archive_read_support_format_mtree.c revision 307139
1/*-
2 * Copyright (c) 2003-2007 Tim Kientzle
3 * Copyright (c) 2008 Joerg Sonnenberger
4 * Copyright (c) 2011-2012 Michihiro NAKAJIMA
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 */
27
28#include "archive_platform.h"
29__FBSDID("$FreeBSD: stable/10/contrib/libarchive/libarchive/archive_read_support_format_mtree.c 307139 2016-10-12 10:28:54Z mm $");
30
31#ifdef HAVE_SYS_STAT_H
32#include <sys/stat.h>
33#endif
34#ifdef HAVE_ERRNO_H
35#include <errno.h>
36#endif
37#ifdef HAVE_FCNTL_H
38#include <fcntl.h>
39#endif
40#include <stddef.h>
41/* #include <stdint.h> */ /* See archive_platform.h */
42#ifdef HAVE_STDLIB_H
43#include <stdlib.h>
44#endif
45#ifdef HAVE_STRING_H
46#include <string.h>
47#endif
48
49#include "archive.h"
50#include "archive_entry.h"
51#include "archive_private.h"
52#include "archive_read_private.h"
53#include "archive_string.h"
54#include "archive_pack_dev.h"
55
56#ifndef O_BINARY
57#define	O_BINARY 0
58#endif
59#ifndef O_CLOEXEC
60#define O_CLOEXEC	0
61#endif
62
63#define	MTREE_HAS_DEVICE	0x0001
64#define	MTREE_HAS_FFLAGS	0x0002
65#define	MTREE_HAS_GID		0x0004
66#define	MTREE_HAS_GNAME		0x0008
67#define	MTREE_HAS_MTIME		0x0010
68#define	MTREE_HAS_NLINK		0x0020
69#define	MTREE_HAS_PERM		0x0040
70#define	MTREE_HAS_SIZE		0x0080
71#define	MTREE_HAS_TYPE		0x0100
72#define	MTREE_HAS_UID		0x0200
73#define	MTREE_HAS_UNAME		0x0400
74
75#define	MTREE_HAS_OPTIONAL	0x0800
76#define	MTREE_HAS_NOCHANGE	0x1000 /* FreeBSD specific */
77
78struct mtree_option {
79	struct mtree_option *next;
80	char *value;
81};
82
83struct mtree_entry {
84	struct mtree_entry *next;
85	struct mtree_option *options;
86	char *name;
87	char full;
88	char used;
89};
90
91struct mtree {
92	struct archive_string	 line;
93	size_t			 buffsize;
94	char			*buff;
95	int64_t			 offset;
96	int			 fd;
97	int			 archive_format;
98	const char		*archive_format_name;
99	struct mtree_entry	*entries;
100	struct mtree_entry	*this_entry;
101	struct archive_string	 current_dir;
102	struct archive_string	 contents_name;
103
104	struct archive_entry_linkresolver *resolver;
105
106	int64_t			 cur_size;
107	char checkfs;
108};
109
110static int	bid_keycmp(const char *, const char *, ssize_t);
111static int	cleanup(struct archive_read *);
112static int	detect_form(struct archive_read *, int *);
113static int	mtree_bid(struct archive_read *, int);
114static int	parse_file(struct archive_read *, struct archive_entry *,
115		    struct mtree *, struct mtree_entry *, int *);
116static void	parse_escapes(char *, struct mtree_entry *);
117static int	parse_line(struct archive_read *, struct archive_entry *,
118		    struct mtree *, struct mtree_entry *, int *);
119static int	parse_keyword(struct archive_read *, struct mtree *,
120		    struct archive_entry *, struct mtree_option *, int *);
121static int	read_data(struct archive_read *a,
122		    const void **buff, size_t *size, int64_t *offset);
123static ssize_t	readline(struct archive_read *, struct mtree *, char **, ssize_t);
124static int	skip(struct archive_read *a);
125static int	read_header(struct archive_read *,
126		    struct archive_entry *);
127static int64_t	mtree_atol10(char **);
128static int64_t	mtree_atol8(char **);
129static int64_t	mtree_atol(char **);
130
131/*
132 * There's no standard for TIME_T_MAX/TIME_T_MIN.  So we compute them
133 * here.  TODO: Move this to configure time, but be careful
134 * about cross-compile environments.
135 */
136static int64_t
137get_time_t_max(void)
138{
139#if defined(TIME_T_MAX)
140	return TIME_T_MAX;
141#else
142	/* ISO C allows time_t to be a floating-point type,
143	   but POSIX requires an integer type.  The following
144	   should work on any system that follows the POSIX
145	   conventions. */
146	if (((time_t)0) < ((time_t)-1)) {
147		/* Time_t is unsigned */
148		return (~(time_t)0);
149	} else {
150		/* Time_t is signed. */
151		/* Assume it's the same as int64_t or int32_t */
152		if (sizeof(time_t) == sizeof(int64_t)) {
153			return (time_t)INT64_MAX;
154		} else {
155			return (time_t)INT32_MAX;
156		}
157	}
158#endif
159}
160
161static int64_t
162get_time_t_min(void)
163{
164#if defined(TIME_T_MIN)
165	return TIME_T_MIN;
166#else
167	if (((time_t)0) < ((time_t)-1)) {
168		/* Time_t is unsigned */
169		return (time_t)0;
170	} else {
171		/* Time_t is signed. */
172		if (sizeof(time_t) == sizeof(int64_t)) {
173			return (time_t)INT64_MIN;
174		} else {
175			return (time_t)INT32_MIN;
176		}
177	}
178#endif
179}
180
181static int
182archive_read_format_mtree_options(struct archive_read *a,
183    const char *key, const char *val)
184{
185	struct mtree *mtree;
186
187	mtree = (struct mtree *)(a->format->data);
188	if (strcmp(key, "checkfs")  == 0) {
189		/* Allows to read information missing from the mtree from the file system */
190		if (val == NULL || val[0] == 0) {
191			mtree->checkfs = 0;
192		} else {
193			mtree->checkfs = 1;
194		}
195		return (ARCHIVE_OK);
196	}
197
198	/* Note: The "warn" return is just to inform the options
199	 * supervisor that we didn't handle it.  It will generate
200	 * a suitable error if no one used this option. */
201	return (ARCHIVE_WARN);
202}
203
204static void
205free_options(struct mtree_option *head)
206{
207	struct mtree_option *next;
208
209	for (; head != NULL; head = next) {
210		next = head->next;
211		free(head->value);
212		free(head);
213	}
214}
215
216int
217archive_read_support_format_mtree(struct archive *_a)
218{
219	struct archive_read *a = (struct archive_read *)_a;
220	struct mtree *mtree;
221	int r;
222
223	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
224	    ARCHIVE_STATE_NEW, "archive_read_support_format_mtree");
225
226	mtree = (struct mtree *)malloc(sizeof(*mtree));
227	if (mtree == NULL) {
228		archive_set_error(&a->archive, ENOMEM,
229		    "Can't allocate mtree data");
230		return (ARCHIVE_FATAL);
231	}
232	memset(mtree, 0, sizeof(*mtree));
233	mtree->fd = -1;
234
235	r = __archive_read_register_format(a, mtree, "mtree",
236           mtree_bid, archive_read_format_mtree_options, read_header, read_data, skip, NULL, cleanup, NULL, NULL);
237
238	if (r != ARCHIVE_OK)
239		free(mtree);
240	return (ARCHIVE_OK);
241}
242
243static int
244cleanup(struct archive_read *a)
245{
246	struct mtree *mtree;
247	struct mtree_entry *p, *q;
248
249	mtree = (struct mtree *)(a->format->data);
250
251	p = mtree->entries;
252	while (p != NULL) {
253		q = p->next;
254		free(p->name);
255		free_options(p->options);
256		free(p);
257		p = q;
258	}
259	archive_string_free(&mtree->line);
260	archive_string_free(&mtree->current_dir);
261	archive_string_free(&mtree->contents_name);
262	archive_entry_linkresolver_free(mtree->resolver);
263
264	free(mtree->buff);
265	free(mtree);
266	(a->format->data) = NULL;
267	return (ARCHIVE_OK);
268}
269
270static ssize_t
271get_line_size(const char *b, ssize_t avail, ssize_t *nlsize)
272{
273	ssize_t len;
274
275	len = 0;
276	while (len < avail) {
277		switch (*b) {
278		case '\0':/* Non-ascii character or control character. */
279			if (nlsize != NULL)
280				*nlsize = 0;
281			return (-1);
282		case '\r':
283			if (avail-len > 1 && b[1] == '\n') {
284				if (nlsize != NULL)
285					*nlsize = 2;
286				return (len+2);
287			}
288			/* FALL THROUGH */
289		case '\n':
290			if (nlsize != NULL)
291				*nlsize = 1;
292			return (len+1);
293		default:
294			b++;
295			len++;
296			break;
297		}
298	}
299	if (nlsize != NULL)
300		*nlsize = 0;
301	return (avail);
302}
303
304/*
305 *  <---------------- ravail --------------------->
306 *  <-- diff ------> <---  avail ----------------->
307 *                   <---- len ----------->
308 * | Previous lines | line being parsed  nl extra |
309 *                  ^
310 *                  b
311 *
312 */
313static ssize_t
314next_line(struct archive_read *a,
315    const char **b, ssize_t *avail, ssize_t *ravail, ssize_t *nl)
316{
317	ssize_t len;
318	int quit;
319
320	quit = 0;
321	if (*avail == 0) {
322		*nl = 0;
323		len = 0;
324	} else
325		len = get_line_size(*b, *avail, nl);
326	/*
327	 * Read bytes more while it does not reach the end of line.
328	 */
329	while (*nl == 0 && len == *avail && !quit) {
330		ssize_t diff = *ravail - *avail;
331		size_t nbytes_req = (*ravail+1023) & ~1023U;
332		ssize_t tested;
333
334		/* Increase reading bytes if it is not enough to at least
335		 * new two lines. */
336		if (nbytes_req < (size_t)*ravail + 160)
337			nbytes_req <<= 1;
338
339		*b = __archive_read_ahead(a, nbytes_req, avail);
340		if (*b == NULL) {
341			if (*ravail >= *avail)
342				return (0);
343			/* Reading bytes reaches the end of file. */
344			*b = __archive_read_ahead(a, *avail, avail);
345			quit = 1;
346		}
347		*ravail = *avail;
348		*b += diff;
349		*avail -= diff;
350		tested = len;/* Skip some bytes we already determinated. */
351		len = get_line_size(*b + len, *avail - len, nl);
352		if (len >= 0)
353			len += tested;
354	}
355	return (len);
356}
357
358/*
359 * Compare characters with a mtree keyword.
360 * Returns the length of a mtree keyword if matched.
361 * Returns 0 if not matched.
362 */
363static int
364bid_keycmp(const char *p, const char *key, ssize_t len)
365{
366	int match_len = 0;
367
368	while (len > 0 && *p && *key) {
369		if (*p == *key) {
370			--len;
371			++p;
372			++key;
373			++match_len;
374			continue;
375		}
376		return (0);/* Not match */
377	}
378	if (*key != '\0')
379		return (0);/* Not match */
380
381	/* A following character should be specified characters */
382	if (p[0] == '=' || p[0] == ' ' || p[0] == '\t' ||
383	    p[0] == '\n' || p[0] == '\r' ||
384	   (p[0] == '\\' && (p[1] == '\n' || p[1] == '\r')))
385		return (match_len);
386	return (0);/* Not match */
387}
388
389/*
390 * Test whether the characters 'p' has is mtree keyword.
391 * Returns the length of a detected keyword.
392 * Returns 0 if any keywords were not found.
393 */
394static int
395bid_keyword(const char *p,  ssize_t len)
396{
397	static const char *keys_c[] = {
398		"content", "contents", "cksum", NULL
399	};
400	static const char *keys_df[] = {
401		"device", "flags", NULL
402	};
403	static const char *keys_g[] = {
404		"gid", "gname", NULL
405	};
406	static const char *keys_il[] = {
407		"ignore", "inode", "link", NULL
408	};
409	static const char *keys_m[] = {
410		"md5", "md5digest", "mode", NULL
411	};
412	static const char *keys_no[] = {
413		"nlink", "nochange", "optional", NULL
414	};
415	static const char *keys_r[] = {
416		"resdevice", "rmd160", "rmd160digest", NULL
417	};
418	static const char *keys_s[] = {
419		"sha1", "sha1digest",
420		"sha256", "sha256digest",
421		"sha384", "sha384digest",
422		"sha512", "sha512digest",
423		"size", NULL
424	};
425	static const char *keys_t[] = {
426		"tags", "time", "type", NULL
427	};
428	static const char *keys_u[] = {
429		"uid", "uname",	NULL
430	};
431	const char **keys;
432	int i;
433
434	switch (*p) {
435	case 'c': keys = keys_c; break;
436	case 'd': case 'f': keys = keys_df; break;
437	case 'g': keys = keys_g; break;
438	case 'i': case 'l': keys = keys_il; break;
439	case 'm': keys = keys_m; break;
440	case 'n': case 'o': keys = keys_no; break;
441	case 'r': keys = keys_r; break;
442	case 's': keys = keys_s; break;
443	case 't': keys = keys_t; break;
444	case 'u': keys = keys_u; break;
445	default: return (0);/* Unknown key */
446	}
447
448	for (i = 0; keys[i] != NULL; i++) {
449		int l = bid_keycmp(p, keys[i], len);
450		if (l > 0)
451			return (l);
452	}
453	return (0);/* Unknown key */
454}
455
456/*
457 * Test whether there is a set of mtree keywords.
458 * Returns the number of keyword.
459 * Returns -1 if we got incorrect sequence.
460 * This function expects a set of "<space characters>keyword=value".
461 * When "unset" is specified, expects a set of "<space characters>keyword".
462 */
463static int
464bid_keyword_list(const char *p,  ssize_t len, int unset, int last_is_path)
465{
466	int l;
467	int keycnt = 0;
468
469	while (len > 0 && *p) {
470		int blank = 0;
471
472		/* Test whether there are blank characters in the line. */
473		while (len >0 && (*p == ' ' || *p == '\t')) {
474			++p;
475			--len;
476			blank = 1;
477		}
478		if (*p == '\n' || *p == '\r')
479			break;
480		if (p[0] == '\\' && (p[1] == '\n' || p[1] == '\r'))
481			break;
482		if (!blank && !last_is_path) /* No blank character. */
483			return (-1);
484		if (last_is_path && len == 0)
485				return (keycnt);
486
487		if (unset) {
488			l = bid_keycmp(p, "all", len);
489			if (l > 0)
490				return (1);
491		}
492		/* Test whether there is a correct key in the line. */
493		l = bid_keyword(p, len);
494		if (l == 0)
495			return (-1);/* Unknown keyword was found. */
496		p += l;
497		len -= l;
498		keycnt++;
499
500		/* Skip value */
501		if (*p == '=') {
502			int value = 0;
503			++p;
504			--len;
505			while (len > 0 && *p != ' ' && *p != '\t') {
506				++p;
507				--len;
508				value = 1;
509			}
510			/* A keyword should have a its value unless
511			 * "/unset" operation. */
512			if (!unset && value == 0)
513				return (-1);
514		}
515	}
516	return (keycnt);
517}
518
519static int
520bid_entry(const char *p, ssize_t len, ssize_t nl, int *last_is_path)
521{
522	int f = 0;
523	static const unsigned char safe_char[256] = {
524		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 00 - 0F */
525		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 10 - 1F */
526		/* !"$%&'()*+,-./  EXCLUSION:( )(#) */
527		0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 20 - 2F */
528		/* 0123456789:;<>?  EXCLUSION:(=) */
529		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, /* 30 - 3F */
530		/* @ABCDEFGHIJKLMNO */
531		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 40 - 4F */
532		/* PQRSTUVWXYZ[\]^_  */
533		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 50 - 5F */
534		/* `abcdefghijklmno */
535		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 60 - 6F */
536		/* pqrstuvwxyz{|}~ */
537		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, /* 70 - 7F */
538		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 80 - 8F */
539		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 90 - 9F */
540		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* A0 - AF */
541		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* B0 - BF */
542		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* C0 - CF */
543		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* D0 - DF */
544		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* E0 - EF */
545		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* F0 - FF */
546	};
547	ssize_t ll;
548	const char *pp = p;
549	const char * const pp_end = pp + len;
550
551	*last_is_path = 0;
552	/*
553	 * Skip the path-name which is quoted.
554	 */
555	for (;pp < pp_end; ++pp) {
556		if (!safe_char[*(const unsigned char *)pp]) {
557			if (*pp != ' ' && *pp != '\t' && *pp != '\r'
558			    && *pp != '\n')
559				f = 0;
560			break;
561		}
562		f = 1;
563	}
564	ll = pp_end - pp;
565
566	/* If a path-name was not found at the first, try to check
567	 * a mtree format(a.k.a form D) ``NetBSD's mtree -D'' creates,
568	 * which places the path-name at the last. */
569	if (f == 0) {
570		const char *pb = p + len - nl;
571		int name_len = 0;
572		int slash;
573
574		/* The form D accepts only a single line for an entry. */
575		if (pb-2 >= p &&
576		    pb[-1] == '\\' && (pb[-2] == ' ' || pb[-2] == '\t'))
577			return (-1);
578		if (pb-1 >= p && pb[-1] == '\\')
579			return (-1);
580
581		slash = 0;
582		while (p <= --pb && *pb != ' ' && *pb != '\t') {
583			if (!safe_char[*(const unsigned char *)pb])
584				return (-1);
585			name_len++;
586			/* The pathname should have a slash in this
587			 * format. */
588			if (*pb == '/')
589				slash = 1;
590		}
591		if (name_len == 0 || slash == 0)
592			return (-1);
593		/* If '/' is placed at the first in this field, this is not
594		 * a valid filename. */
595		if (pb[1] == '/')
596			return (-1);
597		ll = len - nl - name_len;
598		pp = p;
599		*last_is_path = 1;
600	}
601
602	return (bid_keyword_list(pp, ll, 0, *last_is_path));
603}
604
605#define MAX_BID_ENTRY	3
606
607static int
608mtree_bid(struct archive_read *a, int best_bid)
609{
610	const char *signature = "#mtree";
611	const char *p;
612
613	(void)best_bid; /* UNUSED */
614
615	/* Now let's look at the actual header and see if it matches. */
616	p = __archive_read_ahead(a, strlen(signature), NULL);
617	if (p == NULL)
618		return (-1);
619
620	if (memcmp(p, signature, strlen(signature)) == 0)
621		return (8 * (int)strlen(signature));
622
623	/*
624	 * There is not a mtree signature. Let's try to detect mtree format.
625	 */
626	return (detect_form(a, NULL));
627}
628
629static int
630detect_form(struct archive_read *a, int *is_form_d)
631{
632	const char *p;
633	ssize_t avail, ravail;
634	ssize_t detected_bytes = 0, len, nl;
635	int entry_cnt = 0, multiline = 0;
636	int form_D = 0;/* The archive is generated by `NetBSD mtree -D'
637			* (In this source we call it `form D') . */
638
639	if (is_form_d != NULL)
640		*is_form_d = 0;
641	p = __archive_read_ahead(a, 1, &avail);
642	if (p == NULL)
643		return (-1);
644	ravail = avail;
645	for (;;) {
646		len = next_line(a, &p, &avail, &ravail, &nl);
647		/* The terminal character of the line should be
648		 * a new line character, '\r\n' or '\n'. */
649		if (len <= 0 || nl == 0)
650			break;
651		if (!multiline) {
652			/* Leading whitespace is never significant,
653			 * ignore it. */
654			while (len > 0 && (*p == ' ' || *p == '\t')) {
655				++p;
656				--avail;
657				--len;
658			}
659			/* Skip comment or empty line. */
660			if (p[0] == '#' || p[0] == '\n' || p[0] == '\r') {
661				p += len;
662				avail -= len;
663				continue;
664			}
665		} else {
666			/* A continuance line; the terminal
667			 * character of previous line was '\' character. */
668			if (bid_keyword_list(p, len, 0, 0) <= 0)
669				break;
670			if (multiline == 1)
671				detected_bytes += len;
672			if (p[len-nl-1] != '\\') {
673				if (multiline == 1 &&
674				    ++entry_cnt >= MAX_BID_ENTRY)
675					break;
676				multiline = 0;
677			}
678			p += len;
679			avail -= len;
680			continue;
681		}
682		if (p[0] != '/') {
683			int last_is_path, keywords;
684
685			keywords = bid_entry(p, len, nl, &last_is_path);
686			if (keywords >= 0) {
687				detected_bytes += len;
688				if (form_D == 0) {
689					if (last_is_path)
690						form_D = 1;
691					else if (keywords > 0)
692						/* This line is not `form D'. */
693						form_D = -1;
694				} else if (form_D == 1) {
695					if (!last_is_path && keywords > 0)
696						/* This this is not `form D'
697						 * and We cannot accept mixed
698						 * format. */
699						break;
700				}
701				if (!last_is_path && p[len-nl-1] == '\\')
702					/* This line continues. */
703					multiline = 1;
704				else {
705					/* We've got plenty of correct lines
706					 * to assume that this file is a mtree
707					 * format. */
708					if (++entry_cnt >= MAX_BID_ENTRY)
709						break;
710				}
711			} else
712				break;
713		} else if (strncmp(p, "/set", 4) == 0) {
714			if (bid_keyword_list(p+4, len-4, 0, 0) <= 0)
715				break;
716			/* This line continues. */
717			if (p[len-nl-1] == '\\')
718				multiline = 2;
719		} else if (strncmp(p, "/unset", 6) == 0) {
720			if (bid_keyword_list(p+6, len-6, 1, 0) <= 0)
721				break;
722			/* This line continues. */
723			if (p[len-nl-1] == '\\')
724				multiline = 2;
725		} else
726			break;
727
728		/* Test next line. */
729		p += len;
730		avail -= len;
731	}
732	if (entry_cnt >= MAX_BID_ENTRY || (entry_cnt > 0 && len == 0)) {
733		if (is_form_d != NULL) {
734			if (form_D == 1)
735				*is_form_d = 1;
736		}
737		return (32);
738	}
739
740	return (0);
741}
742
743/*
744 * The extended mtree format permits multiple lines specifying
745 * attributes for each file.  For those entries, only the last line
746 * is actually used.  Practically speaking, that means we have
747 * to read the entire mtree file into memory up front.
748 *
749 * The parsing is done in two steps.  First, it is decided if a line
750 * changes the global defaults and if it is, processed accordingly.
751 * Otherwise, the options of the line are merged with the current
752 * global options.
753 */
754static int
755add_option(struct archive_read *a, struct mtree_option **global,
756    const char *value, size_t len)
757{
758	struct mtree_option *opt;
759
760	if ((opt = malloc(sizeof(*opt))) == NULL) {
761		archive_set_error(&a->archive, errno, "Can't allocate memory");
762		return (ARCHIVE_FATAL);
763	}
764	if ((opt->value = malloc(len + 1)) == NULL) {
765		free(opt);
766		archive_set_error(&a->archive, errno, "Can't allocate memory");
767		return (ARCHIVE_FATAL);
768	}
769	memcpy(opt->value, value, len);
770	opt->value[len] = '\0';
771	opt->next = *global;
772	*global = opt;
773	return (ARCHIVE_OK);
774}
775
776static void
777remove_option(struct mtree_option **global, const char *value, size_t len)
778{
779	struct mtree_option *iter, *last;
780
781	last = NULL;
782	for (iter = *global; iter != NULL; last = iter, iter = iter->next) {
783		if (strncmp(iter->value, value, len) == 0 &&
784		    (iter->value[len] == '\0' ||
785		     iter->value[len] == '='))
786			break;
787	}
788	if (iter == NULL)
789		return;
790	if (last == NULL)
791		*global = iter->next;
792	else
793		last->next = iter->next;
794
795	free(iter->value);
796	free(iter);
797}
798
799static int
800process_global_set(struct archive_read *a,
801    struct mtree_option **global, const char *line)
802{
803	const char *next, *eq;
804	size_t len;
805	int r;
806
807	line += 4;
808	for (;;) {
809		next = line + strspn(line, " \t\r\n");
810		if (*next == '\0')
811			return (ARCHIVE_OK);
812		line = next;
813		next = line + strcspn(line, " \t\r\n");
814		eq = strchr(line, '=');
815		if (eq > next)
816			len = next - line;
817		else
818			len = eq - line;
819
820		remove_option(global, line, len);
821		r = add_option(a, global, line, next - line);
822		if (r != ARCHIVE_OK)
823			return (r);
824		line = next;
825	}
826}
827
828static int
829process_global_unset(struct archive_read *a,
830    struct mtree_option **global, const char *line)
831{
832	const char *next;
833	size_t len;
834
835	line += 6;
836	if (strchr(line, '=') != NULL) {
837		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
838		    "/unset shall not contain `='");
839		return ARCHIVE_FATAL;
840	}
841
842	for (;;) {
843		next = line + strspn(line, " \t\r\n");
844		if (*next == '\0')
845			return (ARCHIVE_OK);
846		line = next;
847		len = strcspn(line, " \t\r\n");
848
849		if (len == 3 && strncmp(line, "all", 3) == 0) {
850			free_options(*global);
851			*global = NULL;
852		} else {
853			remove_option(global, line, len);
854		}
855
856		line += len;
857	}
858}
859
860static int
861process_add_entry(struct archive_read *a, struct mtree *mtree,
862    struct mtree_option **global, const char *line, ssize_t line_len,
863    struct mtree_entry **last_entry, int is_form_d)
864{
865	struct mtree_entry *entry;
866	struct mtree_option *iter;
867	const char *next, *eq, *name, *end;
868	size_t name_len, len;
869	int r, i;
870
871	if ((entry = malloc(sizeof(*entry))) == NULL) {
872		archive_set_error(&a->archive, errno, "Can't allocate memory");
873		return (ARCHIVE_FATAL);
874	}
875	entry->next = NULL;
876	entry->options = NULL;
877	entry->name = NULL;
878	entry->used = 0;
879	entry->full = 0;
880
881	/* Add this entry to list. */
882	if (*last_entry == NULL)
883		mtree->entries = entry;
884	else
885		(*last_entry)->next = entry;
886	*last_entry = entry;
887
888	if (is_form_d) {
889		/* Filename is last item on line. */
890		/* Adjust line_len to trim trailing whitespace */
891		while (line_len > 0) {
892			char last_character = line[line_len - 1];
893			if (last_character == '\r'
894			    || last_character == '\n'
895			    || last_character == '\t'
896			    || last_character == ' ') {
897				line_len--;
898			} else {
899				break;
900			}
901		}
902		/* Name starts after the last whitespace separator */
903		name = line;
904		for (i = 0; i < line_len; i++) {
905			if (line[i] == '\r'
906			    || line[i] == '\n'
907			    || line[i] == '\t'
908			    || line[i] == ' ') {
909				name = line + i + 1;
910			}
911		}
912		name_len = line + line_len - name;
913		end = name;
914	} else {
915		/* Filename is first item on line */
916		name_len = strcspn(line, " \t\r\n");
917		name = line;
918		line += name_len;
919		end = line + line_len;
920	}
921	/* name/name_len is the name within the line. */
922	/* line..end brackets the entire line except the name */
923
924	if ((entry->name = malloc(name_len + 1)) == NULL) {
925		archive_set_error(&a->archive, errno, "Can't allocate memory");
926		return (ARCHIVE_FATAL);
927	}
928
929	memcpy(entry->name, name, name_len);
930	entry->name[name_len] = '\0';
931	parse_escapes(entry->name, entry);
932
933	for (iter = *global; iter != NULL; iter = iter->next) {
934		r = add_option(a, &entry->options, iter->value,
935		    strlen(iter->value));
936		if (r != ARCHIVE_OK)
937			return (r);
938	}
939
940	for (;;) {
941		next = line + strspn(line, " \t\r\n");
942		if (*next == '\0')
943			return (ARCHIVE_OK);
944		if (next >= end)
945			return (ARCHIVE_OK);
946		line = next;
947		next = line + strcspn(line, " \t\r\n");
948		eq = strchr(line, '=');
949		if (eq == NULL || eq > next)
950			len = next - line;
951		else
952			len = eq - line;
953
954		remove_option(&entry->options, line, len);
955		r = add_option(a, &entry->options, line, next - line);
956		if (r != ARCHIVE_OK)
957			return (r);
958		line = next;
959	}
960}
961
962static int
963read_mtree(struct archive_read *a, struct mtree *mtree)
964{
965	ssize_t len;
966	uintmax_t counter;
967	char *p;
968	struct mtree_option *global;
969	struct mtree_entry *last_entry;
970	int r, is_form_d;
971
972	mtree->archive_format = ARCHIVE_FORMAT_MTREE;
973	mtree->archive_format_name = "mtree";
974
975	global = NULL;
976	last_entry = NULL;
977
978	(void)detect_form(a, &is_form_d);
979
980	for (counter = 1; ; ++counter) {
981		len = readline(a, mtree, &p, 65536);
982		if (len == 0) {
983			mtree->this_entry = mtree->entries;
984			free_options(global);
985			return (ARCHIVE_OK);
986		}
987		if (len < 0) {
988			free_options(global);
989			return ((int)len);
990		}
991		/* Leading whitespace is never significant, ignore it. */
992		while (*p == ' ' || *p == '\t') {
993			++p;
994			--len;
995		}
996		/* Skip content lines and blank lines. */
997		if (*p == '#')
998			continue;
999		if (*p == '\r' || *p == '\n' || *p == '\0')
1000			continue;
1001		if (*p != '/') {
1002			r = process_add_entry(a, mtree, &global, p, len,
1003			    &last_entry, is_form_d);
1004		} else if (strncmp(p, "/set", 4) == 0) {
1005			if (p[4] != ' ' && p[4] != '\t')
1006				break;
1007			r = process_global_set(a, &global, p);
1008		} else if (strncmp(p, "/unset", 6) == 0) {
1009			if (p[6] != ' ' && p[6] != '\t')
1010				break;
1011			r = process_global_unset(a, &global, p);
1012		} else
1013			break;
1014
1015		if (r != ARCHIVE_OK) {
1016			free_options(global);
1017			return r;
1018		}
1019	}
1020
1021	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1022	    "Can't parse line %ju", counter);
1023	free_options(global);
1024	return (ARCHIVE_FATAL);
1025}
1026
1027/*
1028 * Read in the entire mtree file into memory on the first request.
1029 * Then use the next unused file to satisfy each header request.
1030 */
1031static int
1032read_header(struct archive_read *a, struct archive_entry *entry)
1033{
1034	struct mtree *mtree;
1035	char *p;
1036	int r, use_next;
1037
1038	mtree = (struct mtree *)(a->format->data);
1039
1040	if (mtree->fd >= 0) {
1041		close(mtree->fd);
1042		mtree->fd = -1;
1043	}
1044
1045	if (mtree->entries == NULL) {
1046		mtree->resolver = archive_entry_linkresolver_new();
1047		if (mtree->resolver == NULL)
1048			return ARCHIVE_FATAL;
1049		archive_entry_linkresolver_set_strategy(mtree->resolver,
1050		    ARCHIVE_FORMAT_MTREE);
1051		r = read_mtree(a, mtree);
1052		if (r != ARCHIVE_OK)
1053			return (r);
1054	}
1055
1056	a->archive.archive_format = mtree->archive_format;
1057	a->archive.archive_format_name = mtree->archive_format_name;
1058
1059	for (;;) {
1060		if (mtree->this_entry == NULL)
1061			return (ARCHIVE_EOF);
1062		if (strcmp(mtree->this_entry->name, "..") == 0) {
1063			mtree->this_entry->used = 1;
1064			if (archive_strlen(&mtree->current_dir) > 0) {
1065				/* Roll back current path. */
1066				p = mtree->current_dir.s
1067				    + mtree->current_dir.length - 1;
1068				while (p >= mtree->current_dir.s && *p != '/')
1069					--p;
1070				if (p >= mtree->current_dir.s)
1071					--p;
1072				mtree->current_dir.length
1073				    = p - mtree->current_dir.s + 1;
1074			}
1075		}
1076		if (!mtree->this_entry->used) {
1077			use_next = 0;
1078			r = parse_file(a, entry, mtree, mtree->this_entry,
1079				&use_next);
1080			if (use_next == 0)
1081				return (r);
1082		}
1083		mtree->this_entry = mtree->this_entry->next;
1084	}
1085}
1086
1087/*
1088 * A single file can have multiple lines contribute specifications.
1089 * Parse as many lines as necessary, then pull additional information
1090 * from a backing file on disk as necessary.
1091 */
1092static int
1093parse_file(struct archive_read *a, struct archive_entry *entry,
1094    struct mtree *mtree, struct mtree_entry *mentry, int *use_next)
1095{
1096	const char *path;
1097	struct stat st_storage, *st;
1098	struct mtree_entry *mp;
1099	struct archive_entry *sparse_entry;
1100	int r = ARCHIVE_OK, r1, parsed_kws;
1101
1102	mentry->used = 1;
1103
1104	/* Initialize reasonable defaults. */
1105	archive_entry_set_filetype(entry, AE_IFREG);
1106	archive_entry_set_size(entry, 0);
1107	archive_string_empty(&mtree->contents_name);
1108
1109	/* Parse options from this line. */
1110	parsed_kws = 0;
1111	r = parse_line(a, entry, mtree, mentry, &parsed_kws);
1112
1113	if (mentry->full) {
1114		archive_entry_copy_pathname(entry, mentry->name);
1115		/*
1116		 * "Full" entries are allowed to have multiple lines
1117		 * and those lines aren't required to be adjacent.  We
1118		 * don't support multiple lines for "relative" entries
1119		 * nor do we make any attempt to merge data from
1120		 * separate "relative" and "full" entries.  (Merging
1121		 * "relative" and "full" entries would require dealing
1122		 * with pathname canonicalization, which is a very
1123		 * tricky subject.)
1124		 */
1125		for (mp = mentry->next; mp != NULL; mp = mp->next) {
1126			if (mp->full && !mp->used
1127			    && strcmp(mentry->name, mp->name) == 0) {
1128				/* Later lines override earlier ones. */
1129				mp->used = 1;
1130				r1 = parse_line(a, entry, mtree, mp,
1131				    &parsed_kws);
1132				if (r1 < r)
1133					r = r1;
1134			}
1135		}
1136	} else {
1137		/*
1138		 * Relative entries require us to construct
1139		 * the full path and possibly update the
1140		 * current directory.
1141		 */
1142		size_t n = archive_strlen(&mtree->current_dir);
1143		if (n > 0)
1144			archive_strcat(&mtree->current_dir, "/");
1145		archive_strcat(&mtree->current_dir, mentry->name);
1146		archive_entry_copy_pathname(entry, mtree->current_dir.s);
1147		if (archive_entry_filetype(entry) != AE_IFDIR)
1148			mtree->current_dir.length = n;
1149	}
1150
1151	if (mtree->checkfs) {
1152		/*
1153		 * Try to open and stat the file to get the real size
1154		 * and other file info.  It would be nice to avoid
1155		 * this here so that getting a listing of an mtree
1156		 * wouldn't require opening every referenced contents
1157		 * file.  But then we wouldn't know the actual
1158		 * contents size, so I don't see a really viable way
1159		 * around this.  (Also, we may want to someday pull
1160		 * other unspecified info from the contents file on
1161		 * disk.)
1162		 */
1163		mtree->fd = -1;
1164		if (archive_strlen(&mtree->contents_name) > 0)
1165			path = mtree->contents_name.s;
1166		else
1167			path = archive_entry_pathname(entry);
1168
1169		if (archive_entry_filetype(entry) == AE_IFREG ||
1170				archive_entry_filetype(entry) == AE_IFDIR) {
1171			mtree->fd = open(path, O_RDONLY | O_BINARY | O_CLOEXEC);
1172			__archive_ensure_cloexec_flag(mtree->fd);
1173			if (mtree->fd == -1 &&
1174				(errno != ENOENT ||
1175				 archive_strlen(&mtree->contents_name) > 0)) {
1176				archive_set_error(&a->archive, errno,
1177						"Can't open %s", path);
1178				r = ARCHIVE_WARN;
1179			}
1180		}
1181
1182		st = &st_storage;
1183		if (mtree->fd >= 0) {
1184			if (fstat(mtree->fd, st) == -1) {
1185				archive_set_error(&a->archive, errno,
1186						"Could not fstat %s", path);
1187				r = ARCHIVE_WARN;
1188				/* If we can't stat it, don't keep it open. */
1189				close(mtree->fd);
1190				mtree->fd = -1;
1191				st = NULL;
1192			}
1193		} else if (lstat(path, st) == -1) {
1194			st = NULL;
1195		}
1196
1197		/*
1198		 * Check for a mismatch between the type in the specification
1199		 * and the type of the contents object on disk.
1200		 */
1201		if (st != NULL) {
1202			if (((st->st_mode & S_IFMT) == S_IFREG &&
1203			      archive_entry_filetype(entry) == AE_IFREG)
1204#ifdef S_IFLNK
1205			  ||((st->st_mode & S_IFMT) == S_IFLNK &&
1206			      archive_entry_filetype(entry) == AE_IFLNK)
1207#endif
1208#ifdef S_IFSOCK
1209			  ||((st->st_mode & S_IFSOCK) == S_IFSOCK &&
1210			      archive_entry_filetype(entry) == AE_IFSOCK)
1211#endif
1212#ifdef S_IFCHR
1213			  ||((st->st_mode & S_IFMT) == S_IFCHR &&
1214			      archive_entry_filetype(entry) == AE_IFCHR)
1215#endif
1216#ifdef S_IFBLK
1217			  ||((st->st_mode & S_IFMT) == S_IFBLK &&
1218			      archive_entry_filetype(entry) == AE_IFBLK)
1219#endif
1220			  ||((st->st_mode & S_IFMT) == S_IFDIR &&
1221			      archive_entry_filetype(entry) == AE_IFDIR)
1222#ifdef S_IFIFO
1223			  ||((st->st_mode & S_IFMT) == S_IFIFO &&
1224			      archive_entry_filetype(entry) == AE_IFIFO)
1225#endif
1226			) {
1227				/* Types match. */
1228			} else {
1229				/* Types don't match; bail out gracefully. */
1230				if (mtree->fd >= 0)
1231					close(mtree->fd);
1232				mtree->fd = -1;
1233				if (parsed_kws & MTREE_HAS_OPTIONAL) {
1234					/* It's not an error for an optional
1235					 * entry to not match disk. */
1236					*use_next = 1;
1237				} else if (r == ARCHIVE_OK) {
1238					archive_set_error(&a->archive,
1239					    ARCHIVE_ERRNO_MISC,
1240					    "mtree specification has different"
1241					    " type for %s",
1242					    archive_entry_pathname(entry));
1243					r = ARCHIVE_WARN;
1244				}
1245				return (r);
1246			}
1247		}
1248
1249		/*
1250		 * If there is a contents file on disk, pick some of the
1251		 * metadata from that file.  For most of these, we only
1252		 * set it from the contents if it wasn't already parsed
1253		 * from the specification.
1254		 */
1255		if (st != NULL) {
1256			if (((parsed_kws & MTREE_HAS_DEVICE) == 0 ||
1257				(parsed_kws & MTREE_HAS_NOCHANGE) != 0) &&
1258				(archive_entry_filetype(entry) == AE_IFCHR ||
1259				 archive_entry_filetype(entry) == AE_IFBLK))
1260				archive_entry_set_rdev(entry, st->st_rdev);
1261			if ((parsed_kws & (MTREE_HAS_GID | MTREE_HAS_GNAME))
1262				== 0 ||
1263			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1264				archive_entry_set_gid(entry, st->st_gid);
1265			if ((parsed_kws & (MTREE_HAS_UID | MTREE_HAS_UNAME))
1266				== 0 ||
1267			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1268				archive_entry_set_uid(entry, st->st_uid);
1269			if ((parsed_kws & MTREE_HAS_MTIME) == 0 ||
1270			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0) {
1271#if HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC
1272				archive_entry_set_mtime(entry, st->st_mtime,
1273						st->st_mtimespec.tv_nsec);
1274#elif HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC
1275				archive_entry_set_mtime(entry, st->st_mtime,
1276						st->st_mtim.tv_nsec);
1277#elif HAVE_STRUCT_STAT_ST_MTIME_N
1278				archive_entry_set_mtime(entry, st->st_mtime,
1279						st->st_mtime_n);
1280#elif HAVE_STRUCT_STAT_ST_UMTIME
1281				archive_entry_set_mtime(entry, st->st_mtime,
1282						st->st_umtime*1000);
1283#elif HAVE_STRUCT_STAT_ST_MTIME_USEC
1284				archive_entry_set_mtime(entry, st->st_mtime,
1285						st->st_mtime_usec*1000);
1286#else
1287				archive_entry_set_mtime(entry, st->st_mtime, 0);
1288#endif
1289			}
1290			if ((parsed_kws & MTREE_HAS_NLINK) == 0 ||
1291			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1292				archive_entry_set_nlink(entry, st->st_nlink);
1293			if ((parsed_kws & MTREE_HAS_PERM) == 0 ||
1294			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1295				archive_entry_set_perm(entry, st->st_mode);
1296			if ((parsed_kws & MTREE_HAS_SIZE) == 0 ||
1297			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1298				archive_entry_set_size(entry, st->st_size);
1299			archive_entry_set_ino(entry, st->st_ino);
1300			archive_entry_set_dev(entry, st->st_dev);
1301
1302			archive_entry_linkify(mtree->resolver, &entry,
1303				&sparse_entry);
1304		} else if (parsed_kws & MTREE_HAS_OPTIONAL) {
1305			/*
1306			 * Couldn't open the entry, stat it or the on-disk type
1307			 * didn't match.  If this entry is optional, just
1308			 * ignore it and read the next header entry.
1309			 */
1310			*use_next = 1;
1311			return ARCHIVE_OK;
1312		}
1313	}
1314
1315	mtree->cur_size = archive_entry_size(entry);
1316	mtree->offset = 0;
1317
1318	return r;
1319}
1320
1321/*
1322 * Each line contains a sequence of keywords.
1323 */
1324static int
1325parse_line(struct archive_read *a, struct archive_entry *entry,
1326    struct mtree *mtree, struct mtree_entry *mp, int *parsed_kws)
1327{
1328	struct mtree_option *iter;
1329	int r = ARCHIVE_OK, r1;
1330
1331	for (iter = mp->options; iter != NULL; iter = iter->next) {
1332		r1 = parse_keyword(a, mtree, entry, iter, parsed_kws);
1333		if (r1 < r)
1334			r = r1;
1335	}
1336	if (r == ARCHIVE_OK && (*parsed_kws & MTREE_HAS_TYPE) == 0) {
1337		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1338		    "Missing type keyword in mtree specification");
1339		return (ARCHIVE_WARN);
1340	}
1341	return (r);
1342}
1343
1344/*
1345 * Device entries have one of the following forms:
1346 *  - raw dev_t
1347 *  - format,major,minor[,subdevice]
1348 * When parsing succeeded, `pdev' will contain the appropriate dev_t value.
1349 */
1350
1351/* strsep() is not in C90, but strcspn() is. */
1352/* Taken from http://unixpapa.com/incnote/string.html */
1353static char *
1354la_strsep(char **sp, const char *sep)
1355{
1356	char *p, *s;
1357	if (sp == NULL || *sp == NULL || **sp == '\0')
1358		return(NULL);
1359	s = *sp;
1360	p = s + strcspn(s, sep);
1361	if (*p != '\0')
1362		*p++ = '\0';
1363	*sp = p;
1364	return(s);
1365}
1366
1367static int
1368parse_device(dev_t *pdev, struct archive *a, char *val)
1369{
1370#define MAX_PACK_ARGS 3
1371	unsigned long numbers[MAX_PACK_ARGS];
1372	char *p, *dev;
1373	int argc;
1374	pack_t *pack;
1375	dev_t result;
1376	const char *error = NULL;
1377
1378	memset(pdev, 0, sizeof(*pdev));
1379	if ((dev = strchr(val, ',')) != NULL) {
1380		/*
1381		 * Device's major/minor are given in a specified format.
1382		 * Decode and pack it accordingly.
1383		 */
1384		*dev++ = '\0';
1385		if ((pack = pack_find(val)) == NULL) {
1386			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1387			    "Unknown format `%s'", val);
1388			return ARCHIVE_WARN;
1389		}
1390		argc = 0;
1391		while ((p = la_strsep(&dev, ",")) != NULL) {
1392			if (*p == '\0') {
1393				archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1394				    "Missing number");
1395				return ARCHIVE_WARN;
1396			}
1397			if (argc >= MAX_PACK_ARGS) {
1398				archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1399				    "Too many arguments");
1400				return ARCHIVE_WARN;
1401			}
1402			numbers[argc++] = (unsigned long)mtree_atol(&p);
1403		}
1404		if (argc < 2) {
1405			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1406			    "Not enough arguments");
1407			return ARCHIVE_WARN;
1408		}
1409		result = (*pack)(argc, numbers, &error);
1410		if (error != NULL) {
1411			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1412			    "%s", error);
1413			return ARCHIVE_WARN;
1414		}
1415	} else {
1416		/* file system raw value. */
1417		result = (dev_t)mtree_atol(&val);
1418	}
1419	*pdev = result;
1420	return ARCHIVE_OK;
1421#undef MAX_PACK_ARGS
1422}
1423
1424/*
1425 * Parse a single keyword and its value.
1426 */
1427static int
1428parse_keyword(struct archive_read *a, struct mtree *mtree,
1429    struct archive_entry *entry, struct mtree_option *opt, int *parsed_kws)
1430{
1431	char *val, *key;
1432
1433	key = opt->value;
1434
1435	if (*key == '\0')
1436		return (ARCHIVE_OK);
1437
1438	if (strcmp(key, "nochange") == 0) {
1439		*parsed_kws |= MTREE_HAS_NOCHANGE;
1440		return (ARCHIVE_OK);
1441	}
1442	if (strcmp(key, "optional") == 0) {
1443		*parsed_kws |= MTREE_HAS_OPTIONAL;
1444		return (ARCHIVE_OK);
1445	}
1446	if (strcmp(key, "ignore") == 0) {
1447		/*
1448		 * The mtree processing is not recursive, so
1449		 * recursion will only happen for explicitly listed
1450		 * entries.
1451		 */
1452		return (ARCHIVE_OK);
1453	}
1454
1455	val = strchr(key, '=');
1456	if (val == NULL) {
1457		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1458		    "Malformed attribute \"%s\" (%d)", key, key[0]);
1459		return (ARCHIVE_WARN);
1460	}
1461
1462	*val = '\0';
1463	++val;
1464
1465	switch (key[0]) {
1466	case 'c':
1467		if (strcmp(key, "content") == 0
1468		    || strcmp(key, "contents") == 0) {
1469			parse_escapes(val, NULL);
1470			archive_strcpy(&mtree->contents_name, val);
1471			break;
1472		}
1473		if (strcmp(key, "cksum") == 0)
1474			break;
1475	case 'd':
1476		if (strcmp(key, "device") == 0) {
1477			/* stat(2) st_rdev field, e.g. the major/minor IDs
1478			 * of a char/block special file */
1479			int r;
1480			dev_t dev;
1481
1482			*parsed_kws |= MTREE_HAS_DEVICE;
1483			r = parse_device(&dev, &a->archive, val);
1484			if (r == ARCHIVE_OK)
1485				archive_entry_set_rdev(entry, dev);
1486			return r;
1487		}
1488	case 'f':
1489		if (strcmp(key, "flags") == 0) {
1490			*parsed_kws |= MTREE_HAS_FFLAGS;
1491			archive_entry_copy_fflags_text(entry, val);
1492			break;
1493		}
1494	case 'g':
1495		if (strcmp(key, "gid") == 0) {
1496			*parsed_kws |= MTREE_HAS_GID;
1497			archive_entry_set_gid(entry, mtree_atol10(&val));
1498			break;
1499		}
1500		if (strcmp(key, "gname") == 0) {
1501			*parsed_kws |= MTREE_HAS_GNAME;
1502			archive_entry_copy_gname(entry, val);
1503			break;
1504		}
1505	case 'i':
1506		if (strcmp(key, "inode") == 0) {
1507			archive_entry_set_ino(entry, mtree_atol10(&val));
1508			break;
1509		}
1510	case 'l':
1511		if (strcmp(key, "link") == 0) {
1512			archive_entry_copy_symlink(entry, val);
1513			break;
1514		}
1515	case 'm':
1516		if (strcmp(key, "md5") == 0 || strcmp(key, "md5digest") == 0)
1517			break;
1518		if (strcmp(key, "mode") == 0) {
1519			if (val[0] >= '0' && val[0] <= '9') {
1520				*parsed_kws |= MTREE_HAS_PERM;
1521				archive_entry_set_perm(entry,
1522				    (mode_t)mtree_atol8(&val));
1523			} else {
1524				archive_set_error(&a->archive,
1525				    ARCHIVE_ERRNO_FILE_FORMAT,
1526				    "Symbolic mode \"%s\" unsupported", val);
1527				return ARCHIVE_WARN;
1528			}
1529			break;
1530		}
1531	case 'n':
1532		if (strcmp(key, "nlink") == 0) {
1533			*parsed_kws |= MTREE_HAS_NLINK;
1534			archive_entry_set_nlink(entry,
1535				(unsigned int)mtree_atol10(&val));
1536			break;
1537		}
1538	case 'r':
1539		if (strcmp(key, "resdevice") == 0) {
1540			/* stat(2) st_dev field, e.g. the device ID where the
1541			 * inode resides */
1542			int r;
1543			dev_t dev;
1544
1545			r = parse_device(&dev, &a->archive, val);
1546			if (r == ARCHIVE_OK)
1547				archive_entry_set_dev(entry, dev);
1548			return r;
1549		}
1550		if (strcmp(key, "rmd160") == 0 ||
1551		    strcmp(key, "rmd160digest") == 0)
1552			break;
1553	case 's':
1554		if (strcmp(key, "sha1") == 0 || strcmp(key, "sha1digest") == 0)
1555			break;
1556		if (strcmp(key, "sha256") == 0 ||
1557		    strcmp(key, "sha256digest") == 0)
1558			break;
1559		if (strcmp(key, "sha384") == 0 ||
1560		    strcmp(key, "sha384digest") == 0)
1561			break;
1562		if (strcmp(key, "sha512") == 0 ||
1563		    strcmp(key, "sha512digest") == 0)
1564			break;
1565		if (strcmp(key, "size") == 0) {
1566			archive_entry_set_size(entry, mtree_atol10(&val));
1567			break;
1568		}
1569	case 't':
1570		if (strcmp(key, "tags") == 0) {
1571			/*
1572			 * Comma delimited list of tags.
1573			 * Ignore the tags for now, but the interface
1574			 * should be extended to allow inclusion/exclusion.
1575			 */
1576			break;
1577		}
1578		if (strcmp(key, "time") == 0) {
1579			int64_t m;
1580			int64_t my_time_t_max = get_time_t_max();
1581			int64_t my_time_t_min = get_time_t_min();
1582			long ns = 0;
1583
1584			*parsed_kws |= MTREE_HAS_MTIME;
1585			m = mtree_atol10(&val);
1586			/* Replicate an old mtree bug:
1587			 * 123456789.1 represents 123456789
1588			 * seconds and 1 nanosecond. */
1589			if (*val == '.') {
1590				++val;
1591				ns = (long)mtree_atol10(&val);
1592			} else
1593				ns = 0;
1594			if (m > my_time_t_max)
1595				m = my_time_t_max;
1596			else if (m < my_time_t_min)
1597				m = my_time_t_min;
1598			archive_entry_set_mtime(entry, (time_t)m, ns);
1599			break;
1600		}
1601		if (strcmp(key, "type") == 0) {
1602			switch (val[0]) {
1603			case 'b':
1604				if (strcmp(val, "block") == 0) {
1605					archive_entry_set_filetype(entry, AE_IFBLK);
1606					break;
1607				}
1608			case 'c':
1609				if (strcmp(val, "char") == 0) {
1610					archive_entry_set_filetype(entry,
1611						AE_IFCHR);
1612					break;
1613				}
1614			case 'd':
1615				if (strcmp(val, "dir") == 0) {
1616					archive_entry_set_filetype(entry,
1617						AE_IFDIR);
1618					break;
1619				}
1620			case 'f':
1621				if (strcmp(val, "fifo") == 0) {
1622					archive_entry_set_filetype(entry,
1623						AE_IFIFO);
1624					break;
1625				}
1626				if (strcmp(val, "file") == 0) {
1627					archive_entry_set_filetype(entry,
1628						AE_IFREG);
1629					break;
1630				}
1631			case 'l':
1632				if (strcmp(val, "link") == 0) {
1633					archive_entry_set_filetype(entry,
1634						AE_IFLNK);
1635					break;
1636				}
1637			default:
1638				archive_set_error(&a->archive,
1639				    ARCHIVE_ERRNO_FILE_FORMAT,
1640				    "Unrecognized file type \"%s\"; "
1641				    "assuming \"file\"", val);
1642				archive_entry_set_filetype(entry, AE_IFREG);
1643				return (ARCHIVE_WARN);
1644			}
1645			*parsed_kws |= MTREE_HAS_TYPE;
1646			break;
1647		}
1648	case 'u':
1649		if (strcmp(key, "uid") == 0) {
1650			*parsed_kws |= MTREE_HAS_UID;
1651			archive_entry_set_uid(entry, mtree_atol10(&val));
1652			break;
1653		}
1654		if (strcmp(key, "uname") == 0) {
1655			*parsed_kws |= MTREE_HAS_UNAME;
1656			archive_entry_copy_uname(entry, val);
1657			break;
1658		}
1659	default:
1660		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1661		    "Unrecognized key %s=%s", key, val);
1662		return (ARCHIVE_WARN);
1663	}
1664	return (ARCHIVE_OK);
1665}
1666
1667static int
1668read_data(struct archive_read *a, const void **buff, size_t *size,
1669    int64_t *offset)
1670{
1671	size_t bytes_to_read;
1672	ssize_t bytes_read;
1673	struct mtree *mtree;
1674
1675	mtree = (struct mtree *)(a->format->data);
1676	if (mtree->fd < 0) {
1677		*buff = NULL;
1678		*offset = 0;
1679		*size = 0;
1680		return (ARCHIVE_EOF);
1681	}
1682	if (mtree->buff == NULL) {
1683		mtree->buffsize = 64 * 1024;
1684		mtree->buff = malloc(mtree->buffsize);
1685		if (mtree->buff == NULL) {
1686			archive_set_error(&a->archive, ENOMEM,
1687			    "Can't allocate memory");
1688			return (ARCHIVE_FATAL);
1689		}
1690	}
1691
1692	*buff = mtree->buff;
1693	*offset = mtree->offset;
1694	if ((int64_t)mtree->buffsize > mtree->cur_size - mtree->offset)
1695		bytes_to_read = (size_t)(mtree->cur_size - mtree->offset);
1696	else
1697		bytes_to_read = mtree->buffsize;
1698	bytes_read = read(mtree->fd, mtree->buff, bytes_to_read);
1699	if (bytes_read < 0) {
1700		archive_set_error(&a->archive, errno, "Can't read");
1701		return (ARCHIVE_WARN);
1702	}
1703	if (bytes_read == 0) {
1704		*size = 0;
1705		return (ARCHIVE_EOF);
1706	}
1707	mtree->offset += bytes_read;
1708	*size = bytes_read;
1709	return (ARCHIVE_OK);
1710}
1711
1712/* Skip does nothing except possibly close the contents file. */
1713static int
1714skip(struct archive_read *a)
1715{
1716	struct mtree *mtree;
1717
1718	mtree = (struct mtree *)(a->format->data);
1719	if (mtree->fd >= 0) {
1720		close(mtree->fd);
1721		mtree->fd = -1;
1722	}
1723	return (ARCHIVE_OK);
1724}
1725
1726/*
1727 * Since parsing backslash sequences always makes strings shorter,
1728 * we can always do this conversion in-place.
1729 */
1730static void
1731parse_escapes(char *src, struct mtree_entry *mentry)
1732{
1733	char *dest = src;
1734	char c;
1735
1736	if (mentry != NULL && strcmp(src, ".") == 0)
1737		mentry->full = 1;
1738
1739	while (*src != '\0') {
1740		c = *src++;
1741		if (c == '/' && mentry != NULL)
1742			mentry->full = 1;
1743		if (c == '\\') {
1744			switch (src[0]) {
1745			case '0':
1746				if (src[1] < '0' || src[1] > '7') {
1747					c = 0;
1748					++src;
1749					break;
1750				}
1751				/* FALLTHROUGH */
1752			case '1':
1753			case '2':
1754			case '3':
1755				if (src[1] >= '0' && src[1] <= '7' &&
1756				    src[2] >= '0' && src[2] <= '7') {
1757					c = (src[0] - '0') << 6;
1758					c |= (src[1] - '0') << 3;
1759					c |= (src[2] - '0');
1760					src += 3;
1761				}
1762				break;
1763			case 'a':
1764				c = '\a';
1765				++src;
1766				break;
1767			case 'b':
1768				c = '\b';
1769				++src;
1770				break;
1771			case 'f':
1772				c = '\f';
1773				++src;
1774				break;
1775			case 'n':
1776				c = '\n';
1777				++src;
1778				break;
1779			case 'r':
1780				c = '\r';
1781				++src;
1782				break;
1783			case 's':
1784				c = ' ';
1785				++src;
1786				break;
1787			case 't':
1788				c = '\t';
1789				++src;
1790				break;
1791			case 'v':
1792				c = '\v';
1793				++src;
1794				break;
1795			case '\\':
1796				c = '\\';
1797				++src;
1798				break;
1799			}
1800		}
1801		*dest++ = c;
1802	}
1803	*dest = '\0';
1804}
1805
1806/*
1807 * Note that this implementation does not (and should not!) obey
1808 * locale settings; you cannot simply substitute strtol here, since
1809 * it does obey locale.
1810 */
1811static int64_t
1812mtree_atol8(char **p)
1813{
1814	int64_t	l, limit, last_digit_limit;
1815	int digit, base;
1816
1817	base = 8;
1818	limit = INT64_MAX / base;
1819	last_digit_limit = INT64_MAX % base;
1820
1821	l = 0;
1822	digit = **p - '0';
1823	while (digit >= 0 && digit < base) {
1824		if (l>limit || (l == limit && digit > last_digit_limit)) {
1825			l = INT64_MAX; /* Truncate on overflow. */
1826			break;
1827		}
1828		l = (l * base) + digit;
1829		digit = *++(*p) - '0';
1830	}
1831	return (l);
1832}
1833
1834/*
1835 * Note that this implementation does not (and should not!) obey
1836 * locale settings; you cannot simply substitute strtol here, since
1837 * it does obey locale.
1838 */
1839static int64_t
1840mtree_atol10(char **p)
1841{
1842	int64_t l, limit, last_digit_limit;
1843	int base, digit, sign;
1844
1845	base = 10;
1846
1847	if (**p == '-') {
1848		sign = -1;
1849		limit = ((uint64_t)(INT64_MAX) + 1) / base;
1850		last_digit_limit = ((uint64_t)(INT64_MAX) + 1) % base;
1851		++(*p);
1852	} else {
1853		sign = 1;
1854		limit = INT64_MAX / base;
1855		last_digit_limit = INT64_MAX % base;
1856	}
1857
1858	l = 0;
1859	digit = **p - '0';
1860	while (digit >= 0 && digit < base) {
1861		if (l > limit || (l == limit && digit > last_digit_limit))
1862			return (sign < 0) ? INT64_MIN : INT64_MAX;
1863		l = (l * base) + digit;
1864		digit = *++(*p) - '0';
1865	}
1866	return (sign < 0) ? -l : l;
1867}
1868
1869/* Parse a hex digit. */
1870static int
1871parsehex(char c)
1872{
1873	if (c >= '0' && c <= '9')
1874		return c - '0';
1875	else if (c >= 'a' && c <= 'f')
1876		return c - 'a';
1877	else if (c >= 'A' && c <= 'F')
1878		return c - 'A';
1879	else
1880		return -1;
1881}
1882
1883/*
1884 * Note that this implementation does not (and should not!) obey
1885 * locale settings; you cannot simply substitute strtol here, since
1886 * it does obey locale.
1887 */
1888static int64_t
1889mtree_atol16(char **p)
1890{
1891	int64_t l, limit, last_digit_limit;
1892	int base, digit, sign;
1893
1894	base = 16;
1895
1896	if (**p == '-') {
1897		sign = -1;
1898		limit = ((uint64_t)(INT64_MAX) + 1) / base;
1899		last_digit_limit = ((uint64_t)(INT64_MAX) + 1) % base;
1900		++(*p);
1901	} else {
1902		sign = 1;
1903		limit = INT64_MAX / base;
1904		last_digit_limit = INT64_MAX % base;
1905	}
1906
1907	l = 0;
1908	digit = parsehex(**p);
1909	while (digit >= 0 && digit < base) {
1910		if (l > limit || (l == limit && digit > last_digit_limit))
1911			return (sign < 0) ? INT64_MIN : INT64_MAX;
1912		l = (l * base) + digit;
1913		digit = parsehex(*++(*p));
1914	}
1915	return (sign < 0) ? -l : l;
1916}
1917
1918static int64_t
1919mtree_atol(char **p)
1920{
1921	if (**p != '0')
1922		return mtree_atol10(p);
1923	if ((*p)[1] == 'x' || (*p)[1] == 'X') {
1924		*p += 2;
1925		return mtree_atol16(p);
1926	}
1927	return mtree_atol8(p);
1928}
1929
1930/*
1931 * Returns length of line (including trailing newline)
1932 * or negative on error.  'start' argument is updated to
1933 * point to first character of line.
1934 */
1935static ssize_t
1936readline(struct archive_read *a, struct mtree *mtree, char **start,
1937    ssize_t limit)
1938{
1939	ssize_t bytes_read;
1940	ssize_t total_size = 0;
1941	ssize_t find_off = 0;
1942	const void *t;
1943	void *nl;
1944	char *u;
1945
1946	/* Accumulate line in a line buffer. */
1947	for (;;) {
1948		/* Read some more. */
1949		t = __archive_read_ahead(a, 1, &bytes_read);
1950		if (t == NULL)
1951			return (0);
1952		if (bytes_read < 0)
1953			return (ARCHIVE_FATAL);
1954		nl = memchr(t, '\n', bytes_read);
1955		/* If we found '\n', trim the read to end exactly there. */
1956		if (nl != NULL) {
1957			bytes_read = ((const char *)nl) - ((const char *)t) + 1;
1958		}
1959		if (total_size + bytes_read + 1 > limit) {
1960			archive_set_error(&a->archive,
1961			    ARCHIVE_ERRNO_FILE_FORMAT,
1962			    "Line too long");
1963			return (ARCHIVE_FATAL);
1964		}
1965		if (archive_string_ensure(&mtree->line,
1966			total_size + bytes_read + 1) == NULL) {
1967			archive_set_error(&a->archive, ENOMEM,
1968			    "Can't allocate working buffer");
1969			return (ARCHIVE_FATAL);
1970		}
1971		/* Append new bytes to string. */
1972		memcpy(mtree->line.s + total_size, t, bytes_read);
1973		__archive_read_consume(a, bytes_read);
1974		total_size += bytes_read;
1975		mtree->line.s[total_size] = '\0';
1976
1977		for (u = mtree->line.s + find_off; *u; ++u) {
1978			if (u[0] == '\n') {
1979				/* Ends with unescaped newline. */
1980				*start = mtree->line.s;
1981				return total_size;
1982			} else if (u[0] == '#') {
1983				/* Ends with comment sequence #...\n */
1984				if (nl == NULL) {
1985					/* But we've not found the \n yet */
1986					break;
1987				}
1988			} else if (u[0] == '\\') {
1989				if (u[1] == '\n') {
1990					/* Trim escaped newline. */
1991					total_size -= 2;
1992					mtree->line.s[total_size] = '\0';
1993					break;
1994				} else if (u[1] != '\0') {
1995					/* Skip the two-char escape sequence */
1996					++u;
1997				}
1998			}
1999		}
2000		find_off = u - mtree->line.s;
2001	}
2002}
2003