tar.c revision 1.23
1/* $NetBSD: tar.c,v 1.23 2002/10/12 15:39:30 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. 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 1.23 2002/10/12 15:39:30 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 u_long tar_chksm(char *, int); 72static char *name_split(char *, int); 73static int ul_oct(u_long, char *, int, int); 74#ifndef NET2_STAT 75static int ull_oct(unsigned long long, 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 */ 84char *gnu_hack_string; /* ././@LongLink hackery */ 85 86/* 87 * tar_endwr() 88 * add the tar trailer of two null blocks 89 * Return: 90 * 0 if ok, -1 otherwise (what wr_skip returns) 91 */ 92 93int 94tar_endwr(void) 95{ 96 return(wr_skip((off_t)(NULLCNT*BLKMULT))); 97} 98 99/* 100 * tar_endrd() 101 * no cleanup needed here, just return size of trailer (for append) 102 * Return: 103 * size of trailer (2 * BLKMULT) 104 */ 105 106off_t 107tar_endrd(void) 108{ 109 return((off_t)(NULLCNT*BLKMULT)); 110} 111 112/* 113 * tar_trail() 114 * Called to determine if a header block is a valid trailer. We are passed 115 * the block, the in_sync flag (which tells us we are in resync mode; 116 * looking for a valid header), and cnt (which starts at zero) which is 117 * used to count the number of empty blocks we have seen so far. 118 * Return: 119 * 0 if a valid trailer, -1 if not a valid trailer, or 1 if the block 120 * could never contain a header. 121 */ 122 123int 124tar_trail(char *buf, int in_resync, int *cnt) 125{ 126 int i; 127 128 /* 129 * look for all zero, trailer is two consecutive blocks of zero 130 */ 131 for (i = 0; i < BLKMULT; ++i) { 132 if (buf[i] != '\0') 133 break; 134 } 135 136 /* 137 * if not all zero it is not a trailer, but MIGHT be a header. 138 */ 139 if (i != BLKMULT) 140 return(-1); 141 142 /* 143 * When given a zero block, we must be careful! 144 * If we are not in resync mode, check for the trailer. Have to watch 145 * out that we do not mis-identify file data as the trailer, so we do 146 * NOT try to id a trailer during resync mode. During resync mode we 147 * might as well throw this block out since a valid header can NEVER be 148 * a block of all 0 (we must have a valid file name). 149 */ 150 if (!in_resync && (++*cnt >= NULLCNT)) 151 return(0); 152 return(1); 153} 154 155/* 156 * ul_oct() 157 * convert an unsigned long to an octal string. many oddball field 158 * termination characters are used by the various versions of tar in the 159 * different fields. term selects which kind to use. str is '0' padded 160 * at the front to len. we are unable to use only one format as many old 161 * tar readers are very cranky about this. 162 * Return: 163 * 0 if the number fit into the string, -1 otherwise 164 */ 165 166static int 167ul_oct(u_long val, char *str, int len, int term) 168{ 169 char *pt; 170 171 /* 172 * term selects the appropriate character(s) for the end of the string 173 */ 174 pt = str + len - 1; 175 switch(term) { 176 case 3: 177 *pt-- = '\0'; 178 break; 179 case 2: 180 *pt-- = ' '; 181 *pt-- = '\0'; 182 break; 183 case 1: 184 *pt-- = ' '; 185 break; 186 case 0: 187 default: 188 *pt-- = '\0'; 189 *pt-- = ' '; 190 break; 191 } 192 193 /* 194 * convert and blank pad if there is space 195 */ 196 while (pt >= str) { 197 *pt-- = '0' + (char)(val & 0x7); 198 if ((val = val >> 3) == (u_long)0) 199 break; 200 } 201 202 while (pt >= str) 203 *pt-- = '0'; 204 if (val != (u_long)0) 205 return(-1); 206 return(0); 207} 208 209#ifndef NET2_STAT 210/* 211 * ull_oct() 212 * convert an unsigned long long to an octal string. one of many oddball 213 * field termination characters are used by the various versions of tar 214 * in the different fields. term selects which kind to use. str is '0' 215 * padded at the front to len. we are unable to use only one format as 216 * many old tar readers are very cranky about this. 217 * Return: 218 * 0 if the number fit into the string, -1 otherwise 219 */ 220 221static int 222ull_oct(unsigned long long val, char *str, int len, int term) 223{ 224 char *pt; 225 226 /* 227 * term selects the appropriate character(s) for the end of the string 228 */ 229 pt = str + len - 1; 230 switch(term) { 231 case 3: 232 *pt-- = '\0'; 233 break; 234 case 2: 235 *pt-- = ' '; 236 *pt-- = '\0'; 237 break; 238 case 1: 239 *pt-- = ' '; 240 break; 241 case 0: 242 default: 243 *pt-- = '\0'; 244 *pt-- = ' '; 245 break; 246 } 247 248 /* 249 * convert and blank pad if there is space 250 */ 251 while (pt >= str) { 252 *pt-- = '0' + (char)(val & 0x7); 253 if ((val = val >> 3) == 0) 254 break; 255 } 256 257 while (pt >= str) 258 *pt-- = '0'; 259 if (val != (unsigned long long)0) 260 return(-1); 261 return(0); 262} 263#endif 264 265/* 266 * tar_chksm() 267 * calculate the checksum for a tar block counting the checksum field as 268 * all blanks (BLNKSUM is that value pre-calculated, the sum of 8 blanks). 269 * NOTE: we use len to short circuit summing 0's on write since we ALWAYS 270 * pad headers with 0. 271 * Return: 272 * unsigned long checksum 273 */ 274 275static u_long 276tar_chksm(char *blk, int len) 277{ 278 char *stop; 279 char *pt; 280 u_long chksm = BLNKSUM; /* initial value is checksum field sum */ 281 282 /* 283 * add the part of the block before the checksum field 284 */ 285 pt = blk; 286 stop = blk + CHK_OFFSET; 287 while (pt < stop) 288 chksm += (u_long)(*pt++ & 0xff); 289 /* 290 * move past the checksum field and keep going, spec counts the 291 * checksum field as the sum of 8 blanks (which is pre-computed as 292 * BLNKSUM). 293 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding 294 * starts, no point in summing zero's) 295 */ 296 pt += CHK_LEN; 297 stop = blk + len; 298 while (pt < stop) 299 chksm += (u_long)(*pt++ & 0xff); 300 return(chksm); 301} 302 303/* 304 * Routines for old BSD style tar (also made portable to sysV tar) 305 */ 306 307/* 308 * tar_id() 309 * determine if a block given to us is a valid tar header (and not a USTAR 310 * header). We have to be on the lookout for those pesky blocks of all 311 * zero's. 312 * Return: 313 * 0 if a tar header, -1 otherwise 314 */ 315 316int 317tar_id(char *blk, int size) 318{ 319 HD_TAR *hd; 320 HD_USTAR *uhd; 321 322 if (size < BLKMULT) 323 return(-1); 324 hd = (HD_TAR *)blk; 325 uhd = (HD_USTAR *)blk; 326 327 /* 328 * check for block of zero's first, a simple and fast test, then make 329 * sure this is not a ustar header by looking for the ustar magic 330 * cookie. We should use TMAGLEN, but some USTAR archive programs are 331 * wrong and create archives missing the \0. Last we check the 332 * checksum. If this is ok we have to assume it is a valid header. 333 */ 334 if (hd->name[0] == '\0') 335 return(-1); 336 if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0) 337 return(-1); 338 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 339 return(-1); 340 return(0); 341} 342 343/* 344 * tar_opt() 345 * handle tar format specific -o options 346 * Return: 347 * 0 if ok -1 otherwise 348 */ 349 350int 351tar_opt(void) 352{ 353 OPLIST *opt; 354 355 while ((opt = opt_next()) != NULL) { 356 if (strcmp(opt->name, TAR_OPTION) || 357 strcmp(opt->value, TAR_NODIR)) { 358 tty_warn(1, 359 "Unknown tar format -o option/value pair %s=%s", 360 opt->name, opt->value); 361 tty_warn(1, 362 "%s=%s is the only supported tar format option", 363 TAR_OPTION, TAR_NODIR); 364 return(-1); 365 } 366 367 /* 368 * we only support one option, and only when writing 369 */ 370 if ((act != APPND) && (act != ARCHIVE)) { 371 tty_warn(1, "%s=%s is only supported when writing.", 372 opt->name, opt->value); 373 return(-1); 374 } 375 tar_nodir = 1; 376 } 377 return(0); 378} 379 380 381/* 382 * tar_rd() 383 * extract the values out of block already determined to be a tar header. 384 * store the values in the ARCHD parameter. 385 * Return: 386 * 0 387 */ 388 389int 390tar_rd(ARCHD *arcn, char *buf) 391{ 392 HD_TAR *hd; 393 char *pt; 394 395 /* 396 * we only get proper sized buffers passed to us 397 */ 398 if (tar_id(buf, BLKMULT) < 0) 399 return(-1); 400 arcn->org_name = arcn->name; 401 arcn->sb.st_nlink = 1; 402 arcn->pat = NULL; 403 404 /* 405 * copy out the name and values in the stat buffer 406 */ 407 hd = (HD_TAR *)buf; 408 if (gnu_hack_string) { 409 arcn->nlen = strlcpy(arcn->name, gnu_hack_string, 410 sizeof(arcn->name)); 411 free(gnu_hack_string); 412 gnu_hack_string = NULL; 413 } else { 414 arcn->nlen = strlcpy(arcn->name, hd->name, sizeof(arcn->name)); 415 } 416 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) & 417 0xfff); 418 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 419 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 420 arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT); 421 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 422 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 423 424 /* 425 * have to look at the last character, it may be a '/' and that is used 426 * to encode this as a directory 427 */ 428 pt = &(arcn->name[arcn->nlen - 1]); 429 arcn->pad = 0; 430 arcn->skip = 0; 431 switch(hd->linkflag) { 432 case SYMTYPE: 433 /* 434 * symbolic link, need to get the link name and set the type in 435 * the st_mode so -v printing will look correct. 436 */ 437 arcn->type = PAX_SLK; 438 arcn->ln_nlen = strlcpy(arcn->ln_name, hd->linkname, 439 sizeof(arcn->ln_name)); 440 arcn->sb.st_mode |= S_IFLNK; 441 break; 442 case LNKTYPE: 443 /* 444 * hard link, need to get the link name, set the type in the 445 * st_mode and st_nlink so -v printing will look better. 446 */ 447 arcn->type = PAX_HLK; 448 arcn->sb.st_nlink = 2; 449 arcn->ln_nlen = strlcpy(arcn->ln_name, hd->linkname, 450 sizeof(arcn->ln_name)); 451 452 /* 453 * no idea of what type this thing really points at, but 454 * we set something for printing only. 455 */ 456 arcn->sb.st_mode |= S_IFREG; 457 break; 458 case LONGLINKTYPE: 459 arcn->type = PAX_GLL; 460 /* FALLTHROUGH */ 461 case LONGNAMETYPE: 462 /* 463 * GNU long link/file; we tag these here and let the 464 * pax internals deal with it -- too ugly otherwise. 465 */ 466 if (hd->linkflag != LONGLINKTYPE) 467 arcn->type = PAX_GLF; 468 arcn->pad = TAR_PAD(arcn->sb.st_size); 469 arcn->skip = arcn->sb.st_size; 470 arcn->ln_name[0] = '\0'; 471 arcn->ln_nlen = 0; 472 break; 473 case AREGTYPE: 474 case REGTYPE: 475 case DIRTYPE: /* see below */ 476 default: 477 /* 478 * If we have a trailing / this is a directory and NOT a file. 479 * Note: V7 tar doesn't actually have DIRTYPE, but it was 480 * reported that V7 archives using USTAR directories do exist. 481 */ 482 arcn->ln_name[0] = '\0'; 483 arcn->ln_nlen = 0; 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 if (!strncmp(hd->magic, "ustar ", 8)) 733 is_oldgnutar = 1; 734 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 735 return(-1); 736 return(0); 737} 738 739/* 740 * ustar_rd() 741 * extract the values out of block already determined to be a ustar header. 742 * store the values in the ARCHD parameter. 743 * Return: 744 * 0 745 */ 746 747int 748ustar_rd(ARCHD *arcn, char *buf) 749{ 750 HD_USTAR *hd; 751 char *dest; 752 int cnt; 753 dev_t devmajor; 754 dev_t devminor; 755 756 /* 757 * we only get proper sized buffers 758 */ 759 if (ustar_id(buf, BLKMULT) < 0) 760 return(-1); 761 arcn->org_name = arcn->name; 762 arcn->sb.st_nlink = 1; 763 arcn->pat = NULL; 764 arcn->nlen = 0; 765 hd = (HD_USTAR *)buf; 766 767 /* 768 * see if the filename is split into two parts. if, so joint the parts. 769 * we copy the prefix first and add a / between the prefix and name. 770 */ 771 dest = arcn->name; 772 if (*(hd->prefix) != '\0') { 773 cnt = strlcpy(arcn->name, hd->prefix, sizeof(arcn->name)); 774 dest += cnt; 775 *dest++ = '/'; 776 cnt++; 777 } 778 arcn->nlen = cnt + strlcpy(dest, hd->name, sizeof(arcn->name) - cnt); 779 780 /* 781 * follow the spec to the letter. we should only have mode bits, strip 782 * off all other crud we may be passed. 783 */ 784 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) & 785 0xfff); 786 arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT); 787 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 788 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 789 790 /* 791 * If we can find the ascii names for gname and uname in the password 792 * and group files we will use the uid's and gid they bind. Otherwise 793 * we use the uid and gid values stored in the header. (This is what 794 * the posix spec wants). 795 */ 796 hd->gname[sizeof(hd->gname) - 1] = '\0'; 797 if (gid_from_group(hd->gname, &(arcn->sb.st_gid)) < 0) 798 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 799 hd->uname[sizeof(hd->uname) - 1] = '\0'; 800 if (uid_from_user(hd->uname, &(arcn->sb.st_uid)) < 0) 801 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 802 803 /* 804 * set the defaults, these may be changed depending on the file type 805 */ 806 arcn->ln_name[0] = '\0'; 807 arcn->ln_nlen = 0; 808 arcn->pad = 0; 809 arcn->skip = 0; 810 arcn->sb.st_rdev = (dev_t)0; 811 812 /* 813 * set the mode and PAX type according to the typeflag in the header 814 */ 815 switch(hd->typeflag) { 816 case FIFOTYPE: 817 arcn->type = PAX_FIF; 818 arcn->sb.st_mode |= S_IFIFO; 819 break; 820 case DIRTYPE: 821 arcn->type = PAX_DIR; 822 arcn->sb.st_mode |= S_IFDIR; 823 arcn->sb.st_nlink = 2; 824 825 /* 826 * Some programs that create ustar archives append a '/' 827 * to the pathname for directories. This clearly violates 828 * ustar specs, but we will silently strip it off anyway. 829 */ 830 if (arcn->name[arcn->nlen - 1] == '/') 831 arcn->name[--arcn->nlen] = '\0'; 832 break; 833 case BLKTYPE: 834 case CHRTYPE: 835 /* 836 * this type requires the rdev field to be set. 837 */ 838 if (hd->typeflag == BLKTYPE) { 839 arcn->type = PAX_BLK; 840 arcn->sb.st_mode |= S_IFBLK; 841 } else { 842 arcn->type = PAX_CHR; 843 arcn->sb.st_mode |= S_IFCHR; 844 } 845 devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT); 846 devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT); 847 arcn->sb.st_rdev = TODEV(devmajor, devminor); 848 break; 849 case SYMTYPE: 850 case LNKTYPE: 851 if (hd->typeflag == SYMTYPE) { 852 arcn->type = PAX_SLK; 853 arcn->sb.st_mode |= S_IFLNK; 854 } else { 855 arcn->type = PAX_HLK; 856 /* 857 * so printing looks better 858 */ 859 arcn->sb.st_mode |= S_IFREG; 860 arcn->sb.st_nlink = 2; 861 } 862 /* 863 * copy the link name 864 */ 865 arcn->ln_nlen = strlcpy(arcn->ln_name, hd->linkname, 866 sizeof(arcn->ln_name)); 867 break; 868 case CONTTYPE: 869 case AREGTYPE: 870 case REGTYPE: 871 default: 872 /* 873 * these types have file data that follows. Set the skip and 874 * pad fields. 875 */ 876 arcn->type = PAX_REG; 877 arcn->pad = TAR_PAD(arcn->sb.st_size); 878 arcn->skip = arcn->sb.st_size; 879 arcn->sb.st_mode |= S_IFREG; 880 break; 881 } 882 return(0); 883} 884 885/* 886 * ustar_wr() 887 * write a ustar header for the file specified in the ARCHD to the archive 888 * Have to check for file types that cannot be stored and file names that 889 * are too long. Be careful of the term (last arg) to ul_oct, we only use 890 * '\0' for the termination character (this is different than picky tar) 891 * ASSUMED: space after header in header block is zero filled 892 * Return: 893 * 0 if file has data to be written after the header, 1 if file has NO 894 * data to write after the header, -1 if archive write failed 895 */ 896 897int 898ustar_wr(ARCHD *arcn) 899{ 900 HD_USTAR *hd; 901 char *pt; 902 char hdblk[sizeof(HD_USTAR)]; 903 const char *user, *group; 904 905 /* 906 * check for those file system types ustar cannot store 907 */ 908 if (arcn->type == PAX_SCK) { 909 tty_warn(1, "Ustar cannot archive a socket %s", arcn->org_name); 910 return(1); 911 } 912 913 /* 914 * check the length of the linkname 915 */ 916 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) || 917 (arcn->type == PAX_HRG)) && (arcn->ln_nlen >= sizeof(hd->linkname))){ 918 tty_warn(1, "Link name too long for ustar %s", arcn->ln_name); 919 return(1); 920 } 921 922 /* 923 * split the path name into prefix and name fields (if needed). if 924 * pt != arcn->name, the name has to be split 925 */ 926 if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) { 927 tty_warn(1, "File name too long for ustar %s", arcn->name); 928 return(1); 929 } 930 931 /* 932 * zero out the header so we don't have to worry about zero fill below 933 */ 934 memset(hdblk, 0, sizeof(hdblk)); 935 hd = (HD_USTAR *)hdblk; 936 arcn->pad = 0L; 937 938 /* 939 * split the name, or zero out the prefix 940 */ 941 if (pt != arcn->name) { 942 /* 943 * name was split, pt points at the / where the split is to 944 * occur, we remove the / and copy the first part to the prefix 945 */ 946 *pt = '\0'; 947 strlcpy(hd->prefix, arcn->name, sizeof(hd->prefix)); 948 *pt++ = '/'; 949 } 950 951 /* 952 * copy the name part. this may be the whole path or the part after 953 * the prefix 954 */ 955 strlcpy(hd->name, pt, sizeof(hd->name)); 956 957 /* 958 * set the fields in the header that are type dependent 959 */ 960 switch(arcn->type) { 961 case PAX_DIR: 962 hd->typeflag = DIRTYPE; 963 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 964 goto out; 965 break; 966 case PAX_CHR: 967 case PAX_BLK: 968 if (arcn->type == PAX_CHR) 969 hd->typeflag = CHRTYPE; 970 else 971 hd->typeflag = BLKTYPE; 972 if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor, 973 sizeof(hd->devmajor), 3) || 974 ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor, 975 sizeof(hd->devminor), 3) || 976 ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 977 goto out; 978 break; 979 case PAX_FIF: 980 hd->typeflag = FIFOTYPE; 981 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 982 goto out; 983 break; 984 case PAX_SLK: 985 case PAX_HLK: 986 case PAX_HRG: 987 if (arcn->type == PAX_SLK) 988 hd->typeflag = SYMTYPE; 989 else 990 hd->typeflag = LNKTYPE; 991 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 992 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3)) 993 goto out; 994 break; 995 case PAX_REG: 996 case PAX_CTG: 997 default: 998 /* 999 * file data with this type, set the padding 1000 */ 1001 if (arcn->type == PAX_CTG) 1002 hd->typeflag = CONTTYPE; 1003 else 1004 hd->typeflag = REGTYPE; 1005 arcn->pad = TAR_PAD(arcn->sb.st_size); 1006 if (OFFT_OCT(arcn->sb.st_size, hd->size, sizeof(hd->size), 3)) { 1007 tty_warn(1,"File is too long for ustar %s", 1008 arcn->org_name); 1009 return(1); 1010 } 1011 break; 1012 } 1013 1014 strncpy(hd->magic, TMAGIC, TMAGLEN); 1015 strncpy(hd->version, TVERSION, TVERSLEN); 1016 1017 /* 1018 * set the remaining fields. Some versions want all 16 bits of mode 1019 * we better humor them (they really do not meet spec though).... 1020 */ 1021 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) || 1022 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3) || 1023 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) || 1024 ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3)) 1025 goto out; 1026 user = user_from_uid(arcn->sb.st_uid, 1); 1027 group = group_from_gid(arcn->sb.st_gid, 1); 1028 strncpy(hd->uname, user ? user : "", sizeof(hd->uname)); 1029 strncpy(hd->gname, group ? group : "", sizeof(hd->gname)); 1030 1031 /* 1032 * calculate and store the checksum write the header to the archive 1033 * return 0 tells the caller to now write the file data, 1 says no data 1034 * needs to be written 1035 */ 1036 if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum, 1037 sizeof(hd->chksum), 3)) 1038 goto out; 1039 if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0) 1040 return(-1); 1041 if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0) 1042 return(-1); 1043 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 1044 return(0); 1045 return(1); 1046 1047 out: 1048 /* 1049 * header field is out of range 1050 */ 1051 tty_warn(1, "Ustar header field is too small for %s", arcn->org_name); 1052 return(1); 1053} 1054 1055/* 1056 * name_split() 1057 * see if the name has to be split for storage in a ustar header. We try 1058 * to fit the entire name in the name field without splitting if we can. 1059 * The split point is always at a / 1060 * Return 1061 * character pointer to split point (always the / that is to be removed 1062 * if the split is not needed, the points is set to the start of the file 1063 * name (it would violate the spec to split there). A NULL is returned if 1064 * the file name is too long 1065 */ 1066 1067static char * 1068name_split(char *name, int len) 1069{ 1070 char *start; 1071 1072 /* 1073 * check to see if the file name is small enough to fit in the name 1074 * field. if so just return a pointer to the name. 1075 */ 1076 if (len < TNMSZ) 1077 return(name); 1078 if (len > (TPFSZ + TNMSZ)) 1079 return(NULL); 1080 1081 /* 1082 * we start looking at the biggest sized piece that fits in the name 1083 * field. We walk forward looking for a slash to split at. The idea is 1084 * to find the biggest piece to fit in the name field (or the smallest 1085 * prefix we can find) (the -1 is correct the biggest piece would 1086 * include the slash between the two parts that gets thrown away) 1087 */ 1088 start = name + len - TNMSZ; 1089 while ((*start != '\0') && (*start != '/')) 1090 ++start; 1091 1092 /* 1093 * if we hit the end of the string, this name cannot be split, so we 1094 * cannot store this file. 1095 */ 1096 if (*start == '\0') 1097 return(NULL); 1098 len = start - name; 1099 1100 /* 1101 * NOTE: /str where the length of str == TNMSZ can not be stored under 1102 * the p1003.1-1990 spec for ustar. We could force a prefix of / and 1103 * the file would then expand on extract to //str. The len == 0 below 1104 * makes this special case follow the spec to the letter. 1105 */ 1106 if ((len >= TPFSZ) || (len == 0)) 1107 return(NULL); 1108 1109 /* 1110 * ok have a split point, return it to the caller 1111 */ 1112 return(start); 1113} 1114 1115/* 1116 * deal with GNU tar -X switch. basically, we go through each line of 1117 * the file, building a string from the "glob" lines in the file into 1118 * RE lines, of the form `/^RE$//', which we pass to rep_add(), which 1119 * will add a empty replacement (exclusion), for the named files. 1120 */ 1121int 1122tar_gnutar_X_compat(path) 1123 const char *path; 1124{ 1125 char *line, sbuf[MAXPATHLEN * 2 + 1 + 5]; 1126 FILE *fp; 1127 int lineno = 0, i, j; 1128 size_t len; 1129 1130 fp = fopen(path, "r"); 1131 if (fp == NULL) { 1132 tty_warn(1, "can not open %s: %s", path, 1133 strerror(errno)); 1134 return(-1); 1135 } 1136 1137 while ((line = fgetln(fp, &len))) { 1138 lineno++; 1139 if (len > MAXPATHLEN) { 1140 tty_warn(0, "pathname too long, line %d of %s", 1141 lineno, path); 1142 } 1143 if (line[len - 1] == '\n') 1144 len--; 1145 for (i = 0, j = 2; i < len; i++) { 1146 /* 1147 * convert glob to regexp, escaping everything 1148 */ 1149 if (line[i] == '*') 1150 sbuf[j++] = '.'; 1151 else if (line[i] == '?') 1152 line[i] = '.'; 1153 else if (!isalnum(line[i]) && !isblank(line[i])) 1154 sbuf[j++] = '\\'; 1155 sbuf[j++] = line[i]; 1156 } 1157 sbuf[0] = sbuf[j + 1] = sbuf[j + 2] = '/'; 1158 sbuf[1] = '^'; 1159 sbuf[j] = '$'; 1160 sbuf[j + 3] = '\0'; 1161 if (rep_add(sbuf) < 0) 1162 return (-1); 1163 } 1164 return (0); 1165} 1166