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 READ_BUFFER_SIZE (1024 * 64) /* Default to 64KB per https://technet.microsoft.com/en-us/library/cc938632.aspx */
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, la_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, la_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, la_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->flags = ARCHIVE_READDISK_MAC_COPYFILE;
402 return (&a->archive);
406 _archive_read_free(struct archive *_a)
408 struct archive_read_disk *a = (struct archive_read_disk *)_a;
413 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
414 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free");
416 if (a->archive.state != ARCHIVE_STATE_CLOSED)
417 r = _archive_read_close(&a->archive);
422 if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL)
423 (a->cleanup_gname)(a->lookup_gname_data);
424 if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL)
425 (a->cleanup_uname)(a->lookup_uname_data);
426 archive_string_free(&a->archive.error_string);
427 archive_entry_free(a->entry);
428 a->archive.magic = 0;
434 _archive_read_close(struct archive *_a)
436 struct archive_read_disk *a = (struct archive_read_disk *)_a;
438 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
439 ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close");
441 if (a->archive.state != ARCHIVE_STATE_FATAL)
442 a->archive.state = ARCHIVE_STATE_CLOSED;
450 setup_symlink_mode(struct archive_read_disk *a, char symlink_mode,
453 a->symlink_mode = symlink_mode;
454 a->follow_symlinks = follow_symlinks;
455 if (a->tree != NULL) {
456 a->tree->initial_symlink_mode = a->symlink_mode;
457 a->tree->symlink_mode = a->symlink_mode;
462 archive_read_disk_set_symlink_logical(struct archive *_a)
464 struct archive_read_disk *a = (struct archive_read_disk *)_a;
465 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
466 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical");
467 setup_symlink_mode(a, 'L', 1);
472 archive_read_disk_set_symlink_physical(struct archive *_a)
474 struct archive_read_disk *a = (struct archive_read_disk *)_a;
475 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
476 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical");
477 setup_symlink_mode(a, 'P', 0);
482 archive_read_disk_set_symlink_hybrid(struct archive *_a)
484 struct archive_read_disk *a = (struct archive_read_disk *)_a;
485 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
486 ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid");
487 setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */
492 archive_read_disk_set_atime_restored(struct archive *_a)
494 struct archive_read_disk *a = (struct archive_read_disk *)_a;
495 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
496 ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime");
497 a->flags |= ARCHIVE_READDISK_RESTORE_ATIME;
499 a->tree->flags |= needsRestoreTimes;
504 archive_read_disk_set_behavior(struct archive *_a, int flags)
506 struct archive_read_disk *a = (struct archive_read_disk *)_a;
509 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
510 ARCHIVE_STATE_ANY, "archive_read_disk_honor_nodump");
514 if (flags & ARCHIVE_READDISK_RESTORE_ATIME)
515 r = archive_read_disk_set_atime_restored(_a);
518 a->tree->flags &= ~needsRestoreTimes;
524 * Trivial implementations of gname/uname lookup functions.
525 * These are normally overridden by the client, but these stub
526 * versions ensure that we always have something that works.
529 trivial_lookup_gname(void *private_data, int64_t gid)
531 (void)private_data; /* UNUSED */
532 (void)gid; /* UNUSED */
537 trivial_lookup_uname(void *private_data, int64_t uid)
539 (void)private_data; /* UNUSED */
540 (void)uid; /* UNUSED */
545 align_num_per_sector(struct tree *t, int64_t size)
549 size += t->current_filesystem->bytesPerSector -1;
550 surplus = size % t->current_filesystem->bytesPerSector;
556 start_next_async_read(struct archive_read_disk *a, struct tree *t)
558 struct la_overlapped *olp;
559 DWORD buffbytes, rbytes;
561 if (t->ol_remaining_bytes == 0)
562 return (ARCHIVE_EOF);
564 olp = &(t->ol[t->ol_idx_doing]);
565 t->ol_idx_doing = (t->ol_idx_doing + 1) % MAX_OVERLAPPED;
567 /* Allocate read buffer. */
568 if (olp->buff == NULL) {
570 size_t s = (size_t)align_num_per_sector(t, READ_BUFFER_SIZE);
571 p = VirtualAlloc(NULL, s, MEM_COMMIT, PAGE_READWRITE);
573 archive_set_error(&a->archive, ENOMEM,
574 "Couldn't allocate memory");
575 a->archive.state = ARCHIVE_STATE_FATAL;
576 return (ARCHIVE_FATAL);
580 olp->_a = &a->archive;
581 olp->ol.hEvent = CreateEventW(NULL, TRUE, FALSE, NULL);
582 if (olp->ol.hEvent == NULL) {
583 la_dosmaperr(GetLastError());
584 archive_set_error(&a->archive, errno,
585 "CreateEvent failed");
586 a->archive.state = ARCHIVE_STATE_FATAL;
587 return (ARCHIVE_FATAL);
590 ResetEvent(olp->ol.hEvent);
592 buffbytes = (DWORD)olp->buff_size;
593 if (buffbytes > t->current_sparse->length)
594 buffbytes = (DWORD)t->current_sparse->length;
597 if (t->current_sparse->offset > t->ol_total) {
598 t->ol_remaining_bytes -=
599 t->current_sparse->offset - t->ol_total;
602 olp->offset = t->current_sparse->offset;
603 olp->ol.Offset = (DWORD)(olp->offset & 0xffffffff);
604 olp->ol.OffsetHigh = (DWORD)(olp->offset >> 32);
606 if (t->ol_remaining_bytes > buffbytes) {
607 olp->bytes_expected = buffbytes;
608 t->ol_remaining_bytes -= buffbytes;
610 olp->bytes_expected = (size_t)t->ol_remaining_bytes;
611 t->ol_remaining_bytes = 0;
613 olp->bytes_transferred = 0;
614 t->current_sparse->offset += buffbytes;
615 t->current_sparse->length -= buffbytes;
616 t->ol_total = t->current_sparse->offset;
617 if (t->current_sparse->length == 0 && t->ol_remaining_bytes > 0)
620 if (!ReadFile(t->entry_fh, olp->buff, buffbytes, &rbytes, &(olp->ol))) {
623 lasterr = GetLastError();
624 if (lasterr == ERROR_HANDLE_EOF) {
625 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
626 "Reading file truncated");
627 a->archive.state = ARCHIVE_STATE_FATAL;
628 return (ARCHIVE_FATAL);
629 } else if (lasterr != ERROR_IO_PENDING) {
630 if (lasterr == ERROR_NO_DATA)
632 else if (lasterr == ERROR_ACCESS_DENIED)
635 la_dosmaperr(lasterr);
636 archive_set_error(&a->archive, errno, "Read error");
637 a->archive.state = ARCHIVE_STATE_FATAL;
638 return (ARCHIVE_FATAL);
641 olp->bytes_transferred = rbytes;
644 return (t->ol_remaining_bytes == 0)? ARCHIVE_EOF: ARCHIVE_OK;
648 cancel_async(struct tree *t)
650 if (t->ol_num_doing != t->ol_num_done) {
651 CancelIo(t->entry_fh);
652 t->ol_num_doing = t->ol_num_done = 0;
657 _archive_read_data_block(struct archive *_a, const void **buff,
658 size_t *size, int64_t *offset)
660 struct archive_read_disk *a = (struct archive_read_disk *)_a;
661 struct tree *t = a->tree;
662 struct la_overlapped *olp;
663 DWORD bytes_transferred;
664 int r = ARCHIVE_FATAL;
666 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
667 "archive_read_data_block");
669 if (t->entry_eof || t->entry_remaining_bytes <= 0) {
671 goto abort_read_data;
675 * Make a request to read the file in asynchronous.
677 if (t->ol_num_doing == 0) {
679 r = start_next_async_read(a, t);
680 if (r == ARCHIVE_FATAL)
681 goto abort_read_data;
684 } while (r == ARCHIVE_OK && t->ol_num_doing < MAX_OVERLAPPED);
686 if ((r = start_next_async_read(a, t)) == ARCHIVE_FATAL)
687 goto abort_read_data;
690 olp = &(t->ol[t->ol_idx_done]);
691 t->ol_idx_done = (t->ol_idx_done + 1) % MAX_OVERLAPPED;
692 if (olp->bytes_transferred)
693 bytes_transferred = (DWORD)olp->bytes_transferred;
694 else if (!GetOverlappedResult(t->entry_fh, &(olp->ol),
695 &bytes_transferred, TRUE)) {
696 la_dosmaperr(GetLastError());
697 archive_set_error(&a->archive, errno,
698 "GetOverlappedResult failed");
699 a->archive.state = ARCHIVE_STATE_FATAL;
701 goto abort_read_data;
705 if (bytes_transferred == 0 ||
706 olp->bytes_expected != bytes_transferred) {
707 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
708 "Reading file truncated");
709 a->archive.state = ARCHIVE_STATE_FATAL;
711 goto abort_read_data;
715 *size = bytes_transferred;
716 *offset = olp->offset;
717 if (olp->offset > t->entry_total)
718 t->entry_remaining_bytes -= olp->offset - t->entry_total;
719 t->entry_total = olp->offset + *size;
720 t->entry_remaining_bytes -= *size;
721 if (t->entry_remaining_bytes == 0) {
722 /* Close the current file descriptor */
723 close_and_restore_time(t->entry_fh, t, &t->restore_time);
724 t->entry_fh = INVALID_HANDLE_VALUE;
732 *offset = t->entry_total;
733 if (t->entry_fh != INVALID_HANDLE_VALUE) {
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;
743 next_entry(struct archive_read_disk *a, struct tree *t,
744 struct archive_entry *entry)
746 const BY_HANDLE_FILE_INFORMATION *st;
747 const BY_HANDLE_FILE_INFORMATION *lst;
755 switch (tree_next(t)) {
756 case TREE_ERROR_FATAL:
757 archive_set_error(&a->archive, t->tree_errno,
758 "%ls: Unable to continue traversing directory tree",
759 tree_current_path(t));
760 a->archive.state = ARCHIVE_STATE_FATAL;
761 return (ARCHIVE_FATAL);
763 archive_set_error(&a->archive, t->tree_errno,
764 "%ls: Couldn't visit directory",
765 tree_current_path(t));
766 return (ARCHIVE_FAILED);
768 return (ARCHIVE_EOF);
769 case TREE_POSTDESCENT:
770 case TREE_POSTASCENT:
773 lst = tree_current_lstat(t);
775 archive_set_error(&a->archive, t->tree_errno,
777 tree_current_path(t));
778 return (ARCHIVE_FAILED);
782 } while (lst == NULL);
784 archive_entry_copy_pathname_w(entry, tree_current_path(t));
787 * Perform path matching.
790 r = archive_match_path_excluded(a->matching, entry);
792 archive_set_error(&(a->archive), errno,
793 "Failed : %s", archive_error_string(a->matching));
797 if (a->excluded_cb_func)
798 a->excluded_cb_func(&(a->archive),
799 a->excluded_cb_data, entry);
800 return (ARCHIVE_RETRY);
805 * Distinguish 'L'/'P'/'H' symlink following.
807 switch(t->symlink_mode) {
809 /* 'H': After the first item, rest like 'P'. */
810 t->symlink_mode = 'P';
811 /* 'H': First item (from command line) like 'L'. */
814 /* 'L': Do descend through a symlink to dir. */
815 descend = tree_current_is_dir(t);
816 /* 'L': Follow symlinks to files. */
817 a->symlink_mode = 'L';
818 a->follow_symlinks = 1;
819 /* 'L': Archive symlinks as targets, if we can. */
820 st = tree_current_stat(t);
821 if (st != NULL && !tree_target_is_same_as_parent(t, st))
823 /* If stat fails, we have a broken symlink;
824 * in that case, don't follow the link. */
827 /* 'P': Don't descend through a symlink to dir. */
828 descend = tree_current_is_physical_dir(t);
829 /* 'P': Don't follow symlinks to files. */
830 a->symlink_mode = 'P';
831 a->follow_symlinks = 0;
832 /* 'P': Archive symlinks as symlinks. */
837 if (update_current_filesystem(a, bhfi_dev(st)) != ARCHIVE_OK) {
838 a->archive.state = ARCHIVE_STATE_FATAL;
839 return (ARCHIVE_FATAL);
841 if (t->initial_filesystem_id == -1)
842 t->initial_filesystem_id = t->current_filesystem_id;
843 if (a->flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS) {
844 if (t->initial_filesystem_id != t->current_filesystem_id)
845 return (ARCHIVE_RETRY);
847 t->descend = descend;
849 tree_archive_entry_copy_bhfi(entry, t, st);
851 /* Save the times to be restored. This must be in before
852 * calling archive_read_disk_descend() or any chance of it,
853 * especially, invoking a callback. */
854 t->restore_time.lastWriteTime = st->ftLastWriteTime;
855 t->restore_time.lastAccessTime = st->ftLastAccessTime;
856 t->restore_time.filetype = archive_entry_filetype(entry);
859 * Perform time matching.
862 r = archive_match_time_excluded(a->matching, entry);
864 archive_set_error(&(a->archive), errno,
865 "Failed : %s", archive_error_string(a->matching));
869 if (a->excluded_cb_func)
870 a->excluded_cb_func(&(a->archive),
871 a->excluded_cb_data, entry);
872 return (ARCHIVE_RETRY);
876 /* Lookup uname/gname */
877 name = archive_read_disk_uname(&(a->archive), archive_entry_uid(entry));
879 archive_entry_copy_uname(entry, name);
880 name = archive_read_disk_gname(&(a->archive), archive_entry_gid(entry));
882 archive_entry_copy_gname(entry, name);
885 * Perform owner matching.
888 r = archive_match_owner_excluded(a->matching, entry);
890 archive_set_error(&(a->archive), errno,
891 "Failed : %s", archive_error_string(a->matching));
895 if (a->excluded_cb_func)
896 a->excluded_cb_func(&(a->archive),
897 a->excluded_cb_data, entry);
898 return (ARCHIVE_RETRY);
903 * Invoke a meta data filter callback.
905 if (a->metadata_filter_func) {
906 if (!a->metadata_filter_func(&(a->archive),
907 a->metadata_filter_data, entry))
908 return (ARCHIVE_RETRY);
911 archive_entry_copy_sourcepath_w(entry, tree_current_access_path(t));
914 if (archive_entry_filetype(entry) == AE_IFREG &&
915 archive_entry_size(entry) > 0) {
916 DWORD flags = FILE_FLAG_BACKUP_SEMANTICS;
918 flags |= FILE_FLAG_OVERLAPPED;
920 flags |= FILE_FLAG_NO_BUFFERING;
922 flags |= FILE_FLAG_SEQUENTIAL_SCAN;
923 t->entry_fh = CreateFileW(tree_current_access_path(t),
924 GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, flags, NULL);
925 if (t->entry_fh == INVALID_HANDLE_VALUE) {
926 la_dosmaperr(GetLastError());
927 archive_set_error(&a->archive, errno,
928 "Couldn't open %ls", tree_current_path(a->tree));
929 return (ARCHIVE_FAILED);
932 /* Find sparse data from the disk. */
933 if (archive_entry_hardlink(entry) == NULL &&
934 (st->dwFileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) != 0)
935 r = setup_sparse_from_disk(a, entry, t->entry_fh);
941 _archive_read_next_header(struct archive *_a, struct archive_entry **entryp)
944 struct archive_read_disk *a = (struct archive_read_disk *)_a;
946 ret = _archive_read_next_header2(_a, a->entry);
952 _archive_read_next_header2(struct archive *_a, struct archive_entry *entry)
954 struct archive_read_disk *a = (struct archive_read_disk *)_a;
958 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
959 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
960 "archive_read_next_header2");
963 if (t->entry_fh != INVALID_HANDLE_VALUE) {
965 close_and_restore_time(t->entry_fh, t, &t->restore_time);
966 t->entry_fh = INVALID_HANDLE_VALUE;
969 while ((r = next_entry(a, t, entry)) == ARCHIVE_RETRY)
970 archive_entry_clear(entry);
973 * EOF and FATAL are persistent at this layer. By
974 * modifying the state, we guarantee that future calls to
975 * read a header or read data will fail.
979 a->archive.state = ARCHIVE_STATE_EOF;
984 if (archive_entry_filetype(entry) == AE_IFREG) {
985 t->entry_remaining_bytes = archive_entry_size(entry);
986 t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0;
988 setup_sparse(a, entry) != ARCHIVE_OK)
989 return (ARCHIVE_FATAL);
991 t->entry_remaining_bytes = 0;
994 t->ol_idx_doing = t->ol_idx_done = 0;
995 t->ol_num_doing = t->ol_num_done = 0;
996 t->ol_remaining_bytes = t->entry_remaining_bytes;
998 a->archive.state = ARCHIVE_STATE_DATA;
1003 a->archive.state = ARCHIVE_STATE_FATAL;
1007 __archive_reset_read_data(&a->archive);
1012 setup_sparse(struct archive_read_disk *a, struct archive_entry *entry)
1014 struct tree *t = a->tree;
1015 int64_t aligned, length, offset;
1018 t->sparse_count = archive_entry_sparse_reset(entry);
1019 if (t->sparse_count+1 > t->sparse_list_size) {
1020 free(t->sparse_list);
1021 t->sparse_list_size = t->sparse_count + 1;
1022 t->sparse_list = malloc(sizeof(t->sparse_list[0]) *
1023 t->sparse_list_size);
1024 if (t->sparse_list == NULL) {
1025 t->sparse_list_size = 0;
1026 archive_set_error(&a->archive, ENOMEM,
1027 "Can't allocate data");
1028 a->archive.state = ARCHIVE_STATE_FATAL;
1029 return (ARCHIVE_FATAL);
1033 * Get sparse list and make sure those offsets and lengths are
1034 * aligned by a sector size.
1036 for (i = 0; i < t->sparse_count; i++) {
1037 archive_entry_sparse_next(entry, &offset, &length);
1038 aligned = align_num_per_sector(t, offset);
1039 if (aligned != offset) {
1040 aligned -= t->current_filesystem->bytesPerSector;
1041 length += offset - aligned;
1043 t->sparse_list[i].offset = aligned;
1044 aligned = align_num_per_sector(t, length);
1045 t->sparse_list[i].length = aligned;
1048 aligned = align_num_per_sector(t, archive_entry_size(entry));
1050 t->sparse_list[i].offset = 0;
1051 t->sparse_list[i].length = aligned;
1055 t->sparse_list[i].offset = aligned;
1056 t->sparse_list[i].length = 0;
1057 for (i = 0; i < last; i++) {
1058 if ((t->sparse_list[i].offset +
1059 t->sparse_list[i].length) <=
1060 t->sparse_list[i+1].offset)
1063 * Now sparse_list[i+1] is overlapped by sparse_list[i].
1066 length = t->sparse_list[i+1].offset -
1067 t->sparse_list[i].offset;
1068 t->sparse_list[i+1].offset = t->sparse_list[i].offset;
1069 t->sparse_list[i+1].length += length;
1070 /* Remove sparse_list[i]. */
1071 for (j = i; j < last; j++) {
1072 t->sparse_list[j].offset =
1073 t->sparse_list[j+1].offset;
1074 t->sparse_list[j].length =
1075 t->sparse_list[j+1].length;
1080 t->current_sparse = t->sparse_list;
1082 return (ARCHIVE_OK);
1086 archive_read_disk_set_matching(struct archive *_a, struct archive *_ma,
1087 void (*_excluded_func)(struct archive *, void *, struct archive_entry *),
1090 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1091 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1092 ARCHIVE_STATE_ANY, "archive_read_disk_set_matching");
1094 a->excluded_cb_func = _excluded_func;
1095 a->excluded_cb_data = _client_data;
1096 return (ARCHIVE_OK);
1100 archive_read_disk_set_metadata_filter_callback(struct archive *_a,
1101 int (*_metadata_filter_func)(struct archive *, void *,
1102 struct archive_entry *), void *_client_data)
1104 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1106 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY,
1107 "archive_read_disk_set_metadata_filter_callback");
1109 a->metadata_filter_func = _metadata_filter_func;
1110 a->metadata_filter_data = _client_data;
1111 return (ARCHIVE_OK);
1115 archive_read_disk_can_descend(struct archive *_a)
1117 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1118 struct tree *t = a->tree;
1120 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1121 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1122 "archive_read_disk_can_descend");
1124 return (t->visit_type == TREE_REGULAR && t->descend);
1128 * Called by the client to mark the directory just returned from
1129 * tree_next() as needing to be visited.
1132 archive_read_disk_descend(struct archive *_a)
1134 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1135 struct tree *t = a->tree;
1137 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1138 ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA,
1139 "archive_read_disk_descend");
1141 if (t->visit_type != TREE_REGULAR || !t->descend)
1142 return (ARCHIVE_OK);
1144 if (tree_current_is_physical_dir(t)) {
1145 tree_push(t, t->basename, t->full_path.s,
1146 t->current_filesystem_id,
1147 bhfi_dev(&(t->lst)), bhfi_ino(&(t->lst)),
1149 t->stack->flags |= isDir;
1150 } else if (tree_current_is_dir(t)) {
1151 tree_push(t, t->basename, t->full_path.s,
1152 t->current_filesystem_id,
1153 bhfi_dev(&(t->st)), bhfi_ino(&(t->st)),
1155 t->stack->flags |= isDirLink;
1158 return (ARCHIVE_OK);
1162 archive_read_disk_open(struct archive *_a, const char *pathname)
1164 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1165 struct archive_wstring wpath;
1168 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1169 ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1170 "archive_read_disk_open");
1171 archive_clear_error(&a->archive);
1173 /* Make a wchar_t string from a char string. */
1174 archive_string_init(&wpath);
1175 if (archive_wstring_append_from_mbs(&wpath, pathname,
1176 strlen(pathname)) != 0) {
1177 if (errno == ENOMEM)
1178 archive_set_error(&a->archive, ENOMEM,
1179 "Can't allocate memory");
1181 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1182 "Can't convert a path to a wchar_t string");
1183 a->archive.state = ARCHIVE_STATE_FATAL;
1184 ret = ARCHIVE_FATAL;
1186 ret = _archive_read_disk_open_w(_a, wpath.s);
1188 archive_wstring_free(&wpath);
1193 archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1195 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1197 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC,
1198 ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED,
1199 "archive_read_disk_open_w");
1200 archive_clear_error(&a->archive);
1202 return (_archive_read_disk_open_w(_a, pathname));
1206 _archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname)
1208 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1210 if (a->tree != NULL)
1211 a->tree = tree_reopen(a->tree, pathname,
1212 a->flags & ARCHIVE_READDISK_RESTORE_ATIME);
1214 a->tree = tree_open(pathname, a->symlink_mode,
1215 a->flags & ARCHIVE_READDISK_RESTORE_ATIME);
1216 if (a->tree == NULL) {
1217 archive_set_error(&a->archive, ENOMEM,
1218 "Can't allocate directory traversal data");
1219 a->archive.state = ARCHIVE_STATE_FATAL;
1220 return (ARCHIVE_FATAL);
1222 a->archive.state = ARCHIVE_STATE_HEADER;
1224 return (ARCHIVE_OK);
1228 * Return a current filesystem ID which is index of the filesystem entry
1229 * you've visited through archive_read_disk.
1232 archive_read_disk_current_filesystem(struct archive *_a)
1234 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1236 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1237 "archive_read_disk_current_filesystem");
1239 return (a->tree->current_filesystem_id);
1243 update_current_filesystem(struct archive_read_disk *a, int64_t dev)
1245 struct tree *t = a->tree;
1248 if (t->current_filesystem != NULL &&
1249 t->current_filesystem->dev == dev)
1250 return (ARCHIVE_OK);
1252 for (i = 0; i < t->max_filesystem_id; i++) {
1253 if (t->filesystem_table[i].dev == dev) {
1254 /* There is the filesystem ID we've already generated. */
1255 t->current_filesystem_id = i;
1256 t->current_filesystem = &(t->filesystem_table[i]);
1257 return (ARCHIVE_OK);
1262 * There is a new filesystem, we generate a new ID for.
1264 fid = t->max_filesystem_id++;
1265 if (t->max_filesystem_id > t->allocated_filesystem) {
1269 s = t->max_filesystem_id * 2;
1270 p = realloc(t->filesystem_table,
1271 s * sizeof(*t->filesystem_table));
1273 archive_set_error(&a->archive, ENOMEM,
1274 "Can't allocate tar data");
1275 return (ARCHIVE_FATAL);
1277 t->filesystem_table = (struct filesystem *)p;
1278 t->allocated_filesystem = (int)s;
1280 t->current_filesystem_id = fid;
1281 t->current_filesystem = &(t->filesystem_table[fid]);
1282 t->current_filesystem->dev = dev;
1284 return (setup_current_filesystem(a));
1288 * Returns 1 if current filesystem is generated filesystem, 0 if it is not
1289 * or -1 if it is unknown.
1292 archive_read_disk_current_filesystem_is_synthetic(struct archive *_a)
1294 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1296 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1297 "archive_read_disk_current_filesystem");
1299 return (a->tree->current_filesystem->synthetic);
1303 * Returns 1 if current filesystem is remote filesystem, 0 if it is not
1304 * or -1 if it is unknown.
1307 archive_read_disk_current_filesystem_is_remote(struct archive *_a)
1309 struct archive_read_disk *a = (struct archive_read_disk *)_a;
1311 archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA,
1312 "archive_read_disk_current_filesystem");
1314 return (a->tree->current_filesystem->remote);
1318 * If symlink is broken, statfs or statvfs will fail.
1319 * Use its directory path instead.
1322 safe_path_for_statfs(struct tree *t)
1324 const wchar_t *path;
1325 wchar_t *cp, *p = NULL;
1327 path = tree_current_access_path(t);
1328 if (tree_current_stat(t) == NULL) {
1330 cp = wcsrchr(p, '/');
1331 if (cp != NULL && wcslen(cp) >= 2) {
1342 * Get conditions of synthetic and remote on Windows
1345 setup_current_filesystem(struct archive_read_disk *a)
1347 struct tree *t = a->tree;
1351 t->current_filesystem->synthetic = -1;/* Not supported */
1352 path = safe_path_for_statfs(t);
1353 if (!GetVolumePathNameW(path, vol, sizeof(vol)/sizeof(vol[0]))) {
1355 t->current_filesystem->remote = -1;
1356 t->current_filesystem->bytesPerSector = 0;
1357 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1358 "GetVolumePathName failed: %d", (int)GetLastError());
1359 return (ARCHIVE_FAILED);
1362 switch (GetDriveTypeW(vol)) {
1364 case DRIVE_NO_ROOT_DIR:
1365 t->current_filesystem->remote = -1;
1368 t->current_filesystem->remote = 1;
1371 t->current_filesystem->remote = 0;
1375 if (!GetDiskFreeSpaceW(vol, NULL,
1376 &(t->current_filesystem->bytesPerSector), NULL, NULL)) {
1377 t->current_filesystem->bytesPerSector = 0;
1378 archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
1379 "GetDiskFreeSpace failed: %d", (int)GetLastError());
1380 return (ARCHIVE_FAILED);
1383 return (ARCHIVE_OK);
1387 close_and_restore_time(HANDLE h, struct tree *t, struct restore_time *rt)
1392 if (h == INVALID_HANDLE_VALUE && AE_IFLNK == rt->filetype)
1395 /* Close a file descriptor.
1396 * It will not be used for SetFileTime() because it has been opened
1397 * by a read only mode.
1399 if (h != INVALID_HANDLE_VALUE)
1401 if ((t->flags & needsRestoreTimes) == 0)
1404 handle = CreateFileW(rt->full_path, FILE_WRITE_ATTRIBUTES,
1405 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
1406 if (handle == INVALID_HANDLE_VALUE) {
1411 if (SetFileTime(handle, NULL, &rt->lastAccessTime,
1412 &rt->lastWriteTime) == 0) {
1417 CloseHandle(handle);
1422 * Add a directory path to the current stack.
1425 tree_push(struct tree *t, const wchar_t *path, const wchar_t *full_path,
1426 int filesystem_id, int64_t dev, int64_t ino, struct restore_time *rt)
1428 struct tree_entry *te;
1430 te = calloc(1, sizeof(*te));
1431 te->next = t->stack;
1432 te->parent = t->current;
1434 te->depth = te->parent->depth + 1;
1436 archive_string_init(&te->name);
1437 archive_wstrcpy(&te->name, path);
1438 archive_string_init(&te->full_path);
1439 archive_wstrcpy(&te->full_path, full_path);
1440 te->flags = needsDescent | needsOpen | needsAscent;
1441 te->filesystem_id = filesystem_id;
1444 te->dirname_length = t->dirname_length;
1445 te->full_path_dir_length = t->full_path_dir_length;
1446 te->restore_time.full_path = te->full_path.s;
1448 te->restore_time.lastWriteTime = rt->lastWriteTime;
1449 te->restore_time.lastAccessTime = rt->lastAccessTime;
1450 te->restore_time.filetype = rt->filetype;
1455 * Append a name to the current dir path.
1458 tree_append(struct tree *t, const wchar_t *name, size_t name_length)
1462 t->path.s[t->dirname_length] = L'\0';
1463 t->path.length = t->dirname_length;
1464 /* Strip trailing '/' from name, unless entire name is "/". */
1465 while (name_length > 1 && name[name_length - 1] == L'/')
1468 /* Resize pathname buffer as needed. */
1469 size_needed = name_length + t->dirname_length + 2;
1470 archive_wstring_ensure(&t->path, size_needed);
1471 /* Add a separating '/' if it's needed. */
1472 if (t->dirname_length > 0 &&
1473 t->path.s[archive_strlen(&t->path)-1] != L'/')
1474 archive_wstrappend_wchar(&t->path, L'/');
1475 t->basename = t->path.s + archive_strlen(&t->path);
1476 archive_wstrncat(&t->path, name, name_length);
1477 t->restore_time.full_path = t->basename;
1478 if (t->full_path_dir_length > 0) {
1479 t->full_path.s[t->full_path_dir_length] = L'\0';
1480 t->full_path.length = t->full_path_dir_length;
1481 size_needed = name_length + t->full_path_dir_length + 2;
1482 archive_wstring_ensure(&t->full_path, size_needed);
1483 /* Add a separating '\' if it's needed. */
1484 if (t->full_path.s[archive_strlen(&t->full_path)-1] != L'\\')
1485 archive_wstrappend_wchar(&t->full_path, L'\\');
1486 archive_wstrncat(&t->full_path, name, name_length);
1487 t->restore_time.full_path = t->full_path.s;
1492 * Open a directory tree for traversal.
1494 static struct tree *
1495 tree_open(const wchar_t *path, int symlink_mode, int restore_time)
1499 t = calloc(1, sizeof(*t));
1500 archive_string_init(&(t->full_path));
1501 archive_string_init(&t->path);
1502 archive_wstring_ensure(&t->path, 15);
1503 t->initial_symlink_mode = symlink_mode;
1504 return (tree_reopen(t, path, restore_time));
1507 static struct tree *
1508 tree_reopen(struct tree *t, const wchar_t *path, int restore_time)
1510 struct archive_wstring ws;
1511 wchar_t *pathname, *p, *base;
1513 t->flags = (restore_time != 0)?needsRestoreTimes:0;
1516 t->full_path_dir_length = 0;
1517 t->dirname_length = 0;
1521 t->d = INVALID_HANDLE_VALUE;
1522 t->symlink_mode = t->initial_symlink_mode;
1523 archive_string_empty(&(t->full_path));
1524 archive_string_empty(&t->path);
1525 t->entry_fh = INVALID_HANDLE_VALUE;
1527 t->entry_remaining_bytes = 0;
1528 t->initial_filesystem_id = -1;
1530 /* Get wchar_t strings from char strings. */
1531 archive_string_init(&ws);
1532 archive_wstrcpy(&ws, path);
1534 /* Get a full-path-name. */
1535 p = __la_win_permissive_name_w(pathname);
1538 archive_wstrcpy(&(t->full_path), p);
1541 /* Convert path separators from '\' to '/' */
1542 for (p = pathname; *p != L'\0'; ++p) {
1548 /* First item is set up a lot like a symlink traversal. */
1549 /* printf("Looking for wildcard in %s\n", path); */
1550 if ((base[0] == L'/' && base[1] == L'/' &&
1551 base[2] == L'?' && base[3] == L'/' &&
1552 (wcschr(base+4, L'*') || wcschr(base+4, L'?'))) ||
1553 (!(base[0] == L'/' && base[1] == L'/' &&
1554 base[2] == L'?' && base[3] == L'/') &&
1555 (wcschr(base, L'*') || wcschr(base, L'?')))) {
1556 // It has a wildcard in it...
1557 // Separate the last element.
1558 p = wcsrchr(base, L'/');
1561 tree_append(t, base, p - base);
1562 t->dirname_length = archive_strlen(&t->path);
1565 p = wcsrchr(t->full_path.s, L'\\');
1568 t->full_path.length = wcslen(t->full_path.s);
1569 t->full_path_dir_length = archive_strlen(&t->full_path);
1572 tree_push(t, base, t->full_path.s, 0, 0, 0, NULL);
1573 archive_wstring_free(&ws);
1574 t->stack->flags = needsFirstVisit;
1576 * Debug flag for Direct IO(No buffering) or Async IO.
1577 * Those dependent on environment variable switches
1578 * will be removed until next release.
1582 if ((e = getenv("LIBARCHIVE_DIRECT_IO")) != NULL) {
1587 fprintf(stderr, "LIBARCHIVE_DIRECT_IO=%s\n",
1588 (t->direct_io)?"Enabled":"Disabled");
1590 t->direct_io = DIRECT_IO;
1591 if ((e = getenv("LIBARCHIVE_ASYNC_IO")) != NULL) {
1596 fprintf(stderr, "LIBARCHIVE_ASYNC_IO=%s\n",
1597 (t->async_io)?"Enabled":"Disabled");
1599 t->async_io = ASYNC_IO;
1603 archive_wstring_free(&ws);
1609 tree_descent(struct tree *t)
1611 t->dirname_length = archive_strlen(&t->path);
1612 t->full_path_dir_length = archive_strlen(&t->full_path);
1618 * We've finished a directory; ascend back to the parent.
1621 tree_ascend(struct tree *t)
1623 struct tree_entry *te;
1627 close_and_restore_time(INVALID_HANDLE_VALUE, t, &te->restore_time);
1632 * Pop the working stack.
1635 tree_pop(struct tree *t)
1637 struct tree_entry *te;
1639 t->full_path.s[t->full_path_dir_length] = L'\0';
1640 t->full_path.length = t->full_path_dir_length;
1641 t->path.s[t->dirname_length] = L'\0';
1642 t->path.length = t->dirname_length;
1643 if (t->stack == t->current && t->current != NULL)
1644 t->current = t->current->parent;
1646 t->stack = te->next;
1647 t->dirname_length = te->dirname_length;
1648 t->basename = t->path.s + t->dirname_length;
1649 t->full_path_dir_length = te->full_path_dir_length;
1650 while (t->basename[0] == L'/')
1652 archive_wstring_free(&te->name);
1653 archive_wstring_free(&te->full_path);
1658 * Get the next item in the tree traversal.
1661 tree_next(struct tree *t)
1665 while (t->stack != NULL) {
1666 /* If there's an open dir, get the next entry from there. */
1667 if (t->d != INVALID_HANDLE_VALUE) {
1668 r = tree_dir_next_windows(t, NULL);
1674 if (t->stack->flags & needsFirstVisit) {
1675 wchar_t *d = t->stack->name.s;
1676 t->stack->flags &= ~needsFirstVisit;
1677 if (!(d[0] == L'/' && d[1] == L'/' &&
1678 d[2] == L'?' && d[3] == L'/') &&
1679 (wcschr(d, L'*') || wcschr(d, L'?'))) {
1680 r = tree_dir_next_windows(t, d);
1685 HANDLE h = FindFirstFileW(d, &t->_findData);
1686 if (h == INVALID_HANDLE_VALUE) {
1687 la_dosmaperr(GetLastError());
1688 t->tree_errno = errno;
1689 t->visit_type = TREE_ERROR_DIR;
1690 return (t->visit_type);
1692 t->findData = &t->_findData;
1695 /* Top stack item needs a regular visit. */
1696 t->current = t->stack;
1697 tree_append(t, t->stack->name.s,
1698 archive_strlen(&(t->stack->name)));
1699 //t->dirname_length = t->path_length;
1701 t->stack->flags &= ~needsFirstVisit;
1702 return (t->visit_type = TREE_REGULAR);
1703 } else if (t->stack->flags & needsDescent) {
1704 /* Top stack item is dir to descend into. */
1705 t->current = t->stack;
1706 tree_append(t, t->stack->name.s,
1707 archive_strlen(&(t->stack->name)));
1708 t->stack->flags &= ~needsDescent;
1709 r = tree_descent(t);
1714 t->visit_type = TREE_POSTDESCENT;
1715 return (t->visit_type);
1716 } else if (t->stack->flags & needsOpen) {
1717 t->stack->flags &= ~needsOpen;
1718 r = tree_dir_next_windows(t, L"*");
1722 } else if (t->stack->flags & needsAscent) {
1723 /* Top stack item is dir and we're done with it. */
1726 t->visit_type = r != 0 ? r : TREE_POSTASCENT;
1727 return (t->visit_type);
1729 /* Top item on stack is dead. */
1731 t->flags &= ~hasLstat;
1732 t->flags &= ~hasStat;
1735 return (t->visit_type = 0);
1739 tree_dir_next_windows(struct tree *t, const wchar_t *pattern)
1741 const wchar_t *name;
1746 if (pattern != NULL) {
1747 struct archive_wstring pt;
1749 archive_string_init(&pt);
1750 archive_wstring_ensure(&pt,
1751 archive_strlen(&(t->full_path))
1752 + 2 + wcslen(pattern));
1753 archive_wstring_copy(&pt, &(t->full_path));
1754 archive_wstrappend_wchar(&pt, L'\\');
1755 archive_wstrcat(&pt, pattern);
1756 t->d = FindFirstFileW(pt.s, &t->_findData);
1757 archive_wstring_free(&pt);
1758 if (t->d == INVALID_HANDLE_VALUE) {
1759 la_dosmaperr(GetLastError());
1760 t->tree_errno = errno;
1761 r = tree_ascend(t); /* Undo "chdir" */
1763 t->visit_type = r != 0 ? r : TREE_ERROR_DIR;
1764 return (t->visit_type);
1766 t->findData = &t->_findData;
1768 } else if (!FindNextFileW(t->d, &t->_findData)) {
1770 t->d = INVALID_HANDLE_VALUE;
1774 name = t->findData->cFileName;
1775 namelen = wcslen(name);
1776 t->flags &= ~hasLstat;
1777 t->flags &= ~hasStat;
1778 if (name[0] == L'.' && name[1] == L'\0')
1780 if (name[0] == L'.' && name[1] == L'.' && name[2] == L'\0')
1782 tree_append(t, name, namelen);
1783 return (t->visit_type = TREE_REGULAR);
1787 #define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000)
1789 fileTimeToUtc(const FILETIME *filetime, time_t *t, long *ns)
1793 utc.HighPart = filetime->dwHighDateTime;
1794 utc.LowPart = filetime->dwLowDateTime;
1795 if (utc.QuadPart >= EPOC_TIME) {
1796 utc.QuadPart -= EPOC_TIME;
1797 /* milli seconds base */
1798 *t = (time_t)(utc.QuadPart / 10000000);
1799 /* nano seconds base */
1800 *ns = (long)(utc.QuadPart % 10000000) * 100;
1808 entry_copy_bhfi(struct archive_entry *entry, const wchar_t *path,
1809 const WIN32_FIND_DATAW *findData,
1810 const BY_HANDLE_FILE_INFORMATION *bhfi)
1816 fileTimeToUtc(&bhfi->ftLastAccessTime, &secs, &nsecs);
1817 archive_entry_set_atime(entry, secs, nsecs);
1818 fileTimeToUtc(&bhfi->ftLastWriteTime, &secs, &nsecs);
1819 archive_entry_set_mtime(entry, secs, nsecs);
1820 fileTimeToUtc(&bhfi->ftCreationTime, &secs, &nsecs);
1821 archive_entry_set_birthtime(entry, secs, nsecs);
1822 archive_entry_set_ctime(entry, secs, nsecs);
1823 archive_entry_set_dev(entry, bhfi_dev(bhfi));
1824 archive_entry_set_ino64(entry, bhfi_ino(bhfi));
1825 if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
1826 archive_entry_set_nlink(entry, bhfi->nNumberOfLinks + 1);
1828 archive_entry_set_nlink(entry, bhfi->nNumberOfLinks);
1829 archive_entry_set_size(entry,
1830 (((int64_t)bhfi->nFileSizeHigh) << 32)
1831 + bhfi->nFileSizeLow);
1832 archive_entry_set_uid(entry, 0);
1833 archive_entry_set_gid(entry, 0);
1834 archive_entry_set_rdev(entry, 0);
1836 mode = S_IRUSR | S_IRGRP | S_IROTH;
1837 if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_READONLY) == 0)
1838 mode |= S_IWUSR | S_IWGRP | S_IWOTH;
1839 if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) &&
1841 findData->dwReserved0 == IO_REPARSE_TAG_SYMLINK)
1843 else if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
1844 mode |= S_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH;
1849 p = wcsrchr(path, L'.');
1850 if (p != NULL && wcslen(p) == 4) {
1852 case L'B': case L'b':
1853 if ((p[2] == L'A' || p[2] == L'a' ) &&
1854 (p[3] == L'T' || p[3] == L't' ))
1855 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1857 case L'C': case L'c':
1858 if (((p[2] == L'M' || p[2] == L'm' ) &&
1859 (p[3] == L'D' || p[3] == L'd' )))
1860 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1862 case L'E': case L'e':
1863 if ((p[2] == L'X' || p[2] == L'x' ) &&
1864 (p[3] == L'E' || p[3] == L'e' ))
1865 mode |= S_IXUSR | S_IXGRP | S_IXOTH;
1872 archive_entry_set_mode(entry, mode);
1876 tree_archive_entry_copy_bhfi(struct archive_entry *entry, struct tree *t,
1877 const BY_HANDLE_FILE_INFORMATION *bhfi)
1879 entry_copy_bhfi(entry, tree_current_path(t), t->findData, bhfi);
1883 tree_current_file_information(struct tree *t, BY_HANDLE_FILE_INFORMATION *st,
1888 DWORD flag = FILE_FLAG_BACKUP_SEMANTICS;
1890 if (sim_lstat && tree_current_is_physical_link(t))
1891 flag |= FILE_FLAG_OPEN_REPARSE_POINT;
1892 h = CreateFileW(tree_current_access_path(t), 0, FILE_SHARE_READ, NULL,
1893 OPEN_EXISTING, flag, NULL);
1894 if (h == INVALID_HANDLE_VALUE) {
1895 la_dosmaperr(GetLastError());
1896 t->tree_errno = errno;
1899 r = GetFileInformationByHandle(h, st);
1905 * Get the stat() data for the entry just returned from tree_next().
1907 static const BY_HANDLE_FILE_INFORMATION *
1908 tree_current_stat(struct tree *t)
1910 if (!(t->flags & hasStat)) {
1911 if (!tree_current_file_information(t, &t->st, 0))
1913 t->flags |= hasStat;
1919 * Get the lstat() data for the entry just returned from tree_next().
1921 static const BY_HANDLE_FILE_INFORMATION *
1922 tree_current_lstat(struct tree *t)
1924 if (!(t->flags & hasLstat)) {
1925 if (!tree_current_file_information(t, &t->lst, 1))
1927 t->flags |= hasLstat;
1933 * Test whether current entry is a dir or link to a dir.
1936 tree_current_is_dir(struct tree *t)
1939 return (t->findData->dwFileAttributes
1940 & FILE_ATTRIBUTE_DIRECTORY);
1945 * Test whether current entry is a physical directory. Usually, we
1946 * already have at least one of stat() or lstat() in memory, so we
1947 * use tricks to try to avoid an extra trip to the disk.
1950 tree_current_is_physical_dir(struct tree *t)
1952 if (tree_current_is_physical_link(t))
1954 return (tree_current_is_dir(t));
1958 * Test whether current entry is a symbolic link.
1961 tree_current_is_physical_link(struct tree *t)
1964 return ((t->findData->dwFileAttributes
1965 & FILE_ATTRIBUTE_REPARSE_POINT) &&
1966 (t->findData->dwReserved0
1967 == IO_REPARSE_TAG_SYMLINK));
1972 * Test whether the same file has been in the tree as its parent.
1975 tree_target_is_same_as_parent(struct tree *t,
1976 const BY_HANDLE_FILE_INFORMATION *st)
1978 struct tree_entry *te;
1979 int64_t dev = bhfi_dev(st);
1980 int64_t ino = bhfi_ino(st);
1982 for (te = t->current->parent; te != NULL; te = te->parent) {
1983 if (te->dev == dev && te->ino == ino)
1990 * Return the access path for the entry just returned from tree_next().
1992 static const wchar_t *
1993 tree_current_access_path(struct tree *t)
1995 return (t->full_path.s);
1999 * Return the full path for the entry just returned from tree_next().
2001 static const wchar_t *
2002 tree_current_path(struct tree *t)
2008 * Terminate the traversal.
2011 tree_close(struct tree *t)
2016 if (t->entry_fh != INVALID_HANDLE_VALUE) {
2018 close_and_restore_time(t->entry_fh, t, &t->restore_time);
2019 t->entry_fh = INVALID_HANDLE_VALUE;
2021 /* Close the handle of FindFirstFileW */
2022 if (t->d != INVALID_HANDLE_VALUE) {
2024 t->d = INVALID_HANDLE_VALUE;
2027 /* Release anything remaining in the stack. */
2028 while (t->stack != NULL)
2033 * Release any resources.
2036 tree_free(struct tree *t)
2042 archive_wstring_free(&t->path);
2043 archive_wstring_free(&t->full_path);
2044 free(t->sparse_list);
2045 free(t->filesystem_table);
2046 for (i = 0; i < MAX_OVERLAPPED; i++) {
2048 VirtualFree(t->ol[i].buff, 0, MEM_RELEASE);
2049 CloseHandle(t->ol[i].ol.hEvent);
2056 * Populate the archive_entry with metadata from the disk.
2059 archive_read_disk_entry_from_file(struct archive *_a,
2060 struct archive_entry *entry, int fd, const struct stat *st)
2062 struct archive_read_disk *a = (struct archive_read_disk *)_a;
2063 const wchar_t *path;
2064 const wchar_t *wname;
2067 BY_HANDLE_FILE_INFORMATION bhfi;
2068 DWORD fileAttributes = 0;
2071 archive_clear_error(_a);
2072 wname = archive_entry_sourcepath_w(entry);
2074 wname = archive_entry_pathname_w(entry);
2075 if (wname == NULL) {
2076 archive_set_error(&a->archive, EINVAL,
2077 "Can't get a wide character version of the path");
2078 return (ARCHIVE_FAILED);
2080 path = __la_win_permissive_name_w(wname);
2084 * Get metadata through GetFileInformationByHandle().
2087 h = (HANDLE)_get_osfhandle(fd);
2088 r = GetFileInformationByHandle(h, &bhfi);
2090 la_dosmaperr(GetLastError());
2091 archive_set_error(&a->archive, errno,
2092 "Can't GetFileInformationByHandle");
2093 return (ARCHIVE_FAILED);
2095 entry_copy_bhfi(entry, path, NULL, &bhfi);
2097 WIN32_FIND_DATAW findData;
2098 DWORD flag, desiredAccess;
2100 h = FindFirstFileW(path, &findData);
2101 if (h == INVALID_HANDLE_VALUE) {
2102 la_dosmaperr(GetLastError());
2103 archive_set_error(&a->archive, errno,
2104 "Can't FindFirstFileW");
2105 return (ARCHIVE_FAILED);
2109 flag = FILE_FLAG_BACKUP_SEMANTICS;
2110 if (!a->follow_symlinks &&
2111 (findData.dwFileAttributes
2112 & FILE_ATTRIBUTE_REPARSE_POINT) &&
2113 (findData.dwReserved0 == IO_REPARSE_TAG_SYMLINK)) {
2114 flag |= FILE_FLAG_OPEN_REPARSE_POINT;
2116 } else if (findData.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {
2119 desiredAccess = GENERIC_READ;
2121 h = CreateFileW(path, desiredAccess, FILE_SHARE_READ, NULL,
2122 OPEN_EXISTING, flag, NULL);
2123 if (h == INVALID_HANDLE_VALUE) {
2124 la_dosmaperr(GetLastError());
2125 archive_set_error(&a->archive, errno,
2126 "Can't CreateFileW");
2127 return (ARCHIVE_FAILED);
2129 r = GetFileInformationByHandle(h, &bhfi);
2131 la_dosmaperr(GetLastError());
2132 archive_set_error(&a->archive, errno,
2133 "Can't GetFileInformationByHandle");
2135 return (ARCHIVE_FAILED);
2137 entry_copy_bhfi(entry, path, &findData, &bhfi);
2139 fileAttributes = bhfi.dwFileAttributes;
2141 archive_entry_copy_stat(entry, st);
2142 h = INVALID_HANDLE_VALUE;
2145 /* Lookup uname/gname */
2146 name = archive_read_disk_uname(_a, archive_entry_uid(entry));
2148 archive_entry_copy_uname(entry, name);
2149 name = archive_read_disk_gname(_a, archive_entry_gid(entry));
2151 archive_entry_copy_gname(entry, name);
2154 * Can this file be sparse file ?
2156 if (archive_entry_filetype(entry) != AE_IFREG
2157 || archive_entry_size(entry) <= 0
2158 || archive_entry_hardlink(entry) != NULL) {
2159 if (h != INVALID_HANDLE_VALUE && fd < 0)
2161 return (ARCHIVE_OK);
2164 if (h == INVALID_HANDLE_VALUE) {
2166 h = (HANDLE)_get_osfhandle(fd);
2168 h = CreateFileW(path, GENERIC_READ, FILE_SHARE_READ, NULL,
2169 OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
2170 if (h == INVALID_HANDLE_VALUE) {
2171 la_dosmaperr(GetLastError());
2172 archive_set_error(&a->archive, errno,
2173 "Can't CreateFileW");
2174 return (ARCHIVE_FAILED);
2177 r = GetFileInformationByHandle(h, &bhfi);
2179 la_dosmaperr(GetLastError());
2180 archive_set_error(&a->archive, errno,
2181 "Can't GetFileInformationByHandle");
2182 if (h != INVALID_HANDLE_VALUE && fd < 0)
2184 return (ARCHIVE_FAILED);
2186 fileAttributes = bhfi.dwFileAttributes;
2189 /* Sparse file must be set a mark, FILE_ATTRIBUTE_SPARSE_FILE */
2190 if ((fileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) == 0) {
2193 return (ARCHIVE_OK);
2196 r = setup_sparse_from_disk(a, entry, h);
2204 * Windows sparse interface.
2206 #if defined(__MINGW32__) && !defined(FSCTL_QUERY_ALLOCATED_RANGES)
2207 #define FSCTL_QUERY_ALLOCATED_RANGES 0x940CF
2209 LARGE_INTEGER FileOffset;
2210 LARGE_INTEGER Length;
2211 } FILE_ALLOCATED_RANGE_BUFFER;
2215 setup_sparse_from_disk(struct archive_read_disk *a,
2216 struct archive_entry *entry, HANDLE handle)
2218 FILE_ALLOCATED_RANGE_BUFFER range, *outranges = NULL;
2219 size_t outranges_size;
2220 int64_t entry_size = archive_entry_size(entry);
2221 int exit_sts = ARCHIVE_OK;
2223 range.FileOffset.QuadPart = 0;
2224 range.Length.QuadPart = entry_size;
2225 outranges_size = 2048;
2226 outranges = (FILE_ALLOCATED_RANGE_BUFFER *)malloc(outranges_size);
2227 if (outranges == NULL) {
2228 archive_set_error(&a->archive, ENOMEM,
2229 "Couldn't allocate memory");
2230 exit_sts = ARCHIVE_FATAL;
2231 goto exit_setup_sparse;
2239 ret = DeviceIoControl(handle,
2240 FSCTL_QUERY_ALLOCATED_RANGES,
2241 &range, sizeof(range), outranges,
2242 (DWORD)outranges_size, &retbytes, NULL);
2243 if (ret == 0 && GetLastError() == ERROR_MORE_DATA) {
2245 outranges_size *= 2;
2246 outranges = (FILE_ALLOCATED_RANGE_BUFFER *)
2247 malloc(outranges_size);
2248 if (outranges == NULL) {
2249 archive_set_error(&a->archive, ENOMEM,
2250 "Couldn't allocate memory");
2251 exit_sts = ARCHIVE_FATAL;
2252 goto exit_setup_sparse;
2262 n = retbytes / sizeof(outranges[0]);
2264 outranges[0].FileOffset.QuadPart == 0 &&
2265 outranges[0].Length.QuadPart == entry_size)
2266 break;/* This is not sparse. */
2267 for (i = 0; i < n; i++)
2268 archive_entry_sparse_add_entry(entry,
2269 outranges[i].FileOffset.QuadPart,
2270 outranges[i].Length.QuadPart);
2271 range.FileOffset.QuadPart =
2272 outranges[n-1].FileOffset.QuadPart
2273 + outranges[n-1].Length.QuadPart;
2274 range.Length.QuadPart =
2275 entry_size - range.FileOffset.QuadPart;
2276 if (range.Length.QuadPart > 0)
2279 /* The entire file is a hole. Add one data block of size 0 at the end. */
2280 archive_entry_sparse_add_entry(entry,
2286 la_dosmaperr(GetLastError());
2287 archive_set_error(&a->archive, errno,
2288 "DeviceIoControl Failed: %lu", GetLastError());
2289 exit_sts = ARCHIVE_FAILED;
2290 goto exit_setup_sparse;