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"
57 # include "wimlib/win32.h" /* for realpath() equivalent */
59 #include "wimlib/xml.h"
60 #include "wimlib/wim.h"
68 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
69 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000
70 #define WIMLIB_EXTRACT_MASK_PUBLIC 0x3fffffff
72 /* Given a WIM dentry in the tree to be extracted, resolve all streams in the
73 * corresponding inode and set 'out_refcnt' in each to 0. */
75 dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_ctx)
77 struct apply_ctx *ctx = _ctx;
78 struct wim_inode *inode = dentry->d_inode;
79 struct wim_lookup_table_entry *lte;
83 if (dentry->extraction_skipped)
86 /* Special case: when extracting from a pipe, the WIM lookup table is
87 * initially empty, so "resolving" an inode's streams is initially not
88 * possible. However, we still need to keep track of which streams,
89 * identified by SHA1 message digests, need to be extracted, so we
90 * "resolve" the inode's streams anyway by allocating new entries. */
91 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
93 ret = inode_resolve_ltes(inode, ctx->wim->lookup_table, force);
96 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
97 lte = inode_stream_lte_resolved(inode, i);
105 is_linked_extraction(const struct apply_ctx *ctx)
107 return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
108 WIMLIB_EXTRACT_FLAG_SYMLINK));
112 can_extract_named_data_streams(const struct apply_ctx *ctx)
114 return ctx->supported_features.named_data_streams &&
115 !is_linked_extraction(ctx);
119 ref_stream_to_extract(struct wim_lookup_table_entry *lte,
120 struct wim_dentry *dentry, struct apply_ctx *ctx)
125 /* Tally the size only for each extraction of the stream (not hard
127 if (!(dentry->d_inode->i_visited &&
128 ctx->supported_features.hard_links) &&
129 (!is_linked_extraction(ctx) || (lte->out_refcnt == 0 &&
130 lte->extracted_file == NULL)))
132 ctx->progress.extract.total_bytes += wim_resource_size(lte);
133 ctx->progress.extract.num_streams++;
136 /* Add stream to the extraction_list only one time, even if it's going
137 * to be extracted to multiple locations. */
138 if (lte->out_refcnt == 0) {
139 list_add_tail(<e->extraction_list, &ctx->stream_list);
140 ctx->num_streams_remaining++;
143 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
144 struct wim_dentry **lte_dentries;
146 /* Append dentry to this stream's array of dentries referencing
147 * it. Use inline array to avoid memory allocation until the
148 * number of dentries becomes too large. */
149 if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) {
150 lte_dentries = lte->inline_lte_dentries;
152 struct wim_dentry **prev_lte_dentries;
153 size_t alloc_lte_dentries;
155 if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) {
156 prev_lte_dentries = NULL;
157 alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries);
159 prev_lte_dentries = lte->lte_dentries;
160 alloc_lte_dentries = lte->alloc_lte_dentries;
163 if (lte->out_refcnt == alloc_lte_dentries) {
164 alloc_lte_dentries *= 2;
165 lte_dentries = REALLOC(prev_lte_dentries,
167 sizeof(lte_dentries[0]));
169 return WIMLIB_ERR_NOMEM;
170 if (prev_lte_dentries == NULL) {
172 lte->inline_lte_dentries,
173 sizeof(lte->inline_lte_dentries));
175 lte->lte_dentries = lte_dentries;
176 lte->alloc_lte_dentries = alloc_lte_dentries;
178 lte_dentries = lte->lte_dentries;
180 lte_dentries[lte->out_refcnt] = dentry;
186 /* Given a WIM dentry in the tree to be extracted, iterate through streams that
187 * need to be extracted. For each one, add it to the list of streams to be
188 * extracted (ctx->stream_list) if not already done so, and also update the
189 * progress information (ctx->progress) with the stream. Furthermore, if doing
190 * a sequential extraction, build a mapping from each the stream to the dentries
193 * This uses the i_visited member of the inodes (assumed to be 0 initially). */
195 dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx)
197 struct apply_ctx *ctx = _ctx;
198 struct wim_inode *inode = dentry->d_inode;
201 /* Don't process dentries marked as skipped. */
202 if (dentry->extraction_skipped)
205 /* The unnamed data stream will always be extracted, except in an
207 if (!inode_is_encrypted_directory(inode)) {
208 ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode),
214 /* Named data streams will be extracted only if supported in the current
215 * extraction mode and volume, and to avoid complications, if not doing
216 * a linked extraction. */
217 if (can_extract_named_data_streams(ctx)) {
218 for (u16 i = 0; i < inode->i_num_ads; i++) {
219 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
221 ret = ref_stream_to_extract(inode->i_ads_entries[i].lte,
227 inode->i_visited = 1;
231 /* Inform library user of progress of stream extraction following the successful
232 * extraction of a copy of the stream specified by @lte. */
234 update_extract_progress(struct apply_ctx *ctx,
235 const struct wim_lookup_table_entry *lte)
237 wimlib_progress_func_t progress_func = ctx->progress_func;
238 union wimlib_progress_info *progress = &ctx->progress;
240 progress->extract.completed_bytes += wim_resource_size(lte);
242 progress->extract.completed_bytes >= ctx->next_progress)
244 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress);
245 if (progress->extract.completed_bytes >=
246 progress->extract.total_bytes)
248 ctx->next_progress = ~0ULL;
250 ctx->next_progress += progress->extract.total_bytes / 128;
251 if (ctx->next_progress > progress->extract.total_bytes)
252 ctx->next_progress = progress->extract.total_bytes;
258 /* Extract a symbolic link (not directly as reparse data), handling fixing up
259 * the target of absolute symbolic links and updating the extract progress.
261 * @inode must specify the WIM inode for a symbolic link or junction reparse
264 * @lte_override overrides the resource used as the reparse data for the
267 extract_symlink(const tchar *path, struct apply_ctx *ctx,
268 struct wim_inode *inode,
269 struct wim_lookup_table_entry *lte_override)
271 ssize_t bufsize = ctx->ops->path_max;
272 tchar target[bufsize];
278 /* If absolute symbolic link fixups requested, reserve space in the link
279 * target buffer for the absolute path of the target directory. */
280 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
282 buf += ctx->realtarget_nchars;
283 bufsize -= ctx->realtarget_nchars;
286 /* Translate the WIM inode's reparse data into the link target. */
287 sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override);
290 return WIMLIB_ERR_READLINK;
294 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
297 /* Fix absolute symbolic link target to point into the
298 * actual extraction destination. */
299 tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars);
300 fixed_target = target;
302 /* Keep same link target. */
306 /* Call into the apply_operations to create the symbolic link. */
307 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
309 ret = ctx->ops->create_symlink(fixed_target, path, ctx);
311 ERROR_WITH_ERRNO("Failed to create symlink "
312 "\"%"TS"\" => \"%"TS"\"", path, fixed_target);
316 /* Account for reparse data consumed. */
317 update_extract_progress(ctx,
318 (lte_override ? lte_override :
319 inode_unnamed_lte_resolved(inode)));
322 #endif /* !__WIN32__ */
324 /* Create a file, directory, or symbolic link. */
326 extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode)
331 if (ctx->supported_features.symlink_reparse_points &&
332 !ctx->supported_features.reparse_points &&
333 inode_is_symlink(inode))
335 ret = extract_symlink(path, ctx, inode, NULL);
337 #endif /* !__WIN32__ */
338 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
339 ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie);
341 ERROR_WITH_ERRNO("Failed to create the directory "
344 } else if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) &&
345 ctx->ops->extract_encrypted_stream_creates_file &&
346 ctx->supported_features.encrypted_files) {
347 ret = ctx->ops->extract_encrypted_stream(
348 path, inode_unnamed_lte_resolved(inode), ctx);
350 ERROR_WITH_ERRNO("Failed to create and extract "
351 "encrypted file \"%"TS"\"", path);
354 ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie);
356 ERROR_WITH_ERRNO("Failed to create the file "
364 extract_hardlink(const tchar *oldpath, const tchar *newpath,
365 struct apply_ctx *ctx)
369 DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath);
370 ret = ctx->ops->create_hardlink(oldpath, newpath, ctx);
372 ERROR_WITH_ERRNO("Failed to create hardlink "
373 "\"%"TS"\" => \"%"TS"\"",
381 try_extract_rpfix(u8 *rpbuf,
383 const wchar_t *extract_root_realpath,
384 unsigned extract_root_realpath_nchars)
386 struct reparse_data rpdata;
388 size_t target_nchars;
389 size_t stripped_nchars;
390 wchar_t *stripped_target;
391 wchar_t stripped_target_nchars;
394 utf16lechar *new_target;
395 utf16lechar *new_print_name;
396 size_t new_target_nchars;
397 size_t new_print_name_nchars;
400 ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata);
404 if (extract_root_realpath[0] == L'\0' ||
405 extract_root_realpath[1] != L':' ||
406 extract_root_realpath[2] != L'\\')
407 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
409 ret = parse_substitute_name(rpdata.substitute_name,
410 rpdata.substitute_name_nbytes,
414 stripped_nchars = ret;
415 target = rpdata.substitute_name;
416 target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar);
417 stripped_target = target + stripped_nchars;
418 stripped_target_nchars = target_nchars - stripped_nchars;
420 new_target = alloca((6 + extract_root_realpath_nchars +
421 stripped_target_nchars) * sizeof(utf16lechar));
424 if (stripped_nchars == 6) {
425 /* Include \??\ prefix if it was present before */
426 p = wmempcpy(p, L"\\??\\", 4);
429 /* Print name excludes the \??\ if present. */
431 if (stripped_nchars != 0) {
432 /* Get drive letter from real path to extract root, if a drive
433 * letter was present before. */
434 *p++ = extract_root_realpath[0];
435 *p++ = extract_root_realpath[1];
437 /* Copy the rest of the extract root */
438 p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2);
440 /* Append the stripped target */
441 p = wmempcpy(p, stripped_target, stripped_target_nchars);
442 new_target_nchars = p - new_target;
443 new_print_name_nchars = p - new_print_name;
445 if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE ||
446 new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE)
447 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
449 rpdata.substitute_name = new_target;
450 rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar);
451 rpdata.print_name = new_print_name;
452 rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar);
453 return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p);
455 #endif /* __WIN32__ */
457 /* Set reparse data on extracted file or directory that has
458 * FILE_ATTRIBUTE_REPARSE_POINT set. */
460 extract_reparse_data(const tchar *path, struct apply_ctx *ctx,
461 struct wim_inode *inode,
462 struct wim_lookup_table_entry *lte_override)
465 u8 rpbuf[REPARSE_POINT_MAX_SIZE];
468 ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override);
473 /* Fix up target of absolute symbolic link or junction points so
474 * that they point into the actual extraction target. */
475 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
476 (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
477 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) &&
478 !inode->i_not_rpfixed)
480 ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget,
481 ctx->realtarget_nchars);
482 if (ret && !(ctx->extract_flags &
483 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
485 WARNING("Reparse point fixup of \"%"TS"\" "
494 ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx);
496 /* On Windows, the SeCreateSymbolicLink privilege is required to create
497 * symbolic links. To be more friendly towards non-Administrator users,
498 * we merely warn the user if symbolic links cannot be created due to
499 * insufficient permissions or privileges, unless
500 * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided. */
502 if (ret && inode_is_symlink(inode) &&
503 (errno == EACCES || errno == EPERM) &&
504 !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
506 WARNING("Can't set reparse data on \"%"TS"\": "
508 " You may be trying to "
509 "extract a symbolic link without the\n"
510 " SeCreateSymbolicLink privilege, "
511 "which by default non-Administrator\n"
512 " accounts do not have.",
520 /* Account for reparse data consumed. */
521 update_extract_progress(ctx,
522 (lte_override ? lte_override :
523 inode_unnamed_lte_resolved(inode)));
527 ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path);
532 * Extract zero or more streams to a file.
534 * This function operates slightly differently depending on whether @lte_spec is
535 * NULL or not. When @lte_spec is NULL, the behavior is to extract the default
536 * file contents (unnamed stream), and, if named data streams are supported in
537 * the extract mode and volume, any named data streams. When @lte_spec is NULL,
538 * the behavior is to extract only all copies of the stream @lte_spec, and in
539 * addition use @lte_spec to set the reparse data or create the symbolic link if
543 * Path to file to extract (as can be passed to apply_operations
548 * WIM dentry that corresponds to the file being extracted.
550 * If non-NULL, specifies the lookup table entry for a stream to extract,
551 * and only that stream will be extracted (although there may be more than
552 * one instance of it).
554 * Used only if @lte_spec != NULL; it is passed to the extraction functions
555 * rather than @lte_spec, allowing the location of the stream to be
556 * overridden. (This is used when the WIM is being read from a nonseekable
557 * file, such as a pipe, when streams need to be used more than once; each
558 * such stream is extracted to a temporary file.)
561 extract_streams(const tchar *path, struct apply_ctx *ctx,
562 struct wim_dentry *dentry,
563 struct wim_lookup_table_entry *lte_spec,
564 struct wim_lookup_table_entry *lte_override)
566 struct wim_inode *inode = dentry->d_inode;
567 struct wim_lookup_table_entry *lte;
568 file_spec_t file_spec;
571 if (dentry->was_hardlinked)
578 for (unsigned i = 0; i < SHA1_HASH_SIZE; i++)
579 p += sprintf(p, "%02x", lte_override->hash[i]);
580 DEBUG("Extracting stream SHA1=%s to \"%"TS"\"",
581 sha1_str, path, inode->i_ino);
583 DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino);
587 if (ctx->ops->uses_cookies)
588 file_spec.cookie = inode->extract_cookie;
590 file_spec.path = path;
592 /* Unnamed data stream. */
593 lte = inode_unnamed_lte_resolved(inode);
594 if (lte && (!lte_spec || lte == lte_spec)) {
597 if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
598 FILE_ATTRIBUTE_REPARSE_POINT)))
600 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED &&
601 ctx->supported_features.encrypted_files) {
602 if (!ctx->ops->extract_encrypted_stream_creates_file) {
603 ret = ctx->ops->extract_encrypted_stream(
609 ret = ctx->ops->extract_unnamed_stream(
610 file_spec, lte, ctx);
614 update_extract_progress(ctx, lte);
616 else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
619 if (ctx->supported_features.reparse_points)
620 ret = extract_reparse_data(path, ctx, inode, lte);
622 else if ((inode_is_symlink(inode) &&
623 ctx->supported_features.symlink_reparse_points))
624 ret = extract_symlink(path, ctx, inode, lte);
631 /* Named data streams. */
632 if (can_extract_named_data_streams(ctx)) {
633 for (u16 i = 0; i < inode->i_num_ads; i++) {
634 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
636 if (!ads_entry_is_named_stream(entry))
641 if (lte_spec && lte_spec != lte)
645 ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name,
646 entry->stream_name_nbytes / 2,
650 update_extract_progress(ctx, lte);
656 ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path);
660 /* Set attributes on an extracted file or directory if supported by the
661 * extraction mode. */
663 extract_file_attributes(const tchar *path, struct apply_ctx *ctx,
664 struct wim_dentry *dentry, unsigned pass)
668 if (ctx->ops->set_file_attributes &&
669 !(dentry == ctx->extract_root && ctx->root_dentry_is_special)) {
670 u32 attributes = dentry->d_inode->i_attributes;
672 /* Clear unsupported attributes. */
673 attributes &= ctx->supported_attributes_mask;
675 if ((attributes & FILE_ATTRIBUTE_DIRECTORY &&
676 !ctx->supported_features.encrypted_directories) ||
677 (!(attributes & FILE_ATTRIBUTE_DIRECTORY) &&
678 !ctx->supported_features.encrypted_files))
680 attributes &= ~FILE_ATTRIBUTE_ENCRYPTED;
684 attributes = FILE_ATTRIBUTE_NORMAL;
686 ret = ctx->ops->set_file_attributes(path, attributes, ctx, pass);
688 ERROR_WITH_ERRNO("Failed to set attributes on "
697 /* Set or remove the short (DOS) name on an extracted file or directory if
698 * supported by the extraction mode. Since DOS names are unimportant and it's
699 * easy to run into problems setting them on Windows (SetFileShortName()
700 * requires SE_RESTORE privilege, which only the Administrator can request, and
701 * also requires DELETE access to the file), failure is ignored unless
702 * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set. */
704 extract_short_name(const tchar *path, struct apply_ctx *ctx,
705 struct wim_dentry *dentry)
709 /* The root of the dentry tree being extracted may not be extracted to
710 * its original name, so its short name should be ignored. */
711 if (dentry == ctx->extract_root)
714 if (ctx->supported_features.short_names) {
715 ret = ctx->ops->set_short_name(path,
717 dentry->short_name_nbytes / 2,
719 if (ret && (ctx->extract_flags &
720 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES))
722 ERROR_WITH_ERRNO("Failed to set short name of "
730 /* Set security descriptor, UNIX data, or neither on an extracted file, taking
731 * into account the current extraction mode and flags. */
733 extract_security(const tchar *path, struct apply_ctx *ctx,
734 struct wim_dentry *dentry)
737 struct wim_inode *inode = dentry->d_inode;
739 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS)
742 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
746 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
747 struct wimlib_unix_data data;
749 ret = inode_get_unix_data(inode, &data, NULL);
753 ret = ctx->ops->set_unix_data(path, &data, ctx);
755 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
756 ERROR_WITH_ERRNO("Failed to set UNIX owner, "
757 "group, and/or mode on "
761 WARNING_WITH_ERRNO("Failed to set UNIX owner, "
762 "group, and/or/mode on "
768 #endif /* __WIN32__ */
769 if (ctx->supported_features.security_descriptors &&
770 inode->i_security_id != -1)
772 const struct wim_security_data *sd;
776 sd = wim_const_security_data(ctx->wim);
777 desc = sd->descriptors[inode->i_security_id];
778 desc_size = sd->sizes[inode->i_security_id];
780 ret = ctx->ops->set_security_descriptor(path, desc,
783 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
784 ERROR_WITH_ERRNO("Failed to set security "
785 "descriptor on \"%"TS"\"", path);
789 if (errno != EACCES) {
790 WARNING_WITH_ERRNO("Failed to set "
791 "security descriptor "
792 "on \"%"TS"\"", path);
795 ctx->no_security_descriptors++;
802 /* Set timestamps on an extracted file. Failure is warning-only unless
803 * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */
805 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
806 struct wim_dentry *dentry)
808 struct wim_inode *inode = dentry->d_inode;
811 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
814 if (ctx->ops->set_timestamps) {
815 ret = ctx->ops->set_timestamps(path,
816 inode->i_creation_time,
817 inode->i_last_write_time,
818 inode->i_last_access_time,
821 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
822 ERROR_WITH_ERRNO("Failed to set timestamps "
823 "on \"%"TS"\"", path);
826 WARNING_WITH_ERRNO("Failed to set timestamps "
827 "on \"%"TS"\"", path);
834 /* Check whether the extraction of a dentry should be skipped completely. */
836 dentry_is_supported(struct wim_dentry *dentry,
837 const struct wim_features *supported_features)
839 struct wim_inode *inode = dentry->d_inode;
841 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
842 return supported_features->reparse_points ||
843 (inode_is_symlink(inode) &&
844 supported_features->symlink_reparse_points);
846 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
847 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
848 return supported_features->encrypted_directories != 0;
850 return supported_features->encrypted_files != 0;
855 /* Given a WIM dentry to extract, build the path to which to extract it, in the
856 * format understood by the callbacks in the apply_operations being used.
858 * Write the resulting path into @path, which must have room for at least
859 * ctx->ops->max_path characters including the null-terminator.
861 * Return %true if successful; %false if this WIM dentry doesn't actually need
862 * to be extracted or if the calculated path exceeds ctx->ops->max_path
865 * This function clobbers the tmp_list member of @dentry and its ancestors up
866 * until the extraction root. */
868 build_extraction_path(tchar path[], struct wim_dentry *dentry,
869 struct apply_ctx *ctx)
872 LIST_HEAD(ancestor_list);
874 const tchar *target_prefix;
875 size_t target_prefix_nchars;
876 struct wim_dentry *d;
878 if (dentry->extraction_skipped)
881 path_nchars = ctx->ops->path_prefix_nchars;
883 if (ctx->ops->requires_realtarget_in_paths) {
884 target_prefix = ctx->realtarget;
885 target_prefix_nchars = ctx->realtarget_nchars;
886 } else if (ctx->ops->requires_target_in_paths) {
887 target_prefix = ctx->target;
888 target_prefix_nchars = ctx->target_nchars;
890 target_prefix = NULL;
891 target_prefix_nchars = 0;
893 path_nchars += target_prefix_nchars;
895 for (d = dentry; d != ctx->extract_root; d = d->parent) {
896 path_nchars += d->extraction_name_nchars + 1;
897 list_add(&d->tmp_list, &ancestor_list);
900 path_nchars++; /* null terminator */
902 if (path_nchars > ctx->ops->path_max) {
903 WARNING("\"%"TS"\": Path too long to extract",
904 dentry_full_path(dentry));
908 p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
909 p = tmempcpy(p, target_prefix, target_prefix_nchars);
910 list_for_each_entry(d, &ancestor_list, tmp_list) {
911 *p++ = ctx->ops->path_separator;
912 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
915 wimlib_assert(p - path == path_nchars);
920 get_num_path_components(const tchar *path, tchar path_separator)
922 unsigned num_components = 0;
924 /* Ignore drive letter. */
925 if (path[0] != L'\0' && path[1] == L':')
930 while (*path == path_separator)
934 while (*path && *path != path_separator)
937 return num_components;
941 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
942 struct apply_ctx *ctx, struct wim_dentry *dentry)
944 size_t num_raw_path_components;
945 const struct wim_dentry *d;
946 size_t num_target_path_components;
951 num_raw_path_components = 0;
952 for (d = dentry; d != ctx->extract_root; d = d->parent)
953 num_raw_path_components++;
955 if (ctx->ops->requires_realtarget_in_paths)
956 num_target_path_components = get_num_path_components(ctx->realtarget,
957 ctx->ops->path_separator);
958 else if (ctx->ops->requires_target_in_paths)
959 num_target_path_components = get_num_path_components(ctx->target,
960 ctx->ops->path_separator);
962 num_target_path_components = 0;
964 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
965 wimlib_assert(num_target_path_components > 0);
966 num_raw_path_components++;
967 num_target_path_components--;
970 p_old = oldpath + ctx->ops->path_prefix_nchars;
972 if (p_old[0] != L'\0' && p_old[1] == ':')
975 while (*p_old == ctx->ops->path_separator)
977 while (--num_target_path_components) {
978 while (*p_old != ctx->ops->path_separator)
980 while (*p_old == ctx->ops->path_separator)
984 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
986 p = &symlink_target[0];
987 while (num_raw_path_components--) {
990 *p++ = ctx->ops->path_separator;
993 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
994 newpath, symlink_target);
995 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
997 ERROR_WITH_ERRNO("Failed to create symlink "
998 "\"%"TS"\" => \"%"TS"\"",
999 newpath, symlink_target);
1004 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
1005 * data streams, reparse data (including symbolic links), timestamps, and
1006 * security descriptors. Basically, everything that doesn't require reading
1007 * non-metadata resources from the WIM file and isn't delayed until the final
1010 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
1011 struct apply_ctx *ctx)
1013 struct wim_inode *inode = dentry->d_inode;
1015 const tchar *oldpath;
1017 if (unlikely(is_linked_extraction(ctx))) {
1018 struct wim_lookup_table_entry *unnamed_lte;
1020 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1021 if (unnamed_lte && unnamed_lte->extracted_file) {
1022 oldpath = unnamed_lte->extracted_file;
1023 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
1030 /* Create hard link if this dentry corresponds to an already-extracted
1032 if (inode->i_extracted_file) {
1033 oldpath = inode->i_extracted_file;
1037 /* Skip symlinks unless they can be extracted as reparse points rather
1038 * than created directly. */
1039 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
1042 /* Create this file or directory unless it's the extraction root, which
1043 * was already created if necessary. */
1044 if (dentry != ctx->extract_root) {
1045 ret = extract_inode(path, ctx, inode);
1050 /* Create empty named data streams. */
1051 if (can_extract_named_data_streams(ctx)) {
1052 for (u16 i = 0; i < inode->i_num_ads; i++) {
1053 file_spec_t file_spec;
1054 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1056 if (!ads_entry_is_named_stream(entry))
1060 if (ctx->ops->uses_cookies)
1061 file_spec.cookie = inode->extract_cookie;
1063 file_spec.path = path;
1064 ret = ctx->ops->extract_named_stream(file_spec,
1066 entry->stream_name_nbytes / 2,
1069 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1070 "empty named data stream",
1077 /* Set file attributes (if supported). */
1078 ret = extract_file_attributes(path, ctx, dentry, 0);
1082 /* Set or remove file short name (if supported). */
1083 ret = extract_short_name(path, ctx, dentry);
1087 /* If inode has multiple links and hard links are supported in this
1088 * extraction mode and volume, save the path to the extracted file in
1089 * case it's needed to create a hard link. */
1090 if (unlikely(is_linked_extraction(ctx))) {
1091 struct wim_lookup_table_entry *unnamed_lte;
1093 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1095 unnamed_lte->extracted_file = TSTRDUP(path);
1096 if (!unnamed_lte->extracted_file)
1097 return WIMLIB_ERR_NOMEM;
1099 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1100 inode->i_extracted_file = TSTRDUP(path);
1101 if (!inode->i_extracted_file)
1102 return WIMLIB_ERR_NOMEM;
1107 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1110 dentry->was_hardlinked = 1;
1114 ret = extract_hardlink(oldpath, path, ctx);
1117 dentry->was_hardlinked = 1;
1121 /* This is a wrapper around do_dentry_extract_skeleton() that handles building
1122 * the path, doing short name reordering. This is also idempotent; dentries
1123 * already processed have skeleton_extracted set and no action is taken. See
1124 * apply_operations.requires_short_name_reordering for more details about short
1125 * name reordering. */
1127 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1129 struct apply_ctx *ctx = _ctx;
1130 tchar path[ctx->ops->path_max];
1131 struct wim_dentry *orig_dentry;
1132 struct wim_dentry *other_dentry;
1135 if (dentry->skeleton_extracted)
1139 if (ctx->supported_features.short_names
1140 && ctx->ops->requires_short_name_reordering
1141 && !dentry_has_short_name(dentry)
1142 && !dentry->d_inode->i_dos_name_extracted)
1144 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1145 if (dentry_has_short_name(other_dentry)
1146 && !other_dentry->skeleton_extracted
1147 && other_dentry->in_extraction_tree)
1149 DEBUG("Creating %"TS" before %"TS" "
1150 "to guarantee correct DOS name extraction",
1151 dentry_full_path(other_dentry),
1152 dentry_full_path(dentry));
1153 orig_dentry = dentry;
1154 dentry = other_dentry;
1160 if (!build_extraction_path(path, dentry, ctx))
1162 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1166 dentry->skeleton_extracted = 1;
1169 dentry = orig_dentry;
1173 dentry->d_inode->i_dos_name_extracted = 1;
1178 dentry_extract_dir_skeleton(struct wim_dentry *dentry, void *_ctx)
1180 if (dentry->d_inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1181 return dentry_extract_skeleton(dentry, _ctx);
1185 /* Create a file or directory, then immediately extract all streams. This
1186 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1187 * may not be read sequentially by this function. */
1189 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1191 struct apply_ctx *ctx = _ctx;
1192 tchar path[ctx->ops->path_max];
1195 ret = dentry_extract_skeleton(dentry, ctx);
1199 if (!build_extraction_path(path, dentry, ctx))
1202 return extract_streams(path, ctx, dentry, NULL, NULL);
1205 /* Extract all instances of the stream @lte that are being extracted in this
1206 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1207 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1208 * extracted multiple times, it is extracted to a temporary file first.
1210 * This is intended for use with sequential extraction of a WIM image
1211 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1213 extract_stream_instances(struct wim_lookup_table_entry *lte,
1214 struct apply_ctx *ctx, bool can_seek)
1216 struct wim_dentry **lte_dentries;
1217 struct wim_lookup_table_entry *lte_tmp = NULL;
1218 struct wim_lookup_table_entry *lte_override;
1219 tchar *stream_tmp_filename = NULL;
1220 tchar path[ctx->ops->path_max];
1224 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1225 lte_dentries = lte->inline_lte_dentries;
1227 lte_dentries = lte->lte_dentries;
1229 if (likely(can_seek || lte->out_refcnt < 2)) {
1232 /* Need to extract stream to temporary file. */
1236 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1237 if (!stream_tmp_filename) {
1238 ERROR_WITH_ERRNO("Failed to create temporary filename");
1239 ret = WIMLIB_ERR_OPEN;
1243 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1245 ret = WIMLIB_ERR_NOMEM;
1246 goto out_free_stream_tmp_filename;
1248 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1249 lte_tmp->file_on_disk = stream_tmp_filename;
1250 lte_override = lte_tmp;
1252 raw_fd = topen(stream_tmp_filename,
1253 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1255 ERROR_WITH_ERRNO("Failed to open temporary file");
1256 ret = WIMLIB_ERR_OPEN;
1257 goto out_free_lte_tmp;
1259 filedes_init(&fd, raw_fd);
1260 ret = extract_wim_resource_to_fd(lte, &fd,
1261 wim_resource_size(lte));
1262 if (filedes_close(&fd) && !ret)
1263 ret = WIMLIB_ERR_WRITE;
1265 goto out_unlink_stream_tmp_file;
1268 /* Extract all instances of the stream, reading either from the stream
1269 * in the WIM file or from the temporary file containing the stream.
1270 * dentry->tmp_flag is used to ensure that each dentry is processed only
1271 * once regardless of how many times this stream appears in the streams
1272 * of the corresponding inode. */
1273 for (i = 0; i < lte->out_refcnt; i++) {
1274 struct wim_dentry *dentry = lte_dentries[i];
1276 if (dentry->tmp_flag)
1278 if (!build_extraction_path(path, dentry, ctx))
1280 ret = extract_streams(path, ctx, dentry,
1283 goto out_clear_tmp_flags;
1284 dentry->tmp_flag = 1;
1287 out_clear_tmp_flags:
1288 for (i = 0; i < lte->out_refcnt; i++)
1289 lte_dentries[i]->tmp_flag = 0;
1290 out_unlink_stream_tmp_file:
1291 if (stream_tmp_filename)
1292 tunlink(stream_tmp_filename);
1295 out_free_stream_tmp_filename:
1296 FREE(stream_tmp_filename);
1301 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1302 * structure and empty files were already created. This relies on the
1303 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1304 * stream that was constructed earlier. Streams are extracted exactly in the
1305 * order of the stream list; however, unless the WIM's file descriptor is
1306 * detected to be non-seekable, streams may be read from the WIM file more than
1307 * one time if multiple copies need to be extracted. */
1309 extract_stream_list(struct apply_ctx *ctx)
1311 struct wim_lookup_table_entry *lte;
1315 can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
1316 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1317 ret = extract_stream_instances(lte, ctx, can_seek);
1324 #define PWM_ALLOW_WIM_HDR 0x00001
1325 #define PWM_SILENT_EOF 0x00002
1327 /* Read the header from a stream in a pipable WIM. */
1329 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1330 int flags, struct wim_header_disk *hdr_ret)
1333 struct pwm_stream_hdr stream_hdr;
1334 struct wim_header_disk pwm_hdr;
1338 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
1342 if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
1343 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
1344 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
1345 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
1349 lte->resource_location = RESOURCE_NONEXISTENT;
1350 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
1354 if (buf.stream_hdr.magic != PWM_STREAM_MAGIC) {
1355 ERROR("Data read on pipe is invalid (expected stream header).");
1356 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1359 lte->resource_entry.original_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
1360 copy_hash(lte->hash, buf.stream_hdr.hash);
1361 lte->resource_entry.flags = le32_to_cpu(buf.stream_hdr.flags);
1362 lte->resource_entry.offset = pwm->in_fd.offset;
1363 lte->resource_location = RESOURCE_IN_WIM;
1365 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1366 lte->compression_type = pwm->compression_type;
1367 lte->resource_entry.size = 0;
1369 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1370 lte->resource_entry.size = lte->resource_entry.original_size;
1372 lte->is_pipable = 1;
1376 if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1377 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1382 skip_pwm_chunk_cb(const void *chunk, size_t chunk_size, void *_ctx)
1387 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1389 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1391 return read_partial_wim_resource(lte,
1392 wim_resource_size(lte),
1394 wim_resource_chunk_size(lte),
1396 WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS,
1401 extract_streams_from_pipe(struct apply_ctx *ctx)
1403 struct wim_lookup_table_entry *found_lte;
1404 struct wim_lookup_table_entry *needed_lte;
1405 struct wim_lookup_table *lookup_table;
1406 struct wim_header_disk pwm_hdr;
1410 ret = WIMLIB_ERR_NOMEM;
1411 found_lte = new_lookup_table_entry();
1415 lookup_table = ctx->wim->lookup_table;
1416 pwm_flags = PWM_ALLOW_WIM_HDR;
1417 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1418 pwm_flags |= PWM_SILENT_EOF;
1419 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
1420 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
1421 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
1422 if (ctx->progress_func)
1423 ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1425 while (ctx->num_streams_remaining) {
1426 ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags,
1429 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1430 (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1434 goto out_free_found_lte;
1437 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1438 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1439 && (needed_lte = lookup_resource(lookup_table, found_lte->hash))
1440 && (needed_lte->out_refcnt))
1442 copy_resource_entry(&needed_lte->resource_entry,
1443 &found_lte->resource_entry);
1444 needed_lte->resource_location = found_lte->resource_location;
1445 needed_lte->wim = found_lte->wim;
1446 needed_lte->compression_type = found_lte->compression_type;
1447 needed_lte->is_pipable = found_lte->is_pipable;
1449 ret = extract_stream_instances(needed_lte, ctx, false);
1451 goto out_free_found_lte;
1452 ctx->num_streams_remaining--;
1453 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1454 ret = skip_pwm_stream(found_lte);
1456 goto out_free_found_lte;
1458 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
1459 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
1461 if (part_number != ctx->progress.extract.part_number ||
1462 total_parts != ctx->progress.extract.total_parts ||
1463 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
1466 ctx->progress.extract.part_number = part_number;
1467 ctx->progress.extract.total_parts = total_parts;
1468 memcpy(ctx->progress.extract.guid,
1469 pwm_hdr.guid, WIM_GID_LEN);
1470 if (ctx->progress_func) {
1472 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1481 free_lookup_table_entry(found_lte);
1490 /* Finish extracting a file, directory, or symbolic link by setting file
1491 * security and timestamps. */
1493 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1495 struct apply_ctx *ctx = _ctx;
1497 tchar path[ctx->ops->path_max];
1499 if (!build_extraction_path(path, dentry, ctx))
1502 ret = extract_security(path, ctx, dentry);
1506 if (ctx->ops->requires_final_set_attributes_pass) {
1507 /* Set file attributes (if supported). */
1508 ret = extract_file_attributes(path, ctx, dentry, 1);
1513 return extract_timestamps(path, ctx, dentry);
1517 * Extract a WIM dentry to standard output.
1519 * This obviously doesn't make sense in all cases. We return an error if the
1520 * dentry does not correspond to a regular file. Otherwise we extract the
1521 * unnamed data stream only.
1524 extract_dentry_to_stdout(struct wim_dentry *dentry)
1527 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1528 FILE_ATTRIBUTE_DIRECTORY))
1530 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1531 "extracted to standard output", dentry_full_path(dentry));
1532 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1534 struct wim_lookup_table_entry *lte;
1536 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1538 struct filedes _stdout;
1539 filedes_init(&_stdout, STDOUT_FILENO);
1540 ret = extract_wim_resource_to_fd(lte, &_stdout,
1541 wim_resource_size(lte));
1548 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1550 static const utf16lechar replacement_char = cpu_to_le16('?');
1554 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1560 for (i = 0; i < num_chars; i++) {
1563 case cpu_to_le16('\\'):
1564 case cpu_to_le16(':'):
1565 case cpu_to_le16('*'):
1566 case cpu_to_le16('?'):
1567 case cpu_to_le16('"'):
1568 case cpu_to_le16('<'):
1569 case cpu_to_le16('>'):
1570 case cpu_to_le16('|'):
1572 case cpu_to_le16('/'):
1573 case cpu_to_le16('\0'):
1575 name[i] = replacement_char;
1582 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1583 name[num_chars - 1] == cpu_to_le16('.'))
1586 name[num_chars - 1] = replacement_char;
1595 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1597 const utf16lechar *file_name = dentry->file_name;
1598 return file_name != NULL &&
1599 file_name[0] == cpu_to_le16('.') &&
1600 (file_name[1] == cpu_to_le16('\0') ||
1601 (file_name[1] == cpu_to_le16('.') &&
1602 file_name[2] == cpu_to_le16('\0')));
1606 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1608 dentry->extraction_skipped = 1;
1613 * dentry_calculate_extraction_path-
1615 * Calculate the actual filename component at which a WIM dentry will be
1616 * extracted, handling invalid filenames "properly".
1618 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1619 * UNIX, converted into the platform's multibyte encoding). However, if the
1620 * file name contains characters that are not valid on the current platform or
1621 * has some other format that is not valid, leave dentry->extraction_name as
1622 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1623 * not be extracted, unless the appropriate flag
1624 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1625 * which case a substitute filename will be created and set instead.
1627 * Conflicts with case-insensitive names on Windows are handled similarly; see
1631 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1633 struct apply_ctx *ctx = _args;
1636 dentry->in_extraction_tree = 1;
1638 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1641 if (!dentry_is_supported(dentry, &ctx->supported_features))
1644 if (dentry_is_dot_or_dotdot(dentry)) {
1645 /* WIM files shouldn't contain . or .. entries. But if they are
1646 * there, don't attempt to extract them. */
1647 WARNING("Skipping extraction of unexpected . or .. file "
1648 "\"%"TS"\"", dentry_full_path(dentry));
1653 if (!ctx->ops->supports_case_sensitive_filenames)
1655 struct wim_dentry *other;
1656 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1657 case_insensitive_conflict_list)
1659 if (ctx->extract_flags &
1660 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1661 WARNING("\"%"TS"\" has the same "
1662 "case-insensitive name as "
1663 "\"%"TS"\"; extracting "
1664 "dummy name instead",
1665 dentry_full_path(dentry),
1666 dentry_full_path(other));
1669 WARNING("Not extracting \"%"TS"\": "
1670 "has same case-insensitive "
1671 "name as \"%"TS"\"",
1672 dentry_full_path(dentry),
1673 dentry_full_path(other));
1678 #else /* __WIN32__ */
1679 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1680 #endif /* !__WIN32__ */
1682 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1684 dentry->extraction_name = dentry->file_name;
1685 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1688 return utf16le_to_tstr(dentry->file_name,
1689 dentry->file_name_nbytes,
1690 &dentry->extraction_name,
1691 &dentry->extraction_name_nchars);
1694 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1696 WARNING("\"%"TS"\" has an invalid filename "
1697 "that is not supported on this platform; "
1698 "extracting dummy name instead",
1699 dentry_full_path(dentry));
1702 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1703 "that is not supported on this platform",
1704 dentry_full_path(dentry));
1711 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1713 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1714 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1717 size_t tchar_nchars;
1719 tchar_name = utf16_name_copy;
1720 tchar_nchars = dentry->file_name_nbytes / 2;
1722 ret = utf16le_to_tstr(utf16_name_copy,
1723 dentry->file_name_nbytes,
1724 &tchar_name, &tchar_nchars);
1728 size_t fixed_name_num_chars = tchar_nchars;
1729 tchar fixed_name[tchar_nchars + 50];
1731 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1732 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1733 T(" (invalid filename #%lu)"),
1734 ++ctx->invalid_sequence);
1738 dentry->extraction_name = memdup(fixed_name,
1739 2 * fixed_name_num_chars + 2);
1740 if (!dentry->extraction_name)
1741 return WIMLIB_ERR_NOMEM;
1742 dentry->extraction_name_nchars = fixed_name_num_chars;
1747 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1751 /* Clean up dentry and inode structure after extraction. */
1753 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1755 struct wim_inode *inode = dentry->d_inode;
1757 dentry->in_extraction_tree = 0;
1758 dentry->extraction_skipped = 0;
1759 dentry->was_hardlinked = 0;
1760 dentry->skeleton_extracted = 0;
1761 inode->i_visited = 0;
1762 FREE(inode->i_extracted_file);
1763 inode->i_extracted_file = NULL;
1764 inode->i_dos_name_extracted = 0;
1765 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1766 FREE(dentry->extraction_name);
1767 dentry->extraction_name = NULL;
1771 /* Tally features necessary to extract a dentry and the corresponding inode. */
1773 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1775 struct wim_features *features = _features;
1776 struct wim_inode *inode = dentry->d_inode;
1778 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1779 features->archive_files++;
1780 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1781 features->hidden_files++;
1782 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1783 features->system_files++;
1784 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1785 features->compressed_files++;
1786 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1787 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1788 features->encrypted_directories++;
1790 features->encrypted_files++;
1792 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1793 features->not_context_indexed_files++;
1794 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1795 features->sparse_files++;
1796 if (inode_has_named_stream(inode))
1797 features->named_data_streams++;
1798 if (inode->i_visited)
1799 features->hard_links++;
1800 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1801 features->reparse_points++;
1802 if (inode_is_symlink(inode))
1803 features->symlink_reparse_points++;
1805 features->other_reparse_points++;
1807 if (inode->i_security_id != -1)
1808 features->security_descriptors++;
1809 if (dentry->short_name_nbytes)
1810 features->short_names++;
1811 if (inode_has_unix_data(inode))
1812 features->unix_data++;
1813 inode->i_visited = 1;
1817 /* Tally the features necessary to extract a dentry tree. */
1819 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1821 memset(features, 0, sizeof(struct wim_features));
1822 for_dentry_in_tree(root, dentry_tally_features, features);
1823 dentry_tree_clear_inode_visited(root);
1827 compute_supported_attributes_mask(const struct wim_features *supported_features)
1831 if (!supported_features->archive_files)
1832 mask &= ~FILE_ATTRIBUTE_ARCHIVE;
1834 if (!supported_features->hidden_files)
1835 mask &= ~FILE_ATTRIBUTE_HIDDEN;
1837 if (!supported_features->system_files)
1838 mask &= ~FILE_ATTRIBUTE_SYSTEM;
1840 if (!supported_features->not_context_indexed_files)
1841 mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED;
1843 if (!supported_features->compressed_files)
1844 mask &= ~FILE_ATTRIBUTE_COMPRESSED;
1846 if (!supported_features->sparse_files)
1847 mask &= ~FILE_ATTRIBUTE_SPARSE_FILE;
1849 if (!supported_features->reparse_points)
1850 mask &= ~FILE_ATTRIBUTE_REPARSE_POINT;
1856 do_feature_check(const struct wim_features *required_features,
1857 const struct wim_features *supported_features,
1859 const struct apply_operations *ops,
1860 const tchar *wim_source_path)
1863 const tchar *mode = T("this extraction mode");
1865 if (wim_source_path[0] == '\0')
1866 loc = T("the WIM image");
1868 loc = wim_source_path;
1870 /* We're an archive program, so theoretically we can do what we want
1871 * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't
1872 * bother the user about it. */
1874 if (required_features->archive_files && !supported_features->archive_files)
1877 "%lu files in %"TS" are marked as archived, but this attribute\n"
1878 " is not supported in %"TS".",
1879 required_features->archive_files, loc, mode);
1883 if (required_features->hidden_files && !supported_features->hidden_files)
1886 "%lu files in %"TS" are marked as hidden, but this\n"
1887 " attribute is not supported in %"TS".",
1888 required_features->hidden_files, loc, mode);
1891 if (required_features->system_files && !supported_features->system_files)
1894 "%lu files in %"TS" are marked as system files,\n"
1895 " but this attribute is not supported in %"TS".",
1896 required_features->system_files, loc, mode);
1899 if (required_features->compressed_files && !supported_features->compressed_files)
1902 "%lu files in %"TS" are marked as being transparently\n"
1903 " compressed, but transparent compression is not supported in\n"
1904 " %"TS". These files will be extracted as uncompressed.",
1905 required_features->compressed_files, loc, mode);
1908 if (required_features->encrypted_files && !supported_features->encrypted_files)
1911 "%lu files in %"TS" are marked as being encrypted,\n"
1912 " but encryption is not supported in %"TS". These files\n"
1913 " will not be extracted.",
1914 required_features->encrypted_files, loc, mode);
1917 if (required_features->encrypted_directories &&
1918 !supported_features->encrypted_directories)
1921 "%lu directories in %"TS" are marked as being encrypted,\n"
1922 " but encryption is not supported in %"TS".\n"
1923 " These directories will be extracted as unencrypted.",
1924 required_features->encrypted_directories, loc, mode);
1927 if (required_features->not_context_indexed_files &&
1928 !supported_features->not_context_indexed_files)
1931 "%lu files in %"TS" are marked as not content indexed,\n"
1932 " but this attribute is not supported in %"TS".",
1933 required_features->not_context_indexed_files, loc, mode);
1936 if (required_features->sparse_files && !supported_features->sparse_files)
1939 "%lu files in %"TS" are marked as sparse, but creating\n"
1940 " sparse files is not supported in %"TS". These files\n"
1941 " will be extracted as non-sparse.",
1942 required_features->sparse_files, loc, mode);
1945 if (required_features->named_data_streams &&
1946 !supported_features->named_data_streams)
1949 "%lu files in %"TS" contain one or more alternate (named)\n"
1950 " data streams, which are not supported in %"TS".\n"
1951 " Alternate data streams will NOT be extracted.",
1952 required_features->named_data_streams, loc, mode);
1955 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1956 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1957 required_features->named_data_streams &&
1958 supported_features->named_data_streams)
1961 "%lu files in %"TS" contain one or more alternate (named)\n"
1962 " data streams, which are not supported in linked extraction mode.\n"
1963 " Alternate data streams will NOT be extracted.",
1964 required_features->named_data_streams, loc);
1967 if (required_features->hard_links && !supported_features->hard_links)
1970 "%lu files in %"TS" are hard links, but hard links are\n"
1971 " not supported in %"TS". Hard links will be extracted as\n"
1972 " duplicate copies of the linked files.",
1973 required_features->hard_links, loc, mode);
1976 if (required_features->reparse_points && !supported_features->reparse_points)
1978 if (supported_features->symlink_reparse_points) {
1979 if (required_features->other_reparse_points) {
1981 "%lu files in %"TS" are reparse points that are neither\n"
1982 " symbolic links nor junction points and are not supported in\n"
1983 " %"TS". These reparse points will not be extracted.",
1984 required_features->other_reparse_points, loc,
1989 "%lu files in %"TS" are reparse points, which are\n"
1990 " not supported in %"TS" and will not be extracted.",
1991 required_features->reparse_points, loc, mode);
1995 if (required_features->security_descriptors &&
1996 !supported_features->security_descriptors)
1999 "%lu files in %"TS" have Windows NT security descriptors,\n"
2000 " but extracting security descriptors is not supported in\n"
2001 " %"TS". No security descriptors will be extracted.",
2002 required_features->security_descriptors, loc, mode);
2005 if (required_features->short_names && !supported_features->short_names)
2008 "%lu files in %"TS" have short (DOS) names, but\n"
2009 " extracting short names is not supported in %"TS".\n"
2010 " Short names will not be extracted.\n",
2011 required_features->short_names, loc, mode);
2014 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
2015 required_features->unix_data && !supported_features->unix_data)
2017 ERROR("Extracting UNIX data is not supported in %"TS, mode);
2018 return WIMLIB_ERR_UNSUPPORTED;
2020 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
2021 required_features->short_names && !supported_features->short_names)
2023 ERROR("Extracting short names is not supported in %"TS"", mode);
2024 return WIMLIB_ERR_UNSUPPORTED;
2026 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
2027 !ops->set_timestamps)
2029 ERROR("Extracting timestamps is not supported in %"TS"", mode);
2030 return WIMLIB_ERR_UNSUPPORTED;
2032 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
2033 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
2034 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
2035 required_features->security_descriptors &&
2036 !supported_features->security_descriptors)
2038 ERROR("Extracting security descriptors is not supported in %"TS, mode);
2039 return WIMLIB_ERR_UNSUPPORTED;
2042 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
2043 !supported_features->hard_links)
2045 ERROR("Hard link extraction mode requested, but "
2046 "%"TS" does not support hard links!", mode);
2047 return WIMLIB_ERR_UNSUPPORTED;
2050 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
2051 required_features->symlink_reparse_points &&
2052 !(supported_features->symlink_reparse_points ||
2053 supported_features->reparse_points))
2055 ERROR("Extracting symbolic links is not supported in %"TS, mode);
2056 return WIMLIB_ERR_UNSUPPORTED;
2059 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
2060 !supported_features->symlink_reparse_points)
2062 ERROR("Symbolic link extraction mode requested, but "
2063 "%"TS" does not support symbolic "
2065 return WIMLIB_ERR_UNSUPPORTED;
2071 do_extract_warnings(struct apply_ctx *ctx)
2073 if (ctx->partial_security_descriptors == 0 &&
2074 ctx->no_security_descriptors == 0)
2077 WARNING("Extraction to \"%"TS"\" complete, but with one or more warnings:",
2079 if (ctx->partial_security_descriptors != 0) {
2080 WARNING("- Could only partially set the security descriptor\n"
2081 " on %lu files or directories.",
2082 ctx->partial_security_descriptors);
2084 if (ctx->no_security_descriptors != 0) {
2085 WARNING("- Could not set security descriptor at all\n"
2086 " on %lu files or directories.",
2087 ctx->no_security_descriptors);
2090 WARNING("To fully restore all security descriptors, run the program\n"
2091 " with Administrator rights.");
2096 * extract_tree - Extract a file or directory tree from the currently selected
2099 * @wim: WIMStruct for the WIM file, with the desired image selected
2100 * (as wim->current_image).
2103 * "Canonical" (i.e. no leading or trailing slashes, path
2104 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
2105 * extract. An empty string means the full image.
2108 * Filesystem path to extract the file or directory tree to.
2109 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
2112 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
2113 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
2114 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
2118 * If non-NULL, progress function for the extraction. The messages
2119 * that may be sent in this function are:
2121 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
2122 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
2123 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
2124 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
2125 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
2126 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
2127 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
2128 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
2129 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
2131 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
2134 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
2135 int extract_flags, wimlib_progress_func_t progress_func)
2137 struct wim_dentry *root;
2138 struct wim_features required_features;
2139 struct apply_ctx ctx;
2141 struct wim_lookup_table_entry *lte;
2143 /* Start initializing the apply_ctx. */
2144 memset(&ctx, 0, sizeof(struct apply_ctx));
2146 ctx.extract_flags = extract_flags;
2147 ctx.target = target;
2148 ctx.target_nchars = tstrlen(target);
2149 ctx.progress_func = progress_func;
2150 if (progress_func) {
2151 ctx.progress.extract.wimfile_name = wim->filename;
2152 ctx.progress.extract.image = wim->current_image;
2153 ctx.progress.extract.extract_flags = (extract_flags &
2154 WIMLIB_EXTRACT_MASK_PUBLIC);
2155 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
2156 wim->current_image);
2157 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
2158 ctx.progress.extract.target = target;
2160 INIT_LIST_HEAD(&ctx.stream_list);
2162 /* Translate the path to extract into the corresponding
2163 * `struct wim_dentry', which will be the root of the
2164 * "dentry tree" to extract. */
2165 root = get_dentry(wim, wim_source_path);
2167 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
2168 wim_source_path, wim->current_image);
2169 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
2173 ctx.extract_root = root;
2175 /* Select the appropriate apply_operations based on the
2176 * platform and extract_flags. */
2178 ctx.ops = &win32_apply_ops;
2180 ctx.ops = &unix_apply_ops;
2184 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
2185 ctx.ops = &ntfs_3g_apply_ops;
2188 /* Call the start_extract() callback. This gives the apply_operations
2189 * implementation a chance to do any setup needed to access the volume.
2190 * Furthermore, it's expected to set the supported features of this
2191 * extraction mode (ctx.supported_features), which are determined at
2192 * runtime as they may vary depending on the actual volume. These
2193 * features are then compared with the actual features extracting this
2194 * dentry tree requires. Some mismatches will merely produce warnings
2195 * and the unsupported data will be ignored; others will produce errors.
2197 ret = ctx.ops->start_extract(target, &ctx);
2201 dentry_tree_get_features(root, &required_features);
2202 ret = do_feature_check(&required_features, &ctx.supported_features,
2203 extract_flags, ctx.ops, wim_source_path);
2205 goto out_finish_or_abort_extract;
2207 ctx.supported_attributes_mask =
2208 compute_supported_attributes_mask(&ctx.supported_features);
2210 /* Figure out whether the root dentry is being extracted to the root of
2211 * a volume and therefore needs to be treated "specially", for example
2212 * not being explicitly created and not having attributes set. */
2213 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2214 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2216 /* Calculate the actual filename component of each extracted dentry. In
2217 * the process, set the dentry->extraction_skipped flag on dentries that
2218 * are being skipped for some reason (e.g. invalid filename). */
2219 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
2221 goto out_dentry_reset_needs_extraction;
2223 /* Build the list of the streams that need to be extracted and
2224 * initialize ctx.progress.extract with stream information. */
2225 ret = for_dentry_in_tree(ctx.extract_root,
2226 dentry_resolve_and_zero_lte_refcnt, &ctx);
2228 goto out_dentry_reset_needs_extraction;
2230 ret = for_dentry_in_tree(ctx.extract_root,
2231 dentry_add_streams_to_extract, &ctx);
2233 goto out_teardown_stream_list;
2235 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2236 /* When extracting from a pipe, the number of bytes of data to
2237 * extract can't be determined in the normal way (examining the
2238 * lookup table), since at this point all we have is a set of
2239 * SHA1 message digests of streams that need to be extracted.
2240 * However, we can get a reasonably accurate estimate by taking
2241 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2242 * data. This does assume that a full image is being extracted,
2243 * but currently there is no API for doing otherwise. (Also,
2244 * subtract <HARDLINKBYTES> from this if hard links are
2245 * supported by the extraction mode.) */
2246 ctx.progress.extract.total_bytes =
2247 wim_info_get_image_total_bytes(wim->wim_info,
2248 wim->current_image);
2249 if (ctx.supported_features.hard_links) {
2250 ctx.progress.extract.total_bytes -=
2251 wim_info_get_image_hard_link_bytes(wim->wim_info,
2252 wim->current_image);
2256 /* Handle the special case of extracting a file to standard
2257 * output. In that case, "root" should be a single file, not a
2258 * directory tree. (If not, extract_dentry_to_stdout() will
2259 * return an error.) */
2260 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
2261 ret = extract_dentry_to_stdout(root);
2262 goto out_teardown_stream_list;
2265 /* If a sequential extraction was specified, sort the streams to be
2266 * extracted by their position in the WIM file so that the WIM file can
2267 * be read sequentially. */
2268 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2269 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2270 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2272 ret = sort_stream_list_by_sequential_order(
2274 offsetof(struct wim_lookup_table_entry,
2277 goto out_teardown_stream_list;
2280 if (ctx.ops->realpath_works_on_nonexisting_files &&
2281 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2282 ctx.ops->requires_realtarget_in_paths))
2284 ctx.realtarget = realpath(target, NULL);
2285 if (!ctx.realtarget) {
2286 ret = WIMLIB_ERR_NOMEM;
2287 goto out_teardown_stream_list;
2289 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2292 if (progress_func) {
2293 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2294 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2298 if (!ctx.root_dentry_is_special)
2300 tchar path[ctx.ops->path_max];
2301 if (build_extraction_path(path, root, &ctx))
2303 ret = extract_inode(path, &ctx, root->d_inode);
2305 goto out_free_realtarget;
2309 /* If we need to fix up the targets of absolute symbolic links
2310 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2311 * be absolute, use realpath() (or its replacement on Windows) to get
2312 * the absolute path to the extraction target. Note that this requires
2313 * the target directory to exist, unless
2314 * realpath_works_on_nonexisting_files is set in the apply_operations.
2316 if (!ctx.realtarget &&
2317 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2318 required_features.symlink_reparse_points) ||
2319 ctx.ops->requires_realtarget_in_paths))
2321 ctx.realtarget = realpath(target, NULL);
2322 if (!ctx.realtarget) {
2323 ret = WIMLIB_ERR_NOMEM;
2324 goto out_free_realtarget;
2326 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2329 if (ctx.ops->requires_short_name_reordering) {
2330 ret = for_dentry_in_tree(root, dentry_extract_dir_skeleton,
2333 goto out_free_realtarget;
2336 /* Finally, the important part: extract the tree of files. */
2337 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2338 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2339 /* Sequential extraction requested, so two passes are needed
2340 * (one for directory structure, one for streams.) */
2342 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2345 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2346 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2348 goto out_free_realtarget;
2351 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2353 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2354 ret = extract_streams_from_pipe(&ctx);
2356 ret = extract_stream_list(&ctx);
2358 goto out_free_realtarget;
2360 /* Sequential extraction was not requested, so we can make do
2361 * with one pass where we both create the files and extract
2364 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2366 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2368 goto out_free_realtarget;
2370 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2374 /* If the total number of bytes to extract was miscalculated, just jump
2375 * to the calculated number in order to avoid confusing the progress
2376 * function. This should only occur when extracting from a pipe. */
2377 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2379 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2380 "extracted %"PRIu64,
2381 ctx.progress.extract.total_bytes,
2382 ctx.progress.extract.completed_bytes);
2384 if (progress_func &&
2385 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2387 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2388 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2391 /* Apply security descriptors and timestamps. This is done at the end,
2392 * and in a depth-first manner, to prevent timestamps from getting
2393 * changed by subsequent extract operations and to minimize the chance
2394 * of the restored security descriptors getting in our way. */
2396 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2398 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2400 goto out_free_realtarget;
2402 if (progress_func) {
2403 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2404 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2408 do_extract_warnings(&ctx);
2411 out_free_realtarget:
2412 FREE(ctx.realtarget);
2413 out_teardown_stream_list:
2414 /* Free memory allocated as part of the mapping from each
2415 * wim_lookup_table_entry to the dentries that reference it. */
2416 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2417 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2418 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2419 FREE(lte->lte_dentries);
2420 out_dentry_reset_needs_extraction:
2421 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2422 out_finish_or_abort_extract:
2424 if (ctx.ops->abort_extract)
2425 ctx.ops->abort_extract(&ctx);
2427 if (ctx.ops->finish_extract)
2428 ret = ctx.ops->finish_extract(&ctx);
2434 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2435 * extract flags make sense. */
2437 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2441 /* Empty destination path? */
2442 if (cmd->fs_dest_path[0] == T('\0'))
2443 return WIMLIB_ERR_INVALID_PARAM;
2445 extract_flags = cmd->extract_flags;
2447 /* Check for invalid flag combinations */
2448 if ((extract_flags &
2449 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2450 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2451 WIMLIB_EXTRACT_FLAG_HARDLINK))
2452 return WIMLIB_ERR_INVALID_PARAM;
2454 if ((extract_flags &
2455 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2456 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2457 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2458 return WIMLIB_ERR_INVALID_PARAM;
2460 if ((extract_flags &
2461 (WIMLIB_EXTRACT_FLAG_RPFIX |
2462 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2463 WIMLIB_EXTRACT_FLAG_NORPFIX))
2464 return WIMLIB_ERR_INVALID_PARAM;
2466 if ((extract_flags &
2467 (WIMLIB_EXTRACT_FLAG_RESUME |
2468 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2469 return WIMLIB_ERR_INVALID_PARAM;
2471 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2472 #ifndef WITH_NTFS_3G
2473 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2474 " we cannot apply a WIM image directly to a NTFS volume.");
2475 return WIMLIB_ERR_UNSUPPORTED;
2479 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2480 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2482 /* Do reparse point fixups by default if the WIM header says
2483 * they are enabled and we are extracting a full image. */
2484 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2485 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2488 /* TODO: Since UNIX data entries are stored in the file resources, in a
2489 * completely sequential extraction they may come up before the
2490 * corresponding file or symbolic link data. This needs to be handled
2492 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2493 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2494 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2495 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2497 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2498 WARNING("Setting UNIX file/owner group may "
2499 "be impossible on some\n"
2501 "when applying from a pipe.");
2503 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2504 WARNING("Disabling sequential extraction for "
2509 cmd->extract_flags = extract_flags;
2514 /* Internal function to execute extraction commands for a WIM image. The paths
2515 * in the extract commands are expected to be already "canonicalized". */
2517 do_wimlib_extract_files(WIMStruct *wim,
2519 struct wimlib_extract_command *cmds,
2521 wimlib_progress_func_t progress_func)
2524 bool found_link_cmd = false;
2525 bool found_nolink_cmd = false;
2527 /* Select the image from which we are extracting files */
2528 ret = select_wim_image(wim, image);
2532 /* Make sure there are no streams in the WIM that have not been
2533 * checksummed yet. */
2534 ret = wim_checksum_unhashed_streams(wim);
2538 /* Check for problems with the extraction commands */
2539 for (size_t i = 0; i < num_cmds; i++) {
2540 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2543 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2544 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2545 found_link_cmd = true;
2547 found_nolink_cmd = true;
2549 if (found_link_cmd && found_nolink_cmd) {
2550 ERROR("Symlink or hardlink extraction mode must "
2551 "be set on all extraction commands");
2552 return WIMLIB_ERR_INVALID_PARAM;
2556 /* Execute the extraction commands */
2557 for (size_t i = 0; i < num_cmds; i++) {
2558 ret = extract_tree(wim,
2559 cmds[i].wim_source_path,
2560 cmds[i].fs_dest_path,
2561 cmds[i].extract_flags,
2569 /* API function documented in wimlib.h */
2571 wimlib_extract_files(WIMStruct *wim,
2573 const struct wimlib_extract_command *cmds,
2575 int default_extract_flags,
2576 wimlib_progress_func_t progress_func)
2579 struct wimlib_extract_command *cmds_copy;
2582 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2587 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2589 return WIMLIB_ERR_NOMEM;
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);
2630 * Extracts an image from a WIM file.
2632 * @wim: WIMStruct for the WIM file.
2634 * @image: Number of the single image to extract.
2636 * @target: Directory or NTFS volume to extract the image to.
2638 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2640 * @progress_func: If non-NULL, a progress function to be called
2643 * Returns 0 on success; nonzero on failure.
2646 extract_single_image(WIMStruct *wim, int image,
2647 const tchar *target, int extract_flags,
2648 wimlib_progress_func_t progress_func)
2651 tchar *target_copy = canonicalize_fs_path(target);
2653 return WIMLIB_ERR_NOMEM;
2654 struct wimlib_extract_command cmd = {
2655 .wim_source_path = T(""),
2656 .fs_dest_path = target_copy,
2657 .extract_flags = extract_flags,
2659 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2664 static const tchar * const filename_forbidden_chars =
2673 /* This function checks if it is okay to use a WIM image's name as a directory
2676 image_name_ok_as_dir(const tchar *image_name)
2678 return image_name && *image_name &&
2679 !tstrpbrk(image_name, filename_forbidden_chars) &&
2680 tstrcmp(image_name, T(".")) &&
2681 tstrcmp(image_name, T(".."));
2684 /* Extracts all images from the WIM to the directory @target, with the images
2685 * placed in subdirectories named by their image names. */
2687 extract_all_images(WIMStruct *wim,
2688 const tchar *target,
2690 wimlib_progress_func_t progress_func)
2692 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2693 size_t output_path_len = tstrlen(target);
2694 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2697 const tchar *image_name;
2700 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2702 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2703 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2704 return WIMLIB_ERR_INVALID_PARAM;
2707 if (tstat(target, &stbuf)) {
2708 if (errno == ENOENT) {
2709 if (tmkdir(target, 0755)) {
2710 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2711 return WIMLIB_ERR_MKDIR;
2714 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2715 return WIMLIB_ERR_STAT;
2717 } else if (!S_ISDIR(stbuf.st_mode)) {
2718 ERROR("\"%"TS"\" is not a directory", target);
2719 return WIMLIB_ERR_NOTDIR;
2722 tmemcpy(buf, target, output_path_len);
2723 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2724 for (image = 1; image <= wim->hdr.image_count; image++) {
2725 image_name = wimlib_get_image_name(wim, image);
2726 if (image_name_ok_as_dir(image_name)) {
2727 tstrcpy(buf + output_path_len + 1, image_name);
2729 /* Image name is empty or contains forbidden characters.
2730 * Use image number instead. */
2731 tsprintf(buf + output_path_len + 1, T("%d"), image);
2733 ret = extract_single_image(wim, image, buf, extract_flags,
2742 do_wimlib_extract_image(WIMStruct *wim,
2744 const tchar *target,
2746 wimlib_progress_func_t progress_func)
2750 if (image == WIMLIB_ALL_IMAGES) {
2751 ret = extract_all_images(wim, target, extract_flags,
2754 ret = extract_single_image(wim, image, target, extract_flags,
2758 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2759 WIMLIB_EXTRACT_FLAG_HARDLINK))
2761 for_lookup_table_entry(wim->lookup_table,
2762 lte_free_extracted_file,
2768 /* API function documented in wimlib.h */
2770 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2771 const tchar *target, int extract_flags,
2772 wimlib_progress_func_t progress_func)
2776 struct filedes *in_fd;
2780 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2782 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2783 return WIMLIB_ERR_INVALID_PARAM;
2785 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2787 /* Read the WIM header from the pipe and get a WIMStruct to represent
2788 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2789 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2790 * an empty lookup table, no XML data read, and no filename set. */
2791 ret = open_wim_as_WIMStruct(&pipe_fd,
2792 WIMLIB_OPEN_FLAG_FROM_PIPE,
2793 &pwm, progress_func);
2797 /* Sanity check to make sure this is a pipable WIM. */
2798 if (pwm->hdr.magic != PWM_MAGIC) {
2799 ERROR("The WIM being read from file descriptor %d "
2800 "is not pipable!", pipe_fd);
2801 ret = WIMLIB_ERR_NOT_PIPABLE;
2802 goto out_wimlib_free;
2805 /* Sanity check to make sure the first part of a pipable split WIM is
2806 * sent over the pipe first. */
2807 if (pwm->hdr.part_number != 1) {
2808 ERROR("The first part of the split WIM must be "
2809 "sent over the pipe first.");
2810 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2811 goto out_wimlib_free;
2814 in_fd = &pwm->in_fd;
2815 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2817 /* As mentioned, the WIMStruct we created from the pipe does not have
2818 * XML data yet. Fix this by reading the extra copy of the XML data
2819 * that directly follows the header in pipable WIMs. (Note: see
2820 * write_pipable_wim() for more details about the format of pipable
2823 struct wim_lookup_table_entry xml_lte;
2824 ret = read_pwm_stream_header(pwm, &xml_lte, 0, NULL);
2826 goto out_wimlib_free;
2828 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2830 ERROR("Expected XML data, but found non-metadata "
2832 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2833 goto out_wimlib_free;
2836 copy_resource_entry(&pwm->hdr.xml_res_entry,
2837 &xml_lte.resource_entry);
2839 ret = read_wim_xml_data(pwm);
2841 goto out_wimlib_free;
2842 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2843 ERROR("Image count in XML data is not the same as in WIM header.");
2844 ret = WIMLIB_ERR_XML;
2845 goto out_wimlib_free;
2849 /* Get image index (this may use the XML data that was just read to
2850 * resolve an image name). */
2851 if (image_num_or_name) {
2852 image = wimlib_resolve_image(pwm, image_num_or_name);
2853 if (image == WIMLIB_NO_IMAGE) {
2854 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2856 ret = WIMLIB_ERR_INVALID_IMAGE;
2857 goto out_wimlib_free;
2858 } else if (image == WIMLIB_ALL_IMAGES) {
2859 ERROR("Applying all images from a pipe is not supported.");
2860 ret = WIMLIB_ERR_INVALID_IMAGE;
2861 goto out_wimlib_free;
2864 if (pwm->hdr.image_count != 1) {
2865 ERROR("No image was specified, but the pipable WIM "
2866 "did not contain exactly 1 image");
2867 ret = WIMLIB_ERR_INVALID_IMAGE;
2868 goto out_wimlib_free;
2873 /* Load the needed metadata resource. */
2874 for (i = 1; i <= pwm->hdr.image_count; i++) {
2875 struct wim_lookup_table_entry *metadata_lte;
2876 struct wim_image_metadata *imd;
2878 metadata_lte = new_lookup_table_entry();
2879 if (!metadata_lte) {
2880 ret = WIMLIB_ERR_NOMEM;
2881 goto out_wimlib_free;
2884 ret = read_pwm_stream_header(pwm, metadata_lte, 0, NULL);
2885 imd = pwm->image_metadata[i - 1];
2886 imd->metadata_lte = metadata_lte;
2888 goto out_wimlib_free;
2890 if (!(metadata_lte->resource_entry.flags &
2891 WIM_RESHDR_FLAG_METADATA))
2893 ERROR("Expected metadata resource, but found "
2894 "non-metadata stream.");
2895 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2896 goto out_wimlib_free;
2900 /* Metadata resource is for the images being extracted.
2901 * Parse it and save the metadata in memory. */
2902 ret = read_metadata_resource(pwm, imd);
2904 goto out_wimlib_free;
2907 /* Metadata resource is not for the image being
2908 * extracted. Skip over it. */
2909 ret = skip_pwm_stream(metadata_lte);
2911 goto out_wimlib_free;
2914 /* Extract the image. */
2915 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2916 ret = do_wimlib_extract_image(pwm, image, target,
2917 extract_flags, progress_func);
2918 /* Clean up and return. */
2924 /* API function documented in wimlib.h */
2926 wimlib_extract_image(WIMStruct *wim,
2928 const tchar *target,
2930 wimlib_progress_func_t progress_func)
2932 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2933 return do_wimlib_extract_image(wim, image, target, extract_flags,