xref: /netbsd-src/bin/pax/tar.c (revision fdecd6a253f999ae92b139670d9e15cc9df4497c)
1 /*	$NetBSD: tar.c,v 1.8 1997/06/25 22:44:00 kleink 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 #ifndef lint
41 #if 0
42 static char sccsid[] = "@(#)tar.c	8.2 (Berkeley) 4/18/94";
43 #else
44 static char rcsid[] = "$NetBSD: tar.c,v 1.8 1997/06/25 22:44:00 kleink Exp $";
45 #endif
46 #endif /* not lint */
47 
48 #include <sys/types.h>
49 #include <sys/time.h>
50 #include <sys/stat.h>
51 #include <sys/param.h>
52 #include <string.h>
53 #include <stdio.h>
54 #include <ctype.h>
55 #include <unistd.h>
56 #include <stdlib.h>
57 #include "pax.h"
58 #include "extern.h"
59 #include "tar.h"
60 
61 /*
62  * Routines for reading, writing and header identify of various versions of tar
63  */
64 
65 static u_long tar_chksm __P((char *, int));
66 static char *name_split __P((char *, int));
67 static int ul_oct __P((u_long, char *, int, int));
68 #ifndef NET2_STAT
69 static int uqd_oct __P((u_quad_t, char *, int, int));
70 #endif
71 
72 /*
73  * Routines common to all versions of tar
74  */
75 
76 static int tar_nodir;			/* do not write dirs under old tar */
77 
78 /*
79  * tar_endwr()
80  *	add the tar trailer of two null blocks
81  * Return:
82  *	0 if ok, -1 otherwise (what wr_skip returns)
83  */
84 
85 #if __STDC__
86 int
87 tar_endwr(void)
88 #else
89 int
90 tar_endwr()
91 #endif
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 
103 #if __STDC__
104 off_t
105 tar_endrd(void)
106 #else
107 off_t
108 tar_endrd()
109 #endif
110 {
111 	return((off_t)(NULLCNT*BLKMULT));
112 }
113 
114 /*
115  * tar_trail()
116  *	Called to determine if a header block is a valid trailer. We are passed
117  *	the block, the in_sync flag (which tells us we are in resync mode;
118  *	looking for a valid header), and cnt (which starts at zero) which is
119  *	used to count the number of empty blocks we have seen so far.
120  * Return:
121  *	0 if a valid trailer, -1 if not a valid trailer, or 1 if the block
122  *	could never contain a header.
123  */
124 
125 #if __STDC__
126 int
127 tar_trail(char *buf, int in_resync, int *cnt)
128 #else
129 int
130 tar_trail(buf, in_resync, cnt)
131 	char *buf;
132 	int in_resync;
133 	int *cnt;
134 #endif
135 {
136 	int i;
137 
138 	/*
139 	 * look for all zero, trailer is two consecutive blocks of zero
140 	 */
141 	for (i = 0; i < BLKMULT; ++i) {
142 		if (buf[i] != '\0')
143 			break;
144 	}
145 
146 	/*
147 	 * if not all zero it is not a trailer, but MIGHT be a header.
148 	 */
149 	if (i != BLKMULT)
150 		return(-1);
151 
152 	/*
153 	 * When given a zero block, we must be careful!
154 	 * If we are not in resync mode, check for the trailer. Have to watch
155 	 * out that we do not mis-identify file data as the trailer, so we do
156 	 * NOT try to id a trailer during resync mode. During resync mode we
157 	 * might as well throw this block out since a valid header can NEVER be
158 	 * a block of all 0 (we must have a valid file name).
159 	 */
160 	if (!in_resync && (++*cnt >= NULLCNT))
161 		return(0);
162 	return(1);
163 }
164 
165 /*
166  * ul_oct()
167  *	convert an unsigned long to an octal string. many oddball field
168  *	termination characters are used by the various versions of tar in the
169  *	different fields. term selects which kind to use. str is '0' padded
170  *	at the front to len. we are unable to use only one format as many old
171  *	tar readers are very cranky about this.
172  * Return:
173  *	0 if the number fit into the string, -1 otherwise
174  */
175 
176 #if __STDC__
177 static int
178 ul_oct(u_long val, char *str, int len, int term)
179 #else
180 static int
181 ul_oct(val, str, len, term)
182 	u_long val;
183 	char *str;
184 	int len;
185 	int term;
186 #endif
187 {
188 	char *pt;
189 
190 	/*
191 	 * term selects the appropriate character(s) for the end of the string
192 	 */
193 	pt = str + len - 1;
194 	switch(term) {
195 	case 3:
196 		*pt-- = '\0';
197 		break;
198 	case 2:
199 		*pt-- = ' ';
200 		*pt-- = '\0';
201 		break;
202 	case 1:
203 		*pt-- = ' ';
204 		break;
205 	case 0:
206 	default:
207 		*pt-- = '\0';
208 		*pt-- = ' ';
209 		break;
210 	}
211 
212 	/*
213 	 * convert and blank pad if there is space
214 	 */
215 	while (pt >= str) {
216 		*pt-- = '0' + (char)(val & 0x7);
217 		if ((val = val >> 3) == (u_long)0)
218 			break;
219 	}
220 
221 	while (pt >= str)
222 		*pt-- = '0';
223 	if (val != (u_long)0)
224 		return(-1);
225 	return(0);
226 }
227 
228 #ifndef NET2_STAT
229 /*
230  * uqd_oct()
231  *	convert an u_quad_t to an octal string. one of many oddball field
232  *	termination characters are used by the various versions of tar in the
233  *	different fields. term selects which kind to use. str is '0' padded
234  *	at the front to len. we are unable to use only one format as many old
235  *	tar readers are very cranky about this.
236  * Return:
237  *	0 if the number fit into the string, -1 otherwise
238  */
239 
240 #if __STDC__
241 static int
242 uqd_oct(u_quad_t val, char *str, int len, int term)
243 #else
244 static int
245 uqd_oct(val, str, len, term)
246 	u_quad_t val;
247 	char *str;
248 	int len;
249 	int term;
250 #endif
251 {
252 	char *pt;
253 
254 	/*
255 	 * term selects the appropriate character(s) for the end of the string
256 	 */
257 	pt = str + len - 1;
258 	switch(term) {
259 	case 3:
260 		*pt-- = '\0';
261 		break;
262 	case 2:
263 		*pt-- = ' ';
264 		*pt-- = '\0';
265 		break;
266 	case 1:
267 		*pt-- = ' ';
268 		break;
269 	case 0:
270 	default:
271 		*pt-- = '\0';
272 		*pt-- = ' ';
273 		break;
274 	}
275 
276 	/*
277 	 * convert and blank pad if there is space
278 	 */
279 	while (pt >= str) {
280 		*pt-- = '0' + (char)(val & 0x7);
281 		if ((val = val >> 3) == 0)
282 			break;
283 	}
284 
285 	while (pt >= str)
286 		*pt-- = '0';
287 	if (val != (u_quad_t)0)
288 		return(-1);
289 	return(0);
290 }
291 #endif
292 
293 /*
294  * tar_chksm()
295  *	calculate the checksum for a tar block counting the checksum field as
296  *	all blanks (BLNKSUM is that value pre-calculated, the sume of 8 blanks).
297  *	NOTE: we use len to short circuit summing 0's on write since we ALWAYS
298  *	pad headers with 0.
299  * Return:
300  *	unsigned long checksum
301  */
302 
303 #if __STDC__
304 static u_long
305 tar_chksm(char *blk, int len)
306 #else
307 static u_long
308 tar_chksm(blk, len)
309 	char *blk;
310 	int len;
311 #endif
312 {
313 	char *stop;
314 	char *pt;
315 	u_long chksm = BLNKSUM;	/* inital value is checksum field sum */
316 
317 	/*
318 	 * add the part of the block before the checksum field
319 	 */
320 	pt = blk;
321 	stop = blk + CHK_OFFSET;
322 	while (pt < stop)
323 		chksm += (u_long)(*pt++ & 0xff);
324 	/*
325 	 * move past the checksum field and keep going, spec counts the
326 	 * checksum field as the sum of 8 blanks (which is pre-computed as
327 	 * BLNKSUM).
328 	 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding
329 	 * starts, no point in summing zero's)
330 	 */
331 	pt += CHK_LEN;
332 	stop = blk + len;
333 	while (pt < stop)
334 		chksm += (u_long)(*pt++ & 0xff);
335 	return(chksm);
336 }
337 
338 /*
339  * Routines for old BSD style tar (also made portable to sysV tar)
340  */
341 
342 /*
343  * tar_id()
344  *	determine if a block given to us is a valid tar header (and not a USTAR
345  *	header). We have to be on the lookout for those pesky blocks of	all
346  *	zero's.
347  * Return:
348  *	0 if a tar header, -1 otherwise
349  */
350 
351 #if __STDC__
352 int
353 tar_id(char *blk, int size)
354 #else
355 int
356 tar_id(blk, size)
357 	char *blk;
358 	int size;
359 #endif
360 {
361 	HD_TAR *hd;
362 	HD_USTAR *uhd;
363 
364 	if (size < BLKMULT)
365 		return(-1);
366 	hd = (HD_TAR *)blk;
367 	uhd = (HD_USTAR *)blk;
368 
369 	/*
370 	 * check for block of zero's first, a simple and fast test, then make
371 	 * sure this is not a ustar header by looking for the ustar magic
372 	 * cookie. We should use TMAGLEN, but some USTAR archive programs are
373 	 * wrong and create archives missing the \0. Last we check the
374 	 * checksum. If this is ok we have to assume it is a valid header.
375 	 */
376 	if (hd->name[0] == '\0')
377 		return(-1);
378 	if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0)
379 		return(-1);
380 	if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT))
381 		return(-1);
382 	return(0);
383 }
384 
385 /*
386  * tar_opt()
387  *	handle tar format specific -o options
388  * Return:
389  *	0 if ok -1 otherwise
390  */
391 
392 #if __STDC__
393 int
394 tar_opt(void)
395 #else
396 int
397 tar_opt()
398 #endif
399 {
400 	OPLIST *opt;
401 
402 	while ((opt = opt_next()) != NULL) {
403 		if (strcmp(opt->name, TAR_OPTION) ||
404 		    strcmp(opt->value, TAR_NODIR)) {
405 			warn(1, "Unknown tar format -o option/value pair %s=%s",
406 			    opt->name, opt->value);
407 			warn(1,"%s=%s is the only supported tar format option",
408 			    TAR_OPTION, TAR_NODIR);
409 			return(-1);
410 		}
411 
412 		/*
413 		 * we only support one option, and only when writing
414 		 */
415 		if ((act != APPND) && (act != ARCHIVE)) {
416 			warn(1, "%s=%s is only supported when writing.",
417 			    opt->name, opt->value);
418 			return(-1);
419 		}
420 		tar_nodir = 1;
421 	}
422 	return(0);
423 }
424 
425 
426 /*
427  * tar_rd()
428  *	extract the values out of block already determined to be a tar header.
429  *	store the values in the ARCHD parameter.
430  * Return:
431  *	0
432  */
433 
434 #if __STDC__
435 int
436 tar_rd(ARCHD *arcn, char *buf)
437 #else
438 int
439 tar_rd(arcn, buf)
440 	ARCHD *arcn;
441 	char *buf;
442 #endif
443 {
444 	HD_TAR *hd;
445 	char *pt;
446 
447 	/*
448 	 * we only get proper sized buffers passed to us
449 	 */
450 	if (tar_id(buf, BLKMULT) < 0)
451 		return(-1);
452 	arcn->org_name = arcn->name;
453 	arcn->sb.st_nlink = 1;
454 	arcn->pat = NULL;
455 
456 	/*
457 	 * copy out the name and values in the stat buffer
458 	 */
459 	hd = (HD_TAR *)buf;
460 	arcn->nlen = l_strncpy(arcn->name, hd->name, sizeof(hd->name));
461 	arcn->name[arcn->nlen] = '\0';
462 	arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) &
463 	    0xfff);
464 	arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT);
465 	arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT);
466 	arcn->sb.st_size = (size_t)asc_ul(hd->size, sizeof(hd->size), OCT);
467 	arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT);
468 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
469 
470 	/*
471 	 * have to look at the last character, it may be a '/' and that is used
472 	 * to encode this as a directory
473 	 */
474 	pt = &(arcn->name[arcn->nlen - 1]);
475 	arcn->pad = 0;
476 	arcn->skip = 0;
477 	switch(hd->linkflag) {
478 	case SYMTYPE:
479 		/*
480 		 * symbolic link, need to get the link name and set the type in
481 		 * the st_mode so -v printing will look correct.
482 		 */
483 		arcn->type = PAX_SLK;
484 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
485 			sizeof(hd->linkname));
486 		arcn->ln_name[arcn->ln_nlen] = '\0';
487 		arcn->sb.st_mode |= S_IFLNK;
488 		break;
489 	case LNKTYPE:
490 		/*
491 		 * hard link, need to get the link name, set the type in the
492 		 * st_mode and st_nlink so -v printing will look better.
493 		 */
494 		arcn->type = PAX_HLK;
495 		arcn->sb.st_nlink = 2;
496 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
497 			sizeof(hd->linkname));
498 		arcn->ln_name[arcn->ln_nlen] = '\0';
499 
500 		/*
501 		 * no idea of what type this thing really points at, but
502 		 * we set something for printing only.
503 		 */
504 		arcn->sb.st_mode |= S_IFREG;
505 		break;
506 	case AREGTYPE:
507 	case REGTYPE:
508 	default:
509 		/*
510 		 * If we have a trailing / this is a directory and NOT a file.
511 		 */
512 		arcn->ln_name[0] = '\0';
513 		arcn->ln_nlen = 0;
514 		if (*pt == '/') {
515 			/*
516 			 * it is a directory, set the mode for -v printing
517 			 */
518 			arcn->type = PAX_DIR;
519 			arcn->sb.st_mode |= S_IFDIR;
520 			arcn->sb.st_nlink = 2;
521 		} else {
522 			/*
523 			 * have a file that will be followed by data. Set the
524 			 * skip value to the size field and caluculate the size
525 			 * of the padding.
526 			 */
527 			arcn->type = PAX_REG;
528 			arcn->sb.st_mode |= S_IFREG;
529 			arcn->pad = TAR_PAD(arcn->sb.st_size);
530 			arcn->skip = arcn->sb.st_size;
531 		}
532 		break;
533 	}
534 
535 	/*
536 	 * strip off any trailing slash.
537 	 */
538 	if (*pt == '/') {
539 		*pt = '\0';
540 		--arcn->nlen;
541 	}
542 	return(0);
543 }
544 
545 /*
546  * tar_wr()
547  *	write a tar header for the file specified in the ARCHD to the archive.
548  *	Have to check for file types that cannot be stored and file names that
549  *	are too long. Be careful of the term (last arg) to ul_oct, each field
550  *	of tar has it own spec for the termination character(s).
551  *	ASSUMED: space after header in header block is zero filled
552  * Return:
553  *	0 if file has data to be written after the header, 1 if file has NO
554  *	data to write after the header, -1 if archive write failed
555  */
556 
557 #if __STDC__
558 int
559 tar_wr(ARCHD *arcn)
560 #else
561 int
562 tar_wr(arcn)
563 	ARCHD *arcn;
564 #endif
565 {
566 	HD_TAR *hd;
567 	int len;
568 	char hdblk[sizeof(HD_TAR)];
569 
570 	/*
571 	 * check for those file system types which tar cannot store
572 	 */
573 	switch(arcn->type) {
574 	case PAX_DIR:
575 		/*
576 		 * user asked that dirs not be written to the archive
577 		 */
578 		if (tar_nodir)
579 			return(1);
580 		break;
581 	case PAX_CHR:
582 		warn(1, "Tar cannot archive a character device %s",
583 		    arcn->org_name);
584 		return(1);
585 	case PAX_BLK:
586 		warn(1, "Tar cannot archive a block device %s", arcn->org_name);
587 		return(1);
588 	case PAX_SCK:
589 		warn(1, "Tar cannot archive a socket %s", arcn->org_name);
590 		return(1);
591 	case PAX_FIF:
592 		warn(1, "Tar cannot archive a fifo %s", arcn->org_name);
593 		return(1);
594 	case PAX_SLK:
595 	case PAX_HLK:
596 	case PAX_HRG:
597 		if (arcn->ln_nlen > sizeof(hd->linkname)) {
598 			warn(1,"Link name too long for tar %s", arcn->ln_name);
599 			return(1);
600 		}
601 		break;
602 	case PAX_REG:
603 	case PAX_CTG:
604 	default:
605 		break;
606 	}
607 
608 	/*
609 	 * check file name len, remember extra char for dirs (the / at the end)
610 	 */
611 	len = arcn->nlen;
612 	if (arcn->type == PAX_DIR)
613 		++len;
614 	if (len > sizeof(hd->name)) {
615 		warn(1, "File name too long for tar %s", arcn->name);
616 		return(1);
617 	}
618 
619 	/*
620 	 * copy the data out of the ARCHD into the tar header based on the type
621 	 * of the file. Remember many tar readers want the unused fields to be
622 	 * padded with zero. We set the linkflag field (type), the linkname
623 	 * (or zero if not used),the size, and set the padding (if any) to be
624 	 * added after the file data (0 for all other types, as they only have
625 	 * a header)
626 	 */
627 	hd = (HD_TAR *)hdblk;
628 	zf_strncpy(hd->name, arcn->name, sizeof(hd->name));
629 	arcn->pad = 0;
630 
631 	if (arcn->type == PAX_DIR) {
632 		/*
633 		 * directories are the same as files, except have a filename
634 		 * that ends with a /, we add the slash here. No data follows,
635 		 * dirs, so no pad.
636 		 */
637 		hd->linkflag = AREGTYPE;
638 		memset(hd->linkname, 0, sizeof(hd->linkname));
639 		hd->name[len-1] = '/';
640 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
641 			goto out;
642 	} else if (arcn->type == PAX_SLK) {
643 		/*
644 		 * no data follows this file, so no pad
645 		 */
646 		hd->linkflag = SYMTYPE;
647 		zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname));
648 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
649 			goto out;
650 	} else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) {
651 		/*
652 		 * no data follows this file, so no pad
653 		 */
654 		hd->linkflag = LNKTYPE;
655 		zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname));
656 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
657 			goto out;
658 	} else {
659 		/*
660 		 * data follows this file, so set the pad
661 		 */
662 		hd->linkflag = AREGTYPE;
663 		memset(hd->linkname, 0, sizeof(hd->linkname));
664 #		ifdef NET2_STAT
665 		if (ul_oct((u_long)arcn->sb.st_size, hd->size,
666 		    sizeof(hd->size), 1)) {
667 #		else
668 		if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size,
669 		    sizeof(hd->size), 1)) {
670 #		endif
671 			warn(1,"File is too large for tar %s", arcn->org_name);
672 			return(1);
673 		}
674 		arcn->pad = TAR_PAD(arcn->sb.st_size);
675 	}
676 
677 	/*
678 	 * copy those fields that are independent of the type
679 	 */
680 	if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) ||
681 	    ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) ||
682 	    ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) ||
683 	    ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1))
684 		goto out;
685 
686 	/*
687 	 * calculate and add the checksum, then write the header. A return of
688 	 * 0 tells the caller to now write the file data, 1 says no data needs
689 	 * to be written
690 	 */
691 	if (ul_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum,
692 	    sizeof(hd->chksum), 2))
693 		goto out;
694 	if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0)
695 		return(-1);
696 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0)
697 		return(-1);
698 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
699 		return(0);
700 	return(1);
701 
702     out:
703 	/*
704 	 * header field is out of range
705 	 */
706 	warn(1, "Tar header field is too small for %s", arcn->org_name);
707 	return(1);
708 }
709 
710 /*
711  * Routines for POSIX ustar
712  */
713 
714 /*
715  * ustar_strd()
716  *	initialization for ustar read
717  * Return:
718  *	0 if ok, -1 otherwise
719  */
720 
721 #if __STDC__
722 int
723 ustar_strd(void)
724 #else
725 int
726 ustar_strd()
727 #endif
728 {
729 	if ((usrtb_start() < 0) || (grptb_start() < 0))
730 		return(-1);
731 	return(0);
732 }
733 
734 /*
735  * ustar_stwr()
736  *	initialization for ustar write
737  * Return:
738  *	0 if ok, -1 otherwise
739  */
740 
741 #if __STDC__
742 int
743 ustar_stwr(void)
744 #else
745 int
746 ustar_stwr()
747 #endif
748 {
749 	if ((uidtb_start() < 0) || (gidtb_start() < 0))
750 		return(-1);
751 	return(0);
752 }
753 
754 /*
755  * ustar_id()
756  *	determine if a block given to us is a valid ustar header. We have to
757  *	be on the lookout for those pesky blocks of all zero's
758  * Return:
759  *	0 if a ustar header, -1 otherwise
760  */
761 
762 #if __STDC__
763 int
764 ustar_id(char *blk, int size)
765 #else
766 int
767 ustar_id(blk, size)
768 	char *blk;
769 	int size;
770 #endif
771 {
772 	HD_USTAR *hd;
773 
774 	if (size < BLKMULT)
775 		return(-1);
776 	hd = (HD_USTAR *)blk;
777 
778 	/*
779 	 * check for block of zero's first, a simple and fast test then check
780 	 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive
781 	 * programs are fouled up and create archives missing the \0. Last we
782 	 * check the checksum. If ok we have to assume it is a valid header.
783 	 */
784 	if (hd->name[0] == '\0')
785 		return(-1);
786 	if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0)
787 		return(-1);
788 	if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT))
789 		return(-1);
790 	return(0);
791 }
792 
793 /*
794  * ustar_rd()
795  *	extract the values out of block already determined to be a ustar header.
796  *	store the values in the ARCHD parameter.
797  * Return:
798  *	0
799  */
800 
801 #if __STDC__
802 int
803 ustar_rd(ARCHD *arcn, char *buf)
804 #else
805 int
806 ustar_rd(arcn, buf)
807 	ARCHD *arcn;
808 	char *buf;
809 #endif
810 {
811 	HD_USTAR *hd;
812 	char *dest;
813 	int cnt = 0;
814 	dev_t devmajor;
815 	dev_t devminor;
816 
817 	/*
818 	 * we only get proper sized buffers
819 	 */
820 	if (ustar_id(buf, BLKMULT) < 0)
821 		return(-1);
822 	arcn->org_name = arcn->name;
823 	arcn->sb.st_nlink = 1;
824 	arcn->pat = NULL;
825 	hd = (HD_USTAR *)buf;
826 
827 	/*
828 	 * see if the filename is split into two parts. if, so joint the parts.
829 	 * we copy the prefix first and add a / between the prefix and name.
830 	 */
831 	dest = arcn->name;
832 	if (*(hd->prefix) != '\0') {
833 		cnt = l_strncpy(arcn->name, hd->prefix, sizeof(hd->prefix));
834 		dest = arcn->name + arcn->nlen;
835 		*dest++ = '/';
836 		cnt++;
837 	}
838 	arcn->nlen = l_strncpy(dest, hd->name, sizeof(hd->name));
839 	arcn->nlen += cnt;
840 	arcn->name[arcn->nlen] = '\0';
841 
842 	/*
843 	 * follow the spec to the letter. we should only have mode bits, strip
844 	 * off all other crud we may be passed.
845 	 */
846 	arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) &
847 	    0xfff);
848 	arcn->sb.st_size = (size_t)asc_ul(hd->size, sizeof(hd->size), OCT);
849 	arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT);
850 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
851 
852 	/*
853 	 * If we can find the ascii names for gname and uname in the password
854 	 * and group files we will use the uid's and gid they bind. Otherwise
855 	 * we use the uid and gid values stored in the header. (This is what
856 	 * the posix spec wants).
857 	 */
858 	hd->gname[sizeof(hd->gname) - 1] = '\0';
859 	if (gid_name(hd->gname, &(arcn->sb.st_gid)) < 0)
860 		arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT);
861 	hd->uname[sizeof(hd->uname) - 1] = '\0';
862 	if (uid_name(hd->uname, &(arcn->sb.st_uid)) < 0)
863 		arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT);
864 
865 	/*
866 	 * set the defaults, these may be changed depending on the file type
867 	 */
868 	arcn->ln_name[0] = '\0';
869 	arcn->ln_nlen = 0;
870 	arcn->pad = 0;
871 	arcn->skip = 0;
872 	arcn->sb.st_rdev = (dev_t)0;
873 
874 	/*
875 	 * set the mode and PAX type according to the typeflag in the header
876 	 */
877 	switch(hd->typeflag) {
878 	case FIFOTYPE:
879 		arcn->type = PAX_FIF;
880 		arcn->sb.st_mode |= S_IFIFO;
881 		break;
882 	case DIRTYPE:
883 		arcn->type = PAX_DIR;
884 		arcn->sb.st_mode |= S_IFDIR;
885 		arcn->sb.st_nlink = 2;
886 
887 		/*
888 		 * Some programs that create ustar archives append a '/'
889 		 * to the pathname for directories. This clearly violates
890 		 * ustar specs, but we will silently strip it off anyway.
891 		 */
892 		if (arcn->name[arcn->nlen - 1] == '/')
893 			arcn->name[--arcn->nlen] = '\0';
894 		break;
895 	case BLKTYPE:
896 	case CHRTYPE:
897 		/*
898 		 * this type requires the rdev field to be set.
899 		 */
900 		if (hd->typeflag == BLKTYPE) {
901 			arcn->type = PAX_BLK;
902 			arcn->sb.st_mode |= S_IFBLK;
903 		} else {
904 			arcn->type = PAX_CHR;
905 			arcn->sb.st_mode |= S_IFCHR;
906 		}
907 		devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT);
908 		devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT);
909 		arcn->sb.st_rdev = TODEV(devmajor, devminor);
910 		break;
911 	case SYMTYPE:
912 	case LNKTYPE:
913 		if (hd->typeflag == SYMTYPE) {
914 			arcn->type = PAX_SLK;
915 			arcn->sb.st_mode |= S_IFLNK;
916 		} else {
917 			arcn->type = PAX_HLK;
918 			/*
919 			 * so printing looks better
920 			 */
921 			arcn->sb.st_mode |= S_IFREG;
922 			arcn->sb.st_nlink = 2;
923 		}
924 		/*
925 		 * copy the link name
926 		 */
927 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
928 			sizeof(hd->linkname));
929 		arcn->ln_name[arcn->ln_nlen] = '\0';
930 		break;
931 	case CONTTYPE:
932 	case AREGTYPE:
933 	case REGTYPE:
934 	default:
935 		/*
936 		 * these types have file data that follows. Set the skip and
937 		 * pad fields.
938 		 */
939 		arcn->type = PAX_REG;
940 		arcn->pad = TAR_PAD(arcn->sb.st_size);
941 		arcn->skip = arcn->sb.st_size;
942 		arcn->sb.st_mode |= S_IFREG;
943 		break;
944 	}
945 	return(0);
946 }
947 
948 /*
949  * ustar_wr()
950  *	write a ustar header for the file specified in the ARCHD to the archive
951  *	Have to check for file types that cannot be stored and file names that
952  *	are too long. Be careful of the term (last arg) to ul_oct, we only use
953  *	'\0' for the termination character (this is different than picky tar)
954  *	ASSUMED: space after header in header block is zero filled
955  * Return:
956  *	0 if file has data to be written after the header, 1 if file has NO
957  *	data to write after the header, -1 if archive write failed
958  */
959 
960 #if __STDC__
961 int
962 ustar_wr(ARCHD *arcn)
963 #else
964 int
965 ustar_wr(arcn)
966 	ARCHD *arcn;
967 #endif
968 {
969 	HD_USTAR *hd;
970 	char *pt;
971 	char hdblk[sizeof(HD_USTAR)];
972 
973 	/*
974 	 * check for those file system types ustar cannot store
975 	 */
976 	if (arcn->type == PAX_SCK) {
977 		warn(1, "Ustar cannot archive a socket %s", arcn->org_name);
978 		return(1);
979 	}
980 
981 	/*
982 	 * check the length of the linkname
983 	 */
984 	if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
985 	    (arcn->type == PAX_HRG)) && (arcn->ln_nlen > sizeof(hd->linkname))){
986 		warn(1, "Link name too long for ustar %s", arcn->ln_name);
987 		return(1);
988 	}
989 
990 	/*
991 	 * split the path name into prefix and name fields (if needed). if
992 	 * pt != arcn->name, the name has to be split
993 	 */
994 	if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) {
995 		warn(1, "File name too long for ustar %s", arcn->name);
996 		return(1);
997 	}
998 	hd = (HD_USTAR *)hdblk;
999 	arcn->pad = 0L;
1000 
1001 	/*
1002 	 * split the name, or zero out the prefix
1003 	 */
1004 	if (pt != arcn->name) {
1005 		/*
1006 		 * name was split, pt points at the / where the split is to
1007 		 * occur, we remove the / and copy the first part to the prefix
1008 		 */
1009 		*pt = '\0';
1010 		zf_strncpy(hd->prefix, arcn->name, sizeof(hd->prefix));
1011 		*pt++ = '/';
1012 	} else
1013 		memset(hd->prefix, 0, sizeof(hd->prefix));
1014 
1015 	/*
1016 	 * copy the name part. this may be the whole path or the part after
1017 	 * the prefix
1018 	 */
1019 	zf_strncpy(hd->name, pt, sizeof(hd->name));
1020 
1021 	/*
1022 	 * set the fields in the header that are type dependent
1023 	 */
1024 	switch(arcn->type) {
1025 	case PAX_DIR:
1026 		hd->typeflag = DIRTYPE;
1027 		memset(hd->linkname, 0, sizeof(hd->linkname));
1028 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1029 		memset(hd->devminor, 0, sizeof(hd->devminor));
1030 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1031 			goto out;
1032 		break;
1033 	case PAX_CHR:
1034 	case PAX_BLK:
1035 		if (arcn->type == PAX_CHR)
1036 			hd->typeflag = CHRTYPE;
1037 		else
1038 			hd->typeflag = BLKTYPE;
1039 		memset(hd->linkname, 0, sizeof(hd->linkname));
1040 		if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor,
1041 		   sizeof(hd->devmajor), 3) ||
1042 		   ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor,
1043 		   sizeof(hd->devminor), 3) ||
1044 		   ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1045 			goto out;
1046 		break;
1047 	case PAX_FIF:
1048 		hd->typeflag = FIFOTYPE;
1049 		memset(hd->linkname, 0, sizeof(hd->linkname));
1050 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1051 		memset(hd->devminor, 0, sizeof(hd->devminor));
1052 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1053 			goto out;
1054 		break;
1055 	case PAX_SLK:
1056 	case PAX_HLK:
1057 	case PAX_HRG:
1058 		if (arcn->type == PAX_SLK)
1059 			hd->typeflag = SYMTYPE;
1060 		else
1061 			hd->typeflag = LNKTYPE;
1062 		zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname));
1063 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1064 		memset(hd->devminor, 0, sizeof(hd->devminor));
1065 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1066 			goto out;
1067 		break;
1068 	case PAX_REG:
1069 	case PAX_CTG:
1070 	default:
1071 		/*
1072 		 * file data with this type, set the padding
1073 		 */
1074 		if (arcn->type == PAX_CTG)
1075 			hd->typeflag = CONTTYPE;
1076 		else
1077 			hd->typeflag = REGTYPE;
1078 		memset(hd->linkname, 0, sizeof(hd->linkname));
1079 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1080 		memset(hd->devminor, 0, sizeof(hd->devminor));
1081 		arcn->pad = TAR_PAD(arcn->sb.st_size);
1082 #		ifdef NET2_STAT
1083 		if (ul_oct((u_long)arcn->sb.st_size, hd->size,
1084 		    sizeof(hd->size), 3)) {
1085 #		else
1086 		if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size,
1087 		    sizeof(hd->size), 3)) {
1088 #		endif
1089 			warn(1,"File is too long for ustar %s",arcn->org_name);
1090 			return(1);
1091 		}
1092 		break;
1093 	}
1094 
1095 	zf_strncpy(hd->magic, TMAGIC, TMAGLEN);
1096 	zf_strncpy(hd->version, TVERSION, TVERSLEN);
1097 
1098 	/*
1099 	 * set the remaining fields. Some versions want all 16 bits of mode
1100 	 * we better humor them (they really do not meet spec though)....
1101 	 */
1102 	if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) ||
1103 	    ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3)  ||
1104 	    ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) ||
1105 	    ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3))
1106 		goto out;
1107 	zf_strncpy(hd->uname,name_uid(arcn->sb.st_uid, 0),sizeof(hd->uname));
1108 	zf_strncpy(hd->gname,name_gid(arcn->sb.st_gid, 0),sizeof(hd->gname));
1109 
1110 	/*
1111 	 * calculate and store the checksum write the header to the archive
1112 	 * return 0 tells the caller to now write the file data, 1 says no data
1113 	 * needs to be written
1114 	 */
1115 	if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum,
1116 	   sizeof(hd->chksum), 3))
1117 		goto out;
1118 	if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0)
1119 		return(-1);
1120 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0)
1121 		return(-1);
1122 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
1123 		return(0);
1124 	return(1);
1125 
1126     out:
1127     	/*
1128 	 * header field is out of range
1129 	 */
1130 	warn(1, "Ustar header field is too small for %s", arcn->org_name);
1131 	return(1);
1132 }
1133 
1134 /*
1135  * name_split()
1136  *	see if the name has to be split for storage in a ustar header. We try
1137  *	to fit the entire name in the name field without splitting if we can.
1138  *	The split point is always at a /
1139  * Return
1140  *	character pointer to split point (always the / that is to be removed
1141  *	if the split is not needed, the points is set to the start of the file
1142  *	name (it would violate the spec to split there). A NULL is returned if
1143  *	the file name is too long
1144  */
1145 
1146 #if __STDC__
1147 static char *
1148 name_split(char *name, int len)
1149 #else
1150 static char *
1151 name_split(name, len)
1152 	char *name;
1153 	int len;
1154 #endif
1155 {
1156 	char *start;
1157 
1158 	/*
1159 	 * check to see if the file name is small enough to fit in the name
1160 	 * field. if so just return a pointer to the name.
1161 	 */
1162 	if (len <= TNMSZ)
1163 		return(name);
1164 	if (len > (TPFSZ + TNMSZ + 1))
1165 		return(NULL);
1166 
1167 	/*
1168 	 * we start looking at the biggest sized piece that fits in the name
1169 	 * field. We walk foward looking for a slash to split at. The idea is
1170 	 * to find the biggest piece to fit in the name field (or the smallest
1171 	 * prefix we can find) (the -1 is correct the biggest piece would
1172 	 * include the slash between the two parts that gets thrown away)
1173 	 */
1174 	start = name + len - TNMSZ - 1;
1175 	while ((*start != '\0') && (*start != '/'))
1176 		++start;
1177 
1178 	/*
1179 	 * if we hit the end of the string, this name cannot be split, so we
1180 	 * cannot store this file.
1181 	 */
1182 	if (*start == '\0')
1183 		return(NULL);
1184 	len = start - name;
1185 
1186 	/*
1187 	 * NOTE: /str where the length of str == TNMSZ can not be stored under
1188 	 * the p1003.1-1990 spec for ustar. We could force a prefix of / and
1189 	 * the file would then expand on extract to //str. The len == 0 below
1190 	 * makes this special case follow the spec to the letter.
1191 	 */
1192 	if ((len > TPFSZ) || (len == 0))
1193 		return(NULL);
1194 
1195 	/*
1196 	 * ok have a split point, return it to the caller
1197 	 */
1198 	return(start);
1199 }
1200