archive_read_support_format_mtree.c revision 328828
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 328828 2018-02-03 02:17:25Z 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
78#define	MTREE_HASHTABLE_SIZE 1024
79
80#define	MAX_LINE_LEN		(1024 * 1024)
81
82struct mtree_option {
83	struct mtree_option *next;
84	char *value;
85};
86
87struct mtree_entry {
88	struct mtree_entry *next;
89	struct mtree_option *options;
90	char *name;
91	char full;
92	char used;
93	unsigned int name_hash;
94	struct mtree_entry *hashtable_next;
95};
96
97struct mtree {
98	struct archive_string	 line;
99	size_t			 buffsize;
100	char			*buff;
101	int64_t			 offset;
102	int			 fd;
103	int			 archive_format;
104	const char		*archive_format_name;
105	struct mtree_entry	*entries;
106	struct mtree_entry	*this_entry;
107	struct mtree_entry	*entry_hashtable[MTREE_HASHTABLE_SIZE];
108	struct archive_string	 current_dir;
109	struct archive_string	 contents_name;
110
111	struct archive_entry_linkresolver *resolver;
112
113	int64_t			 cur_size;
114	char checkfs;
115};
116
117static int	bid_keycmp(const char *, const char *, ssize_t);
118static int	cleanup(struct archive_read *);
119static int	detect_form(struct archive_read *, int *);
120static unsigned int	hash(const char *);
121static int	mtree_bid(struct archive_read *, int);
122static int	parse_file(struct archive_read *, struct archive_entry *,
123		    struct mtree *, struct mtree_entry *, int *);
124static void	parse_escapes(char *, struct mtree_entry *);
125static int	parse_line(struct archive_read *, struct archive_entry *,
126		    struct mtree *, struct mtree_entry *, int *);
127static int	parse_keyword(struct archive_read *, struct mtree *,
128		    struct archive_entry *, struct mtree_option *, int *);
129static int	read_data(struct archive_read *a,
130		    const void **buff, size_t *size, int64_t *offset);
131static ssize_t	readline(struct archive_read *, struct mtree *, char **, ssize_t);
132static int	skip(struct archive_read *a);
133static int	read_header(struct archive_read *,
134		    struct archive_entry *);
135static int64_t	mtree_atol(char **, int base);
136
137/*
138 * There's no standard for TIME_T_MAX/TIME_T_MIN.  So we compute them
139 * here.  TODO: Move this to configure time, but be careful
140 * about cross-compile environments.
141 */
142static int64_t
143get_time_t_max(void)
144{
145#if defined(TIME_T_MAX)
146	return TIME_T_MAX;
147#else
148	/* ISO C allows time_t to be a floating-point type,
149	   but POSIX requires an integer type.  The following
150	   should work on any system that follows the POSIX
151	   conventions. */
152	if (((time_t)0) < ((time_t)-1)) {
153		/* Time_t is unsigned */
154		return (~(time_t)0);
155	} else {
156		/* Time_t is signed. */
157		/* Assume it's the same as int64_t or int32_t */
158		if (sizeof(time_t) == sizeof(int64_t)) {
159			return (time_t)INT64_MAX;
160		} else {
161			return (time_t)INT32_MAX;
162		}
163	}
164#endif
165}
166
167static int64_t
168get_time_t_min(void)
169{
170#if defined(TIME_T_MIN)
171	return TIME_T_MIN;
172#else
173	if (((time_t)0) < ((time_t)-1)) {
174		/* Time_t is unsigned */
175		return (time_t)0;
176	} else {
177		/* Time_t is signed. */
178		if (sizeof(time_t) == sizeof(int64_t)) {
179			return (time_t)INT64_MIN;
180		} else {
181			return (time_t)INT32_MIN;
182		}
183	}
184#endif
185}
186
187static int
188archive_read_format_mtree_options(struct archive_read *a,
189    const char *key, const char *val)
190{
191	struct mtree *mtree;
192
193	mtree = (struct mtree *)(a->format->data);
194	if (strcmp(key, "checkfs")  == 0) {
195		/* Allows to read information missing from the mtree from the file system */
196		if (val == NULL || val[0] == 0) {
197			mtree->checkfs = 0;
198		} else {
199			mtree->checkfs = 1;
200		}
201		return (ARCHIVE_OK);
202	}
203
204	/* Note: The "warn" return is just to inform the options
205	 * supervisor that we didn't handle it.  It will generate
206	 * a suitable error if no one used this option. */
207	return (ARCHIVE_WARN);
208}
209
210static void
211free_options(struct mtree_option *head)
212{
213	struct mtree_option *next;
214
215	for (; head != NULL; head = next) {
216		next = head->next;
217		free(head->value);
218		free(head);
219	}
220}
221
222int
223archive_read_support_format_mtree(struct archive *_a)
224{
225	struct archive_read *a = (struct archive_read *)_a;
226	struct mtree *mtree;
227	int r;
228
229	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
230	    ARCHIVE_STATE_NEW, "archive_read_support_format_mtree");
231
232	mtree = (struct mtree *)calloc(1, sizeof(*mtree));
233	if (mtree == NULL) {
234		archive_set_error(&a->archive, ENOMEM,
235		    "Can't allocate mtree data");
236		return (ARCHIVE_FATAL);
237	}
238	mtree->fd = -1;
239
240	r = __archive_read_register_format(a, mtree, "mtree",
241           mtree_bid, archive_read_format_mtree_options, read_header, read_data, skip, NULL, cleanup, NULL, NULL);
242
243	if (r != ARCHIVE_OK)
244		free(mtree);
245	return (ARCHIVE_OK);
246}
247
248static int
249cleanup(struct archive_read *a)
250{
251	struct mtree *mtree;
252	struct mtree_entry *p, *q;
253
254	mtree = (struct mtree *)(a->format->data);
255
256	p = mtree->entries;
257	while (p != NULL) {
258		q = p->next;
259		free(p->name);
260		free_options(p->options);
261		free(p);
262		p = q;
263	}
264	archive_string_free(&mtree->line);
265	archive_string_free(&mtree->current_dir);
266	archive_string_free(&mtree->contents_name);
267	archive_entry_linkresolver_free(mtree->resolver);
268
269	free(mtree->buff);
270	free(mtree);
271	(a->format->data) = NULL;
272	return (ARCHIVE_OK);
273}
274
275static ssize_t
276get_line_size(const char *b, ssize_t avail, ssize_t *nlsize)
277{
278	ssize_t len;
279
280	len = 0;
281	while (len < avail) {
282		switch (*b) {
283		case '\0':/* Non-ascii character or control character. */
284			if (nlsize != NULL)
285				*nlsize = 0;
286			return (-1);
287		case '\r':
288			if (avail-len > 1 && b[1] == '\n') {
289				if (nlsize != NULL)
290					*nlsize = 2;
291				return (len+2);
292			}
293			/* FALL THROUGH */
294		case '\n':
295			if (nlsize != NULL)
296				*nlsize = 1;
297			return (len+1);
298		default:
299			b++;
300			len++;
301			break;
302		}
303	}
304	if (nlsize != NULL)
305		*nlsize = 0;
306	return (avail);
307}
308
309/*
310 *  <---------------- ravail --------------------->
311 *  <-- diff ------> <---  avail ----------------->
312 *                   <---- len ----------->
313 * | Previous lines | line being parsed  nl extra |
314 *                  ^
315 *                  b
316 *
317 */
318static ssize_t
319next_line(struct archive_read *a,
320    const char **b, ssize_t *avail, ssize_t *ravail, ssize_t *nl)
321{
322	ssize_t len;
323	int quit;
324
325	quit = 0;
326	if (*avail == 0) {
327		*nl = 0;
328		len = 0;
329	} else
330		len = get_line_size(*b, *avail, nl);
331	/*
332	 * Read bytes more while it does not reach the end of line.
333	 */
334	while (*nl == 0 && len == *avail && !quit) {
335		ssize_t diff = *ravail - *avail;
336		size_t nbytes_req = (*ravail+1023) & ~1023U;
337		ssize_t tested;
338
339		/*
340		 * Place an arbitrary limit on the line length.
341		 * mtree is almost free-form input and without line length limits,
342		 * it can consume a lot of memory.
343		 */
344		if (len >= MAX_LINE_LEN)
345			return (-1);
346
347		/* Increase reading bytes if it is not enough to at least
348		 * new two lines. */
349		if (nbytes_req < (size_t)*ravail + 160)
350			nbytes_req <<= 1;
351
352		*b = __archive_read_ahead(a, nbytes_req, avail);
353		if (*b == NULL) {
354			if (*ravail >= *avail)
355				return (0);
356			/* Reading bytes reaches the end of file. */
357			*b = __archive_read_ahead(a, *avail, avail);
358			quit = 1;
359		}
360		*ravail = *avail;
361		*b += diff;
362		*avail -= diff;
363		tested = len;/* Skip some bytes we already determinated. */
364		len = get_line_size(*b + len, *avail - len, nl);
365		if (len >= 0)
366			len += tested;
367	}
368	return (len);
369}
370
371/*
372 * Compare characters with a mtree keyword.
373 * Returns the length of a mtree keyword if matched.
374 * Returns 0 if not matched.
375 */
376static int
377bid_keycmp(const char *p, const char *key, ssize_t len)
378{
379	int match_len = 0;
380
381	while (len > 0 && *p && *key) {
382		if (*p == *key) {
383			--len;
384			++p;
385			++key;
386			++match_len;
387			continue;
388		}
389		return (0);/* Not match */
390	}
391	if (*key != '\0')
392		return (0);/* Not match */
393
394	/* A following character should be specified characters */
395	if (p[0] == '=' || p[0] == ' ' || p[0] == '\t' ||
396	    p[0] == '\n' || p[0] == '\r' ||
397	   (p[0] == '\\' && (p[1] == '\n' || p[1] == '\r')))
398		return (match_len);
399	return (0);/* Not match */
400}
401
402/*
403 * Test whether the characters 'p' has is mtree keyword.
404 * Returns the length of a detected keyword.
405 * Returns 0 if any keywords were not found.
406 */
407static int
408bid_keyword(const char *p,  ssize_t len)
409{
410	static const char * const keys_c[] = {
411		"content", "contents", "cksum", NULL
412	};
413	static const char * const keys_df[] = {
414		"device", "flags", NULL
415	};
416	static const char * const keys_g[] = {
417		"gid", "gname", NULL
418	};
419	static const char * const keys_il[] = {
420		"ignore", "inode", "link", NULL
421	};
422	static const char * const keys_m[] = {
423		"md5", "md5digest", "mode", NULL
424	};
425	static const char * const keys_no[] = {
426		"nlink", "nochange", "optional", NULL
427	};
428	static const char * const keys_r[] = {
429		"resdevice", "rmd160", "rmd160digest", NULL
430	};
431	static const char * const keys_s[] = {
432		"sha1", "sha1digest",
433		"sha256", "sha256digest",
434		"sha384", "sha384digest",
435		"sha512", "sha512digest",
436		"size", NULL
437	};
438	static const char * const keys_t[] = {
439		"tags", "time", "type", NULL
440	};
441	static const char * const keys_u[] = {
442		"uid", "uname",	NULL
443	};
444	const char * const *keys;
445	int i;
446
447	switch (*p) {
448	case 'c': keys = keys_c; break;
449	case 'd': case 'f': keys = keys_df; break;
450	case 'g': keys = keys_g; break;
451	case 'i': case 'l': keys = keys_il; break;
452	case 'm': keys = keys_m; break;
453	case 'n': case 'o': keys = keys_no; break;
454	case 'r': keys = keys_r; break;
455	case 's': keys = keys_s; break;
456	case 't': keys = keys_t; break;
457	case 'u': keys = keys_u; break;
458	default: return (0);/* Unknown key */
459	}
460
461	for (i = 0; keys[i] != NULL; i++) {
462		int l = bid_keycmp(p, keys[i], len);
463		if (l > 0)
464			return (l);
465	}
466	return (0);/* Unknown key */
467}
468
469/*
470 * Test whether there is a set of mtree keywords.
471 * Returns the number of keyword.
472 * Returns -1 if we got incorrect sequence.
473 * This function expects a set of "<space characters>keyword=value".
474 * When "unset" is specified, expects a set of "<space characters>keyword".
475 */
476static int
477bid_keyword_list(const char *p,  ssize_t len, int unset, int last_is_path)
478{
479	int l;
480	int keycnt = 0;
481
482	while (len > 0 && *p) {
483		int blank = 0;
484
485		/* Test whether there are blank characters in the line. */
486		while (len >0 && (*p == ' ' || *p == '\t')) {
487			++p;
488			--len;
489			blank = 1;
490		}
491		if (*p == '\n' || *p == '\r')
492			break;
493		if (p[0] == '\\' && (p[1] == '\n' || p[1] == '\r'))
494			break;
495		if (!blank && !last_is_path) /* No blank character. */
496			return (-1);
497		if (last_is_path && len == 0)
498				return (keycnt);
499
500		if (unset) {
501			l = bid_keycmp(p, "all", len);
502			if (l > 0)
503				return (1);
504		}
505		/* Test whether there is a correct key in the line. */
506		l = bid_keyword(p, len);
507		if (l == 0)
508			return (-1);/* Unknown keyword was found. */
509		p += l;
510		len -= l;
511		keycnt++;
512
513		/* Skip value */
514		if (*p == '=') {
515			int value = 0;
516			++p;
517			--len;
518			while (len > 0 && *p != ' ' && *p != '\t') {
519				++p;
520				--len;
521				value = 1;
522			}
523			/* A keyword should have a its value unless
524			 * "/unset" operation. */
525			if (!unset && value == 0)
526				return (-1);
527		}
528	}
529	return (keycnt);
530}
531
532static int
533bid_entry(const char *p, ssize_t len, ssize_t nl, int *last_is_path)
534{
535	int f = 0;
536	static const unsigned char safe_char[256] = {
537		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 00 - 0F */
538		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 10 - 1F */
539		/* !"$%&'()*+,-./  EXCLUSION:( )(#) */
540		0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 20 - 2F */
541		/* 0123456789:;<>?  EXCLUSION:(=) */
542		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, /* 30 - 3F */
543		/* @ABCDEFGHIJKLMNO */
544		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 40 - 4F */
545		/* PQRSTUVWXYZ[\]^_  */
546		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 50 - 5F */
547		/* `abcdefghijklmno */
548		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 60 - 6F */
549		/* pqrstuvwxyz{|}~ */
550		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, /* 70 - 7F */
551		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 80 - 8F */
552		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 90 - 9F */
553		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* A0 - AF */
554		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* B0 - BF */
555		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* C0 - CF */
556		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* D0 - DF */
557		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* E0 - EF */
558		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* F0 - FF */
559	};
560	ssize_t ll;
561	const char *pp = p;
562	const char * const pp_end = pp + len;
563
564	*last_is_path = 0;
565	/*
566	 * Skip the path-name which is quoted.
567	 */
568	for (;pp < pp_end; ++pp) {
569		if (!safe_char[*(const unsigned char *)pp]) {
570			if (*pp != ' ' && *pp != '\t' && *pp != '\r'
571			    && *pp != '\n')
572				f = 0;
573			break;
574		}
575		f = 1;
576	}
577	ll = pp_end - pp;
578
579	/* If a path-name was not found at the first, try to check
580	 * a mtree format(a.k.a form D) ``NetBSD's mtree -D'' creates,
581	 * which places the path-name at the last. */
582	if (f == 0) {
583		const char *pb = p + len - nl;
584		int name_len = 0;
585		int slash;
586
587		/* The form D accepts only a single line for an entry. */
588		if (pb-2 >= p &&
589		    pb[-1] == '\\' && (pb[-2] == ' ' || pb[-2] == '\t'))
590			return (-1);
591		if (pb-1 >= p && pb[-1] == '\\')
592			return (-1);
593
594		slash = 0;
595		while (p <= --pb && *pb != ' ' && *pb != '\t') {
596			if (!safe_char[*(const unsigned char *)pb])
597				return (-1);
598			name_len++;
599			/* The pathname should have a slash in this
600			 * format. */
601			if (*pb == '/')
602				slash = 1;
603		}
604		if (name_len == 0 || slash == 0)
605			return (-1);
606		/* If '/' is placed at the first in this field, this is not
607		 * a valid filename. */
608		if (pb[1] == '/')
609			return (-1);
610		ll = len - nl - name_len;
611		pp = p;
612		*last_is_path = 1;
613	}
614
615	return (bid_keyword_list(pp, ll, 0, *last_is_path));
616}
617
618#define MAX_BID_ENTRY	3
619
620static int
621mtree_bid(struct archive_read *a, int best_bid)
622{
623	const char *signature = "#mtree";
624	const char *p;
625
626	(void)best_bid; /* UNUSED */
627
628	/* Now let's look at the actual header and see if it matches. */
629	p = __archive_read_ahead(a, strlen(signature), NULL);
630	if (p == NULL)
631		return (-1);
632
633	if (memcmp(p, signature, strlen(signature)) == 0)
634		return (8 * (int)strlen(signature));
635
636	/*
637	 * There is not a mtree signature. Let's try to detect mtree format.
638	 */
639	return (detect_form(a, NULL));
640}
641
642static int
643detect_form(struct archive_read *a, int *is_form_d)
644{
645	const char *p;
646	ssize_t avail, ravail;
647	ssize_t detected_bytes = 0, len, nl;
648	int entry_cnt = 0, multiline = 0;
649	int form_D = 0;/* The archive is generated by `NetBSD mtree -D'
650			* (In this source we call it `form D') . */
651
652	if (is_form_d != NULL)
653		*is_form_d = 0;
654	p = __archive_read_ahead(a, 1, &avail);
655	if (p == NULL)
656		return (-1);
657	ravail = avail;
658	for (;;) {
659		len = next_line(a, &p, &avail, &ravail, &nl);
660		/* The terminal character of the line should be
661		 * a new line character, '\r\n' or '\n'. */
662		if (len <= 0 || nl == 0)
663			break;
664		if (!multiline) {
665			/* Leading whitespace is never significant,
666			 * ignore it. */
667			while (len > 0 && (*p == ' ' || *p == '\t')) {
668				++p;
669				--avail;
670				--len;
671			}
672			/* Skip comment or empty line. */
673			if (p[0] == '#' || p[0] == '\n' || p[0] == '\r') {
674				p += len;
675				avail -= len;
676				continue;
677			}
678		} else {
679			/* A continuance line; the terminal
680			 * character of previous line was '\' character. */
681			if (bid_keyword_list(p, len, 0, 0) <= 0)
682				break;
683			if (multiline == 1)
684				detected_bytes += len;
685			if (p[len-nl-1] != '\\') {
686				if (multiline == 1 &&
687				    ++entry_cnt >= MAX_BID_ENTRY)
688					break;
689				multiline = 0;
690			}
691			p += len;
692			avail -= len;
693			continue;
694		}
695		if (p[0] != '/') {
696			int last_is_path, keywords;
697
698			keywords = bid_entry(p, len, nl, &last_is_path);
699			if (keywords >= 0) {
700				detected_bytes += len;
701				if (form_D == 0) {
702					if (last_is_path)
703						form_D = 1;
704					else if (keywords > 0)
705						/* This line is not `form D'. */
706						form_D = -1;
707				} else if (form_D == 1) {
708					if (!last_is_path && keywords > 0)
709						/* This this is not `form D'
710						 * and We cannot accept mixed
711						 * format. */
712						break;
713				}
714				if (!last_is_path && p[len-nl-1] == '\\')
715					/* This line continues. */
716					multiline = 1;
717				else {
718					/* We've got plenty of correct lines
719					 * to assume that this file is a mtree
720					 * format. */
721					if (++entry_cnt >= MAX_BID_ENTRY)
722						break;
723				}
724			} else
725				break;
726		} else if (len > 4 && strncmp(p, "/set", 4) == 0) {
727			if (bid_keyword_list(p+4, len-4, 0, 0) <= 0)
728				break;
729			/* This line continues. */
730			if (p[len-nl-1] == '\\')
731				multiline = 2;
732		} else if (len > 6 && strncmp(p, "/unset", 6) == 0) {
733			if (bid_keyword_list(p+6, len-6, 1, 0) <= 0)
734				break;
735			/* This line continues. */
736			if (p[len-nl-1] == '\\')
737				multiline = 2;
738		} else
739			break;
740
741		/* Test next line. */
742		p += len;
743		avail -= len;
744	}
745	if (entry_cnt >= MAX_BID_ENTRY || (entry_cnt > 0 && len == 0)) {
746		if (is_form_d != NULL) {
747			if (form_D == 1)
748				*is_form_d = 1;
749		}
750		return (32);
751	}
752
753	return (0);
754}
755
756/*
757 * The extended mtree format permits multiple lines specifying
758 * attributes for each file.  For those entries, only the last line
759 * is actually used.  Practically speaking, that means we have
760 * to read the entire mtree file into memory up front.
761 *
762 * The parsing is done in two steps.  First, it is decided if a line
763 * changes the global defaults and if it is, processed accordingly.
764 * Otherwise, the options of the line are merged with the current
765 * global options.
766 */
767static int
768add_option(struct archive_read *a, struct mtree_option **global,
769    const char *value, size_t len)
770{
771	struct mtree_option *opt;
772
773	if ((opt = malloc(sizeof(*opt))) == NULL) {
774		archive_set_error(&a->archive, errno, "Can't allocate memory");
775		return (ARCHIVE_FATAL);
776	}
777	if ((opt->value = malloc(len + 1)) == NULL) {
778		free(opt);
779		archive_set_error(&a->archive, errno, "Can't allocate memory");
780		return (ARCHIVE_FATAL);
781	}
782	memcpy(opt->value, value, len);
783	opt->value[len] = '\0';
784	opt->next = *global;
785	*global = opt;
786	return (ARCHIVE_OK);
787}
788
789static void
790remove_option(struct mtree_option **global, const char *value, size_t len)
791{
792	struct mtree_option *iter, *last;
793
794	last = NULL;
795	for (iter = *global; iter != NULL; last = iter, iter = iter->next) {
796		if (strncmp(iter->value, value, len) == 0 &&
797		    (iter->value[len] == '\0' ||
798		     iter->value[len] == '='))
799			break;
800	}
801	if (iter == NULL)
802		return;
803	if (last == NULL)
804		*global = iter->next;
805	else
806		last->next = iter->next;
807
808	free(iter->value);
809	free(iter);
810}
811
812static int
813process_global_set(struct archive_read *a,
814    struct mtree_option **global, const char *line)
815{
816	const char *next, *eq;
817	size_t len;
818	int r;
819
820	line += 4;
821	for (;;) {
822		next = line + strspn(line, " \t\r\n");
823		if (*next == '\0')
824			return (ARCHIVE_OK);
825		line = next;
826		next = line + strcspn(line, " \t\r\n");
827		eq = strchr(line, '=');
828		if (eq > next)
829			len = next - line;
830		else
831			len = eq - line;
832
833		remove_option(global, line, len);
834		r = add_option(a, global, line, next - line);
835		if (r != ARCHIVE_OK)
836			return (r);
837		line = next;
838	}
839}
840
841static int
842process_global_unset(struct archive_read *a,
843    struct mtree_option **global, const char *line)
844{
845	const char *next;
846	size_t len;
847
848	line += 6;
849	if (strchr(line, '=') != NULL) {
850		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
851		    "/unset shall not contain `='");
852		return ARCHIVE_FATAL;
853	}
854
855	for (;;) {
856		next = line + strspn(line, " \t\r\n");
857		if (*next == '\0')
858			return (ARCHIVE_OK);
859		line = next;
860		len = strcspn(line, " \t\r\n");
861
862		if (len == 3 && strncmp(line, "all", 3) == 0) {
863			free_options(*global);
864			*global = NULL;
865		} else {
866			remove_option(global, line, len);
867		}
868
869		line += len;
870	}
871}
872
873static int
874process_add_entry(struct archive_read *a, struct mtree *mtree,
875    struct mtree_option **global, const char *line, ssize_t line_len,
876    struct mtree_entry **last_entry, int is_form_d)
877{
878	struct mtree_entry *entry, *ht_iter;
879	struct mtree_option *iter;
880	const char *next, *eq, *name, *end;
881	size_t name_len, len;
882	int r, i;
883	unsigned int ht_idx;
884
885	if ((entry = malloc(sizeof(*entry))) == NULL) {
886		archive_set_error(&a->archive, errno, "Can't allocate memory");
887		return (ARCHIVE_FATAL);
888	}
889	entry->next = NULL;
890	entry->options = NULL;
891	entry->name = NULL;
892	entry->used = 0;
893	entry->full = 0;
894	entry->name_hash = 0;
895	entry->hashtable_next = NULL;
896
897	/* Add this entry to list. */
898	if (*last_entry == NULL)
899		mtree->entries = entry;
900	else
901		(*last_entry)->next = entry;
902	*last_entry = entry;
903
904	if (is_form_d) {
905		/* Filename is last item on line. */
906		/* Adjust line_len to trim trailing whitespace */
907		while (line_len > 0) {
908			char last_character = line[line_len - 1];
909			if (last_character == '\r'
910			    || last_character == '\n'
911			    || last_character == '\t'
912			    || last_character == ' ') {
913				line_len--;
914			} else {
915				break;
916			}
917		}
918		/* Name starts after the last whitespace separator */
919		name = line;
920		for (i = 0; i < line_len; i++) {
921			if (line[i] == '\r'
922			    || line[i] == '\n'
923			    || line[i] == '\t'
924			    || line[i] == ' ') {
925				name = line + i + 1;
926			}
927		}
928		name_len = line + line_len - name;
929		end = name;
930	} else {
931		/* Filename is first item on line */
932		name_len = strcspn(line, " \t\r\n");
933		name = line;
934		line += name_len;
935		end = line + line_len;
936	}
937	/* name/name_len is the name within the line. */
938	/* line..end brackets the entire line except the name */
939
940	if ((entry->name = malloc(name_len + 1)) == NULL) {
941		archive_set_error(&a->archive, errno, "Can't allocate memory");
942		return (ARCHIVE_FATAL);
943	}
944
945	memcpy(entry->name, name, name_len);
946	entry->name[name_len] = '\0';
947	parse_escapes(entry->name, entry);
948	entry->name_hash = hash(entry->name);
949
950	ht_idx = entry->name_hash % MTREE_HASHTABLE_SIZE;
951	if ((ht_iter = mtree->entry_hashtable[ht_idx]) != NULL) {
952		while (ht_iter->hashtable_next)
953			ht_iter = ht_iter->hashtable_next;
954		ht_iter->hashtable_next = entry;
955	} else {
956		mtree->entry_hashtable[ht_idx] = entry;
957	}
958
959	for (iter = *global; iter != NULL; iter = iter->next) {
960		r = add_option(a, &entry->options, iter->value,
961		    strlen(iter->value));
962		if (r != ARCHIVE_OK)
963			return (r);
964	}
965
966	for (;;) {
967		next = line + strspn(line, " \t\r\n");
968		if (*next == '\0')
969			return (ARCHIVE_OK);
970		if (next >= end)
971			return (ARCHIVE_OK);
972		line = next;
973		next = line + strcspn(line, " \t\r\n");
974		eq = strchr(line, '=');
975		if (eq == NULL || eq > next)
976			len = next - line;
977		else
978			len = eq - line;
979
980		remove_option(&entry->options, line, len);
981		r = add_option(a, &entry->options, line, next - line);
982		if (r != ARCHIVE_OK)
983			return (r);
984		line = next;
985	}
986}
987
988static int
989read_mtree(struct archive_read *a, struct mtree *mtree)
990{
991	ssize_t len;
992	uintmax_t counter;
993	char *p;
994	struct mtree_option *global;
995	struct mtree_entry *last_entry;
996	int r, is_form_d;
997
998	mtree->archive_format = ARCHIVE_FORMAT_MTREE;
999	mtree->archive_format_name = "mtree";
1000
1001	global = NULL;
1002	last_entry = NULL;
1003
1004	(void)detect_form(a, &is_form_d);
1005
1006	for (counter = 1; ; ++counter) {
1007		len = readline(a, mtree, &p, 65536);
1008		if (len == 0) {
1009			mtree->this_entry = mtree->entries;
1010			free_options(global);
1011			return (ARCHIVE_OK);
1012		}
1013		if (len < 0) {
1014			free_options(global);
1015			return ((int)len);
1016		}
1017		/* Leading whitespace is never significant, ignore it. */
1018		while (*p == ' ' || *p == '\t') {
1019			++p;
1020			--len;
1021		}
1022		/* Skip content lines and blank lines. */
1023		if (*p == '#')
1024			continue;
1025		if (*p == '\r' || *p == '\n' || *p == '\0')
1026			continue;
1027		if (*p != '/') {
1028			r = process_add_entry(a, mtree, &global, p, len,
1029			    &last_entry, is_form_d);
1030		} else if (len > 4 && strncmp(p, "/set", 4) == 0) {
1031			if (p[4] != ' ' && p[4] != '\t')
1032				break;
1033			r = process_global_set(a, &global, p);
1034		} else if (len > 6 && strncmp(p, "/unset", 6) == 0) {
1035			if (p[6] != ' ' && p[6] != '\t')
1036				break;
1037			r = process_global_unset(a, &global, p);
1038		} else
1039			break;
1040
1041		if (r != ARCHIVE_OK) {
1042			free_options(global);
1043			return r;
1044		}
1045	}
1046
1047	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1048	    "Can't parse line %ju", counter);
1049	free_options(global);
1050	return (ARCHIVE_FATAL);
1051}
1052
1053/*
1054 * Read in the entire mtree file into memory on the first request.
1055 * Then use the next unused file to satisfy each header request.
1056 */
1057static int
1058read_header(struct archive_read *a, struct archive_entry *entry)
1059{
1060	struct mtree *mtree;
1061	char *p;
1062	int r, use_next;
1063
1064	mtree = (struct mtree *)(a->format->data);
1065
1066	if (mtree->fd >= 0) {
1067		close(mtree->fd);
1068		mtree->fd = -1;
1069	}
1070
1071	if (mtree->entries == NULL) {
1072		mtree->resolver = archive_entry_linkresolver_new();
1073		if (mtree->resolver == NULL)
1074			return ARCHIVE_FATAL;
1075		archive_entry_linkresolver_set_strategy(mtree->resolver,
1076		    ARCHIVE_FORMAT_MTREE);
1077		r = read_mtree(a, mtree);
1078		if (r != ARCHIVE_OK)
1079			return (r);
1080	}
1081
1082	a->archive.archive_format = mtree->archive_format;
1083	a->archive.archive_format_name = mtree->archive_format_name;
1084
1085	for (;;) {
1086		if (mtree->this_entry == NULL)
1087			return (ARCHIVE_EOF);
1088		if (strcmp(mtree->this_entry->name, "..") == 0) {
1089			mtree->this_entry->used = 1;
1090			if (archive_strlen(&mtree->current_dir) > 0) {
1091				/* Roll back current path. */
1092				p = mtree->current_dir.s
1093				    + mtree->current_dir.length - 1;
1094				while (p >= mtree->current_dir.s && *p != '/')
1095					--p;
1096				if (p >= mtree->current_dir.s)
1097					--p;
1098				mtree->current_dir.length
1099				    = p - mtree->current_dir.s + 1;
1100			}
1101		}
1102		if (!mtree->this_entry->used) {
1103			use_next = 0;
1104			r = parse_file(a, entry, mtree, mtree->this_entry,
1105				&use_next);
1106			if (use_next == 0)
1107				return (r);
1108		}
1109		mtree->this_entry = mtree->this_entry->next;
1110	}
1111}
1112
1113/*
1114 * A single file can have multiple lines contribute specifications.
1115 * Parse as many lines as necessary, then pull additional information
1116 * from a backing file on disk as necessary.
1117 */
1118static int
1119parse_file(struct archive_read *a, struct archive_entry *entry,
1120    struct mtree *mtree, struct mtree_entry *mentry, int *use_next)
1121{
1122	const char *path;
1123	struct stat st_storage, *st;
1124	struct mtree_entry *mp;
1125	struct archive_entry *sparse_entry;
1126	int r = ARCHIVE_OK, r1, parsed_kws;
1127
1128	mentry->used = 1;
1129
1130	/* Initialize reasonable defaults. */
1131	archive_entry_set_filetype(entry, AE_IFREG);
1132	archive_entry_set_size(entry, 0);
1133	archive_string_empty(&mtree->contents_name);
1134
1135	/* Parse options from this line. */
1136	parsed_kws = 0;
1137	r = parse_line(a, entry, mtree, mentry, &parsed_kws);
1138
1139	if (mentry->full) {
1140		archive_entry_copy_pathname(entry, mentry->name);
1141		/*
1142		 * "Full" entries are allowed to have multiple lines
1143		 * and those lines aren't required to be adjacent.  We
1144		 * don't support multiple lines for "relative" entries
1145		 * nor do we make any attempt to merge data from
1146		 * separate "relative" and "full" entries.  (Merging
1147		 * "relative" and "full" entries would require dealing
1148		 * with pathname canonicalization, which is a very
1149		 * tricky subject.)
1150		 */
1151		for (mp = mentry->hashtable_next; mp != NULL; mp = mp->hashtable_next) {
1152			if (mp->full && !mp->used
1153					&& mentry->name_hash == mp->name_hash
1154					&& strcmp(mentry->name, mp->name) == 0) {
1155				/* Later lines override earlier ones. */
1156				mp->used = 1;
1157				r1 = parse_line(a, entry, mtree, mp,
1158				    &parsed_kws);
1159				if (r1 < r)
1160					r = r1;
1161			}
1162		}
1163	} else {
1164		/*
1165		 * Relative entries require us to construct
1166		 * the full path and possibly update the
1167		 * current directory.
1168		 */
1169		size_t n = archive_strlen(&mtree->current_dir);
1170		if (n > 0)
1171			archive_strcat(&mtree->current_dir, "/");
1172		archive_strcat(&mtree->current_dir, mentry->name);
1173		archive_entry_copy_pathname(entry, mtree->current_dir.s);
1174		if (archive_entry_filetype(entry) != AE_IFDIR)
1175			mtree->current_dir.length = n;
1176	}
1177
1178	if (mtree->checkfs) {
1179		/*
1180		 * Try to open and stat the file to get the real size
1181		 * and other file info.  It would be nice to avoid
1182		 * this here so that getting a listing of an mtree
1183		 * wouldn't require opening every referenced contents
1184		 * file.  But then we wouldn't know the actual
1185		 * contents size, so I don't see a really viable way
1186		 * around this.  (Also, we may want to someday pull
1187		 * other unspecified info from the contents file on
1188		 * disk.)
1189		 */
1190		mtree->fd = -1;
1191		if (archive_strlen(&mtree->contents_name) > 0)
1192			path = mtree->contents_name.s;
1193		else
1194			path = archive_entry_pathname(entry);
1195
1196		if (archive_entry_filetype(entry) == AE_IFREG ||
1197				archive_entry_filetype(entry) == AE_IFDIR) {
1198			mtree->fd = open(path, O_RDONLY | O_BINARY | O_CLOEXEC);
1199			__archive_ensure_cloexec_flag(mtree->fd);
1200			if (mtree->fd == -1 &&
1201				(errno != ENOENT ||
1202				 archive_strlen(&mtree->contents_name) > 0)) {
1203				archive_set_error(&a->archive, errno,
1204						"Can't open %s", path);
1205				r = ARCHIVE_WARN;
1206			}
1207		}
1208
1209		st = &st_storage;
1210		if (mtree->fd >= 0) {
1211			if (fstat(mtree->fd, st) == -1) {
1212				archive_set_error(&a->archive, errno,
1213						"Could not fstat %s", path);
1214				r = ARCHIVE_WARN;
1215				/* If we can't stat it, don't keep it open. */
1216				close(mtree->fd);
1217				mtree->fd = -1;
1218				st = NULL;
1219			}
1220		} else if (lstat(path, st) == -1) {
1221			st = NULL;
1222		}
1223
1224		/*
1225		 * Check for a mismatch between the type in the specification
1226		 * and the type of the contents object on disk.
1227		 */
1228		if (st != NULL) {
1229			if (((st->st_mode & S_IFMT) == S_IFREG &&
1230			      archive_entry_filetype(entry) == AE_IFREG)
1231#ifdef S_IFLNK
1232			  ||((st->st_mode & S_IFMT) == S_IFLNK &&
1233			      archive_entry_filetype(entry) == AE_IFLNK)
1234#endif
1235#ifdef S_IFSOCK
1236			  ||((st->st_mode & S_IFSOCK) == S_IFSOCK &&
1237			      archive_entry_filetype(entry) == AE_IFSOCK)
1238#endif
1239#ifdef S_IFCHR
1240			  ||((st->st_mode & S_IFMT) == S_IFCHR &&
1241			      archive_entry_filetype(entry) == AE_IFCHR)
1242#endif
1243#ifdef S_IFBLK
1244			  ||((st->st_mode & S_IFMT) == S_IFBLK &&
1245			      archive_entry_filetype(entry) == AE_IFBLK)
1246#endif
1247			  ||((st->st_mode & S_IFMT) == S_IFDIR &&
1248			      archive_entry_filetype(entry) == AE_IFDIR)
1249#ifdef S_IFIFO
1250			  ||((st->st_mode & S_IFMT) == S_IFIFO &&
1251			      archive_entry_filetype(entry) == AE_IFIFO)
1252#endif
1253			) {
1254				/* Types match. */
1255			} else {
1256				/* Types don't match; bail out gracefully. */
1257				if (mtree->fd >= 0)
1258					close(mtree->fd);
1259				mtree->fd = -1;
1260				if (parsed_kws & MTREE_HAS_OPTIONAL) {
1261					/* It's not an error for an optional
1262					 * entry to not match disk. */
1263					*use_next = 1;
1264				} else if (r == ARCHIVE_OK) {
1265					archive_set_error(&a->archive,
1266					    ARCHIVE_ERRNO_MISC,
1267					    "mtree specification has different"
1268					    " type for %s",
1269					    archive_entry_pathname(entry));
1270					r = ARCHIVE_WARN;
1271				}
1272				return (r);
1273			}
1274		}
1275
1276		/*
1277		 * If there is a contents file on disk, pick some of the
1278		 * metadata from that file.  For most of these, we only
1279		 * set it from the contents if it wasn't already parsed
1280		 * from the specification.
1281		 */
1282		if (st != NULL) {
1283			if (((parsed_kws & MTREE_HAS_DEVICE) == 0 ||
1284				(parsed_kws & MTREE_HAS_NOCHANGE) != 0) &&
1285				(archive_entry_filetype(entry) == AE_IFCHR ||
1286				 archive_entry_filetype(entry) == AE_IFBLK))
1287				archive_entry_set_rdev(entry, st->st_rdev);
1288			if ((parsed_kws & (MTREE_HAS_GID | MTREE_HAS_GNAME))
1289				== 0 ||
1290			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1291				archive_entry_set_gid(entry, st->st_gid);
1292			if ((parsed_kws & (MTREE_HAS_UID | MTREE_HAS_UNAME))
1293				== 0 ||
1294			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1295				archive_entry_set_uid(entry, st->st_uid);
1296			if ((parsed_kws & MTREE_HAS_MTIME) == 0 ||
1297			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0) {
1298#if HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC
1299				archive_entry_set_mtime(entry, st->st_mtime,
1300						st->st_mtimespec.tv_nsec);
1301#elif HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC
1302				archive_entry_set_mtime(entry, st->st_mtime,
1303						st->st_mtim.tv_nsec);
1304#elif HAVE_STRUCT_STAT_ST_MTIME_N
1305				archive_entry_set_mtime(entry, st->st_mtime,
1306						st->st_mtime_n);
1307#elif HAVE_STRUCT_STAT_ST_UMTIME
1308				archive_entry_set_mtime(entry, st->st_mtime,
1309						st->st_umtime*1000);
1310#elif HAVE_STRUCT_STAT_ST_MTIME_USEC
1311				archive_entry_set_mtime(entry, st->st_mtime,
1312						st->st_mtime_usec*1000);
1313#else
1314				archive_entry_set_mtime(entry, st->st_mtime, 0);
1315#endif
1316			}
1317			if ((parsed_kws & MTREE_HAS_NLINK) == 0 ||
1318			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1319				archive_entry_set_nlink(entry, st->st_nlink);
1320			if ((parsed_kws & MTREE_HAS_PERM) == 0 ||
1321			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1322				archive_entry_set_perm(entry, st->st_mode);
1323			if ((parsed_kws & MTREE_HAS_SIZE) == 0 ||
1324			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1325				archive_entry_set_size(entry, st->st_size);
1326			archive_entry_set_ino(entry, st->st_ino);
1327			archive_entry_set_dev(entry, st->st_dev);
1328
1329			archive_entry_linkify(mtree->resolver, &entry,
1330				&sparse_entry);
1331		} else if (parsed_kws & MTREE_HAS_OPTIONAL) {
1332			/*
1333			 * Couldn't open the entry, stat it or the on-disk type
1334			 * didn't match.  If this entry is optional, just
1335			 * ignore it and read the next header entry.
1336			 */
1337			*use_next = 1;
1338			return ARCHIVE_OK;
1339		}
1340	}
1341
1342	mtree->cur_size = archive_entry_size(entry);
1343	mtree->offset = 0;
1344
1345	return r;
1346}
1347
1348/*
1349 * Each line contains a sequence of keywords.
1350 */
1351static int
1352parse_line(struct archive_read *a, struct archive_entry *entry,
1353    struct mtree *mtree, struct mtree_entry *mp, int *parsed_kws)
1354{
1355	struct mtree_option *iter;
1356	int r = ARCHIVE_OK, r1;
1357
1358	for (iter = mp->options; iter != NULL; iter = iter->next) {
1359		r1 = parse_keyword(a, mtree, entry, iter, parsed_kws);
1360		if (r1 < r)
1361			r = r1;
1362	}
1363	if (r == ARCHIVE_OK && (*parsed_kws & MTREE_HAS_TYPE) == 0) {
1364		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1365		    "Missing type keyword in mtree specification");
1366		return (ARCHIVE_WARN);
1367	}
1368	return (r);
1369}
1370
1371/*
1372 * Device entries have one of the following forms:
1373 *  - raw dev_t
1374 *  - format,major,minor[,subdevice]
1375 * When parsing succeeded, `pdev' will contain the appropriate dev_t value.
1376 */
1377
1378/* strsep() is not in C90, but strcspn() is. */
1379/* Taken from http://unixpapa.com/incnote/string.html */
1380static char *
1381la_strsep(char **sp, const char *sep)
1382{
1383	char *p, *s;
1384	if (sp == NULL || *sp == NULL || **sp == '\0')
1385		return(NULL);
1386	s = *sp;
1387	p = s + strcspn(s, sep);
1388	if (*p != '\0')
1389		*p++ = '\0';
1390	*sp = p;
1391	return(s);
1392}
1393
1394static int
1395parse_device(dev_t *pdev, struct archive *a, char *val)
1396{
1397#define MAX_PACK_ARGS 3
1398	unsigned long numbers[MAX_PACK_ARGS];
1399	char *p, *dev;
1400	int argc;
1401	pack_t *pack;
1402	dev_t result;
1403	const char *error = NULL;
1404
1405	memset(pdev, 0, sizeof(*pdev));
1406	if ((dev = strchr(val, ',')) != NULL) {
1407		/*
1408		 * Device's major/minor are given in a specified format.
1409		 * Decode and pack it accordingly.
1410		 */
1411		*dev++ = '\0';
1412		if ((pack = pack_find(val)) == NULL) {
1413			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1414			    "Unknown format `%s'", val);
1415			return ARCHIVE_WARN;
1416		}
1417		argc = 0;
1418		while ((p = la_strsep(&dev, ",")) != NULL) {
1419			if (*p == '\0') {
1420				archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1421				    "Missing number");
1422				return ARCHIVE_WARN;
1423			}
1424			if (argc >= MAX_PACK_ARGS) {
1425				archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1426				    "Too many arguments");
1427				return ARCHIVE_WARN;
1428			}
1429			numbers[argc++] = (unsigned long)mtree_atol(&p, 0);
1430		}
1431		if (argc < 2) {
1432			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1433			    "Not enough arguments");
1434			return ARCHIVE_WARN;
1435		}
1436		result = (*pack)(argc, numbers, &error);
1437		if (error != NULL) {
1438			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1439			    "%s", error);
1440			return ARCHIVE_WARN;
1441		}
1442	} else {
1443		/* file system raw value. */
1444		result = (dev_t)mtree_atol(&val, 0);
1445	}
1446	*pdev = result;
1447	return ARCHIVE_OK;
1448#undef MAX_PACK_ARGS
1449}
1450
1451/*
1452 * Parse a single keyword and its value.
1453 */
1454static int
1455parse_keyword(struct archive_read *a, struct mtree *mtree,
1456    struct archive_entry *entry, struct mtree_option *opt, int *parsed_kws)
1457{
1458	char *val, *key;
1459
1460	key = opt->value;
1461
1462	if (*key == '\0')
1463		return (ARCHIVE_OK);
1464
1465	if (strcmp(key, "nochange") == 0) {
1466		*parsed_kws |= MTREE_HAS_NOCHANGE;
1467		return (ARCHIVE_OK);
1468	}
1469	if (strcmp(key, "optional") == 0) {
1470		*parsed_kws |= MTREE_HAS_OPTIONAL;
1471		return (ARCHIVE_OK);
1472	}
1473	if (strcmp(key, "ignore") == 0) {
1474		/*
1475		 * The mtree processing is not recursive, so
1476		 * recursion will only happen for explicitly listed
1477		 * entries.
1478		 */
1479		return (ARCHIVE_OK);
1480	}
1481
1482	val = strchr(key, '=');
1483	if (val == NULL) {
1484		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1485		    "Malformed attribute \"%s\" (%d)", key, key[0]);
1486		return (ARCHIVE_WARN);
1487	}
1488
1489	*val = '\0';
1490	++val;
1491
1492	switch (key[0]) {
1493	case 'c':
1494		if (strcmp(key, "content") == 0
1495		    || strcmp(key, "contents") == 0) {
1496			parse_escapes(val, NULL);
1497			archive_strcpy(&mtree->contents_name, val);
1498			break;
1499		}
1500		if (strcmp(key, "cksum") == 0)
1501			break;
1502		__LA_FALLTHROUGH;
1503	case 'd':
1504		if (strcmp(key, "device") == 0) {
1505			/* stat(2) st_rdev field, e.g. the major/minor IDs
1506			 * of a char/block special file */
1507			int r;
1508			dev_t dev;
1509
1510			*parsed_kws |= MTREE_HAS_DEVICE;
1511			r = parse_device(&dev, &a->archive, val);
1512			if (r == ARCHIVE_OK)
1513				archive_entry_set_rdev(entry, dev);
1514			return r;
1515		}
1516		__LA_FALLTHROUGH;
1517	case 'f':
1518		if (strcmp(key, "flags") == 0) {
1519			*parsed_kws |= MTREE_HAS_FFLAGS;
1520			archive_entry_copy_fflags_text(entry, val);
1521			break;
1522		}
1523		__LA_FALLTHROUGH;
1524	case 'g':
1525		if (strcmp(key, "gid") == 0) {
1526			*parsed_kws |= MTREE_HAS_GID;
1527			archive_entry_set_gid(entry, mtree_atol(&val, 10));
1528			break;
1529		}
1530		if (strcmp(key, "gname") == 0) {
1531			*parsed_kws |= MTREE_HAS_GNAME;
1532			archive_entry_copy_gname(entry, val);
1533			break;
1534		}
1535		__LA_FALLTHROUGH;
1536	case 'i':
1537		if (strcmp(key, "inode") == 0) {
1538			archive_entry_set_ino(entry, mtree_atol(&val, 10));
1539			break;
1540		}
1541		__LA_FALLTHROUGH;
1542	case 'l':
1543		if (strcmp(key, "link") == 0) {
1544			archive_entry_copy_symlink(entry, val);
1545			break;
1546		}
1547		__LA_FALLTHROUGH;
1548	case 'm':
1549		if (strcmp(key, "md5") == 0 || strcmp(key, "md5digest") == 0)
1550			break;
1551		if (strcmp(key, "mode") == 0) {
1552			if (val[0] >= '0' && val[0] <= '7') {
1553				*parsed_kws |= MTREE_HAS_PERM;
1554				archive_entry_set_perm(entry,
1555				    (mode_t)mtree_atol(&val, 8));
1556			} else {
1557				archive_set_error(&a->archive,
1558				    ARCHIVE_ERRNO_FILE_FORMAT,
1559				    "Symbolic or non-octal mode \"%s\" unsupported", val);
1560				return ARCHIVE_WARN;
1561			}
1562			break;
1563		}
1564		__LA_FALLTHROUGH;
1565	case 'n':
1566		if (strcmp(key, "nlink") == 0) {
1567			*parsed_kws |= MTREE_HAS_NLINK;
1568			archive_entry_set_nlink(entry,
1569				(unsigned int)mtree_atol(&val, 10));
1570			break;
1571		}
1572		__LA_FALLTHROUGH;
1573	case 'r':
1574		if (strcmp(key, "resdevice") == 0) {
1575			/* stat(2) st_dev field, e.g. the device ID where the
1576			 * inode resides */
1577			int r;
1578			dev_t dev;
1579
1580			r = parse_device(&dev, &a->archive, val);
1581			if (r == ARCHIVE_OK)
1582				archive_entry_set_dev(entry, dev);
1583			return r;
1584		}
1585		if (strcmp(key, "rmd160") == 0 ||
1586		    strcmp(key, "rmd160digest") == 0)
1587			break;
1588		__LA_FALLTHROUGH;
1589	case 's':
1590		if (strcmp(key, "sha1") == 0 || strcmp(key, "sha1digest") == 0)
1591			break;
1592		if (strcmp(key, "sha256") == 0 ||
1593		    strcmp(key, "sha256digest") == 0)
1594			break;
1595		if (strcmp(key, "sha384") == 0 ||
1596		    strcmp(key, "sha384digest") == 0)
1597			break;
1598		if (strcmp(key, "sha512") == 0 ||
1599		    strcmp(key, "sha512digest") == 0)
1600			break;
1601		if (strcmp(key, "size") == 0) {
1602			archive_entry_set_size(entry, mtree_atol(&val, 10));
1603			break;
1604		}
1605		__LA_FALLTHROUGH;
1606	case 't':
1607		if (strcmp(key, "tags") == 0) {
1608			/*
1609			 * Comma delimited list of tags.
1610			 * Ignore the tags for now, but the interface
1611			 * should be extended to allow inclusion/exclusion.
1612			 */
1613			break;
1614		}
1615		if (strcmp(key, "time") == 0) {
1616			int64_t m;
1617			int64_t my_time_t_max = get_time_t_max();
1618			int64_t my_time_t_min = get_time_t_min();
1619			long ns = 0;
1620
1621			*parsed_kws |= MTREE_HAS_MTIME;
1622			m = mtree_atol(&val, 10);
1623			/* Replicate an old mtree bug:
1624			 * 123456789.1 represents 123456789
1625			 * seconds and 1 nanosecond. */
1626			if (*val == '.') {
1627				++val;
1628				ns = (long)mtree_atol(&val, 10);
1629				if (ns < 0)
1630					ns = 0;
1631				else if (ns > 999999999)
1632					ns = 999999999;
1633			}
1634			if (m > my_time_t_max)
1635				m = my_time_t_max;
1636			else if (m < my_time_t_min)
1637				m = my_time_t_min;
1638			archive_entry_set_mtime(entry, (time_t)m, ns);
1639			break;
1640		}
1641		if (strcmp(key, "type") == 0) {
1642			switch (val[0]) {
1643			case 'b':
1644				if (strcmp(val, "block") == 0) {
1645					archive_entry_set_filetype(entry, AE_IFBLK);
1646					break;
1647				}
1648				__LA_FALLTHROUGH;
1649			case 'c':
1650				if (strcmp(val, "char") == 0) {
1651					archive_entry_set_filetype(entry,
1652						AE_IFCHR);
1653					break;
1654				}
1655				__LA_FALLTHROUGH;
1656			case 'd':
1657				if (strcmp(val, "dir") == 0) {
1658					archive_entry_set_filetype(entry,
1659						AE_IFDIR);
1660					break;
1661				}
1662				__LA_FALLTHROUGH;
1663			case 'f':
1664				if (strcmp(val, "fifo") == 0) {
1665					archive_entry_set_filetype(entry,
1666						AE_IFIFO);
1667					break;
1668				}
1669				if (strcmp(val, "file") == 0) {
1670					archive_entry_set_filetype(entry,
1671						AE_IFREG);
1672					break;
1673				}
1674				__LA_FALLTHROUGH;
1675			case 'l':
1676				if (strcmp(val, "link") == 0) {
1677					archive_entry_set_filetype(entry,
1678						AE_IFLNK);
1679					break;
1680				}
1681				__LA_FALLTHROUGH;
1682			default:
1683				archive_set_error(&a->archive,
1684				    ARCHIVE_ERRNO_FILE_FORMAT,
1685				    "Unrecognized file type \"%s\"; "
1686				    "assuming \"file\"", val);
1687				archive_entry_set_filetype(entry, AE_IFREG);
1688				return (ARCHIVE_WARN);
1689			}
1690			*parsed_kws |= MTREE_HAS_TYPE;
1691			break;
1692		}
1693		__LA_FALLTHROUGH;
1694	case 'u':
1695		if (strcmp(key, "uid") == 0) {
1696			*parsed_kws |= MTREE_HAS_UID;
1697			archive_entry_set_uid(entry, mtree_atol(&val, 10));
1698			break;
1699		}
1700		if (strcmp(key, "uname") == 0) {
1701			*parsed_kws |= MTREE_HAS_UNAME;
1702			archive_entry_copy_uname(entry, val);
1703			break;
1704		}
1705		__LA_FALLTHROUGH;
1706	default:
1707		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1708		    "Unrecognized key %s=%s", key, val);
1709		return (ARCHIVE_WARN);
1710	}
1711	return (ARCHIVE_OK);
1712}
1713
1714static int
1715read_data(struct archive_read *a, const void **buff, size_t *size,
1716    int64_t *offset)
1717{
1718	size_t bytes_to_read;
1719	ssize_t bytes_read;
1720	struct mtree *mtree;
1721
1722	mtree = (struct mtree *)(a->format->data);
1723	if (mtree->fd < 0) {
1724		*buff = NULL;
1725		*offset = 0;
1726		*size = 0;
1727		return (ARCHIVE_EOF);
1728	}
1729	if (mtree->buff == NULL) {
1730		mtree->buffsize = 64 * 1024;
1731		mtree->buff = malloc(mtree->buffsize);
1732		if (mtree->buff == NULL) {
1733			archive_set_error(&a->archive, ENOMEM,
1734			    "Can't allocate memory");
1735			return (ARCHIVE_FATAL);
1736		}
1737	}
1738
1739	*buff = mtree->buff;
1740	*offset = mtree->offset;
1741	if ((int64_t)mtree->buffsize > mtree->cur_size - mtree->offset)
1742		bytes_to_read = (size_t)(mtree->cur_size - mtree->offset);
1743	else
1744		bytes_to_read = mtree->buffsize;
1745	bytes_read = read(mtree->fd, mtree->buff, bytes_to_read);
1746	if (bytes_read < 0) {
1747		archive_set_error(&a->archive, errno, "Can't read");
1748		return (ARCHIVE_WARN);
1749	}
1750	if (bytes_read == 0) {
1751		*size = 0;
1752		return (ARCHIVE_EOF);
1753	}
1754	mtree->offset += bytes_read;
1755	*size = bytes_read;
1756	return (ARCHIVE_OK);
1757}
1758
1759/* Skip does nothing except possibly close the contents file. */
1760static int
1761skip(struct archive_read *a)
1762{
1763	struct mtree *mtree;
1764
1765	mtree = (struct mtree *)(a->format->data);
1766	if (mtree->fd >= 0) {
1767		close(mtree->fd);
1768		mtree->fd = -1;
1769	}
1770	return (ARCHIVE_OK);
1771}
1772
1773/*
1774 * Since parsing backslash sequences always makes strings shorter,
1775 * we can always do this conversion in-place.
1776 */
1777static void
1778parse_escapes(char *src, struct mtree_entry *mentry)
1779{
1780	char *dest = src;
1781	char c;
1782
1783	if (mentry != NULL && strcmp(src, ".") == 0)
1784		mentry->full = 1;
1785
1786	while (*src != '\0') {
1787		c = *src++;
1788		if (c == '/' && mentry != NULL)
1789			mentry->full = 1;
1790		if (c == '\\') {
1791			switch (src[0]) {
1792			case '0':
1793				if (src[1] < '0' || src[1] > '7') {
1794					c = 0;
1795					++src;
1796					break;
1797				}
1798				/* FALLTHROUGH */
1799			case '1':
1800			case '2':
1801			case '3':
1802				if (src[1] >= '0' && src[1] <= '7' &&
1803				    src[2] >= '0' && src[2] <= '7') {
1804					c = (src[0] - '0') << 6;
1805					c |= (src[1] - '0') << 3;
1806					c |= (src[2] - '0');
1807					src += 3;
1808				}
1809				break;
1810			case 'a':
1811				c = '\a';
1812				++src;
1813				break;
1814			case 'b':
1815				c = '\b';
1816				++src;
1817				break;
1818			case 'f':
1819				c = '\f';
1820				++src;
1821				break;
1822			case 'n':
1823				c = '\n';
1824				++src;
1825				break;
1826			case 'r':
1827				c = '\r';
1828				++src;
1829				break;
1830			case 's':
1831				c = ' ';
1832				++src;
1833				break;
1834			case 't':
1835				c = '\t';
1836				++src;
1837				break;
1838			case 'v':
1839				c = '\v';
1840				++src;
1841				break;
1842			case '\\':
1843				c = '\\';
1844				++src;
1845				break;
1846			}
1847		}
1848		*dest++ = c;
1849	}
1850	*dest = '\0';
1851}
1852
1853/* Parse a hex digit. */
1854static int
1855parsedigit(char c)
1856{
1857	if (c >= '0' && c <= '9')
1858		return c - '0';
1859	else if (c >= 'a' && c <= 'f')
1860		return c - 'a';
1861	else if (c >= 'A' && c <= 'F')
1862		return c - 'A';
1863	else
1864		return -1;
1865}
1866
1867/*
1868 * Note that this implementation does not (and should not!) obey
1869 * locale settings; you cannot simply substitute strtol here, since
1870 * it does obey locale.
1871 */
1872static int64_t
1873mtree_atol(char **p, int base)
1874{
1875	int64_t l, limit;
1876	int digit, last_digit_limit;
1877
1878	if (base == 0) {
1879		if (**p != '0')
1880			base = 10;
1881		else if ((*p)[1] == 'x' || (*p)[1] == 'X') {
1882			*p += 2;
1883			base = 16;
1884		} else {
1885			base = 8;
1886		}
1887	}
1888
1889	if (**p == '-') {
1890		limit = INT64_MIN / base;
1891		last_digit_limit = INT64_MIN % base;
1892		++(*p);
1893
1894		l = 0;
1895		digit = parsedigit(**p);
1896		while (digit >= 0 && digit < base) {
1897			if (l < limit || (l == limit && digit > last_digit_limit))
1898				return INT64_MIN;
1899			l = (l * base) - digit;
1900			digit = parsedigit(*++(*p));
1901		}
1902		return l;
1903	} else {
1904		limit = INT64_MAX / base;
1905		last_digit_limit = INT64_MAX % base;
1906
1907		l = 0;
1908		digit = parsedigit(**p);
1909		while (digit >= 0 && digit < base) {
1910			if (l > limit || (l == limit && digit > last_digit_limit))
1911				return INT64_MAX;
1912			l = (l * base) + digit;
1913			digit = parsedigit(*++(*p));
1914		}
1915		return l;
1916	}
1917}
1918
1919/*
1920 * Returns length of line (including trailing newline)
1921 * or negative on error.  'start' argument is updated to
1922 * point to first character of line.
1923 */
1924static ssize_t
1925readline(struct archive_read *a, struct mtree *mtree, char **start,
1926    ssize_t limit)
1927{
1928	ssize_t bytes_read;
1929	ssize_t total_size = 0;
1930	ssize_t find_off = 0;
1931	const void *t;
1932	void *nl;
1933	char *u;
1934
1935	/* Accumulate line in a line buffer. */
1936	for (;;) {
1937		/* Read some more. */
1938		t = __archive_read_ahead(a, 1, &bytes_read);
1939		if (t == NULL)
1940			return (0);
1941		if (bytes_read < 0)
1942			return (ARCHIVE_FATAL);
1943		nl = memchr(t, '\n', bytes_read);
1944		/* If we found '\n', trim the read to end exactly there. */
1945		if (nl != NULL) {
1946			bytes_read = ((const char *)nl) - ((const char *)t) + 1;
1947		}
1948		if (total_size + bytes_read + 1 > limit) {
1949			archive_set_error(&a->archive,
1950			    ARCHIVE_ERRNO_FILE_FORMAT,
1951			    "Line too long");
1952			return (ARCHIVE_FATAL);
1953		}
1954		if (archive_string_ensure(&mtree->line,
1955			total_size + bytes_read + 1) == NULL) {
1956			archive_set_error(&a->archive, ENOMEM,
1957			    "Can't allocate working buffer");
1958			return (ARCHIVE_FATAL);
1959		}
1960		/* Append new bytes to string. */
1961		memcpy(mtree->line.s + total_size, t, bytes_read);
1962		__archive_read_consume(a, bytes_read);
1963		total_size += bytes_read;
1964		mtree->line.s[total_size] = '\0';
1965
1966		for (u = mtree->line.s + find_off; *u; ++u) {
1967			if (u[0] == '\n') {
1968				/* Ends with unescaped newline. */
1969				*start = mtree->line.s;
1970				return total_size;
1971			} else if (u[0] == '#') {
1972				/* Ends with comment sequence #...\n */
1973				if (nl == NULL) {
1974					/* But we've not found the \n yet */
1975					break;
1976				}
1977			} else if (u[0] == '\\') {
1978				if (u[1] == '\n') {
1979					/* Trim escaped newline. */
1980					total_size -= 2;
1981					mtree->line.s[total_size] = '\0';
1982					break;
1983				} else if (u[1] != '\0') {
1984					/* Skip the two-char escape sequence */
1985					++u;
1986				}
1987			}
1988		}
1989		find_off = u - mtree->line.s;
1990	}
1991}
1992
1993static unsigned int
1994hash(const char *p)
1995{
1996	/* A 32-bit version of Peter Weinberger's (PJW) hash algorithm,
1997	   as used by ELF for hashing function names. */
1998	unsigned g, h = 0;
1999	while (*p != '\0') {
2000		h = (h << 4) + *p++;
2001		if ((g = h & 0xF0000000) != 0) {
2002			h ^= g >> 24;
2003			h &= 0x0FFFFFFF;
2004		}
2005	}
2006	return h;
2007}
2008