xref: /openbsd/bin/pax/tar.c (revision 07ea8d15)
1 /*	$OpenBSD: tar.c,v 1.4 1996/08/27 03:53:14 tholo 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. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  */
40 
41 #ifndef lint
42 #if 0
43 static char sccsid[] = "@(#)tar.c	8.2 (Berkeley) 4/18/94";
44 #else
45 static char rcsid[] = "$OpenBSD: tar.c,v 1.4 1996/08/27 03:53:14 tholo 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 #include <string.h>
54 #include <stdio.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((register char *, register int));
66 static char *name_split __P((register char *, register int));
67 static int ul_oct __P((u_long, register char *, register int, int));
68 #ifndef NET2_STAT
69 static int uqd_oct __P((u_quad_t, register char *, register 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(register char *buf, register int in_resync, register int *cnt)
128 #else
129 int
130 tar_trail(buf, in_resync, cnt)
131 	register char *buf;
132 	register int in_resync;
133 	register int *cnt;
134 #endif
135 {
136 	register 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 BLANK 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, register char *str, register int len, int term)
179 #else
180 static int
181 ul_oct(val, str, len, term)
182 	u_long val;
183 	register char *str;
184 	register int len;
185 	int term;
186 #endif
187 {
188 	register 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-- = ' ';
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 BLANK 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, register char *str, register int len, int term)
243 #else
244 static int
245 uqd_oct(val, str, len, term)
246 	u_quad_t val;
247 	register char *str;
248 	register int len;
249 	int term;
250 #endif
251 {
252 	register 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-- = ' ';
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(register char *blk, register int len)
306 #else
307 static u_long
308 tar_chksm(blk, len)
309 	register char *blk;
310 	register int len;
311 #endif
312 {
313 	register char *stop;
314 	register 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(register char *blk, int size)
354 #else
355 int
356 tar_id(blk, size)
357 	register char *blk;
358 	int size;
359 #endif
360 {
361 	register HD_TAR *hd;
362 	register 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 			paxwarn(1, "Unknown tar format -o option/value pair %s=%s",
406 			    opt->name, opt->value);
407 			paxwarn(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 			paxwarn(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(register ARCHD *arcn, register char *buf)
437 #else
438 int
439 tar_rd(arcn, buf)
440 	register ARCHD *arcn;
441 	register char *buf;
442 #endif
443 {
444 	register HD_TAR *hd;
445 	register 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) - 1);
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) - 1);
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) - 1);
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(register ARCHD *arcn)
560 #else
561 int
562 tar_wr(arcn)
563 	register ARCHD *arcn;
564 #endif
565 {
566 	register 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 		paxwarn(1, "Tar cannot archive a character device %s",
583 		    arcn->org_name);
584 		return(1);
585 	case PAX_BLK:
586 		paxwarn(1, "Tar cannot archive a block device %s", arcn->org_name);
587 		return(1);
588 	case PAX_SCK:
589 		paxwarn(1, "Tar cannot archive a socket %s", arcn->org_name);
590 		return(1);
591 	case PAX_FIF:
592 		paxwarn(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 			paxwarn(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 		paxwarn(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) - 1);
629 	arcn->name[sizeof(hd->name) - 1] = '\0';
630 	arcn->pad = 0;
631 
632 	if (arcn->type == PAX_DIR) {
633 		/*
634 		 * directories are the same as files, except have a filename
635 		 * that ends with a /, we add the slash here. No data follows,
636 		 * dirs, so no pad.
637 		 */
638 		hd->linkflag = AREGTYPE;
639 		memset(hd->linkname, 0, sizeof(hd->linkname));
640 		hd->name[len-1] = '/';
641 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
642 			goto out;
643 	} else if (arcn->type == PAX_SLK) {
644 		/*
645 		 * no data follows this file, so no pad
646 		 */
647 		hd->linkflag = SYMTYPE;
648 		zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname) - 1);
649 		hd->linkname[sizeof(hd->linkname) - 1] = '\0';
650 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
651 			goto out;
652 	} else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) {
653 		/*
654 		 * no data follows this file, so no pad
655 		 */
656 		hd->linkflag = LNKTYPE;
657 		zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname) - 1);
658 		hd->linkname[sizeof(hd->linkname) - 1] = '\0';
659 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
660 			goto out;
661 	} else {
662 		/*
663 		 * data follows this file, so set the pad
664 		 */
665 		hd->linkflag = AREGTYPE;
666 		memset(hd->linkname, 0, sizeof(hd->linkname));
667 #		ifdef NET2_STAT
668 		if (ul_oct((u_long)arcn->sb.st_size, hd->size,
669 		    sizeof(hd->size), 1)) {
670 #		else
671 		if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size,
672 		    sizeof(hd->size), 1)) {
673 #		endif
674 			paxwarn(1,"File is too large for tar %s", arcn->org_name);
675 			return(1);
676 		}
677 		arcn->pad = TAR_PAD(arcn->sb.st_size);
678 	}
679 
680 	/*
681 	 * copy those fields that are independent of the type
682 	 */
683 	if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) ||
684 	    ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) ||
685 	    ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) ||
686 	    ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1))
687 		goto out;
688 
689 	/*
690 	 * calculate and add the checksum, then write the header. A return of
691 	 * 0 tells the caller to now write the file data, 1 says no data needs
692 	 * to be written
693 	 */
694 	if (ul_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum,
695 	    sizeof(hd->chksum), 2))
696 		goto out;
697 	if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0)
698 		return(-1);
699 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0)
700 		return(-1);
701 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
702 		return(0);
703 	return(1);
704 
705     out:
706 	/*
707 	 * header field is out of range
708 	 */
709 	paxwarn(1, "Tar header field is too small for %s", arcn->org_name);
710 	return(1);
711 }
712 
713 /*
714  * Routines for POSIX ustar
715  */
716 
717 /*
718  * ustar_strd()
719  *	initialization for ustar read
720  * Return:
721  *	0 if ok, -1 otherwise
722  */
723 
724 #if __STDC__
725 int
726 ustar_strd(void)
727 #else
728 int
729 ustar_strd()
730 #endif
731 {
732 	if ((usrtb_start() < 0) || (grptb_start() < 0))
733 		return(-1);
734 	return(0);
735 }
736 
737 /*
738  * ustar_stwr()
739  *	initialization for ustar write
740  * Return:
741  *	0 if ok, -1 otherwise
742  */
743 
744 #if __STDC__
745 int
746 ustar_stwr(void)
747 #else
748 int
749 ustar_stwr()
750 #endif
751 {
752 	if ((uidtb_start() < 0) || (gidtb_start() < 0))
753 		return(-1);
754 	return(0);
755 }
756 
757 /*
758  * ustar_id()
759  *	determine if a block given to us is a valid ustar header. We have to
760  *	be on the lookout for those pesky blocks of all zero's
761  * Return:
762  *	0 if a ustar header, -1 otherwise
763  */
764 
765 #if __STDC__
766 int
767 ustar_id(char *blk, int size)
768 #else
769 int
770 ustar_id(blk, size)
771 	char *blk;
772 	int size;
773 #endif
774 {
775 	register HD_USTAR *hd;
776 
777 	if (size < BLKMULT)
778 		return(-1);
779 	hd = (HD_USTAR *)blk;
780 
781 	/*
782 	 * check for block of zero's first, a simple and fast test then check
783 	 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive
784 	 * programs are fouled up and create archives missing the \0. Last we
785 	 * check the checksum. If ok we have to assume it is a valid header.
786 	 */
787 	if (hd->name[0] == '\0')
788 		return(-1);
789 	if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0)
790 		return(-1);
791 	if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT))
792 		return(-1);
793 	return(0);
794 }
795 
796 /*
797  * ustar_rd()
798  *	extract the values out of block already determined to be a ustar header.
799  *	store the values in the ARCHD parameter.
800  * Return:
801  *	0
802  */
803 
804 #if __STDC__
805 int
806 ustar_rd(register ARCHD *arcn, register char *buf)
807 #else
808 int
809 ustar_rd(arcn, buf)
810 	register ARCHD *arcn;
811 	register char *buf;
812 #endif
813 {
814 	register HD_USTAR *hd;
815 	register char *dest;
816 	register int cnt = 0;
817 	dev_t devmajor;
818 	dev_t devminor;
819 
820 	/*
821 	 * we only get proper sized buffers
822 	 */
823 	if (ustar_id(buf, BLKMULT) < 0)
824 		return(-1);
825 	arcn->org_name = arcn->name;
826 	arcn->sb.st_nlink = 1;
827 	arcn->pat = NULL;
828 	hd = (HD_USTAR *)buf;
829 
830 	/*
831 	 * see if the filename is split into two parts. if, so joint the parts.
832 	 * we copy the prefix first and add a / between the prefix and name.
833 	 */
834 	dest = arcn->name;
835 	if (*(hd->prefix) != '\0') {
836 		cnt = l_strncpy(arcn->name, hd->prefix, sizeof(hd->prefix) - 1);
837 		hd->prefix[sizeof(hd->prefix) - 1] = '\0';
838 		dest = arcn->name + arcn->nlen;
839 		*dest++ = '/';
840 	}
841 	arcn->nlen = l_strncpy(dest, hd->name, sizeof(hd->name) - 1);
842 	arcn->nlen += cnt;
843 	arcn->name[arcn->nlen] = '\0';
844 
845 	/*
846 	 * follow the spec to the letter. we should only have mode bits, strip
847 	 * off all other crud we may be passed.
848 	 */
849 	arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) &
850 	    0xfff);
851 	arcn->sb.st_size = (size_t)asc_ul(hd->size, sizeof(hd->size), OCT);
852 	arcn->sb.st_mtime = (time_t)asc_ul(hd->mtime, sizeof(hd->mtime), OCT);
853 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
854 
855 	/*
856 	 * If we can find the ascii names for gname and uname in the password
857 	 * and group files we will use the uid's and gid they bind. Otherwise
858 	 * we use the uid and gid values stored in the header. (This is what
859 	 * the posix spec wants).
860 	 */
861 	hd->gname[sizeof(hd->gname) - 1] = '\0';
862 	if (gid_name(hd->gname, &(arcn->sb.st_gid)) < 0)
863 		arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT);
864 	hd->uname[sizeof(hd->uname) - 1] = '\0';
865 	if (uid_name(hd->uname, &(arcn->sb.st_uid)) < 0)
866 		arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT);
867 
868 	/*
869 	 * set the defaults, these may be changed depending on the file type
870 	 */
871 	arcn->ln_name[0] = '\0';
872 	arcn->ln_nlen = 0;
873 	arcn->pad = 0;
874 	arcn->skip = 0;
875 	arcn->sb.st_rdev = (dev_t)0;
876 
877 	/*
878 	 * set the mode and PAX type according to the typeflag in the header
879 	 */
880 	switch(hd->typeflag) {
881 	case FIFOTYPE:
882 		arcn->type = PAX_FIF;
883 		arcn->sb.st_mode |= S_IFIFO;
884 		break;
885 	case DIRTYPE:
886 		arcn->type = PAX_DIR;
887 		arcn->sb.st_mode |= S_IFDIR;
888 		arcn->sb.st_nlink = 2;
889 
890 		/*
891 		 * Some programs that create ustar archives append a '/'
892 		 * to the pathname for directories. This clearly violates
893 		 * ustar specs, but we will silently strip it off anyway.
894 		 */
895 		if (arcn->name[arcn->nlen - 1] == '/')
896 			arcn->name[--arcn->nlen] = '\0';
897 		break;
898 	case BLKTYPE:
899 	case CHRTYPE:
900 		/*
901 		 * this type requires the rdev field to be set.
902 		 */
903 		if (hd->typeflag == BLKTYPE) {
904 			arcn->type = PAX_BLK;
905 			arcn->sb.st_mode |= S_IFBLK;
906 		} else {
907 			arcn->type = PAX_CHR;
908 			arcn->sb.st_mode |= S_IFCHR;
909 		}
910 		devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT);
911 		devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT);
912 		arcn->sb.st_rdev = TODEV(devmajor, devminor);
913 		break;
914 	case SYMTYPE:
915 	case LNKTYPE:
916 		if (hd->typeflag == SYMTYPE) {
917 			arcn->type = PAX_SLK;
918 			arcn->sb.st_mode |= S_IFLNK;
919 		} else {
920 			arcn->type = PAX_HLK;
921 			/*
922 			 * so printing looks better
923 			 */
924 			arcn->sb.st_mode |= S_IFREG;
925 			arcn->sb.st_nlink = 2;
926 		}
927 		/*
928 		 * copy the link name
929 		 */
930 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
931 			sizeof(hd->linkname) - 1);
932 		arcn->ln_name[arcn->ln_nlen] = '\0';
933 		break;
934 	case CONTTYPE:
935 	case AREGTYPE:
936 	case REGTYPE:
937 	default:
938 		/*
939 		 * these types have file data that follows. Set the skip and
940 		 * pad fields.
941 		 */
942 		arcn->type = PAX_REG;
943 		arcn->pad = TAR_PAD(arcn->sb.st_size);
944 		arcn->skip = arcn->sb.st_size;
945 		arcn->sb.st_mode |= S_IFREG;
946 		break;
947 	}
948 	return(0);
949 }
950 
951 /*
952  * ustar_wr()
953  *	write a ustar header for the file specified in the ARCHD to the archive
954  *	Have to check for file types that cannot be stored and file names that
955  *	are too long. Be careful of the term (last arg) to ul_oct, we only use
956  *	'\0' for the termination character (this is different than picky tar)
957  *	ASSUMED: space after header in header block is zero filled
958  * Return:
959  *	0 if file has data to be written after the header, 1 if file has NO
960  *	data to write after the header, -1 if archive write failed
961  */
962 
963 #if __STDC__
964 int
965 ustar_wr(register ARCHD *arcn)
966 #else
967 int
968 ustar_wr(arcn)
969 	register ARCHD *arcn;
970 #endif
971 {
972 	register HD_USTAR *hd;
973 	register char *pt;
974 	char hdblk[sizeof(HD_USTAR)];
975 
976 	/*
977 	 * check for those file system types ustar cannot store
978 	 */
979 	if (arcn->type == PAX_SCK) {
980 		paxwarn(1, "Ustar cannot archive a socket %s", arcn->org_name);
981 		return(1);
982 	}
983 
984 	/*
985 	 * check the length of the linkname
986 	 */
987 	if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
988 	    (arcn->type == PAX_HRG)) && (arcn->ln_nlen > sizeof(hd->linkname))){
989 		paxwarn(1, "Link name too long for ustar %s", arcn->ln_name);
990 		return(1);
991 	}
992 
993 	/*
994 	 * split the path name into prefix and name fields (if needed). if
995 	 * pt != arcn->name, the name has to be split
996 	 */
997 	if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) {
998 		paxwarn(1, "File name too long for ustar %s", arcn->name);
999 		return(1);
1000 	}
1001 	hd = (HD_USTAR *)hdblk;
1002 	arcn->pad = 0L;
1003 
1004 	/*
1005 	 * split the name, or zero out the prefix
1006 	 */
1007 	if (pt != arcn->name) {
1008 		/*
1009 		 * name was split, pt points at the / where the split is to
1010 		 * occur, we remove the / and copy the first part to the prefix
1011 		 */
1012 		*pt = '\0';
1013 		zf_strncpy(hd->prefix, arcn->name, sizeof(hd->prefix) - 1);
1014 		*pt++ = '/';
1015 	} else
1016 		memset(hd->prefix, 0, sizeof(hd->prefix));
1017 
1018 	/*
1019 	 * copy the name part. this may be the whole path or the part after
1020 	 * the prefix
1021 	 */
1022 	zf_strncpy(hd->name, pt, sizeof(hd->name) - 1);
1023 	hd->name[sizeof(hd->name) - 1] = '\0';
1024 
1025 	/*
1026 	 * set the fields in the header that are type dependent
1027 	 */
1028 	switch(arcn->type) {
1029 	case PAX_DIR:
1030 		hd->typeflag = DIRTYPE;
1031 		memset(hd->linkname, 0, sizeof(hd->linkname));
1032 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1033 		memset(hd->devminor, 0, sizeof(hd->devminor));
1034 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1035 			goto out;
1036 		break;
1037 	case PAX_CHR:
1038 	case PAX_BLK:
1039 		if (arcn->type == PAX_CHR)
1040 			hd->typeflag = CHRTYPE;
1041 		else
1042 			hd->typeflag = BLKTYPE;
1043 		memset(hd->linkname, 0, sizeof(hd->linkname));
1044 		if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor,
1045 		   sizeof(hd->devmajor), 3) ||
1046 		   ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor,
1047 		   sizeof(hd->devminor), 3) ||
1048 		   ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1049 			goto out;
1050 		break;
1051 	case PAX_FIF:
1052 		hd->typeflag = FIFOTYPE;
1053 		memset(hd->linkname, 0, sizeof(hd->linkname));
1054 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1055 		memset(hd->devminor, 0, sizeof(hd->devminor));
1056 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1057 			goto out;
1058 		break;
1059 	case PAX_SLK:
1060 	case PAX_HLK:
1061 	case PAX_HRG:
1062 		if (arcn->type == PAX_SLK)
1063 			hd->typeflag = SYMTYPE;
1064 		else
1065 			hd->typeflag = LNKTYPE;
1066 		zf_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname) - 1);
1067 		hd->linkname[sizeof(hd->linkname) - 1] = '\0';
1068 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1069 		memset(hd->devminor, 0, sizeof(hd->devminor));
1070 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
1071 			goto out;
1072 		break;
1073 	case PAX_REG:
1074 	case PAX_CTG:
1075 	default:
1076 		/*
1077 		 * file data with this type, set the padding
1078 		 */
1079 		if (arcn->type == PAX_CTG)
1080 			hd->typeflag = CONTTYPE;
1081 		else
1082 			hd->typeflag = REGTYPE;
1083 		memset(hd->linkname, 0, sizeof(hd->linkname));
1084 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
1085 		memset(hd->devminor, 0, sizeof(hd->devminor));
1086 		arcn->pad = TAR_PAD(arcn->sb.st_size);
1087 #		ifdef NET2_STAT
1088 		if (ul_oct((u_long)arcn->sb.st_size, hd->size,
1089 		    sizeof(hd->size), 3)) {
1090 #		else
1091 		if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size,
1092 		    sizeof(hd->size), 3)) {
1093 #		endif
1094 			paxwarn(1,"File is too long for ustar %s",arcn->org_name);
1095 			return(1);
1096 		}
1097 		break;
1098 	}
1099 
1100 	zf_strncpy(hd->magic, TMAGIC, TMAGLEN);
1101 	zf_strncpy(hd->version, TVERSION, TVERSLEN);
1102 
1103 	/*
1104 	 * set the remaining fields. Some versions want all 16 bits of mode
1105 	 * we better humor them (they really do not meet spec though)....
1106 	 */
1107 	if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) ||
1108 	    ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3)  ||
1109 	    ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) ||
1110 	    ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3))
1111 		goto out;
1112 	zf_strncpy(hd->uname,name_uid(arcn->sb.st_uid, 0),sizeof(hd->uname));
1113 	zf_strncpy(hd->gname,name_gid(arcn->sb.st_gid, 0),sizeof(hd->gname));
1114 
1115 	/*
1116 	 * calculate and store the checksum write the header to the archive
1117 	 * return 0 tells the caller to now write the file data, 1 says no data
1118 	 * needs to be written
1119 	 */
1120 	if (ul_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum,
1121 	   sizeof(hd->chksum), 3))
1122 		goto out;
1123 	if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0)
1124 		return(-1);
1125 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0)
1126 		return(-1);
1127 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
1128 		return(0);
1129 	return(1);
1130 
1131     out:
1132     	/*
1133 	 * header field is out of range
1134 	 */
1135 	paxwarn(1, "Ustar header field is too small for %s", arcn->org_name);
1136 	return(1);
1137 }
1138 
1139 /*
1140  * name_split()
1141  *	see if the name has to be split for storage in a ustar header. We try
1142  *	to fit the entire name in the name field without splitting if we can.
1143  *	The split point is always at a /
1144  * Return
1145  *	character pointer to split point (always the / that is to be removed
1146  *	if the split is not needed, the points is set to the start of the file
1147  *	name (it would violate the spec to split there). A NULL is returned if
1148  *	the file name is too long
1149  */
1150 
1151 #if __STDC__
1152 static char *
1153 name_split(register char *name, register int len)
1154 #else
1155 static char *
1156 name_split(name, len)
1157 	register char *name;
1158 	register int len;
1159 #endif
1160 {
1161 	register char *start;
1162 
1163 	/*
1164 	 * check to see if the file name is small enough to fit in the name
1165 	 * field. if so just return a pointer to the name.
1166 	 */
1167 	if (len <= TNMSZ)
1168 		return(name);
1169 	if (len > (TPFSZ + TNMSZ + 1))
1170 		return(NULL);
1171 
1172 	/*
1173 	 * we start looking at the biggest sized piece that fits in the name
1174 	 * field. We walk foward looking for a slash to split at. The idea is
1175 	 * to find the biggest piece to fit in the name field (or the smallest
1176 	 * prefix we can find) (the -1 is correct the biggest piece would
1177 	 * include the slash between the two parts that gets thrown away)
1178 	 */
1179 	start = name + len - TNMSZ - 1;
1180 	while ((*start != '\0') && (*start != '/'))
1181 		++start;
1182 
1183 	/*
1184 	 * if we hit the end of the string, this name cannot be split, so we
1185 	 * cannot store this file.
1186 	 */
1187 	if (*start == '\0')
1188 		return(NULL);
1189 	len = start - name;
1190 
1191 	/*
1192 	 * NOTE: /str where the length of str == TNMSZ can not be stored under
1193 	 * the p1003.1-1990 spec for ustar. We could force a prefix of / and
1194 	 * the file would then expand on extract to //str. The len == 0 below
1195 	 * makes this special case follow the spec to the letter.
1196 	 */
1197 	if ((len > TPFSZ) || (len == 0))
1198 		return(NULL);
1199 
1200 	/*
1201 	 * ok have a split point, return it to the caller
1202 	 */
1203 	return(start);
1204 }
1205