tar.c revision 1.32
1/* $NetBSD: tar.c,v 1.32 2002/10/27 20:48:15 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. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 */ 39 40#include <sys/cdefs.h> 41#if defined(__RCSID) && !defined(lint) 42#if 0 43static char sccsid[] = "@(#)tar.c 8.2 (Berkeley) 4/18/94"; 44#else 45__RCSID("$NetBSD: tar.c,v 1.32 2002/10/27 20:48:15 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 *); 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 79 80/* 81 * Routines common to all versions of tar 82 */ 83 84static int tar_nodir; /* do not write dirs under old tar */ 85int is_gnutar; /* behave like gnu tar; enable gnu 86 * extensions and skip end-ofvolume 87 * checks 88 */ 89static int seen_gnu_warning; /* Have we warned yet? */ 90static char *gnu_hack_string; /* ././@LongLink hackery */ 91static int gnu_hack_len; /* len of gnu_hack_string */ 92char *gnu_name_string; /* ././@LongLink hackery name */ 93char *gnu_link_string; /* ././@LongLink hackery link */ 94 95/* 96 * tar_endwr() 97 * add the tar trailer of two null blocks 98 * Return: 99 * 0 if ok, -1 otherwise (what wr_skip returns) 100 */ 101 102int 103tar_endwr(void) 104{ 105 return(wr_skip((off_t)(NULLCNT*BLKMULT))); 106} 107 108/* 109 * tar_endrd() 110 * no cleanup needed here, just return size of trailer (for append) 111 * Return: 112 * size of trailer (2 * BLKMULT) 113 */ 114 115off_t 116tar_endrd(void) 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 132int 133tar_trail(char *buf, int in_resync, int *cnt) 134{ 135 int i; 136 137 /* 138 * look for all zero, trailer is two consecutive blocks of zero 139 */ 140 for (i = 0; i < BLKMULT; ++i) { 141 if (buf[i] != '\0') 142 break; 143 } 144 145 /* 146 * if not all zero it is not a trailer, but MIGHT be a header. 147 */ 148 if (i != BLKMULT) 149 return(-1); 150 151 /* 152 * When given a zero block, we must be careful! 153 * If we are not in resync mode, check for the trailer. Have to watch 154 * out that we do not mis-identify file data as the trailer, so we do 155 * NOT try to id a trailer during resync mode. During resync mode we 156 * might as well throw this block out since a valid header can NEVER be 157 * a block of all 0 (we must have a valid file name). 158 */ 159 if (!in_resync && (++*cnt >= NULLCNT)) 160 return(0); 161 return(1); 162} 163 164/* 165 * ul_oct() 166 * convert an unsigned long to an octal string. many oddball field 167 * termination characters are used by the various versions of tar in the 168 * different fields. term selects which kind to use. str is '0' padded 169 * at the front to len. we are unable to use only one format as many old 170 * tar readers are very cranky about this. 171 * Return: 172 * 0 if the number fit into the string, -1 otherwise 173 */ 174 175static int 176ul_oct(u_long val, char *str, int len, int term) 177{ 178 char *pt; 179 180 /* 181 * term selects the appropriate character(s) for the end of the string 182 */ 183 pt = str + len - 1; 184 switch(term) { 185 case 3: 186 *pt-- = '\0'; 187 break; 188 case 2: 189 *pt-- = ' '; 190 *pt-- = '\0'; 191 break; 192 case 1: 193 *pt-- = ' '; 194 break; 195 case 0: 196 default: 197 *pt-- = '\0'; 198 *pt-- = ' '; 199 break; 200 } 201 202 /* 203 * convert and blank pad if there is space 204 */ 205 while (pt >= str) { 206 *pt-- = '0' + (char)(val & 0x7); 207 if ((val = val >> 3) == (u_long)0) 208 break; 209 } 210 211 while (pt >= str) 212 *pt-- = '0'; 213 if (val != (u_long)0) 214 return(-1); 215 return(0); 216} 217 218#if !defined(NET2_STAT) && !defined(_LP64) 219/* 220 * ull_oct() 221 * convert an unsigned long long to an octal string. one of many oddball 222 * field termination characters are used by the various versions of tar 223 * in the different fields. term selects which kind to use. str is '0' 224 * padded at the front to len. we are unable to use only one format as 225 * many old tar readers are very cranky about this. 226 * Return: 227 * 0 if the number fit into the string, -1 otherwise 228 */ 229 230static int 231ull_oct(unsigned long long val, char *str, int len, int term) 232{ 233 char *pt; 234 235 /* 236 * term selects the appropriate character(s) for the end of the string 237 */ 238 pt = str + len - 1; 239 switch(term) { 240 case 3: 241 *pt-- = '\0'; 242 break; 243 case 2: 244 *pt-- = ' '; 245 *pt-- = '\0'; 246 break; 247 case 1: 248 *pt-- = ' '; 249 break; 250 case 0: 251 default: 252 *pt-- = '\0'; 253 *pt-- = ' '; 254 break; 255 } 256 257 /* 258 * convert and blank pad if there is space 259 */ 260 while (pt >= str) { 261 *pt-- = '0' + (char)(val & 0x7); 262 if ((val = val >> 3) == 0) 263 break; 264 } 265 266 while (pt >= str) 267 *pt-- = '0'; 268 if (val != (unsigned long long)0) 269 return(-1); 270 return(0); 271} 272#endif 273 274/* 275 * tar_chksm() 276 * calculate the checksum for a tar block counting the checksum field as 277 * all blanks (BLNKSUM is that value pre-calculated, the sum of 8 blanks). 278 * NOTE: we use len to short circuit summing 0's on write since we ALWAYS 279 * pad headers with 0. 280 * Return: 281 * unsigned long checksum 282 */ 283 284static u_long 285tar_chksm(char *blk, int len) 286{ 287 char *stop; 288 char *pt; 289 u_long chksm = BLNKSUM; /* initial value is checksum field sum */ 290 291 /* 292 * add the part of the block before the checksum field 293 */ 294 pt = blk; 295 stop = blk + CHK_OFFSET; 296 while (pt < stop) 297 chksm += (u_long)(*pt++ & 0xff); 298 /* 299 * move past the checksum field and keep going, spec counts the 300 * checksum field as the sum of 8 blanks (which is pre-computed as 301 * BLNKSUM). 302 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding 303 * starts, no point in summing zero's) 304 */ 305 pt += CHK_LEN; 306 stop = blk + len; 307 while (pt < stop) 308 chksm += (u_long)(*pt++ & 0xff); 309 return(chksm); 310} 311 312/* 313 * Routines for old BSD style tar (also made portable to sysV tar) 314 */ 315 316/* 317 * tar_id() 318 * determine if a block given to us is a valid tar header (and not a USTAR 319 * header). We have to be on the lookout for those pesky blocks of all 320 * zero's. 321 * Return: 322 * 0 if a tar header, -1 otherwise 323 */ 324 325int 326tar_id(char *blk, int size) 327{ 328 HD_TAR *hd; 329 HD_USTAR *uhd; 330 331 if (size < BLKMULT) 332 return(-1); 333 hd = (HD_TAR *)blk; 334 uhd = (HD_USTAR *)blk; 335 336 /* 337 * check for block of zero's first, a simple and fast test, then make 338 * sure this is not a ustar header by looking for the ustar magic 339 * cookie. We should use TMAGLEN, but some USTAR archive programs are 340 * wrong and create archives missing the \0. Last we check the 341 * checksum. If this is ok we have to assume it is a valid header. 342 */ 343 if (hd->name[0] == '\0') 344 return(-1); 345 if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0) 346 return(-1); 347 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 348 return(-1); 349 return(0); 350} 351 352/* 353 * tar_opt() 354 * handle tar format specific -o options 355 * Return: 356 * 0 if ok -1 otherwise 357 */ 358 359int 360tar_opt(void) 361{ 362 OPLIST *opt; 363 364 while ((opt = opt_next()) != NULL) { 365 if (strcmp(opt->name, TAR_OPTION) || 366 strcmp(opt->value, TAR_NODIR)) { 367 tty_warn(1, 368 "Unknown tar format -o option/value pair %s=%s", 369 opt->name, opt->value); 370 tty_warn(1, 371 "%s=%s is the only supported tar format option", 372 TAR_OPTION, TAR_NODIR); 373 return(-1); 374 } 375 376 /* 377 * we only support one option, and only when writing 378 */ 379 if ((act != APPND) && (act != ARCHIVE)) { 380 tty_warn(1, "%s=%s is only supported when writing.", 381 opt->name, opt->value); 382 return(-1); 383 } 384 tar_nodir = 1; 385 } 386 return(0); 387} 388 389 390/* 391 * tar_rd() 392 * extract the values out of block already determined to be a tar header. 393 * store the values in the ARCHD parameter. 394 * Return: 395 * 0 396 */ 397 398int 399tar_rd(ARCHD *arcn, char *buf) 400{ 401 HD_TAR *hd; 402 char *pt; 403 404 /* 405 * we only get proper sized buffers passed to us 406 */ 407 if (tar_id(buf, BLKMULT) < 0) 408 return(-1); 409 memset(arcn, 0, sizeof(*arcn)); 410 arcn->org_name = arcn->name; 411 arcn->pat = NULL; 412 arcn->sb.st_nlink = 1; 413 414 /* 415 * copy out the name and values in the stat buffer 416 */ 417 hd = (HD_TAR *)buf; 418 if (hd->linkflag != LONGLINKTYPE && hd->linkflag != LONGNAMETYPE) { 419 arcn->nlen = expandname(arcn->name, sizeof(arcn->name), 420 &gnu_name_string, hd->name); 421 arcn->ln_nlen = expandname(arcn->ln_name, sizeof(arcn->ln_name), 422 &gnu_link_string, hd->linkname); 423 } 424 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) & 425 0xfff); 426 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 427 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 428 arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT); 429 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 430 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 431 432 /* 433 * have to look at the last character, it may be a '/' and that is used 434 * to encode this as a directory 435 */ 436 pt = &(arcn->name[arcn->nlen - 1]); 437 arcn->pad = 0; 438 arcn->skip = 0; 439 switch(hd->linkflag) { 440 case SYMTYPE: 441 /* 442 * symbolic link, need to get the link name and set the type in 443 * the st_mode so -v printing will look correct. 444 */ 445 arcn->type = PAX_SLK; 446 arcn->sb.st_mode |= S_IFLNK; 447 break; 448 case LNKTYPE: 449 /* 450 * hard link, need to get the link name, set the type in the 451 * st_mode and st_nlink so -v printing will look better. 452 */ 453 arcn->type = PAX_HLK; 454 arcn->sb.st_nlink = 2; 455 456 /* 457 * no idea of what type this thing really points at, but 458 * we set something for printing only. 459 */ 460 arcn->sb.st_mode |= S_IFREG; 461 break; 462 case LONGLINKTYPE: 463 arcn->type = PAX_GLL; 464 /* FALLTHROUGH */ 465 case LONGNAMETYPE: 466 /* 467 * GNU long link/file; we tag these here and let the 468 * pax internals deal with it -- too ugly otherwise. 469 */ 470 if (hd->linkflag != LONGLINKTYPE) 471 arcn->type = PAX_GLF; 472 arcn->pad = TAR_PAD(arcn->sb.st_size); 473 arcn->skip = arcn->sb.st_size; 474 break; 475 case AREGTYPE: 476 case REGTYPE: 477 case DIRTYPE: /* see below */ 478 default: 479 /* 480 * If we have a trailing / this is a directory and NOT a file. 481 * Note: V7 tar doesn't actually have DIRTYPE, but it was 482 * reported that V7 archives using USTAR directories do exist. 483 */ 484 if (*pt == '/' || hd->linkflag == DIRTYPE) { 485 /* 486 * it is a directory, set the mode for -v printing 487 */ 488 arcn->type = PAX_DIR; 489 arcn->sb.st_mode |= S_IFDIR; 490 arcn->sb.st_nlink = 2; 491 } else { 492 /* 493 * have a file that will be followed by data. Set the 494 * skip value to the size field and calculate the size 495 * of the padding. 496 */ 497 arcn->type = PAX_REG; 498 arcn->sb.st_mode |= S_IFREG; 499 arcn->pad = TAR_PAD(arcn->sb.st_size); 500 arcn->skip = arcn->sb.st_size; 501 } 502 break; 503 } 504 505 /* 506 * strip off any trailing slash. 507 */ 508 if (*pt == '/') { 509 *pt = '\0'; 510 --arcn->nlen; 511 } 512 return(0); 513} 514 515/* 516 * tar_wr() 517 * write a tar header for the file specified in the ARCHD to the archive. 518 * Have to check for file types that cannot be stored and file names that 519 * are too long. Be careful of the term (last arg) to ul_oct, each field 520 * of tar has it own spec for the termination character(s). 521 * ASSUMED: space after header in header block is zero filled 522 * Return: 523 * 0 if file has data to be written after the header, 1 if file has NO 524 * data to write after the header, -1 if archive write failed 525 */ 526 527int 528tar_wr(ARCHD *arcn) 529{ 530 HD_TAR *hd; 531 int len; 532 char hdblk[sizeof(HD_TAR)]; 533 534 /* 535 * check for those file system types which tar cannot store 536 */ 537 switch(arcn->type) { 538 case PAX_DIR: 539 /* 540 * user asked that dirs not be written to the archive 541 */ 542 if (tar_nodir) 543 return(1); 544 break; 545 case PAX_CHR: 546 tty_warn(1, "Tar cannot archive a character device %s", 547 arcn->org_name); 548 return(1); 549 case PAX_BLK: 550 tty_warn(1, 551 "Tar cannot archive a block device %s", arcn->org_name); 552 return(1); 553 case PAX_SCK: 554 tty_warn(1, "Tar cannot archive a socket %s", arcn->org_name); 555 return(1); 556 case PAX_FIF: 557 tty_warn(1, "Tar cannot archive a fifo %s", arcn->org_name); 558 return(1); 559 case PAX_SLK: 560 case PAX_HLK: 561 case PAX_HRG: 562 if (arcn->ln_nlen > sizeof(hd->linkname)) { 563 tty_warn(1,"Link name too long for tar %s", 564 arcn->ln_name); 565 return(1); 566 } 567 break; 568 case PAX_REG: 569 case PAX_CTG: 570 default: 571 break; 572 } 573 574 /* 575 * check file name len, remember extra char for dirs (the / at the end) 576 */ 577 len = arcn->nlen; 578 if (arcn->type == PAX_DIR) 579 ++len; 580 if (len >= sizeof(hd->name)) { 581 tty_warn(1, "File name too long for tar %s", arcn->name); 582 return(1); 583 } 584 585 /* 586 * copy the data out of the ARCHD into the tar header based on the type 587 * of the file. Remember many tar readers want the unused fields to be 588 * padded with zero. We set the linkflag field (type), the linkname 589 * (or zero if not used),the size, and set the padding (if any) to be 590 * added after the file data (0 for all other types, as they only have 591 * a header) 592 */ 593 memset(hdblk, 0, sizeof(hdblk)); 594 hd = (HD_TAR *)hdblk; 595 strlcpy(hd->name, arcn->name, sizeof(hd->name)); 596 arcn->pad = 0; 597 598 if (arcn->type == PAX_DIR) { 599 /* 600 * directories are the same as files, except have a filename 601 * that ends with a /, we add the slash here. No data follows, 602 * dirs, so no pad. 603 */ 604 hd->linkflag = AREGTYPE; 605 hd->name[len-1] = '/'; 606 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 607 goto out; 608 } else if (arcn->type == PAX_SLK) { 609 /* 610 * no data follows this file, so no pad 611 */ 612 hd->linkflag = SYMTYPE; 613 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 614 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 615 goto out; 616 } else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) { 617 /* 618 * no data follows this file, so no pad 619 */ 620 hd->linkflag = LNKTYPE; 621 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 622 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 623 goto out; 624 } else { 625 /* 626 * data follows this file, so set the pad 627 */ 628 hd->linkflag = AREGTYPE; 629 if (OFFT_OCT(arcn->sb.st_size, hd->size, sizeof(hd->size), 1)) { 630 tty_warn(1,"File is too large for tar %s", 631 arcn->org_name); 632 return(1); 633 } 634 arcn->pad = TAR_PAD(arcn->sb.st_size); 635 } 636 637 /* 638 * copy those fields that are independent of the type 639 */ 640 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) || 641 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) || 642 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) || 643 ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1)) 644 goto out; 645 646 /* 647 * calculate and add the checksum, then write the header. A return of 648 * 0 tells the caller to now write the file data, 1 says no data needs 649 * to be written 650 */ 651 if (ul_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum, 652 sizeof(hd->chksum), 3)) 653 goto out; /* XXX Something's wrong here 654 * because a zero-byte file can 655 * cause this to be done and 656 * yet the resulting warning 657 * seems incorrect */ 658 659 if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0) 660 return(-1); 661 if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0) 662 return(-1); 663 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 664 return(0); 665 return(1); 666 667 out: 668 /* 669 * header field is out of range 670 */ 671 tty_warn(1, "Tar header field is too small for %s", arcn->org_name); 672 return(1); 673} 674 675/* 676 * Routines for POSIX ustar 677 */ 678 679/* 680 * ustar_strd() 681 * initialization for ustar read 682 * Return: 683 * 0 if ok, -1 otherwise 684 */ 685 686int 687ustar_strd(void) 688{ 689 return(0); 690} 691 692/* 693 * ustar_stwr() 694 * initialization for ustar write 695 * Return: 696 * 0 if ok, -1 otherwise 697 */ 698 699int 700ustar_stwr(void) 701{ 702 return(0); 703} 704 705/* 706 * ustar_id() 707 * determine if a block given to us is a valid ustar header. We have to 708 * be on the lookout for those pesky blocks of all zero's 709 * Return: 710 * 0 if a ustar header, -1 otherwise 711 */ 712 713int 714ustar_id(char *blk, int size) 715{ 716 HD_USTAR *hd; 717 718 if (size < BLKMULT) 719 return(-1); 720 hd = (HD_USTAR *)blk; 721 722 /* 723 * check for block of zero's first, a simple and fast test then check 724 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive 725 * programs are fouled up and create archives missing the \0. Last we 726 * check the checksum. If ok we have to assume it is a valid header. 727 */ 728 if (hd->name[0] == '\0') 729 return(-1); 730 if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0) 731 return(-1); 732 /* This is GNU tar */ 733 if (strncmp(hd->magic, "ustar ", 8) == 0 && !is_gnutar && 734 !seen_gnu_warning) { 735 seen_gnu_warning = 1; 736 tty_warn(0, 737 "Trying to read GNU tar archive with extensions off"); 738 } 739 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 740 return(-1); 741 return(0); 742} 743 744/* 745 * ustar_rd() 746 * extract the values out of block already determined to be a ustar header. 747 * store the values in the ARCHD parameter. 748 * Return: 749 * 0 750 */ 751 752int 753ustar_rd(ARCHD *arcn, char *buf) 754{ 755 HD_USTAR *hd; 756 char *dest; 757 int cnt; 758 dev_t devmajor; 759 dev_t devminor; 760 761 /* 762 * we only get proper sized buffers 763 */ 764 if (ustar_id(buf, BLKMULT) < 0) 765 return(-1); 766 767 memset(arcn, 0, sizeof(*arcn)); 768 arcn->org_name = arcn->name; 769 arcn->pat = NULL; 770 arcn->sb.st_nlink = 1; 771 hd = (HD_USTAR *)buf; 772 773 /* 774 * see if the filename is split into two parts. if, so joint the parts. 775 * we copy the prefix first and add a / between the prefix and name. 776 */ 777 dest = arcn->name; 778 if (*(hd->prefix) != '\0') { 779 cnt = strlcpy(arcn->name, hd->prefix, sizeof(arcn->name)); 780 dest += cnt; 781 *dest++ = '/'; 782 cnt++; 783 } else { 784 cnt = 0; 785 } 786 787 if (hd->typeflag != LONGLINKTYPE && hd->typeflag != LONGNAMETYPE) { 788 arcn->nlen = expandname(dest, sizeof(arcn->name) - cnt, 789 &gnu_name_string, hd->name); 790 arcn->ln_nlen = expandname(arcn->ln_name, sizeof(arcn->ln_name), 791 &gnu_link_string, hd->linkname); 792 } 793 794 /* 795 * follow the spec to the letter. we should only have mode bits, strip 796 * off all other crud we may be passed. 797 */ 798 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) & 799 0xfff); 800 arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT); 801 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 802 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 803 804 /* 805 * If we can find the ascii names for gname and uname in the password 806 * and group files we will use the uid's and gid they bind. Otherwise 807 * we use the uid and gid values stored in the header. (This is what 808 * the posix spec wants). 809 */ 810 hd->gname[sizeof(hd->gname) - 1] = '\0'; 811 if (gid_from_group(hd->gname, &(arcn->sb.st_gid)) < 0) 812 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 813 hd->uname[sizeof(hd->uname) - 1] = '\0'; 814 if (uid_from_user(hd->uname, &(arcn->sb.st_uid)) < 0) 815 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 816 817 /* 818 * set the defaults, these may be changed depending on the file type 819 */ 820 arcn->pad = 0; 821 arcn->skip = 0; 822 arcn->sb.st_rdev = (dev_t)0; 823 824 /* 825 * set the mode and PAX type according to the typeflag in the header 826 */ 827 switch(hd->typeflag) { 828 case FIFOTYPE: 829 arcn->type = PAX_FIF; 830 arcn->sb.st_mode |= S_IFIFO; 831 break; 832 case DIRTYPE: 833 arcn->type = PAX_DIR; 834 arcn->sb.st_mode |= S_IFDIR; 835 arcn->sb.st_nlink = 2; 836 837 /* 838 * Some programs that create ustar archives append a '/' 839 * to the pathname for directories. This clearly violates 840 * ustar specs, but we will silently strip it off anyway. 841 */ 842 if (arcn->name[arcn->nlen - 1] == '/') 843 arcn->name[--arcn->nlen] = '\0'; 844 break; 845 case BLKTYPE: 846 case CHRTYPE: 847 /* 848 * this type requires the rdev field to be set. 849 */ 850 if (hd->typeflag == BLKTYPE) { 851 arcn->type = PAX_BLK; 852 arcn->sb.st_mode |= S_IFBLK; 853 } else { 854 arcn->type = PAX_CHR; 855 arcn->sb.st_mode |= S_IFCHR; 856 } 857 devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT); 858 devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT); 859 arcn->sb.st_rdev = TODEV(devmajor, devminor); 860 break; 861 case SYMTYPE: 862 case LNKTYPE: 863 if (hd->typeflag == SYMTYPE) { 864 arcn->type = PAX_SLK; 865 arcn->sb.st_mode |= S_IFLNK; 866 } else { 867 arcn->type = PAX_HLK; 868 /* 869 * so printing looks better 870 */ 871 arcn->sb.st_mode |= S_IFREG; 872 arcn->sb.st_nlink = 2; 873 } 874 break; 875 case LONGLINKTYPE: 876 if (is_gnutar) 877 arcn->type = PAX_GLL; 878 /* FALLTHROUGH */ 879 case LONGNAMETYPE: 880 if (is_gnutar) { 881 /* 882 * GNU long link/file; we tag these here and let the 883 * pax internals deal with it -- too ugly otherwise. 884 */ 885 if (hd->typeflag != LONGLINKTYPE) 886 arcn->type = PAX_GLF; 887 arcn->pad = TAR_PAD(arcn->sb.st_size); 888 arcn->skip = arcn->sb.st_size; 889 } else { 890 tty_warn(1, "GNU Long %s found in posix ustar archive.", 891 hd->typeflag == LONGLINKTYPE ? "Link" : "File"); 892 } 893 break; 894 case CONTTYPE: 895 case AREGTYPE: 896 case REGTYPE: 897 default: 898 /* 899 * these types have file data that follows. Set the skip and 900 * pad fields. 901 */ 902 arcn->type = PAX_REG; 903 arcn->pad = TAR_PAD(arcn->sb.st_size); 904 arcn->skip = arcn->sb.st_size; 905 arcn->sb.st_mode |= S_IFREG; 906 break; 907 } 908 return(0); 909} 910 911static int 912expandname(char *buf, size_t len, char **gnu_name, const char *name) 913{ 914 if (*gnu_name) { 915 len = strlcpy(buf, *gnu_name, len); 916 free(*gnu_name); 917 *gnu_name = NULL; 918 } else { 919 len = strlcpy(buf, name, len); 920 } 921 return len; 922} 923 924static void 925longlink(ARCHD *arcn) 926{ 927 ARCHD larc; 928 929 memset(&larc, 0, sizeof(larc)); 930 931 switch (arcn->type) { 932 case PAX_SLK: 933 case PAX_HRG: 934 case PAX_HLK: 935 larc.type = PAX_GLL; 936 larc.ln_nlen = strlcpy(larc.ln_name, "././@LongLink", 937 sizeof(larc.ln_name)); 938 gnu_hack_string = arcn->ln_name; 939 gnu_hack_len = arcn->ln_nlen + 1; 940 break; 941 default: 942 larc.nlen = strlcpy(larc.name, "././@LongLink", 943 sizeof(larc.name)); 944 gnu_hack_string = arcn->name; 945 gnu_hack_len = arcn->nlen + 1; 946 larc.type = PAX_GLF; 947 } 948 /* 949 * We need a longlink now. 950 */ 951 ustar_wr(&larc); 952} 953 954/* 955 * ustar_wr() 956 * write a ustar header for the file specified in the ARCHD to the archive 957 * Have to check for file types that cannot be stored and file names that 958 * are too long. Be careful of the term (last arg) to ul_oct, we only use 959 * '\0' for the termination character (this is different than picky tar) 960 * ASSUMED: space after header in header block is zero filled 961 * Return: 962 * 0 if file has data to be written after the header, 1 if file has NO 963 * data to write after the header, -1 if archive write failed 964 */ 965 966int 967ustar_wr(ARCHD *arcn) 968{ 969 HD_USTAR *hd; 970 char *pt; 971 char hdblk[sizeof(HD_USTAR)]; 972 const char *user, *group; 973 974 /* 975 * check for those file system types ustar cannot store 976 */ 977 if (arcn->type == PAX_SCK) { 978 tty_warn(1, "Ustar cannot archive a socket %s", arcn->org_name); 979 return(1); 980 } 981 982 /* 983 * check the length of the linkname 984 */ 985 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) || 986 (arcn->type == PAX_HRG)) && 987 (arcn->ln_nlen >= sizeof(hd->linkname))){ 988 if (is_gnutar) { 989 longlink(arcn); 990 } else { 991 tty_warn(1, "Link name too long for ustar %s", 992 arcn->ln_name); 993 return(1); 994 } 995 } 996 997 /* 998 * split the path name into prefix and name fields (if needed). if 999 * pt != arcn->name, the name has to be split 1000 */ 1001 if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) { 1002 if (is_gnutar) { 1003 longlink(arcn); 1004 pt = arcn->name; 1005 } else { 1006 tty_warn(1, "File name too long for ustar %s", 1007 arcn->name); 1008 return(1); 1009 } 1010 } 1011 1012 /* 1013 * zero out the header so we don't have to worry about zero fill below 1014 */ 1015 memset(hdblk, 0, sizeof(hdblk)); 1016 hd = (HD_USTAR *)hdblk; 1017 arcn->pad = 0L; 1018 1019 /* 1020 * split the name, or zero out the prefix 1021 */ 1022 if (pt != arcn->name) { 1023 /* 1024 * name was split, pt points at the / where the split is to 1025 * occur, we remove the / and copy the first part to the prefix 1026 */ 1027 *pt = '\0'; 1028 strlcpy(hd->prefix, arcn->name, sizeof(hd->prefix)); 1029 *pt++ = '/'; 1030 } 1031 1032 /* 1033 * copy the name part. this may be the whole path or the part after 1034 * the prefix 1035 */ 1036 strlcpy(hd->name, pt, sizeof(hd->name)); 1037 1038 /* 1039 * set the fields in the header that are type dependent 1040 */ 1041 switch(arcn->type) { 1042 case PAX_DIR: 1043 hd->typeflag = DIRTYPE; 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 if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor, 1054 sizeof(hd->devmajor), 3) || 1055 ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor, 1056 sizeof(hd->devminor), 3) || 1057 ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1058 goto out; 1059 break; 1060 case PAX_FIF: 1061 hd->typeflag = FIFOTYPE; 1062 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 1063 goto out; 1064 break; 1065 case PAX_GLL: 1066 case PAX_SLK: 1067 case PAX_HLK: 1068 case PAX_HRG: 1069 if (arcn->type == PAX_SLK) 1070 hd->typeflag = SYMTYPE; 1071 else if (arcn->type == PAX_GLL) 1072 hd->typeflag = LONGLINKTYPE; 1073 else 1074 hd->typeflag = LNKTYPE; 1075 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 1076 if (ul_oct((u_long)gnu_hack_len, hd->size, 1077 sizeof(hd->size), 3)) 1078 goto out; 1079 break; 1080 case PAX_GLF: 1081 case PAX_REG: 1082 case PAX_CTG: 1083 default: 1084 /* 1085 * file data with this type, set the padding 1086 */ 1087 if (arcn->type == PAX_GLF) { 1088 hd->typeflag = LONGNAMETYPE; 1089 arcn->pad = TAR_PAD(gnu_hack_len); 1090 if (OFFT_OCT((u_long)gnu_hack_len, hd->size, 1091 sizeof(hd->size), 3)) { 1092 tty_warn(1,"File is too long for ustar %s", 1093 arcn->org_name); 1094 return(1); 1095 } 1096 } else { 1097 if (arcn->type == PAX_CTG) 1098 hd->typeflag = CONTTYPE; 1099 else 1100 hd->typeflag = REGTYPE; 1101 arcn->pad = TAR_PAD(arcn->sb.st_size); 1102 if (OFFT_OCT(arcn->sb.st_size, hd->size, 1103 sizeof(hd->size), 3)) { 1104 tty_warn(1,"File is too long for ustar %s", 1105 arcn->org_name); 1106 return(1); 1107 } 1108 } 1109 break; 1110 } 1111 1112 strncpy(hd->magic, TMAGIC, TMAGLEN); 1113 if (is_gnutar) 1114 hd->magic[TMAGLEN - 1] = hd->magic[TMAGLEN] = ' '; 1115 else 1116 strncpy(hd->version, TVERSION, TVERSLEN); 1117 1118 /* 1119 * set the remaining fields. Some versions want all 16 bits of mode 1120 * we better humor them (they really do not meet spec though).... 1121 */ 1122 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) || 1123 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3) || 1124 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) || 1125 ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3)) 1126 goto out; 1127 user = user_from_uid(arcn->sb.st_uid, 1); 1128 group = group_from_gid(arcn->sb.st_gid, 1); 1129 strncpy(hd->uname, user ? user : "", sizeof(hd->uname)); 1130 strncpy(hd->gname, group ? group : "", sizeof(hd->gname)); 1131 1132 /* 1133 * calculate and store the checksum write the header to the archive 1134 * return 0 tells the caller to now write the file data, 1 says no data 1135 * needs to be written 1136 */ 1137 if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum, 1138 sizeof(hd->chksum), 3)) 1139 goto out; 1140 if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0) 1141 return(-1); 1142 if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0) 1143 return(-1); 1144 if (gnu_hack_string) { 1145 int res = wr_rdbuf(gnu_hack_string, gnu_hack_len); 1146 int pad = gnu_hack_len; 1147 gnu_hack_string = NULL; 1148 gnu_hack_len = 0; 1149 if (res < 0) 1150 return(-1); 1151 if (wr_skip((off_t)(BLKMULT - (pad % BLKMULT))) < 0) 1152 return(-1); 1153 } 1154 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 1155 return(0); 1156 return(1); 1157 1158 out: 1159 /* 1160 * header field is out of range 1161 */ 1162 tty_warn(1, "Ustar header field is too small for %s", arcn->org_name); 1163 return(1); 1164} 1165 1166/* 1167 * name_split() 1168 * see if the name has to be split for storage in a ustar header. We try 1169 * to fit the entire name in the name field without splitting if we can. 1170 * The split point is always at a / 1171 * Return 1172 * character pointer to split point (always the / that is to be removed 1173 * if the split is not needed, the points is set to the start of the file 1174 * name (it would violate the spec to split there). A NULL is returned if 1175 * the file name is too long 1176 */ 1177 1178static char * 1179name_split(char *name, int len) 1180{ 1181 char *start; 1182 1183 /* 1184 * check to see if the file name is small enough to fit in the name 1185 * field. if so just return a pointer to the name. 1186 */ 1187 if (len < TNMSZ) 1188 return(name); 1189 if (len > (TPFSZ + TNMSZ)) 1190 return(NULL); 1191 1192 /* 1193 * we start looking at the biggest sized piece that fits in the name 1194 * field. We walk forward looking for a slash to split at. The idea is 1195 * to find the biggest piece to fit in the name field (or the smallest 1196 * prefix we can find) (the -1 is correct the biggest piece would 1197 * include the slash between the two parts that gets thrown away) 1198 */ 1199 start = name + len - TNMSZ; 1200 while ((*start != '\0') && (*start != '/')) 1201 ++start; 1202 1203 /* 1204 * if we hit the end of the string, this name cannot be split, so we 1205 * cannot store this file. 1206 */ 1207 if (*start == '\0') 1208 return(NULL); 1209 len = start - name; 1210 1211 /* 1212 * NOTE: /str where the length of str == TNMSZ can not be stored under 1213 * the p1003.1-1990 spec for ustar. We could force a prefix of / and 1214 * the file would then expand on extract to //str. The len == 0 below 1215 * makes this special case follow the spec to the letter. 1216 */ 1217 if ((len >= TPFSZ) || (len == 0)) 1218 return(NULL); 1219 1220 /* 1221 * ok have a split point, return it to the caller 1222 */ 1223 return(start); 1224} 1225 1226/* 1227 * deal with GNU tar -X switch. basically, we go through each line of 1228 * the file, building a string from the "glob" lines in the file into 1229 * RE lines, of the form `/^RE$//', which we pass to rep_add(), which 1230 * will add a empty replacement (exclusion), for the named files. 1231 */ 1232int 1233tar_gnutar_X_compat(path) 1234 const char *path; 1235{ 1236 char *line, sbuf[MAXPATHLEN * 2 + 1 + 5]; 1237 FILE *fp; 1238 int lineno = 0, i, j; 1239 size_t len; 1240 1241 fp = fopen(path, "r"); 1242 if (fp == NULL) { 1243 tty_warn(1, "can not open %s: %s", path, 1244 strerror(errno)); 1245 return(-1); 1246 } 1247 1248 while ((line = fgetln(fp, &len))) { 1249 lineno++; 1250 if (len > MAXPATHLEN) { 1251 tty_warn(0, "pathname too long, line %d of %s", 1252 lineno, path); 1253 } 1254 if (line[len - 1] == '\n') 1255 len--; 1256 for (i = 0, j = 2; i < len; i++) { 1257 /* 1258 * convert glob to regexp, escaping everything 1259 */ 1260 if (line[i] == '*') 1261 sbuf[j++] = '.'; 1262 else if (line[i] == '?') 1263 line[i] = '.'; 1264 else if (!isalnum(line[i]) && !isblank(line[i])) 1265 sbuf[j++] = '\\'; 1266 sbuf[j++] = line[i]; 1267 } 1268 sbuf[0] = sbuf[j + 1] = sbuf[j + 2] = '/'; 1269 sbuf[1] = '^'; 1270 sbuf[j] = '$'; 1271 sbuf[j + 3] = '\0'; 1272 if (rep_add(sbuf) < 0) 1273 return (-1); 1274 } 1275 return (0); 1276} 1277