2 * Copyright (c) 2003-2009 Tim Kientzle
3 * Copyright (c) 2010-2012 Michihiro NAKAJIMA
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
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.
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.
27 #include "archive_platform.h"
28 __FBSDID("$FreeBSD$");
30 #if defined(_WIN32) && !defined(__CYGWIN__)
41 #include "archive_string.h"
42 #include "archive_entry.h"
43 #include "archive_private.h"
44 #include "archive_read_disk_private.h"
49 #ifndef IO_REPARSE_TAG_SYMLINK
50 /* Old SDKs do not provide IO_REPARSE_TAG_SYMLINK */
51 #define IO_REPARSE_TAG_SYMLINK 0xA000000CL
55 * This is a new directory-walking system that addresses a number
56 * of problems I've had with fts(3). In particular, it has no
57 * pathname-length limits (other than the size of 'int'), handles
58 * deep logical traversals, uses considerably less memory, and has
59 * an opaque interface (easier to modify in the future).
61 * Internally, it keeps a single list of "tree_entry" items that
62 * represent filesystem objects that require further attention.
63 * Non-directories are not kept in memory: they are pulled from
64 * readdir(), returned to the client, then freed as soon as possible.
65 * Any directory entry to be traversed gets pushed onto the stack.
67 * There is surprisingly little information that needs to be kept for
68 * each item on the stack. Just the name, depth (represented here as the
69 * string length of the parent directory's pathname), and some markers
70 * indicating how to get back to the parent (via chdir("..") for a
71 * regular dir or via fchdir(2) for a symlink).
75 const wchar_t *full_path;
76 FILETIME lastWriteTime;
77 FILETIME lastAccessTime;
83 struct tree_entry *next;
84 struct tree_entry *parent;
85 size_t full_path_dir_length;
86 struct archive_wstring name;
87 struct archive_wstring full_path;
88 size_t dirname_length;
93 /* How to restore time of a directory. */
94 struct restore_time restore_time;
101 DWORD bytesPerSector;
104 /* Definitions for tree_entry.flags bitmap. */
105 #define isDir 1 /* This entry is a regular directory. */
106 #define isDirLink 2 /* This entry is a symbolic link to a directory. */
107 #define needsFirstVisit 4 /* This is an initial entry. */
108 #define needsDescent 8 /* This entry needs to be previsited. */
109 #define needsOpen 16 /* This is a directory that needs to be opened. */
110 #define needsAscent 32 /* This entry needs to be postvisited. */
113 * On Windows, "first visit" is handled as a pattern to be handed to
114 * _findfirst(). This is consistent with Windows conventions that
115 * file patterns are handled within the application. On Posix,
116 * "first visit" is just returned to the client.
119 #define MAX_OVERLAPPED 8
120 #define BUFFER_SIZE (1024 * 8)
121 #define DIRECT_IO 0/* Disabled */
122 #define ASYNC_IO 1/* Enabled */
125 * Local data for this package.
128 struct tree_entry *stack;
129 struct tree_entry *current;
131 WIN32_FIND_DATAW _findData;
132 WIN32_FIND_DATAW *findData;
135 /* Error code from last failed operation. */
138 /* A full path with "\\?\" prefix. */
139 struct archive_wstring full_path;
140 size_t full_path_dir_length;
141 /* Dynamically-sized buffer for holding path */
142 struct archive_wstring path;
144 /* Last path element */
145 const wchar_t *basename;
146 /* Leading dir length */
147 size_t dirname_length;
151 BY_HANDLE_FILE_INFORMATION lst;
152 BY_HANDLE_FILE_INFORMATION st;
154 /* How to restore time of a file. */
155 struct restore_time restore_time;
157 struct entry_sparse {
160 } *sparse_list, *current_sparse;
162 int sparse_list_size;
164 char initial_symlink_mode;
166 struct filesystem *current_filesystem;
167 struct filesystem *filesystem_table;
168 int initial_filesystem_id;
169 int current_filesystem_id;
170 int max_filesystem_id;
171 int allocated_filesytem;
175 int64_t entry_remaining_bytes;
182 int64_t ol_remaining_bytes;
184 struct la_overlapped {
190 size_t bytes_expected;
191 size_t bytes_transferred;
192 } ol[MAX_OVERLAPPED];
197 #define bhfi_dev(bhfi) ((bhfi)->dwVolumeSerialNumber)
198 /* Treat FileIndex as i-node. We should remove a sequence number
199 * which is high-16-bits of nFileIndexHigh. */
200 #define bhfi_ino(bhfi) \
201 ((((int64_t)((bhfi)->nFileIndexHigh & 0x0000FFFFUL)) << 32) \
202 + (bhfi)->nFileIndexLow)
204 /* Definitions for tree.flags bitmap. */
205 #define hasStat 16 /* The st entry is valid. */
206 #define hasLstat 32 /* The lst entry is valid. */
207 #define needsRestoreTimes 128
210 tree_dir_next_windows(struct tree *t, const wchar_t *pattern);
212 /* Initiate/terminate a tree traversal. */
213 static struct tree *tree_open(const wchar_t *, int, int);
214 static struct tree *tree_reopen(struct tree *, const wchar_t *, int);
215 static void tree_close(struct tree *);
216 static void tree_free(struct tree *);
217 static void tree_push(struct tree *, const wchar_t *, const wchar_t *,
218 int, int64_t, int64_t, struct restore_time *);
221 * tree_next() returns Zero if there is no next entry, non-zero if
222 * there is. Note that directories are visited three times.
223 * Directories are always visited first as part of enumerating their
224 * parent; that is a "regular" visit. If tree_descend() is invoked at
225 * that time, the directory is added to a work list and will
226 * subsequently be visited two more times: once just after descending
227 * into the directory ("postdescent") and again just after ascending
228 * back to the parent ("postascent").
230 * TREE_ERROR_DIR is returned if the descent failed (because the
231 * directory couldn't be opened, for instance). This is returned
232 * instead of TREE_POSTDESCENT/TREE_POSTASCENT. TREE_ERROR_DIR is not a
233 * fatal error, but it does imply that the relevant subtree won't be
234 * visited. TREE_ERROR_FATAL is returned for an error that left the
235 * traversal completely hosed. Right now, this is only returned for
236 * chdir() failures during ascent.
238 #define TREE_REGULAR 1
239 #define TREE_POSTDESCENT 2
240 #define TREE_POSTASCENT 3
241 #define TREE_ERROR_DIR -1
242 #define TREE_ERROR_FATAL -2
244 static int tree_next(struct tree *);
247 * Return information about the current entry.
251 * The current full pathname, length of the full pathname, and a name
252 * that can be used to access the file. Because tree does use chdir
253 * extensively, the access path is almost never the same as the full
257 static const wchar_t *tree_current_path(struct tree *);
258 static const wchar_t *tree_current_access_path(struct tree *);
261 * Request the lstat() or stat() data for the current path. Since the
262 * tree package needs to do some of this anyway, and caches the
263 * results, you should take advantage of it here if you need it rather
264 * than make a redundant stat() or lstat() call of your own.
266 static const BY_HANDLE_FILE_INFORMATION *tree_current_stat(struct tree *);
267 static const BY_HANDLE_FILE_INFORMATION *tree_current_lstat(struct tree *);
269 /* The following functions use tricks to avoid a certain number of
270 * stat()/lstat() calls. */
271 /* "is_physical_dir" is equivalent to S_ISDIR(tree_current_lstat()->st_mode) */
272 static int tree_current_is_physical_dir(struct tree *);
273 /* "is_physical_link" is equivalent to S_ISLNK(tree_current_lstat()->st_mode) */
274 static int tree_current_is_physical_link(struct tree *);
275 /* Instead of archive_entry_copy_stat for BY_HANDLE_FILE_INFORMATION */
276 static void tree_archive_entry_copy_bhfi(struct archive_entry *,
277 struct tree *, const BY_HANDLE_FILE_INFORMATION *);
278 /* "is_dir" is equivalent to S_ISDIR(tree_current_stat()->st_mode) */
279 static int tree_current_is_dir(struct tree *);
280 static int update_current_filesystem(struct archive_read_disk *a,
282 static int setup_current_filesystem(struct archive_read_disk *);
283 static int tree_target_is_same_as_parent(struct tree *,
284 const BY_HANDLE_FILE_INFORMATION *);
286 static int _archive_read_disk_open_w(struct archive *, const wchar_t *);
287 static int _archive_read_free(struct archive *);
288 static int _archive_read_close(struct archive *);
289 static int _archive_read_data_block(struct archive *,
290 const void **, size_t *, int64_t *);
291 static int _archive_read_next_header(struct archive *,
292 struct archive_entry **);
293 static int _archive_read_next_header2(struct archive *,
294 struct archive_entry *);
295 static const char *trivial_lookup_gname(void *, int64_t gid);
296 static const char *trivial_lookup_uname(void *, int64_t uid);
297 static int setup_sparse(struct archive_read_disk *, struct archive_entry *);
298 static int close_and_restore_time(HANDLE, struct tree *,
299 struct restore_time *);
300 static int setup_sparse_from_disk(struct archive_read_disk *,
301 struct archive_entry *, HANDLE);
305 static struct archive_vtable *
306 archive_read_disk_vtable(void)
308 static struct archive_vtable av;
309 static int inited = 0;
312 av.archive_free = _archive_read_free;
313 av.archive_close = _archive_read_close;
314 av.archive_read_data_block = _archive_read_data_block;
315 av.archive_read_next_header = _archive_read_next_header;
316 av.archive_read_next_header2 = _archive_read_next_header2;
323 archive_read_disk_gname(struct archive *_a, int64_t gid)
325 struct archive_read_disk *a = (struct archive_read_disk *)_a;
326 if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
327 ARCHIVE_STATE_ANY, "archive_read_disk_gname"))
329 if (a->lookup_gname == NULL)
331 return ((*a->lookup_gname)(a->lookup_gname_data, gid));
335 archive_read_disk_uname(struct archive *_a, int64_t uid)
337 struct archive_read_disk *a = (struct archive_read_disk *)_a;
338 if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
339 ARCHIVE_STATE_ANY, "archive_read_disk_uname"))
341 if (a->lookup_uname == NULL)
343 return ((*a->lookup_uname)(a->lookup_uname_data, uid));
347 archive_read_disk_set_gname_lookup(struct archive *_a,
349 const char * (*lookup_gname)(void *private, int64_t gid),
350 void (*cleanup_gname)(void *private))
352 struct archive_read_disk *a = (struct archive_read_disk *)_a;
353 archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC,
354 ARCHIVE_STATE_ANY, "archive_read_disk_set_gname_lookup");
356 if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL)
357 (a->cleanup_gname)(a->lookup_gname_data);
359 a->lookup_gname = lookup_gname;
360 a->cleanup_gname = cleanup_gname;
361 a->lookup_gname_data = private_data;
366 archive_read_disk_set_uname_lookup(struct archive *_a,
368 const char * (*lookup_uname)(void *private, int64_t uid),
369 void (*cleanup_uname)(void *private))
371 struct archive_read_disk *a = (struct archive_read_disk *)_a;
372 archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC,
373 ARCHIVE_STATE_ANY, "archive_read_disk_set_uname_lookup");
375 if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL)
376 (a->cleanup_uname)(a->lookup_uname_data);
378 a->lookup_uname = lookup_uname;
379 a->cleanup_uname = cleanup_uname;
380 a->lookup_uname_data = private_data;
385 * Create a new archive_read_disk object and initialize it with global state.
388 archive_read_disk_new(void)
390 struct archive_read_disk *a;
392 a = (struct archive_read_disk *)malloc(sizeof(*a));
395 memset(a, 0, sizeof(*a));
396 a->archive.magic = ARCHIVE_READ_DISK_MAGIC;
397 a->archive.state = ARCHIVE_STATE_NEW;
398 a->archive.vtable = archive_read_disk_vtable();
399 a->entry = archive_entry_new2(&a->archive);
400 a->lookup_uname = trivial_lookup_uname;
401 a->lookup_gname = trivial_lookup_gname;
402 a->enable_copyfile = 1;
403 a->traverse_mount_points = 1;
404 return (&a->archive);
408 _archive_read_free(struct archive *_a)
410 struct archive_read_disk *a = (struct archive_read_disk *)_a;
415 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
416 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
418 if (a->archive.state != ARCHIVE_STATE_CLOSED)
419 r = _archive_read_close(&a->archive);
424 if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL)
425 (a->cleanup_gname)(a->lookup_gname_data);
426 if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL)
427 (a->cleanup_uname)(a->lookup_uname_data);
428 archive_string_free(&a->archive.error_string);
429 archive_entry_free(a->entry);
430 a->archive.magic = 0;
436 _archive_read_close(struct archive *_a)
438 struct archive_read_disk *a = (struct archive_read_disk *)_a;
440 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
441 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
443 if (a->archive.state != ARCHIVE_STATE_FATAL)
444 a->archive.state = ARCHIVE_STATE_CLOSED;
452 setup_symlink_mode(struct archive_read_disk *a, char symlink_mode,
455 a->symlink_mode = symlink_mode;
456 a->follow_symlinks = follow_symlinks;
457 if (a->tree != NULL) {
458 a->tree->initial_symlink_mode = a->symlink_mode;
459 a->tree->symlink_mode = a->symlink_mode;
464 archive_read_disk_set_symlink_logical(struct archive *_a)
466 struct archive_read_disk *a = (struct archive_read_disk *)_a;
467 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
468 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical");
469 setup_symlink_mode(a, 'L', 1);
474 archive_read_disk_set_symlink_physical(struct archive *_a)
476 struct archive_read_disk *a = (struct archive_read_disk *)_a;
477 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
478 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical");
479 setup_symlink_mode(a, 'P', 0);
484 archive_read_disk_set_symlink_hybrid(struct archive *_a)
486 struct archive_read_disk *a = (struct archive_read_disk *)_a;
487 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
488 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid");
489 setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */
494 archive_read_disk_set_atime_restored(struct archive *_a)
496 struct archive_read_disk *a = (struct archive_read_disk *)_a;
497 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
498 ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime");
501 a->tree->flags |= needsRestoreTimes;
506 archive_read_disk_set_behavior(struct archive *_a, int flags)
508 struct archive_read_disk *a = (struct archive_read_disk *)_a;
511 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
512 ARCHIVE_STATE_ANY, "archive_read_disk_honor_nodump");
514 if (flags & ARCHIVE_READDISK_RESTORE_ATIME)
515 r = archive_read_disk_set_atime_restored(_a);
519 a->tree->flags &= ~needsRestoreTimes;
521 if (flags & ARCHIVE_READDISK_HONOR_NODUMP)
525 if (flags & ARCHIVE_READDISK_MAC_COPYFILE)
526 a->enable_copyfile = 1;
528 a->enable_copyfile = 0;
529 if (flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS)
530 a->traverse_mount_points = 0;
532 a->traverse_mount_points = 1;
537 * Trivial implementations of gname/uname lookup functions.
538 * These are normally overridden by the client, but these stub
539 * versions ensure that we always have something that works.
542 trivial_lookup_gname(void *private_data, int64_t gid)
544 (void)private_data; /* UNUSED */
545 (void)gid; /* UNUSED */
550 trivial_lookup_uname(void *private_data, int64_t uid)
552 (void)private_data; /* UNUSED */
553 (void)uid; /* UNUSED */
558 align_num_per_sector(struct tree *t, int64_t size)
562 size += t->current_filesystem->bytesPerSector -1;
563 surplus = size % t->current_filesystem->bytesPerSector;
569 start_next_async_read(struct archive_read_disk *a, struct tree *t)
571 struct la_overlapped *olp;
572 DWORD buffbytes, rbytes;
574 if (t->ol_remaining_bytes == 0)
575 return (ARCHIVE_EOF);
577 olp = &(t->ol[t->ol_idx_doing]);
578 t->ol_idx_doing = (t->ol_idx_doing + 1) % MAX_OVERLAPPED;
580 /* Allocate read buffer. */
581 if (olp->buff == NULL) {
583 size_t s = (size_t)align_num_per_sector(t, BUFFER_SIZE);
584 p = VirtualAlloc(NULL, s, MEM_COMMIT, PAGE_READWRITE);
586 archive_set_error(&a->archive, ENOMEM,
587 "Couldn't allocate memory");
588 a->archive.state = ARCHIVE_STATE_FATAL;
589 return (ARCHIVE_FATAL);
593 olp->_a = &a->archive;
594 olp->ol.hEvent = CreateEventW(NULL, TRUE, FALSE, NULL);
595 if (olp->ol.hEvent == NULL) {
596 la_dosmaperr(GetLastError());
597 archive_set_error(&a->archive, errno,
598 "CreateEvent failed");
599 a->archive.state = ARCHIVE_STATE_FATAL;
600 return (ARCHIVE_FATAL);
603 ResetEvent(olp->ol.hEvent);
605 buffbytes = (DWORD)olp->buff_size;
606 if (buffbytes > t->current_sparse->length)
607 buffbytes = (DWORD)t->current_sparse->length;
610 if (t->current_sparse->offset > t->ol_total) {
611 t->ol_remaining_bytes -=
612 t->current_sparse->offset - t->ol_total;
615 olp->offset = t->current_sparse->offset;
616 olp->ol.Offset = (DWORD)(olp->offset & 0xffffffff);
617 olp->ol.OffsetHigh = (DWORD)(olp->offset >> 32);
619 if (t->ol_remaining_bytes > buffbytes) {
620 olp->bytes_expected = buffbytes;
621 t->ol_remaining_bytes -= buffbytes;
623 olp->bytes_expected = (size_t)t->ol_remaining_bytes;
624 t->ol_remaining_bytes = 0;
626 olp->bytes_transferred = 0;
627 t->current_sparse->offset += buffbytes;
628 t->current_sparse->length -= buffbytes;
629 t->ol_total = t->current_sparse->offset;
630 if (t->current_sparse->length == 0 && t->ol_remaining_bytes > 0)
633 if (!ReadFile(t->entry_fh, olp->buff, buffbytes, &rbytes, &(olp->ol))) {
636 lasterr = GetLastError();
637 if (lasterr == ERROR_HANDLE_EOF) {
638 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
639 "Reading file truncated");
640 a->archive.state = ARCHIVE_STATE_FATAL;
641 return (ARCHIVE_FATAL);
642 } else if (lasterr != ERROR_IO_PENDING) {
643 if (lasterr == ERROR_NO_DATA)
645 else if (lasterr == ERROR_ACCESS_DENIED)
648 la_dosmaperr(lasterr);
649 archive_set_error(&a->archive, errno, "Read error");
650 a->archive.state = ARCHIVE_STATE_FATAL;
651 return (ARCHIVE_FATAL);
654 olp->bytes_transferred = rbytes;
657 return (t->ol_remaining_bytes == 0)? ARCHIVE_EOF: ARCHIVE_OK;
661 cancel_async(struct tree *t)
663 if (t->ol_num_doing != t->ol_num_done) {
664 CancelIo(t->entry_fh);
665 t->ol_num_doing = t->ol_num_done = 0;
670 _archive_read_data_block(struct archive *_a, const void **buff,
671 size_t *size, int64_t *offset)
673 struct archive_read_disk *a = (struct archive_read_disk *)_a;
674 struct tree *t = a->tree;
675 struct la_overlapped *olp;
676 DWORD bytes_transferred;
677 int r = ARCHIVE_FATAL;
679 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
680 "archive_read_data_block");
682 if (t->entry_eof || t->entry_remaining_bytes <= 0) {
684 goto abort_read_data;
688 * Make a request to read the file in asynchronous.
690 if (t->ol_num_doing == 0) {
692 r = start_next_async_read(a, t);
693 if (r == ARCHIVE_FATAL)
694 goto abort_read_data;
697 } while (r == ARCHIVE_OK && t->ol_num_doing < MAX_OVERLAPPED);
699 if (start_next_async_read(a, t) == ARCHIVE_FATAL)
700 goto abort_read_data;
703 olp = &(t->ol[t->ol_idx_done]);
704 t->ol_idx_done = (t->ol_idx_done + 1) % MAX_OVERLAPPED;
705 if (olp->bytes_transferred)
706 bytes_transferred = (DWORD)olp->bytes_transferred;
707 else if (!GetOverlappedResult(t->entry_fh, &(olp->ol),
708 &bytes_transferred, TRUE)) {
709 la_dosmaperr(GetLastError());
710 archive_set_error(&a->archive, errno,
711 "GetOverlappedResult failed");
712 a->archive.state = ARCHIVE_STATE_FATAL;
714 goto abort_read_data;
718 if (bytes_transferred == 0 ||
719 olp->bytes_expected != bytes_transferred) {
720 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
721 "Reading file truncated");
722 a->archive.state = ARCHIVE_STATE_FATAL;
724 goto abort_read_data;
728 *size = bytes_transferred;
729 *offset = olp->offset;
730 if (olp->offset > t->entry_total)
731 t->entry_remaining_bytes -= olp->offset - t->entry_total;
732 t->entry_total = olp->offset + *size;
733 t->entry_remaining_bytes -= *size;
734 if (t->entry_remaining_bytes == 0) {
735 /* Close the current file descriptor */
736 close_and_restore_time(t->entry_fh, t, &t->restore_time);
737 t->entry_fh = INVALID_HANDLE_VALUE;
745 *offset = t->entry_total;
746 if (t->entry_fh != INVALID_HANDLE_VALUE) {
748 /* Close the current file descriptor */
749 close_and_restore_time(t->entry_fh, t, &t->restore_time);
750 t->entry_fh = INVALID_HANDLE_VALUE;
756 next_entry(struct archive_read_disk *a, struct tree *t,
757 struct archive_entry *entry)
759 const BY_HANDLE_FILE_INFORMATION *st;
760 const BY_HANDLE_FILE_INFORMATION *lst;
768 switch (tree_next(t)) {
769 case TREE_ERROR_FATAL:
770 archive_set_error(&a->archive, t->tree_errno,
771 "%ls: Unable to continue traversing directory tree",
772 tree_current_path(t));
773 a->archive.state = ARCHIVE_STATE_FATAL;
774 return (ARCHIVE_FATAL);
776 archive_set_error(&a->archive, t->tree_errno,
777 "%ls: Couldn't visit directory",
778 tree_current_path(t));
779 return (ARCHIVE_FAILED);
781 return (ARCHIVE_EOF);
782 case TREE_POSTDESCENT:
783 case TREE_POSTASCENT:
786 lst = tree_current_lstat(t);
788 archive_set_error(&a->archive, t->tree_errno,
790 tree_current_path(t));
791 return (ARCHIVE_FAILED);
795 } while (lst == NULL);
797 archive_entry_copy_pathname_w(entry, tree_current_path(t));
800 * Perform path matching.
803 r = archive_match_path_excluded(a->matching, entry);
805 archive_set_error(&(a->archive), errno,
806 "Faild : %s", archive_error_string(a->matching));
810 if (a->excluded_cb_func)
811 a->excluded_cb_func(&(a->archive),
812 a->excluded_cb_data, entry);
813 return (ARCHIVE_RETRY);
818 * Distinguish 'L'/'P'/'H' symlink following.
820 switch(t->symlink_mode) {
822 /* 'H': After the first item, rest like 'P'. */
823 t->symlink_mode = 'P';
824 /* 'H': First item (from command line) like 'L'. */
827 /* 'L': Do descend through a symlink to dir. */
828 descend = tree_current_is_dir(t);
829 /* 'L': Follow symlinks to files. */
830 a->symlink_mode = 'L';
831 a->follow_symlinks = 1;
832 /* 'L': Archive symlinks as targets, if we can. */
833 st = tree_current_stat(t);
834 if (st != NULL && !tree_target_is_same_as_parent(t, st))
836 /* If stat fails, we have a broken symlink;
837 * in that case, don't follow the link. */
840 /* 'P': Don't descend through a symlink to dir. */
841 descend = tree_current_is_physical_dir(t);
842 /* 'P': Don't follow symlinks to files. */
843 a->symlink_mode = 'P';
844 a->follow_symlinks = 0;
845 /* 'P': Archive symlinks as symlinks. */
850 if (update_current_filesystem(a, bhfi_dev(st)) != ARCHIVE_OK) {
851 a->archive.state = ARCHIVE_STATE_FATAL;
852 return (ARCHIVE_FATAL);
854 if (t->initial_filesystem_id == -1)
855 t->initial_filesystem_id = t->current_filesystem_id;
856 if (!a->traverse_mount_points) {
857 if (t->initial_filesystem_id != t->current_filesystem_id)
858 return (ARCHIVE_RETRY);
860 t->descend = descend;
862 tree_archive_entry_copy_bhfi(entry, t, st);
864 /* Save the times to be restored. This must be in before
865 * calling archive_read_disk_descend() or any chance of it,
866 * especially, invokng a callback. */
867 t->restore_time.lastWriteTime = st->ftLastWriteTime;
868 t->restore_time.lastAccessTime = st->ftLastAccessTime;
869 t->restore_time.filetype = archive_entry_filetype(entry);
872 * Perform time matching.
875 r = archive_match_time_excluded(a->matching, entry);
877 archive_set_error(&(a->archive), errno,
878 "Faild : %s", archive_error_string(a->matching));
882 if (a->excluded_cb_func)
883 a->excluded_cb_func(&(a->archive),
884 a->excluded_cb_data, entry);
885 return (ARCHIVE_RETRY);
889 /* Lookup uname/gname */
890 name = archive_read_disk_uname(&(a->archive), archive_entry_uid(entry));
892 archive_entry_copy_uname(entry, name);
893 name = archive_read_disk_gname(&(a->archive), archive_entry_gid(entry));
895 archive_entry_copy_gname(entry, name);
898 * Perform owner matching.
901 r = archive_match_owner_excluded(a->matching, entry);
903 archive_set_error(&(a->archive), errno,
904 "Faild : %s", archive_error_string(a->matching));
908 if (a->excluded_cb_func)
909 a->excluded_cb_func(&(a->archive),
910 a->excluded_cb_data, entry);
911 return (ARCHIVE_RETRY);
916 * Invoke a meta data filter callback.
918 if (a->metadata_filter_func) {
919 if (!a->metadata_filter_func(&(a->archive),
920 a->metadata_filter_data, entry))
921 return (ARCHIVE_RETRY);
924 archive_entry_copy_sourcepath_w(entry, tree_current_access_path(t));
927 if (archive_entry_filetype(entry) == AE_IFREG &&
928 archive_entry_size(entry) > 0) {
929 DWORD flags = FILE_FLAG_BACKUP_SEMANTICS;
931 flags |= FILE_FLAG_OVERLAPPED;
933 flags |= FILE_FLAG_NO_BUFFERING;
935 flags |= FILE_FLAG_SEQUENTIAL_SCAN;
936 t->entry_fh = CreateFileW(tree_current_access_path(t),
937 GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, flags, NULL);
938 if (t->entry_fh == INVALID_HANDLE_VALUE) {
939 archive_set_error(&a->archive, errno,
940 "Couldn't open %ls", tree_current_path(a->tree));
941 return (ARCHIVE_FAILED);
944 /* Find sparse data from the disk. */
945 if (archive_entry_hardlink(entry) == NULL &&
946 (st->dwFileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) != 0)
947 r = setup_sparse_from_disk(a, entry, t->entry_fh);
953 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
956 struct archive_read_disk *a = (struct archive_read_disk *)_a;
958 ret = _archive_read_next_header2(_a, a->entry);
964 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
966 struct archive_read_disk *a = (struct archive_read_disk *)_a;
970 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
971 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
972 "archive_read_next_header2");
975 if (t->entry_fh != INVALID_HANDLE_VALUE) {
977 close_and_restore_time(t->entry_fh, t, &t->restore_time);
978 t->entry_fh = INVALID_HANDLE_VALUE;
981 while ((r = next_entry(a, t, entry)) == ARCHIVE_RETRY)
982 archive_entry_clear(entry);
985 * EOF and FATAL are persistent at this layer. By
986 * modifying the state, we guarantee that future calls to
987 * read a header or read data will fail.
991 a->archive.state = ARCHIVE_STATE_EOF;
996 if (archive_entry_filetype(entry) == AE_IFREG) {
997 t->entry_remaining_bytes = archive_entry_size(entry);
998 t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0;
1000 setup_sparse(a, entry) != ARCHIVE_OK)
1001 return (ARCHIVE_FATAL);
1003 t->entry_remaining_bytes = 0;
1006 t->ol_idx_doing = t->ol_idx_done = 0;
1007 t->ol_num_doing = t->ol_num_done = 0;
1008 t->ol_remaining_bytes = t->entry_remaining_bytes;
1010 a->archive.state = ARCHIVE_STATE_DATA;
1015 a->archive.state = ARCHIVE_STATE_FATAL;
1019 __archive_reset_read_data(&a->archive);
1024 setup_sparse(struct archive_read_disk *a, struct archive_entry *entry)
1026 struct tree *t = a->tree;
1027 int64_t aligned, length, offset;
1030 t->sparse_count = archive_entry_sparse_reset(entry);
1031 if (t->sparse_count+1 > t->sparse_list_size) {
1032 free(t->sparse_list);
1033 t->sparse_list_size = t->sparse_count + 1;
1034 t->sparse_list = malloc(sizeof(t->sparse_list[0]) *
1035 t->sparse_list_size);
1036 if (t->sparse_list == NULL) {
1037 t->sparse_list_size = 0;
1038 archive_set_error(&a->archive, ENOMEM,
1039 "Can't allocate data");
1040 a->archive.state = ARCHIVE_STATE_FATAL;
1041 return (ARCHIVE_FATAL);
1045 * Get sparse list and make sure those offsets and lengths are
1046 * aligned by a sector size.
1048 for (i = 0; i < t->sparse_count; i++) {
1049 archive_entry_sparse_next(entry, &offset, &length);
1050 aligned = align_num_per_sector(t, offset);
1051 if (aligned != offset) {
1052 aligned -= t->current_filesystem->bytesPerSector;
1053 length += offset - aligned;
1055 t->sparse_list[i].offset = aligned;
1056 aligned = align_num_per_sector(t, length);
1057 t->sparse_list[i].length = aligned;
1060 aligned = align_num_per_sector(t, archive_entry_size(entry));
1062 t->sparse_list[i].offset = 0;
1063 t->sparse_list[i].length = aligned;
1067 t->sparse_list[i].offset = aligned;
1068 t->sparse_list[i].length = 0;
1069 for (i = 0; i < last; i++) {
1070 if ((t->sparse_list[i].offset +
1071 t->sparse_list[i].length) <=
1072 t->sparse_list[i+1].offset)
1075 * Now sparse_list[i+1] is overlapped by sparse_list[i].
1078 length = t->sparse_list[i+1].offset -
1079 t->sparse_list[i].offset;
1080 t->sparse_list[i+1].offset = t->sparse_list[i].offset;
1081 t->sparse_list[i+1].length += length;
1082 /* Remove sparse_list[i]. */
1083 for (j = i; j < last; j++) {
1084 t->sparse_list[j].offset =
1085 t->sparse_list[j+1].offset;
1086 t->sparse_list[j].length =
1087 t->sparse_list[j+1].length;
1092 t->current_sparse = t->sparse_list;
1094 return (ARCHIVE_OK);
1098 archive_read_disk_set_matching(struct archive *_a, struct archive *_ma,
1099 void (*_excluded_func)(struct archive *, void *, struct archive_entry *),
1102 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1103 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1104 ARCHIVE_STATE_ANY, "archive_read_disk_set_matching");
1106 a->excluded_cb_func = _excluded_func;
1107 a->excluded_cb_data = _client_data;
1108 return (ARCHIVE_OK);
1112 archive_read_disk_set_metadata_filter_callback(struct archive *_a,
1113 int (*_metadata_filter_func)(struct archive *, void *,
1114 struct archive_entry *), void *_client_data)
1116 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1118 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY,
1119 "archive_read_disk_set_metadata_filter_callback");
1121 a->metadata_filter_func = _metadata_filter_func;
1122 a->metadata_filter_data = _client_data;
1123 return (ARCHIVE_OK);
1127 archive_read_disk_can_descend(struct archive *_a)
1129 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1130 struct tree *t = a->tree;
1132 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1133 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1134 "archive_read_disk_can_descend");
1136 return (t->visit_type == TREE_REGULAR && t->descend);
1140 * Called by the client to mark the directory just returned from
1141 * tree_next() as needing to be visited.
1144 archive_read_disk_descend(struct archive *_a)
1146 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1147 struct tree *t = a->tree;
1149 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1150 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1151 "archive_read_disk_descend");
1153 if (t->visit_type != TREE_REGULAR || !t->descend)
1154 return (ARCHIVE_OK);
1156 if (tree_current_is_physical_dir(t)) {
1157 tree_push(t, t->basename, t->full_path.s,
1158 t->current_filesystem_id,
1159 bhfi_dev(&(t->lst)), bhfi_ino(&(t->lst)),
1161 t->stack->flags |= isDir;
1162 } else if (tree_current_is_dir(t)) {
1163 tree_push(t, t->basename, t->full_path.s,
1164 t->current_filesystem_id,
1165 bhfi_dev(&(t->st)), bhfi_ino(&(t->st)),
1167 t->stack->flags |= isDirLink;
1170 return (ARCHIVE_OK);
1174 archive_read_disk_open(struct archive *_a, const char *pathname)
1176 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1177 struct archive_wstring wpath;
1180 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1181 ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1182 "archive_read_disk_open");
1183 archive_clear_error(&a->archive);
1185 /* Make a wchar_t string from a char string. */
1186 archive_string_init(&wpath);
1187 if (archive_wstring_append_from_mbs(&wpath, pathname,
1188 strlen(pathname)) != 0) {
1189 if (errno == ENOMEM)
1190 archive_set_error(&a->archive, ENOMEM,
1191 "Can't allocate memory");
1193 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1194 "Can't convert a path to a wchar_t string");
1195 a->archive.state = ARCHIVE_STATE_FATAL;
1196 ret = ARCHIVE_FATAL;
1198 ret = _archive_read_disk_open_w(_a, wpath.s);
1200 archive_wstring_free(&wpath);
1205 archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1207 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1209 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1210 ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1211 "archive_read_disk_open_w");
1212 archive_clear_error(&a->archive);
1214 return (_archive_read_disk_open_w(_a, pathname));
1218 _archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1220 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1222 if (a->tree != NULL)
1223 a->tree = tree_reopen(a->tree, pathname, a->restore_time);
1225 a->tree = tree_open(pathname, a->symlink_mode, a->restore_time);
1226 if (a->tree == NULL) {
1227 archive_set_error(&a->archive, ENOMEM,
1228 "Can't allocate directory traversal data");
1229 a->archive.state = ARCHIVE_STATE_FATAL;
1230 return (ARCHIVE_FATAL);
1232 a->archive.state = ARCHIVE_STATE_HEADER;
1234 return (ARCHIVE_OK);
1238 * Return a current filesystem ID which is index of the filesystem entry
1239 * you've visited through archive_read_disk.
1242 archive_read_disk_current_filesystem(struct archive *_a)
1244 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1246 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1247 "archive_read_disk_current_filesystem");
1249 return (a->tree->current_filesystem_id);
1253 update_current_filesystem(struct archive_read_disk *a, int64_t dev)
1255 struct tree *t = a->tree;
1258 if (t->current_filesystem != NULL &&
1259 t->current_filesystem->dev == dev)
1260 return (ARCHIVE_OK);
1262 for (i = 0; i < t->max_filesystem_id; i++) {
1263 if (t->filesystem_table[i].dev == dev) {
1264 /* There is the filesytem ID we've already generated. */
1265 t->current_filesystem_id = i;
1266 t->current_filesystem = &(t->filesystem_table[i]);
1267 return (ARCHIVE_OK);
1272 * There is a new filesytem, we generate a new ID for.
1274 fid = t->max_filesystem_id++;
1275 if (t->max_filesystem_id > t->allocated_filesytem) {
1279 s = t->max_filesystem_id * 2;
1280 p = realloc(t->filesystem_table,
1281 s * sizeof(*t->filesystem_table));
1283 archive_set_error(&a->archive, ENOMEM,
1284 "Can't allocate tar data");
1285 return (ARCHIVE_FATAL);
1287 t->filesystem_table = (struct filesystem *)p;
1288 t->allocated_filesytem = (int)s;
1290 t->current_filesystem_id = fid;
1291 t->current_filesystem = &(t->filesystem_table[fid]);
1292 t->current_filesystem->dev = dev;
1294 return (setup_current_filesystem(a));
1298 * Returns 1 if current filesystem is generated filesystem, 0 if it is not
1299 * or -1 if it is unknown.
1302 archive_read_disk_current_filesystem_is_synthetic(struct archive *_a)
1304 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1306 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1307 "archive_read_disk_current_filesystem");
1309 return (a->tree->current_filesystem->synthetic);
1313 * Returns 1 if current filesystem is remote filesystem, 0 if it is not
1314 * or -1 if it is unknown.
1317 archive_read_disk_current_filesystem_is_remote(struct archive *_a)
1319 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1321 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1322 "archive_read_disk_current_filesystem");
1324 return (a->tree->current_filesystem->remote);
1328 * If symlink is broken, statfs or statvfs will fail.
1329 * Use its directory path instead.
1332 safe_path_for_statfs(struct tree *t)
1334 const wchar_t *path;
1335 wchar_t *cp, *p = NULL;
1337 path = tree_current_access_path(t);
1338 if (tree_current_stat(t) == NULL) {
1340 cp = wcsrchr(p, '/');
1341 if (cp != NULL && wcslen(cp) >= 2) {
1352 * Get conditions of synthetic and remote on Windows
1355 setup_current_filesystem(struct archive_read_disk *a)
1357 struct tree *t = a->tree;
1361 t->current_filesystem->synthetic = -1;/* Not supported */
1362 path = safe_path_for_statfs(t);
1363 if (!GetVolumePathNameW(path, vol, sizeof(vol)/sizeof(vol[0]))) {
1365 t->current_filesystem->remote = -1;
1366 t->current_filesystem->bytesPerSector = 0;
1367 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1368 "GetVolumePathName failed: %d", (int)GetLastError());
1369 return (ARCHIVE_FAILED);
1372 switch (GetDriveTypeW(vol)) {
1374 case DRIVE_NO_ROOT_DIR:
1375 t->current_filesystem->remote = -1;
1378 t->current_filesystem->remote = 1;
1381 t->current_filesystem->remote = 0;
1385 if (!GetDiskFreeSpaceW(vol, NULL,
1386 &(t->current_filesystem->bytesPerSector), NULL, NULL)) {
1387 t->current_filesystem->bytesPerSector = 0;
1388 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1389 "GetDiskFreeSpace failed: %d", (int)GetLastError());
1390 return (ARCHIVE_FAILED);
1393 return (ARCHIVE_OK);
1397 close_and_restore_time(HANDLE h, struct tree *t, struct restore_time *rt)
1402 if (h == INVALID_HANDLE_VALUE && AE_IFLNK == rt->filetype)
1405 /* Close a file descritor.
1406 * It will not be used for SetFileTime() because it has been opened
1407 * by a read only mode.
1409 if (h != INVALID_HANDLE_VALUE)
1411 if ((t->flags & needsRestoreTimes) == 0)
1414 handle = CreateFileW(rt->full_path, FILE_WRITE_ATTRIBUTES,
1415 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
1416 if (handle == INVALID_HANDLE_VALUE) {
1421 if (SetFileTime(handle, NULL, &rt->lastAccessTime,
1422 &rt->lastWriteTime) == 0) {
1427 CloseHandle(handle);
1432 * Add a directory path to the current stack.
1435 tree_push(struct tree *t, const wchar_t *path, const wchar_t *full_path,
1436 int filesystem_id, int64_t dev, int64_t ino, struct restore_time *rt)
1438 struct tree_entry *te;
1440 te = malloc(sizeof(*te));
1441 memset(te, 0, sizeof(*te));
1442 te->next = t->stack;
1443 te->parent = t->current;
1445 te->depth = te->parent->depth + 1;
1447 archive_string_init(&te->name);
1448 archive_wstrcpy(&te->name, path);
1449 archive_string_init(&te->full_path);
1450 archive_wstrcpy(&te->full_path, full_path);
1451 te->flags = needsDescent | needsOpen | needsAscent;
1452 te->filesystem_id = filesystem_id;
1455 te->dirname_length = t->dirname_length;
1456 te->full_path_dir_length = t->full_path_dir_length;
1457 te->restore_time.full_path = te->full_path.s;
1459 te->restore_time.lastWriteTime = rt->lastWriteTime;
1460 te->restore_time.lastAccessTime = rt->lastAccessTime;
1461 te->restore_time.filetype = rt->filetype;
1466 * Append a name to the current dir path.
1469 tree_append(struct tree *t, const wchar_t *name, size_t name_length)
1473 t->path.s[t->dirname_length] = L'\0';
1474 t->path.length = t->dirname_length;
1475 /* Strip trailing '/' from name, unless entire name is "/". */
1476 while (name_length > 1 && name[name_length - 1] == L'/')
1479 /* Resize pathname buffer as needed. */
1480 size_needed = name_length + t->dirname_length + 2;
1481 archive_wstring_ensure(&t->path, size_needed);
1482 /* Add a separating '/' if it's needed. */
1483 if (t->dirname_length > 0 &&
1484 t->path.s[archive_strlen(&t->path)-1] != L'/')
1485 archive_wstrappend_wchar(&t->path, L'/');
1486 t->basename = t->path.s + archive_strlen(&t->path);
1487 archive_wstrncat(&t->path, name, name_length);
1488 t->restore_time.full_path = t->basename;
1489 if (t->full_path_dir_length > 0) {
1490 t->full_path.s[t->full_path_dir_length] = L'\0';
1491 t->full_path.length = t->full_path_dir_length;
1492 size_needed = name_length + t->full_path_dir_length + 2;
1493 archive_wstring_ensure(&t->full_path, size_needed);
1494 /* Add a separating '\' if it's needed. */
1495 if (t->full_path.s[archive_strlen(&t->full_path)-1] != L'\\')
1496 archive_wstrappend_wchar(&t->full_path, L'\\');
1497 archive_wstrncat(&t->full_path, name, name_length);
1498 t->restore_time.full_path = t->full_path.s;
1503 * Open a directory tree for traversal.
1505 static struct tree *
1506 tree_open(const wchar_t *path, int symlink_mode, int restore_time)
1510 t = malloc(sizeof(*t));
1511 memset(t, 0, sizeof(*t));
1512 archive_string_init(&(t->full_path));
1513 archive_string_init(&t->path);
1514 archive_wstring_ensure(&t->path, 15);
1515 t->initial_symlink_mode = symlink_mode;
1516 return (tree_reopen(t, path, restore_time));
1519 static struct tree *
1520 tree_reopen(struct tree *t, const wchar_t *path, int restore_time)
1522 struct archive_wstring ws;
1523 wchar_t *pathname, *p, *base;
1525 t->flags = (restore_time)?needsRestoreTimes:0;
1528 t->full_path_dir_length = 0;
1529 t->dirname_length = 0;
1533 t->d = INVALID_HANDLE_VALUE;
1534 t->symlink_mode = t->initial_symlink_mode;
1535 archive_string_empty(&(t->full_path));
1536 archive_string_empty(&t->path);
1537 t->entry_fh = INVALID_HANDLE_VALUE;
1539 t->entry_remaining_bytes = 0;
1540 t->initial_filesystem_id = -1;
1542 /* Get wchar_t strings from char strings. */
1543 archive_string_init(&ws);
1544 archive_wstrcpy(&ws, path);
1546 /* Get a full-path-name. */
1547 p = __la_win_permissive_name_w(pathname);
1550 archive_wstrcpy(&(t->full_path), p);
1553 /* Convert path separators from '\' to '/' */
1554 for (p = pathname; *p != L'\0'; ++p) {
1560 /* First item is set up a lot like a symlink traversal. */
1561 /* printf("Looking for wildcard in %s\n", path); */
1562 if ((base[0] == L'/' && base[1] == L'/' &&
1563 base[2] == L'?' && base[3] == L'/' &&
1564 (wcschr(base+4, L'*') || wcschr(base+4, L'?'))) ||
1565 (!(base[0] == L'/' && base[1] == L'/' &&
1566 base[2] == L'?' && base[3] == L'/') &&
1567 (wcschr(base, L'*') || wcschr(base, L'?')))) {
1568 // It has a wildcard in it...
1569 // Separate the last element.
1570 p = wcsrchr(base, L'/');
1573 tree_append(t, base, p - base);
1574 t->dirname_length = archive_strlen(&t->path);
1577 p = wcsrchr(t->full_path.s, L'\\');
1580 t->full_path.length = wcslen(t->full_path.s);
1581 t->full_path_dir_length = archive_strlen(&t->full_path);
1584 tree_push(t, base, t->full_path.s, 0, 0, 0, NULL);
1585 archive_wstring_free(&ws);
1586 t->stack->flags = needsFirstVisit;
1588 * Debug flag for Direct IO(No buffering) or Async IO.
1589 * Those dependent on environment variable switches
1590 * will be removed until next release.
1594 if ((e = getenv("LIBARCHIVE_DIRECT_IO")) != NULL) {
1599 fprintf(stderr, "LIBARCHIVE_DIRECT_IO=%s\n",
1600 (t->direct_io)?"Enabled":"Disabled");
1602 t->direct_io = DIRECT_IO;
1603 if ((e = getenv("LIBARCHIVE_ASYNC_IO")) != NULL) {
1608 fprintf(stderr, "LIBARCHIVE_ASYNC_IO=%s\n",
1609 (t->async_io)?"Enabled":"Disabled");
1611 t->async_io = ASYNC_IO;
1615 archive_wstring_free(&ws);
1621 tree_descent(struct tree *t)
1623 t->dirname_length = archive_strlen(&t->path);
1624 t->full_path_dir_length = archive_strlen(&t->full_path);
1630 * We've finished a directory; ascend back to the parent.
1633 tree_ascend(struct tree *t)
1635 struct tree_entry *te;
1639 close_and_restore_time(INVALID_HANDLE_VALUE, t, &te->restore_time);
1644 * Pop the working stack.
1647 tree_pop(struct tree *t)
1649 struct tree_entry *te;
1651 t->full_path.s[t->full_path_dir_length] = L'\0';
1652 t->full_path.length = t->full_path_dir_length;
1653 t->path.s[t->dirname_length] = L'\0';
1654 t->path.length = t->dirname_length;
1655 if (t->stack == t->current && t->current != NULL)
1656 t->current = t->current->parent;
1658 t->stack = te->next;
1659 t->dirname_length = te->dirname_length;
1660 t->basename = t->path.s + t->dirname_length;
1661 t->full_path_dir_length = te->full_path_dir_length;
1662 while (t->basename[0] == L'/')
1664 archive_wstring_free(&te->name);
1665 archive_wstring_free(&te->full_path);
1670 * Get the next item in the tree traversal.
1673 tree_next(struct tree *t)
1677 while (t->stack != NULL) {
1678 /* If there's an open dir, get the next entry from there. */
1679 if (t->d != INVALID_HANDLE_VALUE) {
1680 r = tree_dir_next_windows(t, NULL);
1686 if (t->stack->flags & needsFirstVisit) {
1687 wchar_t *d = t->stack->name.s;
1688 t->stack->flags &= ~needsFirstVisit;
1689 if (!(d[0] == L'/' && d[1] == L'/' &&
1690 d[2] == L'?' && d[3] == L'/') &&
1691 (wcschr(d, L'*') || wcschr(d, L'?'))) {
1692 r = tree_dir_next_windows(t, d);
1697 HANDLE h = FindFirstFileW(d, &t->_findData);
1698 if (h == INVALID_HANDLE_VALUE) {
1699 la_dosmaperr(GetLastError());
1700 t->tree_errno = errno;
1701 t->visit_type = TREE_ERROR_DIR;
1702 return (t->visit_type);
1704 t->findData = &t->_findData;
1707 /* Top stack item needs a regular visit. */
1708 t->current = t->stack;
1709 tree_append(t, t->stack->name.s,
1710 archive_strlen(&(t->stack->name)));
1711 //t->dirname_length = t->path_length;
1713 t->stack->flags &= ~needsFirstVisit;
1714 return (t->visit_type = TREE_REGULAR);
1715 } else if (t->stack->flags & needsDescent) {
1716 /* Top stack item is dir to descend into. */
1717 t->current = t->stack;
1718 tree_append(t, t->stack->name.s,
1719 archive_strlen(&(t->stack->name)));
1720 t->stack->flags &= ~needsDescent;
1721 r = tree_descent(t);
1726 t->visit_type = TREE_POSTDESCENT;
1727 return (t->visit_type);
1728 } else if (t->stack->flags & needsOpen) {
1729 t->stack->flags &= ~needsOpen;
1730 r = tree_dir_next_windows(t, L"*");
1734 } else if (t->stack->flags & needsAscent) {
1735 /* Top stack item is dir and we're done with it. */
1738 t->visit_type = r != 0 ? r : TREE_POSTASCENT;
1739 return (t->visit_type);
1741 /* Top item on stack is dead. */
1743 t->flags &= ~hasLstat;
1744 t->flags &= ~hasStat;
1747 return (t->visit_type = 0);
1751 tree_dir_next_windows(struct tree *t, const wchar_t *pattern)
1753 const wchar_t *name;
1758 if (pattern != NULL) {
1759 struct archive_wstring pt;
1761 archive_string_init(&pt);
1762 archive_wstring_ensure(&pt,
1763 archive_strlen(&(t->full_path))
1764 + 2 + wcslen(pattern));
1765 archive_wstring_copy(&pt, &(t->full_path));
1766 archive_wstrappend_wchar(&pt, L'\\');
1767 archive_wstrcat(&pt, pattern);
1768 t->d = FindFirstFileW(pt.s, &t->_findData);
1769 archive_wstring_free(&pt);
1770 if (t->d == INVALID_HANDLE_VALUE) {
1771 la_dosmaperr(GetLastError());
1772 t->tree_errno = errno;
1773 r = tree_ascend(t); /* Undo "chdir" */
1775 t->visit_type = r != 0 ? r : TREE_ERROR_DIR;
1776 return (t->visit_type);
1778 t->findData = &t->_findData;
1780 } else if (!FindNextFileW(t->d, &t->_findData)) {
1782 t->d = INVALID_HANDLE_VALUE;
1786 name = t->findData->cFileName;
1787 namelen = wcslen(name);
1788 t->flags &= ~hasLstat;
1789 t->flags &= ~hasStat;
1790 if (name[0] == L'.' && name[1] == L'\0')
1792 if (name[0] == L'.' && name[1] == L'.' && name[2] == L'\0')
1794 tree_append(t, name, namelen);
1795 return (t->visit_type = TREE_REGULAR);
1799 #define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000)
1801 fileTimeToUtc(const FILETIME *filetime, time_t *t, long *ns)
1805 utc.HighPart = filetime->dwHighDateTime;
1806 utc.LowPart = filetime->dwLowDateTime;
1807 if (utc.QuadPart >= EPOC_TIME) {
1808 utc.QuadPart -= EPOC_TIME;
1809 /* milli seconds base */
1810 *t = (time_t)(utc.QuadPart / 10000000);
1811 /* nano seconds base */
1812 *ns = (long)(utc.QuadPart % 10000000) * 100;
1820 entry_copy_bhfi(struct archive_entry *entry, const wchar_t *path,
1821 const WIN32_FIND_DATAW *findData,
1822 const BY_HANDLE_FILE_INFORMATION *bhfi)
1828 fileTimeToUtc(&bhfi->ftLastAccessTime, &secs, &nsecs);
1829 archive_entry_set_atime(entry, secs, nsecs);
1830 fileTimeToUtc(&bhfi->ftLastWriteTime, &secs, &nsecs);
1831 archive_entry_set_mtime(entry, secs, nsecs);
1832 fileTimeToUtc(&bhfi->ftCreationTime, &secs, &nsecs);
1833 archive_entry_set_birthtime(entry, secs, nsecs);
1834 archive_entry_set_ctime(entry, secs, nsecs);
1835 archive_entry_set_dev(entry, bhfi_dev(bhfi));
1836 archive_entry_set_ino64(entry, bhfi_ino(bhfi));
1837 if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
1838 archive_entry_set_nlink(entry, bhfi->nNumberOfLinks + 1);
1840 archive_entry_set_nlink(entry, bhfi->nNumberOfLinks);
1841 archive_entry_set_size(entry,
1842 (((int64_t)bhfi->nFileSizeHigh) << 32)
1843 + bhfi->nFileSizeLow);
1844 archive_entry_set_uid(entry, 0);
1845 archive_entry_set_gid(entry, 0);
1846 archive_entry_set_rdev(entry, 0);
1848 mode = S_IRUSR | S_IRGRP | S_IROTH;
1849 if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_READONLY) == 0)
1850 mode |= S_IWUSR | S_IWGRP | S_IWOTH;
1851 if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) &&
1853 findData->dwReserved0 == IO_REPARSE_TAG_SYMLINK)
1855 else if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
1856 mode |= S_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH;
1861 p = wcsrchr(path, L'.');
1862 if (p != NULL && wcslen(p) == 4) {
1864 case L'B': case L'b':
1865 if ((p[2] == L'A' || p[2] == L'a' ) &&
1866 (p[3] == L'T' || p[3] == L't' ))
1867 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1869 case L'C': case L'c':
1870 if (((p[2] == L'M' || p[2] == L'm' ) &&
1871 (p[3] == L'D' || p[3] == L'd' )))
1872 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1874 case L'E': case L'e':
1875 if ((p[2] == L'X' || p[2] == L'x' ) &&
1876 (p[3] == L'E' || p[3] == L'e' ))
1877 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1884 archive_entry_set_mode(entry, mode);
1888 tree_archive_entry_copy_bhfi(struct archive_entry *entry, struct tree *t,
1889 const BY_HANDLE_FILE_INFORMATION *bhfi)
1891 entry_copy_bhfi(entry, tree_current_path(t), t->findData, bhfi);
1895 tree_current_file_information(struct tree *t, BY_HANDLE_FILE_INFORMATION *st,
1900 DWORD flag = FILE_FLAG_BACKUP_SEMANTICS;
1902 if (sim_lstat && tree_current_is_physical_link(t))
1903 flag |= FILE_FLAG_OPEN_REPARSE_POINT;
1904 h = CreateFileW(tree_current_access_path(t), 0, FILE_SHARE_READ, NULL,
1905 OPEN_EXISTING, flag, NULL);
1906 if (h == INVALID_HANDLE_VALUE) {
1907 la_dosmaperr(GetLastError());
1908 t->tree_errno = errno;
1911 r = GetFileInformationByHandle(h, st);
1917 * Get the stat() data for the entry just returned from tree_next().
1919 static const BY_HANDLE_FILE_INFORMATION *
1920 tree_current_stat(struct tree *t)
1922 if (!(t->flags & hasStat)) {
1923 if (!tree_current_file_information(t, &t->st, 0))
1925 t->flags |= hasStat;
1931 * Get the lstat() data for the entry just returned from tree_next().
1933 static const BY_HANDLE_FILE_INFORMATION *
1934 tree_current_lstat(struct tree *t)
1936 if (!(t->flags & hasLstat)) {
1937 if (!tree_current_file_information(t, &t->lst, 1))
1939 t->flags |= hasLstat;
1945 * Test whether current entry is a dir or link to a dir.
1948 tree_current_is_dir(struct tree *t)
1951 return (t->findData->dwFileAttributes
1952 & FILE_ATTRIBUTE_DIRECTORY);
1957 * Test whether current entry is a physical directory. Usually, we
1958 * already have at least one of stat() or lstat() in memory, so we
1959 * use tricks to try to avoid an extra trip to the disk.
1962 tree_current_is_physical_dir(struct tree *t)
1964 if (tree_current_is_physical_link(t))
1966 return (tree_current_is_dir(t));
1970 * Test whether current entry is a symbolic link.
1973 tree_current_is_physical_link(struct tree *t)
1976 return ((t->findData->dwFileAttributes
1977 & FILE_ATTRIBUTE_REPARSE_POINT) &&
1978 (t->findData->dwReserved0
1979 == IO_REPARSE_TAG_SYMLINK));
1984 * Test whether the same file has been in the tree as its parent.
1987 tree_target_is_same_as_parent(struct tree *t,
1988 const BY_HANDLE_FILE_INFORMATION *st)
1990 struct tree_entry *te;
1991 int64_t dev = bhfi_dev(st);
1992 int64_t ino = bhfi_ino(st);
1994 for (te = t->current->parent; te != NULL; te = te->parent) {
1995 if (te->dev == dev && te->ino == ino)
2002 * Return the access path for the entry just returned from tree_next().
2004 static const wchar_t *
2005 tree_current_access_path(struct tree *t)
2007 return (t->full_path.s);
2011 * Return the full path for the entry just returned from tree_next().
2013 static const wchar_t *
2014 tree_current_path(struct tree *t)
2020 * Terminate the traversal.
2023 tree_close(struct tree *t)
2028 if (t->entry_fh != INVALID_HANDLE_VALUE) {
2030 close_and_restore_time(t->entry_fh, t, &t->restore_time);
2031 t->entry_fh = INVALID_HANDLE_VALUE;
2033 /* Close the handle of FindFirstFileW */
2034 if (t->d != INVALID_HANDLE_VALUE) {
2036 t->d = INVALID_HANDLE_VALUE;
2039 /* Release anything remaining in the stack. */
2040 while (t->stack != NULL)
2045 * Release any resources.
2048 tree_free(struct tree *t)
2054 archive_wstring_free(&t->path);
2055 archive_wstring_free(&t->full_path);
2056 free(t->sparse_list);
2057 free(t->filesystem_table);
2058 for (i = 0; i < MAX_OVERLAPPED; i++) {
2060 VirtualFree(t->ol[i].buff, 0, MEM_RELEASE);
2061 CloseHandle(t->ol[i].ol.hEvent);
2068 * Populate the archive_entry with metadata from the disk.
2071 archive_read_disk_entry_from_file(struct archive *_a,
2072 struct archive_entry *entry, int fd, const struct stat *st)
2074 struct archive_read_disk *a = (struct archive_read_disk *)_a;
2075 const wchar_t *path;
2076 const wchar_t *wname;
2079 BY_HANDLE_FILE_INFORMATION bhfi;
2080 DWORD fileAttributes = 0;
2083 archive_clear_error(_a);
2084 wname = archive_entry_sourcepath_w(entry);
2086 wname = archive_entry_pathname_w(entry);
2087 if (wname == NULL) {
2088 archive_set_error(&a->archive, EINVAL,
2089 "Can't get a wide character version of the path");
2090 return (ARCHIVE_FAILED);
2092 path = __la_win_permissive_name_w(wname);
2096 * Get metadata through GetFileInformationByHandle().
2099 h = (HANDLE)_get_osfhandle(fd);
2100 r = GetFileInformationByHandle(h, &bhfi);
2102 la_dosmaperr(GetLastError());
2103 archive_set_error(&a->archive, errno,
2104 "Can't GetFileInformationByHandle");
2105 return (ARCHIVE_FAILED);
2107 entry_copy_bhfi(entry, path, NULL, &bhfi);
2109 WIN32_FIND_DATAW findData;
2110 DWORD flag, desiredAccess;
2112 h = FindFirstFileW(path, &findData);
2113 if (h == INVALID_HANDLE_VALUE) {
2114 la_dosmaperr(GetLastError());
2115 archive_set_error(&a->archive, errno,
2116 "Can't FindFirstFileW");
2117 return (ARCHIVE_FAILED);
2121 flag = FILE_FLAG_BACKUP_SEMANTICS;
2122 if (!a->follow_symlinks &&
2123 (findData.dwFileAttributes
2124 & FILE_ATTRIBUTE_REPARSE_POINT) &&
2125 (findData.dwReserved0 == IO_REPARSE_TAG_SYMLINK)) {
2126 flag |= FILE_FLAG_OPEN_REPARSE_POINT;
2128 } else if (findData.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {
2131 desiredAccess = GENERIC_READ;
2133 h = CreateFileW(path, desiredAccess, FILE_SHARE_READ, NULL,
2134 OPEN_EXISTING, flag, NULL);
2135 if (h == INVALID_HANDLE_VALUE) {
2136 la_dosmaperr(GetLastError());
2137 archive_set_error(&a->archive, errno,
2138 "Can't CreateFileW");
2139 return (ARCHIVE_FAILED);
2141 r = GetFileInformationByHandle(h, &bhfi);
2143 la_dosmaperr(GetLastError());
2144 archive_set_error(&a->archive, errno,
2145 "Can't GetFileInformationByHandle");
2147 return (ARCHIVE_FAILED);
2149 entry_copy_bhfi(entry, path, &findData, &bhfi);
2151 fileAttributes = bhfi.dwFileAttributes;
2153 archive_entry_copy_stat(entry, st);
2154 h = INVALID_HANDLE_VALUE;
2157 /* Lookup uname/gname */
2158 name = archive_read_disk_uname(_a, archive_entry_uid(entry));
2160 archive_entry_copy_uname(entry, name);
2161 name = archive_read_disk_gname(_a, archive_entry_gid(entry));
2163 archive_entry_copy_gname(entry, name);
2166 * Can this file be sparse file ?
2168 if (archive_entry_filetype(entry) != AE_IFREG
2169 || archive_entry_size(entry) <= 0
2170 || archive_entry_hardlink(entry) != NULL) {
2171 if (h != INVALID_HANDLE_VALUE && fd < 0)
2173 return (ARCHIVE_OK);
2176 if (h == INVALID_HANDLE_VALUE) {
2178 h = (HANDLE)_get_osfhandle(fd);
2180 h = CreateFileW(path, GENERIC_READ, FILE_SHARE_READ, NULL,
2181 OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
2182 if (h == INVALID_HANDLE_VALUE) {
2183 la_dosmaperr(GetLastError());
2184 archive_set_error(&a->archive, errno,
2185 "Can't CreateFileW");
2186 return (ARCHIVE_FAILED);
2189 r = GetFileInformationByHandle(h, &bhfi);
2191 la_dosmaperr(GetLastError());
2192 archive_set_error(&a->archive, errno,
2193 "Can't GetFileInformationByHandle");
2194 if (h != INVALID_HANDLE_VALUE && fd < 0)
2196 return (ARCHIVE_FAILED);
2198 fileAttributes = bhfi.dwFileAttributes;
2201 /* Sparse file must be set a mark, FILE_ATTRIBUTE_SPARSE_FILE */
2202 if ((fileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) == 0) {
2205 return (ARCHIVE_OK);
2208 r = setup_sparse_from_disk(a, entry, h);
2216 * Windows sparse interface.
2218 #if defined(__MINGW32__) && !defined(FSCTL_QUERY_ALLOCATED_RANGES)
2219 #define FSCTL_QUERY_ALLOCATED_RANGES 0x940CF
2221 LARGE_INTEGER FileOffset;
2222 LARGE_INTEGER Length;
2223 } FILE_ALLOCATED_RANGE_BUFFER;
2227 setup_sparse_from_disk(struct archive_read_disk *a,
2228 struct archive_entry *entry, HANDLE handle)
2230 FILE_ALLOCATED_RANGE_BUFFER range, *outranges = NULL;
2231 size_t outranges_size;
2232 int64_t entry_size = archive_entry_size(entry);
2233 int exit_sts = ARCHIVE_OK;
2235 range.FileOffset.QuadPart = 0;
2236 range.Length.QuadPart = entry_size;
2237 outranges_size = 2048;
2238 outranges = (FILE_ALLOCATED_RANGE_BUFFER *)malloc(outranges_size);
2239 if (outranges == NULL) {
2240 archive_set_error(&a->archive, ENOMEM,
2241 "Couldn't allocate memory");
2242 exit_sts = ARCHIVE_FATAL;
2243 goto exit_setup_sparse;
2251 ret = DeviceIoControl(handle,
2252 FSCTL_QUERY_ALLOCATED_RANGES,
2253 &range, sizeof(range), outranges,
2254 (DWORD)outranges_size, &retbytes, NULL);
2255 if (ret == 0 && GetLastError() == ERROR_MORE_DATA) {
2257 outranges_size *= 2;
2258 outranges = (FILE_ALLOCATED_RANGE_BUFFER *)
2259 malloc(outranges_size);
2260 if (outranges == NULL) {
2261 archive_set_error(&a->archive, ENOMEM,
2262 "Couldn't allocate memory");
2263 exit_sts = ARCHIVE_FATAL;
2264 goto exit_setup_sparse;
2274 n = retbytes / sizeof(outranges[0]);
2276 outranges[0].FileOffset.QuadPart == 0 &&
2277 outranges[0].Length.QuadPart == entry_size)
2278 break;/* This is not sparse. */
2279 for (i = 0; i < n; i++)
2280 archive_entry_sparse_add_entry(entry,
2281 outranges[i].FileOffset.QuadPart,
2282 outranges[i].Length.QuadPart);
2283 range.FileOffset.QuadPart =
2284 outranges[n-1].FileOffset.QuadPart
2285 + outranges[n-1].Length.QuadPart;
2286 range.Length.QuadPart =
2287 entry_size - range.FileOffset.QuadPart;
2288 if (range.Length.QuadPart > 0)
2291 /* The remaining data is hole. */
2292 archive_entry_sparse_add_entry(entry,
2293 range.FileOffset.QuadPart,
2294 range.Length.QuadPart);
2298 la_dosmaperr(GetLastError());
2299 archive_set_error(&a->archive, errno,
2300 "DeviceIoControl Failed: %lu", GetLastError());
2301 exit_sts = ARCHIVE_FAILED;
2302 goto exit_setup_sparse;