1 /*-
2  * Copyright (c) 2003-2009 Tim Kientzle
3  * Copyright (c) 2010-2012 Michihiro NAKAJIMA
4  * Copyright (c) 2016 Martin Matuska
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
17  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19  * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
20  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26  */
27 
28 #include "archive_platform.h"
29 
30 /* This is the tree-walking code for POSIX systems. */
31 #if !defined(_WIN32) || defined(__CYGWIN__)
32 
33 #ifdef HAVE_SYS_TYPES_H
34 #include <sys/types.h>
35 #endif
36 #ifdef HAVE_SYS_EXTATTR_H
37 #include <sys/extattr.h>
38 #endif
39 #ifdef HAVE_SYS_IOCTL_H
40 #include <sys/ioctl.h>
41 #endif
42 #ifdef HAVE_SYS_PARAM_H
43 #include <sys/param.h>
44 #endif
45 #ifdef HAVE_SYS_STAT_H
46 #include <sys/stat.h>
47 #endif
48 #if defined(HAVE_SYS_XATTR_H)
49 #include <sys/xattr.h>
50 #elif defined(HAVE_ATTR_XATTR_H)
51 #include <attr/xattr.h>
52 #endif
53 #ifdef HAVE_SYS_EA_H
54 #include <sys/ea.h>
55 #endif
56 #ifdef HAVE_COPYFILE_H
57 #include <copyfile.h>
58 #endif
59 #ifdef HAVE_ERRNO_H
60 #include <errno.h>
61 #endif
62 #ifdef HAVE_FCNTL_H
63 #include <fcntl.h>
64 #endif
65 #ifdef HAVE_LIMITS_H
66 #include <limits.h>
67 #endif
68 #ifdef HAVE_LINUX_TYPES_H
69 #include <linux/types.h>
70 #endif
71 #ifdef HAVE_LINUX_FIEMAP_H
72 #include <linux/fiemap.h>
73 #endif
74 #ifdef HAVE_LINUX_FS_H
75 #include <linux/fs.h>
76 #endif
77 /*
78  * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h.
79  * As the include guards don't agree, the order of include is important.
80  */
81 #ifdef HAVE_LINUX_EXT2_FS_H
82 #include <linux/ext2_fs.h>      /* for Linux file flags */
83 #endif
84 #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__)
85 #include <ext2fs/ext2_fs.h>     /* Linux file flags, broken on Cygwin */
86 #endif
87 #ifdef HAVE_PATHS_H
88 #include <paths.h>
89 #endif
90 #ifdef HAVE_UNISTD_H
91 #include <unistd.h>
92 #endif
93 
94 #include "archive.h"
95 #include "archive_entry.h"
96 #include "archive_private.h"
97 #include "archive_read_disk_private.h"
98 
99 #ifndef O_CLOEXEC
100 #define O_CLOEXEC	0
101 #endif
102 
103 static int setup_mac_metadata(struct archive_read_disk *,
104     struct archive_entry *, int *fd);
105 #ifdef ARCHIVE_XATTR_FREEBSD
106 static int setup_xattrs_namespace(struct archive_read_disk *,
107     struct archive_entry *, int *, int);
108 #endif
109 static int setup_xattrs(struct archive_read_disk *,
110     struct archive_entry *, int *fd);
111 static int setup_sparse(struct archive_read_disk *,
112     struct archive_entry *, int *fd);
113 #if defined(HAVE_LINUX_FIEMAP_H)
114 static int setup_sparse_fiemap(struct archive_read_disk *,
115     struct archive_entry *, int *fd);
116 #endif
117 
118 #if !ARCHIVE_ACL_SUPPORT
119 int
archive_read_disk_entry_setup_acls(struct archive_read_disk * a,struct archive_entry * entry,int * fd)120 archive_read_disk_entry_setup_acls(struct archive_read_disk *a,
121     struct archive_entry *entry, int *fd)
122 {
123 	(void)a;      /* UNUSED */
124 	(void)entry;  /* UNUSED */
125 	(void)fd;     /* UNUSED */
126 	return (ARCHIVE_OK);
127 }
128 #endif
129 
130 /*
131  * Enter working directory and return working pathname of archive_entry.
132  * If a pointer to an integer is provided and its value is below zero
133  * open a file descriptor on this pathname.
134  */
135 const char *
archive_read_disk_entry_setup_path(struct archive_read_disk * a,struct archive_entry * entry,int * fd)136 archive_read_disk_entry_setup_path(struct archive_read_disk *a,
137     struct archive_entry *entry, int *fd)
138 {
139 	const char *path;
140 
141 	path = archive_entry_sourcepath(entry);
142 
143 	if (path == NULL || (a->tree != NULL &&
144 	    a->tree_enter_working_dir(a->tree) != 0))
145 		path = archive_entry_pathname(entry);
146 	if (path == NULL) {
147 		archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
148 		   "Couldn't determine path");
149 	} else if (fd != NULL && *fd < 0 && a->tree != NULL &&
150 	    (a->follow_symlinks || archive_entry_filetype(entry) != AE_IFLNK)) {
151 		*fd = a->open_on_current_dir(a->tree, path,
152 		    O_RDONLY | O_NONBLOCK);
153 	}
154 	return (path);
155 }
156 
157 int
archive_read_disk_entry_from_file(struct archive * _a,struct archive_entry * entry,int fd,const struct stat * st)158 archive_read_disk_entry_from_file(struct archive *_a,
159     struct archive_entry *entry,
160     int fd,
161     const struct stat *st)
162 {
163 	struct archive_read_disk *a = (struct archive_read_disk *)_a;
164 	const char *path, *name;
165 	struct stat s;
166 	int initial_fd = fd;
167 	int r, r1;
168 
169 	archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY,
170 		"archive_read_disk_entry_from_file");
171 
172 	archive_clear_error(_a);
173 	path = archive_entry_sourcepath(entry);
174 	if (path == NULL)
175 		path = archive_entry_pathname(entry);
176 
177 	if (a->tree == NULL) {
178 		if (st == NULL) {
179 #if HAVE_FSTAT
180 			if (fd >= 0) {
181 				if (fstat(fd, &s) != 0) {
182 					archive_set_error(&a->archive, errno,
183 					    "Can't fstat");
184 					return (ARCHIVE_FAILED);
185 				}
186 			} else
187 #endif
188 #if HAVE_LSTAT
189 			if (!a->follow_symlinks) {
190 				if (lstat(path, &s) != 0) {
191 					archive_set_error(&a->archive, errno,
192 					    "Can't lstat %s", path);
193 					return (ARCHIVE_FAILED);
194 				}
195 			} else
196 #endif
197 			if (la_stat(path, &s) != 0) {
198 				archive_set_error(&a->archive, errno,
199 				    "Can't stat %s", path);
200 				return (ARCHIVE_FAILED);
201 			}
202 			st = &s;
203 		}
204 		archive_entry_copy_stat(entry, st);
205 	}
206 
207 	/* Lookup uname/gname */
208 	name = archive_read_disk_uname(_a, archive_entry_uid(entry));
209 	if (name != NULL)
210 		archive_entry_copy_uname(entry, name);
211 	name = archive_read_disk_gname(_a, archive_entry_gid(entry));
212 	if (name != NULL)
213 		archive_entry_copy_gname(entry, name);
214 
215 #ifdef HAVE_STRUCT_STAT_ST_FLAGS
216 	/* On FreeBSD, we get flags for free with the stat. */
217 	/* TODO: Does this belong in copy_stat()? */
218 	if ((a->flags & ARCHIVE_READDISK_NO_FFLAGS) == 0 && st->st_flags != 0)
219 		archive_entry_set_fflags(entry, st->st_flags, 0);
220 #endif
221 
222 #if (defined(FS_IOC_GETFLAGS) && defined(HAVE_WORKING_FS_IOC_GETFLAGS)) || \
223     (defined(EXT2_IOC_GETFLAGS) && defined(HAVE_WORKING_EXT2_IOC_GETFLAGS))
224 	/* Linux requires an extra ioctl to pull the flags.  Although
225 	 * this is an extra step, it has a nice side-effect: We get an
226 	 * open file descriptor which we can use in the subsequent lookups. */
227 	if ((a->flags & ARCHIVE_READDISK_NO_FFLAGS) == 0 &&
228 	    (S_ISREG(st->st_mode) || S_ISDIR(st->st_mode))) {
229 		if (fd < 0) {
230 			if (a->tree != NULL)
231 				fd = a->open_on_current_dir(a->tree, path,
232 					O_RDONLY | O_NONBLOCK | O_CLOEXEC);
233 			else
234 				fd = open(path, O_RDONLY | O_NONBLOCK |
235 						O_CLOEXEC);
236 			__archive_ensure_cloexec_flag(fd);
237 		}
238 		if (fd >= 0) {
239 			int stflags;
240 			r = ioctl(fd,
241 #if defined(FS_IOC_GETFLAGS)
242 			    FS_IOC_GETFLAGS,
243 #else
244 			    EXT2_IOC_GETFLAGS,
245 #endif
246 			    &stflags);
247 			if (r == 0 && stflags != 0)
248 				archive_entry_set_fflags(entry, stflags, 0);
249 		}
250 	}
251 #endif
252 
253 #if defined(HAVE_READLINK) || defined(HAVE_READLINKAT)
254 	if (S_ISLNK(st->st_mode)) {
255 		size_t linkbuffer_len = st->st_size;
256 		char *linkbuffer;
257 		int lnklen;
258 
259 		linkbuffer = malloc(linkbuffer_len + 1);
260 		if (linkbuffer == NULL) {
261 			archive_set_error(&a->archive, ENOMEM,
262 			    "Couldn't read link data");
263 			return (ARCHIVE_FAILED);
264 		}
265 		if (a->tree != NULL) {
266 #ifdef HAVE_READLINKAT
267 			lnklen = readlinkat(a->tree_current_dir_fd(a->tree),
268 			    path, linkbuffer, linkbuffer_len);
269 #else
270 			if (a->tree_enter_working_dir(a->tree) != 0) {
271 				archive_set_error(&a->archive, errno,
272 				    "Couldn't read link data");
273 				free(linkbuffer);
274 				return (ARCHIVE_FAILED);
275 			}
276 			lnklen = readlink(path, linkbuffer, linkbuffer_len);
277 #endif /* HAVE_READLINKAT */
278 		} else
279 			lnklen = readlink(path, linkbuffer, linkbuffer_len);
280 		if (lnklen < 0) {
281 			archive_set_error(&a->archive, errno,
282 			    "Couldn't read link data");
283 			free(linkbuffer);
284 			return (ARCHIVE_FAILED);
285 		}
286 		linkbuffer[lnklen] = '\0';
287 		archive_entry_set_symlink(entry, linkbuffer);
288 		free(linkbuffer);
289 	}
290 #endif /* HAVE_READLINK || HAVE_READLINKAT */
291 
292 	r = 0;
293 	if ((a->flags & ARCHIVE_READDISK_NO_ACL) == 0)
294 		r = archive_read_disk_entry_setup_acls(a, entry, &fd);
295 	if ((a->flags & ARCHIVE_READDISK_NO_XATTR) == 0) {
296 		r1 = setup_xattrs(a, entry, &fd);
297 		if (r1 < r)
298 			r = r1;
299 	}
300 	if (a->flags & ARCHIVE_READDISK_MAC_COPYFILE) {
301 		r1 = setup_mac_metadata(a, entry, &fd);
302 		if (r1 < r)
303 			r = r1;
304 	}
305 	if ((a->flags & ARCHIVE_READDISK_NO_SPARSE) == 0) {
306 		r1 = setup_sparse(a, entry, &fd);
307 		if (r1 < r)
308 			r = r1;
309 	}
310 
311 	/* If we opened the file earlier in this function, close it. */
312 	if (initial_fd != fd)
313 		close(fd);
314 	return (r);
315 }
316 
317 #if defined(__APPLE__) && defined(HAVE_COPYFILE_H)
318 /*
319  * The Mac OS "copyfile()" API copies the extended metadata for a
320  * file into a separate file in AppleDouble format (see RFC 1740).
321  *
322  * Mac OS tar and cpio implementations store this extended
323  * metadata as a separate entry just before the regular entry
324  * with a "._" prefix added to the filename.
325  *
326  * Note that this is currently done unconditionally; the tar program has
327  * an option to discard this information before the archive is written.
328  *
329  * TODO: If there's a failure, report it and return ARCHIVE_WARN.
330  */
331 static int
setup_mac_metadata(struct archive_read_disk * a,struct archive_entry * entry,int * fd)332 setup_mac_metadata(struct archive_read_disk *a,
333     struct archive_entry *entry, int *fd)
334 {
335 	int tempfd = -1;
336 	int copyfile_flags = COPYFILE_NOFOLLOW | COPYFILE_ACL | COPYFILE_XATTR;
337 	struct stat copyfile_stat;
338 	int ret = ARCHIVE_OK;
339 	void *buff = NULL;
340 	int have_attrs;
341 	const char *name, *tempdir;
342 	struct archive_string tempfile;
343 
344 	(void)fd; /* UNUSED */
345 
346 	name = archive_read_disk_entry_setup_path(a, entry, NULL);
347 	if (name == NULL)
348 		return (ARCHIVE_WARN);
349 
350 	/* Short-circuit if there's nothing to do. */
351 	have_attrs = copyfile(name, NULL, 0, copyfile_flags | COPYFILE_CHECK);
352 	if (have_attrs == -1) {
353 		archive_set_error(&a->archive, errno,
354 			"Could not check extended attributes");
355 		return (ARCHIVE_WARN);
356 	}
357 	if (have_attrs == 0)
358 		return (ARCHIVE_OK);
359 
360 	tempdir = NULL;
361 	if (issetugid() == 0)
362 		tempdir = getenv("TMPDIR");
363 	if (tempdir == NULL)
364 		tempdir = _PATH_TMP;
365 	archive_string_init(&tempfile);
366 	archive_strcpy(&tempfile, tempdir);
367 	archive_strcat(&tempfile, "tar.md.XXXXXX");
368 	tempfd = mkstemp(tempfile.s);
369 	if (tempfd < 0) {
370 		archive_set_error(&a->archive, errno,
371 		    "Could not open extended attribute file");
372 		ret = ARCHIVE_WARN;
373 		goto cleanup;
374 	}
375 	__archive_ensure_cloexec_flag(tempfd);
376 
377 	/* XXX I wish copyfile() could pack directly to a memory
378 	 * buffer; that would avoid the temp file here.  For that
379 	 * matter, it would be nice if fcopyfile() actually worked,
380 	 * that would reduce the many open/close races here. */
381 	if (copyfile(name, tempfile.s, 0, copyfile_flags | COPYFILE_PACK)) {
382 		archive_set_error(&a->archive, errno,
383 		    "Could not pack extended attributes");
384 		ret = ARCHIVE_WARN;
385 		goto cleanup;
386 	}
387 	if (fstat(tempfd, &copyfile_stat)) {
388 		archive_set_error(&a->archive, errno,
389 		    "Could not check size of extended attributes");
390 		ret = ARCHIVE_WARN;
391 		goto cleanup;
392 	}
393 	buff = malloc(copyfile_stat.st_size);
394 	if (buff == NULL) {
395 		archive_set_error(&a->archive, errno,
396 		    "Could not allocate memory for extended attributes");
397 		ret = ARCHIVE_WARN;
398 		goto cleanup;
399 	}
400 	if (copyfile_stat.st_size != read(tempfd, buff, copyfile_stat.st_size)) {
401 		archive_set_error(&a->archive, errno,
402 		    "Could not read extended attributes into memory");
403 		ret = ARCHIVE_WARN;
404 		goto cleanup;
405 	}
406 	archive_entry_copy_mac_metadata(entry, buff, copyfile_stat.st_size);
407 
408 cleanup:
409 	if (tempfd >= 0) {
410 		close(tempfd);
411 		unlink(tempfile.s);
412 	}
413 	archive_string_free(&tempfile);
414 	free(buff);
415 	return (ret);
416 }
417 
418 #else
419 
420 /*
421  * Stub implementation for non-Mac systems.
422  */
423 static int
setup_mac_metadata(struct archive_read_disk * a,struct archive_entry * entry,int * fd)424 setup_mac_metadata(struct archive_read_disk *a,
425     struct archive_entry *entry, int *fd)
426 {
427 	(void)a; /* UNUSED */
428 	(void)entry; /* UNUSED */
429 	(void)fd; /* UNUSED */
430 	return (ARCHIVE_OK);
431 }
432 #endif
433 
434 #if ARCHIVE_XATTR_LINUX || ARCHIVE_XATTR_DARWIN || ARCHIVE_XATTR_AIX
435 
436 /*
437  * Linux, Darwin and AIX extended attribute support.
438  *
439  * TODO:  By using a stack-allocated buffer for the first
440  * call to getxattr(), we might be able to avoid the second
441  * call entirely.  We only need the second call if the
442  * stack-allocated buffer is too small.  But a modest buffer
443  * of 1024 bytes or so will often be big enough.  Same applies
444  * to listxattr().
445  */
446 
447 
448 static int
setup_xattr(struct archive_read_disk * a,struct archive_entry * entry,const char * name,int fd,const char * accpath)449 setup_xattr(struct archive_read_disk *a,
450     struct archive_entry *entry, const char *name, int fd, const char *accpath)
451 {
452 	ssize_t size;
453 	void *value = NULL;
454 
455 
456 	if (fd >= 0) {
457 #if ARCHIVE_XATTR_LINUX
458 		size = fgetxattr(fd, name, NULL, 0);
459 #elif ARCHIVE_XATTR_DARWIN
460 		size = fgetxattr(fd, name, NULL, 0, 0, 0);
461 #elif ARCHIVE_XATTR_AIX
462 		size = fgetea(fd, name, NULL, 0);
463 #endif
464 	} else if (!a->follow_symlinks) {
465 #if ARCHIVE_XATTR_LINUX
466 		size = lgetxattr(accpath, name, NULL, 0);
467 #elif ARCHIVE_XATTR_DARWIN
468 		size = getxattr(accpath, name, NULL, 0, 0, XATTR_NOFOLLOW);
469 #elif ARCHIVE_XATTR_AIX
470 		size = lgetea(accpath, name, NULL, 0);
471 #endif
472 	} else {
473 #if ARCHIVE_XATTR_LINUX
474 		size = getxattr(accpath, name, NULL, 0);
475 #elif ARCHIVE_XATTR_DARWIN
476 		size = getxattr(accpath, name, NULL, 0, 0, 0);
477 #elif ARCHIVE_XATTR_AIX
478 		size = getea(accpath, name, NULL, 0);
479 #endif
480 	}
481 
482 	if (size == -1) {
483 		archive_set_error(&a->archive, errno,
484 		    "Couldn't query extended attribute");
485 		return (ARCHIVE_WARN);
486 	}
487 
488 	if (size > 0 && (value = malloc(size)) == NULL) {
489 		archive_set_error(&a->archive, errno, "Out of memory");
490 		return (ARCHIVE_FATAL);
491 	}
492 
493 
494 	if (fd >= 0) {
495 #if ARCHIVE_XATTR_LINUX
496 		size = fgetxattr(fd, name, value, size);
497 #elif ARCHIVE_XATTR_DARWIN
498 		size = fgetxattr(fd, name, value, size, 0, 0);
499 #elif ARCHIVE_XATTR_AIX
500 		size = fgetea(fd, name, value, size);
501 #endif
502 	} else if (!a->follow_symlinks) {
503 #if ARCHIVE_XATTR_LINUX
504 		size = lgetxattr(accpath, name, value, size);
505 #elif ARCHIVE_XATTR_DARWIN
506 		size = getxattr(accpath, name, value, size, 0, XATTR_NOFOLLOW);
507 #elif ARCHIVE_XATTR_AIX
508 		size = lgetea(accpath, name, value, size);
509 #endif
510 	} else {
511 #if ARCHIVE_XATTR_LINUX
512 		size = getxattr(accpath, name, value, size);
513 #elif ARCHIVE_XATTR_DARWIN
514 		size = getxattr(accpath, name, value, size, 0, 0);
515 #elif ARCHIVE_XATTR_AIX
516 		size = getea(accpath, name, value, size);
517 #endif
518 	}
519 
520 	if (size == -1) {
521 		archive_set_error(&a->archive, errno,
522 		    "Couldn't read extended attribute");
523 		return (ARCHIVE_WARN);
524 	}
525 
526 	archive_entry_xattr_add_entry(entry, name, value, size);
527 
528 	free(value);
529 	return (ARCHIVE_OK);
530 }
531 
532 static int
setup_xattrs(struct archive_read_disk * a,struct archive_entry * entry,int * fd)533 setup_xattrs(struct archive_read_disk *a,
534     struct archive_entry *entry, int *fd)
535 {
536 	char *list, *p;
537 	const char *path;
538 	ssize_t list_size;
539 
540 	path = NULL;
541 
542 	if (*fd < 0) {
543 		path = archive_read_disk_entry_setup_path(a, entry, fd);
544 		if (path == NULL)
545 			return (ARCHIVE_WARN);
546 	}
547 
548 	if (*fd >= 0) {
549 #if ARCHIVE_XATTR_LINUX
550 		list_size = flistxattr(*fd, NULL, 0);
551 #elif ARCHIVE_XATTR_DARWIN
552 		list_size = flistxattr(*fd, NULL, 0, 0);
553 #elif ARCHIVE_XATTR_AIX
554 		list_size = flistea(*fd, NULL, 0);
555 #endif
556 	} else if (!a->follow_symlinks) {
557 #if ARCHIVE_XATTR_LINUX
558 		list_size = llistxattr(path, NULL, 0);
559 #elif ARCHIVE_XATTR_DARWIN
560 		list_size = listxattr(path, NULL, 0, XATTR_NOFOLLOW);
561 #elif ARCHIVE_XATTR_AIX
562 		list_size = llistea(path, NULL, 0);
563 #endif
564 	} else {
565 #if ARCHIVE_XATTR_LINUX
566 		list_size = listxattr(path, NULL, 0);
567 #elif ARCHIVE_XATTR_DARWIN
568 		list_size = listxattr(path, NULL, 0, 0);
569 #elif ARCHIVE_XATTR_AIX
570 		list_size = listea(path, NULL, 0);
571 #endif
572 	}
573 
574 	if (list_size == -1) {
575 		if (errno == ENOTSUP || errno == ENOSYS)
576 			return (ARCHIVE_OK);
577 		archive_set_error(&a->archive, errno,
578 			"Couldn't list extended attributes");
579 		return (ARCHIVE_WARN);
580 	}
581 
582 	if (list_size == 0)
583 		return (ARCHIVE_OK);
584 
585 	if ((list = malloc(list_size)) == NULL) {
586 		archive_set_error(&a->archive, errno, "Out of memory");
587 		return (ARCHIVE_FATAL);
588 	}
589 
590 	if (*fd >= 0) {
591 #if ARCHIVE_XATTR_LINUX
592 		list_size = flistxattr(*fd, list, list_size);
593 #elif ARCHIVE_XATTR_DARWIN
594 		list_size = flistxattr(*fd, list, list_size, 0);
595 #elif ARCHIVE_XATTR_AIX
596 		list_size = flistea(*fd, list, list_size);
597 #endif
598 	} else if (!a->follow_symlinks) {
599 #if ARCHIVE_XATTR_LINUX
600 		list_size = llistxattr(path, list, list_size);
601 #elif ARCHIVE_XATTR_DARWIN
602 		list_size = listxattr(path, list, list_size, XATTR_NOFOLLOW);
603 #elif ARCHIVE_XATTR_AIX
604 		list_size = llistea(path, list, list_size);
605 #endif
606 	} else {
607 #if ARCHIVE_XATTR_LINUX
608 		list_size = listxattr(path, list, list_size);
609 #elif ARCHIVE_XATTR_DARWIN
610 		list_size = listxattr(path, list, list_size, 0);
611 #elif ARCHIVE_XATTR_AIX
612 		list_size = listea(path, list, list_size);
613 #endif
614 	}
615 
616 	if (list_size == -1) {
617 		archive_set_error(&a->archive, errno,
618 			"Couldn't retrieve extended attributes");
619 		free(list);
620 		return (ARCHIVE_WARN);
621 	}
622 
623 	for (p = list; (p - list) < list_size; p += strlen(p) + 1) {
624 #if ARCHIVE_XATTR_LINUX
625 		/* Linux: skip POSIX.1e ACL extended attributes */
626 		if (strncmp(p, "system.", 7) == 0 &&
627 		   (strcmp(p + 7, "posix_acl_access") == 0 ||
628 		    strcmp(p + 7, "posix_acl_default") == 0))
629 			continue;
630 		if (strncmp(p, "trusted.SGI_", 12) == 0 &&
631 		   (strcmp(p + 12, "ACL_DEFAULT") == 0 ||
632 		    strcmp(p + 12, "ACL_FILE") == 0))
633 			continue;
634 
635 		/* Linux: xfsroot namespace is obsolete and unsupported */
636 		if (strncmp(p, "xfsroot.", 8) == 0)
637 			continue;
638 #endif
639 		setup_xattr(a, entry, p, *fd, path);
640 	}
641 
642 	free(list);
643 	return (ARCHIVE_OK);
644 }
645 
646 #elif ARCHIVE_XATTR_FREEBSD
647 
648 /*
649  * FreeBSD extattr interface.
650  */
651 
652 /* TODO: Implement this.  Follow the Linux model above, but
653  * with FreeBSD-specific system calls, of course.  Be careful
654  * to not include the system extattrs that hold ACLs; we handle
655  * those separately.
656  */
657 static int
658 setup_xattr(struct archive_read_disk *a, struct archive_entry *entry,
659     int namespace, const char *name, const char *fullname, int fd,
660     const char *path);
661 
662 static int
setup_xattr(struct archive_read_disk * a,struct archive_entry * entry,int namespace,const char * name,const char * fullname,int fd,const char * accpath)663 setup_xattr(struct archive_read_disk *a, struct archive_entry *entry,
664     int namespace, const char *name, const char *fullname, int fd,
665     const char *accpath)
666 {
667 	ssize_t size;
668 	void *value = NULL;
669 
670 	if (fd >= 0)
671 		size = extattr_get_fd(fd, namespace, name, NULL, 0);
672 	else if (!a->follow_symlinks)
673 		size = extattr_get_link(accpath, namespace, name, NULL, 0);
674 	else
675 		size = extattr_get_file(accpath, namespace, name, NULL, 0);
676 
677 	if (size == -1) {
678 		archive_set_error(&a->archive, errno,
679 		    "Couldn't query extended attribute");
680 		return (ARCHIVE_WARN);
681 	}
682 
683 	if (size > 0 && (value = malloc(size)) == NULL) {
684 		archive_set_error(&a->archive, errno, "Out of memory");
685 		return (ARCHIVE_FATAL);
686 	}
687 
688 	if (fd >= 0)
689 		size = extattr_get_fd(fd, namespace, name, value, size);
690 	else if (!a->follow_symlinks)
691 		size = extattr_get_link(accpath, namespace, name, value, size);
692 	else
693 		size = extattr_get_file(accpath, namespace, name, value, size);
694 
695 	if (size == -1) {
696 		free(value);
697 		archive_set_error(&a->archive, errno,
698 		    "Couldn't read extended attribute");
699 		return (ARCHIVE_WARN);
700 	}
701 
702 	archive_entry_xattr_add_entry(entry, fullname, value, size);
703 
704 	free(value);
705 	return (ARCHIVE_OK);
706 }
707 
708 static int
setup_xattrs_namespace(struct archive_read_disk * a,struct archive_entry * entry,int * fd,int namespace)709 setup_xattrs_namespace(struct archive_read_disk *a,
710     struct archive_entry *entry, int *fd, int namespace)
711 {
712 	char buff[512];
713 	char *list, *p;
714 	ssize_t list_size;
715 	const char *path;
716 
717 	path = NULL;
718 
719 	if (*fd < 0) {
720 		path = archive_read_disk_entry_setup_path(a, entry, fd);
721 		if (path == NULL)
722 			return (ARCHIVE_WARN);
723 	}
724 
725 	if (*fd >= 0)
726 		list_size = extattr_list_fd(*fd, namespace, NULL, 0);
727 	else if (!a->follow_symlinks)
728 		list_size = extattr_list_link(path, namespace, NULL, 0);
729 	else
730 		list_size = extattr_list_file(path, namespace, NULL, 0);
731 
732 	if (list_size == -1 && errno == EOPNOTSUPP)
733 		return (ARCHIVE_OK);
734 	if (list_size == -1 && errno == EPERM)
735 		return (ARCHIVE_OK);
736 	if (list_size == -1) {
737 		archive_set_error(&a->archive, errno,
738 			"Couldn't list extended attributes");
739 		return (ARCHIVE_WARN);
740 	}
741 
742 	if (list_size == 0)
743 		return (ARCHIVE_OK);
744 
745 	if ((list = malloc(list_size)) == NULL) {
746 		archive_set_error(&a->archive, errno, "Out of memory");
747 		return (ARCHIVE_FATAL);
748 	}
749 
750 	if (*fd >= 0)
751 		list_size = extattr_list_fd(*fd, namespace, list, list_size);
752 	else if (!a->follow_symlinks)
753 		list_size = extattr_list_link(path, namespace, list, list_size);
754 	else
755 		list_size = extattr_list_file(path, namespace, list, list_size);
756 
757 	if (list_size == -1) {
758 		archive_set_error(&a->archive, errno,
759 			"Couldn't retrieve extended attributes");
760 		free(list);
761 		return (ARCHIVE_WARN);
762 	}
763 
764 	p = list;
765 	while ((p - list) < list_size) {
766 		size_t len = 255 & (int)*p;
767 		char *name;
768 
769 		if (namespace == EXTATTR_NAMESPACE_SYSTEM) {
770 			if (!strcmp(p + 1, "nfs4.acl") ||
771 			    !strcmp(p + 1, "posix1e.acl_access") ||
772 			    !strcmp(p + 1, "posix1e.acl_default")) {
773 				p += 1 + len;
774 				continue;
775 			}
776 			strcpy(buff, "system.");
777 		} else {
778 			strcpy(buff, "user.");
779 		}
780 		name = buff + strlen(buff);
781 		memcpy(name, p + 1, len);
782 		name[len] = '\0';
783 		setup_xattr(a, entry, namespace, name, buff, *fd, path);
784 		p += 1 + len;
785 	}
786 
787 	free(list);
788 	return (ARCHIVE_OK);
789 }
790 
791 static int
setup_xattrs(struct archive_read_disk * a,struct archive_entry * entry,int * fd)792 setup_xattrs(struct archive_read_disk *a,
793     struct archive_entry *entry, int *fd)
794 {
795 	int namespaces[2];
796 	int i, res;
797 
798 	namespaces[0] = EXTATTR_NAMESPACE_USER;
799 	namespaces[1] = EXTATTR_NAMESPACE_SYSTEM;
800 
801 	for (i = 0; i < 2; i++) {
802 		res = setup_xattrs_namespace(a, entry, fd,
803 		    namespaces[i]);
804 		switch (res) {
805 			case (ARCHIVE_OK):
806 			case (ARCHIVE_WARN):
807 				break;
808 			default:
809 				return (res);
810 		}
811 	}
812 
813 	return (ARCHIVE_OK);
814 }
815 
816 #else
817 
818 /*
819  * Generic (stub) extended attribute support.
820  */
821 static int
setup_xattrs(struct archive_read_disk * a,struct archive_entry * entry,int * fd)822 setup_xattrs(struct archive_read_disk *a,
823     struct archive_entry *entry, int *fd)
824 {
825 	(void)a;     /* UNUSED */
826 	(void)entry; /* UNUSED */
827 	(void)fd;    /* UNUSED */
828 	return (ARCHIVE_OK);
829 }
830 
831 #endif
832 
833 #if defined(HAVE_LINUX_FIEMAP_H)
834 
835 /*
836  * Linux FIEMAP sparse interface.
837  *
838  * The FIEMAP ioctl returns an "extent" for each physical allocation
839  * on disk.  We need to process those to generate a more compact list
840  * of logical file blocks.  We also need to be very careful to use
841  * FIEMAP_FLAG_SYNC here, since there are reports that Linux sometimes
842  * does not report allocations for newly-written data that hasn't
843  * been synced to disk.
844  *
845  * It's important to return a minimal sparse file list because we want
846  * to not trigger sparse file extensions if we don't have to, since
847  * not all readers support them.
848  */
849 
850 static int
setup_sparse_fiemap(struct archive_read_disk * a,struct archive_entry * entry,int * fd)851 setup_sparse_fiemap(struct archive_read_disk *a,
852     struct archive_entry *entry, int *fd)
853 {
854 	char buff[4096];
855 	struct fiemap *fm;
856 	struct fiemap_extent *fe;
857 	int64_t size;
858 	int count, do_fiemap, iters;
859 	int exit_sts = ARCHIVE_OK;
860 	const char *path;
861 
862 	if (archive_entry_filetype(entry) != AE_IFREG
863 	    || archive_entry_size(entry) <= 0
864 	    || archive_entry_hardlink(entry) != NULL)
865 		return (ARCHIVE_OK);
866 
867 	if (*fd < 0) {
868 		path = archive_read_disk_entry_setup_path(a, entry, NULL);
869 		if (path == NULL)
870 			return (ARCHIVE_FAILED);
871 
872 		if (a->tree != NULL)
873 			*fd = a->open_on_current_dir(a->tree, path,
874 				O_RDONLY | O_NONBLOCK | O_CLOEXEC);
875 		else
876 			*fd = open(path, O_RDONLY | O_NONBLOCK | O_CLOEXEC);
877 		if (*fd < 0) {
878 			archive_set_error(&a->archive, errno,
879 			    "Can't open `%s'", path);
880 			return (ARCHIVE_FAILED);
881 		}
882 		__archive_ensure_cloexec_flag(*fd);
883 	}
884 
885 	/* Initialize buffer to avoid the error valgrind complains about. */
886 	memset(buff, 0, sizeof(buff));
887 	count = (sizeof(buff) - sizeof(*fm))/sizeof(*fe);
888 	fm = (struct fiemap *)buff;
889 	fm->fm_start = 0;
890 	fm->fm_length = ~0ULL;;
891 	fm->fm_flags = FIEMAP_FLAG_SYNC;
892 	fm->fm_extent_count = count;
893 	do_fiemap = 1;
894 	size = archive_entry_size(entry);
895 	for (iters = 0; ; ++iters) {
896 		int i, r;
897 
898 		r = ioctl(*fd, FS_IOC_FIEMAP, fm);
899 		if (r < 0) {
900 			/* When something error happens, it is better we
901 			 * should return ARCHIVE_OK because an earlier
902 			 * version(<2.6.28) cannot perform FS_IOC_FIEMAP. */
903 			goto exit_setup_sparse_fiemap;
904 		}
905 		if (fm->fm_mapped_extents == 0) {
906 			if (iters == 0) {
907 				/* Fully sparse file; insert a zero-length "data" entry */
908 				archive_entry_sparse_add_entry(entry, 0, 0);
909 			}
910 			break;
911 		}
912 		fe = fm->fm_extents;
913 		for (i = 0; i < (int)fm->fm_mapped_extents; i++, fe++) {
914 			if (!(fe->fe_flags & FIEMAP_EXTENT_UNWRITTEN)) {
915 				/* The fe_length of the last block does not
916 				 * adjust itself to its size files. */
917 				int64_t length = fe->fe_length;
918 				if (fe->fe_logical + length > (uint64_t)size)
919 					length -= fe->fe_logical + length - size;
920 				if (fe->fe_logical == 0 && length == size) {
921 					/* This is not sparse. */
922 					do_fiemap = 0;
923 					break;
924 				}
925 				if (length > 0)
926 					archive_entry_sparse_add_entry(entry,
927 					    fe->fe_logical, length);
928 			}
929 			if (fe->fe_flags & FIEMAP_EXTENT_LAST)
930 				do_fiemap = 0;
931 		}
932 		if (do_fiemap) {
933 			fe = fm->fm_extents + fm->fm_mapped_extents -1;
934 			fm->fm_start = fe->fe_logical + fe->fe_length;
935 		} else
936 			break;
937 	}
938 exit_setup_sparse_fiemap:
939 	return (exit_sts);
940 }
941 
942 #if !defined(SEEK_HOLE) || !defined(SEEK_DATA)
943 static int
setup_sparse(struct archive_read_disk * a,struct archive_entry * entry,int * fd)944 setup_sparse(struct archive_read_disk *a,
945     struct archive_entry *entry, int *fd)
946 {
947 	return setup_sparse_fiemap(a, entry, fd);
948 }
949 #endif
950 #endif	/* defined(HAVE_LINUX_FIEMAP_H) */
951 
952 #if defined(SEEK_HOLE) && defined(SEEK_DATA)
953 
954 /*
955  * SEEK_HOLE sparse interface (FreeBSD, Linux, Solaris)
956  */
957 
958 static int
setup_sparse(struct archive_read_disk * a,struct archive_entry * entry,int * fd)959 setup_sparse(struct archive_read_disk *a,
960     struct archive_entry *entry, int *fd)
961 {
962 	int64_t size;
963 	off_t initial_off;
964 	off_t off_s, off_e;
965 	int exit_sts = ARCHIVE_OK;
966 	int check_fully_sparse = 0;
967 	const char *path;
968 
969 	if (archive_entry_filetype(entry) != AE_IFREG
970 	    || archive_entry_size(entry) <= 0
971 	    || archive_entry_hardlink(entry) != NULL)
972 		return (ARCHIVE_OK);
973 
974 	/* Does filesystem support the reporting of hole ? */
975 	if (*fd < 0)
976 		path = archive_read_disk_entry_setup_path(a, entry, fd);
977 	else
978 		path = NULL;
979 
980 	if (*fd >= 0) {
981 #ifdef _PC_MIN_HOLE_SIZE
982 		if (fpathconf(*fd, _PC_MIN_HOLE_SIZE) <= 0)
983 			return (ARCHIVE_OK);
984 #endif
985 		initial_off = lseek(*fd, 0, SEEK_CUR);
986 		if (initial_off != 0)
987 			lseek(*fd, 0, SEEK_SET);
988 	} else {
989 		if (path == NULL)
990 			return (ARCHIVE_FAILED);
991 #ifdef _PC_MIN_HOLE_SIZE
992 		if (pathconf(path, _PC_MIN_HOLE_SIZE) <= 0)
993 			return (ARCHIVE_OK);
994 #endif
995 		*fd = open(path, O_RDONLY | O_NONBLOCK | O_CLOEXEC);
996 		if (*fd < 0) {
997 			archive_set_error(&a->archive, errno,
998 			    "Can't open `%s'", path);
999 			return (ARCHIVE_FAILED);
1000 		}
1001 		__archive_ensure_cloexec_flag(*fd);
1002 		initial_off = 0;
1003 	}
1004 
1005 #ifndef _PC_MIN_HOLE_SIZE
1006 	/* Check if the underlying filesystem supports seek hole */
1007 	off_s = lseek(*fd, 0, SEEK_HOLE);
1008 	if (off_s < 0)
1009 #if defined(HAVE_LINUX_FIEMAP_H)
1010 		return setup_sparse_fiemap(a, entry, fd);
1011 #else
1012 		goto exit_setup_sparse;
1013 #endif
1014 	else if (off_s > 0)
1015 		lseek(*fd, 0, SEEK_SET);
1016 #endif
1017 
1018 	off_s = 0;
1019 	size = archive_entry_size(entry);
1020 	while (off_s < size) {
1021 		off_s = lseek(*fd, off_s, SEEK_DATA);
1022 		if (off_s == (off_t)-1) {
1023 			if (errno == ENXIO) {
1024 				/* no more hole */
1025 				if (archive_entry_sparse_count(entry) == 0) {
1026 					/* Potentially a fully-sparse file. */
1027 					check_fully_sparse = 1;
1028 				}
1029 				break;
1030 			}
1031 			archive_set_error(&a->archive, errno,
1032 			    "lseek(SEEK_HOLE) failed");
1033 			exit_sts = ARCHIVE_FAILED;
1034 			goto exit_setup_sparse;
1035 		}
1036 		off_e = lseek(*fd, off_s, SEEK_HOLE);
1037 		if (off_e == (off_t)-1) {
1038 			if (errno == ENXIO) {
1039 				off_e = lseek(*fd, 0, SEEK_END);
1040 				if (off_e != (off_t)-1)
1041 					break;/* no more data */
1042 			}
1043 			archive_set_error(&a->archive, errno,
1044 			    "lseek(SEEK_DATA) failed");
1045 			exit_sts = ARCHIVE_FAILED;
1046 			goto exit_setup_sparse;
1047 		}
1048 		if (off_s == 0 && off_e == size)
1049 			break;/* This is not sparse. */
1050 		archive_entry_sparse_add_entry(entry, off_s,
1051 			off_e - off_s);
1052 		off_s = off_e;
1053 	}
1054 
1055 	if (check_fully_sparse) {
1056 		if (lseek(*fd, 0, SEEK_HOLE) == 0 &&
1057 			lseek(*fd, 0, SEEK_END) == size) {
1058 			/* Fully sparse file; insert a zero-length "data" entry */
1059 			archive_entry_sparse_add_entry(entry, 0, 0);
1060 		}
1061 	}
1062 exit_setup_sparse:
1063 	lseek(*fd, initial_off, SEEK_SET);
1064 	return (exit_sts);
1065 }
1066 
1067 #elif !defined(HAVE_LINUX_FIEMAP_H)
1068 
1069 /*
1070  * Generic (stub) sparse support.
1071  */
1072 static int
setup_sparse(struct archive_read_disk * a,struct archive_entry * entry,int * fd)1073 setup_sparse(struct archive_read_disk *a,
1074     struct archive_entry *entry, int *fd)
1075 {
1076 	(void)a;     /* UNUSED */
1077 	(void)entry; /* UNUSED */
1078 	(void)fd;    /* UNUSED */
1079 	return (ARCHIVE_OK);
1080 }
1081 
1082 #endif
1083 
1084 #endif /* !defined(_WIN32) || defined(__CYGWIN__) */
1085 
1086