xref: /netbsd-src/bin/pax/tar.c (revision ccd9df534e375a4366c5b55f23782053c7a98d82)
1 /*	$NetBSD: tar.c,v 1.75 2019/03/20 03:13:39 gutteridge 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. 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 #if HAVE_NBTOOL_CONFIG_H
37 #include "nbtool_config.h"
38 #endif
39 
40 #include <sys/cdefs.h>
41 #if !defined(lint)
42 #if 0
43 static char sccsid[] = "@(#)tar.c	8.2 (Berkeley) 4/18/94";
44 #else
45 __RCSID("$NetBSD: tar.c,v 1.75 2019/03/20 03:13:39 gutteridge Exp $");
46 #endif
47 #endif /* not lint */
48 
49 #include <sys/types.h>
50 #include <sys/time.h>
51 #include <sys/stat.h>
52 #include <sys/param.h>
53 
54 #include <ctype.h>
55 #include <errno.h>
56 #include <grp.h>
57 #include <pwd.h>
58 #include <stdio.h>
59 #include <stdlib.h>
60 #include <string.h>
61 #include <unistd.h>
62 
63 #include "pax.h"
64 #include "extern.h"
65 #include "tar.h"
66 
67 extern struct stat tst;
68 
69 /*
70  * Routines for reading, writing and header identify of various versions of tar
71  */
72 
73 static int expandname(char *, size_t,  char **, size_t *, const char *, size_t);
74 static void longlink(ARCHD *, int);
75 static uint32_t tar_chksm(char *, int);
76 static char *name_split(char *, int);
77 static int u32_oct(uintmax_t, char *, int, int);
78 static int umax_oct(uintmax_t, char *, int, int);
79 static int tar_gnutar_exclude_one(const char *, size_t);
80 static int check_sum(char *, size_t, char *, size_t, int);
81 
82 /*
83  * Routines common to all versions of tar
84  */
85 
86 static int tar_nodir;			/* do not write dirs under old tar */
87 int is_gnutar;				/* behave like gnu tar; enable gnu
88 					 * extensions and skip end-of-volume
89 					 * checks
90 					 */
91 static int seen_gnu_warning;		/* Have we warned yet? */
92 static char *gnu_hack_string;		/* ././@LongLink hackery */
93 static int gnu_hack_len;		/* len of gnu_hack_string */
94 char *gnu_name_string;			/* ././@LongLink hackery name */
95 char *gnu_link_string;			/* ././@LongLink hackery link */
96 size_t gnu_name_length;			/* ././@LongLink hackery name */
97 size_t gnu_link_length;			/* ././@LongLink hackery link */
98 static int gnu_short_trailer;		/* gnu short trailer */
99 
100 static const char LONG_LINK[] = "././@LongLink";
101 
102 #ifdef _PAX_
103 char DEV_0[] = "/dev/rst0";
104 char DEV_1[] = "/dev/rst1";
105 char DEV_4[] = "/dev/rst4";
106 char DEV_5[] = "/dev/rst5";
107 char DEV_7[] = "/dev/rst7";
108 char DEV_8[] = "/dev/rst8";
109 #endif
110 
111 static int
112 check_sum(char *hd, size_t hdlen, char *bl, size_t bllen, int quiet)
113 {
114 	uint32_t hdck, blck;
115 
116 	hdck = asc_u32(hd, hdlen, OCT);
117 	blck = tar_chksm(bl, bllen);
118 
119 	if (hdck != blck) {
120 		if (!quiet)
121 			tty_warn(0, "Header checksum %o does not match %o",
122 			    hdck, blck);
123 		return -1;
124 	}
125 	return 0;
126 }
127 
128 
129 /*
130  * tar_endwr()
131  *	add the tar trailer of two null blocks
132  * Return:
133  *	0 if ok, -1 otherwise (what wr_skip returns)
134  */
135 
136 int
137 tar_endwr(void)
138 {
139 	return wr_skip((off_t)(NULLCNT * BLKMULT));
140 }
141 
142 /*
143  * tar_endrd()
144  *	no cleanup needed here, just return size of trailer (for append)
145  * Return:
146  *	size of trailer BLKMULT
147  */
148 
149 off_t
150 tar_endrd(void)
151 {
152 	return (off_t)((gnu_short_trailer ? 1 : NULLCNT) * BLKMULT);
153 }
154 
155 /*
156  * tar_trail()
157  *	Called to determine if a header block is a valid trailer. We are passed
158  *	the block, the in_sync flag (which tells us we are in resync mode;
159  *	looking for a valid header), and cnt (which starts at zero) which is
160  *	used to count the number of empty blocks we have seen so far.
161  * Return:
162  *	0 if a valid trailer, -1 if not a valid trailer, or 1 if the block
163  *	could never contain a header.
164  */
165 
166 int
167 tar_trail(char *buf, int in_resync, int *cnt)
168 {
169 	int i;
170 
171 	gnu_short_trailer = 0;
172 	/*
173 	 * look for all zero, trailer is two consecutive blocks of zero
174 	 */
175 	for (i = 0; i < BLKMULT; ++i) {
176 		if (buf[i] != '\0')
177 			break;
178 	}
179 
180 	/*
181 	 * if not all zero it is not a trailer, but MIGHT be a header.
182 	 */
183 	if (i != BLKMULT)
184 		return -1;
185 
186 	/*
187 	 * When given a zero block, we must be careful!
188 	 * If we are not in resync mode, check for the trailer. Have to watch
189 	 * out that we do not mis-identify file data as the trailer, so we do
190 	 * NOT try to id a trailer during resync mode. During resync mode we
191 	 * might as well throw this block out since a valid header can NEVER be
192 	 * a block of all 0 (we must have a valid file name).
193 	 */
194 	if (!in_resync) {
195 		++*cnt;
196 		/*
197 		 * old GNU tar (up through 1.13) only writes one block of
198 		 * trailers, so we pretend we got another
199 		 */
200 		if (is_gnutar) {
201 			gnu_short_trailer = 1;
202 			++*cnt;
203 		}
204 		if (*cnt >= NULLCNT)
205 			return 0;
206 	}
207 	return 1;
208 }
209 
210 /*
211  * u32_oct()
212  *	convert an uintmax_t to an octal string. many oddball field
213  *	termination characters are used by the various versions of tar in the
214  *	different fields. term selects which kind to use. str is '0' padded
215  *	at the front to len. we are unable to use only one format as many old
216  *	tar readers are very cranky about this.
217  * Return:
218  *	0 if the number fit into the string, -1 otherwise
219  */
220 
221 static int
222 u32_oct(uintmax_t val, char *str, int len, int term)
223 {
224 	char *pt;
225 	uint64_t p;
226 
227 	p = val & TOP_HALF;
228 	if (p && p != TOP_HALF)
229 		return -1;
230 
231 	val &= BOTTOM_HALF;
232 
233 	/*
234 	 * term selects the appropriate character(s) for the end of the string
235 	 */
236 	pt = str + len - 1;
237 	switch(term) {
238 	case 3:
239 		*pt-- = '\0';
240 		break;
241 	case 2:
242 		*pt-- = ' ';
243 		*pt-- = '\0';
244 		break;
245 	case 1:
246 		*pt-- = ' ';
247 		break;
248 	case 0:
249 	default:
250 		*pt-- = '\0';
251 		*pt-- = ' ';
252 		break;
253 	}
254 
255 	/*
256 	 * convert and blank pad if there is space
257 	 */
258 	while (pt >= str) {
259 		*pt-- = '0' + (char)(val & 0x7);
260 		if ((val = val >> 3) == 0)
261 			break;
262 	}
263 
264 	while (pt >= str)
265 		*pt-- = '0';
266 	if (val != 0)
267 		return -1;
268 	return 0;
269 }
270 
271 /*
272  * umax_oct()
273  *	convert an unsigned long long to an octal string. one of many oddball
274  *	field termination characters are used by the various versions of tar
275  *	in the different fields. term selects which kind to use. str is '0'
276  *	padded at the front to len. we are unable to use only one format as
277  *	many old tar readers are very cranky about this.
278  * Return:
279  *	0 if the number fit into the string, -1 otherwise
280  */
281 
282 static int
283 umax_oct(uintmax_t val, char *str, int len, int term)
284 {
285 	char *pt;
286 
287 	/*
288 	 * term selects the appropriate character(s) for the end of the string
289 	 */
290 	pt = str + len - 1;
291 	switch(term) {
292 	case 3:
293 		*pt-- = '\0';
294 		break;
295 	case 2:
296 		*pt-- = ' ';
297 		*pt-- = '\0';
298 		break;
299 	case 1:
300 		*pt-- = ' ';
301 		break;
302 	case 0:
303 	default:
304 		*pt-- = '\0';
305 		*pt-- = ' ';
306 		break;
307 	}
308 
309 	/*
310 	 * convert and blank pad if there is space
311 	 */
312 	while (pt >= str) {
313 		*pt-- = '0' + (char)(val & 0x7);
314 		if ((val = val >> 3) == 0)
315 			break;
316 	}
317 
318 	while (pt >= str)
319 		*pt-- = '0';
320 	if (val != 0)
321 		return -1;
322 	return 0;
323 }
324 
325 /*
326  * tar_chksm()
327  *	calculate the checksum for a tar block counting the checksum field as
328  *	all blanks (BLNKSUM is that value pre-calculated, the sum of 8 blanks).
329  *	NOTE: we use len to short circuit summing 0's on write since we ALWAYS
330  *	pad headers with 0.
331  * Return:
332  *	unsigned long checksum
333  */
334 
335 static uint32_t
336 tar_chksm(char *blk, int len)
337 {
338 	char *stop;
339 	char *pt;
340 	uint32_t chksm = BLNKSUM;	/* initial value is checksum field sum */
341 
342 	/*
343 	 * add the part of the block before the checksum field
344 	 */
345 	pt = blk;
346 	stop = blk + CHK_OFFSET;
347 	while (pt < stop)
348 		chksm += (uint32_t)(*pt++ & 0xff);
349 	/*
350 	 * move past the checksum field and keep going, spec counts the
351 	 * checksum field as the sum of 8 blanks (which is pre-computed as
352 	 * BLNKSUM).
353 	 * ASSUMED: len is greater than CHK_OFFSET. (len is where our 0 padding
354 	 * starts, no point in summing zeros)
355 	 */
356 	pt += CHK_LEN;
357 	stop = blk + len;
358 	while (pt < stop)
359 		chksm += (uint32_t)(*pt++ & 0xff);
360 	return chksm;
361 }
362 
363 /*
364  * Routines for old BSD style tar (also made portable to sysV tar)
365  */
366 
367 /*
368  * tar_id()
369  *	determine if a block given to us is a valid tar header (and not a USTAR
370  *	header). We have to be on the lookout for those pesky blocks of	all
371  *	zeros.
372  * Return:
373  *	0 if a tar header, -1 otherwise
374  */
375 
376 int
377 tar_id(char *blk, int size)
378 {
379 	HD_TAR *hd;
380 	HD_USTAR *uhd;
381 	static int is_ustar = -1;
382 
383 	if (size < BLKMULT)
384 		return -1;
385 	hd = (HD_TAR *)blk;
386 	uhd = (HD_USTAR *)blk;
387 
388 	/*
389 	 * check for block of zeros first, a simple and fast test, then make
390 	 * sure this is not a ustar header by looking for the ustar magic
391 	 * cookie. We should use TMAGLEN, but some USTAR archive programs are
392 	 * wrong and create archives missing the \0. Last we check the
393 	 * checksum. If this is ok we have to assume it is a valid header.
394 	 */
395 	if (hd->name[0] == '\0')
396 		return -1;
397 	if (strncmp(uhd->magic, TMAGIC, TMAGLEN - 1) == 0) {
398 		if (is_ustar == -1) {
399 			is_ustar = 1;
400 			return -1;
401 		} else
402 			tty_warn(0,
403 			    "Busted tar archive: has both ustar and old tar "
404 			    "records");
405 	} else
406 		is_ustar = 0;
407 	return check_sum(hd->chksum, sizeof(hd->chksum), blk, BLKMULT, 1);
408 }
409 
410 /*
411  * tar_opt()
412  *	handle tar format specific -o options
413  * Return:
414  *	0 if ok -1 otherwise
415  */
416 
417 int
418 tar_opt(void)
419 {
420 	OPLIST *opt;
421 
422 	while ((opt = opt_next()) != NULL) {
423 		if (strcmp(opt->name, TAR_OPTION) ||
424 		    strcmp(opt->value, TAR_NODIR)) {
425 			tty_warn(1,
426 			    "Unknown tar format -o option/value pair %s=%s",
427 			    opt->name, opt->value);
428 			tty_warn(1,
429 			    "%s=%s is the only supported tar format option",
430 			    TAR_OPTION, TAR_NODIR);
431 			return -1;
432 		}
433 
434 		/*
435 		 * we only support one option, and only when writing
436 		 */
437 		if ((act != APPND) && (act != ARCHIVE)) {
438 			tty_warn(1, "%s=%s is only supported when writing.",
439 			    opt->name, opt->value);
440 			return -1;
441 		}
442 		tar_nodir = 1;
443 	}
444 	return 0;
445 }
446 
447 
448 /*
449  * tar_rd()
450  *	extract the values out of block already determined to be a tar header.
451  *	store the values in the ARCHD parameter.
452  * Return:
453  *	0
454  */
455 
456 int
457 tar_rd(ARCHD *arcn, char *buf)
458 {
459 	HD_TAR *hd;
460 	char *pt;
461 
462 	/*
463 	 * we only get proper sized buffers passed to us
464 	 */
465 	if (tar_id(buf, BLKMULT) < 0)
466 		return -1;
467 	memset(arcn, 0, sizeof(*arcn));
468 	arcn->org_name = arcn->name;
469 	arcn->pat = NULL;
470 	arcn->sb.st_nlink = 1;
471 
472 	/*
473 	 * copy out the name and values in the stat buffer
474 	 */
475 	hd = (HD_TAR *)buf;
476 	if (hd->linkflag != LONGLINKTYPE && hd->linkflag != LONGNAMETYPE) {
477 		arcn->nlen = expandname(arcn->name, sizeof(arcn->name),
478 		    &gnu_name_string, &gnu_name_length, hd->name,
479 		    sizeof(hd->name));
480 		arcn->ln_nlen = expandname(arcn->ln_name, sizeof(arcn->ln_name),
481 		    &gnu_link_string, &gnu_link_length, hd->linkname,
482 		    sizeof(hd->linkname));
483 	}
484 	arcn->sb.st_mode = (mode_t)(asc_u32(hd->mode,sizeof(hd->mode),OCT) &
485 	    0xfff);
486 	arcn->sb.st_uid = (uid_t)asc_u32(hd->uid, sizeof(hd->uid), OCT);
487 	arcn->sb.st_gid = (gid_t)asc_u32(hd->gid, sizeof(hd->gid), OCT);
488 	arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT);
489 	if (arcn->sb.st_size == -1)
490 		return -1;
491 	arcn->sb.st_mtime = (time_t)(int32_t)asc_u32(hd->mtime, sizeof(hd->mtime), OCT);
492 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
493 
494 	/*
495 	 * have to look at the last character, it may be a '/' and that is used
496 	 * to encode this as a directory
497 	 */
498 	pt = &(arcn->name[arcn->nlen - 1]);
499 	arcn->pad = 0;
500 	arcn->skip = 0;
501 	switch(hd->linkflag) {
502 	case SYMTYPE:
503 		/*
504 		 * symbolic link, need to get the link name and set the type in
505 		 * the st_mode so -v printing will look correct.
506 		 */
507 		arcn->type = PAX_SLK;
508 		arcn->sb.st_mode |= S_IFLNK;
509 		break;
510 	case LNKTYPE:
511 		/*
512 		 * hard link, need to get the link name, set the type in the
513 		 * st_mode and st_nlink so -v printing will look better.
514 		 */
515 		arcn->type = PAX_HLK;
516 		arcn->sb.st_nlink = 2;
517 
518 		/*
519 		 * no idea of what type this thing really points at, but
520 		 * we set something for printing only.
521 		 */
522 		arcn->sb.st_mode |= S_IFREG;
523 		break;
524 	case LONGLINKTYPE:
525 	case LONGNAMETYPE:
526 		/*
527 		 * GNU long link/file; we tag these here and let the
528 		 * pax internals deal with it -- too ugly otherwise.
529 		 */
530 		if (hd->linkflag != LONGLINKTYPE)
531 			arcn->type = PAX_GLF;
532 		else
533 			arcn->type = PAX_GLL;
534 		arcn->pad = TAR_PAD(arcn->sb.st_size);
535 		arcn->skip = arcn->sb.st_size;
536 		break;
537 	case AREGTYPE:
538 	case REGTYPE:
539 	case DIRTYPE:	/* see below */
540 	default:
541 		/*
542 		 * If we have a trailing / this is a directory and NOT a file.
543 		 * Note: V7 tar doesn't actually have DIRTYPE, but it was
544 		 * reported that V7 archives using USTAR directories do exist.
545 		 */
546 		if (*pt == '/' || hd->linkflag == DIRTYPE) {
547 			/*
548 			 * it is a directory, set the mode for -v printing
549 			 */
550 			arcn->type = PAX_DIR;
551 			arcn->sb.st_mode |= S_IFDIR;
552 			arcn->sb.st_nlink = 2;
553 		} else {
554 			/*
555 			 * have a file that will be followed by data. Set the
556 			 * skip value to the size field and calculate the size
557 			 * of the padding.
558 			 */
559 			arcn->type = PAX_REG;
560 			arcn->sb.st_mode |= S_IFREG;
561 			arcn->pad = TAR_PAD(arcn->sb.st_size);
562 			arcn->skip = arcn->sb.st_size;
563 		}
564 		break;
565 	}
566 
567 	/*
568 	 * strip off any trailing slash.
569 	 */
570 	if (*pt == '/') {
571 		*pt = '\0';
572 		--arcn->nlen;
573 	}
574 	return 0;
575 }
576 
577 /*
578  * tar_wr()
579  *	write a tar header for the file specified in the ARCHD to the archive.
580  *	Have to check for file types that cannot be stored and file names that
581  *	are too long. Be careful of the term (last arg) to u32_oct, each field
582  *	of tar has it own spec for the termination character(s).
583  *	ASSUMED: space after header in header block is zero filled
584  * Return:
585  *	0 if file has data to be written after the header, 1 if file has NO
586  *	data to write after the header, -1 if archive write failed
587  */
588 
589 int
590 tar_wr(ARCHD *arcn)
591 {
592 	HD_TAR *hd;
593 	int len;
594 	uintmax_t mtime;
595 	char hdblk[sizeof(HD_TAR)];
596 
597 	/*
598 	 * check for those file system types which tar cannot store
599 	 */
600 	switch(arcn->type) {
601 	case PAX_DIR:
602 		/*
603 		 * user asked that dirs not be written to the archive
604 		 */
605 		if (tar_nodir)
606 			return 1;
607 		break;
608 	case PAX_CHR:
609 		tty_warn(1, "Tar cannot archive a character device %s",
610 		    arcn->org_name);
611 		return 1;
612 	case PAX_BLK:
613 		tty_warn(1,
614 		    "Tar cannot archive a block device %s", arcn->org_name);
615 		return 1;
616 	case PAX_SCK:
617 		tty_warn(1, "Tar cannot archive a socket %s", arcn->org_name);
618 		return 1;
619 	case PAX_FIF:
620 		tty_warn(1, "Tar cannot archive a fifo %s", arcn->org_name);
621 		return 1;
622 	case PAX_SLK:
623 	case PAX_HLK:
624 	case PAX_HRG:
625 		if (arcn->ln_nlen > (int)sizeof(hd->linkname)) {
626 			tty_warn(1,"Link name too long for tar %s",
627 			    arcn->ln_name);
628 			return 1;
629 		}
630 		break;
631 	case PAX_REG:
632 	case PAX_CTG:
633 	default:
634 		break;
635 	}
636 
637 	/*
638 	 * check file name len, remember extra char for dirs (the / at the end)
639 	 */
640 	len = arcn->nlen;
641 	if (arcn->type == PAX_DIR)
642 		++len;
643 	if (len >= (int)sizeof(hd->name)) {
644 		tty_warn(1, "File name too long for tar %s", arcn->name);
645 		return 1;
646 	}
647 
648 	/*
649 	 * copy the data out of the ARCHD into the tar header based on the type
650 	 * of the file. Remember many tar readers want the unused fields to be
651 	 * padded with zero. We set the linkflag field (type), the linkname
652 	 * (or zero if not used),the size, and set the padding (if any) to be
653 	 * added after the file data (0 for all other types, as they only have
654 	 * a header)
655 	 */
656 	memset(hdblk, 0, sizeof(hdblk));
657 	hd = (HD_TAR *)hdblk;
658 	strlcpy(hd->name, arcn->name, sizeof(hd->name));
659 	arcn->pad = 0;
660 
661 	if (arcn->type == PAX_DIR) {
662 		/*
663 		 * directories are the same as files, except have a filename
664 		 * that ends with a /, we add the slash here. No data follows,
665 		 * dirs, so no pad.
666 		 */
667 		hd->linkflag = AREGTYPE;
668 		hd->name[len-1] = '/';
669 		if (u32_oct((uintmax_t)0L, hd->size, sizeof(hd->size), 1))
670 			goto out;
671 	} else if (arcn->type == PAX_SLK) {
672 		/*
673 		 * no data follows this file, so no pad
674 		 */
675 		hd->linkflag = SYMTYPE;
676 		strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname));
677 		if (u32_oct((uintmax_t)0L, hd->size, sizeof(hd->size), 1))
678 			goto out;
679 	} else if ((arcn->type == PAX_HLK) || (arcn->type == PAX_HRG)) {
680 		/*
681 		 * no data follows this file, so no pad
682 		 */
683 		hd->linkflag = LNKTYPE;
684 		strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname));
685 		if (u32_oct((uintmax_t)0L, hd->size, sizeof(hd->size), 1))
686 			goto out;
687 	} else {
688 		/*
689 		 * data follows this file, so set the pad
690 		 */
691 		hd->linkflag = AREGTYPE;
692 		if (OFFT_OCT(arcn->sb.st_size, hd->size, sizeof(hd->size), 1)) {
693 			tty_warn(1,"File is too large for tar %s",
694 			    arcn->org_name);
695 			return 1;
696 		}
697 		arcn->pad = TAR_PAD(arcn->sb.st_size);
698 	}
699 
700 	/*
701 	 * copy those fields that are independent of the type
702 	 */
703 	mtime = tst.st_ino ? tst.st_mtime : arcn->sb.st_mtime;
704 	if (u32_oct((uintmax_t)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 0) ||
705 	    u32_oct((uintmax_t)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 0) ||
706 	    u32_oct((uintmax_t)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 0) ||
707 	    u32_oct(mtime, hd->mtime, sizeof(hd->mtime), 1))
708 		goto out;
709 
710 	/*
711 	 * calculate and add the checksum, then write the header. A return of
712 	 * 0 tells the caller to now write the file data, 1 says no data needs
713 	 * to be written
714 	 */
715 	if (u32_oct(tar_chksm(hdblk, sizeof(HD_TAR)), hd->chksum,
716 	    sizeof(hd->chksum), 3))
717 		goto out;			/* XXX Something's wrong here
718 						 * because a zero-byte file can
719 						 * cause this to be done and
720 						 * yet the resulting warning
721 						 * seems incorrect */
722 
723 	if (wr_rdbuf(hdblk, sizeof(HD_TAR)) < 0)
724 		return -1;
725 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_TAR))) < 0)
726 		return -1;
727 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
728 		return 0;
729 	return 1;
730 
731     out:
732 	/*
733 	 * header field is out of range
734 	 */
735 	tty_warn(1, "Tar header field is too small for %s", arcn->org_name);
736 	return 1;
737 }
738 
739 /*
740  * Routines for POSIX ustar
741  */
742 
743 /*
744  * ustar_strd()
745  *	initialization for ustar read
746  * Return:
747  *	0 if ok, -1 otherwise
748  */
749 
750 int
751 ustar_strd(void)
752 {
753 	return 0;
754 }
755 
756 /*
757  * ustar_stwr()
758  *	initialization for ustar write
759  * Return:
760  *	0 if ok, -1 otherwise
761  */
762 
763 int
764 ustar_stwr(void)
765 {
766 	return 0;
767 }
768 
769 /*
770  * ustar_id()
771  *	determine if a block given to us is a valid ustar header. We have to
772  *	be on the lookout for those pesky blocks of all zeros
773  * Return:
774  *	0 if a ustar header, -1 otherwise
775  */
776 
777 int
778 ustar_id(char *blk, int size)
779 {
780 	HD_USTAR *hd;
781 
782 	if (size < BLKMULT)
783 		return -1;
784 	hd = (HD_USTAR *)blk;
785 
786 	/*
787 	 * check for block of zeros first, a simple and fast test then check
788 	 * ustar magic cookie. We should use TMAGLEN, but some USTAR archive
789 	 * programs are fouled up and create archives missing the \0. Last we
790 	 * check the checksum. If ok we have to assume it is a valid header.
791 	 */
792 	if (hd->name[0] == '\0')
793 		return -1;
794 	if (strncmp(hd->magic, TMAGIC, TMAGLEN - 1) != 0)
795 		return -1;
796 	/* This is GNU tar */
797 	if (strncmp(hd->magic, "ustar  ", 8) == 0 && !is_gnutar &&
798 	    !seen_gnu_warning) {
799 		seen_gnu_warning = 1;
800 		tty_warn(0,
801 		    "Trying to read GNU tar archive with GNU extensions and end-of-volume checks off");
802 	}
803 	return check_sum(hd->chksum, sizeof(hd->chksum), blk, BLKMULT, 0);
804 }
805 
806 /*
807  * ustar_rd()
808  *	extract the values out of block already determined to be a ustar header.
809  *	store the values in the ARCHD parameter.
810  * Return:
811  *	0
812  */
813 
814 int
815 ustar_rd(ARCHD *arcn, char *buf)
816 {
817 	HD_USTAR *hd;
818 	char *dest;
819 	int cnt;
820 	dev_t devmajor;
821 	dev_t devminor;
822 
823 	/*
824 	 * we only get proper sized buffers
825 	 */
826 	if (ustar_id(buf, BLKMULT) < 0)
827 		return -1;
828 
829 	memset(arcn, 0, sizeof(*arcn));
830 	arcn->org_name = arcn->name;
831 	arcn->pat = NULL;
832 	arcn->sb.st_nlink = 1;
833 	hd = (HD_USTAR *)buf;
834 
835 	/*
836 	 * see if the filename is split into two parts. if, so joint the parts.
837 	 * we copy the prefix first and add a / between the prefix and name.
838 	 */
839 	dest = arcn->name;
840 	if (*(hd->prefix) != '\0') {
841 		cnt = strlcpy(arcn->name, hd->prefix, sizeof(arcn->name));
842 		dest += cnt;
843 		*dest++ = '/';
844 		cnt++;
845 	} else {
846 		cnt = 0;
847 	}
848 
849 	if (hd->typeflag != LONGLINKTYPE && hd->typeflag != LONGNAMETYPE) {
850 		arcn->nlen = expandname(dest, sizeof(arcn->name) - cnt,
851 		    &gnu_name_string, &gnu_name_length, hd->name,
852 		    sizeof(hd->name)) + cnt;
853 		arcn->ln_nlen = expandname(arcn->ln_name,
854 		    sizeof(arcn->ln_name), &gnu_link_string, &gnu_link_length,
855 		    hd->linkname, sizeof(hd->linkname));
856 	}
857 
858 	/*
859 	 * follow the spec to the letter. we should only have mode bits, strip
860 	 * off all other crud we may be passed.
861 	 */
862 	arcn->sb.st_mode = (mode_t)(asc_u32(hd->mode, sizeof(hd->mode), OCT) &
863 	    0xfff);
864 	arcn->sb.st_size = (off_t)ASC_OFFT(hd->size, sizeof(hd->size), OCT);
865 	if (arcn->sb.st_size == -1)
866 		return -1;
867 	arcn->sb.st_mtime = (time_t)(int32_t)asc_u32(hd->mtime, sizeof(hd->mtime), OCT);
868 	arcn->sb.st_ctime = arcn->sb.st_atime = arcn->sb.st_mtime;
869 
870 	/*
871 	 * If we can find the ascii names for gname and uname in the password
872 	 * and group files we will use the uid's and gid they bind. Otherwise
873 	 * we use the uid and gid values stored in the header. (This is what
874 	 * the posix spec wants).
875 	 */
876 	hd->gname[sizeof(hd->gname) - 1] = '\0';
877 	if (gid_from_group(hd->gname, &(arcn->sb.st_gid)) < 0)
878 		arcn->sb.st_gid = (gid_t)asc_u32(hd->gid, sizeof(hd->gid), OCT);
879 	hd->uname[sizeof(hd->uname) - 1] = '\0';
880 	if (uid_from_user(hd->uname, &(arcn->sb.st_uid)) < 0)
881 		arcn->sb.st_uid = (uid_t)asc_u32(hd->uid, sizeof(hd->uid), OCT);
882 
883 	/*
884 	 * set the defaults, these may be changed depending on the file type
885 	 */
886 	arcn->pad = 0;
887 	arcn->skip = 0;
888 	arcn->sb.st_rdev = (dev_t)0;
889 
890 	/*
891 	 * set the mode and PAX type according to the typeflag in the header
892 	 */
893 	switch(hd->typeflag) {
894 	case FIFOTYPE:
895 		arcn->type = PAX_FIF;
896 		arcn->sb.st_mode |= S_IFIFO;
897 		break;
898 	case DIRTYPE:
899 		arcn->type = PAX_DIR;
900 		arcn->sb.st_mode |= S_IFDIR;
901 		arcn->sb.st_nlink = 2;
902 
903 		/*
904 		 * Some programs that create ustar archives append a '/'
905 		 * to the pathname for directories. This clearly violates
906 		 * ustar specs, but we will silently strip it off anyway.
907 		 */
908 		if (arcn->name[arcn->nlen - 1] == '/')
909 			arcn->name[--arcn->nlen] = '\0';
910 		break;
911 	case BLKTYPE:
912 	case CHRTYPE:
913 		/*
914 		 * this type requires the rdev field to be set.
915 		 */
916 		if (hd->typeflag == BLKTYPE) {
917 			arcn->type = PAX_BLK;
918 			arcn->sb.st_mode |= S_IFBLK;
919 		} else {
920 			arcn->type = PAX_CHR;
921 			arcn->sb.st_mode |= S_IFCHR;
922 		}
923 		devmajor = (dev_t)asc_u32(hd->devmajor,sizeof(hd->devmajor),OCT);
924 		devminor = (dev_t)asc_u32(hd->devminor,sizeof(hd->devminor),OCT);
925 		arcn->sb.st_rdev = TODEV(devmajor, devminor);
926 		break;
927 	case SYMTYPE:
928 	case LNKTYPE:
929 		if (hd->typeflag == SYMTYPE) {
930 			arcn->type = PAX_SLK;
931 			arcn->sb.st_mode |= S_IFLNK;
932 		} else {
933 			arcn->type = PAX_HLK;
934 			/*
935 			 * so printing looks better
936 			 */
937 			arcn->sb.st_mode |= S_IFREG;
938 			arcn->sb.st_nlink = 2;
939 		}
940 		break;
941 	case LONGLINKTYPE:
942 	case LONGNAMETYPE:
943 		if (is_gnutar) {
944 			/*
945 			 * GNU long link/file; we tag these here and let the
946 			 * pax internals deal with it -- too ugly otherwise.
947 			 */
948 			if (hd->typeflag != LONGLINKTYPE)
949 				arcn->type = PAX_GLF;
950 			else
951 				arcn->type = PAX_GLL;
952 			arcn->pad = TAR_PAD(arcn->sb.st_size);
953 			arcn->skip = arcn->sb.st_size;
954 		} else {
955 			tty_warn(1, "GNU Long %s found in posix ustar archive.",
956 			    hd->typeflag == LONGLINKTYPE ? "Link" : "File");
957 		}
958 		break;
959 	case FILEXTYPE:
960 	case GLOBXTYPE:
961 		    tty_warn(0, "%s extended headers posix ustar archive."
962 		    " Extracting as plain files. Following files might be"
963 		    " in the wrong directory or have wrong attributes.",
964 			hd->typeflag == FILEXTYPE ? "File" : "Global");
965 		    /*FALLTHROUGH*/
966 	case CONTTYPE:
967 	case AREGTYPE:
968 	case REGTYPE:
969 	default:
970 		/*
971 		 * these types have file data that follows. Set the skip and
972 		 * pad fields.
973 		 */
974 		arcn->type = PAX_REG;
975 		arcn->pad = TAR_PAD(arcn->sb.st_size);
976 		arcn->skip = arcn->sb.st_size;
977 		arcn->sb.st_mode |= S_IFREG;
978 		break;
979 	}
980 	return 0;
981 }
982 
983 static int
984 expandname(char *buf, size_t len, char **gnu_name, size_t *gnu_length,
985     const char *name, size_t nlen)
986 {
987 	if (*gnu_name) {
988 		len = strlcpy(buf, *gnu_name, len);
989 		free(*gnu_name);
990 		*gnu_name = NULL;
991 		*gnu_length = 0;
992 	} else {
993 		if (len > ++nlen)
994 			len = nlen;
995 		len = strlcpy(buf, name, len);
996 	}
997 	return len;
998 }
999 
1000 static void
1001 longlink(ARCHD *arcn, int type)
1002 {
1003 	ARCHD larc;
1004 
1005 	(void)memset(&larc, 0, sizeof(larc));
1006 
1007 	larc.type = type;
1008 	larc.nlen = strlcpy(larc.name, LONG_LINK, sizeof(larc.name));
1009 
1010 	switch (type) {
1011 	case PAX_GLL:
1012 		gnu_hack_string = arcn->ln_name;
1013 		gnu_hack_len = arcn->ln_nlen + 1;
1014 		break;
1015 	case PAX_GLF:
1016 		gnu_hack_string = arcn->name;
1017 		gnu_hack_len = arcn->nlen + 1;
1018 		break;
1019 	default:
1020 		errx(1, "Invalid type in GNU longlink %d", type);
1021 	}
1022 
1023 	/*
1024 	 * We need a longlink now.
1025 	 */
1026 	ustar_wr(&larc);
1027 }
1028 
1029 /*
1030  * ustar_wr()
1031  *	write a ustar header for the file specified in the ARCHD to the archive
1032  *	Have to check for file types that cannot be stored and file names that
1033  *	are too long. Be careful of the term (last arg) to u32_oct, we only use
1034  *	'\0' for the termination character (this is different than picky tar)
1035  *	ASSUMED: space after header in header block is zero filled
1036  * Return:
1037  *	0 if file has data to be written after the header, 1 if file has NO
1038  *	data to write after the header, -1 if archive write failed
1039  */
1040 
1041 static int
1042 size_err(const char *what, ARCHD *arcn)
1043 {
1044 	/*
1045 	 * header field is out of range
1046 	 */
1047 	tty_warn(1, "Ustar %s header field is too small for %s",
1048 		what, arcn->org_name);
1049 	return 1;
1050 }
1051 
1052 int
1053 ustar_wr(ARCHD *arcn)
1054 {
1055 	HD_USTAR *hd;
1056 	char *pt;
1057 	uintmax_t mtime;
1058 	char hdblk[sizeof(HD_USTAR)];
1059 	const char *user, *group;
1060 
1061 	switch (arcn->type) {
1062 	case PAX_SCK:
1063 		/*
1064 		 * check for those file system types ustar cannot store
1065 		 */
1066 		if (!is_gnutar)
1067 			tty_warn(1, "Ustar cannot archive a socket %s",
1068 			    arcn->org_name);
1069 		return 1;
1070 
1071 	case PAX_SLK:
1072 	case PAX_HLK:
1073 	case PAX_HRG:
1074 		/*
1075 		 * check the length of the linkname
1076 		 */
1077 		if (arcn->ln_nlen >= (int)sizeof(hd->linkname)) {
1078 			if (is_gnutar) {
1079 				longlink(arcn, PAX_GLL);
1080 			} else {
1081 				tty_warn(1, "Link name too long for ustar %s",
1082 				    arcn->ln_name);
1083 				return 1;
1084 			}
1085 		}
1086 		break;
1087 	default:
1088 		break;
1089 	}
1090 
1091 	/*
1092 	 * split the path name into prefix and name fields (if needed). if
1093 	 * pt != arcn->name, the name has to be split
1094 	 */
1095 	if ((pt = name_split(arcn->name, arcn->nlen)) == NULL) {
1096 		if (is_gnutar) {
1097 			longlink(arcn, PAX_GLF);
1098 			pt = arcn->name;
1099 		} else {
1100 			tty_warn(1, "File name too long for ustar %s",
1101 			    arcn->name);
1102 			return 1;
1103 		}
1104 	}
1105 
1106 	/*
1107 	 * zero out the header so we don't have to worry about zero fill below
1108 	 */
1109 	memset(hdblk, 0, sizeof(hdblk));
1110 	hd = (HD_USTAR *)hdblk;
1111 	arcn->pad = 0L;
1112 
1113 	/*
1114 	 * split the name, or zero out the prefix
1115 	 */
1116 	if (pt != arcn->name) {
1117 		/*
1118 		 * name was split, pt points at the / where the split is to
1119 		 * occur, we remove the / and copy the first part to the prefix
1120 		 */
1121 		*pt = '\0';
1122 		strlcpy(hd->prefix, arcn->name, sizeof(hd->prefix));
1123 		*pt++ = '/';
1124 	}
1125 
1126 	/*
1127 	 * copy the name part. this may be the whole path or the part after
1128 	 * the prefix
1129 	 */
1130 	strlcpy(hd->name, pt, sizeof(hd->name));
1131 
1132 	/*
1133 	 * set the fields in the header that are type dependent
1134 	 */
1135 	switch(arcn->type) {
1136 	case PAX_DIR:
1137 		hd->typeflag = DIRTYPE;
1138 		if (u32_oct((uintmax_t)0L, hd->size, sizeof(hd->size), 3))
1139 			return size_err("DIRTYPE", arcn);
1140 		break;
1141 	case PAX_CHR:
1142 	case PAX_BLK:
1143 		if (arcn->type == PAX_CHR)
1144 			hd->typeflag = CHRTYPE;
1145 		else
1146 			hd->typeflag = BLKTYPE;
1147 		if (u32_oct((uintmax_t)MAJOR(arcn->sb.st_rdev), hd->devmajor,
1148 		   sizeof(hd->devmajor), 3) ||
1149 		   u32_oct((uintmax_t)MINOR(arcn->sb.st_rdev), hd->devminor,
1150 		   sizeof(hd->devminor), 3) ||
1151 		   u32_oct((uintmax_t)0L, hd->size, sizeof(hd->size), 3))
1152 			return size_err("DEVTYPE", arcn);
1153 		break;
1154 	case PAX_FIF:
1155 		hd->typeflag = FIFOTYPE;
1156 		if (u32_oct((uintmax_t)0L, hd->size, sizeof(hd->size), 3))
1157 			return size_err("FIFOTYPE", arcn);
1158 		break;
1159 	case PAX_GLL:
1160 	case PAX_SLK:
1161 	case PAX_HLK:
1162 	case PAX_HRG:
1163 		if (arcn->type == PAX_SLK)
1164 			hd->typeflag = SYMTYPE;
1165 		else if (arcn->type == PAX_GLL)
1166 			hd->typeflag = LONGLINKTYPE;
1167 		else
1168 			hd->typeflag = LNKTYPE;
1169 		strlcpy(hd->linkname, arcn->ln_name, sizeof(hd->linkname));
1170 		if (u32_oct((uintmax_t)gnu_hack_len, hd->size,
1171 		    sizeof(hd->size), 3))
1172 			return size_err("LINKTYPE", arcn);
1173 		break;
1174 	case PAX_GLF:
1175 	case PAX_REG:
1176 	case PAX_CTG:
1177 	default:
1178 		/*
1179 		 * file data with this type, set the padding
1180 		 */
1181 		if (arcn->type == PAX_GLF) {
1182 			hd->typeflag = LONGNAMETYPE;
1183 			arcn->pad = TAR_PAD(gnu_hack_len);
1184 			if (OFFT_OCT((uint32_t)gnu_hack_len, hd->size,
1185 			    sizeof(hd->size), 3)) {
1186 				tty_warn(1,"File is too long for ustar %s",
1187 				    arcn->org_name);
1188 				return 1;
1189 			}
1190 		} else {
1191 			if (arcn->type == PAX_CTG)
1192 				hd->typeflag = CONTTYPE;
1193 			else
1194 				hd->typeflag = REGTYPE;
1195 			arcn->pad = TAR_PAD(arcn->sb.st_size);
1196 			if (OFFT_OCT(arcn->sb.st_size, hd->size,
1197 			    sizeof(hd->size), 3)) {
1198 				tty_warn(1,"File is too long for ustar %s",
1199 				    arcn->org_name);
1200 				return 1;
1201 			}
1202 		}
1203 		break;
1204 	}
1205 
1206 	strncpy(hd->magic, TMAGIC, TMAGLEN);
1207 	if (is_gnutar)
1208 		hd->magic[TMAGLEN - 1] = hd->version[0] = ' ';
1209 	else
1210 		strncpy(hd->version, TVERSION, TVERSLEN);
1211 
1212 	/*
1213 	 * set the remaining fields. Some versions want all 16 bits of mode
1214 	 * we better humor them (they really do not meet spec though)....
1215 	 */
1216 	if (u32_oct((uintmax_t)arcn->sb.st_mode, hd->mode, sizeof(hd->mode), 3))
1217 		return size_err("MODE", arcn);
1218 	if (u32_oct((uintmax_t)arcn->sb.st_uid, hd->uid, sizeof(hd->uid), 3))
1219 		return size_err("UID", arcn);
1220 	if (u32_oct((uintmax_t)arcn->sb.st_gid, hd->gid, sizeof(hd->gid), 3))
1221 		return size_err("GID", arcn);
1222 	mtime = tst.st_ino ? tst.st_mtime : arcn->sb.st_mtime;
1223 	if (u32_oct(mtime, hd->mtime, sizeof(hd->mtime), 3))
1224 		return size_err("MTIME", arcn);
1225 	user = user_from_uid(arcn->sb.st_uid, 1);
1226 	group = group_from_gid(arcn->sb.st_gid, 1);
1227 	strncpy(hd->uname, user ? user : "", sizeof(hd->uname));
1228 	strncpy(hd->gname, group ? group : "", sizeof(hd->gname));
1229 
1230 	/*
1231 	 * calculate and store the checksum write the header to the archive
1232 	 * return 0 tells the caller to now write the file data, 1 says no data
1233 	 * needs to be written
1234 	 */
1235 	if (u32_oct(tar_chksm(hdblk, sizeof(HD_USTAR)), hd->chksum,
1236 	   sizeof(hd->chksum), 3))
1237 		return size_err("CHKSUM", arcn);
1238 	if (wr_rdbuf(hdblk, sizeof(HD_USTAR)) < 0)
1239 		return -1;
1240 	if (wr_skip((off_t)(BLKMULT - sizeof(HD_USTAR))) < 0)
1241 		return -1;
1242 	if (gnu_hack_string) {
1243 		int res = wr_rdbuf(gnu_hack_string, gnu_hack_len);
1244 		int pad = gnu_hack_len;
1245 		gnu_hack_string = NULL;
1246 		gnu_hack_len = 0;
1247 		if (res < 0)
1248 			return -1;
1249 		if (wr_skip((off_t)(BLKMULT - (pad % BLKMULT))) < 0)
1250 			return -1;
1251 	}
1252 	if ((arcn->type == PAX_CTG) || (arcn->type == PAX_REG))
1253 		return 0;
1254 	return 1;
1255 }
1256 
1257 /*
1258  * name_split()
1259  *	see if the name has to be split for storage in a ustar header. We try
1260  *	to fit the entire name in the name field without splitting if we can.
1261  *	The split point is always at a /
1262  * Return
1263  *	character pointer to split point (always the / that is to be removed
1264  *	if the split is not needed, the points is set to the start of the file
1265  *	name (it would violate the spec to split there). A NULL is returned if
1266  *	the file name is too long
1267  */
1268 
1269 static char *
1270 name_split(char *name, int len)
1271 {
1272 	char *start;
1273 
1274 	/*
1275 	 * check to see if the file name is small enough to fit in the name
1276 	 * field. if so just return a pointer to the name.
1277 	 */
1278 	if (len < TNMSZ)
1279 		return name;
1280 	/*
1281 	 * GNU tar does not honor the prefix+name mode if the magic
1282 	 * is not "ustar\0". So in GNU tar compatibility mode, we don't
1283 	 * split the filename into prefix+name because we are setting
1284 	 * the magic to "ustar " as GNU tar does. This of course will
1285 	 * end up creating a LongLink record in cases where it does not
1286 	 * really need do, but we are behaving like GNU tar after all.
1287 	 */
1288 	if (is_gnutar || len > (TPFSZ + TNMSZ))
1289 		return NULL;
1290 
1291 	/*
1292 	 * we start looking at the biggest sized piece that fits in the name
1293 	 * field. We walk forward looking for a slash to split at. The idea is
1294 	 * to find the biggest piece to fit in the name field (or the smallest
1295 	 * prefix we can find) (the -1 is correct the biggest piece would
1296 	 * include the slash between the two parts that gets thrown away)
1297 	 */
1298 	start = name + len - TNMSZ;
1299 	while ((*start != '\0') && (*start != '/'))
1300 		++start;
1301 
1302 	/*
1303 	 * if we hit the end of the string, this name cannot be split, so we
1304 	 * cannot store this file.
1305 	 */
1306 	if (*start == '\0')
1307 		return NULL;
1308 	len = start - name;
1309 
1310 	/*
1311 	 * NOTE: /str where the length of str == TNMSZ cannot be stored under
1312 	 * the p1003.1-1990 spec for ustar. We could force a prefix of / and
1313 	 * the file would then expand on extract to //str. The len == 0 below
1314 	 * makes this special case follow the spec to the letter.
1315 	 */
1316 	if ((len >= TPFSZ) || (len == 0))
1317 		return NULL;
1318 
1319 	/*
1320 	 * ok have a split point, return it to the caller
1321 	 */
1322 	return start;
1323 }
1324 
1325 /*
1326  * convert a glob into a RE, and add it to the list.  we convert to
1327  * four different RE's (because we're using BRE's and can't use |
1328  * alternation :-() with this padding:
1329  *	.*\/ and $
1330  *	.*\/ and \/.*
1331  *	^ and $
1332  *	^ and \/.*
1333  */
1334 static int
1335 tar_gnutar_exclude_one(const char *line, size_t len)
1336 {
1337 	/* 2 * buffer len + nul */
1338 	char sbuf[MAXPATHLEN * 2 + 1];
1339 	/* + / + // + .*""/\/ + \/.* */
1340 	char rabuf[MAXPATHLEN * 2 + 1 + 1 + 2 + 4 + 4];
1341 	size_t i;
1342 	int j = 0;
1343 
1344 	if (line[len - 1] == '\n')
1345 		len--;
1346 	for (i = 0; i < len; i++) {
1347 		/*
1348 		 * convert glob to regexp, escaping everything
1349 		 */
1350 		if (line[i] == '*')
1351 			sbuf[j++] = '.';
1352 		else if (line[i] == '?') {
1353 			sbuf[j++] = '.';
1354 			continue;
1355 		} else if (!isalnum((unsigned char)line[i]) &&
1356 		    !isblank((unsigned char)line[i]))
1357 			sbuf[j++] = '\\';
1358 		sbuf[j++] = line[i];
1359 	}
1360 	sbuf[j] = '\0';
1361 	/* don't need the .*\/ ones if we start with /, i guess */
1362 	if (line[0] != '/') {
1363 		(void)snprintf(rabuf, sizeof rabuf, "/.*\\/%s$//", sbuf);
1364 		if (rep_add(rabuf) < 0)
1365 			return (-1);
1366 		(void)snprintf(rabuf, sizeof rabuf, "/.*\\/%s\\/.*//", sbuf);
1367 		if (rep_add(rabuf) < 0)
1368 			return (-1);
1369 	}
1370 
1371 	(void)snprintf(rabuf, sizeof rabuf, "/^%s$//", sbuf);
1372 	if (rep_add(rabuf) < 0)
1373 		return (-1);
1374 	(void)snprintf(rabuf, sizeof rabuf, "/^%s\\/.*//", sbuf);
1375 	if (rep_add(rabuf) < 0)
1376 		return (-1);
1377 
1378 	return (0);
1379 }
1380 
1381 /*
1382  * deal with GNU tar -X/--exclude-from & --exclude switches.  basically,
1383  * we go through each line of the file, building a string from the "glob"
1384  * lines in the file into RE lines, of the form `/^RE$//', which we pass
1385  * to rep_add(), which will add a empty replacement (exclusion), for the
1386  * named files.
1387  */
1388 int
1389 tar_gnutar_minus_minus_exclude(const char *path)
1390 {
1391 	size_t	len = strlen(path);
1392 
1393 	if (len > MAXPATHLEN)
1394 		tty_warn(0, "pathname too long: %s", path);
1395 
1396 	return (tar_gnutar_exclude_one(path, len));
1397 }
1398 
1399 int
1400 tar_gnutar_X_compat(const char *path)
1401 {
1402 	char *line;
1403 	FILE *fp;
1404 	int lineno = 0;
1405 	size_t len;
1406 
1407 	if (path[0] == '-' && path[1] == '\0')
1408 		fp = stdin;
1409 	else {
1410 		fp = fopen(path, "r");
1411 		if (fp == NULL) {
1412 			tty_warn(1, "cannot open %s: %s", path,
1413 			    strerror(errno));
1414 			return -1;
1415 		}
1416 	}
1417 
1418 	while ((line = fgetln(fp, &len))) {
1419 		lineno++;
1420 		if (len > MAXPATHLEN) {
1421 			tty_warn(0, "pathname too long, line %d of %s",
1422 			    lineno, path);
1423 		}
1424 		if (tar_gnutar_exclude_one(line, len))
1425 			return -1;
1426 	}
1427 	if (fp != stdin)
1428 		fclose(fp);
1429 	return 0;
1430 }
1431