if (rep_add(rabuf) < 0) 1299 return (-1); 1300 (void)snprintf(rabuf, sizeof rabuf, "/.*\\/%s\\/.*//", sbuf); 1301 if (rep_add(rabuf) < 0) 1302 return (-1); 1303 } 1304 1305 (void)snprintf(rabuf, sizeof rabuf, "/^%s$//", sbuf); 1306 if (rep_add(rabuf) < 0) 1307 return (-1); 1308 (void)snprintf(rabuf, sizeof rabuf, "/^%s\\/.*//", sbuf); 1309 if (rep_add(rabuf) < 0) 1310 return (-1); 1311 1312 return (0); 1313} 1314 1315/* 1316 * deal with GNU tar -X/--exclude-from & --exclude switchs. basically, 1317 * we go through each line of the file, building a string from the "glob" 1318 * lines in the file into RE lines, of the form `/^RE$//', which we pass 1319 * to rep_add(), which will add a empty replacement (exclusion), for the 1320 * named files. 1321 */ 1322int 1323tar_gnutar_minus_minus_exclude(path) 1324 const char *path; 1325{ 1326 size_t len = strlen(path); 1327 1328 if (len > MAXPATHLEN) 1329 tty_warn(0, "pathname too long: %s", path); 1330 1331 return (tar_gnutar_exclude_one(path, len)); 1332} 1333 1334int 1335tar_gnutar_X_compat(path) 1336 const char *path; 1337{ 1338 char *line; 1339 FILE *fp; 1340 int lineno = 0; 1341 size_t len; 1342 1343 fp = fopen(path, "r"); 1344 if (fp == NULL) { 1345 tty_warn(1, "cannot open %s: %s", path, 1346 strerror(errno)); 1347 return(-1); 1348 } 1349 1350 while ((line = fgetln(fp, &len))) { 1351 lineno++; 1352 if (len > MAXPATHLEN) { 1353 tty_warn(0, "pathname too long, line %d of %s", 1354 lineno, path); 1355 } 1356 if (tar_gnutar_exclude_one(line, len)) 1357 return (-1); 1358 } 1359 return (0); 1360} 1361