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_filesystem;
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 *)calloc(1, sizeof(*a));
395 a->archive.magic = ARCHIVE_READ_DISK_MAGIC;
396 a->archive.state = ARCHIVE_STATE_NEW;
397 a->archive.vtable = archive_read_disk_vtable();
398 a->entry = archive_entry_new2(&a->archive);
399 a->lookup_uname = trivial_lookup_uname;
400 a->lookup_gname = trivial_lookup_gname;
401 a->enable_copyfile = 1;
402 a->traverse_mount_points = 1;
403 return (&a->archive);
407 _archive_read_free(struct archive *_a)
409 struct archive_read_disk *a = (struct archive_read_disk *)_a;
414 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
415 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
417 if (a->archive.state != ARCHIVE_STATE_CLOSED)
418 r = _archive_read_close(&a->archive);
423 if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL)
424 (a->cleanup_gname)(a->lookup_gname_data);
425 if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL)
426 (a->cleanup_uname)(a->lookup_uname_data);
427 archive_string_free(&a->archive.error_string);
428 archive_entry_free(a->entry);
429 a->archive.magic = 0;
435 _archive_read_close(struct archive *_a)
437 struct archive_read_disk *a = (struct archive_read_disk *)_a;
439 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
440 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
442 if (a->archive.state != ARCHIVE_STATE_FATAL)
443 a->archive.state = ARCHIVE_STATE_CLOSED;
451 setup_symlink_mode(struct archive_read_disk *a, char symlink_mode,
454 a->symlink_mode = symlink_mode;
455 a->follow_symlinks = follow_symlinks;
456 if (a->tree != NULL) {
457 a->tree->initial_symlink_mode = a->symlink_mode;
458 a->tree->symlink_mode = a->symlink_mode;
463 archive_read_disk_set_symlink_logical(struct archive *_a)
465 struct archive_read_disk *a = (struct archive_read_disk *)_a;
466 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
467 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical");
468 setup_symlink_mode(a, 'L', 1);
473 archive_read_disk_set_symlink_physical(struct archive *_a)
475 struct archive_read_disk *a = (struct archive_read_disk *)_a;
476 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
477 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical");
478 setup_symlink_mode(a, 'P', 0);
483 archive_read_disk_set_symlink_hybrid(struct archive *_a)
485 struct archive_read_disk *a = (struct archive_read_disk *)_a;
486 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
487 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid");
488 setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */
493 archive_read_disk_set_atime_restored(struct archive *_a)
495 struct archive_read_disk *a = (struct archive_read_disk *)_a;
496 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
497 ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime");
500 a->tree->flags |= needsRestoreTimes;
505 archive_read_disk_set_behavior(struct archive *_a, int flags)
507 struct archive_read_disk *a = (struct archive_read_disk *)_a;
510 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
511 ARCHIVE_STATE_ANY, "archive_read_disk_honor_nodump");
513 if (flags & ARCHIVE_READDISK_RESTORE_ATIME)
514 r = archive_read_disk_set_atime_restored(_a);
518 a->tree->flags &= ~needsRestoreTimes;
520 if (flags & ARCHIVE_READDISK_HONOR_NODUMP)
524 if (flags & ARCHIVE_READDISK_MAC_COPYFILE)
525 a->enable_copyfile = 1;
527 a->enable_copyfile = 0;
528 if (flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS)
529 a->traverse_mount_points = 0;
531 a->traverse_mount_points = 1;
536 * Trivial implementations of gname/uname lookup functions.
537 * These are normally overridden by the client, but these stub
538 * versions ensure that we always have something that works.
541 trivial_lookup_gname(void *private_data, int64_t gid)
543 (void)private_data; /* UNUSED */
544 (void)gid; /* UNUSED */
549 trivial_lookup_uname(void *private_data, int64_t uid)
551 (void)private_data; /* UNUSED */
552 (void)uid; /* UNUSED */
557 align_num_per_sector(struct tree *t, int64_t size)
561 size += t->current_filesystem->bytesPerSector -1;
562 surplus = size % t->current_filesystem->bytesPerSector;
568 start_next_async_read(struct archive_read_disk *a, struct tree *t)
570 struct la_overlapped *olp;
571 DWORD buffbytes, rbytes;
573 if (t->ol_remaining_bytes == 0)
574 return (ARCHIVE_EOF);
576 olp = &(t->ol[t->ol_idx_doing]);
577 t->ol_idx_doing = (t->ol_idx_doing + 1) % MAX_OVERLAPPED;
579 /* Allocate read buffer. */
580 if (olp->buff == NULL) {
582 size_t s = (size_t)align_num_per_sector(t, BUFFER_SIZE);
583 p = VirtualAlloc(NULL, s, MEM_COMMIT, PAGE_READWRITE);
585 archive_set_error(&a->archive, ENOMEM,
586 "Couldn't allocate memory");
587 a->archive.state = ARCHIVE_STATE_FATAL;
588 return (ARCHIVE_FATAL);
592 olp->_a = &a->archive;
593 olp->ol.hEvent = CreateEventW(NULL, TRUE, FALSE, NULL);
594 if (olp->ol.hEvent == NULL) {
595 la_dosmaperr(GetLastError());
596 archive_set_error(&a->archive, errno,
597 "CreateEvent failed");
598 a->archive.state = ARCHIVE_STATE_FATAL;
599 return (ARCHIVE_FATAL);
602 ResetEvent(olp->ol.hEvent);
604 buffbytes = (DWORD)olp->buff_size;
605 if (buffbytes > t->current_sparse->length)
606 buffbytes = (DWORD)t->current_sparse->length;
609 if (t->current_sparse->offset > t->ol_total) {
610 t->ol_remaining_bytes -=
611 t->current_sparse->offset - t->ol_total;
614 olp->offset = t->current_sparse->offset;
615 olp->ol.Offset = (DWORD)(olp->offset & 0xffffffff);
616 olp->ol.OffsetHigh = (DWORD)(olp->offset >> 32);
618 if (t->ol_remaining_bytes > buffbytes) {
619 olp->bytes_expected = buffbytes;
620 t->ol_remaining_bytes -= buffbytes;
622 olp->bytes_expected = (size_t)t->ol_remaining_bytes;
623 t->ol_remaining_bytes = 0;
625 olp->bytes_transferred = 0;
626 t->current_sparse->offset += buffbytes;
627 t->current_sparse->length -= buffbytes;
628 t->ol_total = t->current_sparse->offset;
629 if (t->current_sparse->length == 0 && t->ol_remaining_bytes > 0)
632 if (!ReadFile(t->entry_fh, olp->buff, buffbytes, &rbytes, &(olp->ol))) {
635 lasterr = GetLastError();
636 if (lasterr == ERROR_HANDLE_EOF) {
637 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
638 "Reading file truncated");
639 a->archive.state = ARCHIVE_STATE_FATAL;
640 return (ARCHIVE_FATAL);
641 } else if (lasterr != ERROR_IO_PENDING) {
642 if (lasterr == ERROR_NO_DATA)
644 else if (lasterr == ERROR_ACCESS_DENIED)
647 la_dosmaperr(lasterr);
648 archive_set_error(&a->archive, errno, "Read error");
649 a->archive.state = ARCHIVE_STATE_FATAL;
650 return (ARCHIVE_FATAL);
653 olp->bytes_transferred = rbytes;
656 return (t->ol_remaining_bytes == 0)? ARCHIVE_EOF: ARCHIVE_OK;
660 cancel_async(struct tree *t)
662 if (t->ol_num_doing != t->ol_num_done) {
663 CancelIo(t->entry_fh);
664 t->ol_num_doing = t->ol_num_done = 0;
669 _archive_read_data_block(struct archive *_a, const void **buff,
670 size_t *size, int64_t *offset)
672 struct archive_read_disk *a = (struct archive_read_disk *)_a;
673 struct tree *t = a->tree;
674 struct la_overlapped *olp;
675 DWORD bytes_transferred;
676 int r = ARCHIVE_FATAL;
678 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
679 "archive_read_data_block");
681 if (t->entry_eof || t->entry_remaining_bytes <= 0) {
683 goto abort_read_data;
687 * Make a request to read the file in asynchronous.
689 if (t->ol_num_doing == 0) {
691 r = start_next_async_read(a, t);
692 if (r == ARCHIVE_FATAL)
693 goto abort_read_data;
696 } while (r == ARCHIVE_OK && t->ol_num_doing < MAX_OVERLAPPED);
698 if (start_next_async_read(a, t) == ARCHIVE_FATAL)
699 goto abort_read_data;
702 olp = &(t->ol[t->ol_idx_done]);
703 t->ol_idx_done = (t->ol_idx_done + 1) % MAX_OVERLAPPED;
704 if (olp->bytes_transferred)
705 bytes_transferred = (DWORD)olp->bytes_transferred;
706 else if (!GetOverlappedResult(t->entry_fh, &(olp->ol),
707 &bytes_transferred, TRUE)) {
708 la_dosmaperr(GetLastError());
709 archive_set_error(&a->archive, errno,
710 "GetOverlappedResult failed");
711 a->archive.state = ARCHIVE_STATE_FATAL;
713 goto abort_read_data;
717 if (bytes_transferred == 0 ||
718 olp->bytes_expected != bytes_transferred) {
719 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
720 "Reading file truncated");
721 a->archive.state = ARCHIVE_STATE_FATAL;
723 goto abort_read_data;
727 *size = bytes_transferred;
728 *offset = olp->offset;
729 if (olp->offset > t->entry_total)
730 t->entry_remaining_bytes -= olp->offset - t->entry_total;
731 t->entry_total = olp->offset + *size;
732 t->entry_remaining_bytes -= *size;
733 if (t->entry_remaining_bytes == 0) {
734 /* Close the current file descriptor */
735 close_and_restore_time(t->entry_fh, t, &t->restore_time);
736 t->entry_fh = INVALID_HANDLE_VALUE;
744 *offset = t->entry_total;
745 if (t->entry_fh != INVALID_HANDLE_VALUE) {
747 /* Close the current file descriptor */
748 close_and_restore_time(t->entry_fh, t, &t->restore_time);
749 t->entry_fh = INVALID_HANDLE_VALUE;
755 next_entry(struct archive_read_disk *a, struct tree *t,
756 struct archive_entry *entry)
758 const BY_HANDLE_FILE_INFORMATION *st;
759 const BY_HANDLE_FILE_INFORMATION *lst;
767 switch (tree_next(t)) {
768 case TREE_ERROR_FATAL:
769 archive_set_error(&a->archive, t->tree_errno,
770 "%ls: Unable to continue traversing directory tree",
771 tree_current_path(t));
772 a->archive.state = ARCHIVE_STATE_FATAL;
773 return (ARCHIVE_FATAL);
775 archive_set_error(&a->archive, t->tree_errno,
776 "%ls: Couldn't visit directory",
777 tree_current_path(t));
778 return (ARCHIVE_FAILED);
780 return (ARCHIVE_EOF);
781 case TREE_POSTDESCENT:
782 case TREE_POSTASCENT:
785 lst = tree_current_lstat(t);
787 archive_set_error(&a->archive, t->tree_errno,
789 tree_current_path(t));
790 return (ARCHIVE_FAILED);
794 } while (lst == NULL);
796 archive_entry_copy_pathname_w(entry, tree_current_path(t));
799 * Perform path matching.
802 r = archive_match_path_excluded(a->matching, entry);
804 archive_set_error(&(a->archive), errno,
805 "Failed : %s", archive_error_string(a->matching));
809 if (a->excluded_cb_func)
810 a->excluded_cb_func(&(a->archive),
811 a->excluded_cb_data, entry);
812 return (ARCHIVE_RETRY);
817 * Distinguish 'L'/'P'/'H' symlink following.
819 switch(t->symlink_mode) {
821 /* 'H': After the first item, rest like 'P'. */
822 t->symlink_mode = 'P';
823 /* 'H': First item (from command line) like 'L'. */
826 /* 'L': Do descend through a symlink to dir. */
827 descend = tree_current_is_dir(t);
828 /* 'L': Follow symlinks to files. */
829 a->symlink_mode = 'L';
830 a->follow_symlinks = 1;
831 /* 'L': Archive symlinks as targets, if we can. */
832 st = tree_current_stat(t);
833 if (st != NULL && !tree_target_is_same_as_parent(t, st))
835 /* If stat fails, we have a broken symlink;
836 * in that case, don't follow the link. */
839 /* 'P': Don't descend through a symlink to dir. */
840 descend = tree_current_is_physical_dir(t);
841 /* 'P': Don't follow symlinks to files. */
842 a->symlink_mode = 'P';
843 a->follow_symlinks = 0;
844 /* 'P': Archive symlinks as symlinks. */
849 if (update_current_filesystem(a, bhfi_dev(st)) != ARCHIVE_OK) {
850 a->archive.state = ARCHIVE_STATE_FATAL;
851 return (ARCHIVE_FATAL);
853 if (t->initial_filesystem_id == -1)
854 t->initial_filesystem_id = t->current_filesystem_id;
855 if (!a->traverse_mount_points) {
856 if (t->initial_filesystem_id != t->current_filesystem_id)
857 return (ARCHIVE_RETRY);
859 t->descend = descend;
861 tree_archive_entry_copy_bhfi(entry, t, st);
863 /* Save the times to be restored. This must be in before
864 * calling archive_read_disk_descend() or any chance of it,
865 * especially, invoking a callback. */
866 t->restore_time.lastWriteTime = st->ftLastWriteTime;
867 t->restore_time.lastAccessTime = st->ftLastAccessTime;
868 t->restore_time.filetype = archive_entry_filetype(entry);
871 * Perform time matching.
874 r = archive_match_time_excluded(a->matching, entry);
876 archive_set_error(&(a->archive), errno,
877 "Failed : %s", archive_error_string(a->matching));
881 if (a->excluded_cb_func)
882 a->excluded_cb_func(&(a->archive),
883 a->excluded_cb_data, entry);
884 return (ARCHIVE_RETRY);
888 /* Lookup uname/gname */
889 name = archive_read_disk_uname(&(a->archive), archive_entry_uid(entry));
891 archive_entry_copy_uname(entry, name);
892 name = archive_read_disk_gname(&(a->archive), archive_entry_gid(entry));
894 archive_entry_copy_gname(entry, name);
897 * Perform owner matching.
900 r = archive_match_owner_excluded(a->matching, entry);
902 archive_set_error(&(a->archive), errno,
903 "Failed : %s", archive_error_string(a->matching));
907 if (a->excluded_cb_func)
908 a->excluded_cb_func(&(a->archive),
909 a->excluded_cb_data, entry);
910 return (ARCHIVE_RETRY);
915 * Invoke a meta data filter callback.
917 if (a->metadata_filter_func) {
918 if (!a->metadata_filter_func(&(a->archive),
919 a->metadata_filter_data, entry))
920 return (ARCHIVE_RETRY);
923 archive_entry_copy_sourcepath_w(entry, tree_current_access_path(t));
926 if (archive_entry_filetype(entry) == AE_IFREG &&
927 archive_entry_size(entry) > 0) {
928 DWORD flags = FILE_FLAG_BACKUP_SEMANTICS;
930 flags |= FILE_FLAG_OVERLAPPED;
932 flags |= FILE_FLAG_NO_BUFFERING;
934 flags |= FILE_FLAG_SEQUENTIAL_SCAN;
935 t->entry_fh = CreateFileW(tree_current_access_path(t),
936 GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, flags, NULL);
937 if (t->entry_fh == INVALID_HANDLE_VALUE) {
938 archive_set_error(&a->archive, errno,
939 "Couldn't open %ls", tree_current_path(a->tree));
940 return (ARCHIVE_FAILED);
943 /* Find sparse data from the disk. */
944 if (archive_entry_hardlink(entry) == NULL &&
945 (st->dwFileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) != 0)
946 r = setup_sparse_from_disk(a, entry, t->entry_fh);
952 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
955 struct archive_read_disk *a = (struct archive_read_disk *)_a;
957 ret = _archive_read_next_header2(_a, a->entry);
963 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
965 struct archive_read_disk *a = (struct archive_read_disk *)_a;
969 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
970 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
971 "archive_read_next_header2");
974 if (t->entry_fh != INVALID_HANDLE_VALUE) {
976 close_and_restore_time(t->entry_fh, t, &t->restore_time);
977 t->entry_fh = INVALID_HANDLE_VALUE;
980 while ((r = next_entry(a, t, entry)) == ARCHIVE_RETRY)
981 archive_entry_clear(entry);
984 * EOF and FATAL are persistent at this layer. By
985 * modifying the state, we guarantee that future calls to
986 * read a header or read data will fail.
990 a->archive.state = ARCHIVE_STATE_EOF;
995 if (archive_entry_filetype(entry) == AE_IFREG) {
996 t->entry_remaining_bytes = archive_entry_size(entry);
997 t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0;
999 setup_sparse(a, entry) != ARCHIVE_OK)
1000 return (ARCHIVE_FATAL);
1002 t->entry_remaining_bytes = 0;
1005 t->ol_idx_doing = t->ol_idx_done = 0;
1006 t->ol_num_doing = t->ol_num_done = 0;
1007 t->ol_remaining_bytes = t->entry_remaining_bytes;
1009 a->archive.state = ARCHIVE_STATE_DATA;
1014 a->archive.state = ARCHIVE_STATE_FATAL;
1018 __archive_reset_read_data(&a->archive);
1023 setup_sparse(struct archive_read_disk *a, struct archive_entry *entry)
1025 struct tree *t = a->tree;
1026 int64_t aligned, length, offset;
1029 t->sparse_count = archive_entry_sparse_reset(entry);
1030 if (t->sparse_count+1 > t->sparse_list_size) {
1031 free(t->sparse_list);
1032 t->sparse_list_size = t->sparse_count + 1;
1033 t->sparse_list = malloc(sizeof(t->sparse_list[0]) *
1034 t->sparse_list_size);
1035 if (t->sparse_list == NULL) {
1036 t->sparse_list_size = 0;
1037 archive_set_error(&a->archive, ENOMEM,
1038 "Can't allocate data");
1039 a->archive.state = ARCHIVE_STATE_FATAL;
1040 return (ARCHIVE_FATAL);
1044 * Get sparse list and make sure those offsets and lengths are
1045 * aligned by a sector size.
1047 for (i = 0; i < t->sparse_count; i++) {
1048 archive_entry_sparse_next(entry, &offset, &length);
1049 aligned = align_num_per_sector(t, offset);
1050 if (aligned != offset) {
1051 aligned -= t->current_filesystem->bytesPerSector;
1052 length += offset - aligned;
1054 t->sparse_list[i].offset = aligned;
1055 aligned = align_num_per_sector(t, length);
1056 t->sparse_list[i].length = aligned;
1059 aligned = align_num_per_sector(t, archive_entry_size(entry));
1061 t->sparse_list[i].offset = 0;
1062 t->sparse_list[i].length = aligned;
1066 t->sparse_list[i].offset = aligned;
1067 t->sparse_list[i].length = 0;
1068 for (i = 0; i < last; i++) {
1069 if ((t->sparse_list[i].offset +
1070 t->sparse_list[i].length) <=
1071 t->sparse_list[i+1].offset)
1074 * Now sparse_list[i+1] is overlapped by sparse_list[i].
1077 length = t->sparse_list[i+1].offset -
1078 t->sparse_list[i].offset;
1079 t->sparse_list[i+1].offset = t->sparse_list[i].offset;
1080 t->sparse_list[i+1].length += length;
1081 /* Remove sparse_list[i]. */
1082 for (j = i; j < last; j++) {
1083 t->sparse_list[j].offset =
1084 t->sparse_list[j+1].offset;
1085 t->sparse_list[j].length =
1086 t->sparse_list[j+1].length;
1091 t->current_sparse = t->sparse_list;
1093 return (ARCHIVE_OK);
1097 archive_read_disk_set_matching(struct archive *_a, struct archive *_ma,
1098 void (*_excluded_func)(struct archive *, void *, struct archive_entry *),
1101 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1102 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1103 ARCHIVE_STATE_ANY, "archive_read_disk_set_matching");
1105 a->excluded_cb_func = _excluded_func;
1106 a->excluded_cb_data = _client_data;
1107 return (ARCHIVE_OK);
1111 archive_read_disk_set_metadata_filter_callback(struct archive *_a,
1112 int (*_metadata_filter_func)(struct archive *, void *,
1113 struct archive_entry *), void *_client_data)
1115 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1117 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY,
1118 "archive_read_disk_set_metadata_filter_callback");
1120 a->metadata_filter_func = _metadata_filter_func;
1121 a->metadata_filter_data = _client_data;
1122 return (ARCHIVE_OK);
1126 archive_read_disk_can_descend(struct archive *_a)
1128 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1129 struct tree *t = a->tree;
1131 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1132 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1133 "archive_read_disk_can_descend");
1135 return (t->visit_type == TREE_REGULAR && t->descend);
1139 * Called by the client to mark the directory just returned from
1140 * tree_next() as needing to be visited.
1143 archive_read_disk_descend(struct archive *_a)
1145 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1146 struct tree *t = a->tree;
1148 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1149 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1150 "archive_read_disk_descend");
1152 if (t->visit_type != TREE_REGULAR || !t->descend)
1153 return (ARCHIVE_OK);
1155 if (tree_current_is_physical_dir(t)) {
1156 tree_push(t, t->basename, t->full_path.s,
1157 t->current_filesystem_id,
1158 bhfi_dev(&(t->lst)), bhfi_ino(&(t->lst)),
1160 t->stack->flags |= isDir;
1161 } else if (tree_current_is_dir(t)) {
1162 tree_push(t, t->basename, t->full_path.s,
1163 t->current_filesystem_id,
1164 bhfi_dev(&(t->st)), bhfi_ino(&(t->st)),
1166 t->stack->flags |= isDirLink;
1169 return (ARCHIVE_OK);
1173 archive_read_disk_open(struct archive *_a, const char *pathname)
1175 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1176 struct archive_wstring wpath;
1179 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1180 ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1181 "archive_read_disk_open");
1182 archive_clear_error(&a->archive);
1184 /* Make a wchar_t string from a char string. */
1185 archive_string_init(&wpath);
1186 if (archive_wstring_append_from_mbs(&wpath, pathname,
1187 strlen(pathname)) != 0) {
1188 if (errno == ENOMEM)
1189 archive_set_error(&a->archive, ENOMEM,
1190 "Can't allocate memory");
1192 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1193 "Can't convert a path to a wchar_t string");
1194 a->archive.state = ARCHIVE_STATE_FATAL;
1195 ret = ARCHIVE_FATAL;
1197 ret = _archive_read_disk_open_w(_a, wpath.s);
1199 archive_wstring_free(&wpath);
1204 archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1206 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1208 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1209 ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1210 "archive_read_disk_open_w");
1211 archive_clear_error(&a->archive);
1213 return (_archive_read_disk_open_w(_a, pathname));
1217 _archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1219 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1221 if (a->tree != NULL)
1222 a->tree = tree_reopen(a->tree, pathname, a->restore_time);
1224 a->tree = tree_open(pathname, a->symlink_mode, a->restore_time);
1225 if (a->tree == NULL) {
1226 archive_set_error(&a->archive, ENOMEM,
1227 "Can't allocate directory traversal data");
1228 a->archive.state = ARCHIVE_STATE_FATAL;
1229 return (ARCHIVE_FATAL);
1231 a->archive.state = ARCHIVE_STATE_HEADER;
1233 return (ARCHIVE_OK);
1237 * Return a current filesystem ID which is index of the filesystem entry
1238 * you've visited through archive_read_disk.
1241 archive_read_disk_current_filesystem(struct archive *_a)
1243 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1245 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1246 "archive_read_disk_current_filesystem");
1248 return (a->tree->current_filesystem_id);
1252 update_current_filesystem(struct archive_read_disk *a, int64_t dev)
1254 struct tree *t = a->tree;
1257 if (t->current_filesystem != NULL &&
1258 t->current_filesystem->dev == dev)
1259 return (ARCHIVE_OK);
1261 for (i = 0; i < t->max_filesystem_id; i++) {
1262 if (t->filesystem_table[i].dev == dev) {
1263 /* There is the filesystem ID we've already generated. */
1264 t->current_filesystem_id = i;
1265 t->current_filesystem = &(t->filesystem_table[i]);
1266 return (ARCHIVE_OK);
1271 * There is a new filesystem, we generate a new ID for.
1273 fid = t->max_filesystem_id++;
1274 if (t->max_filesystem_id > t->allocated_filesystem) {
1278 s = t->max_filesystem_id * 2;
1279 p = realloc(t->filesystem_table,
1280 s * sizeof(*t->filesystem_table));
1282 archive_set_error(&a->archive, ENOMEM,
1283 "Can't allocate tar data");
1284 return (ARCHIVE_FATAL);
1286 t->filesystem_table = (struct filesystem *)p;
1287 t->allocated_filesystem = (int)s;
1289 t->current_filesystem_id = fid;
1290 t->current_filesystem = &(t->filesystem_table[fid]);
1291 t->current_filesystem->dev = dev;
1293 return (setup_current_filesystem(a));
1297 * Returns 1 if current filesystem is generated filesystem, 0 if it is not
1298 * or -1 if it is unknown.
1301 archive_read_disk_current_filesystem_is_synthetic(struct archive *_a)
1303 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1305 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1306 "archive_read_disk_current_filesystem");
1308 return (a->tree->current_filesystem->synthetic);
1312 * Returns 1 if current filesystem is remote filesystem, 0 if it is not
1313 * or -1 if it is unknown.
1316 archive_read_disk_current_filesystem_is_remote(struct archive *_a)
1318 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1320 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1321 "archive_read_disk_current_filesystem");
1323 return (a->tree->current_filesystem->remote);
1327 * If symlink is broken, statfs or statvfs will fail.
1328 * Use its directory path instead.
1331 safe_path_for_statfs(struct tree *t)
1333 const wchar_t *path;
1334 wchar_t *cp, *p = NULL;
1336 path = tree_current_access_path(t);
1337 if (tree_current_stat(t) == NULL) {
1339 cp = wcsrchr(p, '/');
1340 if (cp != NULL && wcslen(cp) >= 2) {
1351 * Get conditions of synthetic and remote on Windows
1354 setup_current_filesystem(struct archive_read_disk *a)
1356 struct tree *t = a->tree;
1360 t->current_filesystem->synthetic = -1;/* Not supported */
1361 path = safe_path_for_statfs(t);
1362 if (!GetVolumePathNameW(path, vol, sizeof(vol)/sizeof(vol[0]))) {
1364 t->current_filesystem->remote = -1;
1365 t->current_filesystem->bytesPerSector = 0;
1366 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1367 "GetVolumePathName failed: %d", (int)GetLastError());
1368 return (ARCHIVE_FAILED);
1371 switch (GetDriveTypeW(vol)) {
1373 case DRIVE_NO_ROOT_DIR:
1374 t->current_filesystem->remote = -1;
1377 t->current_filesystem->remote = 1;
1380 t->current_filesystem->remote = 0;
1384 if (!GetDiskFreeSpaceW(vol, NULL,
1385 &(t->current_filesystem->bytesPerSector), NULL, NULL)) {
1386 t->current_filesystem->bytesPerSector = 0;
1387 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1388 "GetDiskFreeSpace failed: %d", (int)GetLastError());
1389 return (ARCHIVE_FAILED);
1392 return (ARCHIVE_OK);
1396 close_and_restore_time(HANDLE h, struct tree *t, struct restore_time *rt)
1401 if (h == INVALID_HANDLE_VALUE && AE_IFLNK == rt->filetype)
1404 /* Close a file descriptor.
1405 * It will not be used for SetFileTime() because it has been opened
1406 * by a read only mode.
1408 if (h != INVALID_HANDLE_VALUE)
1410 if ((t->flags & needsRestoreTimes) == 0)
1413 handle = CreateFileW(rt->full_path, FILE_WRITE_ATTRIBUTES,
1414 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
1415 if (handle == INVALID_HANDLE_VALUE) {
1420 if (SetFileTime(handle, NULL, &rt->lastAccessTime,
1421 &rt->lastWriteTime) == 0) {
1426 CloseHandle(handle);
1431 * Add a directory path to the current stack.
1434 tree_push(struct tree *t, const wchar_t *path, const wchar_t *full_path,
1435 int filesystem_id, int64_t dev, int64_t ino, struct restore_time *rt)
1437 struct tree_entry *te;
1439 te = calloc(1, sizeof(*te));
1440 te->next = t->stack;
1441 te->parent = t->current;
1443 te->depth = te->parent->depth + 1;
1445 archive_string_init(&te->name);
1446 archive_wstrcpy(&te->name, path);
1447 archive_string_init(&te->full_path);
1448 archive_wstrcpy(&te->full_path, full_path);
1449 te->flags = needsDescent | needsOpen | needsAscent;
1450 te->filesystem_id = filesystem_id;
1453 te->dirname_length = t->dirname_length;
1454 te->full_path_dir_length = t->full_path_dir_length;
1455 te->restore_time.full_path = te->full_path.s;
1457 te->restore_time.lastWriteTime = rt->lastWriteTime;
1458 te->restore_time.lastAccessTime = rt->lastAccessTime;
1459 te->restore_time.filetype = rt->filetype;
1464 * Append a name to the current dir path.
1467 tree_append(struct tree *t, const wchar_t *name, size_t name_length)
1471 t->path.s[t->dirname_length] = L'\0';
1472 t->path.length = t->dirname_length;
1473 /* Strip trailing '/' from name, unless entire name is "/". */
1474 while (name_length > 1 && name[name_length - 1] == L'/')
1477 /* Resize pathname buffer as needed. */
1478 size_needed = name_length + t->dirname_length + 2;
1479 archive_wstring_ensure(&t->path, size_needed);
1480 /* Add a separating '/' if it's needed. */
1481 if (t->dirname_length > 0 &&
1482 t->path.s[archive_strlen(&t->path)-1] != L'/')
1483 archive_wstrappend_wchar(&t->path, L'/');
1484 t->basename = t->path.s + archive_strlen(&t->path);
1485 archive_wstrncat(&t->path, name, name_length);
1486 t->restore_time.full_path = t->basename;
1487 if (t->full_path_dir_length > 0) {
1488 t->full_path.s[t->full_path_dir_length] = L'\0';
1489 t->full_path.length = t->full_path_dir_length;
1490 size_needed = name_length + t->full_path_dir_length + 2;
1491 archive_wstring_ensure(&t->full_path, size_needed);
1492 /* Add a separating '\' if it's needed. */
1493 if (t->full_path.s[archive_strlen(&t->full_path)-1] != L'\\')
1494 archive_wstrappend_wchar(&t->full_path, L'\\');
1495 archive_wstrncat(&t->full_path, name, name_length);
1496 t->restore_time.full_path = t->full_path.s;
1501 * Open a directory tree for traversal.
1503 static struct tree *
1504 tree_open(const wchar_t *path, int symlink_mode, int restore_time)
1508 t = calloc(1, sizeof(*t));
1509 archive_string_init(&(t->full_path));
1510 archive_string_init(&t->path);
1511 archive_wstring_ensure(&t->path, 15);
1512 t->initial_symlink_mode = symlink_mode;
1513 return (tree_reopen(t, path, restore_time));
1516 static struct tree *
1517 tree_reopen(struct tree *t, const wchar_t *path, int restore_time)
1519 struct archive_wstring ws;
1520 wchar_t *pathname, *p, *base;
1522 t->flags = (restore_time)?needsRestoreTimes:0;
1525 t->full_path_dir_length = 0;
1526 t->dirname_length = 0;
1530 t->d = INVALID_HANDLE_VALUE;
1531 t->symlink_mode = t->initial_symlink_mode;
1532 archive_string_empty(&(t->full_path));
1533 archive_string_empty(&t->path);
1534 t->entry_fh = INVALID_HANDLE_VALUE;
1536 t->entry_remaining_bytes = 0;
1537 t->initial_filesystem_id = -1;
1539 /* Get wchar_t strings from char strings. */
1540 archive_string_init(&ws);
1541 archive_wstrcpy(&ws, path);
1543 /* Get a full-path-name. */
1544 p = __la_win_permissive_name_w(pathname);
1547 archive_wstrcpy(&(t->full_path), p);
1550 /* Convert path separators from '\' to '/' */
1551 for (p = pathname; *p != L'\0'; ++p) {
1557 /* First item is set up a lot like a symlink traversal. */
1558 /* printf("Looking for wildcard in %s\n", path); */
1559 if ((base[0] == L'/' && base[1] == L'/' &&
1560 base[2] == L'?' && base[3] == L'/' &&
1561 (wcschr(base+4, L'*') || wcschr(base+4, L'?'))) ||
1562 (!(base[0] == L'/' && base[1] == L'/' &&
1563 base[2] == L'?' && base[3] == L'/') &&
1564 (wcschr(base, L'*') || wcschr(base, L'?')))) {
1565 // It has a wildcard in it...
1566 // Separate the last element.
1567 p = wcsrchr(base, L'/');
1570 tree_append(t, base, p - base);
1571 t->dirname_length = archive_strlen(&t->path);
1574 p = wcsrchr(t->full_path.s, L'\\');
1577 t->full_path.length = wcslen(t->full_path.s);
1578 t->full_path_dir_length = archive_strlen(&t->full_path);
1581 tree_push(t, base, t->full_path.s, 0, 0, 0, NULL);
1582 archive_wstring_free(&ws);
1583 t->stack->flags = needsFirstVisit;
1585 * Debug flag for Direct IO(No buffering) or Async IO.
1586 * Those dependent on environment variable switches
1587 * will be removed until next release.
1591 if ((e = getenv("LIBARCHIVE_DIRECT_IO")) != NULL) {
1596 fprintf(stderr, "LIBARCHIVE_DIRECT_IO=%s\n",
1597 (t->direct_io)?"Enabled":"Disabled");
1599 t->direct_io = DIRECT_IO;
1600 if ((e = getenv("LIBARCHIVE_ASYNC_IO")) != NULL) {
1605 fprintf(stderr, "LIBARCHIVE_ASYNC_IO=%s\n",
1606 (t->async_io)?"Enabled":"Disabled");
1608 t->async_io = ASYNC_IO;
1612 archive_wstring_free(&ws);
1618 tree_descent(struct tree *t)
1620 t->dirname_length = archive_strlen(&t->path);
1621 t->full_path_dir_length = archive_strlen(&t->full_path);
1627 * We've finished a directory; ascend back to the parent.
1630 tree_ascend(struct tree *t)
1632 struct tree_entry *te;
1636 close_and_restore_time(INVALID_HANDLE_VALUE, t, &te->restore_time);
1641 * Pop the working stack.
1644 tree_pop(struct tree *t)
1646 struct tree_entry *te;
1648 t->full_path.s[t->full_path_dir_length] = L'\0';
1649 t->full_path.length = t->full_path_dir_length;
1650 t->path.s[t->dirname_length] = L'\0';
1651 t->path.length = t->dirname_length;
1652 if (t->stack == t->current && t->current != NULL)
1653 t->current = t->current->parent;
1655 t->stack = te->next;
1656 t->dirname_length = te->dirname_length;
1657 t->basename = t->path.s + t->dirname_length;
1658 t->full_path_dir_length = te->full_path_dir_length;
1659 while (t->basename[0] == L'/')
1661 archive_wstring_free(&te->name);
1662 archive_wstring_free(&te->full_path);
1667 * Get the next item in the tree traversal.
1670 tree_next(struct tree *t)
1674 while (t->stack != NULL) {
1675 /* If there's an open dir, get the next entry from there. */
1676 if (t->d != INVALID_HANDLE_VALUE) {
1677 r = tree_dir_next_windows(t, NULL);
1683 if (t->stack->flags & needsFirstVisit) {
1684 wchar_t *d = t->stack->name.s;
1685 t->stack->flags &= ~needsFirstVisit;
1686 if (!(d[0] == L'/' && d[1] == L'/' &&
1687 d[2] == L'?' && d[3] == L'/') &&
1688 (wcschr(d, L'*') || wcschr(d, L'?'))) {
1689 r = tree_dir_next_windows(t, d);
1694 HANDLE h = FindFirstFileW(d, &t->_findData);
1695 if (h == INVALID_HANDLE_VALUE) {
1696 la_dosmaperr(GetLastError());
1697 t->tree_errno = errno;
1698 t->visit_type = TREE_ERROR_DIR;
1699 return (t->visit_type);
1701 t->findData = &t->_findData;
1704 /* Top stack item needs a regular visit. */
1705 t->current = t->stack;
1706 tree_append(t, t->stack->name.s,
1707 archive_strlen(&(t->stack->name)));
1708 //t->dirname_length = t->path_length;
1710 t->stack->flags &= ~needsFirstVisit;
1711 return (t->visit_type = TREE_REGULAR);
1712 } else if (t->stack->flags & needsDescent) {
1713 /* Top stack item is dir to descend into. */
1714 t->current = t->stack;
1715 tree_append(t, t->stack->name.s,
1716 archive_strlen(&(t->stack->name)));
1717 t->stack->flags &= ~needsDescent;
1718 r = tree_descent(t);
1723 t->visit_type = TREE_POSTDESCENT;
1724 return (t->visit_type);
1725 } else if (t->stack->flags & needsOpen) {
1726 t->stack->flags &= ~needsOpen;
1727 r = tree_dir_next_windows(t, L"*");
1731 } else if (t->stack->flags & needsAscent) {
1732 /* Top stack item is dir and we're done with it. */
1735 t->visit_type = r != 0 ? r : TREE_POSTASCENT;
1736 return (t->visit_type);
1738 /* Top item on stack is dead. */
1740 t->flags &= ~hasLstat;
1741 t->flags &= ~hasStat;
1744 return (t->visit_type = 0);
1748 tree_dir_next_windows(struct tree *t, const wchar_t *pattern)
1750 const wchar_t *name;
1755 if (pattern != NULL) {
1756 struct archive_wstring pt;
1758 archive_string_init(&pt);
1759 archive_wstring_ensure(&pt,
1760 archive_strlen(&(t->full_path))
1761 + 2 + wcslen(pattern));
1762 archive_wstring_copy(&pt, &(t->full_path));
1763 archive_wstrappend_wchar(&pt, L'\\');
1764 archive_wstrcat(&pt, pattern);
1765 t->d = FindFirstFileW(pt.s, &t->_findData);
1766 archive_wstring_free(&pt);
1767 if (t->d == INVALID_HANDLE_VALUE) {
1768 la_dosmaperr(GetLastError());
1769 t->tree_errno = errno;
1770 r = tree_ascend(t); /* Undo "chdir" */
1772 t->visit_type = r != 0 ? r : TREE_ERROR_DIR;
1773 return (t->visit_type);
1775 t->findData = &t->_findData;
1777 } else if (!FindNextFileW(t->d, &t->_findData)) {
1779 t->d = INVALID_HANDLE_VALUE;
1783 name = t->findData->cFileName;
1784 namelen = wcslen(name);
1785 t->flags &= ~hasLstat;
1786 t->flags &= ~hasStat;
1787 if (name[0] == L'.' && name[1] == L'\0')
1789 if (name[0] == L'.' && name[1] == L'.' && name[2] == L'\0')
1791 tree_append(t, name, namelen);
1792 return (t->visit_type = TREE_REGULAR);
1796 #define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000)
1798 fileTimeToUtc(const FILETIME *filetime, time_t *t, long *ns)
1802 utc.HighPart = filetime->dwHighDateTime;
1803 utc.LowPart = filetime->dwLowDateTime;
1804 if (utc.QuadPart >= EPOC_TIME) {
1805 utc.QuadPart -= EPOC_TIME;
1806 /* milli seconds base */
1807 *t = (time_t)(utc.QuadPart / 10000000);
1808 /* nano seconds base */
1809 *ns = (long)(utc.QuadPart % 10000000) * 100;
1817 entry_copy_bhfi(struct archive_entry *entry, const wchar_t *path,
1818 const WIN32_FIND_DATAW *findData,
1819 const BY_HANDLE_FILE_INFORMATION *bhfi)
1825 fileTimeToUtc(&bhfi->ftLastAccessTime, &secs, &nsecs);
1826 archive_entry_set_atime(entry, secs, nsecs);
1827 fileTimeToUtc(&bhfi->ftLastWriteTime, &secs, &nsecs);
1828 archive_entry_set_mtime(entry, secs, nsecs);
1829 fileTimeToUtc(&bhfi->ftCreationTime, &secs, &nsecs);
1830 archive_entry_set_birthtime(entry, secs, nsecs);
1831 archive_entry_set_ctime(entry, secs, nsecs);
1832 archive_entry_set_dev(entry, bhfi_dev(bhfi));
1833 archive_entry_set_ino64(entry, bhfi_ino(bhfi));
1834 if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
1835 archive_entry_set_nlink(entry, bhfi->nNumberOfLinks + 1);
1837 archive_entry_set_nlink(entry, bhfi->nNumberOfLinks);
1838 archive_entry_set_size(entry,
1839 (((int64_t)bhfi->nFileSizeHigh) << 32)
1840 + bhfi->nFileSizeLow);
1841 archive_entry_set_uid(entry, 0);
1842 archive_entry_set_gid(entry, 0);
1843 archive_entry_set_rdev(entry, 0);
1845 mode = S_IRUSR | S_IRGRP | S_IROTH;
1846 if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_READONLY) == 0)
1847 mode |= S_IWUSR | S_IWGRP | S_IWOTH;
1848 if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) &&
1850 findData->dwReserved0 == IO_REPARSE_TAG_SYMLINK)
1852 else if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
1853 mode |= S_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH;
1858 p = wcsrchr(path, L'.');
1859 if (p != NULL && wcslen(p) == 4) {
1861 case L'B': case L'b':
1862 if ((p[2] == L'A' || p[2] == L'a' ) &&
1863 (p[3] == L'T' || p[3] == L't' ))
1864 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1866 case L'C': case L'c':
1867 if (((p[2] == L'M' || p[2] == L'm' ) &&
1868 (p[3] == L'D' || p[3] == L'd' )))
1869 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1871 case L'E': case L'e':
1872 if ((p[2] == L'X' || p[2] == L'x' ) &&
1873 (p[3] == L'E' || p[3] == L'e' ))
1874 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1881 archive_entry_set_mode(entry, mode);
1885 tree_archive_entry_copy_bhfi(struct archive_entry *entry, struct tree *t,
1886 const BY_HANDLE_FILE_INFORMATION *bhfi)
1888 entry_copy_bhfi(entry, tree_current_path(t), t->findData, bhfi);
1892 tree_current_file_information(struct tree *t, BY_HANDLE_FILE_INFORMATION *st,
1897 DWORD flag = FILE_FLAG_BACKUP_SEMANTICS;
1899 if (sim_lstat && tree_current_is_physical_link(t))
1900 flag |= FILE_FLAG_OPEN_REPARSE_POINT;
1901 h = CreateFileW(tree_current_access_path(t), 0, FILE_SHARE_READ, NULL,
1902 OPEN_EXISTING, flag, NULL);
1903 if (h == INVALID_HANDLE_VALUE) {
1904 la_dosmaperr(GetLastError());
1905 t->tree_errno = errno;
1908 r = GetFileInformationByHandle(h, st);
1914 * Get the stat() data for the entry just returned from tree_next().
1916 static const BY_HANDLE_FILE_INFORMATION *
1917 tree_current_stat(struct tree *t)
1919 if (!(t->flags & hasStat)) {
1920 if (!tree_current_file_information(t, &t->st, 0))
1922 t->flags |= hasStat;
1928 * Get the lstat() data for the entry just returned from tree_next().
1930 static const BY_HANDLE_FILE_INFORMATION *
1931 tree_current_lstat(struct tree *t)
1933 if (!(t->flags & hasLstat)) {
1934 if (!tree_current_file_information(t, &t->lst, 1))
1936 t->flags |= hasLstat;
1942 * Test whether current entry is a dir or link to a dir.
1945 tree_current_is_dir(struct tree *t)
1948 return (t->findData->dwFileAttributes
1949 & FILE_ATTRIBUTE_DIRECTORY);
1954 * Test whether current entry is a physical directory. Usually, we
1955 * already have at least one of stat() or lstat() in memory, so we
1956 * use tricks to try to avoid an extra trip to the disk.
1959 tree_current_is_physical_dir(struct tree *t)
1961 if (tree_current_is_physical_link(t))
1963 return (tree_current_is_dir(t));
1967 * Test whether current entry is a symbolic link.
1970 tree_current_is_physical_link(struct tree *t)
1973 return ((t->findData->dwFileAttributes
1974 & FILE_ATTRIBUTE_REPARSE_POINT) &&
1975 (t->findData->dwReserved0
1976 == IO_REPARSE_TAG_SYMLINK));
1981 * Test whether the same file has been in the tree as its parent.
1984 tree_target_is_same_as_parent(struct tree *t,
1985 const BY_HANDLE_FILE_INFORMATION *st)
1987 struct tree_entry *te;
1988 int64_t dev = bhfi_dev(st);
1989 int64_t ino = bhfi_ino(st);
1991 for (te = t->current->parent; te != NULL; te = te->parent) {
1992 if (te->dev == dev && te->ino == ino)
1999 * Return the access path for the entry just returned from tree_next().
2001 static const wchar_t *
2002 tree_current_access_path(struct tree *t)
2004 return (t->full_path.s);
2008 * Return the full path for the entry just returned from tree_next().
2010 static const wchar_t *
2011 tree_current_path(struct tree *t)
2017 * Terminate the traversal.
2020 tree_close(struct tree *t)
2025 if (t->entry_fh != INVALID_HANDLE_VALUE) {
2027 close_and_restore_time(t->entry_fh, t, &t->restore_time);
2028 t->entry_fh = INVALID_HANDLE_VALUE;
2030 /* Close the handle of FindFirstFileW */
2031 if (t->d != INVALID_HANDLE_VALUE) {
2033 t->d = INVALID_HANDLE_VALUE;
2036 /* Release anything remaining in the stack. */
2037 while (t->stack != NULL)
2042 * Release any resources.
2045 tree_free(struct tree *t)
2051 archive_wstring_free(&t->path);
2052 archive_wstring_free(&t->full_path);
2053 free(t->sparse_list);
2054 free(t->filesystem_table);
2055 for (i = 0; i < MAX_OVERLAPPED; i++) {
2057 VirtualFree(t->ol[i].buff, 0, MEM_RELEASE);
2058 CloseHandle(t->ol[i].ol.hEvent);
2065 * Populate the archive_entry with metadata from the disk.
2068 archive_read_disk_entry_from_file(struct archive *_a,
2069 struct archive_entry *entry, int fd, const struct stat *st)
2071 struct archive_read_disk *a = (struct archive_read_disk *)_a;
2072 const wchar_t *path;
2073 const wchar_t *wname;
2076 BY_HANDLE_FILE_INFORMATION bhfi;
2077 DWORD fileAttributes = 0;
2080 archive_clear_error(_a);
2081 wname = archive_entry_sourcepath_w(entry);
2083 wname = archive_entry_pathname_w(entry);
2084 if (wname == NULL) {
2085 archive_set_error(&a->archive, EINVAL,
2086 "Can't get a wide character version of the path");
2087 return (ARCHIVE_FAILED);
2089 path = __la_win_permissive_name_w(wname);
2093 * Get metadata through GetFileInformationByHandle().
2096 h = (HANDLE)_get_osfhandle(fd);
2097 r = GetFileInformationByHandle(h, &bhfi);
2099 la_dosmaperr(GetLastError());
2100 archive_set_error(&a->archive, errno,
2101 "Can't GetFileInformationByHandle");
2102 return (ARCHIVE_FAILED);
2104 entry_copy_bhfi(entry, path, NULL, &bhfi);
2106 WIN32_FIND_DATAW findData;
2107 DWORD flag, desiredAccess;
2109 h = FindFirstFileW(path, &findData);
2110 if (h == INVALID_HANDLE_VALUE) {
2111 la_dosmaperr(GetLastError());
2112 archive_set_error(&a->archive, errno,
2113 "Can't FindFirstFileW");
2114 return (ARCHIVE_FAILED);
2118 flag = FILE_FLAG_BACKUP_SEMANTICS;
2119 if (!a->follow_symlinks &&
2120 (findData.dwFileAttributes
2121 & FILE_ATTRIBUTE_REPARSE_POINT) &&
2122 (findData.dwReserved0 == IO_REPARSE_TAG_SYMLINK)) {
2123 flag |= FILE_FLAG_OPEN_REPARSE_POINT;
2125 } else if (findData.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {
2128 desiredAccess = GENERIC_READ;
2130 h = CreateFileW(path, desiredAccess, FILE_SHARE_READ, NULL,
2131 OPEN_EXISTING, flag, NULL);
2132 if (h == INVALID_HANDLE_VALUE) {
2133 la_dosmaperr(GetLastError());
2134 archive_set_error(&a->archive, errno,
2135 "Can't CreateFileW");
2136 return (ARCHIVE_FAILED);
2138 r = GetFileInformationByHandle(h, &bhfi);
2140 la_dosmaperr(GetLastError());
2141 archive_set_error(&a->archive, errno,
2142 "Can't GetFileInformationByHandle");
2144 return (ARCHIVE_FAILED);
2146 entry_copy_bhfi(entry, path, &findData, &bhfi);
2148 fileAttributes = bhfi.dwFileAttributes;
2150 archive_entry_copy_stat(entry, st);
2151 h = INVALID_HANDLE_VALUE;
2154 /* Lookup uname/gname */
2155 name = archive_read_disk_uname(_a, archive_entry_uid(entry));
2157 archive_entry_copy_uname(entry, name);
2158 name = archive_read_disk_gname(_a, archive_entry_gid(entry));
2160 archive_entry_copy_gname(entry, name);
2163 * Can this file be sparse file ?
2165 if (archive_entry_filetype(entry) != AE_IFREG
2166 || archive_entry_size(entry) <= 0
2167 || archive_entry_hardlink(entry) != NULL) {
2168 if (h != INVALID_HANDLE_VALUE && fd < 0)
2170 return (ARCHIVE_OK);
2173 if (h == INVALID_HANDLE_VALUE) {
2175 h = (HANDLE)_get_osfhandle(fd);
2177 h = CreateFileW(path, GENERIC_READ, FILE_SHARE_READ, NULL,
2178 OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
2179 if (h == INVALID_HANDLE_VALUE) {
2180 la_dosmaperr(GetLastError());
2181 archive_set_error(&a->archive, errno,
2182 "Can't CreateFileW");
2183 return (ARCHIVE_FAILED);
2186 r = GetFileInformationByHandle(h, &bhfi);
2188 la_dosmaperr(GetLastError());
2189 archive_set_error(&a->archive, errno,
2190 "Can't GetFileInformationByHandle");
2191 if (h != INVALID_HANDLE_VALUE && fd < 0)
2193 return (ARCHIVE_FAILED);
2195 fileAttributes = bhfi.dwFileAttributes;
2198 /* Sparse file must be set a mark, FILE_ATTRIBUTE_SPARSE_FILE */
2199 if ((fileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) == 0) {
2202 return (ARCHIVE_OK);
2205 r = setup_sparse_from_disk(a, entry, h);
2213 * Windows sparse interface.
2215 #if defined(__MINGW32__) && !defined(FSCTL_QUERY_ALLOCATED_RANGES)
2216 #define FSCTL_QUERY_ALLOCATED_RANGES 0x940CF
2218 LARGE_INTEGER FileOffset;
2219 LARGE_INTEGER Length;
2220 } FILE_ALLOCATED_RANGE_BUFFER;
2224 setup_sparse_from_disk(struct archive_read_disk *a,
2225 struct archive_entry *entry, HANDLE handle)
2227 FILE_ALLOCATED_RANGE_BUFFER range, *outranges = NULL;
2228 size_t outranges_size;
2229 int64_t entry_size = archive_entry_size(entry);
2230 int exit_sts = ARCHIVE_OK;
2232 range.FileOffset.QuadPart = 0;
2233 range.Length.QuadPart = entry_size;
2234 outranges_size = 2048;
2235 outranges = (FILE_ALLOCATED_RANGE_BUFFER *)malloc(outranges_size);
2236 if (outranges == NULL) {
2237 archive_set_error(&a->archive, ENOMEM,
2238 "Couldn't allocate memory");
2239 exit_sts = ARCHIVE_FATAL;
2240 goto exit_setup_sparse;
2248 ret = DeviceIoControl(handle,
2249 FSCTL_QUERY_ALLOCATED_RANGES,
2250 &range, sizeof(range), outranges,
2251 (DWORD)outranges_size, &retbytes, NULL);
2252 if (ret == 0 && GetLastError() == ERROR_MORE_DATA) {
2254 outranges_size *= 2;
2255 outranges = (FILE_ALLOCATED_RANGE_BUFFER *)
2256 malloc(outranges_size);
2257 if (outranges == NULL) {
2258 archive_set_error(&a->archive, ENOMEM,
2259 "Couldn't allocate memory");
2260 exit_sts = ARCHIVE_FATAL;
2261 goto exit_setup_sparse;
2271 n = retbytes / sizeof(outranges[0]);
2273 outranges[0].FileOffset.QuadPart == 0 &&
2274 outranges[0].Length.QuadPart == entry_size)
2275 break;/* This is not sparse. */
2276 for (i = 0; i < n; i++)
2277 archive_entry_sparse_add_entry(entry,
2278 outranges[i].FileOffset.QuadPart,
2279 outranges[i].Length.QuadPart);
2280 range.FileOffset.QuadPart =
2281 outranges[n-1].FileOffset.QuadPart
2282 + outranges[n-1].Length.QuadPart;
2283 range.Length.QuadPart =
2284 entry_size - range.FileOffset.QuadPart;
2285 if (range.Length.QuadPart > 0)
2288 /* The remaining data is hole. */
2289 archive_entry_sparse_add_entry(entry,
2290 range.FileOffset.QuadPart,
2291 range.Length.QuadPart);
2295 la_dosmaperr(GetLastError());
2296 archive_set_error(&a->archive, errno,
2297 "DeviceIoControl Failed: %lu", GetLastError());
2298 exit_sts = ARCHIVE_FAILED;
2299 goto exit_setup_sparse;