xref: /freebsd/contrib/libarchive/libarchive/archive_read_support_format_mtree.c (revision 401026e4825a05abba6f945cf1b74b3328876fa2)
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 
30 #ifdef HAVE_SYS_STAT_H
31 #include <sys/stat.h>
32 #endif
33 #ifdef HAVE_ERRNO_H
34 #include <errno.h>
35 #endif
36 #ifdef HAVE_FCNTL_H
37 #include <fcntl.h>
38 #endif
39 #include <stddef.h>
40 /* #include <stdint.h> */ /* See archive_platform.h */
41 #ifdef HAVE_STDLIB_H
42 #include <stdlib.h>
43 #endif
44 #ifdef HAVE_STRING_H
45 #include <string.h>
46 #endif
47 #ifdef HAVE_CTYPE_H
48 #include <ctype.h>
49 #endif
50 
51 #include "archive.h"
52 #include "archive_entry.h"
53 #include "archive_entry_private.h"
54 #include "archive_platform_stat.h"
55 #include "archive_private.h"
56 #include "archive_rb.h"
57 #include "archive_read_private.h"
58 #include "archive_string.h"
59 #include "archive_pack_dev.h"
60 
61 #ifndef O_BINARY
62 #define	O_BINARY 0
63 #endif
64 #ifndef O_CLOEXEC
65 #define O_CLOEXEC	0
66 #endif
67 
68 #define	MTREE_HAS_DEVICE	0x0001
69 #define	MTREE_HAS_FFLAGS	0x0002
70 #define	MTREE_HAS_GID		0x0004
71 #define	MTREE_HAS_GNAME		0x0008
72 #define	MTREE_HAS_MTIME		0x0010
73 #define	MTREE_HAS_NLINK		0x0020
74 #define	MTREE_HAS_PERM		0x0040
75 #define	MTREE_HAS_SIZE		0x0080
76 #define	MTREE_HAS_TYPE		0x0100
77 #define	MTREE_HAS_UID		0x0200
78 #define	MTREE_HAS_UNAME		0x0400
79 
80 #define	MTREE_HAS_OPTIONAL	0x0800
81 #define	MTREE_HAS_NOCHANGE	0x1000 /* FreeBSD specific */
82 
83 #define	MAX_LINE_LEN		(1024 * 1024)
84 
85 struct mtree_option {
86 	struct mtree_option *next;
87 	char *value;
88 };
89 
90 struct mtree_entry {
91 	struct archive_rb_node rbnode;
92 	struct mtree_entry *next_dup;
93 	struct mtree_entry *next;
94 	struct mtree_option *options;
95 	char *name;
96 	char full;
97 	char used;
98 };
99 
100 struct mtree {
101 	struct archive_string	 line;
102 	size_t			 buffsize;
103 	char			*buff;
104 	int64_t			 offset;
105 	int			 fd;
106 	int			 archive_format;
107 	const char		*archive_format_name;
108 	struct mtree_entry	*entries;
109 	struct mtree_entry	*this_entry;
110 	struct archive_rb_tree	 entry_rbtree;
111 	struct archive_string	 current_dir;
112 	struct archive_string	 contents_name;
113 
114 	struct archive_entry_linkresolver *resolver;
115 	struct archive_rb_tree rbtree;
116 
117 	int64_t			 cur_size;
118 	char checkfs;
119 };
120 
121 static int	bid_keycmp(const char *, const char *, ssize_t);
122 static int	cleanup(struct archive_read *);
123 static int	detect_form(struct archive_read *, int *);
124 static int	mtree_bid(struct archive_read *, int);
125 static int	parse_file(struct archive_read *, struct archive_entry *,
126 		    struct mtree *, struct mtree_entry *, int *);
127 static void	parse_escapes(char *, struct mtree_entry *);
128 static int	parse_line(struct archive_read *, struct archive_entry *,
129 		    struct mtree *, struct mtree_entry *, int *);
130 static int	parse_keyword(struct archive_read *, struct mtree *,
131 		    struct archive_entry *, struct mtree_option *, int *);
132 static int	read_data(struct archive_read *a,
133 		    const void **buff, size_t *size, int64_t *offset);
134 static ssize_t	readline(struct archive_read *, struct mtree *, char **, ssize_t);
135 static int	skip(struct archive_read *a);
136 static int	read_header(struct archive_read *,
137 		    struct archive_entry *);
138 static int64_t	mtree_atol(char **, int base);
139 #ifndef HAVE_STRNLEN
140 static size_t	mtree_strnlen(const char *, size_t);
141 #endif
142 
143 /*
144  * There's no standard for TIME_T_MAX/TIME_T_MIN.  So we compute them
145  * here.  TODO: Move this to configure time, but be careful
146  * about cross-compile environments.
147  */
148 static int64_t
get_time_t_max(void)149 get_time_t_max(void)
150 {
151 #if defined(TIME_T_MAX)
152 	return TIME_T_MAX;
153 #else
154 	/* ISO C allows time_t to be a floating-point type,
155 	   but POSIX requires an integer type.  The following
156 	   should work on any system that follows the POSIX
157 	   conventions. */
158 	if (((time_t)0) < ((time_t)-1)) {
159 		/* Time_t is unsigned */
160 		return (~(time_t)0);
161 	} else {
162 		/* Time_t is signed. */
163 		/* Assume it's the same as int64_t or int32_t */
164 		if (sizeof(time_t) == sizeof(int64_t)) {
165 			return (time_t)INT64_MAX;
166 		} else {
167 			return (time_t)INT32_MAX;
168 		}
169 	}
170 #endif
171 }
172 
173 static int64_t
get_time_t_min(void)174 get_time_t_min(void)
175 {
176 #if defined(TIME_T_MIN)
177 	return TIME_T_MIN;
178 #else
179 	if (((time_t)0) < ((time_t)-1)) {
180 		/* Time_t is unsigned */
181 		return (time_t)0;
182 	} else {
183 		/* Time_t is signed. */
184 		if (sizeof(time_t) == sizeof(int64_t)) {
185 			return (time_t)INT64_MIN;
186 		} else {
187 			return (time_t)INT32_MIN;
188 		}
189 	}
190 #endif
191 }
192 
193 #ifdef HAVE_STRNLEN
194 #define mtree_strnlen(a,b) strnlen(a,b)
195 #else
196 static size_t
mtree_strnlen(const char * p,size_t maxlen)197 mtree_strnlen(const char *p, size_t maxlen)
198 {
199 	size_t i;
200 
201 	for (i = 0; i <= maxlen; i++) {
202 		if (p[i] == 0)
203 			break;
204 	}
205 	if (i > maxlen)
206 		return (-1);/* invalid */
207 	return (i);
208 }
209 #endif
210 
211 static int
archive_read_format_mtree_options(struct archive_read * a,const char * key,const char * val)212 archive_read_format_mtree_options(struct archive_read *a,
213     const char *key, const char *val)
214 {
215 	struct mtree *mtree;
216 
217 	mtree = (struct mtree *)(a->format->data);
218 	if (strcmp(key, "checkfs")  == 0) {
219 		/* Allows to read information missing from the mtree from the file system */
220 		if (val == NULL || val[0] == 0) {
221 			mtree->checkfs = 0;
222 		} else {
223 			mtree->checkfs = 1;
224 		}
225 		return (ARCHIVE_OK);
226 	}
227 
228 	/* Note: The "warn" return is just to inform the options
229 	 * supervisor that we didn't handle it.  It will generate
230 	 * a suitable error if no one used this option. */
231 	return (ARCHIVE_WARN);
232 }
233 
234 static void
free_options(struct mtree_option * head)235 free_options(struct mtree_option *head)
236 {
237 	struct mtree_option *next;
238 
239 	for (; head != NULL; head = next) {
240 		next = head->next;
241 		free(head->value);
242 		free(head);
243 	}
244 }
245 
246 static int
mtree_cmp_node(const struct archive_rb_node * n1,const struct archive_rb_node * n2)247 mtree_cmp_node(const struct archive_rb_node *n1,
248     const struct archive_rb_node *n2)
249 {
250 	const struct mtree_entry *e1 = (const struct mtree_entry *)n1;
251 	const struct mtree_entry *e2 = (const struct mtree_entry *)n2;
252 
253 	return (strcmp(e1->name, e2->name));
254 }
255 
256 static int
mtree_cmp_key(const struct archive_rb_node * n,const void * key)257 mtree_cmp_key(const struct archive_rb_node *n, const void *key)
258 {
259 	const struct mtree_entry *e = (const struct mtree_entry *)n;
260 
261 	return (strcmp(e->name, key));
262 }
263 
264 int
archive_read_support_format_mtree(struct archive * _a)265 archive_read_support_format_mtree(struct archive *_a)
266 {
267 	static const struct archive_rb_tree_ops rb_ops = {
268 		mtree_cmp_node, mtree_cmp_key,
269 	};
270 	struct archive_read *a = (struct archive_read *)_a;
271 	struct mtree *mtree;
272 	int r;
273 
274 	archive_check_magic(_a, ARCHIVE_READ_MAGIC,
275 	    ARCHIVE_STATE_NEW, "archive_read_support_format_mtree");
276 
277 	mtree = calloc(1, sizeof(*mtree));
278 	if (mtree == NULL) {
279 		archive_set_error(&a->archive, ENOMEM,
280 		    "Can't allocate mtree data");
281 		return (ARCHIVE_FATAL);
282 	}
283 	mtree->checkfs = 0;
284 	mtree->fd = -1;
285 
286 	__archive_rb_tree_init(&mtree->rbtree, &rb_ops);
287 
288 	r = __archive_read_register_format(a, mtree, "mtree",
289            mtree_bid, archive_read_format_mtree_options, read_header, read_data, skip, NULL, cleanup, NULL, NULL);
290 
291 	if (r != ARCHIVE_OK)
292 		free(mtree);
293 	return (ARCHIVE_OK);
294 }
295 
296 static int
cleanup(struct archive_read * a)297 cleanup(struct archive_read *a)
298 {
299 	struct mtree *mtree;
300 	struct mtree_entry *p, *q;
301 
302 	mtree = (struct mtree *)(a->format->data);
303 
304 	p = mtree->entries;
305 	while (p != NULL) {
306 		q = p->next;
307 		free(p->name);
308 		free_options(p->options);
309 		free(p);
310 		p = q;
311 	}
312 	archive_string_free(&mtree->line);
313 	archive_string_free(&mtree->current_dir);
314 	archive_string_free(&mtree->contents_name);
315 	archive_entry_linkresolver_free(mtree->resolver);
316 
317 	free(mtree->buff);
318 	free(mtree);
319 	(a->format->data) = NULL;
320 	return (ARCHIVE_OK);
321 }
322 
323 static ssize_t
get_line_size(const char * b,ssize_t avail,ssize_t * nlsize)324 get_line_size(const char *b, ssize_t avail, ssize_t *nlsize)
325 {
326 	ssize_t len;
327 
328 	len = 0;
329 	while (len < avail) {
330 		switch (*b) {
331 		case '\0':/* Non-ascii character or control character. */
332 			if (nlsize != NULL)
333 				*nlsize = 0;
334 			return (-1);
335 		case '\r':
336 			if (avail-len > 1 && b[1] == '\n') {
337 				if (nlsize != NULL)
338 					*nlsize = 2;
339 				return (len+2);
340 			}
341 			/* FALL THROUGH */
342 		case '\n':
343 			if (nlsize != NULL)
344 				*nlsize = 1;
345 			return (len+1);
346 		default:
347 			b++;
348 			len++;
349 			break;
350 		}
351 	}
352 	if (nlsize != NULL)
353 		*nlsize = 0;
354 	return (avail);
355 }
356 
357 /*
358  *  <---------------- ravail --------------------->
359  *  <-- diff ------> <---  avail ----------------->
360  *                   <---- len ----------->
361  * | Previous lines | line being parsed  nl extra |
362  *                  ^
363  *                  b
364  *
365  */
366 static ssize_t
next_line(struct archive_read * a,const char ** b,ssize_t * avail,ssize_t * ravail,ssize_t * nl)367 next_line(struct archive_read *a,
368     const char **b, ssize_t *avail, ssize_t *ravail, ssize_t *nl)
369 {
370 	ssize_t len;
371 	int quit;
372 
373 	quit = 0;
374 	if (*avail == 0) {
375 		*nl = 0;
376 		len = 0;
377 	} else
378 		len = get_line_size(*b, *avail, nl);
379 	/*
380 	 * Read bytes more while it does not reach the end of line.
381 	 */
382 	while (*nl == 0 && len == *avail && !quit) {
383 		ssize_t diff = *ravail - *avail;
384 		size_t nbytes_req = (*ravail+1023) & ~1023U;
385 		ssize_t tested;
386 
387 		/*
388 		 * Place an arbitrary limit on the line length.
389 		 * mtree is almost free-form input and without line length limits,
390 		 * it can consume a lot of memory.
391 		 */
392 		if (len >= MAX_LINE_LEN)
393 			return (-1);
394 
395 		/* Increase reading bytes if it is not enough to at least
396 		 * new two lines. */
397 		if (nbytes_req < (size_t)*ravail + 160)
398 			nbytes_req <<= 1;
399 
400 		*b = __archive_read_ahead(a, nbytes_req, avail);
401 		if (*b == NULL) {
402 			if (*ravail >= *avail)
403 				return (0);
404 			/* Reading bytes reaches the end of file. */
405 			*b = __archive_read_ahead(a, *avail, avail);
406 			quit = 1;
407 		}
408 		*ravail = *avail;
409 		*b += diff;
410 		*avail -= diff;
411 		tested = len;/* Skip some bytes we already determined. */
412 		len = get_line_size(*b + len, *avail - len, nl);
413 		if (len >= 0)
414 			len += tested;
415 	}
416 	return (len);
417 }
418 
419 /*
420  * Compare characters with an mtree keyword.
421  * Returns the length of an mtree keyword if matched.
422  * Returns 0 if not matched.
423  */
424 static int
bid_keycmp(const char * p,const char * key,ssize_t len)425 bid_keycmp(const char *p, const char *key, ssize_t len)
426 {
427 	int match_len = 0;
428 
429 	while (len > 0 && *p && *key) {
430 		if (*p == *key) {
431 			--len;
432 			++p;
433 			++key;
434 			++match_len;
435 			continue;
436 		}
437 		return (0);/* Not match */
438 	}
439 	if (*key != '\0')
440 		return (0);/* Not match */
441 
442 	/* A following character should be specified characters */
443 	if (p[0] == '=' || p[0] == ' ' || p[0] == '\t' ||
444 	    p[0] == '\n' || p[0] == '\r' ||
445 	   (p[0] == '\\' && (p[1] == '\n' || p[1] == '\r')))
446 		return (match_len);
447 	return (0);/* Not match */
448 }
449 
450 /*
451  * Test whether the characters 'p' has is mtree keyword.
452  * Returns the length of a detected keyword.
453  * Returns 0 if any keywords were not found.
454  */
455 static int
bid_keyword(const char * p,ssize_t len)456 bid_keyword(const char *p,  ssize_t len)
457 {
458 	static const char * const keys_c[] = {
459 		"content", "contents", "cksum", NULL
460 	};
461 	static const char * const keys_df[] = {
462 		"device", "flags", NULL
463 	};
464 	static const char * const keys_g[] = {
465 		"gid", "gname", NULL
466 	};
467 	static const char * const keys_il[] = {
468 		"ignore", "inode", "link", NULL
469 	};
470 	static const char * const keys_m[] = {
471 		"md5", "md5digest", "mode", NULL
472 	};
473 	static const char * const keys_no[] = {
474 		"nlink", "nochange", "optional", NULL
475 	};
476 	static const char * const keys_r[] = {
477 		"resdevice", "rmd160", "rmd160digest", NULL
478 	};
479 	static const char * const keys_s[] = {
480 		"sha1", "sha1digest",
481 		"sha256", "sha256digest",
482 		"sha384", "sha384digest",
483 		"sha512", "sha512digest",
484 		"size", NULL
485 	};
486 	static const char * const keys_t[] = {
487 		"tags", "time", "type", NULL
488 	};
489 	static const char * const keys_u[] = {
490 		"uid", "uname",	NULL
491 	};
492 	const char * const *keys;
493 	int i;
494 
495 	switch (*p) {
496 	case 'c': keys = keys_c; break;
497 	case 'd': case 'f': keys = keys_df; break;
498 	case 'g': keys = keys_g; break;
499 	case 'i': case 'l': keys = keys_il; break;
500 	case 'm': keys = keys_m; break;
501 	case 'n': case 'o': keys = keys_no; break;
502 	case 'r': keys = keys_r; break;
503 	case 's': keys = keys_s; break;
504 	case 't': keys = keys_t; break;
505 	case 'u': keys = keys_u; break;
506 	default: return (0);/* Unknown key */
507 	}
508 
509 	for (i = 0; keys[i] != NULL; i++) {
510 		int l = bid_keycmp(p, keys[i], len);
511 		if (l > 0)
512 			return (l);
513 	}
514 	return (0);/* Unknown key */
515 }
516 
517 /*
518  * Test whether there is a set of mtree keywords.
519  * Returns the number of keywords.
520  * Returns -1 if we got incorrect sequence.
521  * This function expects a set of "<space characters>keyword=value".
522  * When "unset" is specified, expects a set of "<space characters>keyword".
523  */
524 static int
bid_keyword_list(const char * p,ssize_t len,int unset,int last_is_path)525 bid_keyword_list(const char *p,  ssize_t len, int unset, int last_is_path)
526 {
527 	int l;
528 	int keycnt = 0;
529 
530 	while (len > 0 && *p) {
531 		int blank = 0;
532 
533 		/* Test whether there are blank characters in the line. */
534 		while (len >0 && (*p == ' ' || *p == '\t')) {
535 			++p;
536 			--len;
537 			blank = 1;
538 		}
539 		if (*p == '\n' || *p == '\r')
540 			break;
541 		if (p[0] == '\\' && (p[1] == '\n' || p[1] == '\r'))
542 			break;
543 		if (!blank && !last_is_path) /* No blank character. */
544 			return (-1);
545 		if (last_is_path && len == 0)
546 				return (keycnt);
547 
548 		if (unset) {
549 			l = bid_keycmp(p, "all", len);
550 			if (l > 0)
551 				return (1);
552 		}
553 		/* Test whether there is a correct key in the line. */
554 		l = bid_keyword(p, len);
555 		if (l == 0)
556 			return (-1);/* Unknown keyword was found. */
557 		p += l;
558 		len -= l;
559 		keycnt++;
560 
561 		/* Skip value */
562 		if (*p == '=') {
563 			int value = 0;
564 			++p;
565 			--len;
566 			while (len > 0 && *p != ' ' && *p != '\t') {
567 				++p;
568 				--len;
569 				value = 1;
570 			}
571 			/* A keyword should have a its value unless
572 			 * "/unset" operation. */
573 			if (!unset && value == 0)
574 				return (-1);
575 		}
576 	}
577 	return (keycnt);
578 }
579 
580 static int
bid_entry(const char * p,ssize_t len,ssize_t nl,int * last_is_path)581 bid_entry(const char *p, ssize_t len, ssize_t nl, int *last_is_path)
582 {
583 	int f = 0;
584 	static const unsigned char safe_char[256] = {
585 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 00 - 0F */
586 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 10 - 1F */
587 		/* !"$%&'()*+,-./  EXCLUSION:( )(#) */
588 		0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 20 - 2F */
589 		/* 0123456789:;<>?  EXCLUSION:(=) */
590 		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, /* 30 - 3F */
591 		/* @ABCDEFGHIJKLMNO */
592 		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 40 - 4F */
593 		/* PQRSTUVWXYZ[\]^_  */
594 		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 50 - 5F */
595 		/* `abcdefghijklmno */
596 		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 60 - 6F */
597 		/* pqrstuvwxyz{|}~ */
598 		1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, /* 70 - 7F */
599 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 80 - 8F */
600 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 90 - 9F */
601 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* A0 - AF */
602 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* B0 - BF */
603 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* C0 - CF */
604 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* D0 - DF */
605 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* E0 - EF */
606 		0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* F0 - FF */
607 	};
608 	ssize_t ll;
609 	const char *pp = p;
610 	const char * const pp_end = pp + len;
611 
612 	*last_is_path = 0;
613 	/*
614 	 * Skip the path-name which is quoted.
615 	 */
616 	for (;pp < pp_end; ++pp) {
617 		if (!safe_char[*(const unsigned char *)pp]) {
618 			if (*pp != ' ' && *pp != '\t' && *pp != '\r'
619 			    && *pp != '\n')
620 				f = 0;
621 			break;
622 		}
623 		f = 1;
624 	}
625 	ll = pp_end - pp;
626 
627 	/* If a path-name was not found at the first, try to check
628 	 * a mtree format(a.k.a form D) ``NetBSD's mtree -D'' creates,
629 	 * which places the path-name at the last. */
630 	if (f == 0) {
631 		const char *pb = p + len - nl;
632 		int name_len = 0;
633 		int slash;
634 
635 		/* The form D accepts only a single line for an entry. */
636 		if (pb-2 >= p &&
637 		    pb[-1] == '\\' && (pb[-2] == ' ' || pb[-2] == '\t'))
638 			return (-1);
639 		if (pb-1 >= p && pb[-1] == '\\')
640 			return (-1);
641 
642 		slash = 0;
643 		while (p <= --pb && *pb != ' ' && *pb != '\t') {
644 			if (!safe_char[*(const unsigned char *)pb])
645 				return (-1);
646 			name_len++;
647 			/* The pathname should have a slash in this
648 			 * format. */
649 			if (*pb == '/')
650 				slash = 1;
651 		}
652 		if (name_len == 0 || slash == 0)
653 			return (-1);
654 		/* If '/' is placed at the first in this field, this is not
655 		 * a valid filename. */
656 		if (pb[1] == '/')
657 			return (-1);
658 		ll = len - nl - name_len;
659 		pp = p;
660 		*last_is_path = 1;
661 	}
662 
663 	return (bid_keyword_list(pp, ll, 0, *last_is_path));
664 }
665 
666 #define MAX_BID_ENTRY	3
667 
668 static int
mtree_bid(struct archive_read * a,int best_bid)669 mtree_bid(struct archive_read *a, int best_bid)
670 {
671 	const char *signature = "#mtree";
672 	const char *p;
673 
674 	(void)best_bid; /* UNUSED */
675 
676 	/* Now let's look at the actual header and see if it matches. */
677 	p = __archive_read_ahead(a, strlen(signature), NULL);
678 	if (p == NULL)
679 		return (-1);
680 
681 	if (memcmp(p, signature, strlen(signature)) == 0)
682 		return (8 * (int)strlen(signature));
683 
684 	/*
685 	 * There is not a mtree signature. Let's try to detect mtree format.
686 	 */
687 	return (detect_form(a, NULL));
688 }
689 
690 static int
detect_form(struct archive_read * a,int * is_form_d)691 detect_form(struct archive_read *a, int *is_form_d)
692 {
693 	const char *p;
694 	ssize_t avail, ravail;
695 	ssize_t len, nl;
696 	int entry_cnt = 0, multiline = 0;
697 	int form_D = 0;/* The archive is generated by `NetBSD mtree -D'
698 			* (In this source we call it `form D') . */
699 
700 	if (is_form_d != NULL)
701 		*is_form_d = 0;
702 	p = __archive_read_ahead(a, 1, &avail);
703 	if (p == NULL)
704 		return (-1);
705 	ravail = avail;
706 	for (;;) {
707 		len = next_line(a, &p, &avail, &ravail, &nl);
708 		/* The terminal character of the line should be
709 		 * a new line character, '\r\n' or '\n'. */
710 		if (len <= 0 || nl == 0)
711 			break;
712 		if (!multiline) {
713 			/* Leading whitespace is never significant,
714 			 * ignore it. */
715 			while (len > 0 && (*p == ' ' || *p == '\t')) {
716 				++p;
717 				--avail;
718 				--len;
719 			}
720 			/* Skip comment or empty line. */
721 			if (p[0] == '#' || p[0] == '\n' || p[0] == '\r') {
722 				p += len;
723 				avail -= len;
724 				continue;
725 			}
726 		} else {
727 			/* A continuance line; the terminal
728 			 * character of previous line was '\' character. */
729 			if (bid_keyword_list(p, len, 0, 0) <= 0)
730 				break;
731 			if (p[len-nl-1] != '\\') {
732 				if (multiline == 1 &&
733 				    ++entry_cnt >= MAX_BID_ENTRY)
734 					break;
735 				multiline = 0;
736 			}
737 			p += len;
738 			avail -= len;
739 			continue;
740 		}
741 		if (p[0] != '/') {
742 			int last_is_path, keywords;
743 
744 			keywords = bid_entry(p, len, nl, &last_is_path);
745 			if (keywords >= 0) {
746 				if (form_D == 0) {
747 					if (last_is_path)
748 						form_D = 1;
749 					else if (keywords > 0)
750 						/* This line is not `form D'. */
751 						form_D = -1;
752 				} else if (form_D == 1) {
753 					if (!last_is_path && keywords > 0)
754 						/* This this is not `form D'
755 						 * and We cannot accept mixed
756 						 * format. */
757 						break;
758 				}
759 				if (!last_is_path && p[len-nl-1] == '\\')
760 					/* This line continues. */
761 					multiline = 1;
762 				else {
763 					/* We've got plenty of correct lines
764 					 * to assume that this file is an mtree
765 					 * format. */
766 					if (++entry_cnt >= MAX_BID_ENTRY)
767 						break;
768 				}
769 			} else
770 				break;
771 		} else if (len > 4 && strncmp(p, "/set", 4) == 0) {
772 			if (bid_keyword_list(p+4, len-4, 0, 0) <= 0)
773 				break;
774 			/* This line continues. */
775 			if (p[len-nl-1] == '\\')
776 				multiline = 2;
777 		} else if (len > 6 && strncmp(p, "/unset", 6) == 0) {
778 			if (bid_keyword_list(p+6, len-6, 1, 0) <= 0)
779 				break;
780 			/* This line continues. */
781 			if (p[len-nl-1] == '\\')
782 				multiline = 2;
783 		} else
784 			break;
785 
786 		/* Test next line. */
787 		p += len;
788 		avail -= len;
789 	}
790 	if (entry_cnt >= MAX_BID_ENTRY || (entry_cnt > 0 && len == 0)) {
791 		if (is_form_d != NULL) {
792 			if (form_D == 1)
793 				*is_form_d = 1;
794 		}
795 		return (32);
796 	}
797 
798 	return (0);
799 }
800 
801 /*
802  * The extended mtree format permits multiple lines specifying
803  * attributes for each file.  For those entries, only the last line
804  * is actually used.  Practically speaking, that means we have
805  * to read the entire mtree file into memory up front.
806  *
807  * The parsing is done in two steps.  First, it is decided if a line
808  * changes the global defaults and if it is, processed accordingly.
809  * Otherwise, the options of the line are merged with the current
810  * global options.
811  */
812 static int
add_option(struct archive_read * a,struct mtree_option ** global,const char * value,size_t len)813 add_option(struct archive_read *a, struct mtree_option **global,
814     const char *value, size_t len)
815 {
816 	struct mtree_option *opt;
817 
818 	if ((opt = malloc(sizeof(*opt))) == NULL) {
819 		archive_set_error(&a->archive, errno, "Can't allocate memory");
820 		return (ARCHIVE_FATAL);
821 	}
822 	if ((opt->value = malloc(len + 1)) == NULL) {
823 		free(opt);
824 		archive_set_error(&a->archive, errno, "Can't allocate memory");
825 		return (ARCHIVE_FATAL);
826 	}
827 	memcpy(opt->value, value, len);
828 	opt->value[len] = '\0';
829 	opt->next = *global;
830 	*global = opt;
831 	return (ARCHIVE_OK);
832 }
833 
834 static void
remove_option(struct mtree_option ** global,const char * value,size_t len)835 remove_option(struct mtree_option **global, const char *value, size_t len)
836 {
837 	struct mtree_option *iter, *last;
838 
839 	last = NULL;
840 	for (iter = *global; iter != NULL; last = iter, iter = iter->next) {
841 		if (strncmp(iter->value, value, len) == 0 &&
842 		    (iter->value[len] == '\0' ||
843 		     iter->value[len] == '='))
844 			break;
845 	}
846 	if (iter == NULL)
847 		return;
848 	if (last == NULL)
849 		*global = iter->next;
850 	else
851 		last->next = iter->next;
852 
853 	free(iter->value);
854 	free(iter);
855 }
856 
857 static int
process_global_set(struct archive_read * a,struct mtree_option ** global,const char * line)858 process_global_set(struct archive_read *a,
859     struct mtree_option **global, const char *line)
860 {
861 	const char *next, *eq;
862 	size_t len;
863 	int r;
864 
865 	line += 4;
866 	for (;;) {
867 		next = line + strspn(line, " \t\r\n");
868 		if (*next == '\0')
869 			return (ARCHIVE_OK);
870 		line = next;
871 		next = line + strcspn(line, " \t\r\n");
872 		eq = strchr(line, '=');
873 		if (eq > next)
874 			len = next - line;
875 		else
876 			len = eq - line;
877 
878 		remove_option(global, line, len);
879 		r = add_option(a, global, line, next - line);
880 		if (r != ARCHIVE_OK)
881 			return (r);
882 		line = next;
883 	}
884 }
885 
886 static int
process_global_unset(struct archive_read * a,struct mtree_option ** global,const char * line)887 process_global_unset(struct archive_read *a,
888     struct mtree_option **global, const char *line)
889 {
890 	const char *next;
891 	size_t len;
892 
893 	line += 6;
894 	if (strchr(line, '=') != NULL) {
895 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
896 		    "/unset shall not contain `='");
897 		return ARCHIVE_FATAL;
898 	}
899 
900 	for (;;) {
901 		next = line + strspn(line, " \t\r\n");
902 		if (*next == '\0')
903 			return (ARCHIVE_OK);
904 		line = next;
905 		len = strcspn(line, " \t\r\n");
906 
907 		if (len == 3 && strncmp(line, "all", 3) == 0) {
908 			free_options(*global);
909 			*global = NULL;
910 		} else {
911 			remove_option(global, line, len);
912 		}
913 
914 		line += len;
915 	}
916 }
917 
918 static int
process_add_entry(struct archive_read * a,struct mtree * mtree,struct mtree_option ** global,const char * line,ssize_t line_len,struct mtree_entry ** last_entry,int is_form_d)919 process_add_entry(struct archive_read *a, struct mtree *mtree,
920     struct mtree_option **global, const char *line, ssize_t line_len,
921     struct mtree_entry **last_entry, int is_form_d)
922 {
923 	struct mtree_entry *entry;
924 	struct mtree_option *iter;
925 	const char *next, *eq, *name, *end;
926 	size_t name_len, len;
927 	int r, i;
928 
929 	if ((entry = malloc(sizeof(*entry))) == NULL) {
930 		archive_set_error(&a->archive, errno, "Can't allocate memory");
931 		return (ARCHIVE_FATAL);
932 	}
933 	entry->next = NULL;
934 	entry->options = NULL;
935 	entry->name = NULL;
936 	entry->used = 0;
937 	entry->full = 0;
938 
939 	/* Add this entry to list. */
940 	if (*last_entry == NULL)
941 		mtree->entries = entry;
942 	else
943 		(*last_entry)->next = entry;
944 	*last_entry = entry;
945 
946 	if (is_form_d) {
947 		/* Filename is last item on line. */
948 		/* Adjust line_len to trim trailing whitespace */
949 		while (line_len > 0) {
950 			char last_character = line[line_len - 1];
951 			if (last_character == '\r'
952 			    || last_character == '\n'
953 			    || last_character == '\t'
954 			    || last_character == ' ') {
955 				line_len--;
956 			} else {
957 				break;
958 			}
959 		}
960 		/* Name starts after the last whitespace separator */
961 		name = line;
962 		for (i = 0; i < line_len; i++) {
963 			if (line[i] == '\r'
964 			    || line[i] == '\n'
965 			    || line[i] == '\t'
966 			    || line[i] == ' ') {
967 				name = line + i + 1;
968 			}
969 		}
970 		name_len = line + line_len - name;
971 		end = name;
972 	} else {
973 		/* Filename is first item on line */
974 		name_len = strcspn(line, " \t\r\n");
975 		name = line;
976 		line += name_len;
977 		end = line + line_len;
978 	}
979 	/* name/name_len is the name within the line. */
980 	/* line..end brackets the entire line except the name */
981 
982 	if ((entry->name = malloc(name_len + 1)) == NULL) {
983 		archive_set_error(&a->archive, errno, "Can't allocate memory");
984 		return (ARCHIVE_FATAL);
985 	}
986 
987 	memcpy(entry->name, name, name_len);
988 	entry->name[name_len] = '\0';
989 	parse_escapes(entry->name, entry);
990 
991 	entry->next_dup = NULL;
992 	if (entry->full) {
993 		if (!__archive_rb_tree_insert_node(&mtree->rbtree, &entry->rbnode)) {
994 			struct mtree_entry *alt;
995 			alt = (struct mtree_entry *)__archive_rb_tree_find_node(
996 			    &mtree->rbtree, entry->name);
997 			if (alt != NULL) {
998 				while (alt->next_dup)
999 					alt = alt->next_dup;
1000 				alt->next_dup = entry;
1001 			}
1002 		}
1003 	}
1004 
1005 	for (iter = *global; iter != NULL; iter = iter->next) {
1006 		r = add_option(a, &entry->options, iter->value,
1007 		    strlen(iter->value));
1008 		if (r != ARCHIVE_OK)
1009 			return (r);
1010 	}
1011 
1012 	for (;;) {
1013 		next = line + strspn(line, " \t\r\n");
1014 		if (*next == '\0')
1015 			return (ARCHIVE_OK);
1016 		if (next >= end)
1017 			return (ARCHIVE_OK);
1018 		line = next;
1019 		next = line + strcspn(line, " \t\r\n");
1020 		eq = strchr(line, '=');
1021 		if (eq == NULL || eq > next)
1022 			len = next - line;
1023 		else
1024 			len = eq - line;
1025 
1026 		remove_option(&entry->options, line, len);
1027 		r = add_option(a, &entry->options, line, next - line);
1028 		if (r != ARCHIVE_OK)
1029 			return (r);
1030 		line = next;
1031 	}
1032 }
1033 
1034 static int
read_mtree(struct archive_read * a,struct mtree * mtree)1035 read_mtree(struct archive_read *a, struct mtree *mtree)
1036 {
1037 	ssize_t len;
1038 	uintmax_t counter;
1039 	char *p, *s;
1040 	struct mtree_option *global;
1041 	struct mtree_entry *last_entry;
1042 	int r, is_form_d;
1043 
1044 	mtree->archive_format = ARCHIVE_FORMAT_MTREE;
1045 	mtree->archive_format_name = "mtree";
1046 
1047 	global = NULL;
1048 	last_entry = NULL;
1049 
1050 	(void)detect_form(a, &is_form_d);
1051 
1052 	for (counter = 1; ; ++counter) {
1053 		r = ARCHIVE_OK;
1054 		len = readline(a, mtree, &p, 65536);
1055 		if (len == 0) {
1056 			mtree->this_entry = mtree->entries;
1057 			free_options(global);
1058 			return (ARCHIVE_OK);
1059 		}
1060 		if (len < 0) {
1061 			free_options(global);
1062 			return ((int)len);
1063 		}
1064 		/* Leading whitespace is never significant, ignore it. */
1065 		while (*p == ' ' || *p == '\t') {
1066 			++p;
1067 			--len;
1068 		}
1069 		/* Skip content lines and blank lines. */
1070 		if (*p == '#')
1071 			continue;
1072 		if (*p == '\r' || *p == '\n' || *p == '\0')
1073 			continue;
1074 		/* Non-printable characters are not allowed */
1075 		for (s = p;s < p + len - 1; s++) {
1076 			if (!isprint((unsigned char)*s) && *s != '\t') {
1077 				archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1078 					"Non-printable character 0x%02X", (unsigned char)(*s));
1079 				r = ARCHIVE_FATAL;
1080 				break;
1081 			}
1082 		}
1083 		if (r != ARCHIVE_OK)
1084 			break;
1085 		if (*p != '/') {
1086 			r = process_add_entry(a, mtree, &global, p, len,
1087 			    &last_entry, is_form_d);
1088 		} else if (len > 4 && strncmp(p, "/set", 4) == 0) {
1089 			if (p[4] != ' ' && p[4] != '\t')
1090 				break;
1091 			r = process_global_set(a, &global, p);
1092 		} else if (len > 6 && strncmp(p, "/unset", 6) == 0) {
1093 			if (p[6] != ' ' && p[6] != '\t')
1094 				break;
1095 			r = process_global_unset(a, &global, p);
1096 		} else
1097 			break;
1098 
1099 		if (r != ARCHIVE_OK) {
1100 			free_options(global);
1101 			return r;
1102 		}
1103 	}
1104 
1105 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1106 	    "Can't parse line %ju", counter);
1107 	free_options(global);
1108 	return (ARCHIVE_FATAL);
1109 }
1110 
1111 /*
1112  * Read in the entire mtree file into memory on the first request.
1113  * Then use the next unused file to satisfy each header request.
1114  */
1115 static int
read_header(struct archive_read * a,struct archive_entry * entry)1116 read_header(struct archive_read *a, struct archive_entry *entry)
1117 {
1118 	struct mtree *mtree;
1119 	char *p;
1120 	int r, use_next;
1121 
1122 	mtree = (struct mtree *)(a->format->data);
1123 
1124 	if (mtree->fd >= 0) {
1125 		close(mtree->fd);
1126 		mtree->fd = -1;
1127 	}
1128 
1129 	if (mtree->entries == NULL) {
1130 		mtree->resolver = archive_entry_linkresolver_new();
1131 		if (mtree->resolver == NULL)
1132 			return ARCHIVE_FATAL;
1133 		archive_entry_linkresolver_set_strategy(mtree->resolver,
1134 		    ARCHIVE_FORMAT_MTREE);
1135 		r = read_mtree(a, mtree);
1136 		if (r != ARCHIVE_OK)
1137 			return (r);
1138 	}
1139 
1140 	a->archive.archive_format = mtree->archive_format;
1141 	a->archive.archive_format_name = mtree->archive_format_name;
1142 
1143 	for (;;) {
1144 		if (mtree->this_entry == NULL)
1145 			return (ARCHIVE_EOF);
1146 		if (strcmp(mtree->this_entry->name, "..") == 0) {
1147 			mtree->this_entry->used = 1;
1148 			if (archive_strlen(&mtree->current_dir) > 0) {
1149 				/* Roll back current path. */
1150 				p = mtree->current_dir.s
1151 				    + mtree->current_dir.length - 1;
1152 				while (p >= mtree->current_dir.s && *p != '/')
1153 					--p;
1154 				if (p >= mtree->current_dir.s)
1155 					--p;
1156 				mtree->current_dir.length
1157 				    = p - mtree->current_dir.s + 1;
1158 			}
1159 		}
1160 		if (!mtree->this_entry->used) {
1161 			use_next = 0;
1162 			r = parse_file(a, entry, mtree, mtree->this_entry,
1163 				&use_next);
1164 			if (use_next == 0)
1165 				return (r);
1166 		}
1167 		mtree->this_entry = mtree->this_entry->next;
1168 	}
1169 }
1170 
1171 /*
1172  * A single file can have multiple lines contribute specifications.
1173  * Parse as many lines as necessary, then pull additional information
1174  * from a backing file on disk as necessary.
1175  */
1176 static int
parse_file(struct archive_read * a,struct archive_entry * entry,struct mtree * mtree,struct mtree_entry * mentry,int * use_next)1177 parse_file(struct archive_read *a, struct archive_entry *entry,
1178     struct mtree *mtree, struct mtree_entry *mentry, int *use_next)
1179 {
1180 	const char *path;
1181 	la_seek_stat_t st_storage, *st;
1182 	struct mtree_entry *mp;
1183 	struct archive_entry *sparse_entry;
1184 	int r = ARCHIVE_OK, r1, parsed_kws;
1185 
1186 	mentry->used = 1;
1187 
1188 	/* Initialize reasonable defaults. */
1189 	archive_entry_set_filetype(entry, AE_IFREG);
1190 	archive_entry_set_size(entry, 0);
1191 	archive_string_empty(&mtree->contents_name);
1192 
1193 	/* Parse options from this line. */
1194 	parsed_kws = 0;
1195 	r = parse_line(a, entry, mtree, mentry, &parsed_kws);
1196 
1197 	if (mentry->full) {
1198 		archive_entry_copy_pathname(entry, mentry->name);
1199 		/*
1200 		 * "Full" entries are allowed to have multiple lines
1201 		 * and those lines aren't required to be adjacent.  We
1202 		 * don't support multiple lines for "relative" entries
1203 		 * nor do we make any attempt to merge data from
1204 		 * separate "relative" and "full" entries.  (Merging
1205 		 * "relative" and "full" entries would require dealing
1206 		 * with pathname canonicalization, which is a very
1207 		 * tricky subject.)
1208 		 */
1209 		mp = (struct mtree_entry *)__archive_rb_tree_find_node(
1210 		    &mtree->rbtree, mentry->name);
1211 		for (; mp; mp = mp->next_dup) {
1212 			if (mp->full && !mp->used) {
1213 				/* Later lines override earlier ones. */
1214 				mp->used = 1;
1215 				r1 = parse_line(a, entry, mtree, mp, &parsed_kws);
1216 				if (r1 < r)
1217 					r = r1;
1218 			}
1219 		}
1220 	} else {
1221 		/*
1222 		 * Relative entries require us to construct
1223 		 * the full path and possibly update the
1224 		 * current directory.
1225 		 */
1226 		size_t n = archive_strlen(&mtree->current_dir);
1227 		if (n > 0)
1228 			archive_strcat(&mtree->current_dir, "/");
1229 		archive_strcat(&mtree->current_dir, mentry->name);
1230 		archive_entry_copy_pathname(entry, mtree->current_dir.s);
1231 		if (archive_entry_filetype(entry) != AE_IFDIR)
1232 			mtree->current_dir.length = n;
1233 	}
1234 
1235 	if (mtree->checkfs) {
1236 		/*
1237 		 * Try to open and stat the file to get the real size
1238 		 * and other file info.  It would be nice to avoid
1239 		 * this here so that getting a listing of an mtree
1240 		 * wouldn't require opening every referenced contents
1241 		 * file.  But then we wouldn't know the actual
1242 		 * contents size, so I don't see a really viable way
1243 		 * around this.  (Also, we may want to someday pull
1244 		 * other unspecified info from the contents file on
1245 		 * disk.)
1246 		 */
1247 		mtree->fd = -1;
1248 		if (archive_strlen(&mtree->contents_name) > 0)
1249 			path = mtree->contents_name.s;
1250 		else
1251 			path = archive_entry_pathname(entry);
1252 
1253 		if (archive_entry_filetype(entry) == AE_IFREG ||
1254 				archive_entry_filetype(entry) == AE_IFDIR) {
1255 			mtree->fd = open(path, O_RDONLY | O_BINARY | O_CLOEXEC);
1256 			__archive_ensure_cloexec_flag(mtree->fd);
1257 			if (mtree->fd < 0 && (
1258 #if defined(_WIN32) && !defined(__CYGWIN__)
1259         /*
1260          * On Windows, attempting to open a file with an
1261          * invalid name result in EINVAL (Error 22)
1262          */
1263 				(errno != ENOENT && errno != EINVAL)
1264 #else
1265 				errno != ENOENT
1266 #endif
1267         || archive_strlen(&mtree->contents_name) > 0)) {
1268 				archive_set_error(&a->archive, errno,
1269 						"Can't open %s", path);
1270 				r = ARCHIVE_WARN;
1271 			}
1272 		}
1273 
1274 		st = &st_storage;
1275 		if (mtree->fd >= 0) {
1276 			if (la_seek_fstat(mtree->fd, st) == -1) {
1277 				archive_set_error(&a->archive, errno,
1278 						"Could not fstat %s", path);
1279 				r = ARCHIVE_WARN;
1280 				/* If we can't stat it, don't keep it open. */
1281 				close(mtree->fd);
1282 				mtree->fd = -1;
1283 				st = NULL;
1284 			}
1285 		}
1286 #ifdef HAVE_LSTAT
1287 		else if (lstat(path, st) == -1)
1288 #else
1289 		else if (la_seek_stat(path, st) == -1)
1290 #endif
1291 		{
1292 			st = NULL;
1293 		}
1294 
1295 		/*
1296 		 * Check for a mismatch between the type in the specification
1297 		 * and the type of the contents object on disk.
1298 		 */
1299 		if (st != NULL) {
1300 			if (((st->st_mode & S_IFMT) == S_IFREG &&
1301 			      archive_entry_filetype(entry) == AE_IFREG)
1302 #ifdef S_IFLNK
1303 			  ||((st->st_mode & S_IFMT) == S_IFLNK &&
1304 			      archive_entry_filetype(entry) == AE_IFLNK)
1305 #endif
1306 #ifdef S_IFSOCK
1307 			  ||((st->st_mode & S_IFSOCK) == S_IFSOCK &&
1308 			      archive_entry_filetype(entry) == AE_IFSOCK)
1309 #endif
1310 #ifdef S_IFCHR
1311 			  ||((st->st_mode & S_IFMT) == S_IFCHR &&
1312 			      archive_entry_filetype(entry) == AE_IFCHR)
1313 #endif
1314 #ifdef S_IFBLK
1315 			  ||((st->st_mode & S_IFMT) == S_IFBLK &&
1316 			      archive_entry_filetype(entry) == AE_IFBLK)
1317 #endif
1318 			  ||((st->st_mode & S_IFMT) == S_IFDIR &&
1319 			      archive_entry_filetype(entry) == AE_IFDIR)
1320 #ifdef S_IFIFO
1321 			  ||((st->st_mode & S_IFMT) == S_IFIFO &&
1322 			      archive_entry_filetype(entry) == AE_IFIFO)
1323 #endif
1324 			) {
1325 				/* Types match. */
1326 			} else {
1327 				/* Types don't match; bail out gracefully. */
1328 				if (mtree->fd >= 0)
1329 					close(mtree->fd);
1330 				mtree->fd = -1;
1331 				if (parsed_kws & MTREE_HAS_OPTIONAL) {
1332 					/* It's not an error for an optional
1333 					 * entry to not match disk. */
1334 					*use_next = 1;
1335 				} else if (r == ARCHIVE_OK) {
1336 					archive_set_error(&a->archive,
1337 					    ARCHIVE_ERRNO_MISC,
1338 					    "mtree specification has different"
1339 					    " type for %s",
1340 					    archive_entry_pathname(entry));
1341 					r = ARCHIVE_WARN;
1342 				}
1343 				return (r);
1344 			}
1345 		}
1346 
1347 		/*
1348 		 * If there is a contents file on disk, pick some of the
1349 		 * metadata from that file.  For most of these, we only
1350 		 * set it from the contents if it wasn't already parsed
1351 		 * from the specification.
1352 		 */
1353 		if (st != NULL) {
1354 			if (((parsed_kws & MTREE_HAS_DEVICE) == 0 ||
1355 				(parsed_kws & MTREE_HAS_NOCHANGE) != 0) &&
1356 				(archive_entry_filetype(entry) == AE_IFCHR ||
1357 				 archive_entry_filetype(entry) == AE_IFBLK))
1358 				archive_entry_set_rdev(entry, st->st_rdev);
1359 			if ((parsed_kws & (MTREE_HAS_GID | MTREE_HAS_GNAME))
1360 				== 0 ||
1361 			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1362 				archive_entry_set_gid(entry, st->st_gid);
1363 			if ((parsed_kws & (MTREE_HAS_UID | MTREE_HAS_UNAME))
1364 				== 0 ||
1365 			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1366 				archive_entry_set_uid(entry, st->st_uid);
1367 			if ((parsed_kws & MTREE_HAS_MTIME) == 0 ||
1368 			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0) {
1369 #if HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC
1370 				archive_entry_set_mtime(entry, st->st_mtime,
1371 						st->st_mtimespec.tv_nsec);
1372 #elif HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC
1373 				archive_entry_set_mtime(entry, st->st_mtime,
1374 						st->st_mtim.tv_nsec);
1375 #elif HAVE_STRUCT_STAT_ST_MTIME_N
1376 				archive_entry_set_mtime(entry, st->st_mtime,
1377 						st->st_mtime_n);
1378 #elif HAVE_STRUCT_STAT_ST_UMTIME
1379 				archive_entry_set_mtime(entry, st->st_mtime,
1380 						st->st_umtime*1000);
1381 #elif HAVE_STRUCT_STAT_ST_MTIME_USEC
1382 				archive_entry_set_mtime(entry, st->st_mtime,
1383 						st->st_mtime_usec*1000);
1384 #else
1385 				archive_entry_set_mtime(entry, st->st_mtime, 0);
1386 #endif
1387 			}
1388 			if ((parsed_kws & MTREE_HAS_NLINK) == 0 ||
1389 			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1390 				archive_entry_set_nlink(entry, st->st_nlink);
1391 			if ((parsed_kws & MTREE_HAS_PERM) == 0 ||
1392 			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1393 				archive_entry_set_perm(entry, st->st_mode);
1394 			if ((parsed_kws & MTREE_HAS_SIZE) == 0 ||
1395 			    (parsed_kws & MTREE_HAS_NOCHANGE) != 0)
1396 				archive_entry_set_size(entry, st->st_size);
1397 			archive_entry_set_ino(entry, st->st_ino);
1398 			archive_entry_set_dev(entry, st->st_dev);
1399 
1400 			archive_entry_linkify(mtree->resolver, &entry,
1401 				&sparse_entry);
1402 		} else if (parsed_kws & MTREE_HAS_OPTIONAL) {
1403 			/*
1404 			 * Couldn't open the entry, stat it or the on-disk type
1405 			 * didn't match.  If this entry is optional, just
1406 			 * ignore it and read the next header entry.
1407 			 */
1408 			*use_next = 1;
1409 			return ARCHIVE_OK;
1410 		}
1411 	}
1412 
1413 	mtree->cur_size = archive_entry_size(entry);
1414 	mtree->offset = 0;
1415 
1416 	return r;
1417 }
1418 
1419 /*
1420  * Each line contains a sequence of keywords.
1421  */
1422 static int
parse_line(struct archive_read * a,struct archive_entry * entry,struct mtree * mtree,struct mtree_entry * mp,int * parsed_kws)1423 parse_line(struct archive_read *a, struct archive_entry *entry,
1424     struct mtree *mtree, struct mtree_entry *mp, int *parsed_kws)
1425 {
1426 	struct mtree_option *iter;
1427 	int r = ARCHIVE_OK, r1;
1428 
1429 	for (iter = mp->options; iter != NULL; iter = iter->next) {
1430 		r1 = parse_keyword(a, mtree, entry, iter, parsed_kws);
1431 		if (r1 < r)
1432 			r = r1;
1433 	}
1434 	if (r == ARCHIVE_OK && (*parsed_kws & MTREE_HAS_TYPE) == 0) {
1435 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1436 		    "Missing type keyword in mtree specification");
1437 		return (ARCHIVE_WARN);
1438 	}
1439 	return (r);
1440 }
1441 
1442 /*
1443  * Device entries have one of the following forms:
1444  *  - raw dev_t
1445  *  - format,major,minor[,subdevice]
1446  * When parsing succeeded, `pdev' will contain the appropriate dev_t value.
1447  */
1448 
1449 /* strsep() is not in C90, but strcspn() is. */
1450 /* Taken from http://unixpapa.com/incnote/string.html */
1451 static char *
la_strsep(char ** sp,const char * sep)1452 la_strsep(char **sp, const char *sep)
1453 {
1454 	char *p, *s;
1455 	if (sp == NULL || *sp == NULL || **sp == '\0')
1456 		return(NULL);
1457 	s = *sp;
1458 	p = s + strcspn(s, sep);
1459 	if (*p != '\0')
1460 		*p++ = '\0';
1461 	*sp = p;
1462 	return(s);
1463 }
1464 
1465 static int
parse_device(dev_t * pdev,struct archive * a,char * val)1466 parse_device(dev_t *pdev, struct archive *a, char *val)
1467 {
1468 #define MAX_PACK_ARGS 3
1469 	unsigned long numbers[MAX_PACK_ARGS];
1470 	char *p, *dev;
1471 	int argc;
1472 	pack_t *pack;
1473 	dev_t result;
1474 	const char *error = NULL;
1475 
1476 	memset(pdev, 0, sizeof(*pdev));
1477 	if ((dev = strchr(val, ',')) != NULL) {
1478 		/*
1479 		 * Device's major/minor are given in a specified format.
1480 		 * Decode and pack it accordingly.
1481 		 */
1482 		*dev++ = '\0';
1483 		if ((pack = pack_find(val)) == NULL) {
1484 			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1485 			    "Unknown format `%s'", val);
1486 			return ARCHIVE_WARN;
1487 		}
1488 		argc = 0;
1489 		while ((p = la_strsep(&dev, ",")) != NULL) {
1490 			if (*p == '\0') {
1491 				archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1492 				    "Missing number");
1493 				return ARCHIVE_WARN;
1494 			}
1495 			if (argc >= MAX_PACK_ARGS) {
1496 				archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1497 				    "Too many arguments");
1498 				return ARCHIVE_WARN;
1499 			}
1500 			numbers[argc++] = (unsigned long)mtree_atol(&p, 0);
1501 		}
1502 		if (argc < 2) {
1503 			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1504 			    "Not enough arguments");
1505 			return ARCHIVE_WARN;
1506 		}
1507 		result = (*pack)(argc, numbers, &error);
1508 		if (error != NULL) {
1509 			archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
1510 			    "%s", error);
1511 			return ARCHIVE_WARN;
1512 		}
1513 	} else {
1514 		/* file system raw value. */
1515 		result = (dev_t)mtree_atol(&val, 0);
1516 	}
1517 	*pdev = result;
1518 	return ARCHIVE_OK;
1519 #undef MAX_PACK_ARGS
1520 }
1521 
1522 static int
parse_hex_nibble(char c)1523 parse_hex_nibble(char c)
1524 {
1525 	if (c >= '0' && c <= '9')
1526 		return c - '0';
1527 	if (c >= 'a' && c <= 'f')
1528 		return 10 + c - 'a';
1529 #if 0
1530 	/* XXX: Is uppercase something we should support? */
1531 	if (c >= 'A' && c <= 'F')
1532 		return 10 + c - 'A';
1533 #endif
1534 
1535 	return -1;
1536 }
1537 
1538 static int
parse_digest(struct archive_read * a,struct archive_entry * entry,const char * digest,int type)1539 parse_digest(struct archive_read *a, struct archive_entry *entry,
1540     const char *digest, int type)
1541 {
1542 	unsigned char digest_buf[64];
1543 	int high, low;
1544 	size_t i, j, len;
1545 
1546 	switch (type) {
1547 	case ARCHIVE_ENTRY_DIGEST_MD5:
1548 		len = sizeof(entry->digest.md5);
1549 		break;
1550 	case ARCHIVE_ENTRY_DIGEST_RMD160:
1551 		len = sizeof(entry->digest.rmd160);
1552 		break;
1553 	case ARCHIVE_ENTRY_DIGEST_SHA1:
1554 		len = sizeof(entry->digest.sha1);
1555 		break;
1556 	case ARCHIVE_ENTRY_DIGEST_SHA256:
1557 		len = sizeof(entry->digest.sha256);
1558 		break;
1559 	case ARCHIVE_ENTRY_DIGEST_SHA384:
1560 		len = sizeof(entry->digest.sha384);
1561 		break;
1562 	case ARCHIVE_ENTRY_DIGEST_SHA512:
1563 		len = sizeof(entry->digest.sha512);
1564 		break;
1565 	default:
1566 		archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
1567 			"Internal error: Unknown digest type");
1568 		return ARCHIVE_FATAL;
1569 	}
1570 
1571 	if (len > sizeof(digest_buf)) {
1572 		archive_set_error(&a->archive, ARCHIVE_ERRNO_PROGRAMMER,
1573 			"Internal error: Digest storage too large");
1574 		return ARCHIVE_FATAL;
1575 	}
1576 
1577 	len *= 2;
1578 
1579 	if (mtree_strnlen(digest, len+1) != len) {
1580 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1581 				  "incorrect digest length, ignoring");
1582 		return ARCHIVE_WARN;
1583 	}
1584 
1585 	for (i = 0, j = 0; i < len; i += 2, j++) {
1586 		high = parse_hex_nibble(digest[i]);
1587 		low = parse_hex_nibble(digest[i+1]);
1588 		if (high == -1 || low == -1) {
1589 			archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1590 					  "invalid digest data, ignoring");
1591 			return ARCHIVE_WARN;
1592 		}
1593 
1594 		digest_buf[j] = high << 4 | low;
1595 	}
1596 
1597 	return archive_entry_set_digest(entry, type, digest_buf);
1598 }
1599 
1600 /*
1601  * Parse a single keyword and its value.
1602  */
1603 static int
parse_keyword(struct archive_read * a,struct mtree * mtree,struct archive_entry * entry,struct mtree_option * opt,int * parsed_kws)1604 parse_keyword(struct archive_read *a, struct mtree *mtree,
1605     struct archive_entry *entry, struct mtree_option *opt, int *parsed_kws)
1606 {
1607 	char *val, *key;
1608 
1609 	key = opt->value;
1610 
1611 	if (*key == '\0')
1612 		return (ARCHIVE_OK);
1613 
1614 	if (strcmp(key, "nochange") == 0) {
1615 		*parsed_kws |= MTREE_HAS_NOCHANGE;
1616 		return (ARCHIVE_OK);
1617 	}
1618 	if (strcmp(key, "optional") == 0) {
1619 		*parsed_kws |= MTREE_HAS_OPTIONAL;
1620 		return (ARCHIVE_OK);
1621 	}
1622 	if (strcmp(key, "ignore") == 0) {
1623 		/*
1624 		 * The mtree processing is not recursive, so
1625 		 * recursion will only happen for explicitly listed
1626 		 * entries.
1627 		 */
1628 		return (ARCHIVE_OK);
1629 	}
1630 
1631 	val = strchr(key, '=');
1632 	if (val == NULL) {
1633 		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1634 		    "Malformed attribute \"%s\" (%d)", key, key[0]);
1635 		return (ARCHIVE_WARN);
1636 	}
1637 
1638 	*val = '\0';
1639 	++val;
1640 
1641 	switch (key[0]) {
1642 	case 'c':
1643 		if (strcmp(key, "content") == 0
1644 		    || strcmp(key, "contents") == 0) {
1645 			parse_escapes(val, NULL);
1646 			archive_strcpy(&mtree->contents_name, val);
1647 			return (ARCHIVE_OK);
1648 		}
1649 		if (strcmp(key, "cksum") == 0)
1650 			return (ARCHIVE_OK);
1651 		break;
1652 	case 'd':
1653 		if (strcmp(key, "device") == 0) {
1654 			/* stat(2) st_rdev field, e.g. the major/minor IDs
1655 			 * of a char/block special file */
1656 			int r;
1657 			dev_t dev;
1658 
1659 			*parsed_kws |= MTREE_HAS_DEVICE;
1660 			r = parse_device(&dev, &a->archive, val);
1661 			if (r == ARCHIVE_OK)
1662 				archive_entry_set_rdev(entry, dev);
1663 			return r;
1664 		}
1665 		break;
1666 	case 'f':
1667 		if (strcmp(key, "flags") == 0) {
1668 			*parsed_kws |= MTREE_HAS_FFLAGS;
1669 			archive_entry_copy_fflags_text(entry, val);
1670 			return (ARCHIVE_OK);
1671 		}
1672 		break;
1673 	case 'g':
1674 		if (strcmp(key, "gid") == 0) {
1675 			*parsed_kws |= MTREE_HAS_GID;
1676 			archive_entry_set_gid(entry, mtree_atol(&val, 10));
1677 			return (ARCHIVE_OK);
1678 		}
1679 		if (strcmp(key, "gname") == 0) {
1680 			*parsed_kws |= MTREE_HAS_GNAME;
1681 			archive_entry_copy_gname(entry, val);
1682 			return (ARCHIVE_OK);
1683 		}
1684 		break;
1685 	case 'i':
1686 		if (strcmp(key, "inode") == 0) {
1687 			archive_entry_set_ino(entry, mtree_atol(&val, 10));
1688 			return (ARCHIVE_OK);
1689 		}
1690 		break;
1691 	case 'l':
1692 		if (strcmp(key, "link") == 0) {
1693 			parse_escapes(val, NULL);
1694 			archive_entry_copy_symlink(entry, val);
1695 			return (ARCHIVE_OK);
1696 		}
1697 		break;
1698 	case 'm':
1699 		if (strcmp(key, "md5") == 0 || strcmp(key, "md5digest") == 0) {
1700 			return parse_digest(a, entry, val,
1701 			    ARCHIVE_ENTRY_DIGEST_MD5);
1702 		}
1703 		if (strcmp(key, "mode") == 0) {
1704 			if (val[0] < '0' || val[0] > '7') {
1705 				archive_set_error(&a->archive,
1706 				    ARCHIVE_ERRNO_FILE_FORMAT,
1707 				    "Symbolic or non-octal mode \"%s\" unsupported", val);
1708 				return (ARCHIVE_WARN);
1709 			}
1710 			*parsed_kws |= MTREE_HAS_PERM;
1711 			archive_entry_set_perm(entry, (mode_t)mtree_atol(&val, 8));
1712 			return (ARCHIVE_OK);
1713 		}
1714 		break;
1715 	case 'n':
1716 		if (strcmp(key, "nlink") == 0) {
1717 			*parsed_kws |= MTREE_HAS_NLINK;
1718 			archive_entry_set_nlink(entry,
1719 				(unsigned int)mtree_atol(&val, 10));
1720 			return (ARCHIVE_OK);
1721 		}
1722 		break;
1723 	case 'r':
1724 		if (strcmp(key, "resdevice") == 0) {
1725 			/* stat(2) st_dev field, e.g. the device ID where the
1726 			 * inode resides */
1727 			int r;
1728 			dev_t dev;
1729 
1730 			r = parse_device(&dev, &a->archive, val);
1731 			if (r == ARCHIVE_OK)
1732 				archive_entry_set_dev(entry, dev);
1733 			return r;
1734 		}
1735 		if (strcmp(key, "rmd160") == 0 ||
1736 		    strcmp(key, "rmd160digest") == 0) {
1737 			return parse_digest(a, entry, val,
1738 			    ARCHIVE_ENTRY_DIGEST_RMD160);
1739 		}
1740 		break;
1741 	case 's':
1742 		if (strcmp(key, "sha1") == 0 ||
1743 		    strcmp(key, "sha1digest") == 0) {
1744 			return parse_digest(a, entry, val,
1745 			    ARCHIVE_ENTRY_DIGEST_SHA1);
1746 		}
1747 		if (strcmp(key, "sha256") == 0 ||
1748 		    strcmp(key, "sha256digest") == 0) {
1749 			return parse_digest(a, entry, val,
1750 			    ARCHIVE_ENTRY_DIGEST_SHA256);
1751 		}
1752 		if (strcmp(key, "sha384") == 0 ||
1753 		    strcmp(key, "sha384digest") == 0) {
1754 			return parse_digest(a, entry, val,
1755 			    ARCHIVE_ENTRY_DIGEST_SHA384);
1756 		}
1757 		if (strcmp(key, "sha512") == 0 ||
1758 		    strcmp(key, "sha512digest") == 0) {
1759 			return parse_digest(a, entry, val,
1760 			    ARCHIVE_ENTRY_DIGEST_SHA512);
1761 		}
1762 		if (strcmp(key, "size") == 0) {
1763 			archive_entry_set_size(entry, mtree_atol(&val, 10));
1764 			return (ARCHIVE_OK);
1765 		}
1766 		break;
1767 	case 't':
1768 		if (strcmp(key, "tags") == 0) {
1769 			/*
1770 			 * Comma delimited list of tags.
1771 			 * Ignore the tags for now, but the interface
1772 			 * should be extended to allow inclusion/exclusion.
1773 			 */
1774 			return (ARCHIVE_OK);
1775 		}
1776 		if (strcmp(key, "time") == 0) {
1777 			int64_t m;
1778 			int64_t my_time_t_max = get_time_t_max();
1779 			int64_t my_time_t_min = get_time_t_min();
1780 			long ns = 0;
1781 
1782 			*parsed_kws |= MTREE_HAS_MTIME;
1783 			m = mtree_atol(&val, 10);
1784 			/* Replicate an old mtree bug:
1785 			 * 123456789.1 represents 123456789
1786 			 * seconds and 1 nanosecond. */
1787 			if (*val == '.') {
1788 				++val;
1789 				ns = (long)mtree_atol(&val, 10);
1790 				if (ns < 0)
1791 					ns = 0;
1792 				else if (ns > 999999999)
1793 					ns = 999999999;
1794 			}
1795 			if (m > my_time_t_max)
1796 				m = my_time_t_max;
1797 			else if (m < my_time_t_min)
1798 				m = my_time_t_min;
1799 			archive_entry_set_mtime(entry, (time_t)m, ns);
1800 			return (ARCHIVE_OK);
1801 		}
1802 		if (strcmp(key, "type") == 0) {
1803 			switch (val[0]) {
1804 			case 'b':
1805 				if (strcmp(val, "block") == 0) {
1806 					*parsed_kws |= MTREE_HAS_TYPE;
1807 					archive_entry_set_filetype(entry,
1808 						AE_IFBLK);
1809 					return (ARCHIVE_OK);
1810 				}
1811 				break;
1812 			case 'c':
1813 				if (strcmp(val, "char") == 0) {
1814 					*parsed_kws |= MTREE_HAS_TYPE;
1815 					archive_entry_set_filetype(entry,
1816 						AE_IFCHR);
1817 					return (ARCHIVE_OK);
1818 				}
1819 				break;
1820 			case 'd':
1821 				if (strcmp(val, "dir") == 0) {
1822 					*parsed_kws |= MTREE_HAS_TYPE;
1823 					archive_entry_set_filetype(entry,
1824 						AE_IFDIR);
1825 					return (ARCHIVE_OK);
1826 				}
1827 				break;
1828 			case 'f':
1829 				if (strcmp(val, "fifo") == 0) {
1830 					*parsed_kws |= MTREE_HAS_TYPE;
1831 					archive_entry_set_filetype(entry,
1832 						AE_IFIFO);
1833 					return (ARCHIVE_OK);
1834 				}
1835 				if (strcmp(val, "file") == 0) {
1836 					*parsed_kws |= MTREE_HAS_TYPE;
1837 					archive_entry_set_filetype(entry,
1838 						AE_IFREG);
1839 					return (ARCHIVE_OK);
1840 				}
1841 				break;
1842 			case 'l':
1843 				if (strcmp(val, "link") == 0) {
1844 					*parsed_kws |= MTREE_HAS_TYPE;
1845 					archive_entry_set_filetype(entry,
1846 						AE_IFLNK);
1847 					return (ARCHIVE_OK);
1848 				}
1849 				break;
1850 			default:
1851 				break;
1852 			}
1853 			archive_set_error(&a->archive,
1854 			    ARCHIVE_ERRNO_FILE_FORMAT,
1855 			    "Unrecognized file type \"%s\"; "
1856 			    "assuming \"file\"", val);
1857 			archive_entry_set_filetype(entry, AE_IFREG);
1858 			return (ARCHIVE_WARN);
1859 		}
1860 		break;
1861 	case 'u':
1862 		if (strcmp(key, "uid") == 0) {
1863 			*parsed_kws |= MTREE_HAS_UID;
1864 			archive_entry_set_uid(entry, mtree_atol(&val, 10));
1865 			return (ARCHIVE_OK);
1866 		}
1867 		if (strcmp(key, "uname") == 0) {
1868 			*parsed_kws |= MTREE_HAS_UNAME;
1869 			archive_entry_copy_uname(entry, val);
1870 			return (ARCHIVE_OK);
1871 		}
1872 		break;
1873 	default:
1874 		break;
1875 	}
1876 	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
1877 	    "Unrecognized key %s=%s", key, val);
1878 	return (ARCHIVE_WARN);
1879 }
1880 
1881 static int
read_data(struct archive_read * a,const void ** buff,size_t * size,int64_t * offset)1882 read_data(struct archive_read *a, const void **buff, size_t *size,
1883     int64_t *offset)
1884 {
1885 	size_t bytes_to_read;
1886 	ssize_t bytes_read;
1887 	struct mtree *mtree;
1888 
1889 	mtree = (struct mtree *)(a->format->data);
1890 	if (mtree->fd < 0) {
1891 		*buff = NULL;
1892 		*offset = 0;
1893 		*size = 0;
1894 		return (ARCHIVE_EOF);
1895 	}
1896 	if (mtree->buff == NULL) {
1897 		mtree->buffsize = 64 * 1024;
1898 		mtree->buff = malloc(mtree->buffsize);
1899 		if (mtree->buff == NULL) {
1900 			archive_set_error(&a->archive, ENOMEM,
1901 			    "Can't allocate memory");
1902 			return (ARCHIVE_FATAL);
1903 		}
1904 	}
1905 
1906 	*buff = mtree->buff;
1907 	*offset = mtree->offset;
1908 	if ((int64_t)mtree->buffsize > mtree->cur_size - mtree->offset)
1909 		bytes_to_read = (size_t)(mtree->cur_size - mtree->offset);
1910 	else
1911 		bytes_to_read = mtree->buffsize;
1912 	bytes_read = read(mtree->fd, mtree->buff, bytes_to_read);
1913 	if (bytes_read < 0) {
1914 		archive_set_error(&a->archive, errno, "Can't read");
1915 		return (ARCHIVE_WARN);
1916 	}
1917 	if (bytes_read == 0) {
1918 		*size = 0;
1919 		return (ARCHIVE_EOF);
1920 	}
1921 	mtree->offset += bytes_read;
1922 	*size = bytes_read;
1923 	return (ARCHIVE_OK);
1924 }
1925 
1926 /* Skip does nothing except possibly close the contents file. */
1927 static int
skip(struct archive_read * a)1928 skip(struct archive_read *a)
1929 {
1930 	struct mtree *mtree;
1931 
1932 	mtree = (struct mtree *)(a->format->data);
1933 	if (mtree->fd >= 0) {
1934 		close(mtree->fd);
1935 		mtree->fd = -1;
1936 	}
1937 	return (ARCHIVE_OK);
1938 }
1939 
1940 /*
1941  * Since parsing backslash sequences always makes strings shorter,
1942  * we can always do this conversion in-place.
1943  */
1944 static void
parse_escapes(char * src,struct mtree_entry * mentry)1945 parse_escapes(char *src, struct mtree_entry *mentry)
1946 {
1947 	char *dest = src;
1948 	char c;
1949 
1950 	if (mentry != NULL && strcmp(src, ".") == 0)
1951 		mentry->full = 1;
1952 
1953 	while (*src != '\0') {
1954 		c = *src++;
1955 		if (c == '/' && mentry != NULL)
1956 			mentry->full = 1;
1957 		if (c == '\\') {
1958 			switch (src[0]) {
1959 			case '0':
1960 				if (src[1] < '0' || src[1] > '7') {
1961 					c = 0;
1962 					++src;
1963 					break;
1964 				}
1965 				/* FALLTHROUGH */
1966 			case '1':
1967 			case '2':
1968 			case '3':
1969 				if (src[1] >= '0' && src[1] <= '7' &&
1970 				    src[2] >= '0' && src[2] <= '7') {
1971 					c = (src[0] - '0') << 6;
1972 					c |= (src[1] - '0') << 3;
1973 					c |= (src[2] - '0');
1974 					src += 3;
1975 				}
1976 				break;
1977 			case 'a':
1978 				c = '\a';
1979 				++src;
1980 				break;
1981 			case 'b':
1982 				c = '\b';
1983 				++src;
1984 				break;
1985 			case 'f':
1986 				c = '\f';
1987 				++src;
1988 				break;
1989 			case 'n':
1990 				c = '\n';
1991 				++src;
1992 				break;
1993 			case 'r':
1994 				c = '\r';
1995 				++src;
1996 				break;
1997 			case 's':
1998 				c = ' ';
1999 				++src;
2000 				break;
2001 			case 't':
2002 				c = '\t';
2003 				++src;
2004 				break;
2005 			case 'v':
2006 				c = '\v';
2007 				++src;
2008 				break;
2009 			case '\\':
2010 				c = '\\';
2011 				++src;
2012 				break;
2013 			}
2014 		}
2015 		*dest++ = c;
2016 	}
2017 	*dest = '\0';
2018 }
2019 
2020 /* Parse a hex digit. */
2021 static int
parsedigit(char c)2022 parsedigit(char c)
2023 {
2024 	if (c >= '0' && c <= '9')
2025 		return c - '0';
2026 	else if (c >= 'a' && c <= 'f')
2027 		return c - 'a';
2028 	else if (c >= 'A' && c <= 'F')
2029 		return c - 'A';
2030 	else
2031 		return -1;
2032 }
2033 
2034 /*
2035  * Note that this implementation does not (and should not!) obey
2036  * locale settings; you cannot simply substitute strtol here, since
2037  * it does obey locale.
2038  */
2039 static int64_t
mtree_atol(char ** p,int base)2040 mtree_atol(char **p, int base)
2041 {
2042 	int64_t l, limit;
2043 	int digit, last_digit_limit;
2044 
2045 	if (base == 0) {
2046 		if (**p != '0')
2047 			base = 10;
2048 		else if ((*p)[1] == 'x' || (*p)[1] == 'X') {
2049 			*p += 2;
2050 			base = 16;
2051 		} else {
2052 			base = 8;
2053 		}
2054 	}
2055 
2056 	if (**p == '-') {
2057 		limit = INT64_MIN / base;
2058 		last_digit_limit = -(INT64_MIN % base);
2059 		++(*p);
2060 
2061 		l = 0;
2062 		digit = parsedigit(**p);
2063 		while (digit >= 0 && digit < base) {
2064 			if (l < limit || (l == limit && digit >= last_digit_limit))
2065 				return INT64_MIN;
2066 			l = (l * base) - digit;
2067 			digit = parsedigit(*++(*p));
2068 		}
2069 		return l;
2070 	} else {
2071 		limit = INT64_MAX / base;
2072 		last_digit_limit = INT64_MAX % base;
2073 
2074 		l = 0;
2075 		digit = parsedigit(**p);
2076 		while (digit >= 0 && digit < base) {
2077 			if (l > limit || (l == limit && digit > last_digit_limit))
2078 				return INT64_MAX;
2079 			l = (l * base) + digit;
2080 			digit = parsedigit(*++(*p));
2081 		}
2082 		return l;
2083 	}
2084 }
2085 
2086 /*
2087  * Returns length of line (including trailing newline)
2088  * or negative on error.  'start' argument is updated to
2089  * point to first character of line.
2090  */
2091 static ssize_t
readline(struct archive_read * a,struct mtree * mtree,char ** start,ssize_t limit)2092 readline(struct archive_read *a, struct mtree *mtree, char **start,
2093     ssize_t limit)
2094 {
2095 	ssize_t bytes_read;
2096 	ssize_t total_size = 0;
2097 	ssize_t find_off = 0;
2098 	const void *t;
2099 	void *nl;
2100 	char *u;
2101 
2102 	/* Accumulate line in a line buffer. */
2103 	for (;;) {
2104 		/* Read some more. */
2105 		t = __archive_read_ahead(a, 1, &bytes_read);
2106 		if (t == NULL)
2107 			return (0);
2108 		if (bytes_read < 0)
2109 			return (ARCHIVE_FATAL);
2110 		nl = memchr(t, '\n', bytes_read);
2111 		/* If we found '\n', trim the read to end exactly there. */
2112 		if (nl != NULL) {
2113 			bytes_read = ((const char *)nl) - ((const char *)t) + 1;
2114 		}
2115 		if (total_size + bytes_read + 1 > limit) {
2116 			archive_set_error(&a->archive,
2117 			    ARCHIVE_ERRNO_FILE_FORMAT,
2118 			    "Line too long");
2119 			return (ARCHIVE_FATAL);
2120 		}
2121 		if (archive_string_ensure(&mtree->line,
2122 			total_size + bytes_read + 1) == NULL) {
2123 			archive_set_error(&a->archive, ENOMEM,
2124 			    "Can't allocate working buffer");
2125 			return (ARCHIVE_FATAL);
2126 		}
2127 		/* Append new bytes to string. */
2128 		memcpy(mtree->line.s + total_size, t, bytes_read);
2129 		__archive_read_consume(a, bytes_read);
2130 		total_size += bytes_read;
2131 		mtree->line.s[total_size] = '\0';
2132 
2133 		for (u = mtree->line.s + find_off; *u; ++u) {
2134 			if (u[0] == '\n') {
2135 				/* Ends with unescaped newline. */
2136 				/* Check if preceded by '\r' for CRLF handling */
2137 				if (u > mtree->line.s && u[-1] == '\r') {
2138 					/* CRLF ending - remove the '\r' */
2139 					u[-1] = '\n';
2140 					u[0] = '\0';
2141 					total_size--;
2142 				}
2143 				*start = mtree->line.s;
2144 				return total_size;
2145 			} else if (u[0] == '#') {
2146 				/* Ends with comment sequence #...\n */
2147 				if (nl == NULL) {
2148 					/* But we've not found the \n yet */
2149 					break;
2150 				}
2151 			} else if (u[0] == '\\') {
2152 				if (u[1] == '\n') {
2153 					/* Trim escaped newline. */
2154 					total_size -= 2;
2155 					mtree->line.s[total_size] = '\0';
2156 					break;
2157 				} else if (u[1] == '\r' && u[2] == '\n') {
2158 					/* Trim escaped CRLF. */
2159 					total_size -= 3;
2160 					mtree->line.s[total_size] = '\0';
2161 					break;
2162 				} else if (u[1] != '\0') {
2163 					/* Skip the two-char escape sequence */
2164 					++u;
2165 				}
2166 			}
2167 		}
2168 		find_off = u - mtree->line.s;
2169 	}
2170 }
2171