4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012, 2013 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
13 * wimlib is free software; you can redistribute it and/or modify it under the
14 * terms of the GNU General Public License as published by the Free
15 * Software Foundation; either version 3 of the License, or (at your option)
18 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20 * A PARTICULAR PURPOSE. See the GNU General Public License for more
23 * You should have received a copy of the GNU General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
28 * This file provides the API functions wimlib_extract_image(),
29 * wimlib_extract_files(), and wimlib_extract_image_from_pipe(). Internally,
30 * all end up calling extract_tree() zero or more times to extract a tree of
31 * files from the currently selected WIM image to the specified target directory
34 * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
35 * Win32), this file does not itself have code to extract files or directories
36 * to any specific target; instead, it handles generic functionality and relies
37 * on lower-level callback functions declared in `struct apply_operations' to do
38 * the actual extraction.
45 #include "wimlib/apply.h"
46 #include "wimlib/dentry.h"
47 #include "wimlib/encoding.h"
48 #include "wimlib/endianness.h"
49 #include "wimlib/error.h"
50 #include "wimlib/lookup_table.h"
51 #include "wimlib/metadata.h"
52 #include "wimlib/paths.h"
53 #include "wimlib/reparse.h"
54 #include "wimlib/resource.h"
55 #include "wimlib/security.h"
56 #include "wimlib/swm.h"
58 # include "wimlib/win32.h" /* for realpath() equivalent */
60 #include "wimlib/xml.h"
61 #include "wimlib/wim.h"
69 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000
71 #define WIMLIB_EXTRACT_MASK_PUBLIC 0x3fffffff
73 /* Given a WIM dentry in the tree to be extracted, resolve all streams in the
74 * corresponding inode and set 'out_refcnt' in each to 0. */
76 dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_ctx)
78 struct apply_ctx *ctx = _ctx;
79 struct wim_inode *inode = dentry->d_inode;
80 struct wim_lookup_table_entry *lte;
84 if (dentry->extraction_skipped)
87 /* Special case: when extracting from a pipe, the WIM lookup table is
88 * initially empty, so "resolving" an inode's streams is initially not
89 * possible. However, we still need to keep track of which streams,
90 * identified by SHA1 message digests, need to be extracted, so we
91 * "resolve" the inode's streams anyway by allocating new entries. */
92 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
94 ret = inode_resolve_ltes(inode, ctx->wim->lookup_table, force);
97 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
98 lte = inode_stream_lte_resolved(inode, i);
106 is_linked_extraction(const struct apply_ctx *ctx)
108 return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
109 WIMLIB_EXTRACT_FLAG_SYMLINK));
113 can_extract_named_data_streams(const struct apply_ctx *ctx)
115 return ctx->supported_features.named_data_streams &&
116 !is_linked_extraction(ctx);
120 ref_stream_to_extract(struct wim_lookup_table_entry *lte,
121 struct wim_dentry *dentry, struct apply_ctx *ctx)
126 if (likely(!is_linked_extraction(ctx)) || (lte->out_refcnt == 0 &&
127 lte->extracted_file == NULL))
129 ctx->progress.extract.total_bytes += wim_resource_size(lte);
130 ctx->progress.extract.num_streams++;
133 if (lte->out_refcnt == 0) {
134 list_add_tail(<e->extraction_list, &ctx->stream_list);
135 ctx->num_streams_remaining++;
138 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
139 struct wim_dentry **lte_dentries;
141 /* Append dentry to this stream's array of dentries referencing
142 * it. Use inline array to avoid memory allocation until the
143 * number of dentries becomes too large. */
144 if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) {
145 lte_dentries = lte->inline_lte_dentries;
147 struct wim_dentry **prev_lte_dentries;
148 size_t alloc_lte_dentries;
150 if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) {
151 prev_lte_dentries = NULL;
152 alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries);
154 prev_lte_dentries = lte->lte_dentries;
155 alloc_lte_dentries = lte->alloc_lte_dentries;
158 if (lte->out_refcnt == alloc_lte_dentries) {
159 alloc_lte_dentries *= 2;
160 lte_dentries = REALLOC(prev_lte_dentries,
162 sizeof(lte_dentries[0]));
164 return WIMLIB_ERR_NOMEM;
165 if (prev_lte_dentries == NULL) {
167 lte->inline_lte_dentries,
168 sizeof(lte->inline_lte_dentries));
170 lte->lte_dentries = lte_dentries;
171 lte->alloc_lte_dentries = alloc_lte_dentries;
173 lte_dentries = lte->lte_dentries;
175 lte_dentries[lte->out_refcnt] = dentry;
181 /* Given a WIM dentry in the tree to be extracted, iterate through streams that
182 * need to be extracted. For each one, add it to the list of streams to be
183 * extracted (ctx->stream_list) if not already done so, and also update the
184 * progress information (ctx->progress) with the stream. Furthermore, if doing
185 * a sequential extraction, build a mapping from each the stream to the dentries
188 dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx)
190 struct apply_ctx *ctx = _ctx;
191 struct wim_inode *inode = dentry->d_inode;
194 /* Don't process dentries marked as skipped. */
195 if (dentry->extraction_skipped)
198 /* Don't process additional hard links. */
199 if (inode->i_visited && ctx->supported_features.hard_links)
202 /* The unnamed data stream will always be extracted, except in an
204 if (!inode_is_encrypted_directory(inode)) {
205 ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode),
211 /* Named data streams will be extracted only if supported in the current
212 * extraction mode and volume, and to avoid complications, if not doing
213 * a linked extraction. */
214 if (can_extract_named_data_streams(ctx)) {
215 for (u16 i = 0; i < inode->i_num_ads; i++) {
216 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
218 ret = ref_stream_to_extract(inode->i_ads_entries[i].lte,
224 inode->i_visited = 1;
228 /* Inform library user of progress of stream extraction following the successful
229 * extraction of a copy of the stream specified by @lte. */
231 update_extract_progress(struct apply_ctx *ctx,
232 const struct wim_lookup_table_entry *lte)
234 wimlib_progress_func_t progress_func = ctx->progress_func;
235 union wimlib_progress_info *progress = &ctx->progress;
237 progress->extract.completed_bytes += wim_resource_size(lte);
239 progress->extract.completed_bytes >= ctx->next_progress)
241 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress);
242 if (progress->extract.completed_bytes >=
243 progress->extract.total_bytes)
245 ctx->next_progress = ~0ULL;
247 ctx->next_progress += progress->extract.total_bytes / 128;
248 if (ctx->next_progress > progress->extract.total_bytes)
249 ctx->next_progress = progress->extract.total_bytes;
255 /* Extract a symbolic link (not directly as reparse data), handling fixing up
256 * the target of absolute symbolic links and updating the extract progress.
258 * @inode must specify the WIM inode for a symbolic link or junction reparse
261 * @lte_override overrides the resource used as the reparse data for the
264 extract_symlink(const tchar *path, struct apply_ctx *ctx,
265 struct wim_inode *inode,
266 struct wim_lookup_table_entry *lte_override)
268 ssize_t bufsize = ctx->ops->path_max;
269 tchar target[bufsize];
275 /* If absolute symbolic link fixups requested, reserve space in the link
276 * target buffer for the absolute path of the target directory. */
277 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
279 buf += ctx->realtarget_nchars;
280 bufsize -= ctx->realtarget_nchars;
283 /* Translate the WIM inode's reparse data into the link target. */
284 sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override);
287 return WIMLIB_ERR_READLINK;
291 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
294 /* Fix absolute symbolic link target to point into the
295 * actual extraction destination. */
296 tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars);
297 fixed_target = target;
299 /* Keep same link target. */
303 /* Call into the apply_operations to create the symbolic link. */
304 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
306 ret = ctx->ops->create_symlink(fixed_target, path, ctx);
308 ERROR_WITH_ERRNO("Failed to create symlink "
309 "\"%"TS"\" => \"%"TS"\"", path, fixed_target);
313 /* Account for reparse data consumed. */
314 update_extract_progress(ctx,
315 (lte_override ? lte_override :
316 inode_unnamed_lte_resolved(inode)));
319 #endif /* !__WIN32__ */
321 /* Create a file, directory, or symbolic link. */
323 extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode)
328 if (ctx->supported_features.symlink_reparse_points &&
329 !ctx->supported_features.reparse_points &&
330 inode_is_symlink(inode))
332 ret = extract_symlink(path, ctx, inode, NULL);
334 #endif /* !__WIN32__ */
335 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
336 ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie);
338 ERROR_WITH_ERRNO("Failed to create the directory "
342 ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie);
344 ERROR_WITH_ERRNO("Failed to create the file "
352 extract_hardlink(const tchar *oldpath, const tchar *newpath,
353 struct apply_ctx *ctx)
357 DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath);
358 ret = ctx->ops->create_hardlink(oldpath, newpath, ctx);
360 ERROR_WITH_ERRNO("Failed to create hardlink "
361 "\"%"TS"\" => \"%"TS"\"",
369 try_extract_rpfix(u8 *rpbuf,
371 const wchar_t *extract_root_realpath,
372 unsigned extract_root_realpath_nchars)
374 struct reparse_data rpdata;
376 size_t target_nchars;
377 size_t stripped_nchars;
378 wchar_t *stripped_target;
379 wchar_t stripped_target_nchars;
382 utf16lechar *new_target;
383 utf16lechar *new_print_name;
384 size_t new_target_nchars;
385 size_t new_print_name_nchars;
388 ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata);
392 if (extract_root_realpath[0] == L'\0' ||
393 extract_root_realpath[1] != L':' ||
394 extract_root_realpath[2] != L'\\')
395 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
397 ret = parse_substitute_name(rpdata.substitute_name,
398 rpdata.substitute_name_nbytes,
402 stripped_nchars = ret;
403 target = rpdata.substitute_name;
404 target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar);
405 stripped_target = target + stripped_nchars;
406 stripped_target_nchars = target_nchars - stripped_nchars;
408 new_target = alloca((6 + extract_root_realpath_nchars +
409 stripped_target_nchars) * sizeof(utf16lechar));
412 if (stripped_nchars == 6) {
413 /* Include \??\ prefix if it was present before */
414 p = wmempcpy(p, L"\\??\\", 4);
417 /* Print name excludes the \??\ if present. */
419 if (stripped_nchars != 0) {
420 /* Get drive letter from real path to extract root, if a drive
421 * letter was present before. */
422 *p++ = extract_root_realpath[0];
423 *p++ = extract_root_realpath[1];
425 /* Copy the rest of the extract root */
426 p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2);
428 /* Append the stripped target */
429 p = wmempcpy(p, stripped_target, stripped_target_nchars);
430 new_target_nchars = p - new_target;
431 new_print_name_nchars = p - new_print_name;
433 if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE ||
434 new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE)
435 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
437 rpdata.substitute_name = new_target;
438 rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar);
439 rpdata.print_name = new_print_name;
440 rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar);
441 return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p);
443 #endif /* __WIN32__ */
445 /* Set reparse data on extracted file or directory that has
446 * FILE_ATTRIBUTE_REPARSE_POINT set. */
448 extract_reparse_data(const tchar *path, struct apply_ctx *ctx,
449 struct wim_inode *inode,
450 struct wim_lookup_table_entry *lte_override)
453 u8 rpbuf[REPARSE_POINT_MAX_SIZE];
456 ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override);
461 /* Fix up target of absolute symbolic link or junction points so
462 * that they point into the actual extraction target. */
463 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
464 (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
465 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) &&
466 !inode->i_not_rpfixed)
468 ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget,
469 ctx->realtarget_nchars);
470 if (ret && !(ctx->extract_flags &
471 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
473 WARNING("Reparse point fixup of \"%"TS"\" "
482 ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx);
484 /* On Windows, the SeCreateSymbolicLink privilege is required to create
485 * symbolic links. To be more friendly towards non-Administrator users,
486 * we merely warn the user if symbolic links cannot be created due to
487 * insufficient permissions or privileges, unless
488 * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided. */
490 if (ret && inode_is_symlink(inode) &&
491 (errno == EACCES || errno == EPERM) &&
492 !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
494 WARNING("Can't set reparse data on \"%"TS"\": "
496 " You may be trying to "
497 "extract a symbolic link without the\n"
498 " SeCreateSymbolicLink privilege, "
499 "which by default non-Administrator\n"
500 " accounts do not have.",
508 /* Account for reparse data consumed. */
509 update_extract_progress(ctx,
510 (lte_override ? lte_override :
511 inode_unnamed_lte_resolved(inode)));
515 ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path);
520 * Extract zero or more streams to a file.
522 * This function operates slightly differently depending on whether @lte_spec is
523 * NULL or not. When @lte_spec is NULL, the behavior is to extract the default
524 * file contents (unnamed stream), and, if named data streams are supported in
525 * the extract mode and volume, any named data streams. When @lte_spec is NULL,
526 * the behavior is to extract only all copies of the stream @lte_spec, and in
527 * addition use @lte_spec to set the reparse data or create the symbolic link if
531 * Path to file to extract (as can be passed to apply_operations
536 * WIM dentry that corresponds to the file being extracted.
538 * If non-NULL, specifies the lookup table entry for a stream to extract,
539 * and only that stream will be extracted (although there may be more than
540 * one instance of it).
542 * Used only if @lte_spec != NULL; it is passed to the extraction functions
543 * rather than @lte_spec, allowing the location of the stream to be
544 * overridden. (This is used when the WIM is being read from a nonseekable
545 * file, such as a pipe, when streams need to be used more than once; each
546 * such stream is extracted to a temporary file.)
549 extract_streams(const tchar *path, struct apply_ctx *ctx,
550 struct wim_dentry *dentry,
551 struct wim_lookup_table_entry *lte_spec,
552 struct wim_lookup_table_entry *lte_override)
554 struct wim_inode *inode = dentry->d_inode;
555 struct wim_lookup_table_entry *lte;
556 file_spec_t file_spec;
559 if (dentry->was_hardlinked)
566 for (unsigned i = 0; i < SHA1_HASH_SIZE; i++)
567 p += sprintf(p, "%02x", lte_override->hash[i]);
568 DEBUG("Extracting stream SHA1=%s to \"%"TS"\"",
569 sha1_str, path, inode->i_ino);
571 DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino);
575 if (ctx->ops->uses_cookies)
576 file_spec.cookie = inode->extract_cookie;
578 file_spec.path = path;
580 /* Unnamed data stream. */
581 lte = inode_unnamed_lte_resolved(inode);
582 if (lte && (!lte_spec || lte == lte_spec)) {
585 if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
586 FILE_ATTRIBUTE_REPARSE_POINT)))
588 if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) &&
589 ctx->supported_features.encrypted_files)
590 ret = ctx->ops->extract_encrypted_stream(file_spec, lte, ctx);
592 ret = ctx->ops->extract_unnamed_stream(file_spec, lte, ctx);
595 update_extract_progress(ctx, lte);
597 else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
600 if (ctx->supported_features.reparse_points)
601 ret = extract_reparse_data(path, ctx, inode, lte);
603 else if ((inode_is_symlink(inode) &&
604 ctx->supported_features.symlink_reparse_points))
605 ret = extract_symlink(path, ctx, inode, lte);
612 /* Named data streams. */
613 if (can_extract_named_data_streams(ctx)) {
614 for (u16 i = 0; i < inode->i_num_ads; i++) {
615 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
617 if (!ads_entry_is_named_stream(entry))
622 if (lte_spec && lte_spec != lte)
626 ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name,
627 entry->stream_name_nbytes / 2,
631 update_extract_progress(ctx, lte);
637 ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path);
641 /* Set attributes on an extracted file or directory if supported by the
642 * extraction mode. */
644 extract_file_attributes(const tchar *path, struct apply_ctx *ctx,
645 struct wim_dentry *dentry)
649 if (ctx->ops->set_file_attributes &&
650 !(dentry == ctx->extract_root && ctx->root_dentry_is_special)) {
651 u32 attributes = dentry->d_inode->i_attributes;
653 /* Clear unsupported attributes. */
654 attributes &= ctx->supported_attributes_mask;
656 if ((attributes & FILE_ATTRIBUTE_DIRECTORY &&
657 !ctx->supported_features.encrypted_directories) ||
658 (!(attributes & FILE_ATTRIBUTE_DIRECTORY) &&
659 !ctx->supported_features.encrypted_files))
661 attributes &= ~FILE_ATTRIBUTE_ENCRYPTED;
665 attributes = FILE_ATTRIBUTE_NORMAL;
667 ret = ctx->ops->set_file_attributes(path, attributes, ctx);
669 ERROR_WITH_ERRNO("Failed to set attributes on "
678 /* Set or remove the short (DOS) name on an extracted file or directory if
679 * supported by the extraction mode. Since DOS names are unimportant and it's
680 * easy to run into problems setting them on Windows (SetFileShortName()
681 * requires SE_RESTORE privilege, which only the Administrator can request, and
682 * also requires DELETE access to the file), failure is ignored unless
683 * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set. */
685 extract_short_name(const tchar *path, struct apply_ctx *ctx,
686 struct wim_dentry *dentry)
690 /* The root of the dentry tree being extracted may not be extracted to
691 * its original name, so its short name should be ignored. */
692 if (dentry == ctx->extract_root)
695 if (ctx->supported_features.short_names) {
696 ret = ctx->ops->set_short_name(path,
698 dentry->short_name_nbytes / 2,
700 if (ret && (ctx->extract_flags &
701 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES))
703 ERROR_WITH_ERRNO("Failed to set short name of "
711 /* Set security descriptor, UNIX data, or neither on an extracted file, taking
712 * into account the current extraction mode and flags. */
714 extract_security(const tchar *path, struct apply_ctx *ctx,
715 struct wim_dentry *dentry)
718 struct wim_inode *inode = dentry->d_inode;
720 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS)
723 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
727 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
728 struct wimlib_unix_data data;
730 ret = inode_get_unix_data(inode, &data, NULL);
734 ret = ctx->ops->set_unix_data(path, &data, ctx);
736 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
737 ERROR_WITH_ERRNO("Failed to set UNIX owner, "
738 "group, and/or mode on "
742 WARNING_WITH_ERRNO("Failed to set UNIX owner, "
743 "group, and/or/mode on "
749 #endif /* __WIN32__ */
750 if (ctx->supported_features.security_descriptors &&
751 inode->i_security_id != -1)
753 const struct wim_security_data *sd;
757 sd = wim_const_security_data(ctx->wim);
758 desc = sd->descriptors[inode->i_security_id];
759 desc_size = sd->sizes[inode->i_security_id];
761 ret = ctx->ops->set_security_descriptor(path, desc,
764 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
765 ERROR_WITH_ERRNO("Failed to set security "
766 "descriptor on \"%"TS"\"", path);
769 if (errno != EACCES) {
770 WARNING_WITH_ERRNO("Failed to set "
771 "security descriptor "
772 "on \"%"TS"\"", path);
780 /* Set timestamps on an extracted file. Failure is warning-only unless
781 * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */
783 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
784 struct wim_dentry *dentry)
786 struct wim_inode *inode = dentry->d_inode;
789 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
792 if (ctx->ops->set_timestamps) {
793 ret = ctx->ops->set_timestamps(path,
794 inode->i_creation_time,
795 inode->i_last_write_time,
796 inode->i_last_access_time,
799 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
800 ERROR_WITH_ERRNO("Failed to set timestamps "
801 "on \"%"TS"\"", path);
804 WARNING_WITH_ERRNO("Failed to set timestamps "
805 "on \"%"TS"\"", path);
812 /* Check whether the extraction of a dentry should be skipped completely. */
814 dentry_is_supported(struct wim_dentry *dentry,
815 const struct wim_features *supported_features)
817 struct wim_inode *inode = dentry->d_inode;
819 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
820 return supported_features->reparse_points ||
821 (inode_is_symlink(inode) &&
822 supported_features->symlink_reparse_points);
824 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
825 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
826 return supported_features->encrypted_directories != 0;
828 return supported_features->encrypted_files != 0;
833 /* Given a WIM dentry to extract, build the path to which to extract it, in the
834 * format understood by the callbacks in the apply_operations being used.
836 * Write the resulting path into @path, which must have room for at least
837 * ctx->ops->max_path characters including the null-terminator.
839 * Return %true if successful; %false if this WIM dentry doesn't actually need
840 * to be extracted or if the calculated path exceeds ctx->ops->max_path
843 * This function clobbers the tmp_list member of @dentry and its ancestors up
844 * until the extraction root. */
846 build_extraction_path(tchar path[], struct wim_dentry *dentry,
847 struct apply_ctx *ctx)
850 LIST_HEAD(ancestor_list);
852 const tchar *target_prefix;
853 size_t target_prefix_nchars;
854 struct wim_dentry *d;
856 if (dentry->extraction_skipped)
859 path_nchars = ctx->ops->path_prefix_nchars;
861 if (ctx->ops->requires_realtarget_in_paths) {
862 target_prefix = ctx->realtarget;
863 target_prefix_nchars = ctx->realtarget_nchars;
864 } else if (ctx->ops->requires_target_in_paths) {
865 target_prefix = ctx->target;
866 target_prefix_nchars = ctx->target_nchars;
868 target_prefix = NULL;
869 target_prefix_nchars = 0;
871 path_nchars += target_prefix_nchars;
873 for (d = dentry; d != ctx->extract_root; d = d->parent) {
874 path_nchars += d->extraction_name_nchars + 1;
875 list_add(&d->tmp_list, &ancestor_list);
878 path_nchars++; /* null terminator */
880 if (path_nchars > ctx->ops->path_max) {
881 WARNING("\"%"TS"\": Path too long to extract",
882 dentry_full_path(dentry));
886 p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
887 p = tmempcpy(p, target_prefix, target_prefix_nchars);
888 list_for_each_entry(d, &ancestor_list, tmp_list) {
889 *p++ = ctx->ops->path_separator;
890 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
893 wimlib_assert(p - path == path_nchars);
898 get_num_path_components(const tchar *path, tchar path_separator)
900 unsigned num_components = 0;
902 /* Ignore drive letter. */
903 if (path[0] != L'\0' && path[1] == L':')
908 while (*path == path_separator)
912 while (*path && *path != path_separator)
915 return num_components;
919 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
920 struct apply_ctx *ctx, struct wim_dentry *dentry)
922 size_t num_raw_path_components;
923 const struct wim_dentry *d;
924 size_t num_target_path_components;
929 num_raw_path_components = 0;
930 for (d = dentry; d != ctx->extract_root; d = d->parent)
931 num_raw_path_components++;
933 if (ctx->ops->requires_realtarget_in_paths)
934 num_target_path_components = get_num_path_components(ctx->realtarget,
935 ctx->ops->path_separator);
936 else if (ctx->ops->requires_target_in_paths)
937 num_target_path_components = get_num_path_components(ctx->target,
938 ctx->ops->path_separator);
940 num_target_path_components = 0;
942 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
943 wimlib_assert(num_target_path_components > 0);
944 num_raw_path_components++;
945 num_target_path_components--;
948 p_old = oldpath + ctx->ops->path_prefix_nchars;
950 if (p_old[0] != L'\0' && p_old[1] == ':')
953 while (*p_old == ctx->ops->path_separator)
955 while (--num_target_path_components) {
956 while (*p_old != ctx->ops->path_separator)
958 while (*p_old == ctx->ops->path_separator)
962 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
964 p = &symlink_target[0];
965 while (num_raw_path_components--) {
968 *p++ = ctx->ops->path_separator;
971 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
972 newpath, symlink_target);
973 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
975 ERROR_WITH_ERRNO("Failed to create symlink "
976 "\"%"TS"\" => \"%"TS"\"",
977 newpath, symlink_target);
982 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
983 * data streams, reparse data (including symbolic links), timestamps, and
984 * security descriptors. Basically, everything that doesn't require reading
985 * non-metadata resources from the WIM file and isn't delayed until the final
988 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
989 struct apply_ctx *ctx)
991 struct wim_inode *inode = dentry->d_inode;
993 const tchar *oldpath;
995 if (unlikely(is_linked_extraction(ctx))) {
996 struct wim_lookup_table_entry *unnamed_lte;
998 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
999 if (unnamed_lte && unnamed_lte->extracted_file) {
1000 oldpath = unnamed_lte->extracted_file;
1001 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
1008 /* Create hard link if this dentry corresponds to an already-extracted
1010 if (inode->i_extracted_file) {
1011 oldpath = inode->i_extracted_file;
1015 /* Skip symlinks unless they can be extracted as reparse points rather
1016 * than created directly. */
1017 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
1020 /* Create this file or directory unless it's the extraction root, which
1021 * was already created if necessary. */
1022 if (dentry != ctx->extract_root) {
1023 ret = extract_inode(path, ctx, inode);
1028 /* Create empty named data streams. */
1029 if (can_extract_named_data_streams(ctx)) {
1030 for (u16 i = 0; i < inode->i_num_ads; i++) {
1031 file_spec_t file_spec;
1032 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1034 if (!ads_entry_is_named_stream(entry))
1038 if (ctx->ops->uses_cookies)
1039 file_spec.cookie = inode->extract_cookie;
1041 file_spec.path = path;
1042 ret = ctx->ops->extract_named_stream(file_spec,
1044 entry->stream_name_nbytes / 2,
1047 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1048 "empty named data stream",
1055 /* Set file attributes (if supported). */
1056 ret = extract_file_attributes(path, ctx, dentry);
1060 /* Set or remove file short name (if supported). */
1061 ret = extract_short_name(path, ctx, dentry);
1065 /* If inode has multiple links and hard links are supported in this
1066 * extraction mode and volume, save the path to the extracted file in
1067 * case it's needed to create a hard link. */
1068 if (unlikely(is_linked_extraction(ctx))) {
1069 struct wim_lookup_table_entry *unnamed_lte;
1071 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1073 unnamed_lte->extracted_file = TSTRDUP(path);
1074 if (!unnamed_lte->extracted_file)
1075 return WIMLIB_ERR_NOMEM;
1077 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1078 inode->i_extracted_file = TSTRDUP(path);
1079 if (!inode->i_extracted_file)
1080 return WIMLIB_ERR_NOMEM;
1085 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1088 dentry->was_hardlinked = 1;
1092 ret = extract_hardlink(oldpath, path, ctx);
1095 dentry->was_hardlinked = 1;
1100 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1102 struct apply_ctx *ctx = _ctx;
1103 tchar path[ctx->ops->path_max];
1104 struct wim_dentry *orig_dentry;
1105 struct wim_dentry *other_dentry;
1108 /* Here we may re-order the extraction of multiple names (hard links)
1109 * for the same file in the same directory in order to ensure the short
1110 * (DOS) name is set correctly. A short name is always associated with
1111 * exactly one long name, and at least on NTFS, only one long name for a
1112 * file can have a short name associated with it. (More specifically,
1113 * there can be unlimited names in the POSIX namespace, but only one
1114 * name can be in the Win32+DOS namespace, or one name in the Win32
1115 * namespace with a corresponding name in the DOS namespace.) To ensure
1116 * the short name of a file is associated with the correct long name in
1117 * a directory, we extract the long name with a corresponding short name
1118 * before any additional names. This can affect NTFS-3g extraction
1119 * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying
1120 * the long name to associate with a short name) and may affect Win32
1121 * extraction as well (which uses SetFileShortName()). */
1123 if (dentry->skeleton_extracted)
1126 if (ctx->supported_features.short_names
1127 && !dentry_has_short_name(dentry)
1128 && !dentry->d_inode->i_dos_name_extracted)
1130 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1131 if (dentry_has_short_name(other_dentry)
1132 && !other_dentry->skeleton_extracted
1133 && other_dentry->parent == dentry->parent)
1135 DEBUG("Creating %"TS" before %"TS" "
1136 "to guarantee correct DOS name extraction",
1137 dentry_full_path(other_dentry),
1138 dentry_full_path(dentry));
1139 orig_dentry = dentry;
1140 dentry = other_dentry;
1146 if (!build_extraction_path(path, dentry, ctx))
1148 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1152 dentry->skeleton_extracted = 1;
1155 dentry = orig_dentry;
1159 dentry->d_inode->i_dos_name_extracted = 1;
1163 /* Create a file or directory, then immediately extract all streams. This
1164 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1165 * may not be read sequentially by this function. */
1167 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1169 struct apply_ctx *ctx = _ctx;
1170 tchar path[ctx->ops->path_max];
1173 ret = dentry_extract_skeleton(dentry, ctx);
1177 if (!build_extraction_path(path, dentry, ctx))
1180 return extract_streams(path, ctx, dentry, NULL, NULL);
1183 /* Extract all instances of the stream @lte that are being extracted in this
1184 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1185 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1186 * extracted multiple times, it is extracted to a temporary file first.
1188 * This is intended for use with sequential extraction of a WIM image
1189 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1191 extract_stream_instances(struct wim_lookup_table_entry *lte,
1192 struct apply_ctx *ctx, bool can_seek)
1194 struct wim_dentry **lte_dentries;
1195 struct wim_lookup_table_entry *lte_tmp = NULL;
1196 struct wim_lookup_table_entry *lte_override;
1197 tchar *stream_tmp_filename = NULL;
1198 tchar path[ctx->ops->path_max];
1202 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1203 lte_dentries = lte->inline_lte_dentries;
1205 lte_dentries = lte->lte_dentries;
1207 if (likely(can_seek || lte->out_refcnt < 2)) {
1210 /* Need to extract stream to temporary file. */
1214 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1215 if (!stream_tmp_filename) {
1216 ERROR_WITH_ERRNO("Failed to create temporary filename");
1217 ret = WIMLIB_ERR_OPEN;
1221 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1223 ret = WIMLIB_ERR_NOMEM;
1224 goto out_free_stream_tmp_filename;
1226 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1227 lte_tmp->file_on_disk = stream_tmp_filename;
1228 lte_override = lte_tmp;
1230 raw_fd = topen(stream_tmp_filename,
1231 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1233 ERROR_WITH_ERRNO("Failed to open temporary file");
1234 ret = WIMLIB_ERR_OPEN;
1235 goto out_free_lte_tmp;
1237 filedes_init(&fd, raw_fd);
1238 ret = extract_wim_resource_to_fd(lte, &fd,
1239 wim_resource_size(lte));
1240 if (filedes_close(&fd) && !ret)
1241 ret = WIMLIB_ERR_WRITE;
1243 goto out_unlink_stream_tmp_file;
1246 /* Extract all instances of the stream, reading either from the stream
1247 * in the WIM file or from the temporary file containing the stream.
1248 * dentry->tmp_flag is used to ensure that each dentry is processed only
1249 * once regardless of how many times this stream appears in the streams
1250 * of the corresponding inode. */
1251 for (i = 0; i < lte->out_refcnt; i++) {
1252 struct wim_dentry *dentry = lte_dentries[i];
1254 if (dentry->tmp_flag)
1256 if (!build_extraction_path(path, dentry, ctx))
1258 ret = extract_streams(path, ctx, dentry,
1261 goto out_clear_tmp_flags;
1262 dentry->tmp_flag = 1;
1265 out_clear_tmp_flags:
1266 for (i = 0; i < lte->out_refcnt; i++)
1267 lte_dentries[i]->tmp_flag = 0;
1268 out_unlink_stream_tmp_file:
1269 if (stream_tmp_filename)
1270 tunlink(stream_tmp_filename);
1273 out_free_stream_tmp_filename:
1274 FREE(stream_tmp_filename);
1279 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1280 * structure and empty files were already created. This relies on the
1281 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1282 * stream that was constructed earlier. Streams are extracted exactly in the
1283 * order of the stream list; however, unless the WIM's file descriptor is
1284 * detected to be non-seekable, streams may be read from the WIM file more than
1285 * one time if multiple copies need to be extracted. */
1287 extract_stream_list(struct apply_ctx *ctx)
1289 struct wim_lookup_table_entry *lte;
1293 can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
1294 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1295 ret = extract_stream_instances(lte, ctx, can_seek);
1302 #define PWM_ALLOW_WIM_HDR 0x00001
1303 #define PWM_SILENT_EOF 0x00002
1305 /* Read the header from a stream in a pipable WIM. */
1307 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1308 int flags, struct wim_header_disk *hdr_ret)
1311 struct pwm_stream_hdr stream_hdr;
1312 struct wim_header_disk pwm_hdr;
1316 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
1320 if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
1321 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
1322 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
1323 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
1327 lte->resource_location = RESOURCE_NONEXISTENT;
1328 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
1332 if (buf.stream_hdr.magic != PWM_STREAM_MAGIC) {
1333 ERROR("Data read on pipe is invalid (expected stream header).");
1334 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1337 lte->resource_entry.original_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
1338 copy_hash(lte->hash, buf.stream_hdr.hash);
1339 lte->resource_entry.flags = le32_to_cpu(buf.stream_hdr.flags);
1340 lte->resource_entry.offset = pwm->in_fd.offset;
1341 lte->resource_location = RESOURCE_IN_WIM;
1343 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1344 lte->compression_type = pwm->compression_type;
1345 lte->resource_entry.size = 0;
1347 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1348 lte->resource_entry.size = lte->resource_entry.original_size;
1350 lte->is_pipable = 1;
1354 if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1355 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1359 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1361 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1363 return read_partial_wim_resource(lte, wim_resource_size(lte),
1365 WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY,
1370 extract_streams_from_pipe(struct apply_ctx *ctx)
1372 struct wim_lookup_table_entry *found_lte;
1373 struct wim_lookup_table_entry *needed_lte;
1374 struct wim_lookup_table *lookup_table;
1375 struct wim_header_disk pwm_hdr;
1379 ret = WIMLIB_ERR_NOMEM;
1380 found_lte = new_lookup_table_entry();
1384 lookup_table = ctx->wim->lookup_table;
1385 pwm_flags = PWM_ALLOW_WIM_HDR;
1386 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1387 pwm_flags |= PWM_SILENT_EOF;
1388 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
1389 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
1390 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
1391 if (ctx->progress_func)
1392 ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1394 while (ctx->num_streams_remaining) {
1395 ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags,
1398 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1399 (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1403 goto out_free_found_lte;
1406 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1407 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1408 && (needed_lte = __lookup_resource(lookup_table, found_lte->hash))
1409 && (needed_lte->out_refcnt))
1411 copy_resource_entry(&needed_lte->resource_entry,
1412 &found_lte->resource_entry);
1413 needed_lte->resource_location = found_lte->resource_location;
1414 needed_lte->wim = found_lte->wim;
1415 needed_lte->compression_type = found_lte->compression_type;
1416 needed_lte->is_pipable = found_lte->is_pipable;
1418 ret = extract_stream_instances(needed_lte, ctx, false);
1420 goto out_free_found_lte;
1421 ctx->num_streams_remaining--;
1422 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1423 ret = skip_pwm_stream(found_lte);
1425 goto out_free_found_lte;
1427 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
1428 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
1430 if (part_number != ctx->progress.extract.part_number ||
1431 total_parts != ctx->progress.extract.total_parts ||
1432 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
1435 ctx->progress.extract.part_number = part_number;
1436 ctx->progress.extract.total_parts = total_parts;
1437 memcpy(ctx->progress.extract.guid,
1438 pwm_hdr.guid, WIM_GID_LEN);
1439 if (ctx->progress_func) {
1441 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1450 free_lookup_table_entry(found_lte);
1459 /* Finish extracting a file, directory, or symbolic link by setting file
1460 * security and timestamps. */
1462 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1464 struct apply_ctx *ctx = _ctx;
1466 tchar path[ctx->ops->path_max];
1468 if (!build_extraction_path(path, dentry, ctx))
1471 ret = extract_security(path, ctx, dentry);
1475 return extract_timestamps(path, ctx, dentry);
1479 * Extract a WIM dentry to standard output.
1481 * This obviously doesn't make sense in all cases. We return an error if the
1482 * dentry does not correspond to a regular file. Otherwise we extract the
1483 * unnamed data stream only.
1486 extract_dentry_to_stdout(struct wim_dentry *dentry)
1489 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1490 FILE_ATTRIBUTE_DIRECTORY))
1492 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1493 "extracted to standard output", dentry_full_path(dentry));
1494 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1496 struct wim_lookup_table_entry *lte;
1498 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1500 struct filedes _stdout;
1501 filedes_init(&_stdout, STDOUT_FILENO);
1502 ret = extract_wim_resource_to_fd(lte, &_stdout,
1503 wim_resource_size(lte));
1510 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1512 static const utf16lechar replacement_char = cpu_to_le16('?');
1516 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1522 for (i = 0; i < num_chars; i++) {
1525 case cpu_to_le16('\\'):
1526 case cpu_to_le16(':'):
1527 case cpu_to_le16('*'):
1528 case cpu_to_le16('?'):
1529 case cpu_to_le16('"'):
1530 case cpu_to_le16('<'):
1531 case cpu_to_le16('>'):
1532 case cpu_to_le16('|'):
1534 case cpu_to_le16('/'):
1535 case cpu_to_le16('\0'):
1537 name[i] = replacement_char;
1544 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1545 name[num_chars - 1] == cpu_to_le16('.'))
1548 name[num_chars - 1] = replacement_char;
1557 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1559 const utf16lechar *file_name = dentry->file_name;
1560 return file_name != NULL &&
1561 file_name[0] == cpu_to_le16('.') &&
1562 (file_name[1] == cpu_to_le16('\0') ||
1563 (file_name[1] == cpu_to_le16('.') &&
1564 file_name[2] == cpu_to_le16('\0')));
1568 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1570 dentry->extraction_skipped = 1;
1575 * dentry_calculate_extraction_path-
1577 * Calculate the actual filename component at which a WIM dentry will be
1578 * extracted, handling invalid filenames "properly".
1580 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1581 * UNIX, converted into the platform's multibyte encoding). However, if the
1582 * file name contains characters that are not valid on the current platform or
1583 * has some other format that is not valid, leave dentry->extraction_name as
1584 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1585 * not be extracted, unless the appropriate flag
1586 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1587 * which case a substitute filename will be created and set instead.
1589 * Conflicts with case-insensitive names on Windows are handled similarly; see
1593 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1595 struct apply_ctx *ctx = _args;
1598 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1601 if (!dentry_is_supported(dentry, &ctx->supported_features))
1604 if (dentry_is_dot_or_dotdot(dentry)) {
1605 /* WIM files shouldn't contain . or .. entries. But if they are
1606 * there, don't attempt to extract them. */
1607 WARNING("Skipping extraction of unexpected . or .. file "
1608 "\"%"TS"\"", dentry_full_path(dentry));
1613 if (!ctx->ops->supports_case_sensitive_filenames)
1615 struct wim_dentry *other;
1616 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1617 case_insensitive_conflict_list)
1619 if (ctx->extract_flags &
1620 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1621 WARNING("\"%"TS"\" has the same "
1622 "case-insensitive name as "
1623 "\"%"TS"\"; extracting "
1624 "dummy name instead",
1625 dentry_full_path(dentry),
1626 dentry_full_path(other));
1629 WARNING("Not extracting \"%"TS"\": "
1630 "has same case-insensitive "
1631 "name as \"%"TS"\"",
1632 dentry_full_path(dentry),
1633 dentry_full_path(other));
1638 #else /* __WIN32__ */
1639 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1640 #endif /* !__WIN32__ */
1642 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1644 dentry->extraction_name = dentry->file_name;
1645 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1648 return utf16le_to_tstr(dentry->file_name,
1649 dentry->file_name_nbytes,
1650 &dentry->extraction_name,
1651 &dentry->extraction_name_nchars);
1654 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1656 WARNING("\"%"TS"\" has an invalid filename "
1657 "that is not supported on this platform; "
1658 "extracting dummy name instead",
1659 dentry_full_path(dentry));
1662 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1663 "that is not supported on this platform",
1664 dentry_full_path(dentry));
1671 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1673 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1674 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1677 size_t tchar_nchars;
1679 tchar_name = utf16_name_copy;
1680 tchar_nchars = dentry->file_name_nbytes / 2;
1682 ret = utf16le_to_tstr(utf16_name_copy,
1683 dentry->file_name_nbytes,
1684 &tchar_name, &tchar_nchars);
1688 size_t fixed_name_num_chars = tchar_nchars;
1689 tchar fixed_name[tchar_nchars + 50];
1691 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1692 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1693 T(" (invalid filename #%lu)"),
1694 ++ctx->invalid_sequence);
1698 dentry->extraction_name = memdup(fixed_name,
1699 2 * fixed_name_num_chars + 2);
1700 if (!dentry->extraction_name)
1701 return WIMLIB_ERR_NOMEM;
1702 dentry->extraction_name_nchars = fixed_name_num_chars;
1707 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1711 /* Clean up dentry and inode structure after extraction. */
1713 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1715 struct wim_inode *inode = dentry->d_inode;
1717 dentry->extraction_skipped = 0;
1718 dentry->was_hardlinked = 0;
1719 dentry->skeleton_extracted = 0;
1720 inode->i_visited = 0;
1721 FREE(inode->i_extracted_file);
1722 inode->i_extracted_file = NULL;
1723 inode->i_dos_name_extracted = 0;
1724 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1725 FREE(dentry->extraction_name);
1726 dentry->extraction_name = NULL;
1730 /* Tally features necessary to extract a dentry and the corresponding inode. */
1732 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1734 struct wim_features *features = _features;
1735 struct wim_inode *inode = dentry->d_inode;
1737 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1738 features->archive_files++;
1739 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1740 features->hidden_files++;
1741 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1742 features->system_files++;
1743 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1744 features->compressed_files++;
1745 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1746 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1747 features->encrypted_directories++;
1749 features->encrypted_files++;
1751 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1752 features->not_context_indexed_files++;
1753 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1754 features->sparse_files++;
1755 if (inode_has_named_stream(inode))
1756 features->named_data_streams++;
1757 if (inode->i_visited)
1758 features->hard_links++;
1759 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1760 features->reparse_points++;
1761 if (inode_is_symlink(inode))
1762 features->symlink_reparse_points++;
1764 features->other_reparse_points++;
1766 if (inode->i_security_id != -1)
1767 features->security_descriptors++;
1768 if (dentry->short_name_nbytes)
1769 features->short_names++;
1770 if (inode_has_unix_data(inode))
1771 features->unix_data++;
1772 inode->i_visited = 1;
1777 dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
1779 dentry->d_inode->i_visited = 0;
1783 /* Tally the features necessary to extract a dentry tree. */
1785 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1787 memset(features, 0, sizeof(struct wim_features));
1788 for_dentry_in_tree(root, dentry_tally_features, features);
1789 for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
1793 compute_supported_attributes_mask(const struct wim_features *supported_features)
1797 if (!supported_features->archive_files)
1798 mask &= ~FILE_ATTRIBUTE_ARCHIVE;
1800 if (!supported_features->hidden_files)
1801 mask &= ~FILE_ATTRIBUTE_HIDDEN;
1803 if (!supported_features->system_files)
1804 mask &= ~FILE_ATTRIBUTE_SYSTEM;
1806 if (!supported_features->not_context_indexed_files)
1807 mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED;
1809 if (!supported_features->compressed_files)
1810 mask &= ~FILE_ATTRIBUTE_COMPRESSED;
1812 if (!supported_features->sparse_files)
1813 mask &= ~FILE_ATTRIBUTE_SPARSE_FILE;
1815 if (!supported_features->reparse_points)
1816 mask &= ~FILE_ATTRIBUTE_REPARSE_POINT;
1822 do_feature_check(const struct wim_features *required_features,
1823 const struct wim_features *supported_features,
1825 const struct apply_operations *ops,
1826 const tchar *wim_source_path)
1829 const tchar *mode = T("this extraction mode");
1831 if (wim_source_path[0] == '\0')
1832 loc = T("the WIM image");
1834 loc = wim_source_path;
1836 /* We're an archive program, so theoretically we can do what we want
1837 * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't
1838 * bother the user about it. */
1840 if (required_features->archive_files && !supported_features->archive_files)
1843 "%lu files in %"TS" are marked as archived, but this attribute\n"
1844 " is not supported in %"TS".",
1845 required_features->archive_files, loc, mode);
1849 if (required_features->hidden_files && !supported_features->hidden_files)
1852 "%lu files in %"TS" are marked as hidden, but this\n"
1853 " attribute is not supported in %"TS".",
1854 required_features->hidden_files, loc, mode);
1857 if (required_features->system_files && !supported_features->system_files)
1860 "%lu files in %"TS" are marked as system files,\n"
1861 " but this attribute is not supported in %"TS".",
1862 required_features->system_files, loc, mode);
1865 if (required_features->compressed_files && !supported_features->compressed_files)
1868 "%lu files in %"TS" are marked as being transparently\n"
1869 " compressed, but transparent compression is not supported in\n"
1870 " %"TS". These files will be extracted as uncompressed.",
1871 required_features->compressed_files, loc, mode);
1874 if (required_features->encrypted_files && !supported_features->encrypted_files)
1877 "%lu files in %"TS" are marked as being encrypted,\n"
1878 " but encryption is not supported in %"TS". These files\n"
1879 " will not be extracted.",
1880 required_features->encrypted_files, loc, mode);
1883 if (required_features->encrypted_directories &&
1884 !supported_features->encrypted_directories)
1887 "%lu directories in %"TS" are marked as being encrypted,\n"
1888 " but encryption is not supported in %"TS".\n"
1889 " These directories will be extracted as unencrypted.",
1890 required_features->encrypted_directories, loc, mode);
1893 if (required_features->not_context_indexed_files &&
1894 !supported_features->not_context_indexed_files)
1897 "%lu files in %"TS" are marked as not content indexed,\n"
1898 " but this attribute is not supported in %"TS".",
1899 required_features->not_context_indexed_files, loc, mode);
1902 if (required_features->sparse_files && !supported_features->sparse_files)
1905 "%lu files in %"TS" are marked as sparse, but creating\n"
1906 " sparse files is not supported in %"TS". These files\n"
1907 " will be extracted as non-sparse.",
1908 required_features->sparse_files, loc, mode);
1911 if (required_features->named_data_streams &&
1912 !supported_features->named_data_streams)
1915 "%lu files in %"TS" contain one or more alternate (named)\n"
1916 " data streams, which are not supported in %"TS".\n"
1917 " Alternate data streams will NOT be extracted.",
1918 required_features->named_data_streams, loc, mode);
1921 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1922 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1923 required_features->named_data_streams &&
1924 supported_features->named_data_streams)
1927 "%lu files in %"TS" contain one or more alternate (named)\n"
1928 " data streams, which are not supported in linked extraction mode.\n"
1929 " Alternate data streams will NOT be extracted.",
1930 required_features->named_data_streams, loc);
1933 if (required_features->hard_links && !supported_features->hard_links)
1936 "%lu files in %"TS" are hard links, but hard links are\n"
1937 " not supported in %"TS". Hard links will be extracted as\n"
1938 " duplicate copies of the linked files.",
1939 required_features->hard_links, loc, mode);
1942 if (required_features->reparse_points && !supported_features->reparse_points)
1944 if (supported_features->symlink_reparse_points) {
1945 if (required_features->other_reparse_points) {
1947 "%lu files in %"TS" are reparse points that are neither\n"
1948 " symbolic links nor junction points and are not supported in\n"
1949 " %"TS". These reparse points will not be extracted.",
1950 required_features->other_reparse_points, loc,
1955 "%lu files in %"TS" are reparse points, which are\n"
1956 " not supported in %"TS" and will not be extracted.",
1957 required_features->reparse_points, loc, mode);
1961 if (required_features->security_descriptors &&
1962 !supported_features->security_descriptors)
1965 "%lu files in %"TS" have Windows NT security descriptors,\n"
1966 " but extracting security descriptors is not supported in\n"
1967 " %"TS". No security descriptors will be extracted.",
1968 required_features->security_descriptors, loc, mode);
1971 if (required_features->short_names && !supported_features->short_names)
1974 "%lu files in %"TS" have short (DOS) names, but\n"
1975 " extracting short names is not supported in %"TS".\n"
1976 " Short names will not be extracted.\n",
1977 required_features->short_names, loc, mode);
1980 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1981 required_features->unix_data && !supported_features->unix_data)
1983 ERROR("Extracting UNIX data is not supported in %"TS, mode);
1984 return WIMLIB_ERR_UNSUPPORTED;
1986 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
1987 required_features->short_names && !supported_features->short_names)
1989 ERROR("Extracting short names is not supported in %"TS"", mode);
1990 return WIMLIB_ERR_UNSUPPORTED;
1992 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1993 !ops->set_timestamps)
1995 ERROR("Extracting timestamps is not supported in %"TS"", mode);
1996 return WIMLIB_ERR_UNSUPPORTED;
1998 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1999 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
2000 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
2001 required_features->security_descriptors &&
2002 !supported_features->security_descriptors)
2004 ERROR("Extracting security descriptors is not supported in %"TS, mode);
2005 return WIMLIB_ERR_UNSUPPORTED;
2008 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
2009 !supported_features->hard_links)
2011 ERROR("Hard link extraction mode requested, but "
2012 "%"TS" does not support hard links!", mode);
2013 return WIMLIB_ERR_UNSUPPORTED;
2016 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
2017 required_features->symlink_reparse_points &&
2018 !(supported_features->symlink_reparse_points ||
2019 supported_features->reparse_points))
2021 ERROR("Extracting symbolic links is not supported in %"TS, mode);
2022 return WIMLIB_ERR_UNSUPPORTED;
2025 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
2026 !supported_features->symlink_reparse_points)
2028 ERROR("Symbolic link extraction mode requested, but "
2029 "%"TS" does not support symbolic "
2031 return WIMLIB_ERR_UNSUPPORTED;
2037 * extract_tree - Extract a file or directory tree from the currently selected
2040 * @wim: WIMStruct for the WIM file, with the desired image selected
2041 * (as wim->current_image).
2044 * "Canonical" (i.e. no leading or trailing slashes, path
2045 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
2046 * extract. An empty string means the full image.
2049 * Filesystem path to extract the file or directory tree to.
2050 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
2053 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
2054 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
2055 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
2059 * If non-NULL, progress function for the extraction. The messages
2060 * that may be sent in this function are:
2062 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
2063 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
2064 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
2065 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
2066 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
2067 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
2068 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
2069 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
2070 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
2072 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
2075 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
2076 int extract_flags, wimlib_progress_func_t progress_func)
2078 struct wim_dentry *root;
2079 struct wim_features required_features;
2080 struct apply_ctx ctx;
2082 struct wim_lookup_table_entry *lte;
2084 /* Start initializing the apply_ctx. */
2085 memset(&ctx, 0, sizeof(struct apply_ctx));
2087 ctx.extract_flags = extract_flags;
2088 ctx.target = target;
2089 ctx.target_nchars = tstrlen(target);
2090 ctx.progress_func = progress_func;
2091 if (progress_func) {
2092 ctx.progress.extract.wimfile_name = wim->filename;
2093 ctx.progress.extract.image = wim->current_image;
2094 ctx.progress.extract.extract_flags = (extract_flags &
2095 WIMLIB_EXTRACT_MASK_PUBLIC);
2096 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
2097 wim->current_image);
2098 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
2099 ctx.progress.extract.target = target;
2101 INIT_LIST_HEAD(&ctx.stream_list);
2103 /* Translate the path to extract into the corresponding
2104 * `struct wim_dentry', which will be the root of the
2105 * "dentry tree" to extract. */
2106 root = get_dentry(wim, wim_source_path);
2108 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
2109 wim_source_path, wim->current_image);
2110 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
2114 ctx.extract_root = root;
2116 /* Select the appropriate apply_operations based on the
2117 * platform and extract_flags. */
2119 ctx.ops = &win32_apply_ops;
2121 ctx.ops = &unix_apply_ops;
2125 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
2126 ctx.ops = &ntfs_3g_apply_ops;
2129 /* Call the start_extract() callback. This gives the apply_operations
2130 * implementation a chance to do any setup needed to access the volume.
2131 * Furthermore, it's expected to set the supported features of this
2132 * extraction mode (ctx.supported_features), which are determined at
2133 * runtime as they may vary depending on the actual volume. These
2134 * features are then compared with the actual features extracting this
2135 * dentry tree requires. Some mismatches will merely produce warnings
2136 * and the unsupported data will be ignored; others will produce errors.
2138 ret = ctx.ops->start_extract(target, &ctx);
2142 dentry_tree_get_features(root, &required_features);
2143 ret = do_feature_check(&required_features, &ctx.supported_features,
2144 extract_flags, ctx.ops, wim_source_path);
2146 goto out_finish_or_abort_extract;
2148 ctx.supported_attributes_mask =
2149 compute_supported_attributes_mask(&ctx.supported_features);
2151 /* Figure out whether the root dentry is being extracted to the root of
2152 * a volume and therefore needs to be treated "specially", for example
2153 * not being explicitly created and not having attributes set. */
2154 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2155 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2157 /* Calculate the actual filename component of each extracted dentry. In
2158 * the process, set the dentry->extraction_skipped flag on dentries that
2159 * are being skipped for some reason (e.g. invalid filename). */
2160 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
2162 goto out_dentry_reset_needs_extraction;
2164 /* Build the list of the streams that need to be extracted and
2165 * initialize ctx.progress.extract with stream information. */
2166 ret = for_dentry_in_tree(ctx.extract_root,
2167 dentry_resolve_and_zero_lte_refcnt, &ctx);
2169 goto out_dentry_reset_needs_extraction;
2171 ret = for_dentry_in_tree(ctx.extract_root,
2172 dentry_add_streams_to_extract, &ctx);
2174 goto out_teardown_stream_list;
2176 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2177 /* When extracting from a pipe, the number of bytes of data to
2178 * extract can't be determined in the normal way (examining the
2179 * lookup table), since at this point all we have is a set of
2180 * SHA1 message digests of streams that need to be extracted.
2181 * However, we can get a reasonably accurate estimate by taking
2182 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2183 * data. This does assume that a full image is being extracted,
2184 * but currently there is no API for doing otherwise. (Also,
2185 * subtract <HARDLINKBYTES> from this if hard links are
2186 * supported by the extraction mode.) */
2187 ctx.progress.extract.total_bytes =
2188 wim_info_get_image_total_bytes(wim->wim_info,
2189 wim->current_image);
2190 if (ctx.supported_features.hard_links) {
2191 ctx.progress.extract.total_bytes -=
2192 wim_info_get_image_hard_link_bytes(wim->wim_info,
2193 wim->current_image);
2197 /* Handle the special case of extracting a file to standard
2198 * output. In that case, "root" should be a single file, not a
2199 * directory tree. (If not, extract_dentry_to_stdout() will
2200 * return an error.) */
2201 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
2202 ret = extract_dentry_to_stdout(root);
2203 goto out_teardown_stream_list;
2206 /* If a sequential extraction was specified, sort the streams to be
2207 * extracted by their position in the WIM file so that the WIM file can
2208 * be read sequentially. */
2209 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2210 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2211 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2213 ret = sort_stream_list_by_sequential_order(
2215 offsetof(struct wim_lookup_table_entry,
2218 goto out_teardown_stream_list;
2221 if (ctx.ops->realpath_works_on_nonexisting_files &&
2222 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2223 ctx.ops->requires_realtarget_in_paths))
2225 ctx.realtarget = realpath(target, NULL);
2226 if (!ctx.realtarget) {
2227 ret = WIMLIB_ERR_NOMEM;
2228 goto out_teardown_stream_list;
2230 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2233 if (progress_func) {
2234 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2235 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2239 if (!ctx.root_dentry_is_special)
2241 tchar path[ctx.ops->path_max];
2242 if (build_extraction_path(path, root, &ctx))
2244 ret = extract_inode(path, &ctx, root->d_inode);
2246 goto out_free_realtarget;
2250 /* If we need to fix up the targets of absolute symbolic links
2251 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2252 * be absolute, use realpath() (or its replacement on Windows) to get
2253 * the absolute path to the extraction target. Note that this requires
2254 * the target directory to exist, unless
2255 * realpath_works_on_nonexisting_files is set in the apply_operations.
2257 if (!ctx.realtarget &&
2258 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2259 required_features.symlink_reparse_points) ||
2260 ctx.ops->requires_realtarget_in_paths))
2262 ctx.realtarget = realpath(target, NULL);
2263 if (!ctx.realtarget) {
2264 ret = WIMLIB_ERR_NOMEM;
2265 goto out_free_realtarget;
2267 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2270 /* Finally, the important part: extract the tree of files. */
2271 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2272 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2273 /* Sequential extraction requested, so two passes are needed
2274 * (one for directory structure, one for streams.) */
2276 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2279 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2280 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2282 goto out_free_realtarget;
2285 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2287 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2288 ret = extract_streams_from_pipe(&ctx);
2290 ret = extract_stream_list(&ctx);
2292 goto out_free_realtarget;
2294 /* Sequential extraction was not requested, so we can make do
2295 * with one pass where we both create the files and extract
2298 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2300 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2302 goto out_free_realtarget;
2304 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2308 /* If the total number of bytes to extract was miscalculated, just jump
2309 * to the calculated number in order to avoid confusing the progress
2310 * function. This should only occur when extracting from a pipe. */
2311 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2313 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2314 "extracted %"PRIu64,
2315 ctx.progress.extract.total_bytes,
2316 ctx.progress.extract.completed_bytes);
2318 if (progress_func &&
2319 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2321 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2322 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2325 /* Apply security descriptors and timestamps. This is done at the end,
2326 * and in a depth-first manner, to prevent timestamps from getting
2327 * changed by subsequent extract operations and to minimize the chance
2328 * of the restored security descriptors getting in our way. */
2330 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2332 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2334 goto out_free_realtarget;
2336 if (progress_func) {
2337 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2338 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2343 out_free_realtarget:
2344 FREE(ctx.realtarget);
2345 out_teardown_stream_list:
2346 /* Free memory allocated as part of the mapping from each
2347 * wim_lookup_table_entry to the dentries that reference it. */
2348 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2349 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2350 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2351 FREE(lte->lte_dentries);
2352 out_dentry_reset_needs_extraction:
2353 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2354 out_finish_or_abort_extract:
2356 if (ctx.ops->abort_extract)
2357 ctx.ops->abort_extract(&ctx);
2359 if (ctx.ops->finish_extract)
2360 ret = ctx.ops->finish_extract(&ctx);
2366 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2367 * extract flags make sense. */
2369 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2373 /* Empty destination path? */
2374 if (cmd->fs_dest_path[0] == T('\0'))
2375 return WIMLIB_ERR_INVALID_PARAM;
2377 extract_flags = cmd->extract_flags;
2379 /* Check for invalid flag combinations */
2380 if ((extract_flags &
2381 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2382 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2383 WIMLIB_EXTRACT_FLAG_HARDLINK))
2384 return WIMLIB_ERR_INVALID_PARAM;
2386 if ((extract_flags &
2387 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2388 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2389 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2390 return WIMLIB_ERR_INVALID_PARAM;
2392 if ((extract_flags &
2393 (WIMLIB_EXTRACT_FLAG_RPFIX |
2394 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2395 WIMLIB_EXTRACT_FLAG_NORPFIX))
2396 return WIMLIB_ERR_INVALID_PARAM;
2398 if ((extract_flags &
2399 (WIMLIB_EXTRACT_FLAG_RESUME |
2400 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2401 return WIMLIB_ERR_INVALID_PARAM;
2403 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2404 #ifndef WITH_NTFS_3G
2405 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2406 " we cannot apply a WIM image directly to a NTFS volume.");
2407 return WIMLIB_ERR_UNSUPPORTED;
2411 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2412 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2414 /* Do reparse point fixups by default if the WIM header says
2415 * they are enabled and we are extracting a full image. */
2416 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2417 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2420 /* TODO: Since UNIX data entries are stored in the file resources, in a
2421 * completely sequential extraction they may come up before the
2422 * corresponding file or symbolic link data. This needs to be handled
2424 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2425 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2426 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2427 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2429 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2430 WARNING("Setting UNIX file/owner group may "
2431 "be impossible on some\n"
2433 "when applying from a pipe.");
2435 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2436 WARNING("Disabling sequential extraction for "
2441 cmd->extract_flags = extract_flags;
2446 /* Internal function to execute extraction commands for a WIM image. The paths
2447 * in the extract commands are expected to be already "canonicalized". */
2449 do_wimlib_extract_files(WIMStruct *wim,
2451 struct wimlib_extract_command *cmds,
2453 wimlib_progress_func_t progress_func)
2456 bool found_link_cmd = false;
2457 bool found_nolink_cmd = false;
2459 /* Select the image from which we are extracting files */
2460 ret = select_wim_image(wim, image);
2464 /* Make sure there are no streams in the WIM that have not been
2465 * checksummed yet. */
2466 ret = wim_checksum_unhashed_streams(wim);
2470 /* Check for problems with the extraction commands */
2471 for (size_t i = 0; i < num_cmds; i++) {
2472 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2475 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2476 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2477 found_link_cmd = true;
2479 found_nolink_cmd = true;
2481 if (found_link_cmd && found_nolink_cmd) {
2482 ERROR("Symlink or hardlink extraction mode must "
2483 "be set on all extraction commands");
2484 return WIMLIB_ERR_INVALID_PARAM;
2488 /* Execute the extraction commands */
2489 for (size_t i = 0; i < num_cmds; i++) {
2490 ret = extract_tree(wim,
2491 cmds[i].wim_source_path,
2492 cmds[i].fs_dest_path,
2493 cmds[i].extract_flags,
2501 /* API function documented in wimlib.h */
2503 wimlib_extract_files(WIMStruct *wim,
2505 const struct wimlib_extract_command *cmds,
2507 int default_extract_flags,
2508 WIMStruct **additional_swms,
2509 unsigned num_additional_swms,
2510 wimlib_progress_func_t progress_func)
2513 struct wimlib_extract_command *cmds_copy;
2516 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2518 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2525 if (num_additional_swms)
2526 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2528 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2530 ret = WIMLIB_ERR_NOMEM;
2531 goto out_restore_lookup_table;
2534 for (size_t i = 0; i < num_cmds; i++) {
2535 cmds_copy[i].extract_flags = (default_extract_flags |
2536 cmds[i].extract_flags)
2537 & WIMLIB_EXTRACT_MASK_PUBLIC;
2538 all_flags |= cmds_copy[i].extract_flags;
2540 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
2541 if (!cmds_copy[i].wim_source_path) {
2542 ret = WIMLIB_ERR_NOMEM;
2543 goto out_free_cmds_copy;
2546 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
2547 if (!cmds_copy[i].fs_dest_path) {
2548 ret = WIMLIB_ERR_NOMEM;
2549 goto out_free_cmds_copy;
2553 ret = do_wimlib_extract_files(wim, image,
2554 cmds_copy, num_cmds,
2557 if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2558 WIMLIB_EXTRACT_FLAG_HARDLINK))
2560 for_lookup_table_entry(wim->lookup_table,
2561 lte_free_extracted_file, NULL);
2564 for (size_t i = 0; i < num_cmds; i++) {
2565 FREE(cmds_copy[i].wim_source_path);
2566 FREE(cmds_copy[i].fs_dest_path);
2569 out_restore_lookup_table:
2570 if (num_additional_swms)
2571 unmerge_lookup_table(wim);
2577 * Extracts an image from a WIM file.
2579 * @wim: WIMStruct for the WIM file.
2581 * @image: Number of the single image to extract.
2583 * @target: Directory or NTFS volume to extract the image to.
2585 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2587 * @progress_func: If non-NULL, a progress function to be called
2590 * Returns 0 on success; nonzero on failure.
2593 extract_single_image(WIMStruct *wim, int image,
2594 const tchar *target, int extract_flags,
2595 wimlib_progress_func_t progress_func)
2598 tchar *target_copy = canonicalize_fs_path(target);
2600 return WIMLIB_ERR_NOMEM;
2601 struct wimlib_extract_command cmd = {
2602 .wim_source_path = T(""),
2603 .fs_dest_path = target_copy,
2604 .extract_flags = extract_flags,
2606 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2611 static const tchar * const filename_forbidden_chars =
2620 /* This function checks if it is okay to use a WIM image's name as a directory
2623 image_name_ok_as_dir(const tchar *image_name)
2625 return image_name && *image_name &&
2626 !tstrpbrk(image_name, filename_forbidden_chars) &&
2627 tstrcmp(image_name, T(".")) &&
2628 tstrcmp(image_name, T(".."));
2631 /* Extracts all images from the WIM to the directory @target, with the images
2632 * placed in subdirectories named by their image names. */
2634 extract_all_images(WIMStruct *wim,
2635 const tchar *target,
2637 wimlib_progress_func_t progress_func)
2639 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2640 size_t output_path_len = tstrlen(target);
2641 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2644 const tchar *image_name;
2647 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2649 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2650 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2651 return WIMLIB_ERR_INVALID_PARAM;
2654 if (tstat(target, &stbuf)) {
2655 if (errno == ENOENT) {
2656 if (tmkdir(target, 0755)) {
2657 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2658 return WIMLIB_ERR_MKDIR;
2661 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2662 return WIMLIB_ERR_STAT;
2664 } else if (!S_ISDIR(stbuf.st_mode)) {
2665 ERROR("\"%"TS"\" is not a directory", target);
2666 return WIMLIB_ERR_NOTDIR;
2669 tmemcpy(buf, target, output_path_len);
2670 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2671 for (image = 1; image <= wim->hdr.image_count; image++) {
2672 image_name = wimlib_get_image_name(wim, image);
2673 if (image_name_ok_as_dir(image_name)) {
2674 tstrcpy(buf + output_path_len + 1, image_name);
2676 /* Image name is empty or contains forbidden characters.
2677 * Use image number instead. */
2678 tsprintf(buf + output_path_len + 1, T("%d"), image);
2680 ret = extract_single_image(wim, image, buf, extract_flags,
2689 do_wimlib_extract_image(WIMStruct *wim,
2691 const tchar *target,
2693 WIMStruct **additional_swms,
2694 unsigned num_additional_swms,
2695 wimlib_progress_func_t progress_func)
2699 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2700 wimlib_assert(wim->hdr.part_number == 1);
2701 wimlib_assert(num_additional_swms == 0);
2703 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2707 if (num_additional_swms)
2708 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2711 if (image == WIMLIB_ALL_IMAGES) {
2712 ret = extract_all_images(wim, target, extract_flags,
2715 ret = extract_single_image(wim, image, target, extract_flags,
2719 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2720 WIMLIB_EXTRACT_FLAG_HARDLINK))
2722 for_lookup_table_entry(wim->lookup_table,
2723 lte_free_extracted_file,
2726 if (num_additional_swms)
2727 unmerge_lookup_table(wim);
2731 /* API function documented in wimlib.h */
2733 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2734 const tchar *target, int extract_flags,
2735 wimlib_progress_func_t progress_func)
2739 struct filedes *in_fd;
2743 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2745 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2746 return WIMLIB_ERR_INVALID_PARAM;
2748 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2750 /* Read the WIM header from the pipe and get a WIMStruct to represent
2751 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2752 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2753 * an empty lookup table, no XML data read, and no filename set. */
2754 ret = open_wim_as_WIMStruct(&pipe_fd,
2755 WIMLIB_OPEN_FLAG_FROM_PIPE |
2756 WIMLIB_OPEN_FLAG_SPLIT_OK,
2757 &pwm, progress_func);
2761 /* Sanity check to make sure this is a pipable WIM. */
2762 if (pwm->hdr.magic != PWM_MAGIC) {
2763 ERROR("The WIM being read from file descriptor %d "
2764 "is not pipable!", pipe_fd);
2765 ret = WIMLIB_ERR_NOT_PIPABLE;
2766 goto out_wimlib_free;
2769 /* Sanity check to make sure the first part of a pipable split WIM is
2770 * sent over the pipe first. */
2771 if (pwm->hdr.part_number != 1) {
2772 ERROR("The first part of the split WIM must be "
2773 "sent over the pipe first.");
2774 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2775 goto out_wimlib_free;
2778 in_fd = &pwm->in_fd;
2779 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2781 /* As mentioned, the WIMStruct we created from the pipe does not have
2782 * XML data yet. Fix this by reading the extra copy of the XML data
2783 * that directly follows the header in pipable WIMs. (Note: see
2784 * write_pipable_wim() for more details about the format of pipable
2787 struct wim_lookup_table_entry xml_lte;
2788 ret = read_pwm_stream_header(pwm, &xml_lte, 0, NULL);
2790 goto out_wimlib_free;
2792 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2794 ERROR("Expected XML data, but found non-metadata "
2796 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2797 goto out_wimlib_free;
2800 copy_resource_entry(&pwm->hdr.xml_res_entry,
2801 &xml_lte.resource_entry);
2803 ret = read_wim_xml_data(pwm);
2805 goto out_wimlib_free;
2806 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2807 ERROR("Image count in XML data is not the same as in WIM header.");
2808 ret = WIMLIB_ERR_XML;
2809 goto out_wimlib_free;
2813 /* Get image index (this may use the XML data that was just read to
2814 * resolve an image name). */
2815 if (image_num_or_name) {
2816 image = wimlib_resolve_image(pwm, image_num_or_name);
2817 if (image == WIMLIB_NO_IMAGE) {
2818 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2820 ret = WIMLIB_ERR_INVALID_IMAGE;
2821 goto out_wimlib_free;
2822 } else if (image == WIMLIB_ALL_IMAGES) {
2823 ERROR("Applying all images from a pipe is not supported.");
2824 ret = WIMLIB_ERR_INVALID_IMAGE;
2825 goto out_wimlib_free;
2828 if (pwm->hdr.image_count != 1) {
2829 ERROR("No image was specified, but the pipable WIM "
2830 "did not contain exactly 1 image");
2831 ret = WIMLIB_ERR_INVALID_IMAGE;
2832 goto out_wimlib_free;
2837 /* Load the needed metadata resource. */
2838 for (i = 1; i <= pwm->hdr.image_count; i++) {
2839 struct wim_lookup_table_entry *metadata_lte;
2840 struct wim_image_metadata *imd;
2842 metadata_lte = new_lookup_table_entry();
2843 if (!metadata_lte) {
2844 ret = WIMLIB_ERR_NOMEM;
2845 goto out_wimlib_free;
2848 ret = read_pwm_stream_header(pwm, metadata_lte, 0, NULL);
2849 imd = pwm->image_metadata[i - 1];
2850 imd->metadata_lte = metadata_lte;
2852 goto out_wimlib_free;
2854 if (!(metadata_lte->resource_entry.flags &
2855 WIM_RESHDR_FLAG_METADATA))
2857 ERROR("Expected metadata resource, but found "
2858 "non-metadata stream.");
2859 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2860 goto out_wimlib_free;
2864 /* Metadata resource is for the images being extracted.
2865 * Parse it and save the metadata in memory. */
2866 ret = read_metadata_resource(pwm, imd);
2868 goto out_wimlib_free;
2871 /* Metadata resource is not for the image being
2872 * extracted. Skip over it. */
2873 ret = skip_pwm_stream(metadata_lte);
2875 goto out_wimlib_free;
2878 /* Extract the image. */
2879 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2880 ret = do_wimlib_extract_image(pwm, image, target,
2881 extract_flags, NULL, 0, progress_func);
2882 /* Clean up and return. */
2888 /* API function documented in wimlib.h */
2890 wimlib_extract_image(WIMStruct *wim,
2892 const tchar *target,
2894 WIMStruct **additional_swms,
2895 unsigned num_additional_swms,
2896 wimlib_progress_func_t progress_func)
2898 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2899 return do_wimlib_extract_image(wim, image, target, extract_flags,
2900 additional_swms, num_additional_swms,