xref: /freebsd/bin/pax/tar.c (revision 734e82fe33aa764367791a7d603b383996c6b40b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
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. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 #ifndef lint
37 #if 0
38 static char sccsid[] = "@(#)tar.c	8.2 (Berkeley) 4/18/94";
39 #endif
40 #endif /* not lint */
41 #include <sys/cdefs.h>
42 #include <sys/types.h>
43 #include <sys/time.h>
44 #include <sys/stat.h>
45 #include <string.h>
46 #include <stdio.h>
47 #include "pax.h"
48 #include "extern.h"
49 #include "tar.h"
50 
51 /*
52  * Routines for reading, writing and header identify of various versions of tar
53  */
54 
55 static u_long tar_chksm(char *, int);
56 static char *name_split(char *, int);
57 static int ul_oct(u_long, char *, int, int);
58 static int uqd_oct(u_quad_t, char *, int, int);
59 
60 /*
61  * Routines common to all versions of tar
62  */
63 
64 static int tar_nodir;			/* do not write dirs under old tar */
65 
66 /*
67  * tar_endwr()
68  *	add the tar trailer of two null blocks
69  * Return:
70  *	0 if ok, -1 otherwise (what wr_skip returns)
71  */
72 
73 int
74 tar_endwr(void)
75 {
76 	return(wr_skip((off_t)(NULLCNT*BLKMULT)));
77 }
78 
79 /*
80  * tar_endrd()
81  *	no cleanup needed here, just return size of trailer (for append)
82  * Return:
83  *	size of trailer (2 * BLKMULT)
84  */
85 
86 off_t
87 tar_endrd(void)
88 {
89 	return((off_t)(NULLCNT*BLKMULT));
90 }
91 
92 /*
93  * tar_trail()
94  *	Called to determine if a header block is a valid trailer. We are passed
95  *	the block, the in_sync flag (which tells us we are in resync mode;
96  *	looking for a valid header), and cnt (which starts at zero) which is
97  *	used to count the number of empty blocks we have seen so far.
98  * Return:
99  *	0 if a valid trailer, -1 if not a valid trailer, or 1 if the block
100  *	could never contain a header.
101  */
102 
103 int
104 tar_trail(char *buf, int in_resync, int *cnt)
105 {
106 	int i;
107 
108 	/*
109 	 * look for all zero, trailer is two consecutive blocks of zero
110 	 */
111 	for (i = 0; i < BLKMULT; ++i) {
112 		if (buf[i] != '\0')
113 			break;
114 	}
115 
116 	/*
117 	 * if not all zero it is not a trailer, but MIGHT be a header.
118 	 */
119 	if (i != BLKMULT)
120 		return(-1);
121 
122 	/*
123 	 * When given a zero block, we must be careful!
124 	 * If we are not in resync mode, check for the trailer. Have to watch
125 	 * out that we do not mis-identify file data as the trailer, so we do
126 	 * NOT try to id a trailer during resync mode. During resync mode we
127 	 * might as well throw this block out since a valid header can NEVER be
128 	 * a block of all 0 (we must have a valid file name).
129 	 */
130 	if (!in_resync && (++*cnt >= NULLCNT))
131 		return(0);
132 	return(1);
133 }
134 
135 /*
136  * ul_oct()
137  *	convert an unsigned long to an octal string. many oddball field
138  *	termination characters are used by the various versions of tar in the
139  *	different fields. term selects which kind to use. str is '0' padded
140  *	at the front to len. we are unable to use only one format as many old
141  *	tar readers are very cranky about this.
142  * Return:
143  *	0 if the number fit into the string, -1 otherwise
144  */
145 
146 static int
147 ul_oct(u_long val, char *str, int len, int term)
148 {
149 	char *pt;
150 
151 	/*
152 	 * term selects the appropriate character(s) for the end of the string
153 	 */
154 	pt = str + len - 1;
155 	switch(term) {
156 	case 3:
157 		*pt-- = '\0';
158 		break;
159 	case 2:
160 		*pt-- = ' ';
161 		*pt-- = '\0';
162 		break;
163 	case 1:
164 		*pt-- = ' ';
165 		break;
166 	case 0:
167 	default:
168 		*pt-- = '\0';
169 		*pt-- = ' ';
170 		break;
171 	}
172 
173 	/*
174 	 * convert and blank pad if there is space
175 	 */
176 	while (pt >= str) {
177 		*pt-- = '0' + (char)(val & 0x7);
178 		if ((val = val >> 3) == (u_long)0)
179 			break;
180 	}
181 
182 	while (pt >= str)
183 		*pt-- = '0';
184 	if (val != (u_long)0)
185 		return(-1);
186 	return(0);
187 }
188 
189 /*
190  * uqd_oct()
191  *	convert an u_quad_t to an octal string. one of many oddball field
192  *	termination characters are used by the various versions of tar in the
193  *	different fields. term selects which kind to use. str is '0' padded
194  *	at the front to len. we are unable to use only one format as many old
195  *	tar readers are very cranky about this.
196  * Return:
197  *	0 if the number fit into the string, -1 otherwise
198  */
199 
200 static int
201 uqd_oct(u_quad_t val, char *str, int len, int term)
202 {
203 	char *pt;
204 
205 	/*
206 	 * term selects the appropriate character(s) for the end of the string
207 	 */
208 	pt = str + len - 1;
209 	switch(term) {
210 	case 3:
211 		*pt-- = '\0';
212 		break;
213 	case 2:
214 		*pt-- = ' ';
215 		*pt-- = '\0';
216 		break;
217 	case 1:
218 		*pt-- = ' ';
219 		break;
220 	case 0:
221 	default:
222 		*pt-- = '\0';
223 		*pt-- = ' ';
224 		break;
225 	}
226 
227 	/*
228 	 * convert and blank pad if there is space
229 	 */
230 	while (pt >= str) {
231 		*pt-- = '0' + (char)(val & 0x7);
232 		if ((val = val >> 3) == 0)
233 			break;
234 	}
235 
236 	while (pt >= str)
237 		*pt-- = '0';
238 	if (val != (u_quad_t)0)
239 		return(-1);
240 	return(0);
241 }
242 
243 /*
244  * tar_chksm()
245  *	calculate the checksum for a tar block counting the checksum field as
246  *	all blanks (BLNKSUM is that value pre-calculated, the sum of 8 blanks).
247  *	NOTE: we use len to short circuit summing 0's on write since we ALWAYS
248  *	pad headers with 0.
249  * Return:
250  *	unsigned long checksum
251  */
252 
253 static u_long
254 tar_chksm(char *blk, int len)
255 {
256 	char *stop;
257 	char *pt;
258 	u_long chksm = BLNKSUM;	/* initial value is checksum field sum */
259 
260 	/*
261 	 * add the part of the block before the checksum field
262 	 */
263 	pt = blk;
264 	stop = blk + CHK_OFFSET;
265 	while (pt < stop)
266 		chksm += (u_long)(*pt++ & 0xff);
267 	/*
268 	 * move past the checksum field and keep going, spec counts the
269 	 * checksum field as the sum of 8 blanks (which is pre-computed as
270 	 * BLNKSUM).
271 	 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding
272 	 * starts, no point in summing zero's)
273 	 */
274 	pt += CHK_LEN;
275 	stop = blk + len;
276 	while (pt < stop)
277 		chksm += (u_long)(*pt++ & 0xff);
278 	return(chksm);
279 }
280 
281 /*
282  * Routines for old BSD style tar (also made portable to sysV tar)
283  */
284 
285 /*
286  * tar_id()
287  *	determine if a block given to us is a valid tar header (and not a USTAR
288  *	header). We have to be on the lookout for those pesky blocks of all
289  *	zero's.
290  * Return:
291  *	0 if a tar header, -1 otherwise
292  */
293 
294 int
295 tar_id(char *blk, int size)
296 {
297 	HD_TAR *hd;
298 	HD_USTAR *uhd;
299 
300 	if (size < BLKMULT)
301 		return(-1);
302 	hd = (HD_TAR *)blk;
303 	uhd = (HD_USTAR *)blk;
304 
305 	/*
306 	 * check for block of zero's first, a simple and fast test, then make
307 	 * sure this is not a ustar header by looking for the ustar magic
308 	 * cookie. We should use TMAGLEN, but some USTAR archive programs are
309 	 * wrong and create archives missing the \0. Last we check the
310 	 * checksum. If this is ok we have to assume it is a valid header.
311 	 */
312 	if (hd->name[0] == '\0')
313 		return(-1);
314 	if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0)
315 		return(-1);
316 	if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT))
317 		return(-1);
318 	return(0);
319 }
320 
321 /*
322  * tar_opt()
323  *	handle tar format specific -o options
324  * Return:
325  *	0 if ok -1 otherwise
326  */
327 
328 int
329 tar_opt(void)
330 {
331 	OPLIST *opt;
332 
333 	while ((opt = opt_next()) != NULL) {
334 		if (strcmp(opt->name, TAR_OPTION) ||
335 		    strcmp(opt->value, TAR_NODIR)) {
336 			paxwarn(1, "Unknown tar format -o option/value pair %s=%s",
337 			    opt->name, opt->value);
338 			paxwarn(1,"%s=%s is the only supported tar format option",
339 			    TAR_OPTION, TAR_NODIR);
340 			return(-1);
341 		}
342 
343 		/*
344 		 * we only support one option, and only when writing
345 		 */
346 		if ((act != APPND) && (act != ARCHIVE)) {
347 			paxwarn(1, "%s=%s is only supported when writing.",
348 			    opt->name, opt->value);
349 			return(-1);
350 		}
351 		tar_nodir = 1;
352 	}
353 	return(0);
354 }
355 
356 
357 /*
358  * tar_rd()
359  *	extract the values out of block already determined to be a tar header.
360  *	store the values in the ARCHD parameter.
361  * Return:
362  *	0
363  */
364 
365 int
366 tar_rd(ARCHD *arcn, char *buf)
367 {
368 	HD_TAR *hd;
369 	char *pt;
370 
371 	/*
372 	 * we only get proper sized buffers passed to us
373 	 */
374 	if (tar_id(buf, BLKMULT) < 0)
375 		return(-1);
376 	arcn->org_name = arcn->name;
377 	arcn->sb.st_nlink = 1;
378 	arcn->pat = NULL;
379 
380 	/*
381 	 * copy out the name and values in the stat buffer
382 	 */
383 	hd = (HD_TAR *)buf;
384 	/*
385 	 * old tar format specifies the name always be null-terminated,
386 	 * but let's be robust to broken archives.
387 	 * the same applies to handling links below.
388 	 */
389 	arcn->nlen = l_strncpy(arcn->name, hd->name,
390 	    MIN(sizeof(hd->name), sizeof(arcn->name)) - 1);
391 	arcn->name[arcn->nlen] = '\0';
392 	arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode,sizeof(hd->mode),OCT) &
393 	    0xfff);
394 	arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT);
395 	arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT);
396 	arcn->sb.st_size = (off_t)asc_uqd(hd->size, sizeof(hd->size), OCT);
397 	arcn->sb.st_mtime = (time_t)asc_uqd(hd->mtime, sizeof(hd->mtime), OCT);
398 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
399 
400 	/*
401 	 * have to look at the last character, it may be a '/' and that is used
402 	 * to encode this as a directory
403 	 */
404 	pt = &(arcn->name[arcn->nlen - 1]);
405 	arcn->pad = 0;
406 	arcn->skip = 0;
407 	switch(hd->linkflag) {
408 	case SYMTYPE:
409 		/*
410 		 * symbolic link, need to get the link name and set the type in
411 		 * the st_mode so -v printing will look correct.
412 		 */
413 		arcn->type = PAX_SLK;
414 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
415 		    MIN(sizeof(hd->linkname), sizeof(arcn->ln_name)) - 1);
416 		arcn->ln_name[arcn->ln_nlen] = '\0';
417 		arcn->sb.st_mode |= S_IFLNK;
418 		break;
419 	case LNKTYPE:
420 		/*
421 		 * hard link, need to get the link name, set the type in the
422 		 * st_mode and st_nlink so -v printing will look better.
423 		 */
424 		arcn->type = PAX_HLK;
425 		arcn->sb.st_nlink = 2;
426 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
427 		    MIN(sizeof(hd->linkname), sizeof(arcn->ln_name)) - 1);
428 		arcn->ln_name[arcn->ln_nlen] = '\0';
429 
430 		/*
431 		 * no idea of what type this thing really points at, but
432 		 * we set something for printing only.
433 		 */
434 		arcn->sb.st_mode |= S_IFREG;
435 		break;
436 	case DIRTYPE:
437 		/*
438 		 * It is a directory, set the mode for -v printing
439 		 */
440 		arcn->type = PAX_DIR;
441 		arcn->sb.st_mode |= S_IFDIR;
442 		arcn->sb.st_nlink = 2;
443 		arcn->ln_name[0] = '\0';
444 		arcn->ln_nlen = 0;
445 		break;
446 	case AREGTYPE:
447 	case REGTYPE:
448 	default:
449 		/*
450 		 * If we have a trailing / this is a directory and NOT a file.
451 		 */
452 		arcn->ln_name[0] = '\0';
453 		arcn->ln_nlen = 0;
454 		if (*pt == '/') {
455 			/*
456 			 * it is a directory, set the mode for -v printing
457 			 */
458 			arcn->type = PAX_DIR;
459 			arcn->sb.st_mode |= S_IFDIR;
460 			arcn->sb.st_nlink = 2;
461 		} else {
462 			/*
463 			 * have a file that will be followed by data. Set the
464 			 * skip value to the size field and calculate the size
465 			 * of the padding.
466 			 */
467 			arcn->type = PAX_REG;
468 			arcn->sb.st_mode |= S_IFREG;
469 			arcn->pad = TAR_PAD(arcn->sb.st_size);
470 			arcn->skip = arcn->sb.st_size;
471 		}
472 		break;
473 	}
474 
475 	/*
476 	 * strip off any trailing slash.
477 	 */
478 	if (*pt == '/') {
479 		*pt = '\0';
480 		--arcn->nlen;
481 	}
482 	return(0);
483 }
484 
485 /*
486  * tar_wr()
487  *	write a tar header for the file specified in the ARCHD to the archive.
488  *	Have to check for file types that cannot be stored and file names that
489  *	are too long. Be careful of the term (last arg) to ul_oct, each field
490  *	of tar has it own spec for the termination character(s).
491  *	ASSUMED: space after header in header block is zero filled
492  * Return:
493  *	0 if file has data to be written after the header, 1 if file has NO
494  *	data to write after the header, -1 if archive write failed
495  */
496 
497 int
498 tar_wr(ARCHD *arcn)
499 {
500 	HD_TAR *hd;
501 	int len;
502 	HD_TAR hdblk;
503 
504 	/*
505 	 * check for those file system types which tar cannot store
506 	 */
507 	switch(arcn->type) {
508 	case PAX_DIR:
509 		/*
510 		 * user asked that dirs not be written to the archive
511 		 */
512 		if (tar_nodir)
513 			return(1);
514 		break;
515 	case PAX_CHR:
516 		paxwarn(1, "Tar cannot archive a character device %s",
517 		    arcn->org_name);
518 		return(1);
519 	case PAX_BLK:
520 		paxwarn(1, "Tar cannot archive a block device %s", arcn->org_name);
521 		return(1);
522 	case PAX_SCK:
523 		paxwarn(1, "Tar cannot archive a socket %s", arcn->org_name);
524 		return(1);
525 	case PAX_FIF:
526 		paxwarn(1, "Tar cannot archive a fifo %s", arcn->org_name);
527 		return(1);
528 	case PAX_SLK:
529 	case PAX_HLK:
530 	case PAX_HRG:
531 		if (arcn->ln_nlen >= (int)sizeof(hd->linkname)) {
532 			paxwarn(1,"Link name too long for tar %s", arcn->ln_name);
533 			return(1);
534 		}
535 		break;
536 	case PAX_REG:
537 	case PAX_CTG:
538 	default:
539 		break;
540 	}
541 
542 	/*
543 	 * check file name len, remember extra char for dirs (the / at the end)
544 	 */
545 	len = arcn->nlen;
546 	if (arcn->type == PAX_DIR)
547 		++len;
548 	if (len >= (int)sizeof(hd->name)) {
549 		paxwarn(1, "File name too long for tar %s", arcn->name);
550 		return(1);
551 	}
552 
553 	/*
554 	 * Copy the data out of the ARCHD into the tar header based on the type
555 	 * of the file. Remember, many tar readers want all fields to be
556 	 * padded with zero so we zero the header first.  We then set the
557 	 * linkflag field (type), the linkname, the size, and set the padding
558 	 * (if any) to be added after the file data (0 for all other types,
559 	 * as they only have a header).
560 	 */
561 	hd = &hdblk;
562 	l_strncpy(hd->name, arcn->name, sizeof(hd->name) - 1);
563 	hd->name[sizeof(hd->name) - 1] = '\0';
564 	arcn->pad = 0;
565 
566 	if (arcn->type == PAX_DIR) {
567 		/*
568 		 * directories are the same as files, except have a filename
569 		 * that ends with a /, we add the slash here. No data follows,
570 		 * dirs, so no pad.
571 		 */
572 		hd->linkflag = AREGTYPE;
573 		memset(hd->linkname, 0, sizeof(hd->linkname));
574 		hd->name[len-1] = '/';
575 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
576 			goto out;
577 	} else if (arcn->type == PAX_SLK) {
578 		/*
579 		 * no data follows this file, so no pad
580 		 */
581 		hd->linkflag = SYMTYPE;
582 		l_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname) - 1);
583 		hd->linkname[sizeof(hd->linkname) - 1] = '\0';
584 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
585 			goto out;
586 	} else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) {
587 		/*
588 		 * no data follows this file, so no pad
589 		 */
590 		hd->linkflag = LNKTYPE;
591 		l_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname) - 1);
592 		hd->linkname[sizeof(hd->linkname) - 1] = '\0';
593 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 1))
594 			goto out;
595 	} else {
596 		/*
597 		 * data follows this file, so set the pad
598 		 */
599 		hd->linkflag = AREGTYPE;
600 		memset(hd->linkname, 0, sizeof(hd->linkname));
601 		if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size,
602 		    sizeof(hd->size), 1)) {
603 			paxwarn(1,"File is too large for tar %s", arcn->org_name);
604 			return(1);
605 		}
606 		arcn->pad = TAR_PAD(arcn->sb.st_size);
607 	}
608 
609 	/*
610 	 * copy those fields that are independent of the type
611 	 */
612 	if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) ||
613 	    ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) ||
614 	    ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) ||
615 	    ul_oct((u_long)arcn->sb.st_mtime, hd->mtime, sizeof(hd->mtime), 1))
616 		goto out;
617 
618 	/*
619 	 * calculate and add the checksum, then write the header. A return of
620 	 * 0 tells the caller to now write the file data, 1 says no data needs
621 	 * to be written
622 	 */
623 	if (ul_oct(tar_chksm((char *)&hdblk, sizeof(HD_TAR)), hd->chksum,
624 	    sizeof(hd->chksum), 3))
625 		goto out;
626 	if (wr_rdbuf((char *)&hdblk, sizeof(HD_TAR)) < 0)
627 		return(-1);
628 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0)
629 		return(-1);
630 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
631 		return(0);
632 	return(1);
633 
634     out:
635 	/*
636 	 * header field is out of range
637 	 */
638 	paxwarn(1, "Tar header field is too small for %s", arcn->org_name);
639 	return(1);
640 }
641 
642 /*
643  * Routines for POSIX ustar
644  */
645 
646 /*
647  * ustar_strd()
648  *	initialization for ustar read
649  * Return:
650  *	0 if ok, -1 otherwise
651  */
652 
653 int
654 ustar_strd(void)
655 {
656 	if ((usrtb_start() < 0) || (grptb_start() < 0))
657 		return(-1);
658 	return(0);
659 }
660 
661 /*
662  * ustar_stwr()
663  *	initialization for ustar write
664  * Return:
665  *	0 if ok, -1 otherwise
666  */
667 
668 int
669 ustar_stwr(void)
670 {
671 	if ((uidtb_start() < 0) || (gidtb_start() < 0))
672 		return(-1);
673 	return(0);
674 }
675 
676 /*
677  * ustar_id()
678  *	determine if a block given to us is a valid ustar header. We have to
679  *	be on the lookout for those pesky blocks of all zero's
680  * Return:
681  *	0 if a ustar header, -1 otherwise
682  */
683 
684 int
685 ustar_id(char *blk, int size)
686 {
687 	HD_USTAR *hd;
688 
689 	if (size < BLKMULT)
690 		return(-1);
691 	hd = (HD_USTAR *)blk;
692 
693 	/*
694 	 * check for block of zero's first, a simple and fast test then check
695 	 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive
696 	 * programs are fouled up and create archives missing the \0. Last we
697 	 * check the checksum. If ok we have to assume it is a valid header.
698 	 */
699 	if (hd->name[0] == '\0')
700 		return(-1);
701 	if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0)
702 		return(-1);
703 	if (asc_ul(hd->chksum,sizeof(hd->chksum),OCT) != tar_chksm(blk,BLKMULT))
704 		return(-1);
705 	return(0);
706 }
707 
708 /*
709  * ustar_rd()
710  *	extract the values out of block already determined to be a ustar header.
711  *	store the values in the ARCHD parameter.
712  * Return:
713  *	0
714  */
715 
716 int
717 ustar_rd(ARCHD *arcn, char *buf)
718 {
719 	HD_USTAR *hd;
720 	char *dest;
721 	int cnt = 0;
722 	dev_t devmajor;
723 	dev_t devminor;
724 
725 	/*
726 	 * we only get proper sized buffers
727 	 */
728 	if (ustar_id(buf, BLKMULT) < 0)
729 		return(-1);
730 	arcn->org_name = arcn->name;
731 	arcn->sb.st_nlink = 1;
732 	arcn->pat = NULL;
733 	arcn->nlen = 0;
734 	hd = (HD_USTAR *)buf;
735 
736 	/*
737 	 * see if the filename is split into two parts. if, so joint the parts.
738 	 * we copy the prefix first and add a / between the prefix and name.
739 	 */
740 	dest = arcn->name;
741 	if (*(hd->prefix) != '\0') {
742 		cnt = l_strncpy(dest, hd->prefix,
743 		    MIN(sizeof(hd->prefix), sizeof(arcn->name) - 2));
744 		dest += cnt;
745 		*dest++ = '/';
746 		cnt++;
747 	}
748 	/*
749 	 * ustar format specifies the name may be unterminated
750 	 * if it fills the entire field.  this also applies to
751 	 * the prefix and the linkname.
752 	 */
753 	arcn->nlen = cnt + l_strncpy(dest, hd->name,
754 	    MIN(sizeof(hd->name), sizeof(arcn->name) - cnt - 1));
755 	arcn->name[arcn->nlen] = '\0';
756 
757 	/*
758 	 * follow the spec to the letter. we should only have mode bits, strip
759 	 * off all other crud we may be passed.
760 	 */
761 	arcn->sb.st_mode = (mode_t)(asc_ul(hd->mode, sizeof(hd->mode), OCT) &
762 	    0xfff);
763 	arcn->sb.st_size = (off_t)asc_uqd(hd->size, sizeof(hd->size), OCT);
764 	arcn->sb.st_mtime = (time_t)asc_uqd(hd->mtime, sizeof(hd->mtime), OCT);
765 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
766 
767 	/*
768 	 * If we can find the ascii names for gname and uname in the password
769 	 * and group files we will use the uid's and gid they bind. Otherwise
770 	 * we use the uid and gid values stored in the header. (This is what
771 	 * the POSIX spec wants).
772 	 */
773 	hd->gname[sizeof(hd->gname) - 1] = '\0';
774 	if (gid_name(hd->gname, &(arcn->sb.st_gid)) < 0)
775 		arcn->sb.st_gid = (gid_t)asc_ul(hd->gid, sizeof(hd->gid), OCT);
776 	hd->uname[sizeof(hd->uname) - 1] = '\0';
777 	if (uid_name(hd->uname, &(arcn->sb.st_uid)) < 0)
778 		arcn->sb.st_uid = (uid_t)asc_ul(hd->uid, sizeof(hd->uid), OCT);
779 
780 	/*
781 	 * set the defaults, these may be changed depending on the file type
782 	 */
783 	arcn->ln_name[0] = '\0';
784 	arcn->ln_nlen = 0;
785 	arcn->pad = 0;
786 	arcn->skip = 0;
787 	arcn->sb.st_rdev = (dev_t)0;
788 
789 	/*
790 	 * set the mode and PAX type according to the typeflag in the header
791 	 */
792 	switch(hd->typeflag) {
793 	case FIFOTYPE:
794 		arcn->type = PAX_FIF;
795 		arcn->sb.st_mode |= S_IFIFO;
796 		break;
797 	case DIRTYPE:
798 		arcn->type = PAX_DIR;
799 		arcn->sb.st_mode |= S_IFDIR;
800 		arcn->sb.st_nlink = 2;
801 
802 		/*
803 		 * Some programs that create ustar archives append a '/'
804 		 * to the pathname for directories. This clearly violates
805 		 * ustar specs, but we will silently strip it off anyway.
806 		 */
807 		if (arcn->name[arcn->nlen - 1] == '/')
808 			arcn->name[--arcn->nlen] = '\0';
809 		break;
810 	case BLKTYPE:
811 	case CHRTYPE:
812 		/*
813 		 * this type requires the rdev field to be set.
814 		 */
815 		if (hd->typeflag == BLKTYPE) {
816 			arcn->type = PAX_BLK;
817 			arcn->sb.st_mode |= S_IFBLK;
818 		} else {
819 			arcn->type = PAX_CHR;
820 			arcn->sb.st_mode |= S_IFCHR;
821 		}
822 		devmajor = (dev_t)asc_ul(hd->devmajor,sizeof(hd->devmajor),OCT);
823 		devminor = (dev_t)asc_ul(hd->devminor,sizeof(hd->devminor),OCT);
824 		arcn->sb.st_rdev = TODEV(devmajor, devminor);
825 		break;
826 	case SYMTYPE:
827 	case LNKTYPE:
828 		if (hd->typeflag == SYMTYPE) {
829 			arcn->type = PAX_SLK;
830 			arcn->sb.st_mode |= S_IFLNK;
831 		} else {
832 			arcn->type = PAX_HLK;
833 			/*
834 			 * so printing looks better
835 			 */
836 			arcn->sb.st_mode |= S_IFREG;
837 			arcn->sb.st_nlink = 2;
838 		}
839 		/*
840 		 * copy the link name
841 		 */
842 		arcn->ln_nlen = l_strncpy(arcn->ln_name, hd->linkname,
843 		    MIN(sizeof(hd->linkname), sizeof(arcn->ln_name) - 1));
844 		arcn->ln_name[arcn->ln_nlen] = '\0';
845 		break;
846 	case CONTTYPE:
847 	case AREGTYPE:
848 	case REGTYPE:
849 	default:
850 		/*
851 		 * these types have file data that follows. Set the skip and
852 		 * pad fields.
853 		 */
854 		arcn->type = PAX_REG;
855 		arcn->pad = TAR_PAD(arcn->sb.st_size);
856 		arcn->skip = arcn->sb.st_size;
857 		arcn->sb.st_mode |= S_IFREG;
858 		break;
859 	}
860 	return(0);
861 }
862 
863 /*
864  * ustar_wr()
865  *	write a ustar header for the file specified in the ARCHD to the archive
866  *	Have to check for file types that cannot be stored and file names that
867  *	are too long. Be careful of the term (last arg) to ul_oct, we only use
868  *	'\0' for the termination character (this is different than picky tar)
869  *	ASSUMED: space after header in header block is zero filled
870  * Return:
871  *	0 if file has data to be written after the header, 1 if file has NO
872  *	data to write after the header, -1 if archive write failed
873  */
874 
875 int
876 ustar_wr(ARCHD *arcn)
877 {
878 	HD_USTAR *hd;
879 	char *pt;
880 	HD_USTAR hdblk;
881 
882 	/*
883 	 * check for those file system types ustar cannot store
884 	 */
885 	if (arcn->type == PAX_SCK) {
886 		paxwarn(1, "Ustar cannot archive a socket %s", arcn->org_name);
887 		return(1);
888 	}
889 
890 	/*
891 	 * check the length of the linkname
892 	 */
893 	if (((arcn->type == PAX_SLK) || (arcn->type == PAX_HLK) ||
894 	    (arcn->type == PAX_HRG)) &&
895 	    (arcn->ln_nlen > (int)sizeof(hd->linkname))) {
896 		paxwarn(1, "Link name too long for ustar %s", arcn->ln_name);
897 		return(1);
898 	}
899 
900 	/*
901 	 * split the path name into prefix and name fields (if needed). if
902 	 * pt != arcn->name, the name has to be split
903 	 */
904 	if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) {
905 		paxwarn(1, "File name too long for ustar %s", arcn->name);
906 		return(1);
907 	}
908 	hd = &hdblk;
909 	arcn->pad = 0L;
910 
911 	/*
912 	 * split the name, or zero out the prefix
913 	 */
914 	if (pt != arcn->name) {
915 		/*
916 		 * name was split, pt points at the / where the split is to
917 		 * occur, we remove the / and copy the first part to the prefix
918 		 */
919 		*pt = '\0';
920 		l_strncpy(hd->prefix, arcn->name, sizeof(hd->prefix));
921 		*pt++ = '/';
922 	} else
923 		memset(hd->prefix, 0, sizeof(hd->prefix));
924 
925 	/*
926 	 * copy the name part. this may be the whole path or the part after
927 	 * the prefix.  both the name and prefix may fill the entire field.
928 	 */
929 	l_strncpy(hd->name, pt, sizeof(hd->name));
930 
931 	/*
932 	 * set the fields in the header that are type dependent
933 	 */
934 	switch(arcn->type) {
935 	case PAX_DIR:
936 		hd->typeflag = DIRTYPE;
937 		memset(hd->linkname, 0, sizeof(hd->linkname));
938 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
939 		memset(hd->devminor, 0, sizeof(hd->devminor));
940 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
941 			goto out;
942 		break;
943 	case PAX_CHR:
944 	case PAX_BLK:
945 		if (arcn->type == PAX_CHR)
946 			hd->typeflag = CHRTYPE;
947 		else
948 			hd->typeflag = BLKTYPE;
949 		memset(hd->linkname, 0, sizeof(hd->linkname));
950 		if (ul_oct((u_long)MAJOR(arcn->sb.st_rdev), hd->devmajor,
951 		   sizeof(hd->devmajor), 3) ||
952 		   ul_oct((u_long)MINOR(arcn->sb.st_rdev), hd->devminor,
953 		   sizeof(hd->devminor), 3) ||
954 		   ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
955 			goto out;
956 		break;
957 	case PAX_FIF:
958 		hd->typeflag = FIFOTYPE;
959 		memset(hd->linkname, 0, sizeof(hd->linkname));
960 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
961 		memset(hd->devminor, 0, sizeof(hd->devminor));
962 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
963 			goto out;
964 		break;
965 	case PAX_SLK:
966 	case PAX_HLK:
967 	case PAX_HRG:
968 		if (arcn->type == PAX_SLK)
969 			hd->typeflag = SYMTYPE;
970 		else
971 			hd->typeflag = LNKTYPE;
972 		/* the link name may occupy the entire field in ustar */
973 		l_strncpy(hd->linkname,arcn->ln_name, sizeof(hd->linkname));
974 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
975 		memset(hd->devminor, 0, sizeof(hd->devminor));
976 		if (ul_oct((u_long)0L, hd->size, sizeof(hd->size), 3))
977 			goto out;
978 		break;
979 	case PAX_REG:
980 	case PAX_CTG:
981 	default:
982 		/*
983 		 * file data with this type, set the padding
984 		 */
985 		if (arcn->type == PAX_CTG)
986 			hd->typeflag = CONTTYPE;
987 		else
988 			hd->typeflag = REGTYPE;
989 		memset(hd->linkname, 0, sizeof(hd->linkname));
990 		memset(hd->devmajor, 0, sizeof(hd->devmajor));
991 		memset(hd->devminor, 0, sizeof(hd->devminor));
992 		arcn->pad = TAR_PAD(arcn->sb.st_size);
993 		if (uqd_oct((u_quad_t)arcn->sb.st_size, hd->size,
994 		    sizeof(hd->size), 3)) {
995 			paxwarn(1,"File is too long for ustar %s",arcn->org_name);
996 			return(1);
997 		}
998 		break;
999 	}
1000 
1001 	l_strncpy(hd->magic, TMAGIC, TMAGLEN);
1002 	l_strncpy(hd->version, TVERSION, TVERSLEN);
1003 
1004 	/*
1005 	 * set the remaining fields. Some versions want all 16 bits of mode
1006 	 * we better humor them (they really do not meet spec though)....
1007 	 */
1008 	if (ul_oct((u_long)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3) ||
1009 	    ul_oct((u_long)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3)  ||
1010 	    ul_oct((u_long)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3) ||
1011 	    ul_oct((u_long)arcn->sb.st_mtime,hd->mtime,sizeof(hd->mtime),3))
1012 		goto out;
1013 	l_strncpy(hd->uname,name_uid(arcn->sb.st_uid, 0),sizeof(hd->uname));
1014 	l_strncpy(hd->gname,name_gid(arcn->sb.st_gid, 0),sizeof(hd->gname));
1015 
1016 	/*
1017 	 * calculate and store the checksum write the header to the archive
1018 	 * return 0 tells the caller to now write the file data, 1 says no data
1019 	 * needs to be written
1020 	 */
1021 	if (ul_oct(tar_chksm((char *)&hdblk, sizeof(HD_USTAR)), hd->chksum,
1022 	   sizeof(hd->chksum), 3))
1023 		goto out;
1024 	if (wr_rdbuf((char *)&hdblk, sizeof(HD_USTAR)) < 0)
1025 		return(-1);
1026 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0)
1027 		return(-1);
1028 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
1029 		return(0);
1030 	return(1);
1031 
1032     out:
1033     	/*
1034 	 * header field is out of range
1035 	 */
1036 	paxwarn(1, "Ustar header field is too small for %s", arcn->org_name);
1037 	return(1);
1038 }
1039 
1040 /*
1041  * name_split()
1042  *	see if the name has to be split for storage in a ustar header. We try
1043  *	to fit the entire name in the name field without splitting if we can.
1044  *	The split point is always at a /
1045  * Return
1046  *	character pointer to split point (always the / that is to be removed
1047  *	if the split is not needed, the points is set to the start of the file
1048  *	name (it would violate the spec to split there). A NULL is returned if
1049  *	the file name is too long
1050  */
1051 
1052 static char *
1053 name_split(char *name, int len)
1054 {
1055 	char *start;
1056 
1057 	/*
1058 	 * check to see if the file name is small enough to fit in the name
1059 	 * field. if so just return a pointer to the name.
1060 	 */
1061 	if (len <= TNMSZ)
1062 		return(name);
1063 	if (len > TPFSZ + TNMSZ)
1064 		return(NULL);
1065 
1066 	/*
1067 	 * we start looking at the biggest sized piece that fits in the name
1068 	 * field. We walk forward looking for a slash to split at. The idea is
1069 	 * to find the biggest piece to fit in the name field (or the smallest
1070 	 * prefix we can find)
1071 	 */
1072 	start = name + len - TNMSZ;
1073 	while ((*start != '\0') && (*start != '/'))
1074 		++start;
1075 
1076 	/*
1077 	 * if we hit the end of the string, this name cannot be split, so we
1078 	 * cannot store this file.
1079 	 */
1080 	if (*start == '\0')
1081 		return(NULL);
1082 	len = start - name;
1083 
1084 	/*
1085 	 * NOTE: /str where the length of str == TNMSZ can not be stored under
1086 	 * the p1003.1-1990 spec for ustar. We could force a prefix of / and
1087 	 * the file would then expand on extract to //str. The len == 0 below
1088 	 * makes this special case follow the spec to the letter.
1089 	 */
1090 	if ((len > TPFSZ) || (len == 0))
1091 		return(NULL);
1092 
1093 	/*
1094 	 * ok have a split point, return it to the caller
1095 	 */
1096 	return(start);
1097 }
1098