1 /*-
2  * Copyright (c) 2003-2009 Tim Kientzle
3  * Copyright (c) 2010-2012 Michihiro NAKAJIMA
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  *    in this position and unchanged.
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 /* This is the tree-walking code for POSIX systems. */
29 #if !defined(_WIN32) || defined(__CYGWIN__)
30 
31 #include "archive_platform.h"
32 __FBSDID("$FreeBSD$");
33 
34 #ifdef HAVE_SYS_PARAM_H
35 #include <sys/param.h>
36 #endif
37 #ifdef HAVE_SYS_MOUNT_H
38 #include <sys/mount.h>
39 #endif
40 #ifdef HAVE_SYS_STAT_H
41 #include <sys/stat.h>
42 #endif
43 #ifdef HAVE_SYS_STATFS_H
44 #include <sys/statfs.h>
45 #endif
46 #ifdef HAVE_SYS_STATVFS_H
47 #include <sys/statvfs.h>
48 #endif
49 #ifdef HAVE_SYS_TIME_H
50 #include <sys/time.h>
51 #endif
52 #ifdef HAVE_LINUX_MAGIC_H
53 #include <linux/magic.h>
54 #endif
55 #ifdef HAVE_LINUX_FS_H
56 #include <linux/fs.h>
57 #endif
58 /*
59  * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h.
60  * As the include guards don't agree, the order of include is important.
61  */
62 #ifdef HAVE_LINUX_EXT2_FS_H
63 #include <linux/ext2_fs.h>      /* for Linux file flags */
64 #endif
65 #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__)
66 #include <ext2fs/ext2_fs.h>     /* Linux file flags, broken on Cygwin */
67 #endif
68 #ifdef HAVE_DIRECT_H
69 #include <direct.h>
70 #endif
71 #ifdef HAVE_DIRENT_H
72 #include <dirent.h>
73 #endif
74 #ifdef HAVE_ERRNO_H
75 #include <errno.h>
76 #endif
77 #ifdef HAVE_FCNTL_H
78 #include <fcntl.h>
79 #endif
80 #ifdef HAVE_LIMITS_H
81 #include <limits.h>
82 #endif
83 #ifdef HAVE_STDLIB_H
84 #include <stdlib.h>
85 #endif
86 #ifdef HAVE_STRING_H
87 #include <string.h>
88 #endif
89 #ifdef HAVE_UNISTD_H
90 #include <unistd.h>
91 #endif
92 #ifdef HAVE_SYS_IOCTL_H
93 #include <sys/ioctl.h>
94 #endif
95 
96 #include "archive.h"
97 #include "archive_string.h"
98 #include "archive_entry.h"
99 #include "archive_private.h"
100 #include "archive_read_disk_private.h"
101 
102 #ifndef HAVE_FCHDIR
103 #error fchdir function required.
104 #endif
105 #ifndef O_BINARY
106 #define O_BINARY	0
107 #endif
108 #ifndef O_CLOEXEC
109 #define O_CLOEXEC	0
110 #endif
111 
112 /*-
113  * This is a new directory-walking system that addresses a number
114  * of problems I've had with fts(3).  In particular, it has no
115  * pathname-length limits (other than the size of 'int'), handles
116  * deep logical traversals, uses considerably less memory, and has
117  * an opaque interface (easier to modify in the future).
118  *
119  * Internally, it keeps a single list of "tree_entry" items that
120  * represent filesystem objects that require further attention.
121  * Non-directories are not kept in memory: they are pulled from
122  * readdir(), returned to the client, then freed as soon as possible.
123  * Any directory entry to be traversed gets pushed onto the stack.
124  *
125  * There is surprisingly little information that needs to be kept for
126  * each item on the stack.  Just the name, depth (represented here as the
127  * string length of the parent directory's pathname), and some markers
128  * indicating how to get back to the parent (via chdir("..") for a
129  * regular dir or via fchdir(2) for a symlink).
130  */
131 /*
132  * TODO:
133  *    1) Loop checking.
134  *    3) Arbitrary logical traversals by closing/reopening intermediate fds.
135  */
136 
137 struct restore_time {
138 	const char		*name;
139 	time_t			 mtime;
140 	long			 mtime_nsec;
141 	time_t			 atime;
142 	long			 atime_nsec;
143 	mode_t			 filetype;
144 	int			 noatime;
145 };
146 
147 struct tree_entry {
148 	int			 depth;
149 	struct tree_entry	*next;
150 	struct tree_entry	*parent;
151 	struct archive_string	 name;
152 	size_t			 dirname_length;
153 	int64_t			 dev;
154 	int64_t			 ino;
155 	int			 flags;
156 	int			 filesystem_id;
157 	/* How to return back to the parent of a symlink. */
158 	int			 symlink_parent_fd;
159 	/* How to restore time of a directory. */
160 	struct restore_time	 restore_time;
161 };
162 
163 struct filesystem {
164 	int64_t		dev;
165 	int		synthetic;
166 	int		remote;
167 	int		noatime;
168 #if defined(USE_READDIR_R)
169 	size_t		name_max;
170 #endif
171 	long		incr_xfer_size;
172 	long		max_xfer_size;
173 	long		min_xfer_size;
174 	long		xfer_align;
175 
176 	/*
177 	 * Buffer used for reading file contents.
178 	 */
179 	/* Exactly allocated memory pointer. */
180 	unsigned char	*allocation_ptr;
181 	/* Pointer adjusted to the filesystem alignment . */
182 	unsigned char	*buff;
183 	size_t		 buff_size;
184 };
185 
186 /* Definitions for tree_entry.flags bitmap. */
187 #define	isDir		1  /* This entry is a regular directory. */
188 #define	isDirLink	2  /* This entry is a symbolic link to a directory. */
189 #define	needsFirstVisit	4  /* This is an initial entry. */
190 #define	needsDescent	8  /* This entry needs to be previsited. */
191 #define	needsOpen	16 /* This is a directory that needs to be opened. */
192 #define	needsAscent	32 /* This entry needs to be postvisited. */
193 
194 /*
195  * Local data for this package.
196  */
197 struct tree {
198 	struct tree_entry	*stack;
199 	struct tree_entry	*current;
200 	DIR			*d;
201 #define	INVALID_DIR_HANDLE NULL
202 	struct dirent		*de;
203 #if defined(USE_READDIR_R)
204 	struct dirent		*dirent;
205 	size_t			 dirent_allocated;
206 #endif
207 	int			 flags;
208 	int			 visit_type;
209 	/* Error code from last failed operation. */
210 	int			 tree_errno;
211 
212 	/* Dynamically-sized buffer for holding path */
213 	struct archive_string	 path;
214 
215 	/* Last path element */
216 	const char		*basename;
217 	/* Leading dir length */
218 	size_t			 dirname_length;
219 
220 	int			 depth;
221 	int			 openCount;
222 	int			 maxOpenCount;
223 	int			 initial_dir_fd;
224 	int			 working_dir_fd;
225 
226 	struct stat		 lst;
227 	struct stat		 st;
228 	int			 descend;
229 	int			 nlink;
230 	/* How to restore time of a file. */
231 	struct restore_time	 restore_time;
232 
233 	struct entry_sparse {
234 		int64_t		 length;
235 		int64_t		 offset;
236 	}			*sparse_list, *current_sparse;
237 	int			 sparse_count;
238 	int			 sparse_list_size;
239 
240 	char			 initial_symlink_mode;
241 	char			 symlink_mode;
242 	struct filesystem	*current_filesystem;
243 	struct filesystem	*filesystem_table;
244 	int			 initial_filesystem_id;
245 	int			 current_filesystem_id;
246 	int			 max_filesystem_id;
247 	int			 allocated_filesystem;
248 
249 	int			 entry_fd;
250 	int			 entry_eof;
251 	int64_t			 entry_remaining_bytes;
252 	int64_t			 entry_total;
253 	unsigned char		*entry_buff;
254 	size_t			 entry_buff_size;
255 };
256 
257 /* Definitions for tree.flags bitmap. */
258 #define	hasStat		16 /* The st entry is valid. */
259 #define	hasLstat	32 /* The lst entry is valid. */
260 #define	onWorkingDir	64 /* We are on the working dir where we are
261 			    * reading directory entry at this time. */
262 #define	needsRestoreTimes 128
263 #define	onInitialDir	256 /* We are on the initial dir. */
264 
265 static int
266 tree_dir_next_posix(struct tree *t);
267 
268 #ifdef HAVE_DIRENT_D_NAMLEN
269 /* BSD extension; avoids need for a strlen() call. */
270 #define	D_NAMELEN(dp)	(dp)->d_namlen
271 #else
272 #define	D_NAMELEN(dp)	(strlen((dp)->d_name))
273 #endif
274 
275 /* Initiate/terminate a tree traversal. */
276 static struct tree *tree_open(const char *, int, int);
277 static struct tree *tree_reopen(struct tree *, const char *, int);
278 static void tree_close(struct tree *);
279 static void tree_free(struct tree *);
280 static void tree_push(struct tree *, const char *, int, int64_t, int64_t,
281 		struct restore_time *);
282 static int tree_enter_initial_dir(struct tree *);
283 static int tree_enter_working_dir(struct tree *);
284 static int tree_current_dir_fd(struct tree *);
285 
286 /*
287  * tree_next() returns Zero if there is no next entry, non-zero if
288  * there is.  Note that directories are visited three times.
289  * Directories are always visited first as part of enumerating their
290  * parent; that is a "regular" visit.  If tree_descend() is invoked at
291  * that time, the directory is added to a work list and will
292  * subsequently be visited two more times: once just after descending
293  * into the directory ("postdescent") and again just after ascending
294  * back to the parent ("postascent").
295  *
296  * TREE_ERROR_DIR is returned if the descent failed (because the
297  * directory couldn't be opened, for instance).  This is returned
298  * instead of TREE_POSTDESCENT/TREE_POSTASCENT.  TREE_ERROR_DIR is not a
299  * fatal error, but it does imply that the relevant subtree won't be
300  * visited.  TREE_ERROR_FATAL is returned for an error that left the
301  * traversal completely hosed.  Right now, this is only returned for
302  * chdir() failures during ascent.
303  */
304 #define	TREE_REGULAR		1
305 #define	TREE_POSTDESCENT	2
306 #define	TREE_POSTASCENT		3
307 #define	TREE_ERROR_DIR		-1
308 #define	TREE_ERROR_FATAL	-2
309 
310 static int tree_next(struct tree *);
311 
312 /*
313  * Return information about the current entry.
314  */
315 
316 /*
317  * The current full pathname, length of the full pathname, and a name
318  * that can be used to access the file.  Because tree does use chdir
319  * extensively, the access path is almost never the same as the full
320  * current path.
321  *
322  * TODO: On platforms that support it, use openat()-style operations
323  * to eliminate the chdir() operations entirely while still supporting
324  * arbitrarily deep traversals.  This makes access_path troublesome to
325  * support, of course, which means we'll need a rich enough interface
326  * that clients can function without it.  (In particular, we'll need
327  * tree_current_open() that returns an open file descriptor.)
328  *
329  */
330 static const char *tree_current_path(struct tree *);
331 static const char *tree_current_access_path(struct tree *);
332 
333 /*
334  * Request the lstat() or stat() data for the current path.  Since the
335  * tree package needs to do some of this anyway, and caches the
336  * results, you should take advantage of it here if you need it rather
337  * than make a redundant stat() or lstat() call of your own.
338  */
339 static const struct stat *tree_current_stat(struct tree *);
340 static const struct stat *tree_current_lstat(struct tree *);
341 static int	tree_current_is_symblic_link_target(struct tree *);
342 
343 /* The following functions use tricks to avoid a certain number of
344  * stat()/lstat() calls. */
345 /* "is_physical_dir" is equivalent to S_ISDIR(tree_current_lstat()->st_mode) */
346 static int tree_current_is_physical_dir(struct tree *);
347 /* "is_dir" is equivalent to S_ISDIR(tree_current_stat()->st_mode) */
348 static int tree_current_is_dir(struct tree *);
349 static int update_current_filesystem(struct archive_read_disk *a,
350 		    int64_t dev);
351 static int setup_current_filesystem(struct archive_read_disk *);
352 static int tree_target_is_same_as_parent(struct tree *, const struct stat *);
353 
354 static int	_archive_read_disk_open(struct archive *, const char *);
355 static int	_archive_read_free(struct archive *);
356 static int	_archive_read_close(struct archive *);
357 static int	_archive_read_data_block(struct archive *,
358 		    const void **, size_t *, int64_t *);
359 static int	_archive_read_next_header(struct archive *,
360 		    struct archive_entry **);
361 static int	_archive_read_next_header2(struct archive *,
362 		    struct archive_entry *);
363 static const char *trivial_lookup_gname(void *, int64_t gid);
364 static const char *trivial_lookup_uname(void *, int64_t uid);
365 static int	setup_sparse(struct archive_read_disk *, struct archive_entry *);
366 static int	close_and_restore_time(int fd, struct tree *,
367 		    struct restore_time *);
368 static int	open_on_current_dir(struct tree *, const char *, int);
369 static int	tree_dup(int);
370 
371 
372 static struct archive_vtable *
archive_read_disk_vtable(void)373 archive_read_disk_vtable(void)
374 {
375 	static struct archive_vtable av;
376 	static int inited = 0;
377 
378 	if (!inited) {
379 		av.archive_free = _archive_read_free;
380 		av.archive_close = _archive_read_close;
381 		av.archive_read_data_block = _archive_read_data_block;
382 		av.archive_read_next_header = _archive_read_next_header;
383 		av.archive_read_next_header2 = _archive_read_next_header2;
384 		inited = 1;
385 	}
386 	return (&av);
387 }
388 
389 const char *
archive_read_disk_gname(struct archive * _a,la_int64_t gid)390 archive_read_disk_gname(struct archive *_a, la_int64_t gid)
391 {
392 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
393 	if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
394 		ARCHIVE_STATE_ANY, "archive_read_disk_gname"))
395 		return (NULL);
396 	if (a->lookup_gname == NULL)
397 		return (NULL);
398 	return ((*a->lookup_gname)(a->lookup_gname_data, gid));
399 }
400 
401 const char *
archive_read_disk_uname(struct archive * _a,la_int64_t uid)402 archive_read_disk_uname(struct archive *_a, la_int64_t uid)
403 {
404 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
405 	if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
406 		ARCHIVE_STATE_ANY, "archive_read_disk_uname"))
407 		return (NULL);
408 	if (a->lookup_uname == NULL)
409 		return (NULL);
410 	return ((*a->lookup_uname)(a->lookup_uname_data, uid));
411 }
412 
413 int
archive_read_disk_set_gname_lookup(struct archive * _a,void * private_data,const char * (* lookup_gname)(void * private,la_int64_t gid),void (* cleanup_gname)(void * private))414 archive_read_disk_set_gname_lookup(struct archive *_a,
415     void *private_data,
416     const char * (*lookup_gname)(void *private, la_int64_t gid),
417     void (*cleanup_gname)(void *private))
418 {
419 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
420 	archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC,
421 	    ARCHIVE_STATE_ANY, "archive_read_disk_set_gname_lookup");
422 
423 	if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL)
424 		(a->cleanup_gname)(a->lookup_gname_data);
425 
426 	a->lookup_gname = lookup_gname;
427 	a->cleanup_gname = cleanup_gname;
428 	a->lookup_gname_data = private_data;
429 	return (ARCHIVE_OK);
430 }
431 
432 int
archive_read_disk_set_uname_lookup(struct archive * _a,void * private_data,const char * (* lookup_uname)(void * private,la_int64_t uid),void (* cleanup_uname)(void * private))433 archive_read_disk_set_uname_lookup(struct archive *_a,
434     void *private_data,
435     const char * (*lookup_uname)(void *private, la_int64_t uid),
436     void (*cleanup_uname)(void *private))
437 {
438 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
439 	archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC,
440 	    ARCHIVE_STATE_ANY, "archive_read_disk_set_uname_lookup");
441 
442 	if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL)
443 		(a->cleanup_uname)(a->lookup_uname_data);
444 
445 	a->lookup_uname = lookup_uname;
446 	a->cleanup_uname = cleanup_uname;
447 	a->lookup_uname_data = private_data;
448 	return (ARCHIVE_OK);
449 }
450 
451 /*
452  * Create a new archive_read_disk object and initialize it with global state.
453  */
454 struct archive *
archive_read_disk_new(void)455 archive_read_disk_new(void)
456 {
457 	struct archive_read_disk *a;
458 
459 	a = (struct archive_read_disk *)calloc(1, sizeof(*a));
460 	if (a == NULL)
461 		return (NULL);
462 	a->archive.magic = ARCHIVE_READ_DISK_MAGIC;
463 	a->archive.state = ARCHIVE_STATE_NEW;
464 	a->archive.vtable = archive_read_disk_vtable();
465 	a->entry = archive_entry_new2(&a->archive);
466 	a->lookup_uname = trivial_lookup_uname;
467 	a->lookup_gname = trivial_lookup_gname;
468 	a->flags = ARCHIVE_READDISK_MAC_COPYFILE;
469 	a->open_on_current_dir = open_on_current_dir;
470 	a->tree_current_dir_fd = tree_current_dir_fd;
471 	a->tree_enter_working_dir = tree_enter_working_dir;
472 	return (&a->archive);
473 }
474 
475 static int
_archive_read_free(struct archive * _a)476 _archive_read_free(struct archive *_a)
477 {
478 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
479 	int r;
480 
481 	if (_a == NULL)
482 		return (ARCHIVE_OK);
483 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
484 	    ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
485 
486 	if (a->archive.state != ARCHIVE_STATE_CLOSED)
487 		r = _archive_read_close(&a->archive);
488 	else
489 		r = ARCHIVE_OK;
490 
491 	tree_free(a->tree);
492 	if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL)
493 		(a->cleanup_gname)(a->lookup_gname_data);
494 	if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL)
495 		(a->cleanup_uname)(a->lookup_uname_data);
496 	archive_string_free(&a->archive.error_string);
497 	archive_entry_free(a->entry);
498 	a->archive.magic = 0;
499 	__archive_clean(&a->archive);
500 	free(a);
501 	return (r);
502 }
503 
504 static int
_archive_read_close(struct archive * _a)505 _archive_read_close(struct archive *_a)
506 {
507 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
508 
509 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
510 	    ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
511 
512 	if (a->archive.state != ARCHIVE_STATE_FATAL)
513 		a->archive.state = ARCHIVE_STATE_CLOSED;
514 
515 	tree_close(a->tree);
516 
517 	return (ARCHIVE_OK);
518 }
519 
520 static void
setup_symlink_mode(struct archive_read_disk * a,char symlink_mode,int follow_symlinks)521 setup_symlink_mode(struct archive_read_disk *a, char symlink_mode,
522     int follow_symlinks)
523 {
524 	a->symlink_mode = symlink_mode;
525 	a->follow_symlinks = follow_symlinks;
526 	if (a->tree != NULL) {
527 		a->tree->initial_symlink_mode = a->symlink_mode;
528 		a->tree->symlink_mode = a->symlink_mode;
529 	}
530 }
531 
532 int
archive_read_disk_set_symlink_logical(struct archive * _a)533 archive_read_disk_set_symlink_logical(struct archive *_a)
534 {
535 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
536 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
537 	    ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical");
538 	setup_symlink_mode(a, 'L', 1);
539 	return (ARCHIVE_OK);
540 }
541 
542 int
archive_read_disk_set_symlink_physical(struct archive * _a)543 archive_read_disk_set_symlink_physical(struct archive *_a)
544 {
545 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
546 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
547 	    ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical");
548 	setup_symlink_mode(a, 'P', 0);
549 	return (ARCHIVE_OK);
550 }
551 
552 int
archive_read_disk_set_symlink_hybrid(struct archive * _a)553 archive_read_disk_set_symlink_hybrid(struct archive *_a)
554 {
555 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
556 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
557 	    ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid");
558 	setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */
559 	return (ARCHIVE_OK);
560 }
561 
562 int
archive_read_disk_set_atime_restored(struct archive * _a)563 archive_read_disk_set_atime_restored(struct archive *_a)
564 {
565 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
566 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
567 	    ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime");
568 #ifdef HAVE_UTIMES
569 	a->flags |= ARCHIVE_READDISK_RESTORE_ATIME;
570 	if (a->tree != NULL)
571 		a->tree->flags |= needsRestoreTimes;
572 	return (ARCHIVE_OK);
573 #else
574 	/* Display warning and unset flag */
575 	archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
576 	    "Cannot restore access time on this system");
577 	a->flags &= ~ARCHIVE_READDISK_RESTORE_ATIME;
578 	return (ARCHIVE_WARN);
579 #endif
580 }
581 
582 int
archive_read_disk_set_behavior(struct archive * _a,int flags)583 archive_read_disk_set_behavior(struct archive *_a, int flags)
584 {
585 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
586 	int r = ARCHIVE_OK;
587 
588 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
589 	    ARCHIVE_STATE_ANY, "archive_read_disk_honor_nodump");
590 
591 	a->flags = flags;
592 
593 	if (flags & ARCHIVE_READDISK_RESTORE_ATIME)
594 		r = archive_read_disk_set_atime_restored(_a);
595 	else {
596 		if (a->tree != NULL)
597 			a->tree->flags &= ~needsRestoreTimes;
598 	}
599 	return (r);
600 }
601 
602 /*
603  * Trivial implementations of gname/uname lookup functions.
604  * These are normally overridden by the client, but these stub
605  * versions ensure that we always have something that works.
606  */
607 static const char *
trivial_lookup_gname(void * private_data,int64_t gid)608 trivial_lookup_gname(void *private_data, int64_t gid)
609 {
610 	(void)private_data; /* UNUSED */
611 	(void)gid; /* UNUSED */
612 	return (NULL);
613 }
614 
615 static const char *
trivial_lookup_uname(void * private_data,int64_t uid)616 trivial_lookup_uname(void *private_data, int64_t uid)
617 {
618 	(void)private_data; /* UNUSED */
619 	(void)uid; /* UNUSED */
620 	return (NULL);
621 }
622 
623 /*
624  * Allocate memory for the reading buffer adjusted to the filesystem
625  * alignment.
626  */
627 static int
setup_suitable_read_buffer(struct archive_read_disk * a)628 setup_suitable_read_buffer(struct archive_read_disk *a)
629 {
630 	struct tree *t = a->tree;
631 	struct filesystem *cf = t->current_filesystem;
632 	size_t asize;
633 	size_t s;
634 
635 	if (cf->allocation_ptr == NULL) {
636 		/* If we couldn't get a filesystem alignment,
637 		 * we use 4096 as default value but we won't use
638 		 * O_DIRECT to open() and openat() operations. */
639 		long xfer_align = (cf->xfer_align == -1)?4096:cf->xfer_align;
640 
641 		if (cf->max_xfer_size != -1)
642 			asize = cf->max_xfer_size + xfer_align;
643 		else {
644 			long incr = cf->incr_xfer_size;
645 			/* Some platform does not set a proper value to
646 			 * incr_xfer_size.*/
647 			if (incr < 0)
648 				incr = cf->min_xfer_size;
649 			if (cf->min_xfer_size < 0) {
650 				incr = xfer_align;
651 				asize = xfer_align;
652 			} else
653 				asize = cf->min_xfer_size;
654 
655 			/* Increase a buffer size up to 64K bytes in
656 			 * a proper increment size. */
657 			while (asize < 1024*64)
658 				asize += incr;
659 			/* Take a margin to adjust to the filesystem
660 			 * alignment. */
661 			asize += xfer_align;
662 		}
663 		cf->allocation_ptr = malloc(asize);
664 		if (cf->allocation_ptr == NULL) {
665 			archive_set_error(&a->archive, ENOMEM,
666 			    "Couldn't allocate memory");
667 			a->archive.state = ARCHIVE_STATE_FATAL;
668 			return (ARCHIVE_FATAL);
669 		}
670 
671 		/*
672 		 * Calculate proper address for the filesystem.
673 		 */
674 		s = (uintptr_t)cf->allocation_ptr;
675 		s %= xfer_align;
676 		if (s > 0)
677 			s = xfer_align - s;
678 
679 		/*
680 		 * Set a read buffer pointer in the proper alignment of
681 		 * the current filesystem.
682 		 */
683 		cf->buff = cf->allocation_ptr + s;
684 		cf->buff_size = asize - xfer_align;
685 	}
686 	return (ARCHIVE_OK);
687 }
688 
689 static int
_archive_read_data_block(struct archive * _a,const void ** buff,size_t * size,int64_t * offset)690 _archive_read_data_block(struct archive *_a, const void **buff,
691     size_t *size, int64_t *offset)
692 {
693 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
694 	struct tree *t = a->tree;
695 	int r;
696 	ssize_t bytes;
697 	size_t buffbytes;
698 	int empty_sparse_region = 0;
699 
700 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
701 	    "archive_read_data_block");
702 
703 	if (t->entry_eof || t->entry_remaining_bytes <= 0) {
704 		r = ARCHIVE_EOF;
705 		goto abort_read_data;
706 	}
707 
708 	/*
709 	 * Open the current file.
710 	 */
711 	if (t->entry_fd < 0) {
712 		int flags = O_RDONLY | O_BINARY | O_CLOEXEC;
713 
714 		/*
715 		 * Eliminate or reduce cache effects if we can.
716 		 *
717 		 * Carefully consider this to be enabled.
718 		 */
719 #if defined(O_DIRECT) && 0/* Disabled for now */
720 		if (t->current_filesystem->xfer_align != -1 &&
721 		    t->nlink == 1)
722 			flags |= O_DIRECT;
723 #endif
724 #if defined(O_NOATIME)
725 		/*
726 		 * Linux has O_NOATIME flag; use it if we need.
727 		 */
728 		if ((t->flags & needsRestoreTimes) != 0 &&
729 		    t->restore_time.noatime == 0)
730 			flags |= O_NOATIME;
731 		do {
732 #endif
733 			t->entry_fd = open_on_current_dir(t,
734 			    tree_current_access_path(t), flags);
735 			__archive_ensure_cloexec_flag(t->entry_fd);
736 #if defined(O_NOATIME)
737 			/*
738 			 * When we did open the file with O_NOATIME flag,
739 			 * if successful, set 1 to t->restore_time.noatime
740 			 * not to restore an atime of the file later.
741 			 * if failed by EPERM, retry it without O_NOATIME flag.
742 			 */
743 			if (flags & O_NOATIME) {
744 				if (t->entry_fd >= 0)
745 					t->restore_time.noatime = 1;
746 				else if (errno == EPERM) {
747 					flags &= ~O_NOATIME;
748 					continue;
749 				}
750 			}
751 		} while (0);
752 #endif
753 		if (t->entry_fd < 0) {
754 			archive_set_error(&a->archive, errno,
755 			    "Couldn't open %s", tree_current_path(t));
756 			r = ARCHIVE_FAILED;
757 			tree_enter_initial_dir(t);
758 			goto abort_read_data;
759 		}
760 		tree_enter_initial_dir(t);
761 	}
762 
763 	/*
764 	 * Allocate read buffer if not allocated.
765 	 */
766 	if (t->current_filesystem->allocation_ptr == NULL) {
767 		r = setup_suitable_read_buffer(a);
768 		if (r != ARCHIVE_OK) {
769 			a->archive.state = ARCHIVE_STATE_FATAL;
770 			goto abort_read_data;
771 		}
772 	}
773 	t->entry_buff = t->current_filesystem->buff;
774 	t->entry_buff_size = t->current_filesystem->buff_size;
775 
776 	buffbytes = t->entry_buff_size;
777 	if ((int64_t)buffbytes > t->current_sparse->length)
778 		buffbytes = t->current_sparse->length;
779 
780 	if (t->current_sparse->length == 0)
781 		empty_sparse_region = 1;
782 
783 	/*
784 	 * Skip hole.
785 	 * TODO: Should we consider t->current_filesystem->xfer_align?
786 	 */
787 	if (t->current_sparse->offset > t->entry_total) {
788 		if (lseek(t->entry_fd,
789 		    (off_t)t->current_sparse->offset, SEEK_SET) < 0) {
790 			archive_set_error(&a->archive, errno, "Seek error");
791 			r = ARCHIVE_FATAL;
792 			a->archive.state = ARCHIVE_STATE_FATAL;
793 			goto abort_read_data;
794 		}
795 		bytes = t->current_sparse->offset - t->entry_total;
796 		t->entry_remaining_bytes -= bytes;
797 		t->entry_total += bytes;
798 	}
799 
800 	/*
801 	 * Read file contents.
802 	 */
803 	if (buffbytes > 0) {
804 		bytes = read(t->entry_fd, t->entry_buff, buffbytes);
805 		if (bytes < 0) {
806 			archive_set_error(&a->archive, errno, "Read error");
807 			r = ARCHIVE_FATAL;
808 			a->archive.state = ARCHIVE_STATE_FATAL;
809 			goto abort_read_data;
810 		}
811 	} else
812 		bytes = 0;
813 	/*
814 	 * Return an EOF unless we've read a leading empty sparse region, which
815 	 * is used to represent fully-sparse files.
816 	*/
817 	if (bytes == 0 && !empty_sparse_region) {
818 		/* Get EOF */
819 		t->entry_eof = 1;
820 		r = ARCHIVE_EOF;
821 		goto abort_read_data;
822 	}
823 	*buff = t->entry_buff;
824 	*size = bytes;
825 	*offset = t->entry_total;
826 	t->entry_total += bytes;
827 	t->entry_remaining_bytes -= bytes;
828 	if (t->entry_remaining_bytes == 0) {
829 		/* Close the current file descriptor */
830 		close_and_restore_time(t->entry_fd, t, &t->restore_time);
831 		t->entry_fd = -1;
832 		t->entry_eof = 1;
833 	}
834 	t->current_sparse->offset += bytes;
835 	t->current_sparse->length -= bytes;
836 	if (t->current_sparse->length == 0 && !t->entry_eof)
837 		t->current_sparse++;
838 	return (ARCHIVE_OK);
839 
840 abort_read_data:
841 	*buff = NULL;
842 	*size = 0;
843 	*offset = t->entry_total;
844 	if (t->entry_fd >= 0) {
845 		/* Close the current file descriptor */
846 		close_and_restore_time(t->entry_fd, t, &t->restore_time);
847 		t->entry_fd = -1;
848 	}
849 	return (r);
850 }
851 
852 static int
next_entry(struct archive_read_disk * a,struct tree * t,struct archive_entry * entry)853 next_entry(struct archive_read_disk *a, struct tree *t,
854     struct archive_entry *entry)
855 {
856 	const struct stat *st; /* info to use for this entry */
857 	const struct stat *lst;/* lstat() information */
858 	const char *name;
859 	int delayed, delayed_errno, descend, r;
860 	struct archive_string delayed_str;
861 
862 	delayed = ARCHIVE_OK;
863 	delayed_errno = 0;
864 	archive_string_init(&delayed_str);
865 
866 	st = NULL;
867 	lst = NULL;
868 	t->descend = 0;
869 	do {
870 		switch (tree_next(t)) {
871 		case TREE_ERROR_FATAL:
872 			archive_set_error(&a->archive, t->tree_errno,
873 			    "%s: Unable to continue traversing directory tree",
874 			    tree_current_path(t));
875 			a->archive.state = ARCHIVE_STATE_FATAL;
876 			tree_enter_initial_dir(t);
877 			return (ARCHIVE_FATAL);
878 		case TREE_ERROR_DIR:
879 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
880 			    "%s: Couldn't visit directory",
881 			    tree_current_path(t));
882 			tree_enter_initial_dir(t);
883 			return (ARCHIVE_FAILED);
884 		case 0:
885 			tree_enter_initial_dir(t);
886 			return (ARCHIVE_EOF);
887 		case TREE_POSTDESCENT:
888 		case TREE_POSTASCENT:
889 			break;
890 		case TREE_REGULAR:
891 			lst = tree_current_lstat(t);
892 			if (lst == NULL) {
893 			    if (errno == ENOENT && t->depth > 0) {
894 				delayed = ARCHIVE_WARN;
895 				delayed_errno = errno;
896 				if (delayed_str.length == 0) {
897 					archive_string_sprintf(&delayed_str,
898 					    "%s", tree_current_path(t));
899 				} else {
900 					archive_string_sprintf(&delayed_str,
901 					    " %s", tree_current_path(t));
902 				}
903 			    } else {
904 				archive_set_error(&a->archive, errno,
905 				    "%s: Cannot stat",
906 				    tree_current_path(t));
907 				tree_enter_initial_dir(t);
908 				return (ARCHIVE_FAILED);
909 			    }
910 			}
911 			break;
912 		}
913 	} while (lst == NULL);
914 
915 #ifdef __APPLE__
916 	if (a->flags & ARCHIVE_READDISK_MAC_COPYFILE) {
917 		/* If we're using copyfile(), ignore "._XXX" files. */
918 		const char *bname = strrchr(tree_current_path(t), '/');
919 		if (bname == NULL)
920 			bname = tree_current_path(t);
921 		else
922 			++bname;
923 		if (bname[0] == '.' && bname[1] == '_')
924 			return (ARCHIVE_RETRY);
925 	}
926 #endif
927 
928 	archive_entry_copy_pathname(entry, tree_current_path(t));
929 	/*
930 	 * Perform path matching.
931 	 */
932 	if (a->matching) {
933 		r = archive_match_path_excluded(a->matching, entry);
934 		if (r < 0) {
935 			archive_set_error(&(a->archive), errno,
936 			    "Failed : %s", archive_error_string(a->matching));
937 			return (r);
938 		}
939 		if (r) {
940 			if (a->excluded_cb_func)
941 				a->excluded_cb_func(&(a->archive),
942 				    a->excluded_cb_data, entry);
943 			return (ARCHIVE_RETRY);
944 		}
945 	}
946 
947 	/*
948 	 * Distinguish 'L'/'P'/'H' symlink following.
949 	 */
950 	switch(t->symlink_mode) {
951 	case 'H':
952 		/* 'H': After the first item, rest like 'P'. */
953 		t->symlink_mode = 'P';
954 		/* 'H': First item (from command line) like 'L'. */
955 		/* FALLTHROUGH */
956 	case 'L':
957 		/* 'L': Do descend through a symlink to dir. */
958 		descend = tree_current_is_dir(t);
959 		/* 'L': Follow symlinks to files. */
960 		a->symlink_mode = 'L';
961 		a->follow_symlinks = 1;
962 		/* 'L': Archive symlinks as targets, if we can. */
963 		st = tree_current_stat(t);
964 		if (st != NULL && !tree_target_is_same_as_parent(t, st))
965 			break;
966 		/* If stat fails, we have a broken symlink;
967 		 * in that case, don't follow the link. */
968 		/* FALLTHROUGH */
969 	default:
970 		/* 'P': Don't descend through a symlink to dir. */
971 		descend = tree_current_is_physical_dir(t);
972 		/* 'P': Don't follow symlinks to files. */
973 		a->symlink_mode = 'P';
974 		a->follow_symlinks = 0;
975 		/* 'P': Archive symlinks as symlinks. */
976 		st = lst;
977 		break;
978 	}
979 
980 	if (update_current_filesystem(a, st->st_dev) != ARCHIVE_OK) {
981 		a->archive.state = ARCHIVE_STATE_FATAL;
982 		tree_enter_initial_dir(t);
983 		return (ARCHIVE_FATAL);
984 	}
985 	if (t->initial_filesystem_id == -1)
986 		t->initial_filesystem_id = t->current_filesystem_id;
987 	if (a->flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS) {
988 		if (t->initial_filesystem_id != t->current_filesystem_id)
989 			descend = 0;
990 	}
991 	t->descend = descend;
992 
993 	/*
994 	 * Honor nodump flag.
995 	 * If the file is marked with nodump flag, do not return this entry.
996 	 */
997 	if (a->flags & ARCHIVE_READDISK_HONOR_NODUMP) {
998 #if defined(HAVE_STRUCT_STAT_ST_FLAGS) && defined(UF_NODUMP)
999 		if (st->st_flags & UF_NODUMP)
1000 			return (ARCHIVE_RETRY);
1001 #elif (defined(FS_IOC_GETFLAGS) && defined(FS_NODUMP_FL) && \
1002        defined(HAVE_WORKING_FS_IOC_GETFLAGS)) || \
1003       (defined(EXT2_IOC_GETFLAGS) && defined(EXT2_NODUMP_FL) && \
1004        defined(HAVE_WORKING_EXT2_IOC_GETFLAGS))
1005 		if (S_ISREG(st->st_mode) || S_ISDIR(st->st_mode)) {
1006 			int stflags;
1007 
1008 			t->entry_fd = open_on_current_dir(t,
1009 			    tree_current_access_path(t),
1010 			    O_RDONLY | O_NONBLOCK | O_CLOEXEC);
1011 			__archive_ensure_cloexec_flag(t->entry_fd);
1012 			if (t->entry_fd >= 0) {
1013 				r = ioctl(t->entry_fd,
1014 #ifdef FS_IOC_GETFLAGS
1015 				FS_IOC_GETFLAGS,
1016 #else
1017 				EXT2_IOC_GETFLAGS,
1018 #endif
1019 					&stflags);
1020 #ifdef FS_NODUMP_FL
1021 				if (r == 0 && (stflags & FS_NODUMP_FL) != 0)
1022 #else
1023 				if (r == 0 && (stflags & EXT2_NODUMP_FL) != 0)
1024 #endif
1025 					return (ARCHIVE_RETRY);
1026 			}
1027 		}
1028 #endif
1029 	}
1030 
1031 	archive_entry_copy_stat(entry, st);
1032 
1033 	/* Save the times to be restored. This must be in before
1034 	 * calling archive_read_disk_descend() or any chance of it,
1035 	 * especially, invoking a callback. */
1036 	t->restore_time.mtime = archive_entry_mtime(entry);
1037 	t->restore_time.mtime_nsec = archive_entry_mtime_nsec(entry);
1038 	t->restore_time.atime = archive_entry_atime(entry);
1039 	t->restore_time.atime_nsec = archive_entry_atime_nsec(entry);
1040 	t->restore_time.filetype = archive_entry_filetype(entry);
1041 	t->restore_time.noatime = t->current_filesystem->noatime;
1042 
1043 	/*
1044 	 * Perform time matching.
1045 	 */
1046 	if (a->matching) {
1047 		r = archive_match_time_excluded(a->matching, entry);
1048 		if (r < 0) {
1049 			archive_set_error(&(a->archive), errno,
1050 			    "Failed : %s", archive_error_string(a->matching));
1051 			return (r);
1052 		}
1053 		if (r) {
1054 			if (a->excluded_cb_func)
1055 				a->excluded_cb_func(&(a->archive),
1056 				    a->excluded_cb_data, entry);
1057 			return (ARCHIVE_RETRY);
1058 		}
1059 	}
1060 
1061 	/* Lookup uname/gname */
1062 	name = archive_read_disk_uname(&(a->archive), archive_entry_uid(entry));
1063 	if (name != NULL)
1064 		archive_entry_copy_uname(entry, name);
1065 	name = archive_read_disk_gname(&(a->archive), archive_entry_gid(entry));
1066 	if (name != NULL)
1067 		archive_entry_copy_gname(entry, name);
1068 
1069 	/*
1070 	 * Perform owner matching.
1071 	 */
1072 	if (a->matching) {
1073 		r = archive_match_owner_excluded(a->matching, entry);
1074 		if (r < 0) {
1075 			archive_set_error(&(a->archive), errno,
1076 			    "Failed : %s", archive_error_string(a->matching));
1077 			return (r);
1078 		}
1079 		if (r) {
1080 			if (a->excluded_cb_func)
1081 				a->excluded_cb_func(&(a->archive),
1082 				    a->excluded_cb_data, entry);
1083 			return (ARCHIVE_RETRY);
1084 		}
1085 	}
1086 
1087 	/*
1088 	 * Invoke a meta data filter callback.
1089 	 */
1090 	if (a->metadata_filter_func) {
1091 		if (!a->metadata_filter_func(&(a->archive),
1092 		    a->metadata_filter_data, entry))
1093 			return (ARCHIVE_RETRY);
1094 	}
1095 
1096 	/*
1097 	 * Populate the archive_entry with metadata from the disk.
1098 	 */
1099 	archive_entry_copy_sourcepath(entry, tree_current_access_path(t));
1100 	r = archive_read_disk_entry_from_file(&(a->archive), entry,
1101 		t->entry_fd, st);
1102 
1103 	if (r == ARCHIVE_OK) {
1104 		r = delayed;
1105 		if (r != ARCHIVE_OK) {
1106 			archive_string_sprintf(&delayed_str, ": %s",
1107 			    "File removed before we read it");
1108 			archive_set_error(&(a->archive), delayed_errno,
1109 			    "%s", delayed_str.s);
1110 		}
1111 	}
1112 	if (!archive_string_empty(&delayed_str))
1113 		archive_string_free(&delayed_str);
1114 
1115 	return (r);
1116 }
1117 
1118 static int
_archive_read_next_header(struct archive * _a,struct archive_entry ** entryp)1119 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
1120 {
1121 	int ret;
1122 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1123 	*entryp = NULL;
1124 	ret = _archive_read_next_header2(_a, a->entry);
1125 	*entryp = a->entry;
1126 	return ret;
1127 }
1128 
1129 static int
_archive_read_next_header2(struct archive * _a,struct archive_entry * entry)1130 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
1131 {
1132 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1133 	struct tree *t;
1134 	int r;
1135 
1136 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1137 	    ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1138 	    "archive_read_next_header2");
1139 
1140 	t = a->tree;
1141 	if (t->entry_fd >= 0) {
1142 		close_and_restore_time(t->entry_fd, t, &t->restore_time);
1143 		t->entry_fd = -1;
1144 	}
1145 
1146 	archive_entry_clear(entry);
1147 
1148 	for (;;) {
1149 		r = next_entry(a, t, entry);
1150 		if (t->entry_fd >= 0) {
1151 			close(t->entry_fd);
1152 			t->entry_fd = -1;
1153 		}
1154 
1155 		if (r == ARCHIVE_RETRY) {
1156 			archive_entry_clear(entry);
1157 			continue;
1158 		}
1159 		break;
1160 	}
1161 
1162 	/* Return to the initial directory. */
1163 	tree_enter_initial_dir(t);
1164 
1165 	/*
1166 	 * EOF and FATAL are persistent at this layer.  By
1167 	 * modifying the state, we guarantee that future calls to
1168 	 * read a header or read data will fail.
1169 	 */
1170 	switch (r) {
1171 	case ARCHIVE_EOF:
1172 		a->archive.state = ARCHIVE_STATE_EOF;
1173 		break;
1174 	case ARCHIVE_OK:
1175 	case ARCHIVE_WARN:
1176 		/* Overwrite the sourcepath based on the initial directory. */
1177 		archive_entry_copy_sourcepath(entry, tree_current_path(t));
1178 		t->entry_total = 0;
1179 		if (archive_entry_filetype(entry) == AE_IFREG) {
1180 			t->nlink = archive_entry_nlink(entry);
1181 			t->entry_remaining_bytes = archive_entry_size(entry);
1182 			t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0;
1183 			if (!t->entry_eof &&
1184 			    setup_sparse(a, entry) != ARCHIVE_OK)
1185 				return (ARCHIVE_FATAL);
1186 		} else {
1187 			t->entry_remaining_bytes = 0;
1188 			t->entry_eof = 1;
1189 		}
1190 		a->archive.state = ARCHIVE_STATE_DATA;
1191 		break;
1192 	case ARCHIVE_RETRY:
1193 		break;
1194 	case ARCHIVE_FATAL:
1195 		a->archive.state = ARCHIVE_STATE_FATAL;
1196 		break;
1197 	}
1198 
1199 	__archive_reset_read_data(&a->archive);
1200 	return (r);
1201 }
1202 
1203 static int
setup_sparse(struct archive_read_disk * a,struct archive_entry * entry)1204 setup_sparse(struct archive_read_disk *a, struct archive_entry *entry)
1205 {
1206 	struct tree *t = a->tree;
1207 	int64_t length, offset;
1208 	int i;
1209 
1210 	t->sparse_count = archive_entry_sparse_reset(entry);
1211 	if (t->sparse_count+1 > t->sparse_list_size) {
1212 		free(t->sparse_list);
1213 		t->sparse_list_size = t->sparse_count + 1;
1214 		t->sparse_list = malloc(sizeof(t->sparse_list[0]) *
1215 		    t->sparse_list_size);
1216 		if (t->sparse_list == NULL) {
1217 			t->sparse_list_size = 0;
1218 			archive_set_error(&a->archive, ENOMEM,
1219 			    "Can't allocate data");
1220 			a->archive.state = ARCHIVE_STATE_FATAL;
1221 			return (ARCHIVE_FATAL);
1222 		}
1223 	}
1224 	for (i = 0; i < t->sparse_count; i++) {
1225 		archive_entry_sparse_next(entry, &offset, &length);
1226 		t->sparse_list[i].offset = offset;
1227 		t->sparse_list[i].length = length;
1228 	}
1229 	if (i == 0) {
1230 		t->sparse_list[i].offset = 0;
1231 		t->sparse_list[i].length = archive_entry_size(entry);
1232 	} else {
1233 		t->sparse_list[i].offset = archive_entry_size(entry);
1234 		t->sparse_list[i].length = 0;
1235 	}
1236 	t->current_sparse = t->sparse_list;
1237 
1238 	return (ARCHIVE_OK);
1239 }
1240 
1241 int
archive_read_disk_set_matching(struct archive * _a,struct archive * _ma,void (* _excluded_func)(struct archive *,void *,struct archive_entry *),void * _client_data)1242 archive_read_disk_set_matching(struct archive *_a, struct archive *_ma,
1243     void (*_excluded_func)(struct archive *, void *, struct archive_entry *),
1244     void *_client_data)
1245 {
1246 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1247 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1248 	    ARCHIVE_STATE_ANY, "archive_read_disk_set_matching");
1249 	a->matching = _ma;
1250 	a->excluded_cb_func = _excluded_func;
1251 	a->excluded_cb_data = _client_data;
1252 	return (ARCHIVE_OK);
1253 }
1254 
1255 int
archive_read_disk_set_metadata_filter_callback(struct archive * _a,int (* _metadata_filter_func)(struct archive *,void *,struct archive_entry *),void * _client_data)1256 archive_read_disk_set_metadata_filter_callback(struct archive *_a,
1257     int (*_metadata_filter_func)(struct archive *, void *,
1258     struct archive_entry *), void *_client_data)
1259 {
1260 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1261 
1262 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY,
1263 	    "archive_read_disk_set_metadata_filter_callback");
1264 
1265 	a->metadata_filter_func = _metadata_filter_func;
1266 	a->metadata_filter_data = _client_data;
1267 	return (ARCHIVE_OK);
1268 }
1269 
1270 int
archive_read_disk_can_descend(struct archive * _a)1271 archive_read_disk_can_descend(struct archive *_a)
1272 {
1273 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1274 	struct tree *t = a->tree;
1275 
1276 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1277 	    ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1278 	    "archive_read_disk_can_descend");
1279 
1280 	return (t->visit_type == TREE_REGULAR && t->descend);
1281 }
1282 
1283 /*
1284  * Called by the client to mark the directory just returned from
1285  * tree_next() as needing to be visited.
1286  */
1287 int
archive_read_disk_descend(struct archive * _a)1288 archive_read_disk_descend(struct archive *_a)
1289 {
1290 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1291 	struct tree *t = a->tree;
1292 
1293 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1294 	    ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1295 	    "archive_read_disk_descend");
1296 
1297 	if (t->visit_type != TREE_REGULAR || !t->descend)
1298 		return (ARCHIVE_OK);
1299 
1300 	/*
1301 	 * We must not treat the initial specified path as a physical dir,
1302 	 * because if we do then we will try and ascend out of it by opening
1303 	 * ".." which is (a) wrong and (b) causes spurious permissions errors
1304 	 * if ".." is not readable by us. Instead, treat it as if it were a
1305 	 * symlink. (This uses an extra fd, but it can only happen once at the
1306 	 * top level of a traverse.) But we can't necessarily assume t->st is
1307 	 * valid here (though t->lst is), which complicates the logic a
1308 	 * little.
1309 	 */
1310 	if (tree_current_is_physical_dir(t)) {
1311 		tree_push(t, t->basename, t->current_filesystem_id,
1312 		    t->lst.st_dev, t->lst.st_ino, &t->restore_time);
1313 		if (t->stack->parent->parent != NULL)
1314 			t->stack->flags |= isDir;
1315 		else
1316 			t->stack->flags |= isDirLink;
1317 	} else if (tree_current_is_dir(t)) {
1318 		tree_push(t, t->basename, t->current_filesystem_id,
1319 		    t->st.st_dev, t->st.st_ino, &t->restore_time);
1320 		t->stack->flags |= isDirLink;
1321 	}
1322 	t->descend = 0;
1323 	return (ARCHIVE_OK);
1324 }
1325 
1326 int
archive_read_disk_open(struct archive * _a,const char * pathname)1327 archive_read_disk_open(struct archive *_a, const char *pathname)
1328 {
1329 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1330 
1331 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1332 	    ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1333 	    "archive_read_disk_open");
1334 	archive_clear_error(&a->archive);
1335 
1336 	return (_archive_read_disk_open(_a, pathname));
1337 }
1338 
1339 int
archive_read_disk_open_w(struct archive * _a,const wchar_t * pathname)1340 archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1341 {
1342 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1343 	struct archive_string path;
1344 	int ret;
1345 
1346 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1347 	    ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1348 	    "archive_read_disk_open_w");
1349 	archive_clear_error(&a->archive);
1350 
1351 	/* Make a char string from a wchar_t string. */
1352 	archive_string_init(&path);
1353 	if (archive_string_append_from_wcs(&path, pathname,
1354 	    wcslen(pathname)) != 0) {
1355 		if (errno == ENOMEM)
1356 			archive_set_error(&a->archive, ENOMEM,
1357 			    "Can't allocate memory");
1358 		else
1359 			archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1360 			    "Can't convert a path to a char string");
1361 		a->archive.state = ARCHIVE_STATE_FATAL;
1362 		ret = ARCHIVE_FATAL;
1363 	} else
1364 		ret = _archive_read_disk_open(_a, path.s);
1365 
1366 	archive_string_free(&path);
1367 	return (ret);
1368 }
1369 
1370 static int
_archive_read_disk_open(struct archive * _a,const char * pathname)1371 _archive_read_disk_open(struct archive *_a, const char *pathname)
1372 {
1373 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1374 
1375 	if (a->tree != NULL)
1376 		a->tree = tree_reopen(a->tree, pathname,
1377 		    a->flags & ARCHIVE_READDISK_RESTORE_ATIME);
1378 	else
1379 		a->tree = tree_open(pathname, a->symlink_mode,
1380 		    a->flags & ARCHIVE_READDISK_RESTORE_ATIME);
1381 	if (a->tree == NULL) {
1382 		archive_set_error(&a->archive, ENOMEM,
1383 		    "Can't allocate tar data");
1384 		a->archive.state = ARCHIVE_STATE_FATAL;
1385 		return (ARCHIVE_FATAL);
1386 	}
1387 	a->archive.state = ARCHIVE_STATE_HEADER;
1388 
1389 	return (ARCHIVE_OK);
1390 }
1391 
1392 /*
1393  * Return a current filesystem ID which is index of the filesystem entry
1394  * you've visited through archive_read_disk.
1395  */
1396 int
archive_read_disk_current_filesystem(struct archive * _a)1397 archive_read_disk_current_filesystem(struct archive *_a)
1398 {
1399 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1400 
1401 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1402 	    "archive_read_disk_current_filesystem");
1403 
1404 	return (a->tree->current_filesystem_id);
1405 }
1406 
1407 static int
update_current_filesystem(struct archive_read_disk * a,int64_t dev)1408 update_current_filesystem(struct archive_read_disk *a, int64_t dev)
1409 {
1410 	struct tree *t = a->tree;
1411 	int i, fid;
1412 
1413 	if (t->current_filesystem != NULL &&
1414 	    t->current_filesystem->dev == dev)
1415 		return (ARCHIVE_OK);
1416 
1417 	for (i = 0; i < t->max_filesystem_id; i++) {
1418 		if (t->filesystem_table[i].dev == dev) {
1419 			/* There is the filesystem ID we've already generated. */
1420 			t->current_filesystem_id = i;
1421 			t->current_filesystem = &(t->filesystem_table[i]);
1422 			return (ARCHIVE_OK);
1423 		}
1424 	}
1425 
1426 	/*
1427 	 * This is the new filesystem which we have to generate a new ID for.
1428 	 */
1429 	fid = t->max_filesystem_id++;
1430 	if (t->max_filesystem_id > t->allocated_filesystem) {
1431 		size_t s;
1432 		void *p;
1433 
1434 		s = t->max_filesystem_id * 2;
1435 		p = realloc(t->filesystem_table,
1436 		        s * sizeof(*t->filesystem_table));
1437 		if (p == NULL) {
1438 			archive_set_error(&a->archive, ENOMEM,
1439 			    "Can't allocate tar data");
1440 			return (ARCHIVE_FATAL);
1441 		}
1442 		t->filesystem_table = (struct filesystem *)p;
1443 		t->allocated_filesystem = s;
1444 	}
1445 	t->current_filesystem_id = fid;
1446 	t->current_filesystem = &(t->filesystem_table[fid]);
1447 	t->current_filesystem->dev = dev;
1448 	t->current_filesystem->allocation_ptr = NULL;
1449 	t->current_filesystem->buff = NULL;
1450 
1451 	/* Setup the current filesystem properties which depend on
1452 	 * platform specific. */
1453 	return (setup_current_filesystem(a));
1454 }
1455 
1456 /*
1457  * Returns 1 if current filesystem is generated filesystem, 0 if it is not
1458  * or -1 if it is unknown.
1459  */
1460 int
archive_read_disk_current_filesystem_is_synthetic(struct archive * _a)1461 archive_read_disk_current_filesystem_is_synthetic(struct archive *_a)
1462 {
1463 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1464 
1465 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1466 	    "archive_read_disk_current_filesystem");
1467 
1468 	return (a->tree->current_filesystem->synthetic);
1469 }
1470 
1471 /*
1472  * Returns 1 if current filesystem is remote filesystem, 0 if it is not
1473  * or -1 if it is unknown.
1474  */
1475 int
archive_read_disk_current_filesystem_is_remote(struct archive * _a)1476 archive_read_disk_current_filesystem_is_remote(struct archive *_a)
1477 {
1478 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
1479 
1480 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1481 	    "archive_read_disk_current_filesystem");
1482 
1483 	return (a->tree->current_filesystem->remote);
1484 }
1485 
1486 #if defined(_PC_REC_INCR_XFER_SIZE) && defined(_PC_REC_MAX_XFER_SIZE) &&\
1487 	defined(_PC_REC_MIN_XFER_SIZE) && defined(_PC_REC_XFER_ALIGN)
1488 static int
get_xfer_size(struct tree * t,int fd,const char * path)1489 get_xfer_size(struct tree *t, int fd, const char *path)
1490 {
1491 	t->current_filesystem->xfer_align = -1;
1492 	errno = 0;
1493 	if (fd >= 0) {
1494 		t->current_filesystem->incr_xfer_size =
1495 		    fpathconf(fd, _PC_REC_INCR_XFER_SIZE);
1496 		t->current_filesystem->max_xfer_size =
1497 		    fpathconf(fd, _PC_REC_MAX_XFER_SIZE);
1498 		t->current_filesystem->min_xfer_size =
1499 		    fpathconf(fd, _PC_REC_MIN_XFER_SIZE);
1500 		t->current_filesystem->xfer_align =
1501 		    fpathconf(fd, _PC_REC_XFER_ALIGN);
1502 	} else if (path != NULL) {
1503 		t->current_filesystem->incr_xfer_size =
1504 		    pathconf(path, _PC_REC_INCR_XFER_SIZE);
1505 		t->current_filesystem->max_xfer_size =
1506 		    pathconf(path, _PC_REC_MAX_XFER_SIZE);
1507 		t->current_filesystem->min_xfer_size =
1508 		    pathconf(path, _PC_REC_MIN_XFER_SIZE);
1509 		t->current_filesystem->xfer_align =
1510 		    pathconf(path, _PC_REC_XFER_ALIGN);
1511 	}
1512 	/* At least we need an alignment size. */
1513 	if (t->current_filesystem->xfer_align == -1)
1514 		return ((errno == EINVAL)?1:-1);
1515 	else
1516 		return (0);
1517 }
1518 #else
1519 static int
get_xfer_size(struct tree * t,int fd,const char * path)1520 get_xfer_size(struct tree *t, int fd, const char *path)
1521 {
1522 	(void)t; /* UNUSED */
1523 	(void)fd; /* UNUSED */
1524 	(void)path; /* UNUSED */
1525 	return (1);/* Not supported */
1526 }
1527 #endif
1528 
1529 #if defined(HAVE_STATFS) && defined(HAVE_FSTATFS) && defined(MNT_LOCAL) \
1530 	&& !defined(ST_LOCAL)
1531 
1532 /*
1533  * Gather current filesystem properties on FreeBSD, OpenBSD and Mac OS X.
1534  */
1535 static int
setup_current_filesystem(struct archive_read_disk * a)1536 setup_current_filesystem(struct archive_read_disk *a)
1537 {
1538 	struct tree *t = a->tree;
1539 	struct statfs sfs;
1540 #if defined(HAVE_GETVFSBYNAME) && defined(VFCF_SYNTHETIC)
1541 /* TODO: configure should set GETVFSBYNAME_ARG_TYPE to make
1542  * this accurate; some platforms have both and we need the one that's
1543  * used by getvfsbyname()
1544  *
1545  * Then the following would become:
1546  *  #if defined(GETVFSBYNAME_ARG_TYPE)
1547  *   GETVFSBYNAME_ARG_TYPE vfc;
1548  *  #endif
1549  */
1550 #  if defined(HAVE_STRUCT_XVFSCONF)
1551 	struct xvfsconf vfc;
1552 #  else
1553 	struct vfsconf vfc;
1554 #  endif
1555 #endif
1556 	int r, xr = 0;
1557 #if !defined(HAVE_STRUCT_STATFS_F_NAMEMAX)
1558 	long nm;
1559 #endif
1560 
1561 	t->current_filesystem->synthetic = -1;
1562 	t->current_filesystem->remote = -1;
1563 	if (tree_current_is_symblic_link_target(t)) {
1564 #if defined(HAVE_OPENAT)
1565 		/*
1566 		 * Get file system statistics on any directory
1567 		 * where current is.
1568 		 */
1569 		int fd = openat(tree_current_dir_fd(t),
1570 		    tree_current_access_path(t), O_RDONLY | O_CLOEXEC);
1571 		__archive_ensure_cloexec_flag(fd);
1572 		if (fd < 0) {
1573 			archive_set_error(&a->archive, errno,
1574 			    "openat failed");
1575 			return (ARCHIVE_FAILED);
1576 		}
1577 		r = fstatfs(fd, &sfs);
1578 		if (r == 0)
1579 			xr = get_xfer_size(t, fd, NULL);
1580 		close(fd);
1581 #else
1582 		if (tree_enter_working_dir(t) != 0) {
1583 			archive_set_error(&a->archive, errno, "fchdir failed");
1584 			return (ARCHIVE_FAILED);
1585 		}
1586 		r = statfs(tree_current_access_path(t), &sfs);
1587 		if (r == 0)
1588 			xr = get_xfer_size(t, -1, tree_current_access_path(t));
1589 #endif
1590 	} else {
1591 		r = fstatfs(tree_current_dir_fd(t), &sfs);
1592 		if (r == 0)
1593 			xr = get_xfer_size(t, tree_current_dir_fd(t), NULL);
1594 	}
1595 	if (r == -1 || xr == -1) {
1596 		archive_set_error(&a->archive, errno, "statfs failed");
1597 		return (ARCHIVE_FAILED);
1598 	} else if (xr == 1) {
1599 		/* pathconf(_PC_REX_*) operations are not supported. */
1600 		t->current_filesystem->xfer_align = sfs.f_bsize;
1601 		t->current_filesystem->max_xfer_size = -1;
1602 		t->current_filesystem->min_xfer_size = sfs.f_iosize;
1603 		t->current_filesystem->incr_xfer_size = sfs.f_iosize;
1604 	}
1605 	if (sfs.f_flags & MNT_LOCAL)
1606 		t->current_filesystem->remote = 0;
1607 	else
1608 		t->current_filesystem->remote = 1;
1609 
1610 #if defined(HAVE_GETVFSBYNAME) && defined(VFCF_SYNTHETIC)
1611 	r = getvfsbyname(sfs.f_fstypename, &vfc);
1612 	if (r == -1) {
1613 		archive_set_error(&a->archive, errno, "getvfsbyname failed");
1614 		return (ARCHIVE_FAILED);
1615 	}
1616 	if (vfc.vfc_flags & VFCF_SYNTHETIC)
1617 		t->current_filesystem->synthetic = 1;
1618 	else
1619 		t->current_filesystem->synthetic = 0;
1620 #endif
1621 
1622 #if defined(MNT_NOATIME)
1623 	if (sfs.f_flags & MNT_NOATIME)
1624 		t->current_filesystem->noatime = 1;
1625 	else
1626 #endif
1627 		t->current_filesystem->noatime = 0;
1628 
1629 #if defined(USE_READDIR_R)
1630 	/* Set maximum filename length. */
1631 #if defined(HAVE_STRUCT_STATFS_F_NAMEMAX)
1632 	t->current_filesystem->name_max = sfs.f_namemax;
1633 #else
1634 # if defined(_PC_NAME_MAX)
1635 	/* Mac OS X does not have f_namemax in struct statfs. */
1636 	if (tree_current_is_symblic_link_target(t)) {
1637 		if (tree_enter_working_dir(t) != 0) {
1638 			archive_set_error(&a->archive, errno, "fchdir failed");
1639 			return (ARCHIVE_FAILED);
1640 		}
1641 		nm = pathconf(tree_current_access_path(t), _PC_NAME_MAX);
1642 	} else
1643 		nm = fpathconf(tree_current_dir_fd(t), _PC_NAME_MAX);
1644 # else
1645 	nm = -1;
1646 # endif
1647 	if (nm == -1)
1648 		t->current_filesystem->name_max = NAME_MAX;
1649 	else
1650 		t->current_filesystem->name_max = nm;
1651 #endif
1652 #endif /* USE_READDIR_R */
1653 	return (ARCHIVE_OK);
1654 }
1655 
1656 #elif (defined(HAVE_STATVFS) || defined(HAVE_FSTATVFS)) && defined(ST_LOCAL)
1657 
1658 /*
1659  * Gather current filesystem properties on NetBSD
1660  */
1661 static int
setup_current_filesystem(struct archive_read_disk * a)1662 setup_current_filesystem(struct archive_read_disk *a)
1663 {
1664 	struct tree *t = a->tree;
1665 	struct statvfs sfs;
1666 	int r, xr = 0;
1667 
1668 	t->current_filesystem->synthetic = -1;
1669 	if (tree_enter_working_dir(t) != 0) {
1670 		archive_set_error(&a->archive, errno, "fchdir failed");
1671 		return (ARCHIVE_FAILED);
1672 	}
1673 	if (tree_current_is_symblic_link_target(t)) {
1674 		r = statvfs(tree_current_access_path(t), &sfs);
1675 		if (r == 0)
1676 			xr = get_xfer_size(t, -1, tree_current_access_path(t));
1677 	} else {
1678 #ifdef HAVE_FSTATVFS
1679 		r = fstatvfs(tree_current_dir_fd(t), &sfs);
1680 		if (r == 0)
1681 			xr = get_xfer_size(t, tree_current_dir_fd(t), NULL);
1682 #else
1683 		r = statvfs(".", &sfs);
1684 		if (r == 0)
1685 			xr = get_xfer_size(t, -1, ".");
1686 #endif
1687 	}
1688 	if (r == -1 || xr == -1) {
1689 		t->current_filesystem->remote = -1;
1690 		archive_set_error(&a->archive, errno, "statvfs failed");
1691 		return (ARCHIVE_FAILED);
1692 	} else if (xr == 1) {
1693 		/* Usually come here unless NetBSD supports _PC_REC_XFER_ALIGN
1694 		 * for pathconf() function. */
1695 		t->current_filesystem->xfer_align = sfs.f_frsize;
1696 		t->current_filesystem->max_xfer_size = -1;
1697 #if defined(HAVE_STRUCT_STATVFS_F_IOSIZE)
1698 		t->current_filesystem->min_xfer_size = sfs.f_iosize;
1699 		t->current_filesystem->incr_xfer_size = sfs.f_iosize;
1700 #else
1701 		t->current_filesystem->min_xfer_size = sfs.f_bsize;
1702 		t->current_filesystem->incr_xfer_size = sfs.f_bsize;
1703 #endif
1704 	}
1705 	if (sfs.f_flag & ST_LOCAL)
1706 		t->current_filesystem->remote = 0;
1707 	else
1708 		t->current_filesystem->remote = 1;
1709 
1710 #if defined(ST_NOATIME)
1711 	if (sfs.f_flag & ST_NOATIME)
1712 		t->current_filesystem->noatime = 1;
1713 	else
1714 #endif
1715 		t->current_filesystem->noatime = 0;
1716 
1717 	/* Set maximum filename length. */
1718 	t->current_filesystem->name_max = sfs.f_namemax;
1719 	return (ARCHIVE_OK);
1720 }
1721 
1722 #elif defined(HAVE_SYS_STATFS_H) && defined(HAVE_LINUX_MAGIC_H) &&\
1723 	defined(HAVE_STATFS) && defined(HAVE_FSTATFS)
1724 /*
1725  * Note: statfs is deprecated since LSB 3.2
1726  */
1727 
1728 #ifndef CIFS_SUPER_MAGIC
1729 #define CIFS_SUPER_MAGIC 0xFF534D42
1730 #endif
1731 #ifndef DEVFS_SUPER_MAGIC
1732 #define DEVFS_SUPER_MAGIC 0x1373
1733 #endif
1734 
1735 /*
1736  * Gather current filesystem properties on Linux
1737  */
1738 static int
setup_current_filesystem(struct archive_read_disk * a)1739 setup_current_filesystem(struct archive_read_disk *a)
1740 {
1741 	struct tree *t = a->tree;
1742 	struct statfs sfs;
1743 #if defined(HAVE_STATVFS)
1744 	struct statvfs svfs;
1745 #endif
1746 	int r, vr = 0, xr = 0;
1747 
1748 	if (tree_current_is_symblic_link_target(t)) {
1749 #if defined(HAVE_OPENAT)
1750 		/*
1751 		 * Get file system statistics on any directory
1752 		 * where current is.
1753 		 */
1754 		int fd = openat(tree_current_dir_fd(t),
1755 		    tree_current_access_path(t), O_RDONLY | O_CLOEXEC);
1756 		__archive_ensure_cloexec_flag(fd);
1757 		if (fd < 0) {
1758 			archive_set_error(&a->archive, errno,
1759 			    "openat failed");
1760 			return (ARCHIVE_FAILED);
1761 		}
1762 #if defined(HAVE_FSTATVFS)
1763 		vr = fstatvfs(fd, &svfs);/* for f_flag, mount flags */
1764 #endif
1765 		r = fstatfs(fd, &sfs);
1766 		if (r == 0)
1767 			xr = get_xfer_size(t, fd, NULL);
1768 		close(fd);
1769 #else
1770 		if (tree_enter_working_dir(t) != 0) {
1771 			archive_set_error(&a->archive, errno, "fchdir failed");
1772 			return (ARCHIVE_FAILED);
1773 		}
1774 #if defined(HAVE_STATVFS)
1775 		vr = statvfs(tree_current_access_path(t), &svfs);
1776 #endif
1777 		r = statfs(tree_current_access_path(t), &sfs);
1778 		if (r == 0)
1779 			xr = get_xfer_size(t, -1, tree_current_access_path(t));
1780 #endif
1781 	} else {
1782 #ifdef HAVE_FSTATFS
1783 #if defined(HAVE_FSTATVFS)
1784 		vr = fstatvfs(tree_current_dir_fd(t), &svfs);
1785 #endif
1786 		r = fstatfs(tree_current_dir_fd(t), &sfs);
1787 		if (r == 0)
1788 			xr = get_xfer_size(t, tree_current_dir_fd(t), NULL);
1789 #else
1790 		if (tree_enter_working_dir(t) != 0) {
1791 			archive_set_error(&a->archive, errno, "fchdir failed");
1792 			return (ARCHIVE_FAILED);
1793 		}
1794 #if defined(HAVE_STATVFS)
1795 		vr = statvfs(".", &svfs);
1796 #endif
1797 		r = statfs(".", &sfs);
1798 		if (r == 0)
1799 			xr = get_xfer_size(t, -1, ".");
1800 #endif
1801 	}
1802 	if (r == -1 || xr == -1 || vr == -1) {
1803 		t->current_filesystem->synthetic = -1;
1804 		t->current_filesystem->remote = -1;
1805 		archive_set_error(&a->archive, errno, "statfs failed");
1806 		return (ARCHIVE_FAILED);
1807 	} else if (xr == 1) {
1808 		/* pathconf(_PC_REX_*) operations are not supported. */
1809 #if defined(HAVE_STATVFS)
1810 		t->current_filesystem->xfer_align = svfs.f_frsize;
1811 		t->current_filesystem->max_xfer_size = -1;
1812 		t->current_filesystem->min_xfer_size = svfs.f_bsize;
1813 		t->current_filesystem->incr_xfer_size = svfs.f_bsize;
1814 #else
1815 		t->current_filesystem->xfer_align = sfs.f_frsize;
1816 		t->current_filesystem->max_xfer_size = -1;
1817 		t->current_filesystem->min_xfer_size = sfs.f_bsize;
1818 		t->current_filesystem->incr_xfer_size = sfs.f_bsize;
1819 #endif
1820 	}
1821 	switch (sfs.f_type) {
1822 	case AFS_SUPER_MAGIC:
1823 	case CIFS_SUPER_MAGIC:
1824 	case CODA_SUPER_MAGIC:
1825 	case NCP_SUPER_MAGIC:/* NetWare */
1826 	case NFS_SUPER_MAGIC:
1827 	case SMB_SUPER_MAGIC:
1828 		t->current_filesystem->remote = 1;
1829 		t->current_filesystem->synthetic = 0;
1830 		break;
1831 	case DEVFS_SUPER_MAGIC:
1832 	case PROC_SUPER_MAGIC:
1833 	case USBDEVICE_SUPER_MAGIC:
1834 		t->current_filesystem->remote = 0;
1835 		t->current_filesystem->synthetic = 1;
1836 		break;
1837 	default:
1838 		t->current_filesystem->remote = 0;
1839 		t->current_filesystem->synthetic = 0;
1840 		break;
1841 	}
1842 
1843 #if defined(ST_NOATIME)
1844 #if defined(HAVE_STATVFS)
1845 	if (svfs.f_flag & ST_NOATIME)
1846 #else
1847 	if (sfs.f_flag & ST_NOATIME)
1848 #endif
1849 		t->current_filesystem->noatime = 1;
1850 	else
1851 #endif
1852 		t->current_filesystem->noatime = 0;
1853 
1854 #if defined(USE_READDIR_R)
1855 	/* Set maximum filename length. */
1856 	t->current_filesystem->name_max = sfs.f_namelen;
1857 #endif
1858 	return (ARCHIVE_OK);
1859 }
1860 
1861 #elif defined(HAVE_SYS_STATVFS_H) &&\
1862 	(defined(HAVE_STATVFS) || defined(HAVE_FSTATVFS))
1863 
1864 /*
1865  * Gather current filesystem properties on other posix platform.
1866  */
1867 static int
setup_current_filesystem(struct archive_read_disk * a)1868 setup_current_filesystem(struct archive_read_disk *a)
1869 {
1870 	struct tree *t = a->tree;
1871 	struct statvfs sfs;
1872 	int r, xr = 0;
1873 
1874 	t->current_filesystem->synthetic = -1;/* Not supported */
1875 	t->current_filesystem->remote = -1;/* Not supported */
1876 	if (tree_current_is_symblic_link_target(t)) {
1877 #if defined(HAVE_OPENAT)
1878 		/*
1879 		 * Get file system statistics on any directory
1880 		 * where current is.
1881 		 */
1882 		int fd = openat(tree_current_dir_fd(t),
1883 		    tree_current_access_path(t), O_RDONLY | O_CLOEXEC);
1884 		__archive_ensure_cloexec_flag(fd);
1885 		if (fd < 0) {
1886 			archive_set_error(&a->archive, errno,
1887 			    "openat failed");
1888 			return (ARCHIVE_FAILED);
1889 		}
1890 		r = fstatvfs(fd, &sfs);
1891 		if (r == 0)
1892 			xr = get_xfer_size(t, fd, NULL);
1893 		close(fd);
1894 #else
1895 		if (tree_enter_working_dir(t) != 0) {
1896 			archive_set_error(&a->archive, errno, "fchdir failed");
1897 			return (ARCHIVE_FAILED);
1898 		}
1899 		r = statvfs(tree_current_access_path(t), &sfs);
1900 		if (r == 0)
1901 			xr = get_xfer_size(t, -1, tree_current_access_path(t));
1902 #endif
1903 	} else {
1904 #ifdef HAVE_FSTATVFS
1905 		r = fstatvfs(tree_current_dir_fd(t), &sfs);
1906 		if (r == 0)
1907 			xr = get_xfer_size(t, tree_current_dir_fd(t), NULL);
1908 #else
1909 		if (tree_enter_working_dir(t) != 0) {
1910 			archive_set_error(&a->archive, errno, "fchdir failed");
1911 			return (ARCHIVE_FAILED);
1912 		}
1913 		r = statvfs(".", &sfs);
1914 		if (r == 0)
1915 			xr = get_xfer_size(t, -1, ".");
1916 #endif
1917 	}
1918 	if (r == -1 || xr == -1) {
1919 		t->current_filesystem->synthetic = -1;
1920 		t->current_filesystem->remote = -1;
1921 		archive_set_error(&a->archive, errno, "statvfs failed");
1922 		return (ARCHIVE_FAILED);
1923 	} else if (xr == 1) {
1924 		/* pathconf(_PC_REX_*) operations are not supported. */
1925 		t->current_filesystem->xfer_align = sfs.f_frsize;
1926 		t->current_filesystem->max_xfer_size = -1;
1927 		t->current_filesystem->min_xfer_size = sfs.f_bsize;
1928 		t->current_filesystem->incr_xfer_size = sfs.f_bsize;
1929 	}
1930 
1931 #if defined(ST_NOATIME)
1932 	if (sfs.f_flag & ST_NOATIME)
1933 		t->current_filesystem->noatime = 1;
1934 	else
1935 #endif
1936 		t->current_filesystem->noatime = 0;
1937 
1938 #if defined(USE_READDIR_R)
1939 	/* Set maximum filename length. */
1940 	t->current_filesystem->name_max = sfs.f_namemax;
1941 #endif
1942 	return (ARCHIVE_OK);
1943 }
1944 
1945 #else
1946 
1947 /*
1948  * Generic: Gather current filesystem properties.
1949  * TODO: Is this generic function really needed?
1950  */
1951 static int
setup_current_filesystem(struct archive_read_disk * a)1952 setup_current_filesystem(struct archive_read_disk *a)
1953 {
1954 	struct tree *t = a->tree;
1955 #if defined(_PC_NAME_MAX) && defined(USE_READDIR_R)
1956 	long nm;
1957 #endif
1958 	t->current_filesystem->synthetic = -1;/* Not supported */
1959 	t->current_filesystem->remote = -1;/* Not supported */
1960 	t->current_filesystem->noatime = 0;
1961 	(void)get_xfer_size(t, -1, ".");/* Dummy call to avoid build error. */
1962 	t->current_filesystem->xfer_align = -1;/* Unknown */
1963 	t->current_filesystem->max_xfer_size = -1;
1964 	t->current_filesystem->min_xfer_size = -1;
1965 	t->current_filesystem->incr_xfer_size = -1;
1966 
1967 #if defined(USE_READDIR_R)
1968 	/* Set maximum filename length. */
1969 #  if defined(_PC_NAME_MAX)
1970 	if (tree_current_is_symblic_link_target(t)) {
1971 		if (tree_enter_working_dir(t) != 0) {
1972 			archive_set_error(&a->archive, errno, "fchdir failed");
1973 			return (ARCHIVE_FAILED);
1974 		}
1975 		nm = pathconf(tree_current_access_path(t), _PC_NAME_MAX);
1976 	} else
1977 		nm = fpathconf(tree_current_dir_fd(t), _PC_NAME_MAX);
1978 	if (nm == -1)
1979 #  endif /* _PC_NAME_MAX */
1980 		/*
1981 		 * Some systems (HP-UX or others?) incorrectly defined
1982 		 * NAME_MAX macro to be a smaller value.
1983 		 */
1984 #  if defined(NAME_MAX) && NAME_MAX >= 255
1985 		t->current_filesystem->name_max = NAME_MAX;
1986 #  else
1987 		/* No way to get a trusted value of maximum filename
1988 		 * length. */
1989 		t->current_filesystem->name_max = PATH_MAX;
1990 #  endif /* NAME_MAX */
1991 #  if defined(_PC_NAME_MAX)
1992 	else
1993 		t->current_filesystem->name_max = nm;
1994 #  endif /* _PC_NAME_MAX */
1995 #endif /* USE_READDIR_R */
1996 	return (ARCHIVE_OK);
1997 }
1998 
1999 #endif
2000 
2001 static int
close_and_restore_time(int fd,struct tree * t,struct restore_time * rt)2002 close_and_restore_time(int fd, struct tree *t, struct restore_time *rt)
2003 {
2004 #ifndef HAVE_UTIMES
2005 	(void)t; /* UNUSED */
2006 	(void)rt; /* UNUSED */
2007 	return (close(fd));
2008 #else
2009 #if defined(HAVE_FUTIMENS) && !defined(__CYGWIN__)
2010 	struct timespec timespecs[2];
2011 #endif
2012 	struct timeval times[2];
2013 
2014 	if ((t->flags & needsRestoreTimes) == 0 || rt->noatime) {
2015 		if (fd >= 0)
2016 			return (close(fd));
2017 		else
2018 			return (0);
2019 	}
2020 
2021 #if defined(HAVE_FUTIMENS) && !defined(__CYGWIN__)
2022 	timespecs[1].tv_sec = rt->mtime;
2023 	timespecs[1].tv_nsec = rt->mtime_nsec;
2024 
2025 	timespecs[0].tv_sec = rt->atime;
2026 	timespecs[0].tv_nsec = rt->atime_nsec;
2027 	/* futimens() is defined in POSIX.1-2008. */
2028 	if (futimens(fd, timespecs) == 0)
2029 		return (close(fd));
2030 #endif
2031 
2032 	times[1].tv_sec = rt->mtime;
2033 	times[1].tv_usec = rt->mtime_nsec / 1000;
2034 
2035 	times[0].tv_sec = rt->atime;
2036 	times[0].tv_usec = rt->atime_nsec / 1000;
2037 
2038 #if !defined(HAVE_FUTIMENS) && defined(HAVE_FUTIMES) && !defined(__CYGWIN__)
2039 	if (futimes(fd, times) == 0)
2040 		return (close(fd));
2041 #endif
2042 	close(fd);
2043 #if defined(HAVE_FUTIMESAT)
2044 	if (futimesat(tree_current_dir_fd(t), rt->name, times) == 0)
2045 		return (0);
2046 #endif
2047 #ifdef HAVE_LUTIMES
2048 	if (lutimes(rt->name, times) != 0)
2049 #else
2050 	if (AE_IFLNK != rt->filetype && utimes(rt->name, times) != 0)
2051 #endif
2052 		return (-1);
2053 #endif
2054 	return (0);
2055 }
2056 
2057 static int
open_on_current_dir(struct tree * t,const char * path,int flags)2058 open_on_current_dir(struct tree *t, const char *path, int flags)
2059 {
2060 #ifdef HAVE_OPENAT
2061 	return (openat(tree_current_dir_fd(t), path, flags));
2062 #else
2063 	if (tree_enter_working_dir(t) != 0)
2064 		return (-1);
2065 	return (open(path, flags));
2066 #endif
2067 }
2068 
2069 static int
tree_dup(int fd)2070 tree_dup(int fd)
2071 {
2072 	int new_fd;
2073 #ifdef F_DUPFD_CLOEXEC
2074 	static volatile int can_dupfd_cloexec = 1;
2075 
2076 	if (can_dupfd_cloexec) {
2077 		new_fd = fcntl(fd, F_DUPFD_CLOEXEC, 0);
2078 		if (new_fd != -1)
2079 			return (new_fd);
2080 		/* Linux 2.6.18 - 2.6.23 declare F_DUPFD_CLOEXEC,
2081 		 * but it cannot be used. So we have to try dup(). */
2082 		/* We won't try F_DUPFD_CLOEXEC. */
2083 		can_dupfd_cloexec = 0;
2084 	}
2085 #endif /* F_DUPFD_CLOEXEC */
2086 	new_fd = dup(fd);
2087 	__archive_ensure_cloexec_flag(new_fd);
2088 	return (new_fd);
2089 }
2090 
2091 /*
2092  * Add a directory path to the current stack.
2093  */
2094 static void
tree_push(struct tree * t,const char * path,int filesystem_id,int64_t dev,int64_t ino,struct restore_time * rt)2095 tree_push(struct tree *t, const char *path, int filesystem_id,
2096     int64_t dev, int64_t ino, struct restore_time *rt)
2097 {
2098 	struct tree_entry *te;
2099 
2100 	te = calloc(1, sizeof(*te));
2101 	te->next = t->stack;
2102 	te->parent = t->current;
2103 	if (te->parent)
2104 		te->depth = te->parent->depth + 1;
2105 	t->stack = te;
2106 	archive_string_init(&te->name);
2107 	te->symlink_parent_fd = -1;
2108 	archive_strcpy(&te->name, path);
2109 	te->flags = needsDescent | needsOpen | needsAscent;
2110 	te->filesystem_id = filesystem_id;
2111 	te->dev = dev;
2112 	te->ino = ino;
2113 	te->dirname_length = t->dirname_length;
2114 	te->restore_time.name = te->name.s;
2115 	if (rt != NULL) {
2116 		te->restore_time.mtime = rt->mtime;
2117 		te->restore_time.mtime_nsec = rt->mtime_nsec;
2118 		te->restore_time.atime = rt->atime;
2119 		te->restore_time.atime_nsec = rt->atime_nsec;
2120 		te->restore_time.filetype = rt->filetype;
2121 		te->restore_time.noatime = rt->noatime;
2122 	}
2123 }
2124 
2125 /*
2126  * Append a name to the current dir path.
2127  */
2128 static void
tree_append(struct tree * t,const char * name,size_t name_length)2129 tree_append(struct tree *t, const char *name, size_t name_length)
2130 {
2131 	size_t size_needed;
2132 
2133 	t->path.s[t->dirname_length] = '\0';
2134 	t->path.length = t->dirname_length;
2135 	/* Strip trailing '/' from name, unless entire name is "/". */
2136 	while (name_length > 1 && name[name_length - 1] == '/')
2137 		name_length--;
2138 
2139 	/* Resize pathname buffer as needed. */
2140 	size_needed = name_length + t->dirname_length + 2;
2141 	archive_string_ensure(&t->path, size_needed);
2142 	/* Add a separating '/' if it's needed. */
2143 	if (t->dirname_length > 0 && t->path.s[archive_strlen(&t->path)-1] != '/')
2144 		archive_strappend_char(&t->path, '/');
2145 	t->basename = t->path.s + archive_strlen(&t->path);
2146 	archive_strncat(&t->path, name, name_length);
2147 	t->restore_time.name = t->basename;
2148 }
2149 
2150 /*
2151  * Open a directory tree for traversal.
2152  */
2153 static struct tree *
tree_open(const char * path,int symlink_mode,int restore_time)2154 tree_open(const char *path, int symlink_mode, int restore_time)
2155 {
2156 	struct tree *t;
2157 
2158 	if ((t = calloc(1, sizeof(*t))) == NULL)
2159 		return (NULL);
2160 	archive_string_init(&t->path);
2161 	archive_string_ensure(&t->path, 31);
2162 	t->initial_symlink_mode = symlink_mode;
2163 	return (tree_reopen(t, path, restore_time));
2164 }
2165 
2166 static struct tree *
tree_reopen(struct tree * t,const char * path,int restore_time)2167 tree_reopen(struct tree *t, const char *path, int restore_time)
2168 {
2169 #if defined(O_PATH)
2170 	/* Linux */
2171 	const int o_flag = O_PATH;
2172 #elif defined(O_SEARCH)
2173 	/* SunOS */
2174 	const int o_flag = O_SEARCH;
2175 #elif defined(O_EXEC)
2176 	/* FreeBSD */
2177 	const int o_flag = O_EXEC;
2178 #endif
2179 
2180 	t->flags = (restore_time != 0)?needsRestoreTimes:0;
2181 	t->flags |= onInitialDir;
2182 	t->visit_type = 0;
2183 	t->tree_errno = 0;
2184 	t->dirname_length = 0;
2185 	t->depth = 0;
2186 	t->descend = 0;
2187 	t->current = NULL;
2188 	t->d = INVALID_DIR_HANDLE;
2189 	t->symlink_mode = t->initial_symlink_mode;
2190 	archive_string_empty(&t->path);
2191 	t->entry_fd = -1;
2192 	t->entry_eof = 0;
2193 	t->entry_remaining_bytes = 0;
2194 	t->initial_filesystem_id = -1;
2195 
2196 	/* First item is set up a lot like a symlink traversal. */
2197 	tree_push(t, path, 0, 0, 0, NULL);
2198 	t->stack->flags = needsFirstVisit;
2199 	t->maxOpenCount = t->openCount = 1;
2200 	t->initial_dir_fd = open(".", O_RDONLY | O_CLOEXEC);
2201 #if defined(O_PATH) || defined(O_SEARCH) || defined(O_EXEC)
2202 	/*
2203 	 * Most likely reason to fail opening "." is that it's not readable,
2204 	 * so try again for execute. The consequences of not opening this are
2205 	 * unhelpful and unnecessary errors later.
2206 	 */
2207 	if (t->initial_dir_fd < 0)
2208 		t->initial_dir_fd = open(".", o_flag | O_CLOEXEC);
2209 #endif
2210 	__archive_ensure_cloexec_flag(t->initial_dir_fd);
2211 	t->working_dir_fd = tree_dup(t->initial_dir_fd);
2212 	return (t);
2213 }
2214 
2215 static int
tree_descent(struct tree * t)2216 tree_descent(struct tree *t)
2217 {
2218 	int flag, new_fd, r = 0;
2219 
2220 	t->dirname_length = archive_strlen(&t->path);
2221 	flag = O_RDONLY | O_CLOEXEC;
2222 #if defined(O_DIRECTORY)
2223 	flag |= O_DIRECTORY;
2224 #endif
2225 	new_fd = open_on_current_dir(t, t->stack->name.s, flag);
2226 	__archive_ensure_cloexec_flag(new_fd);
2227 	if (new_fd < 0) {
2228 		t->tree_errno = errno;
2229 		r = TREE_ERROR_DIR;
2230 	} else {
2231 		t->depth++;
2232 		/* If it is a link, set up fd for the ascent. */
2233 		if (t->stack->flags & isDirLink) {
2234 			t->stack->symlink_parent_fd = t->working_dir_fd;
2235 			t->openCount++;
2236 			if (t->openCount > t->maxOpenCount)
2237 				t->maxOpenCount = t->openCount;
2238 		} else
2239 			close(t->working_dir_fd);
2240 		/* Renew the current working directory. */
2241 		t->working_dir_fd = new_fd;
2242 		t->flags &= ~onWorkingDir;
2243 	}
2244 	return (r);
2245 }
2246 
2247 /*
2248  * We've finished a directory; ascend back to the parent.
2249  */
2250 static int
tree_ascend(struct tree * t)2251 tree_ascend(struct tree *t)
2252 {
2253 	struct tree_entry *te;
2254 	int new_fd, r = 0, prev_dir_fd;
2255 
2256 	te = t->stack;
2257 	prev_dir_fd = t->working_dir_fd;
2258 	if (te->flags & isDirLink)
2259 		new_fd = te->symlink_parent_fd;
2260 	else {
2261 		new_fd = open_on_current_dir(t, "..", O_RDONLY | O_CLOEXEC);
2262 		__archive_ensure_cloexec_flag(new_fd);
2263 	}
2264 	if (new_fd < 0) {
2265 		t->tree_errno = errno;
2266 		r = TREE_ERROR_FATAL;
2267 	} else {
2268 		/* Renew the current working directory. */
2269 		t->working_dir_fd = new_fd;
2270 		t->flags &= ~onWorkingDir;
2271 		/* Current directory has been changed, we should
2272 		 * close an fd of previous working directory. */
2273 		close_and_restore_time(prev_dir_fd, t, &te->restore_time);
2274 		if (te->flags & isDirLink) {
2275 			t->openCount--;
2276 			te->symlink_parent_fd = -1;
2277 		}
2278 		t->depth--;
2279 	}
2280 	return (r);
2281 }
2282 
2283 /*
2284  * Return to the initial directory where tree_open() was performed.
2285  */
2286 static int
tree_enter_initial_dir(struct tree * t)2287 tree_enter_initial_dir(struct tree *t)
2288 {
2289 	int r = 0;
2290 
2291 	if ((t->flags & onInitialDir) == 0) {
2292 		r = fchdir(t->initial_dir_fd);
2293 		if (r == 0) {
2294 			t->flags &= ~onWorkingDir;
2295 			t->flags |= onInitialDir;
2296 		}
2297 	}
2298 	return (r);
2299 }
2300 
2301 /*
2302  * Restore working directory of directory traversals.
2303  */
2304 static int
tree_enter_working_dir(struct tree * t)2305 tree_enter_working_dir(struct tree *t)
2306 {
2307 	int r = 0;
2308 
2309 	/*
2310 	 * Change the current directory if really needed.
2311 	 * Sometimes this is unneeded when we did not do
2312 	 * descent.
2313 	 */
2314 	if (t->depth > 0 && (t->flags & onWorkingDir) == 0) {
2315 		r = fchdir(t->working_dir_fd);
2316 		if (r == 0) {
2317 			t->flags &= ~onInitialDir;
2318 			t->flags |= onWorkingDir;
2319 		}
2320 	}
2321 	return (r);
2322 }
2323 
2324 static int
tree_current_dir_fd(struct tree * t)2325 tree_current_dir_fd(struct tree *t)
2326 {
2327 	return (t->working_dir_fd);
2328 }
2329 
2330 /*
2331  * Pop the working stack.
2332  */
2333 static void
tree_pop(struct tree * t)2334 tree_pop(struct tree *t)
2335 {
2336 	struct tree_entry *te;
2337 
2338 	t->path.s[t->dirname_length] = '\0';
2339 	t->path.length = t->dirname_length;
2340 	if (t->stack == t->current && t->current != NULL)
2341 		t->current = t->current->parent;
2342 	te = t->stack;
2343 	t->stack = te->next;
2344 	t->dirname_length = te->dirname_length;
2345 	t->basename = t->path.s + t->dirname_length;
2346 	while (t->basename[0] == '/')
2347 		t->basename++;
2348 	archive_string_free(&te->name);
2349 	free(te);
2350 }
2351 
2352 /*
2353  * Get the next item in the tree traversal.
2354  */
2355 static int
tree_next(struct tree * t)2356 tree_next(struct tree *t)
2357 {
2358 	int r;
2359 
2360 	while (t->stack != NULL) {
2361 		/* If there's an open dir, get the next entry from there. */
2362 		if (t->d != INVALID_DIR_HANDLE) {
2363 			r = tree_dir_next_posix(t);
2364 			if (r == 0)
2365 				continue;
2366 			return (r);
2367 		}
2368 
2369 		if (t->stack->flags & needsFirstVisit) {
2370 			/* Top stack item needs a regular visit. */
2371 			t->current = t->stack;
2372 			tree_append(t, t->stack->name.s,
2373 			    archive_strlen(&(t->stack->name)));
2374 			/* t->dirname_length = t->path_length; */
2375 			/* tree_pop(t); */
2376 			t->stack->flags &= ~needsFirstVisit;
2377 			return (t->visit_type = TREE_REGULAR);
2378 		} else if (t->stack->flags & needsDescent) {
2379 			/* Top stack item is dir to descend into. */
2380 			t->current = t->stack;
2381 			tree_append(t, t->stack->name.s,
2382 			    archive_strlen(&(t->stack->name)));
2383 			t->stack->flags &= ~needsDescent;
2384 			r = tree_descent(t);
2385 			if (r != 0) {
2386 				tree_pop(t);
2387 				t->visit_type = r;
2388 			} else
2389 				t->visit_type = TREE_POSTDESCENT;
2390 			return (t->visit_type);
2391 		} else if (t->stack->flags & needsOpen) {
2392 			t->stack->flags &= ~needsOpen;
2393 			r = tree_dir_next_posix(t);
2394 			if (r == 0)
2395 				continue;
2396 			return (r);
2397 		} else if (t->stack->flags & needsAscent) {
2398 		        /* Top stack item is dir and we're done with it. */
2399 			r = tree_ascend(t);
2400 			tree_pop(t);
2401 			t->visit_type = r != 0 ? r : TREE_POSTASCENT;
2402 			return (t->visit_type);
2403 		} else {
2404 			/* Top item on stack is dead. */
2405 			tree_pop(t);
2406 			t->flags &= ~hasLstat;
2407 			t->flags &= ~hasStat;
2408 		}
2409 	}
2410 	return (t->visit_type = 0);
2411 }
2412 
2413 static int
tree_dir_next_posix(struct tree * t)2414 tree_dir_next_posix(struct tree *t)
2415 {
2416 	int r;
2417 	const char *name;
2418 	size_t namelen;
2419 
2420 	if (t->d == NULL) {
2421 #if defined(USE_READDIR_R)
2422 		size_t dirent_size;
2423 #endif
2424 
2425 #if defined(HAVE_FDOPENDIR)
2426 		t->d = fdopendir(tree_dup(t->working_dir_fd));
2427 #else /* HAVE_FDOPENDIR */
2428 		if (tree_enter_working_dir(t) == 0) {
2429 			t->d = opendir(".");
2430 #if HAVE_DIRFD || defined(dirfd)
2431 			__archive_ensure_cloexec_flag(dirfd(t->d));
2432 #endif
2433 		}
2434 #endif /* HAVE_FDOPENDIR */
2435 		if (t->d == NULL) {
2436 			r = tree_ascend(t); /* Undo "chdir" */
2437 			tree_pop(t);
2438 			t->tree_errno = errno;
2439 			t->visit_type = r != 0 ? r : TREE_ERROR_DIR;
2440 			return (t->visit_type);
2441 		}
2442 #if defined(USE_READDIR_R)
2443 		dirent_size = offsetof(struct dirent, d_name) +
2444 		  t->filesystem_table[t->current->filesystem_id].name_max + 1;
2445 		if (t->dirent == NULL || t->dirent_allocated < dirent_size) {
2446 			free(t->dirent);
2447 			t->dirent = malloc(dirent_size);
2448 			if (t->dirent == NULL) {
2449 				closedir(t->d);
2450 				t->d = INVALID_DIR_HANDLE;
2451 				(void)tree_ascend(t);
2452 				tree_pop(t);
2453 				t->tree_errno = ENOMEM;
2454 				t->visit_type = TREE_ERROR_DIR;
2455 				return (t->visit_type);
2456 			}
2457 			t->dirent_allocated = dirent_size;
2458 		}
2459 #endif /* USE_READDIR_R */
2460 	}
2461 	for (;;) {
2462 		errno = 0;
2463 #if defined(USE_READDIR_R)
2464 		r = readdir_r(t->d, t->dirent, &t->de);
2465 #ifdef _AIX
2466 		/* Note: According to the man page, return value 9 indicates
2467 		 * that the readdir_r was not successful and the error code
2468 		 * is set to the global errno variable. And then if the end
2469 		 * of directory entries was reached, the return value is 9
2470 		 * and the third parameter is set to NULL and errno is
2471 		 * unchanged. */
2472 		if (r == 9)
2473 			r = errno;
2474 #endif /* _AIX */
2475 		if (r != 0 || t->de == NULL) {
2476 #else
2477 		t->de = readdir(t->d);
2478 		if (t->de == NULL) {
2479 			r = errno;
2480 #endif
2481 			closedir(t->d);
2482 			t->d = INVALID_DIR_HANDLE;
2483 			if (r != 0) {
2484 				t->tree_errno = r;
2485 				t->visit_type = TREE_ERROR_DIR;
2486 				return (t->visit_type);
2487 			} else
2488 				return (0);
2489 		}
2490 		name = t->de->d_name;
2491 		namelen = D_NAMELEN(t->de);
2492 		t->flags &= ~hasLstat;
2493 		t->flags &= ~hasStat;
2494 		if (name[0] == '.' && name[1] == '\0')
2495 			continue;
2496 		if (name[0] == '.' && name[1] == '.' && name[2] == '\0')
2497 			continue;
2498 		tree_append(t, name, namelen);
2499 		return (t->visit_type = TREE_REGULAR);
2500 	}
2501 }
2502 
2503 
2504 /*
2505  * Get the stat() data for the entry just returned from tree_next().
2506  */
2507 static const struct stat *
2508 tree_current_stat(struct tree *t)
2509 {
2510 	if (!(t->flags & hasStat)) {
2511 #ifdef HAVE_FSTATAT
2512 		if (fstatat(tree_current_dir_fd(t),
2513 		    tree_current_access_path(t), &t->st, 0) != 0)
2514 #else
2515 		if (tree_enter_working_dir(t) != 0)
2516 			return NULL;
2517 		if (la_stat(tree_current_access_path(t), &t->st) != 0)
2518 #endif
2519 			return NULL;
2520 		t->flags |= hasStat;
2521 	}
2522 	return (&t->st);
2523 }
2524 
2525 /*
2526  * Get the lstat() data for the entry just returned from tree_next().
2527  */
2528 static const struct stat *
2529 tree_current_lstat(struct tree *t)
2530 {
2531 	if (!(t->flags & hasLstat)) {
2532 #ifdef HAVE_FSTATAT
2533 		if (fstatat(tree_current_dir_fd(t),
2534 		    tree_current_access_path(t), &t->lst,
2535 		    AT_SYMLINK_NOFOLLOW) != 0)
2536 #else
2537 		if (tree_enter_working_dir(t) != 0)
2538 			return NULL;
2539 		if (lstat(tree_current_access_path(t), &t->lst) != 0)
2540 #endif
2541 			return NULL;
2542 		t->flags |= hasLstat;
2543 	}
2544 	return (&t->lst);
2545 }
2546 
2547 /*
2548  * Test whether current entry is a dir or link to a dir.
2549  */
2550 static int
2551 tree_current_is_dir(struct tree *t)
2552 {
2553 	const struct stat *st;
2554 	/*
2555 	 * If we already have lstat() info, then try some
2556 	 * cheap tests to determine if this is a dir.
2557 	 */
2558 	if (t->flags & hasLstat) {
2559 		/* If lstat() says it's a dir, it must be a dir. */
2560 		st = tree_current_lstat(t);
2561 		if (st == NULL)
2562 			return 0;
2563 		if (S_ISDIR(st->st_mode))
2564 			return 1;
2565 		/* Not a dir; might be a link to a dir. */
2566 		/* If it's not a link, then it's not a link to a dir. */
2567 		if (!S_ISLNK(st->st_mode))
2568 			return 0;
2569 		/*
2570 		 * It's a link, but we don't know what it's a link to,
2571 		 * so we'll have to use stat().
2572 		 */
2573 	}
2574 
2575 	st = tree_current_stat(t);
2576 	/* If we can't stat it, it's not a dir. */
2577 	if (st == NULL)
2578 		return 0;
2579 	/* Use the definitive test.  Hopefully this is cached. */
2580 	return (S_ISDIR(st->st_mode));
2581 }
2582 
2583 /*
2584  * Test whether current entry is a physical directory.  Usually, we
2585  * already have at least one of stat() or lstat() in memory, so we
2586  * use tricks to try to avoid an extra trip to the disk.
2587  */
2588 static int
2589 tree_current_is_physical_dir(struct tree *t)
2590 {
2591 	const struct stat *st;
2592 
2593 	/*
2594 	 * If stat() says it isn't a dir, then it's not a dir.
2595 	 * If stat() data is cached, this check is free, so do it first.
2596 	 */
2597 	if (t->flags & hasStat) {
2598 		st = tree_current_stat(t);
2599 		if (st == NULL)
2600 			return (0);
2601 		if (!S_ISDIR(st->st_mode))
2602 			return (0);
2603 	}
2604 
2605 	/*
2606 	 * Either stat() said it was a dir (in which case, we have
2607 	 * to determine whether it's really a link to a dir) or
2608 	 * stat() info wasn't available.  So we use lstat(), which
2609 	 * hopefully is already cached.
2610 	 */
2611 
2612 	st = tree_current_lstat(t);
2613 	/* If we can't stat it, it's not a dir. */
2614 	if (st == NULL)
2615 		return 0;
2616 	/* Use the definitive test.  Hopefully this is cached. */
2617 	return (S_ISDIR(st->st_mode));
2618 }
2619 
2620 /*
2621  * Test whether the same file has been in the tree as its parent.
2622  */
2623 static int
2624 tree_target_is_same_as_parent(struct tree *t, const struct stat *st)
2625 {
2626 	struct tree_entry *te;
2627 
2628 	for (te = t->current->parent; te != NULL; te = te->parent) {
2629 		if (te->dev == (int64_t)st->st_dev &&
2630 		    te->ino == (int64_t)st->st_ino)
2631 			return (1);
2632 	}
2633 	return (0);
2634 }
2635 
2636 /*
2637  * Test whether the current file is symbolic link target and
2638  * on the other filesystem.
2639  */
2640 static int
2641 tree_current_is_symblic_link_target(struct tree *t)
2642 {
2643 	static const struct stat *lst, *st;
2644 
2645 	lst = tree_current_lstat(t);
2646 	st = tree_current_stat(t);
2647 	return (st != NULL && lst != NULL &&
2648 	    (int64_t)st->st_dev == t->current_filesystem->dev &&
2649 	    st->st_dev != lst->st_dev);
2650 }
2651 
2652 /*
2653  * Return the access path for the entry just returned from tree_next().
2654  */
2655 static const char *
2656 tree_current_access_path(struct tree *t)
2657 {
2658 	return (t->basename);
2659 }
2660 
2661 /*
2662  * Return the full path for the entry just returned from tree_next().
2663  */
2664 static const char *
2665 tree_current_path(struct tree *t)
2666 {
2667 	return (t->path.s);
2668 }
2669 
2670 /*
2671  * Terminate the traversal.
2672  */
2673 static void
2674 tree_close(struct tree *t)
2675 {
2676 
2677 	if (t == NULL)
2678 		return;
2679 	if (t->entry_fd >= 0) {
2680 		close_and_restore_time(t->entry_fd, t, &t->restore_time);
2681 		t->entry_fd = -1;
2682 	}
2683 	/* Close the handle of readdir(). */
2684 	if (t->d != INVALID_DIR_HANDLE) {
2685 		closedir(t->d);
2686 		t->d = INVALID_DIR_HANDLE;
2687 	}
2688 	/* Release anything remaining in the stack. */
2689 	while (t->stack != NULL) {
2690 		if (t->stack->flags & isDirLink)
2691 			close(t->stack->symlink_parent_fd);
2692 		tree_pop(t);
2693 	}
2694 	if (t->working_dir_fd >= 0) {
2695 		close(t->working_dir_fd);
2696 		t->working_dir_fd = -1;
2697 	}
2698 	if (t->initial_dir_fd >= 0) {
2699 		close(t->initial_dir_fd);
2700 		t->initial_dir_fd = -1;
2701 	}
2702 }
2703 
2704 /*
2705  * Release any resources.
2706  */
2707 static void
2708 tree_free(struct tree *t)
2709 {
2710 	int i;
2711 
2712 	if (t == NULL)
2713 		return;
2714 	archive_string_free(&t->path);
2715 #if defined(USE_READDIR_R)
2716 	free(t->dirent);
2717 #endif
2718 	free(t->sparse_list);
2719 	for (i = 0; i < t->max_filesystem_id; i++)
2720 		free(t->filesystem_table[i].allocation_ptr);
2721 	free(t->filesystem_table);
2722 	free(t);
2723 }
2724 
2725 #endif
2726