1/* $OpenBSD: tar.c,v 1.41 2006/03/04 20:24:55 otto Exp $ */ 2/* $NetBSD: tar.c,v 1.5 1995/03/21 09:07:49 cgd Exp $ */ 3 4/*- 5 * Copyright (c) 1992 Keith Muller. 6 * Copyright (c) 1992, 1993 7 * The Regents of the University of California. All rights reserved. 8 * 9 * This code is derived from software contributed to Berkeley by 10 * Keith Muller of the University of California, San Diego. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 */ 36 37#include <sys/cdefs.h> 38#ifndef lint 39#if 0 40static const char sccsid[] = "@(#)tar.c 8.2 (Berkeley) 4/18/94"; 41#else 42__used static const char rcsid[] = "$OpenBSD: tar.c,v 1.41 2006/03/04 20:24:55 otto Exp $"; 43#endif 44#endif /* not lint */ 45 46#include <sys/types.h> 47#include <sys/time.h> 48#include <sys/stat.h> 49#include <sys/param.h> 50#include <string.h> 51#include <stdio.h> 52#include <unistd.h> 53#include <stdlib.h> 54#include "pax.h" 55#include "extern.h" 56#include "tar.h" 57 58/* 59 * Routines for reading, writing and header identify of various versions of tar 60 */ 61 62static size_t expandname(char *, size_t, char **, const char *, size_t); 63static u_long tar_chksm(char *, int); 64static char *name_split(char *, int); 65static int ul_oct(u_long, char *, int, int); 66#ifndef LONG_OFF_T 67static int uqd_oct(u_quad_t, char *, int, int); 68#endif 69 70static uid_t uid_nobody; 71static uid_t uid_warn; 72static gid_t gid_nobody; 73static gid_t gid_warn; 74 75/* 76 * Routines common to all versions of tar 77 */ 78 79static int tar_nodir; /* do not write dirs under old tar */ 80char *gnu_name_string; /* GNU ././@LongLink hackery name */ 81char *gnu_link_string; /* GNU ././@LongLink hackery link */ 82 83/* 84 * tar_endwr() 85 * add the tar trailer of two null blocks 86 * Return: 87 * 0 if ok, -1 otherwise (what wr_skip returns) 88 */ 89 90int 91tar_endwr(void) 92{ 93 return(wr_skip((off_t)(NULLCNT*BLKMULT))); 94} 95 96/* 97 * tar_endrd() 98 * no cleanup needed here, just return size of trailer (for append) 99 * Return: 100 * size of trailer (2 * BLKMULT) 101 */ 102 103off_t 104tar_endrd(void) 105{ 106 return((off_t)(NULLCNT*BLKMULT)); 107} 108 109/* 110 * tar_trail() 111 * Called to determine if a header block is a valid trailer. We are passed 112 * the block, the in_sync flag (which tells us we are in resync mode; 113 * looking for a valid header), and cnt (which starts at zero) which is 114 * used to count the number of empty blocks we have seen so far. 115 * Return: 116 * 0 if a valid trailer, -1 if not a valid trailer, or 1 if the block 117 * could never contain a header. 118 */ 119 120int 121tar_trail(ARCHD *ignore, char *buf, int in_resync, int *cnt) 122{ 123 int i; 124 125 /* 126 * look for all zero, trailer is two consecutive blocks of zero 127 */ 128 for (i = 0; i < BLKMULT; ++i) { 129 if (buf[i] != '\0') 130 break; 131 } 132 133 /* 134 * if not all zero it is not a trailer, but MIGHT be a header. 135 */ 136 if (i != BLKMULT) 137 return(-1); 138 139 /* 140 * When given a zero block, we must be careful! 141 * If we are not in resync mode, check for the trailer. Have to watch 142 * out that we do not mis-identify file data as the trailer, so we do 143 * NOT try to id a trailer during resync mode. During resync mode we 144 * might as well throw this block out since a valid header can NEVER be 145 * a block of all 0 (we must have a valid file name). 146 */ 147 if (!in_resync && (++*cnt >= NULLCNT)) 148 return(0); 149 return(1); 150} 151 152/* 153 * ul_oct() 154 * convert an unsigned long to an octal string. many oddball field 155 * termination characters are used by the various versions of tar in the 156 * different fields. term selects which kind to use. str is '0' padded 157 * at the front to len. we are unable to use only one format as many old 158 * tar readers are very cranky about this. 159 * Return: 160 * 0 if the number fit into the string, -1 otherwise 161 */ 162 163static int 164ul_oct(u_long val, char *str, int len, int term) 165{ 166 char *pt; 167 168 /* 169 * term selects the appropriate character(s) for the end of the string 170 */ 171 pt = str + len - 1; 172 switch (term) { 173 case 3: 174 *pt-- = '\0'; 175 break; 176 case 2: 177 *pt-- = ' '; 178 *pt-- = '\0'; 179 break; 180 case 1: 181 *pt-- = ' '; 182 break; 183 case 0: 184 default: 185 *pt-- = '\0'; 186 *pt-- = ' '; 187 break; 188 } 189 190 /* 191 * convert and blank pad if there is space 192 */ 193 while (pt >= str) { 194 *pt-- = '0' + (char)(val & 0x7); 195 if ((val = val >> 3) == (u_long)0) 196 break; 197 } 198 199 while (pt >= str) 200 *pt-- = '0'; 201 if (val != (u_long)0) 202 return(-1); 203 return(0); 204} 205 206#ifndef LONG_OFF_T 207/* 208 * uqd_oct() 209 * convert an u_quad_t to an octal string. one of many oddball field 210 * termination characters are used by the various versions of tar in the 211 * different fields. term selects which kind to use. str is '0' padded 212 * at the front to len. we are unable to use only one format as many old 213 * tar readers are very cranky about this. 214 * Return: 215 * 0 if the number fit into the string, -1 otherwise 216 */ 217 218static int 219uqd_oct(u_quad_t val, char *str, int len, int term) 220{ 221 char *pt; 222 223 /* 224 * term selects the appropriate character(s) for the end of the string 225 */ 226 pt = str + len - 1; 227 switch (term) { 228 case 3: 229 *pt-- = '\0'; 230 break; 231 case 2: 232 *pt-- = ' '; 233 *pt-- = '\0'; 234 break; 235 case 1: 236 *pt-- = ' '; 237 break; 238 case 0: 239 default: 240 *pt-- = '\0'; 241 *pt-- = ' '; 242 break; 243 } 244 245 /* 246 * convert and blank pad if there is space 247 */ 248 while (pt >= str) { 249 *pt-- = '0' + (char)(val & 0x7); 250 if ((val = val >> 3) == 0) 251 break; 252 } 253 254 while (pt >= str) 255 *pt-- = '0'; 256 if (val != (u_quad_t)0) 257 return(-1); 258 return(0); 259} 260#endif 261 262/* 263 * tar_chksm() 264 * calculate the checksum for a tar block counting the checksum field as 265 * all blanks (BLNKSUM is that value pre-calculated, the sum of 8 blanks). 266 * NOTE: we use len to short circuit summing 0's on write since we ALWAYS 267 * pad headers with 0. 268 * Return: 269 * unsigned long checksum 270 */ 271 272static u_long 273tar_chksm(char *blk, int len) 274{ 275 char *stop; 276 char *pt; 277 u_long chksm = BLNKSUM; /* initial value is checksum field sum */ 278 279 /* 280 * add the part of the block before the checksum field 281 */ 282 pt = blk; 283 stop = blk + CHK_OFFSET; 284 while (pt < stop) 285 chksm += (u_long)(*pt++ & 0xff); 286 /* 287 * move past the checksum field and keep going, spec counts the 288 * checksum field as the sum of 8 blanks (which is pre-computed as 289 * BLNKSUM). 290 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding 291 * starts, no point in summing zero's) 292 */ 293 pt += CHK_LEN; 294 stop = blk + len; 295 while (pt < stop) 296 chksm += (u_long)(*pt++ & 0xff); 297 return(chksm); 298} 299 300/* 301 * Routines for old BSD style tar (also made portable to sysV tar) 302 */ 303 304/* 305 * tar_id() 306 * determine if a block given to us is a valid tar header (and not a USTAR 307 * header). We have to be on the lookout for those pesky blocks of all 308 * zero's. 309 * Return: 310 * 0 if a tar header, -1 otherwise 311 */ 312 313int 314tar_id(char *blk, int size) 315{ 316 HD_TAR *hd; 317 HD_USTAR *uhd; 318 319 if (size < BLKMULT) 320 return(-1); 321 hd = (HD_TAR *)blk; 322 uhd = (HD_USTAR *)blk; 323 324 /* 325 * check for block of zero's first, a simple and fast test, then make 326 * sure this is not a ustar header by looking for the ustar magic 327 * cookie. We should use TMAGLEN, but some USTAR archive programs are 328 * wrong and create archives missing the \0. Last we check the 329 * checksum. If this is ok we have to assume it is a valid header. 330 */ 331 if (hd->name[0] == '\0') 332 return(-1); 333 if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0) 334 return(-1); 335 if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 336 return(-1); 337 force_one_volume = 1; 338 return(0); 339} 340 341/* 342 * tar_opt() 343 * handle tar format specific -o options 344 * Return: 345 * 0 if ok -1 otherwise 346 */ 347 348int 349tar_opt(void) 350{ 351 OPLIST *opt; 352 353 while ((opt = opt_next()) != NULL) { 354 if (strcmp(opt->name, TAR_OPTION) || 355 strcmp(opt->value, TAR_NODIR)) { 356 paxwarn(1, "Unknown tar format -o option/value pair %s=%s", 357 opt->name, opt->value); 358 paxwarn(1,"%s=%s is the only supported tar format option", 359 TAR_OPTION, TAR_NODIR); 360 return(-1); 361 } 362 363 /* 364 * we only support one option, and only when writing 365 */ 366 if ((act != APPND) && (act != ARCHIVE)) { 367 paxwarn(1, "%s=%s is only supported when writing.", 368 opt->name, opt->value); 369 return(-1); 370 } 371 tar_nodir = 1; 372 } 373 return(0); 374} 375 376 377/* 378 * tar_rd() 379 * extract the values out of block already determined to be a tar header. 380 * store the values in the ARCHD parameter. 381 * Return: 382 * 0 383 */ 384 385int 386tar_rd(ARCHD *arcn, char *buf) 387{ 388 HD_TAR *hd; 389 char *pt; 390 391 /* 392 * we only get proper sized buffers passed to us 393 */ 394 if (tar_id(buf, BLKMULT) < 0) 395 return(-1); 396 memset(arcn, 0, sizeof(*arcn)); 397 arcn->org_name = arcn->name; 398 arcn->sb.st_nlink = 1; 399 400 /* 401 * copy out the name and values in the stat buffer 402 */ 403 hd = (HD_TAR *)buf; 404 if (hd->linkflag != LONGLINKTYPE && hd->linkflag != LONGNAMETYPE) { 405 arcn->nlen = expandname(arcn->name, sizeof(arcn->name), 406 &gnu_name_string, hd->name, sizeof(hd->name)); 407 arcn->ln_nlen = expandname(arcn->ln_name, sizeof(arcn->ln_name), 408 &gnu_link_string, hd->linkname, sizeof(hd->linkname)); 409 } 410 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) & 411 0xfff); 412 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 413 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 414#ifdef LONG_OFF_T 415 arcn->sb.st_size = (off_t)asc_ul(hd->size, sizeof(hd->size), OCT); 416#else 417 arcn->sb.st_size = (off_t)asc_uqd(hd->size, sizeof(hd->size), OCT); 418#endif 419 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 420 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 421 422 /* 423 * have to look at the last character, it may be a '/' and that is used 424 * to encode this as a directory 425 */ 426 pt = &(arcn->name[arcn->nlen - 1]); 427 arcn->pad = 0; 428 arcn->skip = 0; 429 switch (hd->linkflag) { 430 case SYMTYPE: 431 /* 432 * symbolic link, need to get the link name and set the type in 433 * the st_mode so -v printing will look correct. 434 */ 435 arcn->type = PAX_SLK; 436 arcn->sb.st_mode |= S_IFLNK; 437 arcn->ln_nlen = strlcpy(arcn->ln_name, hd->linkname, sizeof(arcn->ln_name)); 438 break; 439 case LNKTYPE: 440 /* 441 * hard link, need to get the link name, set the type in the 442 * st_mode and st_nlink so -v printing will look better. 443 */ 444 arcn->type = PAX_HLK; 445 arcn->sb.st_nlink = 2; 446 arcn->ln_nlen = strlcpy(arcn->ln_name, hd->linkname, sizeof(arcn->ln_name)); 447 448 /* 449 * no idea of what type this thing really points at, but 450 * we set something for printing only. 451 */ 452 arcn->sb.st_mode |= S_IFREG; 453 break; 454 case LONGLINKTYPE: 455 case LONGNAMETYPE: 456 /* 457 * GNU long link/file; we tag these here and let the 458 * pax internals deal with it -- too ugly otherwise. 459 */ 460 arcn->type = 461 hd->linkflag == LONGLINKTYPE ? PAX_GLL : PAX_GLF; 462 arcn->pad = TAR_PAD(arcn->sb.st_size); 463 arcn->skip = arcn->sb.st_size; 464 break; 465 case DIRTYPE: 466 /* 467 * It is a directory, set the mode for -v printing 468 */ 469 arcn->type = PAX_DIR; 470 arcn->sb.st_mode |= S_IFDIR; 471 arcn->sb.st_nlink = 2; 472 break; 473 case AREGTYPE: 474 case REGTYPE: 475 default: 476 /* 477 * If we have a trailing / this is a directory and NOT a file. 478 */ 479 arcn->ln_name[0] = '\0'; 480 arcn->ln_nlen = 0; 481 if (*pt == '/') { 482 /* 483 * it is a directory, set the mode for -v printing 484 */ 485 arcn->type = PAX_DIR; 486 arcn->sb.st_mode |= S_IFDIR; 487 arcn->sb.st_nlink = 2; 488 } else { 489 /* 490 * have a file that will be followed by data. Set the 491 * skip value to the size field and calculate the size 492 * of the padding. 493 */ 494 arcn->type = PAX_REG; 495 arcn->sb.st_mode |= S_IFREG; 496 arcn->pad = TAR_PAD(arcn->sb.st_size); 497 arcn->skip = arcn->sb.st_size; 498 } 499 break; 500 } 501 502 /* 503 * strip off any trailing slash. 504 */ 505 if (*pt == '/') { 506 *pt = '\0'; 507 --arcn->nlen; 508 } 509 return(0); 510} 511 512/* 513 * tar_wr() 514 * write a tar header for the file specified in the ARCHD to the archive. 515 * Have to check for file types that cannot be stored and file names that 516 * are too long. Be careful of the term (last arg) to ul_oct, each field 517 * of tar has it own spec for the termination character(s). 518 * ASSUMED: space after header in header block is zero filled 519 * Return: 520 * 0 if file has data to be written after the header, 1 if file has NO 521 * data to write after the header, -1 if archive write failed 522 */ 523 524int 525tar_wr(ARCHD *arcn) 526{ 527 HD_TAR *hd; 528 int len; 529 HD_TAR hdblk; 530 531 /* 532 * check for those file system types which tar cannot store 533 */ 534 switch (arcn->type) { 535 case PAX_DIR: 536 /* 537 * user asked that dirs not be written to the archive 538 */ 539 if (tar_nodir) 540 return(1); 541 break; 542 case PAX_CHR: 543 paxwarn(1, "Tar cannot archive a character device %s", 544 arcn->org_name); 545 return(1); 546 case PAX_BLK: 547 paxwarn(1, "Tar cannot archive a block device %s", arcn->org_name); 548 return(1); 549 case PAX_SCK: 550 paxwarn(1, "Tar cannot archive a socket %s", arcn->org_name); 551 return(1); 552 case PAX_FIF: 553 paxwarn(1, "Tar cannot archive a fifo %s", arcn->org_name); 554 return(1); 555 case PAX_SLK: 556 case PAX_HLK: 557 case PAX_HRG: 558 if (arcn->ln_nlen >= sizeof(hd->linkname)) { 559 paxwarn(1, "Link name too long for tar %s", 560 arcn->ln_name); 561 return(1); 562 } 563 break; 564 case PAX_REG: 565 case PAX_CTG: 566 default: 567 break; 568 } 569 570 /* 571 * check file name len, remember extra char for dirs (the / at the end) 572 */ 573 len = arcn->nlen; 574 if (arcn->type == PAX_DIR) 575 ++len; 576 if (len >= sizeof(hd->name)) { 577 paxwarn(1, "File name too long for tar %s", arcn->name); 578 return(1); 579 } 580 581 /* 582 * Copy the data out of the ARCHD into the tar header based on the type 583 * of the file. Remember, many tar readers want all fields to be 584 * padded with zero so we zero the header first. We then set the 585 * linkflag field (type), the linkname, the size, and set the padding 586 * (if any) to be added after the file data (0 for all other types, 587 * as they only have a header). 588 */ 589 memset(&hdblk, 0, sizeof(hdblk)); 590 hd = (HD_TAR *)&hdblk; 591 strlcpy(hd->name, arcn->name, sizeof(hd->name)); 592 arcn->pad = 0; 593 594 if (arcn->type == PAX_DIR) { 595 /* 596 * directories are the same as files, except have a filename 597 * that ends with a /, we add the slash here. No data follows 598 * dirs, so no pad. 599 */ 600 hd->linkflag = AREGTYPE; 601 hd->name[len-1] = '/'; 602 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 603 goto out; 604 } else if (arcn->type == PAX_SLK) { 605 /* 606 * no data follows this file, so no pad 607 */ 608 hd->linkflag = SYMTYPE; 609 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 610 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 611 goto out; 612 } else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) { 613 /* 614 * no data follows this file, so no pad 615 */ 616 hd->linkflag = LNKTYPE; 617 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 618 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1)) 619 goto out; 620 } else { 621 /* 622 * data follows this file, so set the pad 623 */ 624 hd->linkflag = AREGTYPE; 625# ifdef LONG_OFF_T 626 if (ul_oct((u_long)arcn->sb.st_size, hd->size, 627 sizeof(hd->size), 1)) { 628# else 629 if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size, 630 sizeof(hd->size), 1)) { 631# endif 632 paxwarn(1,"File is too large for tar %s", arcn->org_name); 633 return(1); 634 } 635 arcn->pad = TAR_PAD(arcn->sb.st_size); 636 } 637 638 /* 639 * copy those fields that are independent of the type 640 */ 641 if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) || 642 ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) || 643 ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) || 644 ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1)) 645 goto out; 646 647 /* 648 * calculate and add the checksum, then write the header. A return of 649 * 0 tells the caller to now write the file data, 1 says no data needs 650 * to be written 651 */ 652 if (ul_oct(tar_chksm((char *)&hdblk, sizeof(HD_TAR)), hd->chksum, 653 sizeof(hd->chksum), 3)) 654 goto out; 655 if (wr_rdbuf((char *)&hdblk, sizeof(HD_TAR)) < 0) 656 return(-1); 657 if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0) 658 return(-1); 659 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 660 return(0); 661 return(1); 662 663 out: 664 /* 665 * header field is out of range 666 */ 667 paxwarn(1, "Tar header field is too small for %s", arcn->org_name); 668 return(1); 669} 670 671/* 672 * Routines for POSIX ustar 673 */ 674 675/* 676 * ustar_strd() 677 * initialization for ustar read 678 * Return: 679 * 0 if ok, -1 otherwise 680 */ 681 682int 683ustar_strd(void) 684{ 685 if ((usrtb_start() < 0) || (grptb_start() < 0)) 686 return(-1); 687 return(0); 688} 689 690/* 691 * ustar_stwr() 692 * initialization for ustar write 693 * Return: 694 * 0 if ok, -1 otherwise 695 */ 696 697int 698ustar_stwr(void) 699{ 700 if ((uidtb_start() < 0) || (gidtb_start() < 0)) 701 return(-1); 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 (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT)) 733 return(-1); 734 return(0); 735} 736 737/* 738 * ustar_rd() 739 * extract the values out of block already determined to be a ustar header. 740 * store the values in the ARCHD parameter. 741 * Return: 742 * 0 743 */ 744 745int 746ustar_rd(ARCHD *arcn, char *buf) 747{ 748 HD_USTAR *hd; 749 char *dest; 750 int cnt = 0; 751 dev_t devmajor; 752 dev_t devminor; 753 754 /* 755 * we only get proper sized buffers 756 */ 757 if (ustar_id(buf, BLKMULT) < 0) 758 return(-1); 759 memset(arcn, 0, sizeof(*arcn)); 760 arcn->org_name = arcn->name; 761 arcn->sb.st_nlink = 1; 762 hd = (HD_USTAR *)buf; 763 764 /* 765 * see if the filename is split into two parts. if, so joint the parts. 766 * we copy the prefix first and add a / between the prefix and name. 767 */ 768 dest = arcn->name; 769 if (*(hd->prefix) != '\0') { 770 cnt = strlcpy(dest, hd->prefix, sizeof(arcn->name) - 1); 771 dest += cnt; 772 *dest++ = '/'; 773 cnt++; 774 } else { 775 cnt = 0; 776 } 777 778 if (hd->typeflag != LONGLINKTYPE && hd->typeflag != LONGNAMETYPE) { 779 arcn->nlen = cnt + expandname(dest, sizeof(arcn->name) - cnt, 780 &gnu_name_string, hd->name, sizeof(hd->name)); 781 arcn->ln_nlen = expandname(arcn->ln_name, sizeof(arcn->ln_name), 782 &gnu_link_string, hd->linkname, sizeof(hd->linkname)); 783 } 784 785 /* 786 * follow the spec to the letter. we should only have mode bits, strip 787 * off all other crud we may be passed. 788 */ 789 arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) & 790 0xfff); 791#ifdef LONG_OFF_T 792 arcn->sb.st_size = (off_t)asc_ul(hd->size, sizeof(hd->size), OCT); 793#else 794 arcn->sb.st_size = (off_t)asc_uqd(hd->size, sizeof(hd->size), OCT); 795#endif 796 arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT); 797 arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime; 798 799 /* 800 * If we can find the ascii names for gname and uname in the password 801 * and group files we will use the uid's and gid they bind. Otherwise 802 * we use the uid and gid values stored in the header. (This is what 803 * the POSIX spec wants). 804 */ 805 hd->gname[sizeof(hd->gname) - 1] = '\0'; 806 if (gid_name(hd->gname, &(arcn->sb.st_gid)) < 0) 807 arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT); 808 hd->uname[sizeof(hd->uname) - 1] = '\0'; 809 if (uid_name(hd->uname, &(arcn->sb.st_uid)) < 0) 810 arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT); 811 812 /* 813 * set the defaults, these may be changed depending on the file type 814 */ 815 arcn->pad = 0; 816 arcn->skip = 0; 817 arcn->sb.st_rdev = (dev_t)0; 818 819 /* 820 * set the mode and PAX type according to the typeflag in the header 821 */ 822 switch (hd->typeflag) { 823 case FIFOTYPE: 824 arcn->type = PAX_FIF; 825 arcn->sb.st_mode |= S_IFIFO; 826 break; 827 case DIRTYPE: 828 arcn->type = PAX_DIR; 829 arcn->sb.st_mode |= S_IFDIR; 830 arcn->sb.st_nlink = 2; 831 832 /* 833 * Some programs that create ustar archives append a '/' 834 * to the pathname for directories. This clearly violates 835 * ustar specs, but we will silently strip it off anyway. 836 */ 837 if (arcn->name[arcn->nlen - 1] == '/') 838 arcn->name[--arcn->nlen] = '\0'; 839 break; 840 case BLKTYPE: 841 case CHRTYPE: 842 /* 843 * this type requires the rdev field to be set. 844 */ 845 if (hd->typeflag == BLKTYPE) { 846 arcn->type = PAX_BLK; 847 arcn->sb.st_mode |= S_IFBLK; 848 } else { 849 arcn->type = PAX_CHR; 850 arcn->sb.st_mode |= S_IFCHR; 851 } 852 devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT); 853 devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT); 854 arcn->sb.st_rdev = TODEV(devmajor, devminor); 855 break; 856 case SYMTYPE: 857 case LNKTYPE: 858 if (hd->typeflag == SYMTYPE) { 859 arcn->type = PAX_SLK; 860 arcn->sb.st_mode |= S_IFLNK; 861 } else { 862 arcn->type = PAX_HLK; 863 /* 864 * so printing looks better 865 */ 866 arcn->sb.st_mode |= S_IFREG; 867 arcn->sb.st_nlink = 2; 868 } 869 break; 870 case LONGLINKTYPE: 871 case LONGNAMETYPE: 872 /* 873 * GNU long link/file; we tag these here and let the 874 * pax internals deal with it -- too ugly otherwise. 875 */ 876 arcn->type = 877 hd->typeflag == LONGLINKTYPE ? PAX_GLL : PAX_GLF; 878 arcn->pad = TAR_PAD(arcn->sb.st_size); 879 arcn->skip = arcn->sb.st_size; 880 break; 881 case CONTTYPE: 882 case AREGTYPE: 883 case REGTYPE: 884 default: 885 /* 886 * these types have file data that follows. Set the skip and 887 * pad fields. 888 */ 889 arcn->type = PAX_REG; 890 arcn->pad = TAR_PAD(arcn->sb.st_size); 891 arcn->skip = arcn->sb.st_size; 892 arcn->sb.st_mode |= S_IFREG; 893 break; 894 } 895 return(0); 896} 897 898/* 899 * ustar_wr() 900 * write a ustar header for the file specified in the ARCHD to the archive 901 * Have to check for file types that cannot be stored and file names that 902 * are too long. Be careful of the term (last arg) to ul_oct, we only use 903 * '\0' for the termination character (this is different than picky tar) 904 * ASSUMED: space after header in header block is zero filled 905 * Return: 906 * 0 if file has data to be written after the header, 1 if file has NO 907 * data to write after the header, -1 if archive write failed 908 */ 909 910int 911ustar_wr(ARCHD *arcn) 912{ 913 HD_USTAR *hd; 914 char *pt; 915 char hdblk[sizeof(HD_USTAR)]; 916 mode_t mode12only; 917 int term_char=3; /* orignal setting */ 918 term_char=1; /* To pass conformance tests 274, 301 */ 919 920 /* 921 * check for those file system types ustar cannot store 922 */ 923 if (arcn->type == PAX_SCK) { 924 paxwarn(1, "Ustar cannot archive a socket %s", arcn->org_name); 925 return(1); 926 } 927 928 /* 929 * check the length of the linkname 930 */ 931 if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) || 932 (arcn->type == PAX_HRG)) && (arcn->ln_nlen > sizeof(hd->linkname))){ 933 paxwarn(1, "Link name too long for ustar %s", arcn->ln_name); 934 /* 935 * Conformance: test pax:285 wants error code to be non-zero, and 936 * test tar:12 wants error code from pax to be 0 937 */ 938 return(1); 939 } 940 941 /* 942 * split the path name into prefix and name fields (if needed). if 943 * pt != arcn->name, the name has to be split 944 */ 945 if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) { 946 paxwarn(1, "File name too long for ustar %s", arcn->name); 947 return(1); 948 } 949 950 /* 951 * zero out the header so we don't have to worry about zero fill below 952 */ 953 memset(hdblk, 0, sizeof(hdblk)); 954 hd = (HD_USTAR *)hdblk; 955 arcn->pad = 0L; 956 957 /* To pass conformance tests 274/301, always set these fields to "zero" */ 958 ul_oct(0, hd->devmajor, sizeof(hd->devmajor), term_char); 959 ul_oct(0, hd->devminor, sizeof(hd->devminor), term_char); 960 961 /* 962 * split the name, or zero out the prefix 963 */ 964 if (pt != arcn->name) { 965 /* 966 * name was split, pt points at the / where the split is to 967 * occur, we remove the / and copy the first part to the prefix 968 */ 969 *pt = '\0'; 970 strlcpy(hd->prefix, arcn->name, sizeof(hd->prefix)); 971 *pt++ = '/'; 972 } 973 974 /* 975 * copy the name part. this may be the whole path or the part after 976 * the prefix. both the name and prefix may fill the entire field. 977 */ 978 if (strlen(pt) == sizeof(hd->name)) { /* must account for name just fits in buffer */ 979 strncpy(hd->name, pt, sizeof(hd->name)); 980 } else { 981 strlcpy(hd->name, pt, sizeof(hd->name)); 982 } 983 984 /* 985 * set the fields in the header that are type dependent 986 */ 987 switch (arcn->type) { 988 case PAX_DIR: 989 hd->typeflag = DIRTYPE; 990 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), term_char)) 991 goto out; 992 break; 993 case PAX_CHR: 994 case PAX_BLK: 995 if (arcn->type == PAX_CHR) 996 hd->typeflag = CHRTYPE; 997 else 998 hd->typeflag = BLKTYPE; 999 if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor, 1000 sizeof(hd->devmajor), term_char) || 1001 ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor, 1002 sizeof(hd->devminor), term_char) || 1003 ul_oct((u_long)0L, hd->size, sizeof(hd->size), term_char)) 1004 goto out; 1005 break; 1006 case PAX_FIF: 1007 hd->typeflag = FIFOTYPE; 1008 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), term_char)) 1009 goto out; 1010 break; 1011 case PAX_SLK: 1012 case PAX_HLK: 1013 case PAX_HRG: 1014 if (arcn->type == PAX_SLK) 1015 hd->typeflag = SYMTYPE; 1016 else 1017 hd->typeflag = LNKTYPE; 1018 if (strlen(arcn->ln_name) == sizeof(hd->linkname)) { /* must account for name just fits in buffer */ 1019 strncpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 1020 } else { 1021 strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname)); 1022 } 1023 if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), term_char)) 1024 goto out; 1025 break; 1026 case PAX_REG: 1027 case PAX_CTG: 1028 default: 1029 /* 1030 * file data with this type, set the padding 1031 */ 1032 if (arcn->type == PAX_CTG) 1033 hd->typeflag = CONTTYPE; 1034 else 1035 hd->typeflag = REGTYPE; 1036 arcn->pad = TAR_PAD(arcn->sb.st_size); 1037# ifdef LONG_OFF_T 1038 if (ul_oct((u_long)arcn->sb.st_size, hd->size, 1039 sizeof(hd->size), term_char)) { 1040# else 1041 if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size, 1042 sizeof(hd->size), term_char)) { 1043# endif 1044 paxwarn(1,"File is too long for ustar %s",arcn->org_name); 1045 return(1); 1046 } 1047 break; 1048 } 1049 1050 strncpy(hd->magic, TMAGIC, TMAGLEN); 1051 strncpy(hd->version, TVERSION, TVERSLEN); 1052 1053 /* 1054 * set the remaining fields. Some versions want all 16 bits of mode 1055 * we better humor them (they really do not meet spec though).... 1056 */ 1057 if (ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), term_char)) { 1058 if (uid_nobody == 0) { 1059 if (uid_name("nobody", &uid_nobody) == -1) 1060 goto out; 1061 } 1062 if (uid_warn != arcn->sb.st_uid) { 1063 uid_warn = arcn->sb.st_uid; 1064 paxwarn(1, 1065 "Ustar header field is too small for uid %lu, " 1066 "using nobody", (u_long)arcn->sb.st_uid); 1067 } 1068 if (ul_oct((u_long)uid_nobody, hd->uid, sizeof(hd->uid), term_char)) 1069 goto out; 1070 } 1071 if (ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), term_char)) { 1072 if (gid_nobody == 0) { 1073 if (gid_name("nobody", &gid_nobody) == -1) 1074 goto out; 1075 } 1076 if (gid_warn != arcn->sb.st_gid) { 1077 gid_warn = arcn->sb.st_gid; 1078 paxwarn(1, 1079 "Ustar header field is too small for gid %lu, " 1080 "using nobody", (u_long)arcn->sb.st_gid); 1081 } 1082 if (ul_oct((u_long)gid_nobody, hd->gid, sizeof(hd->gid), term_char)) 1083 goto out; 1084 } 1085 /* However, Unix conformance tests do not like MORE than 12 mode bits: 1086 remove all beyond (see definition of stat.st_mode structure) */ 1087 mode12only = ((u_long)arcn->sb.st_mode) & 0x00000fff; 1088 if (ul_oct((u_long)mode12only, hd->mode, sizeof(hd->mode), term_char) || 1089 ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),term_char)) 1090 goto out; 1091 strncpy(hd->uname, name_uid(arcn->sb.st_uid, 0), sizeof(hd->uname)); 1092 strncpy(hd->gname, name_gid(arcn->sb.st_gid, 0), sizeof(hd->gname)); 1093 1094 /* 1095 * calculate and store the checksum write the header to the archive 1096 * return 0 tells the caller to now write the file data, 1 says no data 1097 * needs to be written 1098 */ 1099 if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum, 1100 sizeof(hd->chksum), term_char)) 1101 goto out; 1102 if (wr_rdbuf((char *)&hdblk, sizeof(HD_USTAR)) < 0) 1103 return(-1); 1104 if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0) 1105 return(-1); 1106 if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG)) 1107 return(0); 1108 return(1); 1109 1110 out: 1111 /* 1112 * header field is out of range 1113 */ 1114 paxwarn(1, "Ustar header field is too small for %s", arcn->org_name); 1115 return(1); 1116} 1117 1118/* 1119 * name_split() 1120 * see if the name has to be split for storage in a ustar header. We try 1121 * to fit the entire name in the name field without splitting if we can. 1122 * The split point is always at a / 1123 * Return 1124 * character pointer to split point (always the / that is to be removed 1125 * if the split is not needed, the points is set to the start of the file 1126 * name (it would violate the spec to split there). A NULL is returned if 1127 * the file name is too long 1128 */ 1129 1130static char * 1131name_split(char *name, int len) 1132{ 1133 char *start; 1134 1135 /* 1136 * check to see if the file name is small enough to fit in the name 1137 * field. if so just return a pointer to the name. 1138 * The strings can fill the complete name and prefix fields 1139 * without a NUL terminator. 1140 */ 1141 if (len <= TNMSZ) 1142 return(name); 1143 if (len > (TPFSZ + TNMSZ + 1)) 1144 return(NULL); 1145 1146 /* 1147 * we start looking at the biggest sized piece that fits in the name 1148 * field. We walk forward looking for a slash to split at. The idea is 1149 * to find the biggest piece to fit in the name field (or the smallest 1150 * prefix we can find) (the -1 is correct the biggest piece would 1151 * include the slash between the two parts that gets thrown away) 1152 */ 1153 start = name + len - TNMSZ - 1; 1154 if ((*start == '/') && (start == name)) 1155 ++start; /* 101 byte paths with leading '/' are dinged otherwise */ 1156 while ((*start != '\0') && (*start != '/')) 1157 ++start; 1158 1159 /* 1160 * if we hit the end of the string, this name cannot be split, so we 1161 * cannot store this file. 1162 */ 1163 if (*start == '\0') 1164 return(NULL); 1165 len = start - name; 1166 1167 /* 1168 * NOTE: /str where the length of str == TNMSZ can not be stored under 1169 * the p1003.1-1990 spec for ustar. We could force a prefix of / and 1170 * the file would then expand on extract to //str. The len == 0 below 1171 * makes this special case follow the spec to the letter. 1172 */ 1173 if ((len > TPFSZ) || (len == 0)) 1174 return(NULL); 1175 1176 /* 1177 * ok have a split point, return it to the caller 1178 */ 1179 return(start); 1180} 1181 1182static size_t 1183expandname(char *buf, size_t len, char **gnu_name, const char *name, 1184 size_t name_len) 1185{ 1186 size_t nlen; 1187 1188 if (*gnu_name) { 1189 /* *gnu_name is NUL terminated */ 1190 if ((nlen = strlcpy(buf, *gnu_name, len)) >= len) 1191 nlen = len - 1; 1192 free(*gnu_name); 1193 *gnu_name = NULL; 1194 } else { 1195 if (name_len < len) { 1196 /* name may not be null terminated: it might be as big as the 1197 field, so copy is limited to the max size of the header field */ 1198 if ((nlen = strlcpy(buf, name, name_len+1)) >= name_len+1) 1199 nlen = name_len; 1200 } else { 1201 if ((nlen = strlcpy(buf, name, len)) >= len) 1202 nlen = len - 1; 1203 } 1204 } 1205 return(nlen); 1206} 1207