4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012, 2013 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
13 * wimlib is free software; you can redistribute it and/or modify it under the
14 * terms of the GNU General Public License as published by the Free
15 * Software Foundation; either version 3 of the License, or (at your option)
18 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20 * A PARTICULAR PURPOSE. See the GNU General Public License for more
23 * You should have received a copy of the GNU General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
28 * This file provides the API functions wimlib_extract_image(),
29 * wimlib_extract_files(), and wimlib_extract_image_from_pipe(). Internally,
30 * all end up calling extract_tree() zero or more times to extract a tree of
31 * files from the currently selected WIM image to the specified target directory
34 * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
35 * Win32), this file does not itself have code to extract files or directories
36 * to any specific target; instead, it handles generic functionality and relies
37 * on lower-level callback functions declared in `struct apply_operations' to do
38 * the actual extraction.
45 #include "wimlib/apply.h"
46 #include "wimlib/dentry.h"
47 #include "wimlib/encoding.h"
48 #include "wimlib/endianness.h"
49 #include "wimlib/error.h"
50 #include "wimlib/lookup_table.h"
51 #include "wimlib/metadata.h"
52 #include "wimlib/paths.h"
53 #include "wimlib/reparse.h"
54 #include "wimlib/resource.h"
55 #include "wimlib/security.h"
56 #include "wimlib/swm.h"
58 # include "wimlib/win32.h" /* for realpath() equivalent */
60 #include "wimlib/xml.h"
61 #include "wimlib/wim.h"
69 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000
71 #define WIMLIB_EXTRACT_MASK_PUBLIC 0x3fffffff
73 /* Given a WIM dentry in the tree to be extracted, resolve all streams in the
74 * corresponding inode and set 'out_refcnt' in each to 0. */
76 dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_ctx)
78 struct apply_ctx *ctx = _ctx;
79 struct wim_inode *inode = dentry->d_inode;
80 struct wim_lookup_table_entry *lte;
84 if (dentry->extraction_skipped)
87 /* Special case: when extracting from a pipe, the WIM lookup table is
88 * initially empty, so "resolving" an inode's streams is initially not
89 * possible. However, we still need to keep track of which streams,
90 * identified by SHA1 message digests, need to be extracted, so we
91 * "resolve" the inode's streams anyway by allocating new entries. */
92 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
94 ret = inode_resolve_ltes(inode, ctx->wim->lookup_table, force);
97 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
98 lte = inode_stream_lte_resolved(inode, i);
106 is_linked_extraction(const struct apply_ctx *ctx)
108 return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
109 WIMLIB_EXTRACT_FLAG_SYMLINK));
113 can_extract_named_data_streams(const struct apply_ctx *ctx)
115 return ctx->supported_features.named_data_streams &&
116 !is_linked_extraction(ctx);
120 ref_stream_to_extract(struct wim_lookup_table_entry *lte,
121 struct wim_dentry *dentry, struct apply_ctx *ctx)
126 if (likely(!is_linked_extraction(ctx)) || (lte->out_refcnt == 0 &&
127 lte->extracted_file == NULL))
129 ctx->progress.extract.total_bytes += wim_resource_size(lte);
130 ctx->progress.extract.num_streams++;
133 if (lte->out_refcnt == 0) {
134 list_add_tail(<e->extraction_list, &ctx->stream_list);
135 ctx->num_streams_remaining++;
138 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
139 struct wim_dentry **lte_dentries;
141 /* Append dentry to this stream's array of dentries referencing
142 * it. Use inline array to avoid memory allocation until the
143 * number of dentries becomes too large. */
144 if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) {
145 lte_dentries = lte->inline_lte_dentries;
147 struct wim_dentry **prev_lte_dentries;
148 size_t alloc_lte_dentries;
150 if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) {
151 prev_lte_dentries = NULL;
152 alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries);
154 prev_lte_dentries = lte->lte_dentries;
155 alloc_lte_dentries = lte->alloc_lte_dentries;
158 if (lte->out_refcnt == alloc_lte_dentries) {
159 alloc_lte_dentries *= 2;
160 lte_dentries = REALLOC(prev_lte_dentries,
162 sizeof(lte_dentries[0]));
164 return WIMLIB_ERR_NOMEM;
165 if (prev_lte_dentries == NULL) {
167 lte->inline_lte_dentries,
168 sizeof(lte->inline_lte_dentries));
170 lte->lte_dentries = lte_dentries;
171 lte->alloc_lte_dentries = alloc_lte_dentries;
173 lte_dentries = lte->lte_dentries;
175 lte_dentries[lte->out_refcnt] = dentry;
181 /* Given a WIM dentry in the tree to be extracted, iterate through streams that
182 * need to be extracted. For each one, add it to the list of streams to be
183 * extracted (ctx->stream_list) if not already done so, and also update the
184 * progress information (ctx->progress) with the stream. Furthermore, if doing
185 * a sequential extraction, build a mapping from each the stream to the dentries
188 dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx)
190 struct apply_ctx *ctx = _ctx;
191 struct wim_inode *inode = dentry->d_inode;
194 /* Don't process dentries marked as skipped. */
195 if (dentry->extraction_skipped)
198 /* Don't process additional hard links. */
199 if (inode->i_visited && ctx->supported_features.hard_links)
202 /* The unnamed data stream will always be extracted, except in an
204 if (!inode_is_encrypted_directory(inode)) {
205 ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode),
211 /* Named data streams will be extracted only if supported in the current
212 * extraction mode and volume, and to avoid complications, if not doing
213 * a linked extraction. */
214 if (can_extract_named_data_streams(ctx)) {
215 for (u16 i = 0; i < inode->i_num_ads; i++) {
216 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
218 ret = ref_stream_to_extract(inode->i_ads_entries[i].lte,
224 inode->i_visited = 1;
228 /* Inform library user of progress of stream extraction following the successful
229 * extraction of a copy of the stream specified by @lte. */
231 update_extract_progress(struct apply_ctx *ctx,
232 const struct wim_lookup_table_entry *lte)
234 wimlib_progress_func_t progress_func = ctx->progress_func;
235 union wimlib_progress_info *progress = &ctx->progress;
237 progress->extract.completed_bytes += wim_resource_size(lte);
239 progress->extract.completed_bytes >= ctx->next_progress)
241 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress);
242 if (progress->extract.completed_bytes >=
243 progress->extract.total_bytes)
245 ctx->next_progress = ~0ULL;
247 ctx->next_progress += progress->extract.total_bytes / 128;
248 if (ctx->next_progress > progress->extract.total_bytes)
249 ctx->next_progress = progress->extract.total_bytes;
255 /* Extract a symbolic link (not directly as reparse data), handling fixing up
256 * the target of absolute symbolic links and updating the extract progress.
258 * @inode must specify the WIM inode for a symbolic link or junction reparse
261 * @lte_override overrides the resource used as the reparse data for the
264 extract_symlink(const tchar *path, struct apply_ctx *ctx,
265 struct wim_inode *inode,
266 struct wim_lookup_table_entry *lte_override)
268 ssize_t bufsize = ctx->ops->path_max;
269 tchar target[bufsize];
275 /* If absolute symbolic link fixups requested, reserve space in the link
276 * target buffer for the absolute path of the target directory. */
277 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
279 buf += ctx->realtarget_nchars;
280 bufsize -= ctx->realtarget_nchars;
283 /* Translate the WIM inode's reparse data into the link target. */
284 sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override);
287 return WIMLIB_ERR_READLINK;
291 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
294 /* Fix absolute symbolic link target to point into the
295 * actual extraction destination. */
296 tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars);
297 fixed_target = target;
299 /* Keep same link target. */
303 /* Call into the apply_operations to create the symbolic link. */
304 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
306 ret = ctx->ops->create_symlink(fixed_target, path, ctx);
308 ERROR_WITH_ERRNO("Failed to create symlink "
309 "\"%"TS"\" => \"%"TS"\"", path, fixed_target);
313 /* Account for reparse data consumed. */
314 update_extract_progress(ctx,
315 (lte_override ? lte_override :
316 inode_unnamed_lte_resolved(inode)));
319 #endif /* !__WIN32__ */
321 /* Create a file, directory, or symbolic link. */
323 extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode)
328 if (ctx->supported_features.symlink_reparse_points &&
329 !ctx->supported_features.reparse_points &&
330 inode_is_symlink(inode))
332 ret = extract_symlink(path, ctx, inode, NULL);
334 #endif /* !__WIN32__ */
335 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
336 ret = ctx->ops->create_directory(path, ctx);
338 ERROR_WITH_ERRNO("Failed to create the directory "
342 ret = ctx->ops->create_file(path, ctx);
344 ERROR_WITH_ERRNO("Failed to create the file "
352 extract_hardlink(const tchar *oldpath, const tchar *newpath,
353 struct apply_ctx *ctx)
357 DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath);
358 ret = ctx->ops->create_hardlink(oldpath, newpath, ctx);
360 ERROR_WITH_ERRNO("Failed to create hardlink "
361 "\"%"TS"\" => \"%"TS"\"",
369 try_extract_rpfix(u8 *rpbuf,
371 const wchar_t *extract_root_realpath,
372 unsigned extract_root_realpath_nchars)
374 struct reparse_data rpdata;
376 size_t target_nchars;
377 size_t stripped_nchars;
378 wchar_t *stripped_target;
379 wchar_t stripped_target_nchars;
382 utf16lechar *new_target;
383 utf16lechar *new_print_name;
384 size_t new_target_nchars;
385 size_t new_print_name_nchars;
388 ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata);
392 if (extract_root_realpath[0] == L'\0' ||
393 extract_root_realpath[1] != L':' ||
394 extract_root_realpath[2] != L'\\')
395 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
397 ret = parse_substitute_name(rpdata.substitute_name,
398 rpdata.substitute_name_nbytes,
402 stripped_nchars = ret;
403 target = rpdata.substitute_name;
404 target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar);
405 stripped_target = target + stripped_nchars;
406 stripped_target_nchars = target_nchars - stripped_nchars;
408 new_target = alloca((6 + extract_root_realpath_nchars +
409 stripped_target_nchars) * sizeof(utf16lechar));
412 if (stripped_nchars == 6) {
413 /* Include \??\ prefix if it was present before */
414 p = wmempcpy(p, L"\\??\\", 4);
417 /* Print name excludes the \??\ if present. */
419 if (stripped_nchars != 0) {
420 /* Get drive letter from real path to extract root, if a drive
421 * letter was present before. */
422 *p++ = extract_root_realpath[0];
423 *p++ = extract_root_realpath[1];
425 /* Copy the rest of the extract root */
426 p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2);
428 /* Append the stripped target */
429 p = wmempcpy(p, stripped_target, stripped_target_nchars);
430 new_target_nchars = p - new_target;
431 new_print_name_nchars = p - new_print_name;
433 if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE ||
434 new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE)
435 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
437 rpdata.substitute_name = new_target;
438 rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar);
439 rpdata.print_name = new_print_name;
440 rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar);
441 return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p);
443 #endif /* __WIN32__ */
445 /* Set reparse data on extracted file or directory that has
446 * FILE_ATTRIBUTE_REPARSE_POINT set. */
448 extract_reparse_data(const tchar *path, struct apply_ctx *ctx,
449 struct wim_inode *inode,
450 struct wim_lookup_table_entry *lte_override)
453 u8 rpbuf[REPARSE_POINT_MAX_SIZE];
456 ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override);
461 /* Fix up target of absolute symbolic link or junction points so
462 * that they point into the actual extraction target. */
463 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
464 (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
465 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) &&
466 !inode->i_not_rpfixed)
468 ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget,
469 ctx->realtarget_nchars);
470 if (ret && !(ctx->extract_flags &
471 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
473 WARNING("Reparse point fixup of \"%"TS"\" "
482 ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx);
484 /* On Windows, the SeCreateSymbolicLink privilege is required to create
485 * symbolic links. To be more friendly towards non-Administrator users,
486 * we merely warn the user if symbolic links cannot be created due to
487 * insufficient permissions or privileges, unless
488 * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided. */
490 if (ret && inode_is_symlink(inode) &&
491 (errno == EACCES || errno == EPERM) &&
492 !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
494 WARNING("Can't set reparse data on \"%"TS"\": "
496 " You may be trying to "
497 "extract a symbolic link without the\n"
498 " SeCreateSymbolicLink privilege, "
499 "which by default non-Administrator\n"
500 " accounts do not have.",
508 /* Account for reparse data consumed. */
509 update_extract_progress(ctx,
510 (lte_override ? lte_override :
511 inode_unnamed_lte_resolved(inode)));
515 ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path);
520 * Extract zero or more streams to a file.
522 * This function operates slightly differently depending on whether @lte_spec is
523 * NULL or not. When @lte_spec is NULL, the behavior is to extract the default
524 * file contents (unnamed stream), and, if named data streams are supported in
525 * the extract mode and volume, any named data streams. When @lte_spec is NULL,
526 * the behavior is to extract only all copies of the stream @lte_spec, and in
527 * addition use @lte_spec to set the reparse data or create the symbolic link if
531 * Path to file to extract (as can be passed to apply_operations
536 * WIM dentry that corresponds to the file being extracted.
538 * If non-NULL, specifies the lookup table entry for a stream to extract,
539 * and only that stream will be extracted (although there may be more than
540 * one instance of it).
542 * Used only if @lte_spec != NULL; it is passed to the extraction functions
543 * rather than @lte_spec, allowing the location of the stream to be
544 * overridden. (This is used when the WIM is being read from a nonseekable
545 * file, such as a pipe, when streams need to be used more than once; each
546 * such stream is extracted to a temporary file.)
549 extract_streams(const tchar *path, struct apply_ctx *ctx,
550 struct wim_dentry *dentry,
551 struct wim_lookup_table_entry *lte_spec,
552 struct wim_lookup_table_entry *lte_override)
554 struct wim_inode *inode = dentry->d_inode;
555 struct wim_lookup_table_entry *lte;
558 if (dentry->was_hardlinked)
565 for (unsigned i = 0; i < SHA1_HASH_SIZE; i++)
566 p += sprintf(p, "%02x", lte_override->hash[i]);
567 DEBUG("Extracting stream SHA1=%s to \"%"TS"\"",
568 sha1_str, path, inode->i_ino);
570 DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino);
574 /* Unnamed data stream. */
575 lte = inode_unnamed_lte_resolved(inode);
576 if (lte && (!lte_spec || lte == lte_spec)) {
579 if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
580 FILE_ATTRIBUTE_REPARSE_POINT)))
582 if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) &&
583 ctx->supported_features.encrypted_files)
584 ret = ctx->ops->extract_encrypted_stream(path, lte, ctx);
586 ret = ctx->ops->extract_unnamed_stream(path, lte, ctx);
589 update_extract_progress(ctx, lte);
591 else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
594 if (ctx->supported_features.reparse_points)
595 ret = extract_reparse_data(path, ctx, inode, lte);
597 else if ((inode_is_symlink(inode) &&
598 ctx->supported_features.symlink_reparse_points))
599 ret = extract_symlink(path, ctx, inode, lte);
606 /* Named data streams. */
607 if (can_extract_named_data_streams(ctx)) {
608 for (u16 i = 0; i < inode->i_num_ads; i++) {
609 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
611 if (!ads_entry_is_named_stream(entry))
616 if (lte_spec && lte_spec != lte)
620 ret = ctx->ops->extract_named_stream(path, entry->stream_name,
621 entry->stream_name_nbytes / 2,
625 update_extract_progress(ctx, lte);
631 ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path);
635 /* Set attributes on an extracted file or directory if supported by the
636 * extraction mode. */
638 extract_file_attributes(const tchar *path, struct apply_ctx *ctx,
639 struct wim_dentry *dentry)
643 if (ctx->ops->set_file_attributes) {
644 if (dentry == ctx->extract_root && ctx->root_dentry_is_special)
646 ret = ctx->ops->set_file_attributes(path,
647 dentry->d_inode->i_attributes,
650 ERROR_WITH_ERRNO("Failed to set attributes on "
659 /* Set or remove the short (DOS) name on an extracted file or directory if
660 * supported by the extraction mode. Since DOS names are unimportant and it's
661 * easy to run into problems setting them on Windows (SetFileShortName()
662 * requires SE_RESTORE privilege, which only the Administrator can request, and
663 * also requires DELETE access to the file), failure is ignored unless
664 * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set. */
666 extract_short_name(const tchar *path, struct apply_ctx *ctx,
667 struct wim_dentry *dentry)
671 /* The root of the dentry tree being extracted may not be extracted to
672 * its original name, so its short name should be ignored. */
673 if (dentry == ctx->extract_root)
676 if (ctx->supported_features.short_names) {
677 ret = ctx->ops->set_short_name(path,
679 dentry->short_name_nbytes / 2,
681 if (ret && (ctx->extract_flags &
682 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES))
684 ERROR_WITH_ERRNO("Failed to set short name of "
692 /* Set security descriptor, UNIX data, or neither on an extracted file, taking
693 * into account the current extraction mode and flags. */
695 extract_security(const tchar *path, struct apply_ctx *ctx,
696 struct wim_dentry *dentry)
699 struct wim_inode *inode = dentry->d_inode;
701 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS)
704 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
708 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
709 struct wimlib_unix_data data;
711 ret = inode_get_unix_data(inode, &data, NULL);
715 ret = ctx->ops->set_unix_data(path, &data, ctx);
717 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
718 ERROR_WITH_ERRNO("Failed to set UNIX owner, "
719 "group, and/or mode on "
723 WARNING_WITH_ERRNO("Failed to set UNIX owner, "
724 "group, and/or/mode on "
730 #endif /* __WIN32__ */
731 if (ctx->supported_features.security_descriptors &&
732 inode->i_security_id != -1)
734 const struct wim_security_data *sd;
738 sd = wim_const_security_data(ctx->wim);
739 desc = sd->descriptors[inode->i_security_id];
740 desc_size = sd->sizes[inode->i_security_id];
742 ret = ctx->ops->set_security_descriptor(path, desc,
744 !!(ctx->extract_flags &
745 WIMLIB_EXTRACT_FLAG_STRICT_ACLS));
747 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
748 ERROR_WITH_ERRNO("Failed to set security "
749 "descriptor on \"%"TS"\"", path);
752 if (errno != EACCES) {
753 WARNING_WITH_ERRNO("Failed to set "
754 "security descriptor "
755 "on \"%"TS"\"", path);
763 /* Set timestamps on an extracted file. Failure is warning-only unless
764 * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */
766 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
767 struct wim_dentry *dentry)
769 struct wim_inode *inode = dentry->d_inode;
772 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
775 if (ctx->ops->set_timestamps) {
776 ret = ctx->ops->set_timestamps(path,
777 inode->i_creation_time,
778 inode->i_last_write_time,
779 inode->i_last_access_time,
782 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
783 ERROR_WITH_ERRNO("Failed to set timestamps "
784 "on \"%"TS"\"", path);
787 WARNING_WITH_ERRNO("Failed to set timestamps "
788 "on \"%"TS"\"", path);
795 /* Check whether the extraction of a dentry should be skipped completely. */
797 dentry_is_supported(struct wim_dentry *dentry,
798 const struct wim_features *supported_features)
800 struct wim_inode *inode = dentry->d_inode;
802 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
803 if (supported_features->reparse_points)
805 if (supported_features->symlink_reparse_points &&
806 inode_is_symlink(inode))
813 /* Given a WIM dentry to extract, build the path to which to extract it, in the
814 * format understood by the callbacks in the apply_operations being used.
816 * Write the resulting path into @path, which must have room for at least
817 * ctx->ops->max_path characters including the null-terminator.
819 * Return %true if successful; %false if this WIM dentry doesn't actually need
820 * to be extracted or if the calculated path exceeds ctx->ops->max_path
823 * This function clobbers the tmp_list member of @dentry and its ancestors up
824 * until the extraction root. */
826 build_extraction_path(tchar path[], struct wim_dentry *dentry,
827 struct apply_ctx *ctx)
830 LIST_HEAD(ancestor_list);
832 const tchar *target_prefix;
833 size_t target_prefix_nchars;
834 struct wim_dentry *d;
836 if (dentry->extraction_skipped)
839 path_nchars = ctx->ops->path_prefix_nchars;
841 if (ctx->ops->requires_realtarget_in_paths) {
842 target_prefix = ctx->realtarget;
843 target_prefix_nchars = ctx->realtarget_nchars;
844 } else if (ctx->ops->requires_target_in_paths) {
845 target_prefix = ctx->target;
846 target_prefix_nchars = ctx->target_nchars;
848 target_prefix = NULL;
849 target_prefix_nchars = 0;
851 path_nchars += target_prefix_nchars;
853 for (d = dentry; d != ctx->extract_root; d = d->parent) {
854 path_nchars += d->extraction_name_nchars + 1;
855 list_add(&d->tmp_list, &ancestor_list);
858 path_nchars++; /* null terminator */
860 if (path_nchars > ctx->ops->path_max) {
861 WARNING("\"%"TS"\": Path too long to extract",
862 dentry_full_path(dentry));
866 p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
867 p = tmempcpy(p, target_prefix, target_prefix_nchars);
868 list_for_each_entry(d, &ancestor_list, tmp_list) {
869 *p++ = ctx->ops->path_separator;
870 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
873 wimlib_assert(p - path == path_nchars);
878 get_num_path_components(const tchar *path, tchar path_separator)
880 unsigned num_components = 0;
883 while (*path == path_separator)
887 while (*path && *path != path_separator)
890 return num_components;
894 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
895 struct apply_ctx *ctx, struct wim_dentry *dentry)
897 size_t num_raw_path_components;
898 const struct wim_dentry *d;
899 size_t num_target_path_components;
904 num_raw_path_components = 0;
905 for (d = dentry; d != ctx->extract_root; d = d->parent)
906 num_raw_path_components++;
908 if (ctx->ops->requires_realtarget_in_paths)
909 num_target_path_components = get_num_path_components(ctx->realtarget,
910 ctx->ops->path_separator);
911 else if (ctx->ops->requires_target_in_paths)
912 num_target_path_components = get_num_path_components(ctx->target,
913 ctx->ops->path_separator);
915 num_target_path_components = 0;
917 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
918 wimlib_assert(num_target_path_components > 0);
919 num_raw_path_components++;
920 num_target_path_components--;
924 while (*p_old == ctx->ops->path_separator)
926 while (--num_target_path_components) {
927 while (*p_old != ctx->ops->path_separator)
929 while (*p_old == ctx->ops->path_separator)
933 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
935 p = &symlink_target[0];
936 while (num_raw_path_components--) {
939 *p++ = ctx->ops->path_separator;
942 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
943 newpath, symlink_target);
944 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
946 ERROR_WITH_ERRNO("Failed to create symlink "
947 "\"%"TS"\" => \"%"TS"\"",
948 newpath, symlink_target);
953 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
954 * data streams, reparse data (including symbolic links), timestamps, and
955 * security descriptors. Basically, everything that doesn't require reading
956 * non-metadata resources from the WIM file and isn't delayed until the final
959 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
960 struct apply_ctx *ctx)
962 struct wim_inode *inode = dentry->d_inode;
964 const tchar *oldpath;
966 if (unlikely(is_linked_extraction(ctx))) {
967 struct wim_lookup_table_entry *unnamed_lte;
969 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
970 if (unnamed_lte && unnamed_lte->extracted_file) {
971 oldpath = unnamed_lte->extracted_file;
972 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
979 /* Create hard link if this dentry corresponds to an already-extracted
981 if (inode->i_extracted_file) {
982 oldpath = inode->i_extracted_file;
986 /* Skip symlinks unless they can be extracted as reparse points rather
987 * than created directly. */
988 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
991 /* Create this file or directory unless it's the extraction root, which
992 * was already created if necessary. */
993 if (dentry != ctx->extract_root) {
994 ret = extract_inode(path, ctx, inode);
999 /* Create empty named data streams. */
1000 if (can_extract_named_data_streams(ctx)) {
1001 for (u16 i = 0; i < inode->i_num_ads; i++) {
1002 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1004 if (!ads_entry_is_named_stream(entry))
1008 ret = ctx->ops->extract_named_stream(path,
1010 entry->stream_name_nbytes / 2,
1013 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1014 "empty named data stream",
1021 /* Set file attributes (if supported). */
1022 ret = extract_file_attributes(path, ctx, dentry);
1026 /* Set or remove file short name (if supported). */
1027 ret = extract_short_name(path, ctx, dentry);
1031 /* If inode has multiple links and hard links are supported in this
1032 * extraction mode and volume, save the path to the extracted file in
1033 * case it's needed to create a hard link. */
1034 if (unlikely(is_linked_extraction(ctx))) {
1035 struct wim_lookup_table_entry *unnamed_lte;
1037 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1039 unnamed_lte->extracted_file = TSTRDUP(path);
1040 if (!unnamed_lte->extracted_file)
1041 return WIMLIB_ERR_NOMEM;
1043 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1044 inode->i_extracted_file = TSTRDUP(path);
1045 if (!inode->i_extracted_file)
1046 return WIMLIB_ERR_NOMEM;
1051 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1054 dentry->was_hardlinked = 1;
1058 ret = extract_hardlink(oldpath, path, ctx);
1061 dentry->was_hardlinked = 1;
1066 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1068 struct apply_ctx *ctx = _ctx;
1069 tchar path[ctx->ops->path_max];
1071 if (!build_extraction_path(path, dentry, ctx))
1073 return do_dentry_extract_skeleton(path, dentry, ctx);
1076 /* Create a file or directory, then immediately extract all streams. This
1077 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1078 * may not be read sequentially by this function. */
1080 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1082 struct apply_ctx *ctx = _ctx;
1083 tchar path[ctx->ops->path_max];
1086 if (!build_extraction_path(path, dentry, ctx))
1089 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1093 return extract_streams(path, ctx, dentry, NULL, NULL);
1096 /* Extract all instances of the stream @lte that are being extracted in this
1097 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1098 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1099 * extracted multiple times, it is extracted to a temporary file first.
1101 * This is intended for use with sequential extraction of a WIM image
1102 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1104 extract_stream_instances(struct wim_lookup_table_entry *lte,
1105 struct apply_ctx *ctx, bool can_seek)
1107 struct wim_dentry **lte_dentries;
1108 struct wim_lookup_table_entry *lte_tmp = NULL;
1109 struct wim_lookup_table_entry *lte_override;
1110 tchar *stream_tmp_filename = NULL;
1111 tchar path[ctx->ops->path_max];
1115 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1116 lte_dentries = lte->inline_lte_dentries;
1118 lte_dentries = lte->lte_dentries;
1120 if (likely(can_seek || lte->out_refcnt < 2)) {
1123 /* Need to extract stream to temporary file. */
1127 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1128 if (!stream_tmp_filename) {
1129 ERROR_WITH_ERRNO("Failed to create temporary filename");
1130 ret = WIMLIB_ERR_OPEN;
1134 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1136 ret = WIMLIB_ERR_NOMEM;
1137 goto out_free_stream_tmp_filename;
1139 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1140 lte_tmp->file_on_disk = stream_tmp_filename;
1141 lte_override = lte_tmp;
1143 raw_fd = topen(stream_tmp_filename,
1144 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1146 ERROR_WITH_ERRNO("Failed to open temporary file");
1147 ret = WIMLIB_ERR_OPEN;
1148 goto out_free_lte_tmp;
1150 filedes_init(&fd, raw_fd);
1151 ret = extract_wim_resource_to_fd(lte, &fd,
1152 wim_resource_size(lte));
1153 if (filedes_close(&fd) && !ret)
1154 ret = WIMLIB_ERR_WRITE;
1156 goto out_unlink_stream_tmp_file;
1159 /* Extract all instances of the stream, reading either from the stream
1160 * in the WIM file or from the temporary file containing the stream.
1161 * dentry->tmp_flag is used to ensure that each dentry is processed only
1162 * once regardless of how many times this stream appears in the streams
1163 * of the corresponding inode. */
1164 for (i = 0; i < lte->out_refcnt; i++) {
1165 struct wim_dentry *dentry = lte_dentries[i];
1167 if (dentry->tmp_flag)
1169 if (!build_extraction_path(path, dentry, ctx))
1171 ret = extract_streams(path, ctx, dentry,
1174 goto out_clear_tmp_flags;
1175 dentry->tmp_flag = 1;
1178 out_clear_tmp_flags:
1179 for (i = 0; i < lte->out_refcnt; i++)
1180 lte_dentries[i]->tmp_flag = 0;
1181 out_unlink_stream_tmp_file:
1182 if (stream_tmp_filename)
1183 tunlink(stream_tmp_filename);
1186 out_free_stream_tmp_filename:
1187 FREE(stream_tmp_filename);
1192 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1193 * structure and empty files were already created. This relies on the
1194 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1195 * stream that was constructed earlier. Streams are extracted exactly in the
1196 * order of the stream list; however, unless the WIM's file descriptor is
1197 * detected to be non-seekable, streams may be read from the WIM file more than
1198 * one time if multiple copies need to be extracted. */
1200 extract_stream_list(struct apply_ctx *ctx)
1202 struct wim_lookup_table_entry *lte;
1206 can_seek = !(lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) == (off_t)-1 &&
1208 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1209 ret = extract_stream_instances(lte, ctx, can_seek);
1216 /* Read the header from a stream in a pipable WIM. */
1218 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1221 struct pwm_stream_hdr stream_hdr;
1224 ret = full_read(&pwm->in_fd, &stream_hdr, sizeof(stream_hdr));
1228 if (allow_header && stream_hdr.magic == PWM_MAGIC) {
1229 u8 buf[WIM_HEADER_DISK_SIZE - sizeof(stream_hdr)];
1230 ret = full_read(&pwm->in_fd, buf, sizeof(buf));
1233 lte->resource_location = RESOURCE_NONEXISTENT;
1237 if (stream_hdr.magic != PWM_STREAM_MAGIC) {
1238 ERROR("Data read on pipe is invalid (expected stream header).");
1239 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1242 lte->resource_entry.original_size = le64_to_cpu(stream_hdr.uncompressed_size);
1243 copy_hash(lte->hash, stream_hdr.hash);
1244 lte->resource_entry.flags = le32_to_cpu(stream_hdr.flags);
1245 lte->resource_entry.offset = pwm->in_fd.offset;
1246 lte->resource_location = RESOURCE_IN_WIM;
1248 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1249 lte->compression_type = pwm->compression_type;
1250 lte->resource_entry.size = 0;
1252 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1253 lte->resource_entry.size = lte->resource_entry.original_size;
1255 lte->is_pipable = 1;
1259 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1263 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1265 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1267 return read_partial_wim_resource(lte, wim_resource_size(lte),
1269 WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY,
1274 extract_streams_from_pipe(struct apply_ctx *ctx)
1276 struct wim_lookup_table_entry *found_lte;
1277 struct wim_lookup_table_entry *needed_lte;
1278 struct wim_lookup_table *lookup_table;
1281 ret = WIMLIB_ERR_NOMEM;
1282 found_lte = new_lookup_table_entry();
1286 lookup_table = ctx->wim->lookup_table;
1288 while (ctx->num_streams_remaining) {
1289 ret = read_pwm_stream_header(ctx->wim, found_lte, true);
1291 goto out_free_found_lte;
1293 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1294 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1295 && (needed_lte = __lookup_resource(lookup_table, found_lte->hash))
1296 && (needed_lte->out_refcnt))
1298 copy_resource_entry(&needed_lte->resource_entry,
1299 &found_lte->resource_entry);
1300 needed_lte->resource_location = found_lte->resource_location;
1301 needed_lte->wim = found_lte->wim;
1302 needed_lte->compression_type = found_lte->compression_type;
1303 needed_lte->is_pipable = found_lte->is_pipable;
1305 ret = extract_stream_instances(needed_lte, ctx, false);
1307 goto out_free_found_lte;
1308 ctx->num_streams_remaining--;
1309 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1310 ret = skip_pwm_stream(found_lte);
1312 goto out_free_found_lte;
1317 free_lookup_table_entry(found_lte);
1322 /* Finish extracting a file, directory, or symbolic link by setting file
1323 * security and timestamps. */
1325 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1327 struct apply_ctx *ctx = _ctx;
1329 tchar path[ctx->ops->path_max];
1331 if (!build_extraction_path(path, dentry, ctx))
1334 ret = extract_security(path, ctx, dentry);
1338 return extract_timestamps(path, ctx, dentry);
1341 /* Sorts a list of streams in ascending order of their offset in the WIM file in
1342 * order to prepare for sequential extraction. */
1344 sort_stream_list_by_wim_position(struct list_head *stream_list)
1346 struct list_head *cur;
1348 struct wim_lookup_table_entry **array;
1353 list_for_each(cur, stream_list)
1355 array_size = num_streams * sizeof(array[0]);
1356 array = MALLOC(array_size);
1358 ERROR("Failed to allocate %zu bytes to sort stream entries",
1360 return WIMLIB_ERR_NOMEM;
1362 cur = stream_list->next;
1363 for (i = 0; i < num_streams; i++) {
1364 array[i] = container_of(cur, struct wim_lookup_table_entry, extraction_list);
1368 qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position);
1370 INIT_LIST_HEAD(stream_list);
1371 for (i = 0; i < num_streams; i++)
1372 list_add_tail(&array[i]->extraction_list, stream_list);
1378 * Extract a WIM dentry to standard output.
1380 * This obviously doesn't make sense in all cases. We return an error if the
1381 * dentry does not correspond to a regular file. Otherwise we extract the
1382 * unnamed data stream only.
1385 extract_dentry_to_stdout(struct wim_dentry *dentry)
1388 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1389 FILE_ATTRIBUTE_DIRECTORY))
1391 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1392 "extracted to standard output", dentry_full_path(dentry));
1393 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1395 struct wim_lookup_table_entry *lte;
1397 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1399 struct filedes _stdout;
1400 filedes_init(&_stdout, STDOUT_FILENO);
1401 ret = extract_wim_resource_to_fd(lte, &_stdout,
1402 wim_resource_size(lte));
1409 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1411 static const utf16lechar replacement_char = cpu_to_le16('?');
1415 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1421 for (i = 0; i < num_chars; i++) {
1424 case cpu_to_le16('\\'):
1425 case cpu_to_le16(':'):
1426 case cpu_to_le16('*'):
1427 case cpu_to_le16('?'):
1428 case cpu_to_le16('"'):
1429 case cpu_to_le16('<'):
1430 case cpu_to_le16('>'):
1431 case cpu_to_le16('|'):
1433 case cpu_to_le16('/'):
1434 case cpu_to_le16('\0'):
1436 name[i] = replacement_char;
1443 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1444 name[num_chars - 1] == cpu_to_le16('.'))
1447 name[num_chars - 1] = replacement_char;
1456 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1458 const utf16lechar *file_name = dentry->file_name;
1459 return file_name != NULL &&
1460 file_name[0] == cpu_to_le16('.') &&
1461 (file_name[1] == cpu_to_le16('\0') ||
1462 (file_name[1] == cpu_to_le16('.') &&
1463 file_name[2] == cpu_to_le16('\0')));
1467 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1469 dentry->extraction_skipped = 1;
1474 * dentry_calculate_extraction_path-
1476 * Calculate the actual filename component at which a WIM dentry will be
1477 * extracted, handling invalid filenames "properly".
1479 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1480 * UNIX, converted into the platform's multibyte encoding). However, if the
1481 * file name contains characters that are not valid on the current platform or
1482 * has some other format that is not valid, leave dentry->extraction_name as
1483 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1484 * not be extracted, unless the appropriate flag
1485 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1486 * which case a substitute filename will be created and set instead.
1488 * Conflicts with case-insensitive names on Windows are handled similarly; see
1492 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1494 struct apply_ctx *ctx = _args;
1497 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1500 if (!dentry_is_supported(dentry, &ctx->supported_features))
1503 if (dentry_is_dot_or_dotdot(dentry)) {
1504 /* WIM files shouldn't contain . or .. entries. But if they are
1505 * there, don't attempt to extract them. */
1506 WARNING("Skipping extraction of unexpected . or .. file "
1507 "\"%"TS"\"", dentry_full_path(dentry));
1512 if (!ctx->ops->supports_case_sensitive_filenames)
1514 struct wim_dentry *other;
1515 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1516 case_insensitive_conflict_list)
1518 if (ctx->extract_flags &
1519 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1520 WARNING("\"%"TS"\" has the same "
1521 "case-insensitive name as "
1522 "\"%"TS"\"; extracting "
1523 "dummy name instead",
1524 dentry_full_path(dentry),
1525 dentry_full_path(other));
1528 WARNING("Not extracting \"%"TS"\": "
1529 "has same case-insensitive "
1530 "name as \"%"TS"\"",
1531 dentry_full_path(dentry),
1532 dentry_full_path(other));
1537 #else /* __WIN32__ */
1538 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1539 #endif /* !__WIN32__ */
1541 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1543 dentry->extraction_name = dentry->file_name;
1544 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1547 return utf16le_to_tstr(dentry->file_name,
1548 dentry->file_name_nbytes,
1549 &dentry->extraction_name,
1550 &dentry->extraction_name_nchars);
1553 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1555 WARNING("\"%"TS"\" has an invalid filename "
1556 "that is not supported on this platform; "
1557 "extracting dummy name instead",
1558 dentry_full_path(dentry));
1561 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1562 "that is not supported on this platform",
1563 dentry_full_path(dentry));
1570 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1572 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1573 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1576 size_t tchar_nchars;
1578 tchar_name = utf16_name_copy;
1579 tchar_nchars = dentry->file_name_nbytes / 2;
1581 ret = utf16le_to_tstr(utf16_name_copy,
1582 dentry->file_name_nbytes,
1583 &tchar_name, &tchar_nchars);
1587 size_t fixed_name_num_chars = tchar_nchars;
1588 tchar fixed_name[tchar_nchars + 50];
1590 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1591 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1592 T(" (invalid filename #%lu)"),
1593 ++ctx->invalid_sequence);
1597 dentry->extraction_name = memdup(fixed_name,
1598 2 * fixed_name_num_chars + 2);
1599 if (!dentry->extraction_name)
1600 return WIMLIB_ERR_NOMEM;
1601 dentry->extraction_name_nchars = fixed_name_num_chars;
1606 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1610 /* Clean up dentry and inode structure after extraction. */
1612 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1614 struct wim_inode *inode = dentry->d_inode;
1616 dentry->extraction_skipped = 0;
1617 dentry->was_hardlinked = 0;
1618 inode->i_visited = 0;
1619 FREE(inode->i_extracted_file);
1620 inode->i_extracted_file = NULL;
1621 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1622 FREE(dentry->extraction_name);
1623 dentry->extraction_name = NULL;
1627 /* Tally features necessary to extract a dentry and the corresponding inode. */
1629 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1631 struct wim_features *features = _features;
1632 struct wim_inode *inode = dentry->d_inode;
1634 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1635 features->archive_files++;
1636 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1637 features->hidden_files++;
1638 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1639 features->system_files++;
1640 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1641 features->compressed_files++;
1642 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED)
1643 features->encrypted_files++;
1644 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1645 features->not_context_indexed_files++;
1646 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1647 features->sparse_files++;
1648 if (inode_has_named_stream(inode))
1649 features->named_data_streams++;
1650 if (inode->i_visited)
1651 features->hard_links++;
1652 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1653 features->reparse_points++;
1654 if (inode_is_symlink(inode))
1655 features->symlink_reparse_points++;
1657 features->other_reparse_points++;
1659 if (inode->i_security_id != -1)
1660 features->security_descriptors++;
1661 if (dentry->short_name_nbytes)
1662 features->short_names++;
1663 if (inode_has_unix_data(inode))
1664 features->unix_data++;
1665 inode->i_visited = 1;
1670 dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
1672 dentry->d_inode->i_visited = 0;
1676 /* Tally the features necessary to extract a dentry tree. */
1678 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1680 memset(features, 0, sizeof(struct wim_features));
1681 for_dentry_in_tree(root, dentry_tally_features, features);
1682 for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
1686 do_feature_check(const struct wim_features *required_features,
1687 const struct wim_features *supported_features,
1689 const struct apply_operations *ops)
1691 if (required_features->archive_files && !supported_features->archive_files)
1694 "%lu files are marked as archived, but this attribute\n"
1695 " is not supported in this extraction mode or volume.",
1696 required_features->archive_files);
1699 if (required_features->hidden_files && !supported_features->hidden_files)
1702 "%lu files are marked as hidden, but this attribute\n"
1703 " is not supported in this extraction mode or volume.",
1704 required_features->hidden_files);
1707 if (required_features->system_files && !supported_features->system_files)
1710 "%lu files are marked as system files, but this attribute\n"
1711 " is not supported in this extraction mode or volume.",
1712 required_features->system_files);
1715 if (required_features->compressed_files && !supported_features->compressed_files)
1718 "%lu files are marked as being transparently compressed, but\n"
1719 " transparent compression is not supported in this extraction\n"
1720 " mode or volume. These files will be extracted as uncompressed.",
1721 required_features->compressed_files);
1724 if (required_features->encrypted_files && !supported_features->encrypted_files)
1727 "%lu files are marked as being encrypted, but encryption is not\n"
1728 " supported in this extraction mode or volume. These files will be\n"
1729 " extracted as raw encrypted data instead.",
1730 required_features->encrypted_files);
1733 if (required_features->not_context_indexed_files &&
1734 !supported_features->not_context_indexed_files)
1737 "%lu files are marked as not content indexed, but this attribute\n"
1738 " is not supported in this extraction mode or volume.",
1739 required_features->not_context_indexed_files);
1742 if (required_features->sparse_files && !supported_features->sparse_files)
1745 "%lu files are marked as sparse, but creating sparse files is not\n"
1746 " supported in this extraction mode or volume. These files will be\n"
1747 " extracted as non-sparse.",
1748 required_features->not_context_indexed_files);
1751 if (required_features->named_data_streams &&
1752 !supported_features->named_data_streams)
1755 "%lu files contain one or more alternate (named) data streams,\n"
1756 " which are not supported in this extraction mode or volume.\n"
1757 " Alternate data streams will NOT be extracted.",
1758 required_features->named_data_streams);
1761 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1762 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1763 required_features->named_data_streams &&
1764 supported_features->named_data_streams)
1767 "%lu files contain one or more alternate (named) data streams,\n"
1768 " which are not supported in linked extraction mode.\n"
1769 " Alternate data streams will NOT be extracted.",
1770 required_features->named_data_streams);
1773 if (required_features->hard_links && !supported_features->hard_links)
1776 "%lu files are hard links, but hard links are not supported in\n"
1777 " this extraction mode or volume. Hard links will be extracted as\n"
1778 " duplicate copies of the linked files.",
1779 required_features->hard_links);
1782 if (required_features->reparse_points && !supported_features->reparse_points)
1784 if (supported_features->symlink_reparse_points) {
1785 if (required_features->other_reparse_points) {
1787 "%lu files are reparse points that are neither symbolic links\n"
1788 " nor junction points and are not supported in this extraction mode\n"
1789 " or volume. These reparse points will not be extracted.",
1790 required_features->other_reparse_points);
1794 "%lu files are reparse points, which are not supported in this\n"
1795 " extraction mode or volume and will not be extracted.",
1796 required_features->reparse_points);
1800 if (required_features->security_descriptors &&
1801 !supported_features->security_descriptors)
1804 "%lu files have Windows NT security descriptors, but extracting\n"
1805 " security descriptors is not supported in this extraction mode\n"
1806 " or volume. No security descriptors will be extracted.",
1807 required_features->security_descriptors);
1810 if (required_features->short_names && !supported_features->short_names)
1813 "%lu files have short (DOS) names, but extracting short names\n"
1814 " is not supported in this extraction mode or volume. Short names\n"
1815 " will not be extracted.\n",
1816 required_features->short_names);
1819 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1820 required_features->unix_data && !supported_features->unix_data)
1822 ERROR("UNIX data not supported in this extraction mode "
1823 "or volume", ops->name);
1824 return WIMLIB_ERR_UNSUPPORTED;
1826 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
1827 required_features->short_names && !supported_features->short_names)
1829 ERROR("Short names are not supported in this extraction "
1830 "mode or volume", ops->name);
1831 return WIMLIB_ERR_UNSUPPORTED;
1833 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1834 !ops->set_timestamps)
1836 ERROR("Timestamps are not supported in this extraction "
1837 "mode or volume", ops->name);
1838 return WIMLIB_ERR_UNSUPPORTED;
1840 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1841 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1842 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1843 required_features->security_descriptors &&
1844 !supported_features->security_descriptors)
1846 ERROR("Security descriptors not supported in this extraction "
1848 return WIMLIB_ERR_UNSUPPORTED;
1851 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
1852 !supported_features->hard_links)
1854 ERROR("Hard link extraction mode requested, but "
1855 "extraction mode or volume does not support hard links!");
1856 return WIMLIB_ERR_UNSUPPORTED;
1859 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
1860 !supported_features->symlink_reparse_points)
1862 ERROR("Symbolic link extraction mode requested, but "
1863 "extraction mode or volume does not support symbolic "
1865 return WIMLIB_ERR_UNSUPPORTED;
1871 * extract_tree - Extract a file or directory tree from the currently selected
1874 * @wim: WIMStruct for the WIM file, with the desired image selected
1875 * (as wim->current_image).
1878 * "Canonical" (i.e. no leading or trailing slashes, path
1879 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
1880 * extract. An empty string means the full image.
1883 * Filesystem path to extract the file or directory tree to.
1884 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
1887 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
1888 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
1889 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
1893 * If non-NULL, progress function for the extraction. The messages
1894 * that may be sent in this function are:
1896 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
1897 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
1898 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
1899 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
1900 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
1901 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
1902 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
1903 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
1904 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
1906 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
1909 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
1910 int extract_flags, wimlib_progress_func_t progress_func)
1912 struct wim_dentry *root;
1913 struct wim_features required_features;
1914 struct apply_ctx ctx;
1916 struct wim_lookup_table_entry *lte;
1918 /* Start initializing the apply_ctx. */
1919 memset(&ctx, 0, sizeof(struct apply_ctx));
1921 ctx.extract_flags = extract_flags;
1922 ctx.target = target;
1923 ctx.target_nchars = tstrlen(target);
1924 ctx.progress_func = progress_func;
1925 if (progress_func) {
1926 ctx.progress.extract.wimfile_name = wim->filename;
1927 ctx.progress.extract.image = wim->current_image;
1928 ctx.progress.extract.extract_flags = (extract_flags &
1929 WIMLIB_EXTRACT_MASK_PUBLIC);
1930 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
1931 wim->current_image);
1932 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
1933 ctx.progress.extract.target = target;
1935 INIT_LIST_HEAD(&ctx.stream_list);
1937 /* Translate the path to extract into the corresponding
1938 * `struct wim_dentry', which will be the root of the
1939 * "dentry tree" to extract. */
1940 root = get_dentry(wim, wim_source_path);
1942 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
1943 wim_source_path, wim->current_image);
1944 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1948 ctx.extract_root = root;
1950 /* Select the appropriate apply_operations based on the
1951 * platform and extract_flags. */
1953 ctx.ops = &win32_apply_ops;
1955 ctx.ops = &unix_apply_ops;
1959 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1960 ctx.ops = &ntfs_3g_apply_ops;
1963 /* Call the start_extract() callback. This gives the apply_operations
1964 * implementation a chance to do any setup needed to access the volume.
1965 * Furthermore, it's expected to set the supported features of this
1966 * extraction mode (ctx.supported_features), which are determined at
1967 * runtime as they may vary depending on the actual volume. These
1968 * features are then compared with the actual features extracting this
1969 * dentry tree requires. Some mismatches will merely produce warnings
1970 * and the unsupported data will be ignored; others will produce errors.
1972 ret = ctx.ops->start_extract(target, &ctx);
1976 dentry_tree_get_features(root, &required_features);
1977 ret = do_feature_check(&required_features, &ctx.supported_features,
1978 extract_flags, ctx.ops);
1980 goto out_finish_or_abort_extract;
1982 /* Figure out whether the root dentry is being extracted to the root of
1983 * a volume and therefore needs to be treated "specially", for example
1984 * not being explicitly created and not having attributes set. */
1985 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
1986 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
1988 /* Calculate the actual filename component of each extracted dentry. In
1989 * the process, set the dentry->extraction_skipped flag on dentries that
1990 * are being skipped for some reason (e.g. invalid filename). */
1991 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
1993 goto out_dentry_reset_needs_extraction;
1995 /* Build the list of the streams that need to be extracted and
1996 * initialize ctx.progress.extract with stream information. */
1997 ret = for_dentry_in_tree(ctx.extract_root,
1998 dentry_resolve_and_zero_lte_refcnt, &ctx);
2000 goto out_dentry_reset_needs_extraction;
2002 ret = for_dentry_in_tree(ctx.extract_root,
2003 dentry_add_streams_to_extract, &ctx);
2005 goto out_teardown_stream_list;
2007 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2008 /* When extracting from a pipe, the number of bytes of data to
2009 * extract can't be determined in the normal way (examining the
2010 * lookup table), since at this point all we have is a set of
2011 * SHA1 message digests of streams that need to be extracted.
2012 * However, we can get a reasonably accurate estimate by taking
2013 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2014 * data. This does assume that a full image is being extracted,
2015 * but currently there is no API for doing otherwise. */
2016 ctx.progress.extract.total_bytes =
2017 wim_info_get_image_total_bytes(wim->wim_info,
2018 wim->current_image);
2021 /* Handle the special case of extracting a file to standard
2022 * output. In that case, "root" should be a single file, not a
2023 * directory tree. (If not, extract_dentry_to_stdout() will
2024 * return an error.) */
2025 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT &&
2026 !(extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2027 ret = extract_dentry_to_stdout(root);
2028 goto out_teardown_stream_list;
2031 /* If a sequential extraction was specified, sort the streams to be
2032 * extracted by their position in the WIM file so that the WIM file can
2033 * be read sequentially. */
2034 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2035 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2036 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2038 ret = sort_stream_list_by_wim_position(&ctx.stream_list);
2040 goto out_teardown_stream_list;
2043 if (ctx.ops->realpath_works_on_nonexisting_files &&
2044 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2045 ctx.ops->requires_realtarget_in_paths))
2047 ctx.realtarget = realpath(target, NULL);
2048 if (!ctx.realtarget) {
2049 ret = WIMLIB_ERR_NOMEM;
2050 goto out_teardown_stream_list;
2052 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2055 if (progress_func) {
2056 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2057 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2061 if (!ctx.root_dentry_is_special)
2063 tchar path[ctx.ops->path_max];
2064 if (build_extraction_path(path, root, &ctx))
2066 ret = extract_inode(path, &ctx, root->d_inode);
2068 goto out_free_realtarget;
2072 /* If we need to fix up the targets of absolute symbolic links
2073 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2074 * be absolute, use realpath() (or its replacement on Windows) to get
2075 * the absolute path to the extraction target. Note that this requires
2076 * the target directory to exist, unless
2077 * realpath_works_on_nonexisting_files is set in the apply_operations.
2079 if (!ctx.realtarget &&
2080 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2081 required_features.symlink_reparse_points) ||
2082 ctx.ops->requires_realtarget_in_paths))
2084 ctx.realtarget = realpath(target, NULL);
2085 if (!ctx.realtarget) {
2086 ret = WIMLIB_ERR_NOMEM;
2087 goto out_free_realtarget;
2089 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2092 /* Finally, the important part: extract the tree of files. */
2093 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2094 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2095 /* Sequential extraction requested, so two passes are needed
2096 * (one for directory structure, one for streams.) */
2098 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2100 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2102 goto out_free_realtarget;
2104 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2106 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2107 ret = extract_streams_from_pipe(&ctx);
2109 ret = extract_stream_list(&ctx);
2111 goto out_free_realtarget;
2113 /* Sequential extraction was not requested, so we can make do
2114 * with one pass where we both create the files and extract
2117 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2119 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2121 goto out_free_realtarget;
2123 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2127 /* If the total number of bytes to extract was miscalculated, just jump
2128 * to the calculated number in order to avoid confusing the progress
2129 * function. This should only occur when extracting from a pipe. */
2130 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2132 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2133 "extracted %"PRIu64,
2134 ctx.progress.extract.total_bytes,
2135 ctx.progress.extract.completed_bytes);
2137 if (progress_func &&
2138 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2140 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2141 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2144 /* Apply security descriptors and timestamps. This is done at the end,
2145 * and in a depth-first manner, to prevent timestamps from getting
2146 * changed by subsequent extract operations and to minimize the chance
2147 * of the restored security descriptors getting in our way. */
2149 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2151 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2153 goto out_free_realtarget;
2155 if (progress_func) {
2156 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2157 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2162 out_free_realtarget:
2163 FREE(ctx.realtarget);
2164 out_teardown_stream_list:
2165 /* Free memory allocated as part of the mapping from each
2166 * wim_lookup_table_entry to the dentries that reference it. */
2167 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2168 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2169 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2170 FREE(lte->lte_dentries);
2171 out_dentry_reset_needs_extraction:
2172 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2173 out_finish_or_abort_extract:
2175 if (ctx.ops->abort_extract)
2176 ctx.ops->abort_extract(&ctx);
2178 if (ctx.ops->finish_extract)
2179 ret = ctx.ops->finish_extract(&ctx);
2185 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2186 * extract flags make sense. */
2188 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2192 /* Empty destination path? */
2193 if (cmd->fs_dest_path[0] == T('\0'))
2194 return WIMLIB_ERR_INVALID_PARAM;
2196 extract_flags = cmd->extract_flags;
2198 /* Check for invalid flag combinations */
2199 if ((extract_flags &
2200 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2201 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2202 WIMLIB_EXTRACT_FLAG_HARDLINK))
2203 return WIMLIB_ERR_INVALID_PARAM;
2205 if ((extract_flags &
2206 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2207 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2208 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2209 return WIMLIB_ERR_INVALID_PARAM;
2211 if ((extract_flags &
2212 (WIMLIB_EXTRACT_FLAG_RPFIX |
2213 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2214 WIMLIB_EXTRACT_FLAG_NORPFIX))
2215 return WIMLIB_ERR_INVALID_PARAM;
2217 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2218 #ifndef WITH_NTFS_3G
2219 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2220 " we cannot apply a WIM image directly to a NTFS volume.");
2221 return WIMLIB_ERR_UNSUPPORTED;
2225 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2226 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2228 /* Do reparse point fixups by default if the WIM header says
2229 * they are enabled and we are extracting a full image. */
2230 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2231 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2234 /* TODO: Since UNIX data entries are stored in the file resources, in a
2235 * completely sequential extraction they may come up before the
2236 * corresponding file or symbolic link data. This needs to be handled
2238 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2239 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2240 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2241 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2243 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2244 WARNING("Setting UNIX file/owner group may "
2245 "be impossible on some\n"
2247 "when applying from a pipe.");
2249 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2250 WARNING("Disabling sequential extraction for "
2255 cmd->extract_flags = extract_flags;
2260 /* Internal function to execute extraction commands for a WIM image. The paths
2261 * in the extract commands are expected to be already "canonicalized". */
2263 do_wimlib_extract_files(WIMStruct *wim,
2265 struct wimlib_extract_command *cmds,
2267 wimlib_progress_func_t progress_func)
2270 bool found_link_cmd = false;
2271 bool found_nolink_cmd = false;
2273 /* Select the image from which we are extracting files */
2274 ret = select_wim_image(wim, image);
2278 /* Make sure there are no streams in the WIM that have not been
2279 * checksummed yet. Needed at least because 'unhashed_list' aliases
2280 * 'extraction_list' in `struct wim_lookup_table_entry'. */
2281 ret = wim_checksum_unhashed_streams(wim);
2285 /* Check for problems with the extraction commands */
2286 for (size_t i = 0; i < num_cmds; i++) {
2287 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2290 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2291 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2292 found_link_cmd = true;
2294 found_nolink_cmd = true;
2296 if (found_link_cmd && found_nolink_cmd) {
2297 ERROR("Symlink or hardlink extraction mode must "
2298 "be set on all extraction commands");
2299 return WIMLIB_ERR_INVALID_PARAM;
2303 /* Execute the extraction commands */
2304 for (size_t i = 0; i < num_cmds; i++) {
2305 ret = extract_tree(wim,
2306 cmds[i].wim_source_path,
2307 cmds[i].fs_dest_path,
2308 cmds[i].extract_flags,
2316 /* API function documented in wimlib.h */
2318 wimlib_extract_files(WIMStruct *wim,
2320 const struct wimlib_extract_command *cmds,
2322 int default_extract_flags,
2323 WIMStruct **additional_swms,
2324 unsigned num_additional_swms,
2325 wimlib_progress_func_t progress_func)
2328 struct wimlib_extract_command *cmds_copy;
2331 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2333 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2340 if (num_additional_swms)
2341 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2343 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2345 ret = WIMLIB_ERR_NOMEM;
2346 goto out_restore_lookup_table;
2349 for (size_t i = 0; i < num_cmds; i++) {
2350 cmds_copy[i].extract_flags = (default_extract_flags |
2351 cmds[i].extract_flags)
2352 & WIMLIB_EXTRACT_MASK_PUBLIC;
2353 all_flags |= cmds_copy[i].extract_flags;
2355 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
2356 if (!cmds_copy[i].wim_source_path) {
2357 ret = WIMLIB_ERR_NOMEM;
2358 goto out_free_cmds_copy;
2361 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
2362 if (!cmds_copy[i].fs_dest_path) {
2363 ret = WIMLIB_ERR_NOMEM;
2364 goto out_free_cmds_copy;
2368 ret = do_wimlib_extract_files(wim, image,
2369 cmds_copy, num_cmds,
2372 if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2373 WIMLIB_EXTRACT_FLAG_HARDLINK))
2375 for_lookup_table_entry(wim->lookup_table,
2376 lte_free_extracted_file, NULL);
2379 for (size_t i = 0; i < num_cmds; i++) {
2380 FREE(cmds_copy[i].wim_source_path);
2381 FREE(cmds_copy[i].fs_dest_path);
2384 out_restore_lookup_table:
2385 if (num_additional_swms)
2386 unmerge_lookup_table(wim);
2392 * Extracts an image from a WIM file.
2394 * @wim: WIMStruct for the WIM file.
2396 * @image: Number of the single image to extract.
2398 * @target: Directory or NTFS volume to extract the image to.
2400 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2402 * @progress_func: If non-NULL, a progress function to be called
2405 * Returns 0 on success; nonzero on failure.
2408 extract_single_image(WIMStruct *wim, int image,
2409 const tchar *target, int extract_flags,
2410 wimlib_progress_func_t progress_func)
2413 tchar *target_copy = canonicalize_fs_path(target);
2415 return WIMLIB_ERR_NOMEM;
2416 struct wimlib_extract_command cmd = {
2417 .wim_source_path = T(""),
2418 .fs_dest_path = target_copy,
2419 .extract_flags = extract_flags,
2421 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2426 static const tchar * const filename_forbidden_chars =
2435 /* This function checks if it is okay to use a WIM image's name as a directory
2438 image_name_ok_as_dir(const tchar *image_name)
2440 return image_name && *image_name &&
2441 !tstrpbrk(image_name, filename_forbidden_chars) &&
2442 tstrcmp(image_name, T(".")) &&
2443 tstrcmp(image_name, T(".."));
2446 /* Extracts all images from the WIM to the directory @target, with the images
2447 * placed in subdirectories named by their image names. */
2449 extract_all_images(WIMStruct *wim,
2450 const tchar *target,
2452 wimlib_progress_func_t progress_func)
2454 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2455 size_t output_path_len = tstrlen(target);
2456 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2459 const tchar *image_name;
2462 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2465 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2466 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2467 return WIMLIB_ERR_INVALID_PARAM;
2471 if (tstat(target, &stbuf)) {
2472 if (errno == ENOENT) {
2473 if (tmkdir(target, 0755)) {
2474 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2475 return WIMLIB_ERR_MKDIR;
2478 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2479 return WIMLIB_ERR_STAT;
2481 } else if (!S_ISDIR(stbuf.st_mode)) {
2482 ERROR("\"%"TS"\" is not a directory", target);
2483 return WIMLIB_ERR_NOTDIR;
2486 tmemcpy(buf, target, output_path_len);
2487 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2488 for (image = 1; image <= wim->hdr.image_count; image++) {
2489 image_name = wimlib_get_image_name(wim, image);
2490 if (image_name_ok_as_dir(image_name)) {
2491 tstrcpy(buf + output_path_len + 1, image_name);
2493 /* Image name is empty or contains forbidden characters.
2494 * Use image number instead. */
2495 tsprintf(buf + output_path_len + 1, T("%d"), image);
2497 ret = extract_single_image(wim, image, buf, extract_flags,
2506 do_wimlib_extract_image(WIMStruct *wim,
2508 const tchar *target,
2510 WIMStruct **additional_swms,
2511 unsigned num_additional_swms,
2512 wimlib_progress_func_t progress_func)
2516 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2517 wimlib_assert(wim->hdr.part_number == 1);
2518 wimlib_assert(num_additional_swms == 0);
2520 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2524 if (num_additional_swms)
2525 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2528 if (image == WIMLIB_ALL_IMAGES) {
2529 ret = extract_all_images(wim, target, extract_flags,
2532 ret = extract_single_image(wim, image, target, extract_flags,
2536 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2537 WIMLIB_EXTRACT_FLAG_HARDLINK))
2539 for_lookup_table_entry(wim->lookup_table,
2540 lte_free_extracted_file,
2543 if (num_additional_swms)
2544 unmerge_lookup_table(wim);
2548 /* API function documented in wimlib.h */
2550 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2551 const tchar *target, int extract_flags,
2552 wimlib_progress_func_t progress_func)
2556 struct filedes *in_fd;
2560 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2562 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2563 return WIMLIB_ERR_INVALID_PARAM;
2565 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2567 /* Read the WIM header from the pipe and get a WIMStruct to represent
2568 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2569 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2570 * an empty lookup table, no XML data read, and no filename set. */
2571 ret = open_wim_as_WIMStruct(&pipe_fd,
2572 WIMLIB_OPEN_FLAG_FROM_PIPE |
2573 WIMLIB_OPEN_FLAG_SPLIT_OK,
2574 &pwm, progress_func);
2578 /* Sanity check to make sure this is a pipable WIM. */
2579 if (pwm->hdr.magic != PWM_MAGIC) {
2580 ERROR("The WIM being read from file descriptor %d "
2581 "is not pipable!", pipe_fd);
2582 ret = WIMLIB_ERR_NOT_PIPABLE;
2583 goto out_wimlib_free;
2586 /* Sanity check to make sure the first part of a pipable split WIM is
2587 * sent over the pipe first. */
2588 if (pwm->hdr.part_number != 1) {
2589 ERROR("The first part of the split WIM must be "
2590 "sent over the pipe first.");
2591 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2592 goto out_wimlib_free;
2595 in_fd = &pwm->in_fd;
2596 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2598 /* As mentioned, the WIMStruct we created from the pipe does not have
2599 * XML data yet. Fix this by reading the extra copy of the XML data
2600 * that directly follows the header in pipable WIMs. (Note: see
2601 * write_pipable_wim() for more details about the format of pipable
2604 struct wim_lookup_table_entry xml_lte;
2605 ret = read_pwm_stream_header(pwm, &xml_lte, false);
2607 goto out_wimlib_free;
2609 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2611 ERROR("Expected XML data, but found non-metadata "
2613 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2614 goto out_wimlib_free;
2617 copy_resource_entry(&pwm->hdr.xml_res_entry,
2618 &xml_lte.resource_entry);
2620 ret = read_wim_xml_data(pwm);
2622 goto out_wimlib_free;
2623 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2624 ERROR("Image count in XML data is not the same as in WIM header.");
2625 ret = WIMLIB_ERR_XML;
2626 goto out_wimlib_free;
2630 /* Get image index (this may use the XML data that was just read to
2631 * resolve an image name). */
2632 image = wimlib_resolve_image(pwm, image_num_or_name);
2633 if (image == WIMLIB_NO_IMAGE) {
2634 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2636 ret = WIMLIB_ERR_INVALID_IMAGE;
2637 goto out_wimlib_free;
2638 } else if (image == WIMLIB_ALL_IMAGES) {
2639 ERROR("Applying all images from a pipe is not supported.");
2640 ret = WIMLIB_ERR_INVALID_IMAGE;
2641 goto out_wimlib_free;
2644 /* Load the needed metadata resource. */
2645 for (i = 1; i <= pwm->hdr.image_count; i++) {
2646 struct wim_lookup_table_entry *metadata_lte;
2647 struct wim_image_metadata *imd;
2649 metadata_lte = new_lookup_table_entry();
2650 if (!metadata_lte) {
2651 ret = WIMLIB_ERR_NOMEM;
2652 goto out_wimlib_free;
2655 ret = read_pwm_stream_header(pwm, metadata_lte, false);
2656 imd = pwm->image_metadata[i - 1];
2657 imd->metadata_lte = metadata_lte;
2659 goto out_wimlib_free;
2661 if (!(metadata_lte->resource_entry.flags &
2662 WIM_RESHDR_FLAG_METADATA))
2664 ERROR("Expected metadata resource, but found "
2665 "non-metadata stream.");
2666 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2667 goto out_wimlib_free;
2671 /* Metadata resource is for the images being extracted.
2672 * Parse it and save the metadata in memory. */
2673 ret = read_metadata_resource(pwm, imd);
2675 goto out_wimlib_free;
2678 /* Metadata resource is not for the image being
2679 * extracted. Skip over it. */
2680 ret = skip_pwm_stream(metadata_lte);
2682 goto out_wimlib_free;
2685 /* Extract the image. */
2686 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2687 ret = do_wimlib_extract_image(pwm, image, target,
2688 extract_flags, NULL, 0, progress_func);
2689 /* Clean up and return. */
2695 /* API function documented in wimlib.h */
2697 wimlib_extract_image(WIMStruct *wim,
2699 const tchar *target,
2701 WIMStruct **additional_swms,
2702 unsigned num_additional_swms,
2703 wimlib_progress_func_t progress_func)
2705 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2706 return do_wimlib_extract_image(wim, image, target, extract_flags,
2707 additional_swms, num_additional_swms,