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;
903 while (*path == path_separator)
907 while (*path && *path != path_separator)
910 return num_components;
914 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
915 struct apply_ctx *ctx, struct wim_dentry *dentry)
917 size_t num_raw_path_components;
918 const struct wim_dentry *d;
919 size_t num_target_path_components;
924 num_raw_path_components = 0;
925 for (d = dentry; d != ctx->extract_root; d = d->parent)
926 num_raw_path_components++;
928 if (ctx->ops->requires_realtarget_in_paths)
929 num_target_path_components = get_num_path_components(ctx->realtarget,
930 ctx->ops->path_separator);
931 else if (ctx->ops->requires_target_in_paths)
932 num_target_path_components = get_num_path_components(ctx->target,
933 ctx->ops->path_separator);
935 num_target_path_components = 0;
937 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
938 wimlib_assert(num_target_path_components > 0);
939 num_raw_path_components++;
940 num_target_path_components--;
944 while (*p_old == ctx->ops->path_separator)
946 while (--num_target_path_components) {
947 while (*p_old != ctx->ops->path_separator)
949 while (*p_old == ctx->ops->path_separator)
953 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
955 p = &symlink_target[0];
956 while (num_raw_path_components--) {
959 *p++ = ctx->ops->path_separator;
962 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
963 newpath, symlink_target);
964 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
966 ERROR_WITH_ERRNO("Failed to create symlink "
967 "\"%"TS"\" => \"%"TS"\"",
968 newpath, symlink_target);
973 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
974 * data streams, reparse data (including symbolic links), timestamps, and
975 * security descriptors. Basically, everything that doesn't require reading
976 * non-metadata resources from the WIM file and isn't delayed until the final
979 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
980 struct apply_ctx *ctx)
982 struct wim_inode *inode = dentry->d_inode;
984 const tchar *oldpath;
986 if (unlikely(is_linked_extraction(ctx))) {
987 struct wim_lookup_table_entry *unnamed_lte;
989 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
990 if (unnamed_lte && unnamed_lte->extracted_file) {
991 oldpath = unnamed_lte->extracted_file;
992 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
999 /* Create hard link if this dentry corresponds to an already-extracted
1001 if (inode->i_extracted_file) {
1002 oldpath = inode->i_extracted_file;
1006 /* Skip symlinks unless they can be extracted as reparse points rather
1007 * than created directly. */
1008 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
1011 /* Create this file or directory unless it's the extraction root, which
1012 * was already created if necessary. */
1013 if (dentry != ctx->extract_root) {
1014 ret = extract_inode(path, ctx, inode);
1019 /* Create empty named data streams. */
1020 if (can_extract_named_data_streams(ctx)) {
1021 for (u16 i = 0; i < inode->i_num_ads; i++) {
1022 file_spec_t file_spec;
1023 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1025 if (!ads_entry_is_named_stream(entry))
1029 if (ctx->ops->uses_cookies)
1030 file_spec.cookie = inode->extract_cookie;
1032 file_spec.path = path;
1033 ret = ctx->ops->extract_named_stream(file_spec,
1035 entry->stream_name_nbytes / 2,
1038 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1039 "empty named data stream",
1046 /* Set file attributes (if supported). */
1047 ret = extract_file_attributes(path, ctx, dentry);
1051 /* Set or remove file short name (if supported). */
1052 ret = extract_short_name(path, ctx, dentry);
1056 /* If inode has multiple links and hard links are supported in this
1057 * extraction mode and volume, save the path to the extracted file in
1058 * case it's needed to create a hard link. */
1059 if (unlikely(is_linked_extraction(ctx))) {
1060 struct wim_lookup_table_entry *unnamed_lte;
1062 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1064 unnamed_lte->extracted_file = TSTRDUP(path);
1065 if (!unnamed_lte->extracted_file)
1066 return WIMLIB_ERR_NOMEM;
1068 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1069 inode->i_extracted_file = TSTRDUP(path);
1070 if (!inode->i_extracted_file)
1071 return WIMLIB_ERR_NOMEM;
1076 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1079 dentry->was_hardlinked = 1;
1083 ret = extract_hardlink(oldpath, path, ctx);
1086 dentry->was_hardlinked = 1;
1091 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1093 struct apply_ctx *ctx = _ctx;
1094 tchar path[ctx->ops->path_max];
1095 struct wim_dentry *orig_dentry;
1096 struct wim_dentry *other_dentry;
1099 /* Here we may re-order the extraction of multiple names (hard links)
1100 * for the same file in the same directory in order to ensure the short
1101 * (DOS) name is set correctly. A short name is always associated with
1102 * exactly one long name, and at least on NTFS, only one long name for a
1103 * file can have a short name associated with it. (More specifically,
1104 * there can be unlimited names in the POSIX namespace, but only one
1105 * name can be in the Win32+DOS namespace, or one name in the Win32
1106 * namespace with a corresponding name in the DOS namespace.) To ensure
1107 * the short name of a file is associated with the correct long name in
1108 * a directory, we extract the long name with a corresponding short name
1109 * before any additional names. This can affect NTFS-3g extraction
1110 * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying
1111 * the long name to associate with a short name) and may affect Win32
1112 * extraction as well (which uses SetFileShortName()). */
1114 if (dentry->skeleton_extracted)
1117 if (ctx->supported_features.short_names
1118 && !dentry_has_short_name(dentry)
1119 && !dentry->d_inode->i_dos_name_extracted)
1121 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1122 if (dentry_has_short_name(other_dentry)
1123 && !other_dentry->skeleton_extracted
1124 && other_dentry->parent == dentry->parent)
1126 DEBUG("Creating %"TS" before %"TS" "
1127 "to guarantee correct DOS name extraction",
1128 dentry_full_path(other_dentry),
1129 dentry_full_path(dentry));
1130 orig_dentry = dentry;
1131 dentry = other_dentry;
1137 if (!build_extraction_path(path, dentry, ctx))
1139 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1143 dentry->skeleton_extracted = 1;
1146 dentry = orig_dentry;
1150 dentry->d_inode->i_dos_name_extracted = 1;
1154 /* Create a file or directory, then immediately extract all streams. This
1155 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1156 * may not be read sequentially by this function. */
1158 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1160 struct apply_ctx *ctx = _ctx;
1161 tchar path[ctx->ops->path_max];
1164 ret = dentry_extract_skeleton(dentry, ctx);
1168 if (!build_extraction_path(path, dentry, ctx))
1171 return extract_streams(path, ctx, dentry, NULL, NULL);
1174 /* Extract all instances of the stream @lte that are being extracted in this
1175 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1176 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1177 * extracted multiple times, it is extracted to a temporary file first.
1179 * This is intended for use with sequential extraction of a WIM image
1180 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1182 extract_stream_instances(struct wim_lookup_table_entry *lte,
1183 struct apply_ctx *ctx, bool can_seek)
1185 struct wim_dentry **lte_dentries;
1186 struct wim_lookup_table_entry *lte_tmp = NULL;
1187 struct wim_lookup_table_entry *lte_override;
1188 tchar *stream_tmp_filename = NULL;
1189 tchar path[ctx->ops->path_max];
1193 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1194 lte_dentries = lte->inline_lte_dentries;
1196 lte_dentries = lte->lte_dentries;
1198 if (likely(can_seek || lte->out_refcnt < 2)) {
1201 /* Need to extract stream to temporary file. */
1205 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1206 if (!stream_tmp_filename) {
1207 ERROR_WITH_ERRNO("Failed to create temporary filename");
1208 ret = WIMLIB_ERR_OPEN;
1212 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1214 ret = WIMLIB_ERR_NOMEM;
1215 goto out_free_stream_tmp_filename;
1217 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1218 lte_tmp->file_on_disk = stream_tmp_filename;
1219 lte_override = lte_tmp;
1221 raw_fd = topen(stream_tmp_filename,
1222 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1224 ERROR_WITH_ERRNO("Failed to open temporary file");
1225 ret = WIMLIB_ERR_OPEN;
1226 goto out_free_lte_tmp;
1228 filedes_init(&fd, raw_fd);
1229 ret = extract_wim_resource_to_fd(lte, &fd,
1230 wim_resource_size(lte));
1231 if (filedes_close(&fd) && !ret)
1232 ret = WIMLIB_ERR_WRITE;
1234 goto out_unlink_stream_tmp_file;
1237 /* Extract all instances of the stream, reading either from the stream
1238 * in the WIM file or from the temporary file containing the stream.
1239 * dentry->tmp_flag is used to ensure that each dentry is processed only
1240 * once regardless of how many times this stream appears in the streams
1241 * of the corresponding inode. */
1242 for (i = 0; i < lte->out_refcnt; i++) {
1243 struct wim_dentry *dentry = lte_dentries[i];
1245 if (dentry->tmp_flag)
1247 if (!build_extraction_path(path, dentry, ctx))
1249 ret = extract_streams(path, ctx, dentry,
1252 goto out_clear_tmp_flags;
1253 dentry->tmp_flag = 1;
1256 out_clear_tmp_flags:
1257 for (i = 0; i < lte->out_refcnt; i++)
1258 lte_dentries[i]->tmp_flag = 0;
1259 out_unlink_stream_tmp_file:
1260 if (stream_tmp_filename)
1261 tunlink(stream_tmp_filename);
1264 out_free_stream_tmp_filename:
1265 FREE(stream_tmp_filename);
1270 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1271 * structure and empty files were already created. This relies on the
1272 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1273 * stream that was constructed earlier. Streams are extracted exactly in the
1274 * order of the stream list; however, unless the WIM's file descriptor is
1275 * detected to be non-seekable, streams may be read from the WIM file more than
1276 * one time if multiple copies need to be extracted. */
1278 extract_stream_list(struct apply_ctx *ctx)
1280 struct wim_lookup_table_entry *lte;
1284 can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
1285 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1286 ret = extract_stream_instances(lte, ctx, can_seek);
1293 #define PWM_ALLOW_WIM_HDR 0x00001
1294 #define PWM_SILENT_EOF 0x00002
1296 /* Read the header from a stream in a pipable WIM. */
1298 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1299 int flags, struct wim_header_disk *hdr_ret)
1302 struct pwm_stream_hdr stream_hdr;
1303 struct wim_header_disk pwm_hdr;
1307 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
1311 if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
1312 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
1313 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
1314 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
1318 lte->resource_location = RESOURCE_NONEXISTENT;
1319 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
1323 if (buf.stream_hdr.magic != PWM_STREAM_MAGIC) {
1324 ERROR("Data read on pipe is invalid (expected stream header).");
1325 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1328 lte->resource_entry.original_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
1329 copy_hash(lte->hash, buf.stream_hdr.hash);
1330 lte->resource_entry.flags = le32_to_cpu(buf.stream_hdr.flags);
1331 lte->resource_entry.offset = pwm->in_fd.offset;
1332 lte->resource_location = RESOURCE_IN_WIM;
1334 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1335 lte->compression_type = pwm->compression_type;
1336 lte->resource_entry.size = 0;
1338 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1339 lte->resource_entry.size = lte->resource_entry.original_size;
1341 lte->is_pipable = 1;
1345 if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1346 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1350 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1352 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1354 return read_partial_wim_resource(lte, wim_resource_size(lte),
1356 WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY,
1361 extract_streams_from_pipe(struct apply_ctx *ctx)
1363 struct wim_lookup_table_entry *found_lte;
1364 struct wim_lookup_table_entry *needed_lte;
1365 struct wim_lookup_table *lookup_table;
1366 struct wim_header_disk pwm_hdr;
1370 ret = WIMLIB_ERR_NOMEM;
1371 found_lte = new_lookup_table_entry();
1375 lookup_table = ctx->wim->lookup_table;
1376 pwm_flags = PWM_ALLOW_WIM_HDR;
1377 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1378 pwm_flags |= PWM_SILENT_EOF;
1379 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
1380 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
1381 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
1382 if (ctx->progress_func)
1383 ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1385 while (ctx->num_streams_remaining) {
1386 ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags,
1389 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1390 (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1394 goto out_free_found_lte;
1397 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1398 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1399 && (needed_lte = __lookup_resource(lookup_table, found_lte->hash))
1400 && (needed_lte->out_refcnt))
1402 copy_resource_entry(&needed_lte->resource_entry,
1403 &found_lte->resource_entry);
1404 needed_lte->resource_location = found_lte->resource_location;
1405 needed_lte->wim = found_lte->wim;
1406 needed_lte->compression_type = found_lte->compression_type;
1407 needed_lte->is_pipable = found_lte->is_pipable;
1409 ret = extract_stream_instances(needed_lte, ctx, false);
1411 goto out_free_found_lte;
1412 ctx->num_streams_remaining--;
1413 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1414 ret = skip_pwm_stream(found_lte);
1416 goto out_free_found_lte;
1418 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
1419 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
1421 if (part_number != ctx->progress.extract.part_number ||
1422 total_parts != ctx->progress.extract.total_parts ||
1423 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
1426 ctx->progress.extract.part_number = part_number;
1427 ctx->progress.extract.total_parts = total_parts;
1428 memcpy(ctx->progress.extract.guid,
1429 pwm_hdr.guid, WIM_GID_LEN);
1430 if (ctx->progress_func) {
1432 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1441 free_lookup_table_entry(found_lte);
1450 /* Finish extracting a file, directory, or symbolic link by setting file
1451 * security and timestamps. */
1453 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1455 struct apply_ctx *ctx = _ctx;
1457 tchar path[ctx->ops->path_max];
1459 if (!build_extraction_path(path, dentry, ctx))
1462 ret = extract_security(path, ctx, dentry);
1466 return extract_timestamps(path, ctx, dentry);
1470 * Extract a WIM dentry to standard output.
1472 * This obviously doesn't make sense in all cases. We return an error if the
1473 * dentry does not correspond to a regular file. Otherwise we extract the
1474 * unnamed data stream only.
1477 extract_dentry_to_stdout(struct wim_dentry *dentry)
1480 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1481 FILE_ATTRIBUTE_DIRECTORY))
1483 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1484 "extracted to standard output", dentry_full_path(dentry));
1485 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1487 struct wim_lookup_table_entry *lte;
1489 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1491 struct filedes _stdout;
1492 filedes_init(&_stdout, STDOUT_FILENO);
1493 ret = extract_wim_resource_to_fd(lte, &_stdout,
1494 wim_resource_size(lte));
1501 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1503 static const utf16lechar replacement_char = cpu_to_le16('?');
1507 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1513 for (i = 0; i < num_chars; i++) {
1516 case cpu_to_le16('\\'):
1517 case cpu_to_le16(':'):
1518 case cpu_to_le16('*'):
1519 case cpu_to_le16('?'):
1520 case cpu_to_le16('"'):
1521 case cpu_to_le16('<'):
1522 case cpu_to_le16('>'):
1523 case cpu_to_le16('|'):
1525 case cpu_to_le16('/'):
1526 case cpu_to_le16('\0'):
1528 name[i] = replacement_char;
1535 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1536 name[num_chars - 1] == cpu_to_le16('.'))
1539 name[num_chars - 1] = replacement_char;
1548 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1550 const utf16lechar *file_name = dentry->file_name;
1551 return file_name != NULL &&
1552 file_name[0] == cpu_to_le16('.') &&
1553 (file_name[1] == cpu_to_le16('\0') ||
1554 (file_name[1] == cpu_to_le16('.') &&
1555 file_name[2] == cpu_to_le16('\0')));
1559 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1561 dentry->extraction_skipped = 1;
1566 * dentry_calculate_extraction_path-
1568 * Calculate the actual filename component at which a WIM dentry will be
1569 * extracted, handling invalid filenames "properly".
1571 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1572 * UNIX, converted into the platform's multibyte encoding). However, if the
1573 * file name contains characters that are not valid on the current platform or
1574 * has some other format that is not valid, leave dentry->extraction_name as
1575 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1576 * not be extracted, unless the appropriate flag
1577 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1578 * which case a substitute filename will be created and set instead.
1580 * Conflicts with case-insensitive names on Windows are handled similarly; see
1584 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1586 struct apply_ctx *ctx = _args;
1589 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1592 if (!dentry_is_supported(dentry, &ctx->supported_features))
1595 if (dentry_is_dot_or_dotdot(dentry)) {
1596 /* WIM files shouldn't contain . or .. entries. But if they are
1597 * there, don't attempt to extract them. */
1598 WARNING("Skipping extraction of unexpected . or .. file "
1599 "\"%"TS"\"", dentry_full_path(dentry));
1604 if (!ctx->ops->supports_case_sensitive_filenames)
1606 struct wim_dentry *other;
1607 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1608 case_insensitive_conflict_list)
1610 if (ctx->extract_flags &
1611 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1612 WARNING("\"%"TS"\" has the same "
1613 "case-insensitive name as "
1614 "\"%"TS"\"; extracting "
1615 "dummy name instead",
1616 dentry_full_path(dentry),
1617 dentry_full_path(other));
1620 WARNING("Not extracting \"%"TS"\": "
1621 "has same case-insensitive "
1622 "name as \"%"TS"\"",
1623 dentry_full_path(dentry),
1624 dentry_full_path(other));
1629 #else /* __WIN32__ */
1630 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1631 #endif /* !__WIN32__ */
1633 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1635 dentry->extraction_name = dentry->file_name;
1636 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1639 return utf16le_to_tstr(dentry->file_name,
1640 dentry->file_name_nbytes,
1641 &dentry->extraction_name,
1642 &dentry->extraction_name_nchars);
1645 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1647 WARNING("\"%"TS"\" has an invalid filename "
1648 "that is not supported on this platform; "
1649 "extracting dummy name instead",
1650 dentry_full_path(dentry));
1653 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1654 "that is not supported on this platform",
1655 dentry_full_path(dentry));
1662 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1664 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1665 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1668 size_t tchar_nchars;
1670 tchar_name = utf16_name_copy;
1671 tchar_nchars = dentry->file_name_nbytes / 2;
1673 ret = utf16le_to_tstr(utf16_name_copy,
1674 dentry->file_name_nbytes,
1675 &tchar_name, &tchar_nchars);
1679 size_t fixed_name_num_chars = tchar_nchars;
1680 tchar fixed_name[tchar_nchars + 50];
1682 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1683 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1684 T(" (invalid filename #%lu)"),
1685 ++ctx->invalid_sequence);
1689 dentry->extraction_name = memdup(fixed_name,
1690 2 * fixed_name_num_chars + 2);
1691 if (!dentry->extraction_name)
1692 return WIMLIB_ERR_NOMEM;
1693 dentry->extraction_name_nchars = fixed_name_num_chars;
1698 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1702 /* Clean up dentry and inode structure after extraction. */
1704 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1706 struct wim_inode *inode = dentry->d_inode;
1708 dentry->extraction_skipped = 0;
1709 dentry->was_hardlinked = 0;
1710 dentry->skeleton_extracted = 0;
1711 inode->i_visited = 0;
1712 FREE(inode->i_extracted_file);
1713 inode->i_extracted_file = NULL;
1714 inode->i_dos_name_extracted = 0;
1715 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1716 FREE(dentry->extraction_name);
1717 dentry->extraction_name = NULL;
1721 /* Tally features necessary to extract a dentry and the corresponding inode. */
1723 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1725 struct wim_features *features = _features;
1726 struct wim_inode *inode = dentry->d_inode;
1728 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1729 features->archive_files++;
1730 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1731 features->hidden_files++;
1732 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1733 features->system_files++;
1734 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1735 features->compressed_files++;
1736 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1737 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1738 features->encrypted_directories++;
1740 features->encrypted_files++;
1742 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1743 features->not_context_indexed_files++;
1744 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1745 features->sparse_files++;
1746 if (inode_has_named_stream(inode))
1747 features->named_data_streams++;
1748 if (inode->i_visited)
1749 features->hard_links++;
1750 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1751 features->reparse_points++;
1752 if (inode_is_symlink(inode))
1753 features->symlink_reparse_points++;
1755 features->other_reparse_points++;
1757 if (inode->i_security_id != -1)
1758 features->security_descriptors++;
1759 if (dentry->short_name_nbytes)
1760 features->short_names++;
1761 if (inode_has_unix_data(inode))
1762 features->unix_data++;
1763 inode->i_visited = 1;
1768 dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
1770 dentry->d_inode->i_visited = 0;
1774 /* Tally the features necessary to extract a dentry tree. */
1776 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1778 memset(features, 0, sizeof(struct wim_features));
1779 for_dentry_in_tree(root, dentry_tally_features, features);
1780 for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
1784 compute_supported_attributes_mask(const struct wim_features *supported_features)
1788 if (!supported_features->archive_files)
1789 mask &= ~FILE_ATTRIBUTE_ARCHIVE;
1791 if (!supported_features->hidden_files)
1792 mask &= ~FILE_ATTRIBUTE_HIDDEN;
1794 if (!supported_features->system_files)
1795 mask &= ~FILE_ATTRIBUTE_SYSTEM;
1797 if (!supported_features->not_context_indexed_files)
1798 mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED;
1800 if (!supported_features->compressed_files)
1801 mask &= ~FILE_ATTRIBUTE_COMPRESSED;
1803 if (!supported_features->sparse_files)
1804 mask &= ~FILE_ATTRIBUTE_SPARSE_FILE;
1806 if (!supported_features->reparse_points)
1807 mask &= ~FILE_ATTRIBUTE_REPARSE_POINT;
1813 do_feature_check(const struct wim_features *required_features,
1814 const struct wim_features *supported_features,
1816 const struct apply_operations *ops,
1817 const tchar *wim_source_path)
1820 const tchar *mode = T("this extraction mode");
1822 if (wim_source_path[0] == '\0')
1823 loc = T("the WIM image");
1825 loc = wim_source_path;
1827 /* We're an archive program, so theoretically we can do what we want
1828 * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't
1829 * bother the user about it. */
1831 if (required_features->archive_files && !supported_features->archive_files)
1834 "%lu files in %"TS" are marked as archived, but this attribute\n"
1835 " is not supported in %"TS".",
1836 required_features->archive_files, loc, mode);
1840 if (required_features->hidden_files && !supported_features->hidden_files)
1843 "%lu files in %"TS" are marked as hidden, but this\n"
1844 " attribute is not supported in %"TS".",
1845 required_features->hidden_files, loc, mode);
1848 if (required_features->system_files && !supported_features->system_files)
1851 "%lu files in %"TS" are marked as system files,\n"
1852 " but this attribute is not supported in %"TS".",
1853 required_features->system_files, loc, mode);
1856 if (required_features->compressed_files && !supported_features->compressed_files)
1859 "%lu files in %"TS" are marked as being transparently\n"
1860 " compressed, but transparent compression is not supported in\n"
1861 " %"TS". These files will be extracted as uncompressed.",
1862 required_features->compressed_files, loc, mode);
1865 if (required_features->encrypted_files && !supported_features->encrypted_files)
1868 "%lu files in %"TS" are marked as being encrypted,\n"
1869 " but encryption is not supported in %"TS". These files\n"
1870 " will not be extracted.",
1871 required_features->encrypted_files, loc, mode);
1874 if (required_features->encrypted_directories &&
1875 !supported_features->encrypted_directories)
1878 "%lu directories in %"TS" are marked as being encrypted,\n"
1879 " but encryption is not supported in %"TS".\n"
1880 " These directories will be extracted as unencrypted.",
1881 required_features->encrypted_directories, loc, mode);
1884 if (required_features->not_context_indexed_files &&
1885 !supported_features->not_context_indexed_files)
1888 "%lu files in %"TS" are marked as not content indexed,\n"
1889 " but this attribute is not supported in %"TS".",
1890 required_features->not_context_indexed_files, loc, mode);
1893 if (required_features->sparse_files && !supported_features->sparse_files)
1896 "%lu files in %"TS" are marked as sparse, but creating\n"
1897 " sparse files is not supported in %"TS". These files\n"
1898 " will be extracted as non-sparse.",
1899 required_features->sparse_files, loc, mode);
1902 if (required_features->named_data_streams &&
1903 !supported_features->named_data_streams)
1906 "%lu files in %"TS" contain one or more alternate (named)\n"
1907 " data streams, which are not supported in %"TS".\n"
1908 " Alternate data streams will NOT be extracted.",
1909 required_features->named_data_streams, loc, mode);
1912 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1913 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1914 required_features->named_data_streams &&
1915 supported_features->named_data_streams)
1918 "%lu files in %"TS" contain one or more alternate (named)\n"
1919 " data streams, which are not supported in linked extraction mode.\n"
1920 " Alternate data streams will NOT be extracted.",
1921 required_features->named_data_streams, loc);
1924 if (required_features->hard_links && !supported_features->hard_links)
1927 "%lu files in %"TS" are hard links, but hard links are\n"
1928 " not supported in %"TS". Hard links will be extracted as\n"
1929 " duplicate copies of the linked files.",
1930 required_features->hard_links, loc, mode);
1933 if (required_features->reparse_points && !supported_features->reparse_points)
1935 if (supported_features->symlink_reparse_points) {
1936 if (required_features->other_reparse_points) {
1938 "%lu files in %"TS" are reparse points that are neither\n"
1939 " symbolic links nor junction points and are not supported in\n"
1940 " %"TS". These reparse points will not be extracted.",
1941 required_features->other_reparse_points, loc,
1946 "%lu files in %"TS" are reparse points, which are\n"
1947 " not supported in %"TS" and will not be extracted.",
1948 required_features->reparse_points, loc, mode);
1952 if (required_features->security_descriptors &&
1953 !supported_features->security_descriptors)
1956 "%lu files in %"TS" have Windows NT security descriptors,\n"
1957 " but extracting security descriptors is not supported in\n"
1958 " %"TS". No security descriptors will be extracted.",
1959 required_features->security_descriptors, loc, mode);
1962 if (required_features->short_names && !supported_features->short_names)
1965 "%lu files in %"TS" have short (DOS) names, but\n"
1966 " extracting short names is not supported in %"TS".\n"
1967 " Short names will not be extracted.\n",
1968 required_features->short_names, loc, mode);
1971 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1972 required_features->unix_data && !supported_features->unix_data)
1974 ERROR("Extracting UNIX data is not supported in %"TS, mode);
1975 return WIMLIB_ERR_UNSUPPORTED;
1977 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
1978 required_features->short_names && !supported_features->short_names)
1980 ERROR("Extracting short names is not supported in %"TS"", mode);
1981 return WIMLIB_ERR_UNSUPPORTED;
1983 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1984 !ops->set_timestamps)
1986 ERROR("Extracting timestamps is not supported in %"TS"", mode);
1987 return WIMLIB_ERR_UNSUPPORTED;
1989 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1990 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1991 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1992 required_features->security_descriptors &&
1993 !supported_features->security_descriptors)
1995 ERROR("Extracting security descriptors is not supported in %"TS, mode);
1996 return WIMLIB_ERR_UNSUPPORTED;
1999 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
2000 !supported_features->hard_links)
2002 ERROR("Hard link extraction mode requested, but "
2003 "%"TS" does not support hard links!", mode);
2004 return WIMLIB_ERR_UNSUPPORTED;
2007 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
2008 required_features->symlink_reparse_points &&
2009 !(supported_features->symlink_reparse_points ||
2010 supported_features->reparse_points))
2012 ERROR("Extracting symbolic links is not supported in %"TS, mode);
2013 return WIMLIB_ERR_UNSUPPORTED;
2016 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
2017 !supported_features->symlink_reparse_points)
2019 ERROR("Symbolic link extraction mode requested, but "
2020 "%"TS" does not support symbolic "
2022 return WIMLIB_ERR_UNSUPPORTED;
2028 * extract_tree - Extract a file or directory tree from the currently selected
2031 * @wim: WIMStruct for the WIM file, with the desired image selected
2032 * (as wim->current_image).
2035 * "Canonical" (i.e. no leading or trailing slashes, path
2036 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
2037 * extract. An empty string means the full image.
2040 * Filesystem path to extract the file or directory tree to.
2041 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
2044 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
2045 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
2046 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
2050 * If non-NULL, progress function for the extraction. The messages
2051 * that may be sent in this function are:
2053 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
2054 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
2055 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
2056 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
2057 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
2058 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
2059 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
2060 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
2061 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
2063 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
2066 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
2067 int extract_flags, wimlib_progress_func_t progress_func)
2069 struct wim_dentry *root;
2070 struct wim_features required_features;
2071 struct apply_ctx ctx;
2073 struct wim_lookup_table_entry *lte;
2075 /* Start initializing the apply_ctx. */
2076 memset(&ctx, 0, sizeof(struct apply_ctx));
2078 ctx.extract_flags = extract_flags;
2079 ctx.target = target;
2080 ctx.target_nchars = tstrlen(target);
2081 ctx.progress_func = progress_func;
2082 if (progress_func) {
2083 ctx.progress.extract.wimfile_name = wim->filename;
2084 ctx.progress.extract.image = wim->current_image;
2085 ctx.progress.extract.extract_flags = (extract_flags &
2086 WIMLIB_EXTRACT_MASK_PUBLIC);
2087 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
2088 wim->current_image);
2089 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
2090 ctx.progress.extract.target = target;
2092 INIT_LIST_HEAD(&ctx.stream_list);
2094 /* Translate the path to extract into the corresponding
2095 * `struct wim_dentry', which will be the root of the
2096 * "dentry tree" to extract. */
2097 root = get_dentry(wim, wim_source_path);
2099 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
2100 wim_source_path, wim->current_image);
2101 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
2105 ctx.extract_root = root;
2107 /* Select the appropriate apply_operations based on the
2108 * platform and extract_flags. */
2110 ctx.ops = &win32_apply_ops;
2112 ctx.ops = &unix_apply_ops;
2116 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
2117 ctx.ops = &ntfs_3g_apply_ops;
2120 /* Call the start_extract() callback. This gives the apply_operations
2121 * implementation a chance to do any setup needed to access the volume.
2122 * Furthermore, it's expected to set the supported features of this
2123 * extraction mode (ctx.supported_features), which are determined at
2124 * runtime as they may vary depending on the actual volume. These
2125 * features are then compared with the actual features extracting this
2126 * dentry tree requires. Some mismatches will merely produce warnings
2127 * and the unsupported data will be ignored; others will produce errors.
2129 ret = ctx.ops->start_extract(target, &ctx);
2133 dentry_tree_get_features(root, &required_features);
2134 ret = do_feature_check(&required_features, &ctx.supported_features,
2135 extract_flags, ctx.ops, wim_source_path);
2137 goto out_finish_or_abort_extract;
2139 ctx.supported_attributes_mask =
2140 compute_supported_attributes_mask(&ctx.supported_features);
2142 /* Figure out whether the root dentry is being extracted to the root of
2143 * a volume and therefore needs to be treated "specially", for example
2144 * not being explicitly created and not having attributes set. */
2145 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2146 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2148 /* Calculate the actual filename component of each extracted dentry. In
2149 * the process, set the dentry->extraction_skipped flag on dentries that
2150 * are being skipped for some reason (e.g. invalid filename). */
2151 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
2153 goto out_dentry_reset_needs_extraction;
2155 /* Build the list of the streams that need to be extracted and
2156 * initialize ctx.progress.extract with stream information. */
2157 ret = for_dentry_in_tree(ctx.extract_root,
2158 dentry_resolve_and_zero_lte_refcnt, &ctx);
2160 goto out_dentry_reset_needs_extraction;
2162 ret = for_dentry_in_tree(ctx.extract_root,
2163 dentry_add_streams_to_extract, &ctx);
2165 goto out_teardown_stream_list;
2167 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2168 /* When extracting from a pipe, the number of bytes of data to
2169 * extract can't be determined in the normal way (examining the
2170 * lookup table), since at this point all we have is a set of
2171 * SHA1 message digests of streams that need to be extracted.
2172 * However, we can get a reasonably accurate estimate by taking
2173 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2174 * data. This does assume that a full image is being extracted,
2175 * but currently there is no API for doing otherwise. (Also,
2176 * subtract <HARDLINKBYTES> from this if hard links are
2177 * supported by the extraction mode.) */
2178 ctx.progress.extract.total_bytes =
2179 wim_info_get_image_total_bytes(wim->wim_info,
2180 wim->current_image);
2181 if (ctx.supported_features.hard_links) {
2182 ctx.progress.extract.total_bytes -=
2183 wim_info_get_image_hard_link_bytes(wim->wim_info,
2184 wim->current_image);
2188 /* Handle the special case of extracting a file to standard
2189 * output. In that case, "root" should be a single file, not a
2190 * directory tree. (If not, extract_dentry_to_stdout() will
2191 * return an error.) */
2192 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
2193 ret = extract_dentry_to_stdout(root);
2194 goto out_teardown_stream_list;
2197 /* If a sequential extraction was specified, sort the streams to be
2198 * extracted by their position in the WIM file so that the WIM file can
2199 * be read sequentially. */
2200 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2201 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2202 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2204 ret = sort_stream_list_by_sequential_order(
2206 offsetof(struct wim_lookup_table_entry,
2209 goto out_teardown_stream_list;
2212 if (ctx.ops->realpath_works_on_nonexisting_files &&
2213 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2214 ctx.ops->requires_realtarget_in_paths))
2216 ctx.realtarget = realpath(target, NULL);
2217 if (!ctx.realtarget) {
2218 ret = WIMLIB_ERR_NOMEM;
2219 goto out_teardown_stream_list;
2221 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2224 if (progress_func) {
2225 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2226 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2230 if (!ctx.root_dentry_is_special)
2232 tchar path[ctx.ops->path_max];
2233 if (build_extraction_path(path, root, &ctx))
2235 ret = extract_inode(path, &ctx, root->d_inode);
2237 goto out_free_realtarget;
2241 /* If we need to fix up the targets of absolute symbolic links
2242 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2243 * be absolute, use realpath() (or its replacement on Windows) to get
2244 * the absolute path to the extraction target. Note that this requires
2245 * the target directory to exist, unless
2246 * realpath_works_on_nonexisting_files is set in the apply_operations.
2248 if (!ctx.realtarget &&
2249 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2250 required_features.symlink_reparse_points) ||
2251 ctx.ops->requires_realtarget_in_paths))
2253 ctx.realtarget = realpath(target, NULL);
2254 if (!ctx.realtarget) {
2255 ret = WIMLIB_ERR_NOMEM;
2256 goto out_free_realtarget;
2258 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2261 /* Finally, the important part: extract the tree of files. */
2262 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2263 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2264 /* Sequential extraction requested, so two passes are needed
2265 * (one for directory structure, one for streams.) */
2267 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2270 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2271 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2273 goto out_free_realtarget;
2276 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2278 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2279 ret = extract_streams_from_pipe(&ctx);
2281 ret = extract_stream_list(&ctx);
2283 goto out_free_realtarget;
2285 /* Sequential extraction was not requested, so we can make do
2286 * with one pass where we both create the files and extract
2289 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2291 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2293 goto out_free_realtarget;
2295 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2299 /* If the total number of bytes to extract was miscalculated, just jump
2300 * to the calculated number in order to avoid confusing the progress
2301 * function. This should only occur when extracting from a pipe. */
2302 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2304 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2305 "extracted %"PRIu64,
2306 ctx.progress.extract.total_bytes,
2307 ctx.progress.extract.completed_bytes);
2309 if (progress_func &&
2310 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2312 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2313 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2316 /* Apply security descriptors and timestamps. This is done at the end,
2317 * and in a depth-first manner, to prevent timestamps from getting
2318 * changed by subsequent extract operations and to minimize the chance
2319 * of the restored security descriptors getting in our way. */
2321 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2323 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2325 goto out_free_realtarget;
2327 if (progress_func) {
2328 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2329 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2334 out_free_realtarget:
2335 FREE(ctx.realtarget);
2336 out_teardown_stream_list:
2337 /* Free memory allocated as part of the mapping from each
2338 * wim_lookup_table_entry to the dentries that reference it. */
2339 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2340 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2341 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2342 FREE(lte->lte_dentries);
2343 out_dentry_reset_needs_extraction:
2344 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2345 out_finish_or_abort_extract:
2347 if (ctx.ops->abort_extract)
2348 ctx.ops->abort_extract(&ctx);
2350 if (ctx.ops->finish_extract)
2351 ret = ctx.ops->finish_extract(&ctx);
2357 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2358 * extract flags make sense. */
2360 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2364 /* Empty destination path? */
2365 if (cmd->fs_dest_path[0] == T('\0'))
2366 return WIMLIB_ERR_INVALID_PARAM;
2368 extract_flags = cmd->extract_flags;
2370 /* Check for invalid flag combinations */
2371 if ((extract_flags &
2372 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2373 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2374 WIMLIB_EXTRACT_FLAG_HARDLINK))
2375 return WIMLIB_ERR_INVALID_PARAM;
2377 if ((extract_flags &
2378 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2379 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2380 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2381 return WIMLIB_ERR_INVALID_PARAM;
2383 if ((extract_flags &
2384 (WIMLIB_EXTRACT_FLAG_RPFIX |
2385 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2386 WIMLIB_EXTRACT_FLAG_NORPFIX))
2387 return WIMLIB_ERR_INVALID_PARAM;
2389 if ((extract_flags &
2390 (WIMLIB_EXTRACT_FLAG_RESUME |
2391 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2392 return WIMLIB_ERR_INVALID_PARAM;
2394 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2395 #ifndef WITH_NTFS_3G
2396 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2397 " we cannot apply a WIM image directly to a NTFS volume.");
2398 return WIMLIB_ERR_UNSUPPORTED;
2402 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2403 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2405 /* Do reparse point fixups by default if the WIM header says
2406 * they are enabled and we are extracting a full image. */
2407 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2408 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2411 /* TODO: Since UNIX data entries are stored in the file resources, in a
2412 * completely sequential extraction they may come up before the
2413 * corresponding file or symbolic link data. This needs to be handled
2415 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2416 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2417 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2418 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2420 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2421 WARNING("Setting UNIX file/owner group may "
2422 "be impossible on some\n"
2424 "when applying from a pipe.");
2426 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2427 WARNING("Disabling sequential extraction for "
2432 cmd->extract_flags = extract_flags;
2437 /* Internal function to execute extraction commands for a WIM image. The paths
2438 * in the extract commands are expected to be already "canonicalized". */
2440 do_wimlib_extract_files(WIMStruct *wim,
2442 struct wimlib_extract_command *cmds,
2444 wimlib_progress_func_t progress_func)
2447 bool found_link_cmd = false;
2448 bool found_nolink_cmd = false;
2450 /* Select the image from which we are extracting files */
2451 ret = select_wim_image(wim, image);
2455 /* Make sure there are no streams in the WIM that have not been
2456 * checksummed yet. */
2457 ret = wim_checksum_unhashed_streams(wim);
2461 /* Check for problems with the extraction commands */
2462 for (size_t i = 0; i < num_cmds; i++) {
2463 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2466 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2467 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2468 found_link_cmd = true;
2470 found_nolink_cmd = true;
2472 if (found_link_cmd && found_nolink_cmd) {
2473 ERROR("Symlink or hardlink extraction mode must "
2474 "be set on all extraction commands");
2475 return WIMLIB_ERR_INVALID_PARAM;
2479 /* Execute the extraction commands */
2480 for (size_t i = 0; i < num_cmds; i++) {
2481 ret = extract_tree(wim,
2482 cmds[i].wim_source_path,
2483 cmds[i].fs_dest_path,
2484 cmds[i].extract_flags,
2492 /* API function documented in wimlib.h */
2494 wimlib_extract_files(WIMStruct *wim,
2496 const struct wimlib_extract_command *cmds,
2498 int default_extract_flags,
2499 WIMStruct **additional_swms,
2500 unsigned num_additional_swms,
2501 wimlib_progress_func_t progress_func)
2504 struct wimlib_extract_command *cmds_copy;
2507 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2509 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2516 if (num_additional_swms)
2517 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2519 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2521 ret = WIMLIB_ERR_NOMEM;
2522 goto out_restore_lookup_table;
2525 for (size_t i = 0; i < num_cmds; i++) {
2526 cmds_copy[i].extract_flags = (default_extract_flags |
2527 cmds[i].extract_flags)
2528 & WIMLIB_EXTRACT_MASK_PUBLIC;
2529 all_flags |= cmds_copy[i].extract_flags;
2531 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
2532 if (!cmds_copy[i].wim_source_path) {
2533 ret = WIMLIB_ERR_NOMEM;
2534 goto out_free_cmds_copy;
2537 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
2538 if (!cmds_copy[i].fs_dest_path) {
2539 ret = WIMLIB_ERR_NOMEM;
2540 goto out_free_cmds_copy;
2544 ret = do_wimlib_extract_files(wim, image,
2545 cmds_copy, num_cmds,
2548 if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2549 WIMLIB_EXTRACT_FLAG_HARDLINK))
2551 for_lookup_table_entry(wim->lookup_table,
2552 lte_free_extracted_file, NULL);
2555 for (size_t i = 0; i < num_cmds; i++) {
2556 FREE(cmds_copy[i].wim_source_path);
2557 FREE(cmds_copy[i].fs_dest_path);
2560 out_restore_lookup_table:
2561 if (num_additional_swms)
2562 unmerge_lookup_table(wim);
2568 * Extracts an image from a WIM file.
2570 * @wim: WIMStruct for the WIM file.
2572 * @image: Number of the single image to extract.
2574 * @target: Directory or NTFS volume to extract the image to.
2576 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2578 * @progress_func: If non-NULL, a progress function to be called
2581 * Returns 0 on success; nonzero on failure.
2584 extract_single_image(WIMStruct *wim, int image,
2585 const tchar *target, int extract_flags,
2586 wimlib_progress_func_t progress_func)
2589 tchar *target_copy = canonicalize_fs_path(target);
2591 return WIMLIB_ERR_NOMEM;
2592 struct wimlib_extract_command cmd = {
2593 .wim_source_path = T(""),
2594 .fs_dest_path = target_copy,
2595 .extract_flags = extract_flags,
2597 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2602 static const tchar * const filename_forbidden_chars =
2611 /* This function checks if it is okay to use a WIM image's name as a directory
2614 image_name_ok_as_dir(const tchar *image_name)
2616 return image_name && *image_name &&
2617 !tstrpbrk(image_name, filename_forbidden_chars) &&
2618 tstrcmp(image_name, T(".")) &&
2619 tstrcmp(image_name, T(".."));
2622 /* Extracts all images from the WIM to the directory @target, with the images
2623 * placed in subdirectories named by their image names. */
2625 extract_all_images(WIMStruct *wim,
2626 const tchar *target,
2628 wimlib_progress_func_t progress_func)
2630 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2631 size_t output_path_len = tstrlen(target);
2632 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2635 const tchar *image_name;
2638 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2640 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2641 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2642 return WIMLIB_ERR_INVALID_PARAM;
2645 if (tstat(target, &stbuf)) {
2646 if (errno == ENOENT) {
2647 if (tmkdir(target, 0755)) {
2648 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2649 return WIMLIB_ERR_MKDIR;
2652 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2653 return WIMLIB_ERR_STAT;
2655 } else if (!S_ISDIR(stbuf.st_mode)) {
2656 ERROR("\"%"TS"\" is not a directory", target);
2657 return WIMLIB_ERR_NOTDIR;
2660 tmemcpy(buf, target, output_path_len);
2661 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2662 for (image = 1; image <= wim->hdr.image_count; image++) {
2663 image_name = wimlib_get_image_name(wim, image);
2664 if (image_name_ok_as_dir(image_name)) {
2665 tstrcpy(buf + output_path_len + 1, image_name);
2667 /* Image name is empty or contains forbidden characters.
2668 * Use image number instead. */
2669 tsprintf(buf + output_path_len + 1, T("%d"), image);
2671 ret = extract_single_image(wim, image, buf, extract_flags,
2680 do_wimlib_extract_image(WIMStruct *wim,
2682 const tchar *target,
2684 WIMStruct **additional_swms,
2685 unsigned num_additional_swms,
2686 wimlib_progress_func_t progress_func)
2690 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2691 wimlib_assert(wim->hdr.part_number == 1);
2692 wimlib_assert(num_additional_swms == 0);
2694 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2698 if (num_additional_swms)
2699 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2702 if (image == WIMLIB_ALL_IMAGES) {
2703 ret = extract_all_images(wim, target, extract_flags,
2706 ret = extract_single_image(wim, image, target, extract_flags,
2710 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2711 WIMLIB_EXTRACT_FLAG_HARDLINK))
2713 for_lookup_table_entry(wim->lookup_table,
2714 lte_free_extracted_file,
2717 if (num_additional_swms)
2718 unmerge_lookup_table(wim);
2722 /* API function documented in wimlib.h */
2724 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2725 const tchar *target, int extract_flags,
2726 wimlib_progress_func_t progress_func)
2730 struct filedes *in_fd;
2734 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2736 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2737 return WIMLIB_ERR_INVALID_PARAM;
2739 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2741 /* Read the WIM header from the pipe and get a WIMStruct to represent
2742 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2743 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2744 * an empty lookup table, no XML data read, and no filename set. */
2745 ret = open_wim_as_WIMStruct(&pipe_fd,
2746 WIMLIB_OPEN_FLAG_FROM_PIPE |
2747 WIMLIB_OPEN_FLAG_SPLIT_OK,
2748 &pwm, progress_func);
2752 /* Sanity check to make sure this is a pipable WIM. */
2753 if (pwm->hdr.magic != PWM_MAGIC) {
2754 ERROR("The WIM being read from file descriptor %d "
2755 "is not pipable!", pipe_fd);
2756 ret = WIMLIB_ERR_NOT_PIPABLE;
2757 goto out_wimlib_free;
2760 /* Sanity check to make sure the first part of a pipable split WIM is
2761 * sent over the pipe first. */
2762 if (pwm->hdr.part_number != 1) {
2763 ERROR("The first part of the split WIM must be "
2764 "sent over the pipe first.");
2765 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2766 goto out_wimlib_free;
2769 in_fd = &pwm->in_fd;
2770 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2772 /* As mentioned, the WIMStruct we created from the pipe does not have
2773 * XML data yet. Fix this by reading the extra copy of the XML data
2774 * that directly follows the header in pipable WIMs. (Note: see
2775 * write_pipable_wim() for more details about the format of pipable
2778 struct wim_lookup_table_entry xml_lte;
2779 ret = read_pwm_stream_header(pwm, &xml_lte, 0, NULL);
2781 goto out_wimlib_free;
2783 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2785 ERROR("Expected XML data, but found non-metadata "
2787 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2788 goto out_wimlib_free;
2791 copy_resource_entry(&pwm->hdr.xml_res_entry,
2792 &xml_lte.resource_entry);
2794 ret = read_wim_xml_data(pwm);
2796 goto out_wimlib_free;
2797 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2798 ERROR("Image count in XML data is not the same as in WIM header.");
2799 ret = WIMLIB_ERR_XML;
2800 goto out_wimlib_free;
2804 /* Get image index (this may use the XML data that was just read to
2805 * resolve an image name). */
2806 if (image_num_or_name) {
2807 image = wimlib_resolve_image(pwm, image_num_or_name);
2808 if (image == WIMLIB_NO_IMAGE) {
2809 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2811 ret = WIMLIB_ERR_INVALID_IMAGE;
2812 goto out_wimlib_free;
2813 } else if (image == WIMLIB_ALL_IMAGES) {
2814 ERROR("Applying all images from a pipe is not supported.");
2815 ret = WIMLIB_ERR_INVALID_IMAGE;
2816 goto out_wimlib_free;
2819 if (pwm->hdr.image_count != 1) {
2820 ERROR("No image was specified, but the pipable WIM "
2821 "did not contain exactly 1 image");
2822 ret = WIMLIB_ERR_INVALID_IMAGE;
2823 goto out_wimlib_free;
2828 /* Load the needed metadata resource. */
2829 for (i = 1; i <= pwm->hdr.image_count; i++) {
2830 struct wim_lookup_table_entry *metadata_lte;
2831 struct wim_image_metadata *imd;
2833 metadata_lte = new_lookup_table_entry();
2834 if (!metadata_lte) {
2835 ret = WIMLIB_ERR_NOMEM;
2836 goto out_wimlib_free;
2839 ret = read_pwm_stream_header(pwm, metadata_lte, 0, NULL);
2840 imd = pwm->image_metadata[i - 1];
2841 imd->metadata_lte = metadata_lte;
2843 goto out_wimlib_free;
2845 if (!(metadata_lte->resource_entry.flags &
2846 WIM_RESHDR_FLAG_METADATA))
2848 ERROR("Expected metadata resource, but found "
2849 "non-metadata stream.");
2850 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2851 goto out_wimlib_free;
2855 /* Metadata resource is for the images being extracted.
2856 * Parse it and save the metadata in memory. */
2857 ret = read_metadata_resource(pwm, imd);
2859 goto out_wimlib_free;
2862 /* Metadata resource is not for the image being
2863 * extracted. Skip over it. */
2864 ret = skip_pwm_stream(metadata_lte);
2866 goto out_wimlib_free;
2869 /* Extract the image. */
2870 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2871 ret = do_wimlib_extract_image(pwm, image, target,
2872 extract_flags, NULL, 0, progress_func);
2873 /* Clean up and return. */
2879 /* API function documented in wimlib.h */
2881 wimlib_extract_image(WIMStruct *wim,
2883 const tchar *target,
2885 WIMStruct **additional_swms,
2886 unsigned num_additional_swms,
2887 wimlib_progress_func_t progress_func)
2889 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2890 return do_wimlib_extract_image(wim, image, target, extract_flags,
2891 additional_swms, num_additional_swms,