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 /* Tally the size only for each extraction of the stream (not hard
128 if (!(dentry->d_inode->i_visited &&
129 ctx->supported_features.hard_links) &&
130 (!is_linked_extraction(ctx) || (lte->out_refcnt == 0 &&
131 lte->extracted_file == NULL)))
133 ctx->progress.extract.total_bytes += wim_resource_size(lte);
134 ctx->progress.extract.num_streams++;
137 /* Add stream to the extraction_list only one time, even if it's going
138 * to be extracted to multiple locations. */
139 if (lte->out_refcnt == 0) {
140 list_add_tail(<e->extraction_list, &ctx->stream_list);
141 ctx->num_streams_remaining++;
144 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
145 struct wim_dentry **lte_dentries;
147 /* Append dentry to this stream's array of dentries referencing
148 * it. Use inline array to avoid memory allocation until the
149 * number of dentries becomes too large. */
150 if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) {
151 lte_dentries = lte->inline_lte_dentries;
153 struct wim_dentry **prev_lte_dentries;
154 size_t alloc_lte_dentries;
156 if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) {
157 prev_lte_dentries = NULL;
158 alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries);
160 prev_lte_dentries = lte->lte_dentries;
161 alloc_lte_dentries = lte->alloc_lte_dentries;
164 if (lte->out_refcnt == alloc_lte_dentries) {
165 alloc_lte_dentries *= 2;
166 lte_dentries = REALLOC(prev_lte_dentries,
168 sizeof(lte_dentries[0]));
170 return WIMLIB_ERR_NOMEM;
171 if (prev_lte_dentries == NULL) {
173 lte->inline_lte_dentries,
174 sizeof(lte->inline_lte_dentries));
176 lte->lte_dentries = lte_dentries;
177 lte->alloc_lte_dentries = alloc_lte_dentries;
179 lte_dentries = lte->lte_dentries;
181 lte_dentries[lte->out_refcnt] = dentry;
187 /* Given a WIM dentry in the tree to be extracted, iterate through streams that
188 * need to be extracted. For each one, add it to the list of streams to be
189 * extracted (ctx->stream_list) if not already done so, and also update the
190 * progress information (ctx->progress) with the stream. Furthermore, if doing
191 * a sequential extraction, build a mapping from each the stream to the dentries
194 * This uses the i_visited member of the inodes (assumed to be 0 initially). */
196 dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx)
198 struct apply_ctx *ctx = _ctx;
199 struct wim_inode *inode = dentry->d_inode;
202 /* Don't process dentries marked as skipped. */
203 if (dentry->extraction_skipped)
206 /* The unnamed data stream will always be extracted, except in an
208 if (!inode_is_encrypted_directory(inode)) {
209 ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode),
215 /* Named data streams will be extracted only if supported in the current
216 * extraction mode and volume, and to avoid complications, if not doing
217 * a linked extraction. */
218 if (can_extract_named_data_streams(ctx)) {
219 for (u16 i = 0; i < inode->i_num_ads; i++) {
220 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
222 ret = ref_stream_to_extract(inode->i_ads_entries[i].lte,
228 inode->i_visited = 1;
232 /* Inform library user of progress of stream extraction following the successful
233 * extraction of a copy of the stream specified by @lte. */
235 update_extract_progress(struct apply_ctx *ctx,
236 const struct wim_lookup_table_entry *lte)
238 wimlib_progress_func_t progress_func = ctx->progress_func;
239 union wimlib_progress_info *progress = &ctx->progress;
241 progress->extract.completed_bytes += wim_resource_size(lte);
243 progress->extract.completed_bytes >= ctx->next_progress)
245 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress);
246 if (progress->extract.completed_bytes >=
247 progress->extract.total_bytes)
249 ctx->next_progress = ~0ULL;
251 ctx->next_progress += progress->extract.total_bytes / 128;
252 if (ctx->next_progress > progress->extract.total_bytes)
253 ctx->next_progress = progress->extract.total_bytes;
259 /* Extract a symbolic link (not directly as reparse data), handling fixing up
260 * the target of absolute symbolic links and updating the extract progress.
262 * @inode must specify the WIM inode for a symbolic link or junction reparse
265 * @lte_override overrides the resource used as the reparse data for the
268 extract_symlink(const tchar *path, struct apply_ctx *ctx,
269 struct wim_inode *inode,
270 struct wim_lookup_table_entry *lte_override)
272 ssize_t bufsize = ctx->ops->path_max;
273 tchar target[bufsize];
279 /* If absolute symbolic link fixups requested, reserve space in the link
280 * target buffer for the absolute path of the target directory. */
281 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
283 buf += ctx->realtarget_nchars;
284 bufsize -= ctx->realtarget_nchars;
287 /* Translate the WIM inode's reparse data into the link target. */
288 sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override);
291 return WIMLIB_ERR_READLINK;
295 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
298 /* Fix absolute symbolic link target to point into the
299 * actual extraction destination. */
300 tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars);
301 fixed_target = target;
303 /* Keep same link target. */
307 /* Call into the apply_operations to create the symbolic link. */
308 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
310 ret = ctx->ops->create_symlink(fixed_target, path, ctx);
312 ERROR_WITH_ERRNO("Failed to create symlink "
313 "\"%"TS"\" => \"%"TS"\"", path, fixed_target);
317 /* Account for reparse data consumed. */
318 update_extract_progress(ctx,
319 (lte_override ? lte_override :
320 inode_unnamed_lte_resolved(inode)));
323 #endif /* !__WIN32__ */
325 /* Create a file, directory, or symbolic link. */
327 extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode)
332 if (ctx->supported_features.symlink_reparse_points &&
333 !ctx->supported_features.reparse_points &&
334 inode_is_symlink(inode))
336 ret = extract_symlink(path, ctx, inode, NULL);
338 #endif /* !__WIN32__ */
339 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
340 ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie);
342 ERROR_WITH_ERRNO("Failed to create the directory "
345 } else if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) &&
346 ctx->ops->extract_encrypted_stream_creates_file &&
347 ctx->supported_features.encrypted_files) {
348 ret = ctx->ops->extract_encrypted_stream(
349 path, inode_unnamed_lte_resolved(inode), ctx);
351 ERROR_WITH_ERRNO("Failed to create and extract "
352 "encrypted file \"%"TS"\"", path);
355 ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie);
357 ERROR_WITH_ERRNO("Failed to create the file "
365 extract_hardlink(const tchar *oldpath, const tchar *newpath,
366 struct apply_ctx *ctx)
370 DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath);
371 ret = ctx->ops->create_hardlink(oldpath, newpath, ctx);
373 ERROR_WITH_ERRNO("Failed to create hardlink "
374 "\"%"TS"\" => \"%"TS"\"",
382 try_extract_rpfix(u8 *rpbuf,
384 const wchar_t *extract_root_realpath,
385 unsigned extract_root_realpath_nchars)
387 struct reparse_data rpdata;
389 size_t target_nchars;
390 size_t stripped_nchars;
391 wchar_t *stripped_target;
392 wchar_t stripped_target_nchars;
395 utf16lechar *new_target;
396 utf16lechar *new_print_name;
397 size_t new_target_nchars;
398 size_t new_print_name_nchars;
401 ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata);
405 if (extract_root_realpath[0] == L'\0' ||
406 extract_root_realpath[1] != L':' ||
407 extract_root_realpath[2] != L'\\')
408 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
410 ret = parse_substitute_name(rpdata.substitute_name,
411 rpdata.substitute_name_nbytes,
415 stripped_nchars = ret;
416 target = rpdata.substitute_name;
417 target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar);
418 stripped_target = target + stripped_nchars;
419 stripped_target_nchars = target_nchars - stripped_nchars;
421 new_target = alloca((6 + extract_root_realpath_nchars +
422 stripped_target_nchars) * sizeof(utf16lechar));
425 if (stripped_nchars == 6) {
426 /* Include \??\ prefix if it was present before */
427 p = wmempcpy(p, L"\\??\\", 4);
430 /* Print name excludes the \??\ if present. */
432 if (stripped_nchars != 0) {
433 /* Get drive letter from real path to extract root, if a drive
434 * letter was present before. */
435 *p++ = extract_root_realpath[0];
436 *p++ = extract_root_realpath[1];
438 /* Copy the rest of the extract root */
439 p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2);
441 /* Append the stripped target */
442 p = wmempcpy(p, stripped_target, stripped_target_nchars);
443 new_target_nchars = p - new_target;
444 new_print_name_nchars = p - new_print_name;
446 if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE ||
447 new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE)
448 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
450 rpdata.substitute_name = new_target;
451 rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar);
452 rpdata.print_name = new_print_name;
453 rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar);
454 return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p);
456 #endif /* __WIN32__ */
458 /* Set reparse data on extracted file or directory that has
459 * FILE_ATTRIBUTE_REPARSE_POINT set. */
461 extract_reparse_data(const tchar *path, struct apply_ctx *ctx,
462 struct wim_inode *inode,
463 struct wim_lookup_table_entry *lte_override)
466 u8 rpbuf[REPARSE_POINT_MAX_SIZE];
469 ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override);
474 /* Fix up target of absolute symbolic link or junction points so
475 * that they point into the actual extraction target. */
476 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
477 (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
478 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) &&
479 !inode->i_not_rpfixed)
481 ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget,
482 ctx->realtarget_nchars);
483 if (ret && !(ctx->extract_flags &
484 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
486 WARNING("Reparse point fixup of \"%"TS"\" "
495 ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx);
497 /* On Windows, the SeCreateSymbolicLink privilege is required to create
498 * symbolic links. To be more friendly towards non-Administrator users,
499 * we merely warn the user if symbolic links cannot be created due to
500 * insufficient permissions or privileges, unless
501 * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided. */
503 if (ret && inode_is_symlink(inode) &&
504 (errno == EACCES || errno == EPERM) &&
505 !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
507 WARNING("Can't set reparse data on \"%"TS"\": "
509 " You may be trying to "
510 "extract a symbolic link without the\n"
511 " SeCreateSymbolicLink privilege, "
512 "which by default non-Administrator\n"
513 " accounts do not have.",
521 /* Account for reparse data consumed. */
522 update_extract_progress(ctx,
523 (lte_override ? lte_override :
524 inode_unnamed_lte_resolved(inode)));
528 ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path);
533 * Extract zero or more streams to a file.
535 * This function operates slightly differently depending on whether @lte_spec is
536 * NULL or not. When @lte_spec is NULL, the behavior is to extract the default
537 * file contents (unnamed stream), and, if named data streams are supported in
538 * the extract mode and volume, any named data streams. When @lte_spec is NULL,
539 * the behavior is to extract only all copies of the stream @lte_spec, and in
540 * addition use @lte_spec to set the reparse data or create the symbolic link if
544 * Path to file to extract (as can be passed to apply_operations
549 * WIM dentry that corresponds to the file being extracted.
551 * If non-NULL, specifies the lookup table entry for a stream to extract,
552 * and only that stream will be extracted (although there may be more than
553 * one instance of it).
555 * Used only if @lte_spec != NULL; it is passed to the extraction functions
556 * rather than @lte_spec, allowing the location of the stream to be
557 * overridden. (This is used when the WIM is being read from a nonseekable
558 * file, such as a pipe, when streams need to be used more than once; each
559 * such stream is extracted to a temporary file.)
562 extract_streams(const tchar *path, struct apply_ctx *ctx,
563 struct wim_dentry *dentry,
564 struct wim_lookup_table_entry *lte_spec,
565 struct wim_lookup_table_entry *lte_override)
567 struct wim_inode *inode = dentry->d_inode;
568 struct wim_lookup_table_entry *lte;
569 file_spec_t file_spec;
572 if (dentry->was_hardlinked)
579 for (unsigned i = 0; i < SHA1_HASH_SIZE; i++)
580 p += sprintf(p, "%02x", lte_override->hash[i]);
581 DEBUG("Extracting stream SHA1=%s to \"%"TS"\"",
582 sha1_str, path, inode->i_ino);
584 DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino);
588 if (ctx->ops->uses_cookies)
589 file_spec.cookie = inode->extract_cookie;
591 file_spec.path = path;
593 /* Unnamed data stream. */
594 lte = inode_unnamed_lte_resolved(inode);
595 if (lte && (!lte_spec || lte == lte_spec)) {
598 if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
599 FILE_ATTRIBUTE_REPARSE_POINT)))
601 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED &&
602 ctx->supported_features.encrypted_files) {
603 if (!ctx->ops->extract_encrypted_stream_creates_file) {
604 ret = ctx->ops->extract_encrypted_stream(
610 ret = ctx->ops->extract_unnamed_stream(
611 file_spec, lte, ctx);
615 update_extract_progress(ctx, lte);
617 else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
620 if (ctx->supported_features.reparse_points)
621 ret = extract_reparse_data(path, ctx, inode, lte);
623 else if ((inode_is_symlink(inode) &&
624 ctx->supported_features.symlink_reparse_points))
625 ret = extract_symlink(path, ctx, inode, lte);
632 /* Named data streams. */
633 if (can_extract_named_data_streams(ctx)) {
634 for (u16 i = 0; i < inode->i_num_ads; i++) {
635 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
637 if (!ads_entry_is_named_stream(entry))
642 if (lte_spec && lte_spec != lte)
646 ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name,
647 entry->stream_name_nbytes / 2,
651 update_extract_progress(ctx, lte);
657 ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path);
661 /* Set attributes on an extracted file or directory if supported by the
662 * extraction mode. */
664 extract_file_attributes(const tchar *path, struct apply_ctx *ctx,
665 struct wim_dentry *dentry, unsigned pass)
669 if (ctx->ops->set_file_attributes &&
670 !(dentry == ctx->extract_root && ctx->root_dentry_is_special)) {
671 u32 attributes = dentry->d_inode->i_attributes;
673 /* Clear unsupported attributes. */
674 attributes &= ctx->supported_attributes_mask;
676 if ((attributes & FILE_ATTRIBUTE_DIRECTORY &&
677 !ctx->supported_features.encrypted_directories) ||
678 (!(attributes & FILE_ATTRIBUTE_DIRECTORY) &&
679 !ctx->supported_features.encrypted_files))
681 attributes &= ~FILE_ATTRIBUTE_ENCRYPTED;
685 attributes = FILE_ATTRIBUTE_NORMAL;
687 ret = ctx->ops->set_file_attributes(path, attributes, ctx, pass);
689 ERROR_WITH_ERRNO("Failed to set attributes on "
698 /* Set or remove the short (DOS) name on an extracted file or directory if
699 * supported by the extraction mode. Since DOS names are unimportant and it's
700 * easy to run into problems setting them on Windows (SetFileShortName()
701 * requires SE_RESTORE privilege, which only the Administrator can request, and
702 * also requires DELETE access to the file), failure is ignored unless
703 * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set. */
705 extract_short_name(const tchar *path, struct apply_ctx *ctx,
706 struct wim_dentry *dentry)
710 /* The root of the dentry tree being extracted may not be extracted to
711 * its original name, so its short name should be ignored. */
712 if (dentry == ctx->extract_root)
715 if (ctx->supported_features.short_names) {
716 ret = ctx->ops->set_short_name(path,
718 dentry->short_name_nbytes / 2,
720 if (ret && (ctx->extract_flags &
721 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES))
723 ERROR_WITH_ERRNO("Failed to set short name of "
731 /* Set security descriptor, UNIX data, or neither on an extracted file, taking
732 * into account the current extraction mode and flags. */
734 extract_security(const tchar *path, struct apply_ctx *ctx,
735 struct wim_dentry *dentry)
738 struct wim_inode *inode = dentry->d_inode;
740 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS)
743 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
747 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
748 struct wimlib_unix_data data;
750 ret = inode_get_unix_data(inode, &data, NULL);
754 ret = ctx->ops->set_unix_data(path, &data, ctx);
756 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
757 ERROR_WITH_ERRNO("Failed to set UNIX owner, "
758 "group, and/or mode on "
762 WARNING_WITH_ERRNO("Failed to set UNIX owner, "
763 "group, and/or/mode on "
769 #endif /* __WIN32__ */
770 if (ctx->supported_features.security_descriptors &&
771 inode->i_security_id != -1)
773 const struct wim_security_data *sd;
777 sd = wim_const_security_data(ctx->wim);
778 desc = sd->descriptors[inode->i_security_id];
779 desc_size = sd->sizes[inode->i_security_id];
781 ret = ctx->ops->set_security_descriptor(path, desc,
784 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
785 ERROR_WITH_ERRNO("Failed to set security "
786 "descriptor on \"%"TS"\"", path);
790 if (errno != EACCES) {
791 WARNING_WITH_ERRNO("Failed to set "
792 "security descriptor "
793 "on \"%"TS"\"", path);
796 ctx->no_security_descriptors++;
803 /* Set timestamps on an extracted file. Failure is warning-only unless
804 * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */
806 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
807 struct wim_dentry *dentry)
809 struct wim_inode *inode = dentry->d_inode;
812 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
815 if (ctx->ops->set_timestamps) {
816 ret = ctx->ops->set_timestamps(path,
817 inode->i_creation_time,
818 inode->i_last_write_time,
819 inode->i_last_access_time,
822 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
823 ERROR_WITH_ERRNO("Failed to set timestamps "
824 "on \"%"TS"\"", path);
827 WARNING_WITH_ERRNO("Failed to set timestamps "
828 "on \"%"TS"\"", path);
835 /* Check whether the extraction of a dentry should be skipped completely. */
837 dentry_is_supported(struct wim_dentry *dentry,
838 const struct wim_features *supported_features)
840 struct wim_inode *inode = dentry->d_inode;
842 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
843 return supported_features->reparse_points ||
844 (inode_is_symlink(inode) &&
845 supported_features->symlink_reparse_points);
847 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
848 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
849 return supported_features->encrypted_directories != 0;
851 return supported_features->encrypted_files != 0;
856 /* Given a WIM dentry to extract, build the path to which to extract it, in the
857 * format understood by the callbacks in the apply_operations being used.
859 * Write the resulting path into @path, which must have room for at least
860 * ctx->ops->max_path characters including the null-terminator.
862 * Return %true if successful; %false if this WIM dentry doesn't actually need
863 * to be extracted or if the calculated path exceeds ctx->ops->max_path
866 * This function clobbers the tmp_list member of @dentry and its ancestors up
867 * until the extraction root. */
869 build_extraction_path(tchar path[], struct wim_dentry *dentry,
870 struct apply_ctx *ctx)
873 LIST_HEAD(ancestor_list);
875 const tchar *target_prefix;
876 size_t target_prefix_nchars;
877 struct wim_dentry *d;
879 if (dentry->extraction_skipped)
882 path_nchars = ctx->ops->path_prefix_nchars;
884 if (ctx->ops->requires_realtarget_in_paths) {
885 target_prefix = ctx->realtarget;
886 target_prefix_nchars = ctx->realtarget_nchars;
887 } else if (ctx->ops->requires_target_in_paths) {
888 target_prefix = ctx->target;
889 target_prefix_nchars = ctx->target_nchars;
891 target_prefix = NULL;
892 target_prefix_nchars = 0;
894 path_nchars += target_prefix_nchars;
896 for (d = dentry; d != ctx->extract_root; d = d->parent) {
897 path_nchars += d->extraction_name_nchars + 1;
898 list_add(&d->tmp_list, &ancestor_list);
901 path_nchars++; /* null terminator */
903 if (path_nchars > ctx->ops->path_max) {
904 WARNING("\"%"TS"\": Path too long to extract",
905 dentry_full_path(dentry));
909 p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
910 p = tmempcpy(p, target_prefix, target_prefix_nchars);
911 list_for_each_entry(d, &ancestor_list, tmp_list) {
912 *p++ = ctx->ops->path_separator;
913 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
916 wimlib_assert(p - path == path_nchars);
921 get_num_path_components(const tchar *path, tchar path_separator)
923 unsigned num_components = 0;
925 /* Ignore drive letter. */
926 if (path[0] != L'\0' && path[1] == L':')
931 while (*path == path_separator)
935 while (*path && *path != path_separator)
938 return num_components;
942 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
943 struct apply_ctx *ctx, struct wim_dentry *dentry)
945 size_t num_raw_path_components;
946 const struct wim_dentry *d;
947 size_t num_target_path_components;
952 num_raw_path_components = 0;
953 for (d = dentry; d != ctx->extract_root; d = d->parent)
954 num_raw_path_components++;
956 if (ctx->ops->requires_realtarget_in_paths)
957 num_target_path_components = get_num_path_components(ctx->realtarget,
958 ctx->ops->path_separator);
959 else if (ctx->ops->requires_target_in_paths)
960 num_target_path_components = get_num_path_components(ctx->target,
961 ctx->ops->path_separator);
963 num_target_path_components = 0;
965 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
966 wimlib_assert(num_target_path_components > 0);
967 num_raw_path_components++;
968 num_target_path_components--;
971 p_old = oldpath + ctx->ops->path_prefix_nchars;
973 if (p_old[0] != L'\0' && p_old[1] == ':')
976 while (*p_old == ctx->ops->path_separator)
978 while (--num_target_path_components) {
979 while (*p_old != ctx->ops->path_separator)
981 while (*p_old == ctx->ops->path_separator)
985 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
987 p = &symlink_target[0];
988 while (num_raw_path_components--) {
991 *p++ = ctx->ops->path_separator;
994 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
995 newpath, symlink_target);
996 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
998 ERROR_WITH_ERRNO("Failed to create symlink "
999 "\"%"TS"\" => \"%"TS"\"",
1000 newpath, symlink_target);
1005 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
1006 * data streams, reparse data (including symbolic links), timestamps, and
1007 * security descriptors. Basically, everything that doesn't require reading
1008 * non-metadata resources from the WIM file and isn't delayed until the final
1011 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
1012 struct apply_ctx *ctx)
1014 struct wim_inode *inode = dentry->d_inode;
1016 const tchar *oldpath;
1018 if (unlikely(is_linked_extraction(ctx))) {
1019 struct wim_lookup_table_entry *unnamed_lte;
1021 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1022 if (unnamed_lte && unnamed_lte->extracted_file) {
1023 oldpath = unnamed_lte->extracted_file;
1024 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
1031 /* Create hard link if this dentry corresponds to an already-extracted
1033 if (inode->i_extracted_file) {
1034 oldpath = inode->i_extracted_file;
1038 /* Skip symlinks unless they can be extracted as reparse points rather
1039 * than created directly. */
1040 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
1043 /* Create this file or directory unless it's the extraction root, which
1044 * was already created if necessary. */
1045 if (dentry != ctx->extract_root) {
1046 ret = extract_inode(path, ctx, inode);
1051 /* Create empty named data streams. */
1052 if (can_extract_named_data_streams(ctx)) {
1053 for (u16 i = 0; i < inode->i_num_ads; i++) {
1054 file_spec_t file_spec;
1055 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1057 if (!ads_entry_is_named_stream(entry))
1061 if (ctx->ops->uses_cookies)
1062 file_spec.cookie = inode->extract_cookie;
1064 file_spec.path = path;
1065 ret = ctx->ops->extract_named_stream(file_spec,
1067 entry->stream_name_nbytes / 2,
1070 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1071 "empty named data stream",
1078 /* Set file attributes (if supported). */
1079 ret = extract_file_attributes(path, ctx, dentry, 0);
1083 /* Set or remove file short name (if supported). */
1084 ret = extract_short_name(path, ctx, dentry);
1088 /* If inode has multiple links and hard links are supported in this
1089 * extraction mode and volume, save the path to the extracted file in
1090 * case it's needed to create a hard link. */
1091 if (unlikely(is_linked_extraction(ctx))) {
1092 struct wim_lookup_table_entry *unnamed_lte;
1094 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1096 unnamed_lte->extracted_file = TSTRDUP(path);
1097 if (!unnamed_lte->extracted_file)
1098 return WIMLIB_ERR_NOMEM;
1100 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1101 inode->i_extracted_file = TSTRDUP(path);
1102 if (!inode->i_extracted_file)
1103 return WIMLIB_ERR_NOMEM;
1108 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1111 dentry->was_hardlinked = 1;
1115 ret = extract_hardlink(oldpath, path, ctx);
1118 dentry->was_hardlinked = 1;
1123 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1125 struct apply_ctx *ctx = _ctx;
1126 tchar path[ctx->ops->path_max];
1127 struct wim_dentry *orig_dentry;
1128 struct wim_dentry *other_dentry;
1131 /* Here we may re-order the extraction of multiple names (hard links)
1132 * for the same file in the same directory in order to ensure the short
1133 * (DOS) name is set correctly. A short name is always associated with
1134 * exactly one long name, and at least on NTFS, only one long name for a
1135 * file can have a short name associated with it. (More specifically,
1136 * there can be unlimited names in the POSIX namespace, but only one
1137 * name can be in the Win32+DOS namespace, or one name in the Win32
1138 * namespace with a corresponding name in the DOS namespace.) To ensure
1139 * the short name of a file is associated with the correct long name in
1140 * a directory, we extract the long name with a corresponding short name
1141 * before any additional names. This can affect NTFS-3g extraction
1142 * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying
1143 * the long name to associate with a short name) and may affect Win32
1144 * extraction as well (which uses SetFileShortName()). */
1146 if (dentry->skeleton_extracted)
1149 if (ctx->supported_features.short_names
1150 && !dentry_has_short_name(dentry)
1151 && !dentry->d_inode->i_dos_name_extracted)
1153 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1154 if (dentry_has_short_name(other_dentry)
1155 && !other_dentry->skeleton_extracted
1156 && other_dentry->parent == dentry->parent)
1158 DEBUG("Creating %"TS" before %"TS" "
1159 "to guarantee correct DOS name extraction",
1160 dentry_full_path(other_dentry),
1161 dentry_full_path(dentry));
1162 orig_dentry = dentry;
1163 dentry = other_dentry;
1169 if (!build_extraction_path(path, dentry, ctx))
1171 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1175 dentry->skeleton_extracted = 1;
1178 dentry = orig_dentry;
1182 dentry->d_inode->i_dos_name_extracted = 1;
1186 /* Create a file or directory, then immediately extract all streams. This
1187 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1188 * may not be read sequentially by this function. */
1190 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1192 struct apply_ctx *ctx = _ctx;
1193 tchar path[ctx->ops->path_max];
1196 ret = dentry_extract_skeleton(dentry, ctx);
1200 if (!build_extraction_path(path, dentry, ctx))
1203 return extract_streams(path, ctx, dentry, NULL, NULL);
1206 /* Extract all instances of the stream @lte that are being extracted in this
1207 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1208 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1209 * extracted multiple times, it is extracted to a temporary file first.
1211 * This is intended for use with sequential extraction of a WIM image
1212 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1214 extract_stream_instances(struct wim_lookup_table_entry *lte,
1215 struct apply_ctx *ctx, bool can_seek)
1217 struct wim_dentry **lte_dentries;
1218 struct wim_lookup_table_entry *lte_tmp = NULL;
1219 struct wim_lookup_table_entry *lte_override;
1220 tchar *stream_tmp_filename = NULL;
1221 tchar path[ctx->ops->path_max];
1225 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1226 lte_dentries = lte->inline_lte_dentries;
1228 lte_dentries = lte->lte_dentries;
1230 if (likely(can_seek || lte->out_refcnt < 2)) {
1233 /* Need to extract stream to temporary file. */
1237 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1238 if (!stream_tmp_filename) {
1239 ERROR_WITH_ERRNO("Failed to create temporary filename");
1240 ret = WIMLIB_ERR_OPEN;
1244 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1246 ret = WIMLIB_ERR_NOMEM;
1247 goto out_free_stream_tmp_filename;
1249 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1250 lte_tmp->file_on_disk = stream_tmp_filename;
1251 lte_override = lte_tmp;
1253 raw_fd = topen(stream_tmp_filename,
1254 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1256 ERROR_WITH_ERRNO("Failed to open temporary file");
1257 ret = WIMLIB_ERR_OPEN;
1258 goto out_free_lte_tmp;
1260 filedes_init(&fd, raw_fd);
1261 ret = extract_wim_resource_to_fd(lte, &fd,
1262 wim_resource_size(lte));
1263 if (filedes_close(&fd) && !ret)
1264 ret = WIMLIB_ERR_WRITE;
1266 goto out_unlink_stream_tmp_file;
1269 /* Extract all instances of the stream, reading either from the stream
1270 * in the WIM file or from the temporary file containing the stream.
1271 * dentry->tmp_flag is used to ensure that each dentry is processed only
1272 * once regardless of how many times this stream appears in the streams
1273 * of the corresponding inode. */
1274 for (i = 0; i < lte->out_refcnt; i++) {
1275 struct wim_dentry *dentry = lte_dentries[i];
1277 if (dentry->tmp_flag)
1279 if (!build_extraction_path(path, dentry, ctx))
1281 ret = extract_streams(path, ctx, dentry,
1284 goto out_clear_tmp_flags;
1285 dentry->tmp_flag = 1;
1288 out_clear_tmp_flags:
1289 for (i = 0; i < lte->out_refcnt; i++)
1290 lte_dentries[i]->tmp_flag = 0;
1291 out_unlink_stream_tmp_file:
1292 if (stream_tmp_filename)
1293 tunlink(stream_tmp_filename);
1296 out_free_stream_tmp_filename:
1297 FREE(stream_tmp_filename);
1302 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1303 * structure and empty files were already created. This relies on the
1304 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1305 * stream that was constructed earlier. Streams are extracted exactly in the
1306 * order of the stream list; however, unless the WIM's file descriptor is
1307 * detected to be non-seekable, streams may be read from the WIM file more than
1308 * one time if multiple copies need to be extracted. */
1310 extract_stream_list(struct apply_ctx *ctx)
1312 struct wim_lookup_table_entry *lte;
1316 can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
1317 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1318 ret = extract_stream_instances(lte, ctx, can_seek);
1325 #define PWM_ALLOW_WIM_HDR 0x00001
1326 #define PWM_SILENT_EOF 0x00002
1328 /* Read the header from a stream in a pipable WIM. */
1330 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1331 int flags, struct wim_header_disk *hdr_ret)
1334 struct pwm_stream_hdr stream_hdr;
1335 struct wim_header_disk pwm_hdr;
1339 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
1343 if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
1344 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
1345 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
1346 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
1350 lte->resource_location = RESOURCE_NONEXISTENT;
1351 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
1355 if (buf.stream_hdr.magic != PWM_STREAM_MAGIC) {
1356 ERROR("Data read on pipe is invalid (expected stream header).");
1357 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1360 lte->resource_entry.original_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
1361 copy_hash(lte->hash, buf.stream_hdr.hash);
1362 lte->resource_entry.flags = le32_to_cpu(buf.stream_hdr.flags);
1363 lte->resource_entry.offset = pwm->in_fd.offset;
1364 lte->resource_location = RESOURCE_IN_WIM;
1366 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1367 lte->compression_type = pwm->compression_type;
1368 lte->resource_entry.size = 0;
1370 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1371 lte->resource_entry.size = lte->resource_entry.original_size;
1373 lte->is_pipable = 1;
1377 if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1378 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1382 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1384 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1386 return read_partial_wim_resource(lte, wim_resource_size(lte),
1388 WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY,
1393 extract_streams_from_pipe(struct apply_ctx *ctx)
1395 struct wim_lookup_table_entry *found_lte;
1396 struct wim_lookup_table_entry *needed_lte;
1397 struct wim_lookup_table *lookup_table;
1398 struct wim_header_disk pwm_hdr;
1402 ret = WIMLIB_ERR_NOMEM;
1403 found_lte = new_lookup_table_entry();
1407 lookup_table = ctx->wim->lookup_table;
1408 pwm_flags = PWM_ALLOW_WIM_HDR;
1409 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1410 pwm_flags |= PWM_SILENT_EOF;
1411 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
1412 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
1413 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
1414 if (ctx->progress_func)
1415 ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1417 while (ctx->num_streams_remaining) {
1418 ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags,
1421 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1422 (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1426 goto out_free_found_lte;
1429 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1430 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1431 && (needed_lte = __lookup_resource(lookup_table, found_lte->hash))
1432 && (needed_lte->out_refcnt))
1434 copy_resource_entry(&needed_lte->resource_entry,
1435 &found_lte->resource_entry);
1436 needed_lte->resource_location = found_lte->resource_location;
1437 needed_lte->wim = found_lte->wim;
1438 needed_lte->compression_type = found_lte->compression_type;
1439 needed_lte->is_pipable = found_lte->is_pipable;
1441 ret = extract_stream_instances(needed_lte, ctx, false);
1443 goto out_free_found_lte;
1444 ctx->num_streams_remaining--;
1445 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1446 ret = skip_pwm_stream(found_lte);
1448 goto out_free_found_lte;
1450 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
1451 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
1453 if (part_number != ctx->progress.extract.part_number ||
1454 total_parts != ctx->progress.extract.total_parts ||
1455 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
1458 ctx->progress.extract.part_number = part_number;
1459 ctx->progress.extract.total_parts = total_parts;
1460 memcpy(ctx->progress.extract.guid,
1461 pwm_hdr.guid, WIM_GID_LEN);
1462 if (ctx->progress_func) {
1464 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1473 free_lookup_table_entry(found_lte);
1482 /* Finish extracting a file, directory, or symbolic link by setting file
1483 * security and timestamps. */
1485 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1487 struct apply_ctx *ctx = _ctx;
1489 tchar path[ctx->ops->path_max];
1491 if (!build_extraction_path(path, dentry, ctx))
1494 ret = extract_security(path, ctx, dentry);
1498 if (ctx->ops->requires_final_set_attributes_pass) {
1499 /* Set file attributes (if supported). */
1500 ret = extract_file_attributes(path, ctx, dentry, 1);
1505 return extract_timestamps(path, ctx, dentry);
1509 * Extract a WIM dentry to standard output.
1511 * This obviously doesn't make sense in all cases. We return an error if the
1512 * dentry does not correspond to a regular file. Otherwise we extract the
1513 * unnamed data stream only.
1516 extract_dentry_to_stdout(struct wim_dentry *dentry)
1519 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1520 FILE_ATTRIBUTE_DIRECTORY))
1522 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1523 "extracted to standard output", dentry_full_path(dentry));
1524 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1526 struct wim_lookup_table_entry *lte;
1528 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1530 struct filedes _stdout;
1531 filedes_init(&_stdout, STDOUT_FILENO);
1532 ret = extract_wim_resource_to_fd(lte, &_stdout,
1533 wim_resource_size(lte));
1540 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1542 static const utf16lechar replacement_char = cpu_to_le16('?');
1546 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1552 for (i = 0; i < num_chars; i++) {
1555 case cpu_to_le16('\\'):
1556 case cpu_to_le16(':'):
1557 case cpu_to_le16('*'):
1558 case cpu_to_le16('?'):
1559 case cpu_to_le16('"'):
1560 case cpu_to_le16('<'):
1561 case cpu_to_le16('>'):
1562 case cpu_to_le16('|'):
1564 case cpu_to_le16('/'):
1565 case cpu_to_le16('\0'):
1567 name[i] = replacement_char;
1574 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1575 name[num_chars - 1] == cpu_to_le16('.'))
1578 name[num_chars - 1] = replacement_char;
1587 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1589 const utf16lechar *file_name = dentry->file_name;
1590 return file_name != NULL &&
1591 file_name[0] == cpu_to_le16('.') &&
1592 (file_name[1] == cpu_to_le16('\0') ||
1593 (file_name[1] == cpu_to_le16('.') &&
1594 file_name[2] == cpu_to_le16('\0')));
1598 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1600 dentry->extraction_skipped = 1;
1605 * dentry_calculate_extraction_path-
1607 * Calculate the actual filename component at which a WIM dentry will be
1608 * extracted, handling invalid filenames "properly".
1610 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1611 * UNIX, converted into the platform's multibyte encoding). However, if the
1612 * file name contains characters that are not valid on the current platform or
1613 * has some other format that is not valid, leave dentry->extraction_name as
1614 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1615 * not be extracted, unless the appropriate flag
1616 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1617 * which case a substitute filename will be created and set instead.
1619 * Conflicts with case-insensitive names on Windows are handled similarly; see
1623 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1625 struct apply_ctx *ctx = _args;
1628 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1631 if (!dentry_is_supported(dentry, &ctx->supported_features))
1634 if (dentry_is_dot_or_dotdot(dentry)) {
1635 /* WIM files shouldn't contain . or .. entries. But if they are
1636 * there, don't attempt to extract them. */
1637 WARNING("Skipping extraction of unexpected . or .. file "
1638 "\"%"TS"\"", dentry_full_path(dentry));
1643 if (!ctx->ops->supports_case_sensitive_filenames)
1645 struct wim_dentry *other;
1646 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1647 case_insensitive_conflict_list)
1649 if (ctx->extract_flags &
1650 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1651 WARNING("\"%"TS"\" has the same "
1652 "case-insensitive name as "
1653 "\"%"TS"\"; extracting "
1654 "dummy name instead",
1655 dentry_full_path(dentry),
1656 dentry_full_path(other));
1659 WARNING("Not extracting \"%"TS"\": "
1660 "has same case-insensitive "
1661 "name as \"%"TS"\"",
1662 dentry_full_path(dentry),
1663 dentry_full_path(other));
1668 #else /* __WIN32__ */
1669 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1670 #endif /* !__WIN32__ */
1672 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1674 dentry->extraction_name = dentry->file_name;
1675 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1678 return utf16le_to_tstr(dentry->file_name,
1679 dentry->file_name_nbytes,
1680 &dentry->extraction_name,
1681 &dentry->extraction_name_nchars);
1684 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1686 WARNING("\"%"TS"\" has an invalid filename "
1687 "that is not supported on this platform; "
1688 "extracting dummy name instead",
1689 dentry_full_path(dentry));
1692 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1693 "that is not supported on this platform",
1694 dentry_full_path(dentry));
1701 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1703 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1704 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1707 size_t tchar_nchars;
1709 tchar_name = utf16_name_copy;
1710 tchar_nchars = dentry->file_name_nbytes / 2;
1712 ret = utf16le_to_tstr(utf16_name_copy,
1713 dentry->file_name_nbytes,
1714 &tchar_name, &tchar_nchars);
1718 size_t fixed_name_num_chars = tchar_nchars;
1719 tchar fixed_name[tchar_nchars + 50];
1721 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1722 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1723 T(" (invalid filename #%lu)"),
1724 ++ctx->invalid_sequence);
1728 dentry->extraction_name = memdup(fixed_name,
1729 2 * fixed_name_num_chars + 2);
1730 if (!dentry->extraction_name)
1731 return WIMLIB_ERR_NOMEM;
1732 dentry->extraction_name_nchars = fixed_name_num_chars;
1737 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1741 /* Clean up dentry and inode structure after extraction. */
1743 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1745 struct wim_inode *inode = dentry->d_inode;
1747 dentry->extraction_skipped = 0;
1748 dentry->was_hardlinked = 0;
1749 dentry->skeleton_extracted = 0;
1750 inode->i_visited = 0;
1751 FREE(inode->i_extracted_file);
1752 inode->i_extracted_file = NULL;
1753 inode->i_dos_name_extracted = 0;
1754 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1755 FREE(dentry->extraction_name);
1756 dentry->extraction_name = NULL;
1760 /* Tally features necessary to extract a dentry and the corresponding inode. */
1762 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1764 struct wim_features *features = _features;
1765 struct wim_inode *inode = dentry->d_inode;
1767 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1768 features->archive_files++;
1769 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1770 features->hidden_files++;
1771 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1772 features->system_files++;
1773 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1774 features->compressed_files++;
1775 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1776 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1777 features->encrypted_directories++;
1779 features->encrypted_files++;
1781 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1782 features->not_context_indexed_files++;
1783 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1784 features->sparse_files++;
1785 if (inode_has_named_stream(inode))
1786 features->named_data_streams++;
1787 if (inode->i_visited)
1788 features->hard_links++;
1789 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1790 features->reparse_points++;
1791 if (inode_is_symlink(inode))
1792 features->symlink_reparse_points++;
1794 features->other_reparse_points++;
1796 if (inode->i_security_id != -1)
1797 features->security_descriptors++;
1798 if (dentry->short_name_nbytes)
1799 features->short_names++;
1800 if (inode_has_unix_data(inode))
1801 features->unix_data++;
1802 inode->i_visited = 1;
1807 dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
1809 dentry->d_inode->i_visited = 0;
1813 /* Tally the features necessary to extract a dentry tree. */
1815 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1817 memset(features, 0, sizeof(struct wim_features));
1818 for_dentry_in_tree(root, dentry_tally_features, features);
1819 for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
1823 compute_supported_attributes_mask(const struct wim_features *supported_features)
1827 if (!supported_features->archive_files)
1828 mask &= ~FILE_ATTRIBUTE_ARCHIVE;
1830 if (!supported_features->hidden_files)
1831 mask &= ~FILE_ATTRIBUTE_HIDDEN;
1833 if (!supported_features->system_files)
1834 mask &= ~FILE_ATTRIBUTE_SYSTEM;
1836 if (!supported_features->not_context_indexed_files)
1837 mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED;
1839 if (!supported_features->compressed_files)
1840 mask &= ~FILE_ATTRIBUTE_COMPRESSED;
1842 if (!supported_features->sparse_files)
1843 mask &= ~FILE_ATTRIBUTE_SPARSE_FILE;
1845 if (!supported_features->reparse_points)
1846 mask &= ~FILE_ATTRIBUTE_REPARSE_POINT;
1852 do_feature_check(const struct wim_features *required_features,
1853 const struct wim_features *supported_features,
1855 const struct apply_operations *ops,
1856 const tchar *wim_source_path)
1859 const tchar *mode = T("this extraction mode");
1861 if (wim_source_path[0] == '\0')
1862 loc = T("the WIM image");
1864 loc = wim_source_path;
1866 /* We're an archive program, so theoretically we can do what we want
1867 * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't
1868 * bother the user about it. */
1870 if (required_features->archive_files && !supported_features->archive_files)
1873 "%lu files in %"TS" are marked as archived, but this attribute\n"
1874 " is not supported in %"TS".",
1875 required_features->archive_files, loc, mode);
1879 if (required_features->hidden_files && !supported_features->hidden_files)
1882 "%lu files in %"TS" are marked as hidden, but this\n"
1883 " attribute is not supported in %"TS".",
1884 required_features->hidden_files, loc, mode);
1887 if (required_features->system_files && !supported_features->system_files)
1890 "%lu files in %"TS" are marked as system files,\n"
1891 " but this attribute is not supported in %"TS".",
1892 required_features->system_files, loc, mode);
1895 if (required_features->compressed_files && !supported_features->compressed_files)
1898 "%lu files in %"TS" are marked as being transparently\n"
1899 " compressed, but transparent compression is not supported in\n"
1900 " %"TS". These files will be extracted as uncompressed.",
1901 required_features->compressed_files, loc, mode);
1904 if (required_features->encrypted_files && !supported_features->encrypted_files)
1907 "%lu files in %"TS" are marked as being encrypted,\n"
1908 " but encryption is not supported in %"TS". These files\n"
1909 " will not be extracted.",
1910 required_features->encrypted_files, loc, mode);
1913 if (required_features->encrypted_directories &&
1914 !supported_features->encrypted_directories)
1917 "%lu directories in %"TS" are marked as being encrypted,\n"
1918 " but encryption is not supported in %"TS".\n"
1919 " These directories will be extracted as unencrypted.",
1920 required_features->encrypted_directories, loc, mode);
1923 if (required_features->not_context_indexed_files &&
1924 !supported_features->not_context_indexed_files)
1927 "%lu files in %"TS" are marked as not content indexed,\n"
1928 " but this attribute is not supported in %"TS".",
1929 required_features->not_context_indexed_files, loc, mode);
1932 if (required_features->sparse_files && !supported_features->sparse_files)
1935 "%lu files in %"TS" are marked as sparse, but creating\n"
1936 " sparse files is not supported in %"TS". These files\n"
1937 " will be extracted as non-sparse.",
1938 required_features->sparse_files, loc, mode);
1941 if (required_features->named_data_streams &&
1942 !supported_features->named_data_streams)
1945 "%lu files in %"TS" contain one or more alternate (named)\n"
1946 " data streams, which are not supported in %"TS".\n"
1947 " Alternate data streams will NOT be extracted.",
1948 required_features->named_data_streams, loc, mode);
1951 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1952 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1953 required_features->named_data_streams &&
1954 supported_features->named_data_streams)
1957 "%lu files in %"TS" contain one or more alternate (named)\n"
1958 " data streams, which are not supported in linked extraction mode.\n"
1959 " Alternate data streams will NOT be extracted.",
1960 required_features->named_data_streams, loc);
1963 if (required_features->hard_links && !supported_features->hard_links)
1966 "%lu files in %"TS" are hard links, but hard links are\n"
1967 " not supported in %"TS". Hard links will be extracted as\n"
1968 " duplicate copies of the linked files.",
1969 required_features->hard_links, loc, mode);
1972 if (required_features->reparse_points && !supported_features->reparse_points)
1974 if (supported_features->symlink_reparse_points) {
1975 if (required_features->other_reparse_points) {
1977 "%lu files in %"TS" are reparse points that are neither\n"
1978 " symbolic links nor junction points and are not supported in\n"
1979 " %"TS". These reparse points will not be extracted.",
1980 required_features->other_reparse_points, loc,
1985 "%lu files in %"TS" are reparse points, which are\n"
1986 " not supported in %"TS" and will not be extracted.",
1987 required_features->reparse_points, loc, mode);
1991 if (required_features->security_descriptors &&
1992 !supported_features->security_descriptors)
1995 "%lu files in %"TS" have Windows NT security descriptors,\n"
1996 " but extracting security descriptors is not supported in\n"
1997 " %"TS". No security descriptors will be extracted.",
1998 required_features->security_descriptors, loc, mode);
2001 if (required_features->short_names && !supported_features->short_names)
2004 "%lu files in %"TS" have short (DOS) names, but\n"
2005 " extracting short names is not supported in %"TS".\n"
2006 " Short names will not be extracted.\n",
2007 required_features->short_names, loc, mode);
2010 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
2011 required_features->unix_data && !supported_features->unix_data)
2013 ERROR("Extracting UNIX data is not supported in %"TS, mode);
2014 return WIMLIB_ERR_UNSUPPORTED;
2016 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
2017 required_features->short_names && !supported_features->short_names)
2019 ERROR("Extracting short names is not supported in %"TS"", mode);
2020 return WIMLIB_ERR_UNSUPPORTED;
2022 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
2023 !ops->set_timestamps)
2025 ERROR("Extracting timestamps is not supported in %"TS"", mode);
2026 return WIMLIB_ERR_UNSUPPORTED;
2028 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
2029 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
2030 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
2031 required_features->security_descriptors &&
2032 !supported_features->security_descriptors)
2034 ERROR("Extracting security descriptors is not supported in %"TS, mode);
2035 return WIMLIB_ERR_UNSUPPORTED;
2038 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
2039 !supported_features->hard_links)
2041 ERROR("Hard link extraction mode requested, but "
2042 "%"TS" does not support hard links!", mode);
2043 return WIMLIB_ERR_UNSUPPORTED;
2046 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
2047 required_features->symlink_reparse_points &&
2048 !(supported_features->symlink_reparse_points ||
2049 supported_features->reparse_points))
2051 ERROR("Extracting symbolic links is not supported in %"TS, mode);
2052 return WIMLIB_ERR_UNSUPPORTED;
2055 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
2056 !supported_features->symlink_reparse_points)
2058 ERROR("Symbolic link extraction mode requested, but "
2059 "%"TS" does not support symbolic "
2061 return WIMLIB_ERR_UNSUPPORTED;
2067 do_extract_warnings(struct apply_ctx *ctx)
2069 if (ctx->partial_security_descriptors == 0 &&
2070 ctx->no_security_descriptors == 0)
2073 WARNING("Extraction of \"%"TS"\" complete, but with one or more warnings:",
2075 if (ctx->partial_security_descriptors != 0) {
2076 WARNING("- Could only partially set the security descriptor\n"
2077 " on %lu files or directories.",
2078 ctx->partial_security_descriptors);
2080 if (ctx->no_security_descriptors != 0) {
2081 WARNING("- Could not set security descriptor at all\n"
2082 " on %lu files or directories.",
2083 ctx->no_security_descriptors);
2086 WARNING("To fully restore all security descriptors, run the program\n"
2087 " with Administrator rights.");
2092 * extract_tree - Extract a file or directory tree from the currently selected
2095 * @wim: WIMStruct for the WIM file, with the desired image selected
2096 * (as wim->current_image).
2099 * "Canonical" (i.e. no leading or trailing slashes, path
2100 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
2101 * extract. An empty string means the full image.
2104 * Filesystem path to extract the file or directory tree to.
2105 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
2108 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
2109 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
2110 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
2114 * If non-NULL, progress function for the extraction. The messages
2115 * that may be sent in this function are:
2117 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
2118 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
2119 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
2120 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
2121 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
2122 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
2123 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
2124 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
2125 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
2127 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
2130 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
2131 int extract_flags, wimlib_progress_func_t progress_func)
2133 struct wim_dentry *root;
2134 struct wim_features required_features;
2135 struct apply_ctx ctx;
2137 struct wim_lookup_table_entry *lte;
2139 /* Start initializing the apply_ctx. */
2140 memset(&ctx, 0, sizeof(struct apply_ctx));
2142 ctx.extract_flags = extract_flags;
2143 ctx.target = target;
2144 ctx.target_nchars = tstrlen(target);
2145 ctx.progress_func = progress_func;
2146 if (progress_func) {
2147 ctx.progress.extract.wimfile_name = wim->filename;
2148 ctx.progress.extract.image = wim->current_image;
2149 ctx.progress.extract.extract_flags = (extract_flags &
2150 WIMLIB_EXTRACT_MASK_PUBLIC);
2151 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
2152 wim->current_image);
2153 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
2154 ctx.progress.extract.target = target;
2156 INIT_LIST_HEAD(&ctx.stream_list);
2158 /* Translate the path to extract into the corresponding
2159 * `struct wim_dentry', which will be the root of the
2160 * "dentry tree" to extract. */
2161 root = get_dentry(wim, wim_source_path);
2163 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
2164 wim_source_path, wim->current_image);
2165 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
2169 ctx.extract_root = root;
2171 /* Select the appropriate apply_operations based on the
2172 * platform and extract_flags. */
2174 ctx.ops = &win32_apply_ops;
2176 ctx.ops = &unix_apply_ops;
2180 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
2181 ctx.ops = &ntfs_3g_apply_ops;
2184 /* Call the start_extract() callback. This gives the apply_operations
2185 * implementation a chance to do any setup needed to access the volume.
2186 * Furthermore, it's expected to set the supported features of this
2187 * extraction mode (ctx.supported_features), which are determined at
2188 * runtime as they may vary depending on the actual volume. These
2189 * features are then compared with the actual features extracting this
2190 * dentry tree requires. Some mismatches will merely produce warnings
2191 * and the unsupported data will be ignored; others will produce errors.
2193 ret = ctx.ops->start_extract(target, &ctx);
2197 dentry_tree_get_features(root, &required_features);
2198 ret = do_feature_check(&required_features, &ctx.supported_features,
2199 extract_flags, ctx.ops, wim_source_path);
2201 goto out_finish_or_abort_extract;
2203 ctx.supported_attributes_mask =
2204 compute_supported_attributes_mask(&ctx.supported_features);
2206 /* Figure out whether the root dentry is being extracted to the root of
2207 * a volume and therefore needs to be treated "specially", for example
2208 * not being explicitly created and not having attributes set. */
2209 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2210 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2212 /* Calculate the actual filename component of each extracted dentry. In
2213 * the process, set the dentry->extraction_skipped flag on dentries that
2214 * are being skipped for some reason (e.g. invalid filename). */
2215 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
2217 goto out_dentry_reset_needs_extraction;
2219 /* Build the list of the streams that need to be extracted and
2220 * initialize ctx.progress.extract with stream information. */
2221 ret = for_dentry_in_tree(ctx.extract_root,
2222 dentry_resolve_and_zero_lte_refcnt, &ctx);
2224 goto out_dentry_reset_needs_extraction;
2226 ret = for_dentry_in_tree(ctx.extract_root,
2227 dentry_add_streams_to_extract, &ctx);
2229 goto out_teardown_stream_list;
2231 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2232 /* When extracting from a pipe, the number of bytes of data to
2233 * extract can't be determined in the normal way (examining the
2234 * lookup table), since at this point all we have is a set of
2235 * SHA1 message digests of streams that need to be extracted.
2236 * However, we can get a reasonably accurate estimate by taking
2237 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2238 * data. This does assume that a full image is being extracted,
2239 * but currently there is no API for doing otherwise. (Also,
2240 * subtract <HARDLINKBYTES> from this if hard links are
2241 * supported by the extraction mode.) */
2242 ctx.progress.extract.total_bytes =
2243 wim_info_get_image_total_bytes(wim->wim_info,
2244 wim->current_image);
2245 if (ctx.supported_features.hard_links) {
2246 ctx.progress.extract.total_bytes -=
2247 wim_info_get_image_hard_link_bytes(wim->wim_info,
2248 wim->current_image);
2252 /* Handle the special case of extracting a file to standard
2253 * output. In that case, "root" should be a single file, not a
2254 * directory tree. (If not, extract_dentry_to_stdout() will
2255 * return an error.) */
2256 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
2257 ret = extract_dentry_to_stdout(root);
2258 goto out_teardown_stream_list;
2261 /* If a sequential extraction was specified, sort the streams to be
2262 * extracted by their position in the WIM file so that the WIM file can
2263 * be read sequentially. */
2264 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2265 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2266 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2268 ret = sort_stream_list_by_sequential_order(
2270 offsetof(struct wim_lookup_table_entry,
2273 goto out_teardown_stream_list;
2276 if (ctx.ops->realpath_works_on_nonexisting_files &&
2277 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2278 ctx.ops->requires_realtarget_in_paths))
2280 ctx.realtarget = realpath(target, NULL);
2281 if (!ctx.realtarget) {
2282 ret = WIMLIB_ERR_NOMEM;
2283 goto out_teardown_stream_list;
2285 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2288 if (progress_func) {
2289 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2290 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2294 if (!ctx.root_dentry_is_special)
2296 tchar path[ctx.ops->path_max];
2297 if (build_extraction_path(path, root, &ctx))
2299 ret = extract_inode(path, &ctx, root->d_inode);
2301 goto out_free_realtarget;
2305 /* If we need to fix up the targets of absolute symbolic links
2306 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2307 * be absolute, use realpath() (or its replacement on Windows) to get
2308 * the absolute path to the extraction target. Note that this requires
2309 * the target directory to exist, unless
2310 * realpath_works_on_nonexisting_files is set in the apply_operations.
2312 if (!ctx.realtarget &&
2313 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2314 required_features.symlink_reparse_points) ||
2315 ctx.ops->requires_realtarget_in_paths))
2317 ctx.realtarget = realpath(target, NULL);
2318 if (!ctx.realtarget) {
2319 ret = WIMLIB_ERR_NOMEM;
2320 goto out_free_realtarget;
2322 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2325 /* Finally, the important part: extract the tree of files. */
2326 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2327 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2328 /* Sequential extraction requested, so two passes are needed
2329 * (one for directory structure, one for streams.) */
2331 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2334 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2335 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2337 goto out_free_realtarget;
2340 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2342 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2343 ret = extract_streams_from_pipe(&ctx);
2345 ret = extract_stream_list(&ctx);
2347 goto out_free_realtarget;
2349 /* Sequential extraction was not requested, so we can make do
2350 * with one pass where we both create the files and extract
2353 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2355 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2357 goto out_free_realtarget;
2359 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2363 /* If the total number of bytes to extract was miscalculated, just jump
2364 * to the calculated number in order to avoid confusing the progress
2365 * function. This should only occur when extracting from a pipe. */
2366 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2368 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2369 "extracted %"PRIu64,
2370 ctx.progress.extract.total_bytes,
2371 ctx.progress.extract.completed_bytes);
2373 if (progress_func &&
2374 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2376 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2377 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2380 /* Apply security descriptors and timestamps. This is done at the end,
2381 * and in a depth-first manner, to prevent timestamps from getting
2382 * changed by subsequent extract operations and to minimize the chance
2383 * of the restored security descriptors getting in our way. */
2385 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2387 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2389 goto out_free_realtarget;
2391 if (progress_func) {
2392 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2393 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2397 do_extract_warnings(&ctx);
2400 out_free_realtarget:
2401 FREE(ctx.realtarget);
2402 out_teardown_stream_list:
2403 /* Free memory allocated as part of the mapping from each
2404 * wim_lookup_table_entry to the dentries that reference it. */
2405 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2406 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2407 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2408 FREE(lte->lte_dentries);
2409 out_dentry_reset_needs_extraction:
2410 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2411 out_finish_or_abort_extract:
2413 if (ctx.ops->abort_extract)
2414 ctx.ops->abort_extract(&ctx);
2416 if (ctx.ops->finish_extract)
2417 ret = ctx.ops->finish_extract(&ctx);
2423 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2424 * extract flags make sense. */
2426 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2430 /* Empty destination path? */
2431 if (cmd->fs_dest_path[0] == T('\0'))
2432 return WIMLIB_ERR_INVALID_PARAM;
2434 extract_flags = cmd->extract_flags;
2436 /* Check for invalid flag combinations */
2437 if ((extract_flags &
2438 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2439 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2440 WIMLIB_EXTRACT_FLAG_HARDLINK))
2441 return WIMLIB_ERR_INVALID_PARAM;
2443 if ((extract_flags &
2444 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2445 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2446 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2447 return WIMLIB_ERR_INVALID_PARAM;
2449 if ((extract_flags &
2450 (WIMLIB_EXTRACT_FLAG_RPFIX |
2451 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2452 WIMLIB_EXTRACT_FLAG_NORPFIX))
2453 return WIMLIB_ERR_INVALID_PARAM;
2455 if ((extract_flags &
2456 (WIMLIB_EXTRACT_FLAG_RESUME |
2457 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2458 return WIMLIB_ERR_INVALID_PARAM;
2460 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2461 #ifndef WITH_NTFS_3G
2462 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2463 " we cannot apply a WIM image directly to a NTFS volume.");
2464 return WIMLIB_ERR_UNSUPPORTED;
2468 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2469 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2471 /* Do reparse point fixups by default if the WIM header says
2472 * they are enabled and we are extracting a full image. */
2473 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2474 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2477 /* TODO: Since UNIX data entries are stored in the file resources, in a
2478 * completely sequential extraction they may come up before the
2479 * corresponding file or symbolic link data. This needs to be handled
2481 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2482 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2483 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2484 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2486 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2487 WARNING("Setting UNIX file/owner group may "
2488 "be impossible on some\n"
2490 "when applying from a pipe.");
2492 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2493 WARNING("Disabling sequential extraction for "
2498 cmd->extract_flags = extract_flags;
2503 /* Internal function to execute extraction commands for a WIM image. The paths
2504 * in the extract commands are expected to be already "canonicalized". */
2506 do_wimlib_extract_files(WIMStruct *wim,
2508 struct wimlib_extract_command *cmds,
2510 wimlib_progress_func_t progress_func)
2513 bool found_link_cmd = false;
2514 bool found_nolink_cmd = false;
2516 /* Select the image from which we are extracting files */
2517 ret = select_wim_image(wim, image);
2521 /* Make sure there are no streams in the WIM that have not been
2522 * checksummed yet. */
2523 ret = wim_checksum_unhashed_streams(wim);
2527 /* Check for problems with the extraction commands */
2528 for (size_t i = 0; i < num_cmds; i++) {
2529 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2532 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2533 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2534 found_link_cmd = true;
2536 found_nolink_cmd = true;
2538 if (found_link_cmd && found_nolink_cmd) {
2539 ERROR("Symlink or hardlink extraction mode must "
2540 "be set on all extraction commands");
2541 return WIMLIB_ERR_INVALID_PARAM;
2545 /* Execute the extraction commands */
2546 for (size_t i = 0; i < num_cmds; i++) {
2547 ret = extract_tree(wim,
2548 cmds[i].wim_source_path,
2549 cmds[i].fs_dest_path,
2550 cmds[i].extract_flags,
2558 /* API function documented in wimlib.h */
2560 wimlib_extract_files(WIMStruct *wim,
2562 const struct wimlib_extract_command *cmds,
2564 int default_extract_flags,
2565 WIMStruct **additional_swms,
2566 unsigned num_additional_swms,
2567 wimlib_progress_func_t progress_func)
2570 struct wimlib_extract_command *cmds_copy;
2573 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2575 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2582 if (num_additional_swms)
2583 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2585 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2587 ret = WIMLIB_ERR_NOMEM;
2588 goto out_restore_lookup_table;
2591 for (size_t i = 0; i < num_cmds; i++) {
2592 cmds_copy[i].extract_flags = (default_extract_flags |
2593 cmds[i].extract_flags)
2594 & WIMLIB_EXTRACT_MASK_PUBLIC;
2595 all_flags |= cmds_copy[i].extract_flags;
2597 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
2598 if (!cmds_copy[i].wim_source_path) {
2599 ret = WIMLIB_ERR_NOMEM;
2600 goto out_free_cmds_copy;
2603 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
2604 if (!cmds_copy[i].fs_dest_path) {
2605 ret = WIMLIB_ERR_NOMEM;
2606 goto out_free_cmds_copy;
2610 ret = do_wimlib_extract_files(wim, image,
2611 cmds_copy, num_cmds,
2614 if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2615 WIMLIB_EXTRACT_FLAG_HARDLINK))
2617 for_lookup_table_entry(wim->lookup_table,
2618 lte_free_extracted_file, NULL);
2621 for (size_t i = 0; i < num_cmds; i++) {
2622 FREE(cmds_copy[i].wim_source_path);
2623 FREE(cmds_copy[i].fs_dest_path);
2626 out_restore_lookup_table:
2627 if (num_additional_swms)
2628 unmerge_lookup_table(wim);
2634 * Extracts an image from a WIM file.
2636 * @wim: WIMStruct for the WIM file.
2638 * @image: Number of the single image to extract.
2640 * @target: Directory or NTFS volume to extract the image to.
2642 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2644 * @progress_func: If non-NULL, a progress function to be called
2647 * Returns 0 on success; nonzero on failure.
2650 extract_single_image(WIMStruct *wim, int image,
2651 const tchar *target, int extract_flags,
2652 wimlib_progress_func_t progress_func)
2655 tchar *target_copy = canonicalize_fs_path(target);
2657 return WIMLIB_ERR_NOMEM;
2658 struct wimlib_extract_command cmd = {
2659 .wim_source_path = T(""),
2660 .fs_dest_path = target_copy,
2661 .extract_flags = extract_flags,
2663 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2668 static const tchar * const filename_forbidden_chars =
2677 /* This function checks if it is okay to use a WIM image's name as a directory
2680 image_name_ok_as_dir(const tchar *image_name)
2682 return image_name && *image_name &&
2683 !tstrpbrk(image_name, filename_forbidden_chars) &&
2684 tstrcmp(image_name, T(".")) &&
2685 tstrcmp(image_name, T(".."));
2688 /* Extracts all images from the WIM to the directory @target, with the images
2689 * placed in subdirectories named by their image names. */
2691 extract_all_images(WIMStruct *wim,
2692 const tchar *target,
2694 wimlib_progress_func_t progress_func)
2696 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2697 size_t output_path_len = tstrlen(target);
2698 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2701 const tchar *image_name;
2704 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2706 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2707 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2708 return WIMLIB_ERR_INVALID_PARAM;
2711 if (tstat(target, &stbuf)) {
2712 if (errno == ENOENT) {
2713 if (tmkdir(target, 0755)) {
2714 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2715 return WIMLIB_ERR_MKDIR;
2718 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2719 return WIMLIB_ERR_STAT;
2721 } else if (!S_ISDIR(stbuf.st_mode)) {
2722 ERROR("\"%"TS"\" is not a directory", target);
2723 return WIMLIB_ERR_NOTDIR;
2726 tmemcpy(buf, target, output_path_len);
2727 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2728 for (image = 1; image <= wim->hdr.image_count; image++) {
2729 image_name = wimlib_get_image_name(wim, image);
2730 if (image_name_ok_as_dir(image_name)) {
2731 tstrcpy(buf + output_path_len + 1, image_name);
2733 /* Image name is empty or contains forbidden characters.
2734 * Use image number instead. */
2735 tsprintf(buf + output_path_len + 1, T("%d"), image);
2737 ret = extract_single_image(wim, image, buf, extract_flags,
2746 do_wimlib_extract_image(WIMStruct *wim,
2748 const tchar *target,
2750 WIMStruct **additional_swms,
2751 unsigned num_additional_swms,
2752 wimlib_progress_func_t progress_func)
2756 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2757 wimlib_assert(wim->hdr.part_number == 1);
2758 wimlib_assert(num_additional_swms == 0);
2760 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2764 if (num_additional_swms)
2765 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2768 if (image == WIMLIB_ALL_IMAGES) {
2769 ret = extract_all_images(wim, target, extract_flags,
2772 ret = extract_single_image(wim, image, target, extract_flags,
2776 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2777 WIMLIB_EXTRACT_FLAG_HARDLINK))
2779 for_lookup_table_entry(wim->lookup_table,
2780 lte_free_extracted_file,
2783 if (num_additional_swms)
2784 unmerge_lookup_table(wim);
2788 /* API function documented in wimlib.h */
2790 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2791 const tchar *target, int extract_flags,
2792 wimlib_progress_func_t progress_func)
2796 struct filedes *in_fd;
2800 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2802 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2803 return WIMLIB_ERR_INVALID_PARAM;
2805 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2807 /* Read the WIM header from the pipe and get a WIMStruct to represent
2808 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2809 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2810 * an empty lookup table, no XML data read, and no filename set. */
2811 ret = open_wim_as_WIMStruct(&pipe_fd,
2812 WIMLIB_OPEN_FLAG_FROM_PIPE |
2813 WIMLIB_OPEN_FLAG_SPLIT_OK,
2814 &pwm, progress_func);
2818 /* Sanity check to make sure this is a pipable WIM. */
2819 if (pwm->hdr.magic != PWM_MAGIC) {
2820 ERROR("The WIM being read from file descriptor %d "
2821 "is not pipable!", pipe_fd);
2822 ret = WIMLIB_ERR_NOT_PIPABLE;
2823 goto out_wimlib_free;
2826 /* Sanity check to make sure the first part of a pipable split WIM is
2827 * sent over the pipe first. */
2828 if (pwm->hdr.part_number != 1) {
2829 ERROR("The first part of the split WIM must be "
2830 "sent over the pipe first.");
2831 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2832 goto out_wimlib_free;
2835 in_fd = &pwm->in_fd;
2836 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2838 /* As mentioned, the WIMStruct we created from the pipe does not have
2839 * XML data yet. Fix this by reading the extra copy of the XML data
2840 * that directly follows the header in pipable WIMs. (Note: see
2841 * write_pipable_wim() for more details about the format of pipable
2844 struct wim_lookup_table_entry xml_lte;
2845 ret = read_pwm_stream_header(pwm, &xml_lte, 0, NULL);
2847 goto out_wimlib_free;
2849 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2851 ERROR("Expected XML data, but found non-metadata "
2853 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2854 goto out_wimlib_free;
2857 copy_resource_entry(&pwm->hdr.xml_res_entry,
2858 &xml_lte.resource_entry);
2860 ret = read_wim_xml_data(pwm);
2862 goto out_wimlib_free;
2863 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2864 ERROR("Image count in XML data is not the same as in WIM header.");
2865 ret = WIMLIB_ERR_XML;
2866 goto out_wimlib_free;
2870 /* Get image index (this may use the XML data that was just read to
2871 * resolve an image name). */
2872 if (image_num_or_name) {
2873 image = wimlib_resolve_image(pwm, image_num_or_name);
2874 if (image == WIMLIB_NO_IMAGE) {
2875 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2877 ret = WIMLIB_ERR_INVALID_IMAGE;
2878 goto out_wimlib_free;
2879 } else if (image == WIMLIB_ALL_IMAGES) {
2880 ERROR("Applying all images from a pipe is not supported.");
2881 ret = WIMLIB_ERR_INVALID_IMAGE;
2882 goto out_wimlib_free;
2885 if (pwm->hdr.image_count != 1) {
2886 ERROR("No image was specified, but the pipable WIM "
2887 "did not contain exactly 1 image");
2888 ret = WIMLIB_ERR_INVALID_IMAGE;
2889 goto out_wimlib_free;
2894 /* Load the needed metadata resource. */
2895 for (i = 1; i <= pwm->hdr.image_count; i++) {
2896 struct wim_lookup_table_entry *metadata_lte;
2897 struct wim_image_metadata *imd;
2899 metadata_lte = new_lookup_table_entry();
2900 if (!metadata_lte) {
2901 ret = WIMLIB_ERR_NOMEM;
2902 goto out_wimlib_free;
2905 ret = read_pwm_stream_header(pwm, metadata_lte, 0, NULL);
2906 imd = pwm->image_metadata[i - 1];
2907 imd->metadata_lte = metadata_lte;
2909 goto out_wimlib_free;
2911 if (!(metadata_lte->resource_entry.flags &
2912 WIM_RESHDR_FLAG_METADATA))
2914 ERROR("Expected metadata resource, but found "
2915 "non-metadata stream.");
2916 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2917 goto out_wimlib_free;
2921 /* Metadata resource is for the images being extracted.
2922 * Parse it and save the metadata in memory. */
2923 ret = read_metadata_resource(pwm, imd);
2925 goto out_wimlib_free;
2928 /* Metadata resource is not for the image being
2929 * extracted. Skip over it. */
2930 ret = skip_pwm_stream(metadata_lte);
2932 goto out_wimlib_free;
2935 /* Extract the image. */
2936 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2937 ret = do_wimlib_extract_image(pwm, image, target,
2938 extract_flags, NULL, 0, progress_func);
2939 /* Clean up and return. */
2945 /* API function documented in wimlib.h */
2947 wimlib_extract_image(WIMStruct *wim,
2949 const tchar *target,
2951 WIMStruct **additional_swms,
2952 unsigned num_additional_swms,
2953 wimlib_progress_func_t progress_func)
2955 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2956 return do_wimlib_extract_image(wim, image, target, extract_flags,
2957 additional_swms, num_additional_swms,