1/*-
2 * Copyright (c) 2003-2007 Tim Kientzle
3 * Copyright (c) 2008 Joerg Sonnenberger
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 */
26
27#include "archive_platform.h"
28__FBSDID("$FreeBSD: head/lib/libarchive/archive_read_support_format_mtree.c 201165 2009-12-29 05:52:13Z kientzle $");
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
48#include "archive.h"
49#include "archive_entry.h"
50#include "archive_private.h"
51#include "archive_read_private.h"
52#include "archive_string.h"
53
54#ifndef O_BINARY
55#define	O_BINARY 0
56#endif
57
58#define	MTREE_HAS_DEVICE	0x0001
59#define	MTREE_HAS_FFLAGS	0x0002
60#define	MTREE_HAS_GID		0x0004
61#define	MTREE_HAS_GNAME		0x0008
62#define	MTREE_HAS_MTIME		0x0010
63#define	MTREE_HAS_NLINK		0x0020
64#define	MTREE_HAS_PERM		0x0040
65#define	MTREE_HAS_SIZE		0x0080
66#define	MTREE_HAS_TYPE		0x0100
67#define	MTREE_HAS_UID		0x0200
68#define	MTREE_HAS_UNAME		0x0400
69
70#define	MTREE_HAS_OPTIONAL	0x0800
71
72struct mtree_option {
73	struct mtree_option *next;
74	char *value;
75};
76
77struct mtree_entry {
78	struct mtree_entry *next;
79	struct mtree_option *options;
80	char *name;
81	char full;
82	char used;
83};
84
85struct mtree {
86	struct archive_string	 line;
87	size_t			 buffsize;
88	char			*buff;
89	off_t			 offset;
90	int			 fd;
91	int			 filetype;
92	int			 archive_format;
93	const char		*archive_format_name;
94	struct mtree_entry	*entries;
95	struct mtree_entry	*this_entry;
96	struct archive_string	 current_dir;
97	struct archive_string	 contents_name;
98
99	struct archive_entry_linkresolver *resolver;
100
101	off_t			 cur_size, cur_offset;
102};
103
104static int	cleanup(struct archive_read *);
105static int	mtree_bid(struct archive_read *);
106static int	parse_file(struct archive_read *, struct archive_entry *,
107		    struct mtree *, struct mtree_entry *, int *);
108static void	parse_escapes(char *, struct mtree_entry *);
109static int	parse_line(struct archive_read *, struct archive_entry *,
110		    struct mtree *, struct mtree_entry *, int *);
111static int	parse_keyword(struct archive_read *, struct mtree *,
112		    struct archive_entry *, struct mtree_option *, int *);
113static int	read_data(struct archive_read *a,
114		    const void **buff, size_t *size, off_t *offset);
115static ssize_t	readline(struct archive_read *, struct mtree *, char **, ssize_t);
116static int	skip(struct archive_read *a);
117static int	read_header(struct archive_read *,
118		    struct archive_entry *);
119static int64_t	mtree_atol10(char **);
120static int64_t	mtree_atol8(char **);
121static int64_t	mtree_atol(char **);
122
123static void
124free_options(struct mtree_option *head)
125{
126	struct mtree_option *next;
127
128	for (; head != NULL; head = next) {
129		next = head->next;
130		free(head->value);
131		free(head);
132	}
133}
134
135int
136archive_read_support_format_mtree(struct archive *_a)
137{
138	struct archive_read *a = (struct archive_read *)_a;
139	struct mtree *mtree;
140	int r;
141
142	mtree = (struct mtree *)malloc(sizeof(*mtree));
143	if (mtree == NULL) {
144		archive_set_error(&a->archive, ENOMEM,
145		    "Can't allocate mtree data");
146		return (ARCHIVE_FATAL);
147	}
148	memset(mtree, 0, sizeof(*mtree));
149	mtree->fd = -1;
150
151	r = __archive_read_register_format(a, mtree, "mtree",
152	    mtree_bid, NULL, read_header, read_data, skip, cleanup);
153
154	if (r != ARCHIVE_OK)
155		free(mtree);
156	return (ARCHIVE_OK);
157}
158
159static int
160cleanup(struct archive_read *a)
161{
162	struct mtree *mtree;
163	struct mtree_entry *p, *q;
164
165	mtree = (struct mtree *)(a->format->data);
166
167	p = mtree->entries;
168	while (p != NULL) {
169		q = p->next;
170		free(p->name);
171		free_options(p->options);
172		free(p);
173		p = q;
174	}
175	archive_string_free(&mtree->line);
176	archive_string_free(&mtree->current_dir);
177	archive_string_free(&mtree->contents_name);
178	archive_entry_linkresolver_free(mtree->resolver);
179
180	free(mtree->buff);
181	free(mtree);
182	(a->format->data) = NULL;
183	return (ARCHIVE_OK);
184}
185
186
187static int
188mtree_bid(struct archive_read *a)
189{
190	const char *signature = "#mtree";
191	const char *p;
192
193	/* Now let's look at the actual header and see if it matches. */
194	p = __archive_read_ahead(a, strlen(signature), NULL);
195	if (p == NULL)
196		return (-1);
197
198	if (strncmp(p, signature, strlen(signature)) == 0)
199		return (8 * (int)strlen(signature));
200	return (0);
201}
202
203/*
204 * The extended mtree format permits multiple lines specifying
205 * attributes for each file.  For those entries, only the last line
206 * is actually used.  Practically speaking, that means we have
207 * to read the entire mtree file into memory up front.
208 *
209 * The parsing is done in two steps.  First, it is decided if a line
210 * changes the global defaults and if it is, processed accordingly.
211 * Otherwise, the options of the line are merged with the current
212 * global options.
213 */
214static int
215add_option(struct archive_read *a, struct mtree_option **global,
216    const char *value, size_t len)
217{
218	struct mtree_option *option;
219
220	if ((option = malloc(sizeof(*option))) == NULL) {
221		archive_set_error(&a->archive, errno, "Can't allocate memory");
222		return (ARCHIVE_FATAL);
223	}
224	if ((option->value = malloc(len + 1)) == NULL) {
225		free(option);
226		archive_set_error(&a->archive, errno, "Can't allocate memory");
227		return (ARCHIVE_FATAL);
228	}
229	memcpy(option->value, value, len);
230	option->value[len] = '\0';
231	option->next = *global;
232	*global = option;
233	return (ARCHIVE_OK);
234}
235
236static void
237remove_option(struct mtree_option **global, const char *value, size_t len)
238{
239	struct mtree_option *iter, *last;
240
241	last = NULL;
242	for (iter = *global; iter != NULL; last = iter, iter = iter->next) {
243		if (strncmp(iter->value, value, len) == 0 &&
244		    (iter->value[len] == '\0' ||
245		     iter->value[len] == '='))
246			break;
247	}
248	if (iter == NULL)
249		return;
250	if (last == NULL)
251		*global = iter->next;
252	else
253		last->next = iter->next;
254
255	free(iter->value);
256	free(iter);
257}
258
259static int
260process_global_set(struct archive_read *a,
261    struct mtree_option **global, const char *line)
262{
263	const char *next, *eq;
264	size_t len;
265	int r;
266
267	line += 4;
268	for (;;) {
269		next = line + strspn(line, " \t\r\n");
270		if (*next == '\0')
271			return (ARCHIVE_OK);
272		line = next;
273		next = line + strcspn(line, " \t\r\n");
274		eq = strchr(line, '=');
275		if (eq > next)
276			len = next - line;
277		else
278			len = eq - line;
279
280		remove_option(global, line, len);
281		r = add_option(a, global, line, next - line);
282		if (r != ARCHIVE_OK)
283			return (r);
284		line = next;
285	}
286}
287
288static int
289process_global_unset(struct archive_read *a,
290    struct mtree_option **global, const char *line)
291{
292	const char *next;
293	size_t len;
294
295	line += 6;
296	if (strchr(line, '=') != NULL) {
297		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
298		    "/unset shall not contain `='");
299		return ARCHIVE_FATAL;
300	}
301
302	for (;;) {
303		next = line + strspn(line, " \t\r\n");
304		if (*next == '\0')
305			return (ARCHIVE_OK);
306		line = next;
307		len = strcspn(line, " \t\r\n");
308
309		if (len == 3 && strncmp(line, "all", 3) == 0) {
310			free_options(*global);
311			*global = NULL;
312		} else {
313			remove_option(global, line, len);
314		}
315
316		line += len;
317	}
318}
319
320static int
321process_add_entry(struct archive_read *a, struct mtree *mtree,
322    struct mtree_option **global, const char *line,
323    struct mtree_entry **last_entry)
324{
325	struct mtree_entry *entry;
326	struct mtree_option *iter;
327	const char *next, *eq;
328	size_t len;
329	int r;
330
331	if ((entry = malloc(sizeof(*entry))) == NULL) {
332		archive_set_error(&a->archive, errno, "Can't allocate memory");
333		return (ARCHIVE_FATAL);
334	}
335	entry->next = NULL;
336	entry->options = NULL;
337	entry->name = NULL;
338	entry->used = 0;
339	entry->full = 0;
340
341	/* Add this entry to list. */
342	if (*last_entry == NULL)
343		mtree->entries = entry;
344	else
345		(*last_entry)->next = entry;
346	*last_entry = entry;
347
348	len = strcspn(line, " \t\r\n");
349	if ((entry->name = malloc(len + 1)) == NULL) {
350		archive_set_error(&a->archive, errno, "Can't allocate memory");
351		return (ARCHIVE_FATAL);
352	}
353
354	memcpy(entry->name, line, len);
355	entry->name[len] = '\0';
356	parse_escapes(entry->name, entry);
357
358	line += len;
359	for (iter = *global; iter != NULL; iter = iter->next) {
360		r = add_option(a, &entry->options, iter->value,
361		    strlen(iter->value));
362		if (r != ARCHIVE_OK)
363			return (r);
364	}
365
366	for (;;) {
367		next = line + strspn(line, " \t\r\n");
368		if (*next == '\0')
369			return (ARCHIVE_OK);
370		line = next;
371		next = line + strcspn(line, " \t\r\n");
372		eq = strchr(line, '=');
373		if (eq > next)
374			len = next - line;
375		else
376			len = eq - line;
377
378		remove_option(&entry->options, line, len);
379		r = add_option(a, &entry->options, line, next - line);
380		if (r != ARCHIVE_OK)
381			return (r);
382		line = next;
383	}
384}
385
386static int
387read_mtree(struct archive_read *a, struct mtree *mtree)
388{
389	ssize_t len;
390	uintmax_t counter;
391	char *p;
392	struct mtree_option *global;
393	struct mtree_entry *last_entry;
394	int r;
395
396	mtree->archive_format = ARCHIVE_FORMAT_MTREE;
397	mtree->archive_format_name = "mtree";
398
399	global = NULL;
400	last_entry = NULL;
401
402	for (counter = 1; ; ++counter) {
403		len = readline(a, mtree, &p, 256);
404		if (len == 0) {
405			mtree->this_entry = mtree->entries;
406			free_options(global);
407			return (ARCHIVE_OK);
408		}
409		if (len < 0) {
410			free_options(global);
411			return (len);
412		}
413		/* Leading whitespace is never significant, ignore it. */
414		while (*p == ' ' || *p == '\t') {
415			++p;
416			--len;
417		}
418		/* Skip content lines and blank lines. */
419		if (*p == '#')
420			continue;
421		if (*p == '\r' || *p == '\n' || *p == '\0')
422			continue;
423		if (*p != '/') {
424			r = process_add_entry(a, mtree, &global, p,
425			    &last_entry);
426		} else if (strncmp(p, "/set", 4) == 0) {
427			if (p[4] != ' ' && p[4] != '\t')
428				break;
429			r = process_global_set(a, &global, p);
430		} else if (strncmp(p, "/unset", 6) == 0) {
431			if (p[6] != ' ' && p[6] != '\t')
432				break;
433			r = process_global_unset(a, &global, p);
434		} else
435			break;
436
437		if (r != ARCHIVE_OK) {
438			free_options(global);
439			return r;
440		}
441	}
442
443	archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
444	    "Can't parse line %ju", counter);
445	free_options(global);
446	return (ARCHIVE_FATAL);
447}
448
449/*
450 * Read in the entire mtree file into memory on the first request.
451 * Then use the next unused file to satisfy each header request.
452 */
453static int
454read_header(struct archive_read *a, struct archive_entry *entry)
455{
456	struct mtree *mtree;
457	char *p;
458	int r, use_next;
459
460	mtree = (struct mtree *)(a->format->data);
461
462	if (mtree->fd >= 0) {
463		close(mtree->fd);
464		mtree->fd = -1;
465	}
466
467	if (mtree->entries == NULL) {
468		mtree->resolver = archive_entry_linkresolver_new();
469		if (mtree->resolver == NULL)
470			return ARCHIVE_FATAL;
471		archive_entry_linkresolver_set_strategy(mtree->resolver,
472		    ARCHIVE_FORMAT_MTREE);
473		r = read_mtree(a, mtree);
474		if (r != ARCHIVE_OK)
475			return (r);
476	}
477
478	a->archive.archive_format = mtree->archive_format;
479	a->archive.archive_format_name = mtree->archive_format_name;
480
481	for (;;) {
482		if (mtree->this_entry == NULL)
483			return (ARCHIVE_EOF);
484		if (strcmp(mtree->this_entry->name, "..") == 0) {
485			mtree->this_entry->used = 1;
486			if (archive_strlen(&mtree->current_dir) > 0) {
487				/* Roll back current path. */
488				p = mtree->current_dir.s
489				    + mtree->current_dir.length - 1;
490				while (p >= mtree->current_dir.s && *p != '/')
491					--p;
492				if (p >= mtree->current_dir.s)
493					--p;
494				mtree->current_dir.length
495				    = p - mtree->current_dir.s + 1;
496			}
497		}
498		if (!mtree->this_entry->used) {
499			use_next = 0;
500			r = parse_file(a, entry, mtree, mtree->this_entry, &use_next);
501			if (use_next == 0)
502				return (r);
503		}
504		mtree->this_entry = mtree->this_entry->next;
505	}
506}
507
508/*
509 * A single file can have multiple lines contribute specifications.
510 * Parse as many lines as necessary, then pull additional information
511 * from a backing file on disk as necessary.
512 */
513static int
514parse_file(struct archive_read *a, struct archive_entry *entry,
515    struct mtree *mtree, struct mtree_entry *mentry, int *use_next)
516{
517	const char *path;
518	struct stat st_storage, *st;
519	struct mtree_entry *mp;
520	struct archive_entry *sparse_entry;
521	int r = ARCHIVE_OK, r1, parsed_kws, mismatched_type;
522
523	mentry->used = 1;
524
525	/* Initialize reasonable defaults. */
526	mtree->filetype = AE_IFREG;
527	archive_entry_set_size(entry, 0);
528
529	/* Parse options from this line. */
530	parsed_kws = 0;
531	r = parse_line(a, entry, mtree, mentry, &parsed_kws);
532
533	if (mentry->full) {
534		archive_entry_copy_pathname(entry, mentry->name);
535		/*
536		 * "Full" entries are allowed to have multiple lines
537		 * and those lines aren't required to be adjacent.  We
538		 * don't support multiple lines for "relative" entries
539		 * nor do we make any attempt to merge data from
540		 * separate "relative" and "full" entries.  (Merging
541		 * "relative" and "full" entries would require dealing
542		 * with pathname canonicalization, which is a very
543		 * tricky subject.)
544		 */
545		for (mp = mentry->next; mp != NULL; mp = mp->next) {
546			if (mp->full && !mp->used
547			    && strcmp(mentry->name, mp->name) == 0) {
548				/* Later lines override earlier ones. */
549				mp->used = 1;
550				r1 = parse_line(a, entry, mtree, mp,
551				    &parsed_kws);
552				if (r1 < r)
553					r = r1;
554			}
555		}
556	} else {
557		/*
558		 * Relative entries require us to construct
559		 * the full path and possibly update the
560		 * current directory.
561		 */
562		size_t n = archive_strlen(&mtree->current_dir);
563		if (n > 0)
564			archive_strcat(&mtree->current_dir, "/");
565		archive_strcat(&mtree->current_dir, mentry->name);
566		archive_entry_copy_pathname(entry, mtree->current_dir.s);
567		if (archive_entry_filetype(entry) != AE_IFDIR)
568			mtree->current_dir.length = n;
569	}
570
571	/*
572	 * Try to open and stat the file to get the real size
573	 * and other file info.  It would be nice to avoid
574	 * this here so that getting a listing of an mtree
575	 * wouldn't require opening every referenced contents
576	 * file.  But then we wouldn't know the actual
577	 * contents size, so I don't see a really viable way
578	 * around this.  (Also, we may want to someday pull
579	 * other unspecified info from the contents file on
580	 * disk.)
581	 */
582	mtree->fd = -1;
583	if (archive_strlen(&mtree->contents_name) > 0)
584		path = mtree->contents_name.s;
585	else
586		path = archive_entry_pathname(entry);
587
588	if (archive_entry_filetype(entry) == AE_IFREG ||
589	    archive_entry_filetype(entry) == AE_IFDIR) {
590		mtree->fd = open(path, O_RDONLY | O_BINARY);
591		if (mtree->fd == -1 &&
592		    (errno != ENOENT ||
593		     archive_strlen(&mtree->contents_name) > 0)) {
594			archive_set_error(&a->archive, errno,
595			    "Can't open %s", path);
596			r = ARCHIVE_WARN;
597		}
598	}
599
600	st = &st_storage;
601	if (mtree->fd >= 0) {
602		if (fstat(mtree->fd, st) == -1) {
603			archive_set_error(&a->archive, errno,
604			    "Could not fstat %s", path);
605			r = ARCHIVE_WARN;
606			/* If we can't stat it, don't keep it open. */
607			close(mtree->fd);
608			mtree->fd = -1;
609			st = NULL;
610		}
611	} else if (lstat(path, st) == -1) {
612		st = NULL;
613	}
614
615	/*
616	 * If there is a contents file on disk, use that size;
617	 * otherwise leave it as-is (it might have been set from
618	 * the mtree size= keyword).
619	 */
620	if (st != NULL) {
621		mismatched_type = 0;
622		if ((st->st_mode & S_IFMT) == S_IFREG &&
623		    archive_entry_filetype(entry) != AE_IFREG)
624			mismatched_type = 1;
625		if ((st->st_mode & S_IFMT) == S_IFLNK &&
626		    archive_entry_filetype(entry) != AE_IFLNK)
627			mismatched_type = 1;
628		if ((st->st_mode & S_IFSOCK) == S_IFSOCK &&
629		    archive_entry_filetype(entry) != AE_IFSOCK)
630			mismatched_type = 1;
631		if ((st->st_mode & S_IFMT) == S_IFCHR &&
632		    archive_entry_filetype(entry) != AE_IFCHR)
633			mismatched_type = 1;
634		if ((st->st_mode & S_IFMT) == S_IFBLK &&
635		    archive_entry_filetype(entry) != AE_IFBLK)
636			mismatched_type = 1;
637		if ((st->st_mode & S_IFMT) == S_IFDIR &&
638		    archive_entry_filetype(entry) != AE_IFDIR)
639			mismatched_type = 1;
640		if ((st->st_mode & S_IFMT) == S_IFIFO &&
641		    archive_entry_filetype(entry) != AE_IFIFO)
642			mismatched_type = 1;
643
644		if (mismatched_type) {
645			if ((parsed_kws & MTREE_HAS_OPTIONAL) == 0) {
646				archive_set_error(&a->archive,
647				    ARCHIVE_ERRNO_MISC,
648				    "mtree specification has different type for %s",
649				    archive_entry_pathname(entry));
650				r = ARCHIVE_WARN;
651			} else {
652				*use_next = 1;
653			}
654			/* Don't hold a non-regular file open. */
655			if (mtree->fd >= 0)
656				close(mtree->fd);
657			mtree->fd = -1;
658			st = NULL;
659			return r;
660		}
661	}
662
663	if (st != NULL) {
664		if ((parsed_kws & MTREE_HAS_DEVICE) == 0 &&
665		    (archive_entry_filetype(entry) == AE_IFCHR ||
666		     archive_entry_filetype(entry) == AE_IFBLK))
667			archive_entry_set_rdev(entry, st->st_rdev);
668		if ((parsed_kws & (MTREE_HAS_GID | MTREE_HAS_GNAME)) == 0)
669			archive_entry_set_gid(entry, st->st_gid);
670		if ((parsed_kws & (MTREE_HAS_UID | MTREE_HAS_UNAME)) == 0)
671			archive_entry_set_uid(entry, st->st_uid);
672		if ((parsed_kws & MTREE_HAS_MTIME) == 0) {
673#if HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC
674			archive_entry_set_mtime(entry, st->st_mtime,
675			    st->st_mtimespec.tv_nsec);
676#elif HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC
677			archive_entry_set_mtime(entry, st->st_mtime,
678			    st->st_mtim.tv_nsec);
679#elif HAVE_STRUCT_STAT_ST_MTIME_N
680			archive_entry_set_mtime(entry, st->st_mtime,
681			    st->st_mtime_n);
682#elif HAVE_STRUCT_STAT_ST_UMTIME
683			archive_entry_set_mtime(entry, st->st_mtime,
684			    st->st_umtime*1000);
685#elif HAVE_STRUCT_STAT_ST_MTIME_USEC
686			archive_entry_set_mtime(entry, st->st_mtime,
687			    st->st_mtime_usec*1000);
688#else
689			archive_entry_set_mtime(entry, st->st_mtime, 0);
690#endif
691		}
692		if ((parsed_kws & MTREE_HAS_NLINK) == 0)
693			archive_entry_set_nlink(entry, st->st_nlink);
694		if ((parsed_kws & MTREE_HAS_PERM) == 0)
695			archive_entry_set_perm(entry, st->st_mode);
696		if ((parsed_kws & MTREE_HAS_SIZE) == 0)
697			archive_entry_set_size(entry, st->st_size);
698		archive_entry_set_ino(entry, st->st_ino);
699		archive_entry_set_dev(entry, st->st_dev);
700
701		archive_entry_linkify(mtree->resolver, &entry, &sparse_entry);
702	} else if (parsed_kws & MTREE_HAS_OPTIONAL) {
703		/*
704		 * Couldn't open the entry, stat it or the on-disk type
705		 * didn't match.  If this entry is optional, just ignore it
706		 * and read the next header entry.
707		 */
708		*use_next = 1;
709		return ARCHIVE_OK;
710	}
711
712	mtree->cur_size = archive_entry_size(entry);
713	mtree->offset = 0;
714
715	return r;
716}
717
718/*
719 * Each line contains a sequence of keywords.
720 */
721static int
722parse_line(struct archive_read *a, struct archive_entry *entry,
723    struct mtree *mtree, struct mtree_entry *mp, int *parsed_kws)
724{
725	struct mtree_option *iter;
726	int r = ARCHIVE_OK, r1;
727
728	for (iter = mp->options; iter != NULL; iter = iter->next) {
729		r1 = parse_keyword(a, mtree, entry, iter, parsed_kws);
730		if (r1 < r)
731			r = r1;
732	}
733	if ((*parsed_kws & MTREE_HAS_TYPE) == 0) {
734		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
735		    "Missing type keyword in mtree specification");
736		return (ARCHIVE_WARN);
737	}
738	return (r);
739}
740
741/*
742 * Device entries have one of the following forms:
743 * raw dev_t
744 * format,major,minor[,subdevice]
745 *
746 * Just use major and minor, no translation etc is done
747 * between formats.
748 */
749static int
750parse_device(struct archive *a, struct archive_entry *entry, char *val)
751{
752	char *comma1, *comma2;
753
754	comma1 = strchr(val, ',');
755	if (comma1 == NULL) {
756		archive_entry_set_dev(entry, mtree_atol10(&val));
757		return (ARCHIVE_OK);
758	}
759	++comma1;
760	comma2 = strchr(comma1, ',');
761	if (comma2 == NULL) {
762		archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
763		    "Malformed device attribute");
764		return (ARCHIVE_WARN);
765	}
766	++comma2;
767	archive_entry_set_rdevmajor(entry, mtree_atol(&comma1));
768	archive_entry_set_rdevminor(entry, mtree_atol(&comma2));
769	return (ARCHIVE_OK);
770}
771
772/*
773 * Parse a single keyword and its value.
774 */
775static int
776parse_keyword(struct archive_read *a, struct mtree *mtree,
777    struct archive_entry *entry, struct mtree_option *option, int *parsed_kws)
778{
779	char *val, *key;
780
781	key = option->value;
782
783	if (*key == '\0')
784		return (ARCHIVE_OK);
785
786	if (strcmp(key, "optional") == 0) {
787		*parsed_kws |= MTREE_HAS_OPTIONAL;
788		return (ARCHIVE_OK);
789	}
790	if (strcmp(key, "ignore") == 0) {
791		/*
792		 * The mtree processing is not recursive, so
793		 * recursion will only happen for explicitly listed
794		 * entries.
795		 */
796		return (ARCHIVE_OK);
797	}
798
799	val = strchr(key, '=');
800	if (val == NULL) {
801		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
802		    "Malformed attribute \"%s\" (%d)", key, key[0]);
803		return (ARCHIVE_WARN);
804	}
805
806	*val = '\0';
807	++val;
808
809	switch (key[0]) {
810	case 'c':
811		if (strcmp(key, "content") == 0
812		    || strcmp(key, "contents") == 0) {
813			parse_escapes(val, NULL);
814			archive_strcpy(&mtree->contents_name, val);
815			break;
816		}
817		if (strcmp(key, "cksum") == 0)
818			break;
819	case 'd':
820		if (strcmp(key, "device") == 0) {
821			*parsed_kws |= MTREE_HAS_DEVICE;
822			return parse_device(&a->archive, entry, val);
823		}
824	case 'f':
825		if (strcmp(key, "flags") == 0) {
826			*parsed_kws |= MTREE_HAS_FFLAGS;
827			archive_entry_copy_fflags_text(entry, val);
828			break;
829		}
830	case 'g':
831		if (strcmp(key, "gid") == 0) {
832			*parsed_kws |= MTREE_HAS_GID;
833			archive_entry_set_gid(entry, mtree_atol10(&val));
834			break;
835		}
836		if (strcmp(key, "gname") == 0) {
837			*parsed_kws |= MTREE_HAS_GNAME;
838			archive_entry_copy_gname(entry, val);
839			break;
840		}
841	case 'l':
842		if (strcmp(key, "link") == 0) {
843			archive_entry_copy_symlink(entry, val);
844			break;
845		}
846	case 'm':
847		if (strcmp(key, "md5") == 0 || strcmp(key, "md5digest") == 0)
848			break;
849		if (strcmp(key, "mode") == 0) {
850			if (val[0] >= '0' && val[0] <= '9') {
851				*parsed_kws |= MTREE_HAS_PERM;
852				archive_entry_set_perm(entry,
853				    mtree_atol8(&val));
854			} else {
855				archive_set_error(&a->archive,
856				    ARCHIVE_ERRNO_FILE_FORMAT,
857				    "Symbolic mode \"%s\" unsupported", val);
858				return ARCHIVE_WARN;
859			}
860			break;
861		}
862	case 'n':
863		if (strcmp(key, "nlink") == 0) {
864			*parsed_kws |= MTREE_HAS_NLINK;
865			archive_entry_set_nlink(entry, mtree_atol10(&val));
866			break;
867		}
868	case 'r':
869		if (strcmp(key, "rmd160") == 0 ||
870		    strcmp(key, "rmd160digest") == 0)
871			break;
872	case 's':
873		if (strcmp(key, "sha1") == 0 || strcmp(key, "sha1digest") == 0)
874			break;
875		if (strcmp(key, "sha256") == 0 ||
876		    strcmp(key, "sha256digest") == 0)
877			break;
878		if (strcmp(key, "sha384") == 0 ||
879		    strcmp(key, "sha384digest") == 0)
880			break;
881		if (strcmp(key, "sha512") == 0 ||
882		    strcmp(key, "sha512digest") == 0)
883			break;
884		if (strcmp(key, "size") == 0) {
885			archive_entry_set_size(entry, mtree_atol10(&val));
886			break;
887		}
888	case 't':
889		if (strcmp(key, "tags") == 0) {
890			/*
891			 * Comma delimited list of tags.
892			 * Ignore the tags for now, but the interface
893			 * should be extended to allow inclusion/exclusion.
894			 */
895			break;
896		}
897		if (strcmp(key, "time") == 0) {
898			time_t m;
899			long ns;
900
901			*parsed_kws |= MTREE_HAS_MTIME;
902			m = (time_t)mtree_atol10(&val);
903			if (*val == '.') {
904				++val;
905				ns = (long)mtree_atol10(&val);
906			} else
907				ns = 0;
908			archive_entry_set_mtime(entry, m, ns);
909			break;
910		}
911		if (strcmp(key, "type") == 0) {
912			*parsed_kws |= MTREE_HAS_TYPE;
913			switch (val[0]) {
914			case 'b':
915				if (strcmp(val, "block") == 0) {
916					mtree->filetype = AE_IFBLK;
917					break;
918				}
919			case 'c':
920				if (strcmp(val, "char") == 0) {
921					mtree->filetype = AE_IFCHR;
922					break;
923				}
924			case 'd':
925				if (strcmp(val, "dir") == 0) {
926					mtree->filetype = AE_IFDIR;
927					break;
928				}
929			case 'f':
930				if (strcmp(val, "fifo") == 0) {
931					mtree->filetype = AE_IFIFO;
932					break;
933				}
934				if (strcmp(val, "file") == 0) {
935					mtree->filetype = AE_IFREG;
936					break;
937				}
938			case 'l':
939				if (strcmp(val, "link") == 0) {
940					mtree->filetype = AE_IFLNK;
941					break;
942				}
943			default:
944				archive_set_error(&a->archive,
945				    ARCHIVE_ERRNO_FILE_FORMAT,
946				    "Unrecognized file type \"%s\"", val);
947				return (ARCHIVE_WARN);
948			}
949			archive_entry_set_filetype(entry, mtree->filetype);
950			break;
951		}
952	case 'u':
953		if (strcmp(key, "uid") == 0) {
954			*parsed_kws |= MTREE_HAS_UID;
955			archive_entry_set_uid(entry, mtree_atol10(&val));
956			break;
957		}
958		if (strcmp(key, "uname") == 0) {
959			*parsed_kws |= MTREE_HAS_UNAME;
960			archive_entry_copy_uname(entry, val);
961			break;
962		}
963	default:
964		archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
965		    "Unrecognized key %s=%s", key, val);
966		return (ARCHIVE_WARN);
967	}
968	return (ARCHIVE_OK);
969}
970
971static int
972read_data(struct archive_read *a, const void **buff, size_t *size, off_t *offset)
973{
974	size_t bytes_to_read;
975	ssize_t bytes_read;
976	struct mtree *mtree;
977
978	mtree = (struct mtree *)(a->format->data);
979	if (mtree->fd < 0) {
980		*buff = NULL;
981		*offset = 0;
982		*size = 0;
983		return (ARCHIVE_EOF);
984	}
985	if (mtree->buff == NULL) {
986		mtree->buffsize = 64 * 1024;
987		mtree->buff = malloc(mtree->buffsize);
988		if (mtree->buff == NULL) {
989			archive_set_error(&a->archive, ENOMEM,
990			    "Can't allocate memory");
991			return (ARCHIVE_FATAL);
992		}
993	}
994
995	*buff = mtree->buff;
996	*offset = mtree->offset;
997	if ((off_t)mtree->buffsize > mtree->cur_size - mtree->offset)
998		bytes_to_read = mtree->cur_size - mtree->offset;
999	else
1000		bytes_to_read = mtree->buffsize;
1001	bytes_read = read(mtree->fd, mtree->buff, bytes_to_read);
1002	if (bytes_read < 0) {
1003		archive_set_error(&a->archive, errno, "Can't read");
1004		return (ARCHIVE_WARN);
1005	}
1006	if (bytes_read == 0) {
1007		*size = 0;
1008		return (ARCHIVE_EOF);
1009	}
1010	mtree->offset += bytes_read;
1011	*size = bytes_read;
1012	return (ARCHIVE_OK);
1013}
1014
1015/* Skip does nothing except possibly close the contents file. */
1016static int
1017skip(struct archive_read *a)
1018{
1019	struct mtree *mtree;
1020
1021	mtree = (struct mtree *)(a->format->data);
1022	if (mtree->fd >= 0) {
1023		close(mtree->fd);
1024		mtree->fd = -1;
1025	}
1026	return (ARCHIVE_OK);
1027}
1028
1029/*
1030 * Since parsing backslash sequences always makes strings shorter,
1031 * we can always do this conversion in-place.
1032 */
1033static void
1034parse_escapes(char *src, struct mtree_entry *mentry)
1035{
1036	char *dest = src;
1037	char c;
1038
1039	if (mentry != NULL && strcmp(src, ".") == 0)
1040		mentry->full = 1;
1041
1042	while (*src != '\0') {
1043		c = *src++;
1044		if (c == '/' && mentry != NULL)
1045			mentry->full = 1;
1046		if (c == '\\') {
1047			switch (src[0]) {
1048			case '0':
1049				if (src[1] < '0' || src[1] > '7') {
1050					c = 0;
1051					++src;
1052					break;
1053				}
1054				/* FALLTHROUGH */
1055			case '1':
1056			case '2':
1057			case '3':
1058				if (src[1] >= '0' && src[1] <= '7' &&
1059				    src[2] >= '0' && src[2] <= '7') {
1060					c = (src[0] - '0') << 6;
1061					c |= (src[1] - '0') << 3;
1062					c |= (src[2] - '0');
1063					src += 3;
1064				}
1065				break;
1066			case 'a':
1067				c = '\a';
1068				++src;
1069				break;
1070			case 'b':
1071				c = '\b';
1072				++src;
1073				break;
1074			case 'f':
1075				c = '\f';
1076				++src;
1077				break;
1078			case 'n':
1079				c = '\n';
1080				++src;
1081				break;
1082			case 'r':
1083				c = '\r';
1084				++src;
1085				break;
1086			case 's':
1087				c = ' ';
1088				++src;
1089				break;
1090			case 't':
1091				c = '\t';
1092				++src;
1093				break;
1094			case 'v':
1095				c = '\v';
1096				++src;
1097				break;
1098			}
1099		}
1100		*dest++ = c;
1101	}
1102	*dest = '\0';
1103}
1104
1105/*
1106 * Note that this implementation does not (and should not!) obey
1107 * locale settings; you cannot simply substitute strtol here, since
1108 * it does obey locale.
1109 */
1110static int64_t
1111mtree_atol8(char **p)
1112{
1113	int64_t	l, limit, last_digit_limit;
1114	int digit, base;
1115
1116	base = 8;
1117	limit = INT64_MAX / base;
1118	last_digit_limit = INT64_MAX % base;
1119
1120	l = 0;
1121	digit = **p - '0';
1122	while (digit >= 0 && digit < base) {
1123		if (l>limit || (l == limit && digit > last_digit_limit)) {
1124			l = INT64_MAX; /* Truncate on overflow. */
1125			break;
1126		}
1127		l = (l * base) + digit;
1128		digit = *++(*p) - '0';
1129	}
1130	return (l);
1131}
1132
1133/*
1134 * Note that this implementation does not (and should not!) obey
1135 * locale settings; you cannot simply substitute strtol here, since
1136 * it does obey locale.
1137 */
1138static int64_t
1139mtree_atol10(char **p)
1140{
1141	int64_t l, limit, last_digit_limit;
1142	int base, digit, sign;
1143
1144	base = 10;
1145	limit = INT64_MAX / base;
1146	last_digit_limit = INT64_MAX % base;
1147
1148	if (**p == '-') {
1149		sign = -1;
1150		++(*p);
1151	} else
1152		sign = 1;
1153
1154	l = 0;
1155	digit = **p - '0';
1156	while (digit >= 0 && digit < base) {
1157		if (l > limit || (l == limit && digit > last_digit_limit)) {
1158			l = INT64_MAX; /* Truncate on overflow. */
1159			break;
1160		}
1161		l = (l * base) + digit;
1162		digit = *++(*p) - '0';
1163	}
1164	return (sign < 0) ? -l : l;
1165}
1166
1167/*
1168 * Note that this implementation does not (and should not!) obey
1169 * locale settings; you cannot simply substitute strtol here, since
1170 * it does obey locale.
1171 */
1172static int64_t
1173mtree_atol16(char **p)
1174{
1175	int64_t l, limit, last_digit_limit;
1176	int base, digit, sign;
1177
1178	base = 16;
1179	limit = INT64_MAX / base;
1180	last_digit_limit = INT64_MAX % base;
1181
1182	if (**p == '-') {
1183		sign = -1;
1184		++(*p);
1185	} else
1186		sign = 1;
1187
1188	l = 0;
1189	if (**p >= '0' && **p <= '9')
1190		digit = **p - '0';
1191	else if (**p >= 'a' && **p <= 'f')
1192		digit = **p - 'a' + 10;
1193	else if (**p >= 'A' && **p <= 'F')
1194		digit = **p - 'A' + 10;
1195	else
1196		digit = -1;
1197	while (digit >= 0 && digit < base) {
1198		if (l > limit || (l == limit && digit > last_digit_limit)) {
1199			l = INT64_MAX; /* Truncate on overflow. */
1200			break;
1201		}
1202		l = (l * base) + digit;
1203		if (**p >= '0' && **p <= '9')
1204			digit = **p - '0';
1205		else if (**p >= 'a' && **p <= 'f')
1206			digit = **p - 'a' + 10;
1207		else if (**p >= 'A' && **p <= 'F')
1208			digit = **p - 'A' + 10;
1209		else
1210			digit = -1;
1211	}
1212	return (sign < 0) ? -l : l;
1213}
1214
1215static int64_t
1216mtree_atol(char **p)
1217{
1218	if (**p != '0')
1219		return mtree_atol10(p);
1220	if ((*p)[1] == 'x' || (*p)[1] == 'X') {
1221		*p += 2;
1222		return mtree_atol16(p);
1223	}
1224	return mtree_atol8(p);
1225}
1226
1227/*
1228 * Returns length of line (including trailing newline)
1229 * or negative on error.  'start' argument is updated to
1230 * point to first character of line.
1231 */
1232static ssize_t
1233readline(struct archive_read *a, struct mtree *mtree, char **start, ssize_t limit)
1234{
1235	ssize_t bytes_read;
1236	ssize_t total_size = 0;
1237	ssize_t find_off = 0;
1238	const void *t;
1239	const char *s;
1240	void *p;
1241	char *u;
1242
1243	/* Accumulate line in a line buffer. */
1244	for (;;) {
1245		/* Read some more. */
1246		t = __archive_read_ahead(a, 1, &bytes_read);
1247		if (t == NULL)
1248			return (0);
1249		if (bytes_read < 0)
1250			return (ARCHIVE_FATAL);
1251		s = t;  /* Start of line? */
1252		p = memchr(t, '\n', bytes_read);
1253		/* If we found '\n', trim the read. */
1254		if (p != NULL) {
1255			bytes_read = 1 + ((const char *)p) - s;
1256		}
1257		if (total_size + bytes_read + 1 > limit) {
1258			archive_set_error(&a->archive,
1259			    ARCHIVE_ERRNO_FILE_FORMAT,
1260			    "Line too long");
1261			return (ARCHIVE_FATAL);
1262		}
1263		if (archive_string_ensure(&mtree->line,
1264			total_size + bytes_read + 1) == NULL) {
1265			archive_set_error(&a->archive, ENOMEM,
1266			    "Can't allocate working buffer");
1267			return (ARCHIVE_FATAL);
1268		}
1269		memcpy(mtree->line.s + total_size, t, bytes_read);
1270		__archive_read_consume(a, bytes_read);
1271		total_size += bytes_read;
1272		/* Null terminate. */
1273		mtree->line.s[total_size] = '\0';
1274		/* If we found an unescaped '\n', clean up and return. */
1275		for (u = mtree->line.s + find_off; *u; ++u) {
1276			if (u[0] == '\n') {
1277				*start = mtree->line.s;
1278				return total_size;
1279			}
1280			if (u[0] == '#') {
1281				if (p == NULL)
1282					break;
1283				*start = mtree->line.s;
1284				return total_size;
1285			}
1286			if (u[0] != '\\')
1287				continue;
1288			if (u[1] == '\\') {
1289				++u;
1290				continue;
1291			}
1292			if (u[1] == '\n') {
1293				memmove(u, u + 1,
1294				    total_size - (u - mtree->line.s) + 1);
1295				--total_size;
1296				++u;
1297				break;
1298			}
1299			if (u[1] == '\0')
1300				break;
1301		}
1302		find_off = u - mtree->line.s;
1303	}
1304}
1305