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,
745 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
746 ERROR_WITH_ERRNO("Failed to set security "
747 "descriptor on \"%"TS"\"", path);
750 if (errno != EACCES) {
751 WARNING_WITH_ERRNO("Failed to set "
752 "security descriptor "
753 "on \"%"TS"\"", path);
761 /* Set timestamps on an extracted file. Failure is warning-only unless
762 * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */
764 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
765 struct wim_dentry *dentry)
767 struct wim_inode *inode = dentry->d_inode;
770 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
773 if (ctx->ops->set_timestamps) {
774 ret = ctx->ops->set_timestamps(path,
775 inode->i_creation_time,
776 inode->i_last_write_time,
777 inode->i_last_access_time,
780 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
781 ERROR_WITH_ERRNO("Failed to set timestamps "
782 "on \"%"TS"\"", path);
785 WARNING_WITH_ERRNO("Failed to set timestamps "
786 "on \"%"TS"\"", path);
793 /* Check whether the extraction of a dentry should be skipped completely. */
795 dentry_is_supported(struct wim_dentry *dentry,
796 const struct wim_features *supported_features)
798 struct wim_inode *inode = dentry->d_inode;
800 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
801 if (supported_features->reparse_points)
803 if (supported_features->symlink_reparse_points &&
804 inode_is_symlink(inode))
811 /* Given a WIM dentry to extract, build the path to which to extract it, in the
812 * format understood by the callbacks in the apply_operations being used.
814 * Write the resulting path into @path, which must have room for at least
815 * ctx->ops->max_path characters including the null-terminator.
817 * Return %true if successful; %false if this WIM dentry doesn't actually need
818 * to be extracted or if the calculated path exceeds ctx->ops->max_path
821 * This function clobbers the tmp_list member of @dentry and its ancestors up
822 * until the extraction root. */
824 build_extraction_path(tchar path[], struct wim_dentry *dentry,
825 struct apply_ctx *ctx)
828 LIST_HEAD(ancestor_list);
830 const tchar *target_prefix;
831 size_t target_prefix_nchars;
832 struct wim_dentry *d;
834 if (dentry->extraction_skipped)
837 path_nchars = ctx->ops->path_prefix_nchars;
839 if (ctx->ops->requires_realtarget_in_paths) {
840 target_prefix = ctx->realtarget;
841 target_prefix_nchars = ctx->realtarget_nchars;
842 } else if (ctx->ops->requires_target_in_paths) {
843 target_prefix = ctx->target;
844 target_prefix_nchars = ctx->target_nchars;
846 target_prefix = NULL;
847 target_prefix_nchars = 0;
849 path_nchars += target_prefix_nchars;
851 for (d = dentry; d != ctx->extract_root; d = d->parent) {
852 path_nchars += d->extraction_name_nchars + 1;
853 list_add(&d->tmp_list, &ancestor_list);
856 path_nchars++; /* null terminator */
858 if (path_nchars > ctx->ops->path_max) {
859 WARNING("\"%"TS"\": Path too long to extract",
860 dentry_full_path(dentry));
864 p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
865 p = tmempcpy(p, target_prefix, target_prefix_nchars);
866 list_for_each_entry(d, &ancestor_list, tmp_list) {
867 *p++ = ctx->ops->path_separator;
868 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
871 wimlib_assert(p - path == path_nchars);
876 get_num_path_components(const tchar *path, tchar path_separator)
878 unsigned num_components = 0;
881 while (*path == path_separator)
885 while (*path && *path != path_separator)
888 return num_components;
892 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
893 struct apply_ctx *ctx, struct wim_dentry *dentry)
895 size_t num_raw_path_components;
896 const struct wim_dentry *d;
897 size_t num_target_path_components;
902 num_raw_path_components = 0;
903 for (d = dentry; d != ctx->extract_root; d = d->parent)
904 num_raw_path_components++;
906 if (ctx->ops->requires_realtarget_in_paths)
907 num_target_path_components = get_num_path_components(ctx->realtarget,
908 ctx->ops->path_separator);
909 else if (ctx->ops->requires_target_in_paths)
910 num_target_path_components = get_num_path_components(ctx->target,
911 ctx->ops->path_separator);
913 num_target_path_components = 0;
915 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
916 wimlib_assert(num_target_path_components > 0);
917 num_raw_path_components++;
918 num_target_path_components--;
922 while (*p_old == ctx->ops->path_separator)
924 while (--num_target_path_components) {
925 while (*p_old != ctx->ops->path_separator)
927 while (*p_old == ctx->ops->path_separator)
931 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
933 p = &symlink_target[0];
934 while (num_raw_path_components--) {
937 *p++ = ctx->ops->path_separator;
940 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
941 newpath, symlink_target);
942 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
944 ERROR_WITH_ERRNO("Failed to create symlink "
945 "\"%"TS"\" => \"%"TS"\"",
946 newpath, symlink_target);
951 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
952 * data streams, reparse data (including symbolic links), timestamps, and
953 * security descriptors. Basically, everything that doesn't require reading
954 * non-metadata resources from the WIM file and isn't delayed until the final
957 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
958 struct apply_ctx *ctx)
960 struct wim_inode *inode = dentry->d_inode;
962 const tchar *oldpath;
964 if (unlikely(is_linked_extraction(ctx))) {
965 struct wim_lookup_table_entry *unnamed_lte;
967 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
968 if (unnamed_lte && unnamed_lte->extracted_file) {
969 oldpath = unnamed_lte->extracted_file;
970 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
977 /* Create hard link if this dentry corresponds to an already-extracted
979 if (inode->i_extracted_file) {
980 oldpath = inode->i_extracted_file;
984 /* Skip symlinks unless they can be extracted as reparse points rather
985 * than created directly. */
986 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
989 /* Create this file or directory unless it's the extraction root, which
990 * was already created if necessary. */
991 if (dentry != ctx->extract_root) {
992 ret = extract_inode(path, ctx, inode);
997 /* Create empty named data streams. */
998 if (can_extract_named_data_streams(ctx)) {
999 for (u16 i = 0; i < inode->i_num_ads; i++) {
1000 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1002 if (!ads_entry_is_named_stream(entry))
1006 ret = ctx->ops->extract_named_stream(path,
1008 entry->stream_name_nbytes / 2,
1011 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1012 "empty named data stream",
1019 /* Set file attributes (if supported). */
1020 ret = extract_file_attributes(path, ctx, dentry);
1024 /* Set or remove file short name (if supported). */
1025 ret = extract_short_name(path, ctx, dentry);
1029 /* If inode has multiple links and hard links are supported in this
1030 * extraction mode and volume, save the path to the extracted file in
1031 * case it's needed to create a hard link. */
1032 if (unlikely(is_linked_extraction(ctx))) {
1033 struct wim_lookup_table_entry *unnamed_lte;
1035 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1037 unnamed_lte->extracted_file = TSTRDUP(path);
1038 if (!unnamed_lte->extracted_file)
1039 return WIMLIB_ERR_NOMEM;
1041 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1042 inode->i_extracted_file = TSTRDUP(path);
1043 if (!inode->i_extracted_file)
1044 return WIMLIB_ERR_NOMEM;
1049 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1052 dentry->was_hardlinked = 1;
1056 ret = extract_hardlink(oldpath, path, ctx);
1059 dentry->was_hardlinked = 1;
1064 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1066 struct apply_ctx *ctx = _ctx;
1067 tchar path[ctx->ops->path_max];
1068 struct wim_dentry *orig_dentry;
1069 struct wim_dentry *other_dentry;
1072 /* Here we may re-order the extraction of multiple names (hard links)
1073 * for the same file in the same directory in order to ensure the short
1074 * (DOS) name is set correctly. A short name is always associated with
1075 * exactly one long name, and at least on NTFS, only one long name for a
1076 * file can have a short name associated with it. (More specifically,
1077 * there can be unlimited names in the POSIX namespace, but only one
1078 * name can be in the Win32+DOS namespace, or one name in the Win32
1079 * namespace with a corresponding name in the DOS namespace.) To ensure
1080 * the short name of a file is associated with the correct long name in
1081 * a directory, we extract the long name with a corresponding short name
1082 * before any additional names. This can affect NTFS-3g extraction
1083 * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying
1084 * the long name to associate with a short name) and may affect Win32
1085 * extraction as well (which uses SetFileShortName()). */
1087 if (dentry->skeleton_extracted)
1090 if (ctx->supported_features.short_names
1091 && !dentry_has_short_name(dentry)
1092 && !dentry->d_inode->i_dos_name_extracted)
1094 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1095 if (dentry_has_short_name(other_dentry)
1096 && !other_dentry->skeleton_extracted
1097 && other_dentry->parent == dentry->parent)
1099 DEBUG("Creating %"TS" before %"TS" "
1100 "to guarantee correct DOS name extraction",
1101 dentry_full_path(other_dentry),
1102 dentry_full_path(dentry));
1103 orig_dentry = dentry;
1104 dentry = other_dentry;
1110 if (!build_extraction_path(path, dentry, ctx))
1112 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1116 dentry->skeleton_extracted = 1;
1119 dentry = orig_dentry;
1123 dentry->d_inode->i_dos_name_extracted = 1;
1127 /* Create a file or directory, then immediately extract all streams. This
1128 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1129 * may not be read sequentially by this function. */
1131 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1133 struct apply_ctx *ctx = _ctx;
1134 tchar path[ctx->ops->path_max];
1137 ret = dentry_extract_skeleton(dentry, ctx);
1141 if (!build_extraction_path(path, dentry, ctx))
1144 return extract_streams(path, ctx, dentry, NULL, NULL);
1147 /* Extract all instances of the stream @lte that are being extracted in this
1148 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1149 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1150 * extracted multiple times, it is extracted to a temporary file first.
1152 * This is intended for use with sequential extraction of a WIM image
1153 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1155 extract_stream_instances(struct wim_lookup_table_entry *lte,
1156 struct apply_ctx *ctx, bool can_seek)
1158 struct wim_dentry **lte_dentries;
1159 struct wim_lookup_table_entry *lte_tmp = NULL;
1160 struct wim_lookup_table_entry *lte_override;
1161 tchar *stream_tmp_filename = NULL;
1162 tchar path[ctx->ops->path_max];
1166 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1167 lte_dentries = lte->inline_lte_dentries;
1169 lte_dentries = lte->lte_dentries;
1171 if (likely(can_seek || lte->out_refcnt < 2)) {
1174 /* Need to extract stream to temporary file. */
1178 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1179 if (!stream_tmp_filename) {
1180 ERROR_WITH_ERRNO("Failed to create temporary filename");
1181 ret = WIMLIB_ERR_OPEN;
1185 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1187 ret = WIMLIB_ERR_NOMEM;
1188 goto out_free_stream_tmp_filename;
1190 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1191 lte_tmp->file_on_disk = stream_tmp_filename;
1192 lte_override = lte_tmp;
1194 raw_fd = topen(stream_tmp_filename,
1195 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1197 ERROR_WITH_ERRNO("Failed to open temporary file");
1198 ret = WIMLIB_ERR_OPEN;
1199 goto out_free_lte_tmp;
1201 filedes_init(&fd, raw_fd);
1202 ret = extract_wim_resource_to_fd(lte, &fd,
1203 wim_resource_size(lte));
1204 if (filedes_close(&fd) && !ret)
1205 ret = WIMLIB_ERR_WRITE;
1207 goto out_unlink_stream_tmp_file;
1210 /* Extract all instances of the stream, reading either from the stream
1211 * in the WIM file or from the temporary file containing the stream.
1212 * dentry->tmp_flag is used to ensure that each dentry is processed only
1213 * once regardless of how many times this stream appears in the streams
1214 * of the corresponding inode. */
1215 for (i = 0; i < lte->out_refcnt; i++) {
1216 struct wim_dentry *dentry = lte_dentries[i];
1218 if (dentry->tmp_flag)
1220 if (!build_extraction_path(path, dentry, ctx))
1222 ret = extract_streams(path, ctx, dentry,
1225 goto out_clear_tmp_flags;
1226 dentry->tmp_flag = 1;
1229 out_clear_tmp_flags:
1230 for (i = 0; i < lte->out_refcnt; i++)
1231 lte_dentries[i]->tmp_flag = 0;
1232 out_unlink_stream_tmp_file:
1233 if (stream_tmp_filename)
1234 tunlink(stream_tmp_filename);
1237 out_free_stream_tmp_filename:
1238 FREE(stream_tmp_filename);
1243 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1244 * structure and empty files were already created. This relies on the
1245 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1246 * stream that was constructed earlier. Streams are extracted exactly in the
1247 * order of the stream list; however, unless the WIM's file descriptor is
1248 * detected to be non-seekable, streams may be read from the WIM file more than
1249 * one time if multiple copies need to be extracted. */
1251 extract_stream_list(struct apply_ctx *ctx)
1253 struct wim_lookup_table_entry *lte;
1257 can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
1258 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1259 ret = extract_stream_instances(lte, ctx, can_seek);
1266 #define PWM_ALLOW_WIM_HDR 0x00001
1267 #define PWM_SILENT_EOF 0x00002
1269 /* Read the header from a stream in a pipable WIM. */
1271 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1274 struct pwm_stream_hdr stream_hdr;
1277 ret = full_read(&pwm->in_fd, &stream_hdr, sizeof(stream_hdr));
1281 if ((flags & PWM_ALLOW_WIM_HDR) && stream_hdr.magic == PWM_MAGIC) {
1282 u8 buf[WIM_HEADER_DISK_SIZE - sizeof(stream_hdr)];
1283 ret = full_read(&pwm->in_fd, buf, sizeof(buf));
1286 lte->resource_location = RESOURCE_NONEXISTENT;
1290 if (stream_hdr.magic != PWM_STREAM_MAGIC) {
1291 ERROR("Data read on pipe is invalid (expected stream header).");
1292 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1295 lte->resource_entry.original_size = le64_to_cpu(stream_hdr.uncompressed_size);
1296 copy_hash(lte->hash, stream_hdr.hash);
1297 lte->resource_entry.flags = le32_to_cpu(stream_hdr.flags);
1298 lte->resource_entry.offset = pwm->in_fd.offset;
1299 lte->resource_location = RESOURCE_IN_WIM;
1301 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1302 lte->compression_type = pwm->compression_type;
1303 lte->resource_entry.size = 0;
1305 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1306 lte->resource_entry.size = lte->resource_entry.original_size;
1308 lte->is_pipable = 1;
1312 if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1313 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1317 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1319 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1321 return read_partial_wim_resource(lte, wim_resource_size(lte),
1323 WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY,
1328 extract_streams_from_pipe(struct apply_ctx *ctx)
1330 struct wim_lookup_table_entry *found_lte;
1331 struct wim_lookup_table_entry *needed_lte;
1332 struct wim_lookup_table *lookup_table;
1336 ret = WIMLIB_ERR_NOMEM;
1337 found_lte = new_lookup_table_entry();
1341 lookup_table = ctx->wim->lookup_table;
1342 pwm_flags = PWM_ALLOW_WIM_HDR;
1343 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1344 pwm_flags |= PWM_SILENT_EOF;
1345 while (ctx->num_streams_remaining) {
1346 ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags);
1348 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1349 (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1353 goto out_free_found_lte;
1356 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1357 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1358 && (needed_lte = __lookup_resource(lookup_table, found_lte->hash))
1359 && (needed_lte->out_refcnt))
1361 copy_resource_entry(&needed_lte->resource_entry,
1362 &found_lte->resource_entry);
1363 needed_lte->resource_location = found_lte->resource_location;
1364 needed_lte->wim = found_lte->wim;
1365 needed_lte->compression_type = found_lte->compression_type;
1366 needed_lte->is_pipable = found_lte->is_pipable;
1368 ret = extract_stream_instances(needed_lte, ctx, false);
1370 goto out_free_found_lte;
1371 ctx->num_streams_remaining--;
1372 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1373 ret = skip_pwm_stream(found_lte);
1375 goto out_free_found_lte;
1380 free_lookup_table_entry(found_lte);
1389 /* Finish extracting a file, directory, or symbolic link by setting file
1390 * security and timestamps. */
1392 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1394 struct apply_ctx *ctx = _ctx;
1396 tchar path[ctx->ops->path_max];
1398 if (!build_extraction_path(path, dentry, ctx))
1401 ret = extract_security(path, ctx, dentry);
1405 return extract_timestamps(path, ctx, dentry);
1409 * Extract a WIM dentry to standard output.
1411 * This obviously doesn't make sense in all cases. We return an error if the
1412 * dentry does not correspond to a regular file. Otherwise we extract the
1413 * unnamed data stream only.
1416 extract_dentry_to_stdout(struct wim_dentry *dentry)
1419 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1420 FILE_ATTRIBUTE_DIRECTORY))
1422 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1423 "extracted to standard output", dentry_full_path(dentry));
1424 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1426 struct wim_lookup_table_entry *lte;
1428 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1430 struct filedes _stdout;
1431 filedes_init(&_stdout, STDOUT_FILENO);
1432 ret = extract_wim_resource_to_fd(lte, &_stdout,
1433 wim_resource_size(lte));
1440 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1442 static const utf16lechar replacement_char = cpu_to_le16('?');
1446 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1452 for (i = 0; i < num_chars; i++) {
1455 case cpu_to_le16('\\'):
1456 case cpu_to_le16(':'):
1457 case cpu_to_le16('*'):
1458 case cpu_to_le16('?'):
1459 case cpu_to_le16('"'):
1460 case cpu_to_le16('<'):
1461 case cpu_to_le16('>'):
1462 case cpu_to_le16('|'):
1464 case cpu_to_le16('/'):
1465 case cpu_to_le16('\0'):
1467 name[i] = replacement_char;
1474 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1475 name[num_chars - 1] == cpu_to_le16('.'))
1478 name[num_chars - 1] = replacement_char;
1487 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1489 const utf16lechar *file_name = dentry->file_name;
1490 return file_name != NULL &&
1491 file_name[0] == cpu_to_le16('.') &&
1492 (file_name[1] == cpu_to_le16('\0') ||
1493 (file_name[1] == cpu_to_le16('.') &&
1494 file_name[2] == cpu_to_le16('\0')));
1498 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1500 dentry->extraction_skipped = 1;
1505 * dentry_calculate_extraction_path-
1507 * Calculate the actual filename component at which a WIM dentry will be
1508 * extracted, handling invalid filenames "properly".
1510 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1511 * UNIX, converted into the platform's multibyte encoding). However, if the
1512 * file name contains characters that are not valid on the current platform or
1513 * has some other format that is not valid, leave dentry->extraction_name as
1514 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1515 * not be extracted, unless the appropriate flag
1516 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1517 * which case a substitute filename will be created and set instead.
1519 * Conflicts with case-insensitive names on Windows are handled similarly; see
1523 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1525 struct apply_ctx *ctx = _args;
1528 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1531 if (!dentry_is_supported(dentry, &ctx->supported_features))
1534 if (dentry_is_dot_or_dotdot(dentry)) {
1535 /* WIM files shouldn't contain . or .. entries. But if they are
1536 * there, don't attempt to extract them. */
1537 WARNING("Skipping extraction of unexpected . or .. file "
1538 "\"%"TS"\"", dentry_full_path(dentry));
1543 if (!ctx->ops->supports_case_sensitive_filenames)
1545 struct wim_dentry *other;
1546 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1547 case_insensitive_conflict_list)
1549 if (ctx->extract_flags &
1550 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1551 WARNING("\"%"TS"\" has the same "
1552 "case-insensitive name as "
1553 "\"%"TS"\"; extracting "
1554 "dummy name instead",
1555 dentry_full_path(dentry),
1556 dentry_full_path(other));
1559 WARNING("Not extracting \"%"TS"\": "
1560 "has same case-insensitive "
1561 "name as \"%"TS"\"",
1562 dentry_full_path(dentry),
1563 dentry_full_path(other));
1568 #else /* __WIN32__ */
1569 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1570 #endif /* !__WIN32__ */
1572 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1574 dentry->extraction_name = dentry->file_name;
1575 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1578 return utf16le_to_tstr(dentry->file_name,
1579 dentry->file_name_nbytes,
1580 &dentry->extraction_name,
1581 &dentry->extraction_name_nchars);
1584 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1586 WARNING("\"%"TS"\" has an invalid filename "
1587 "that is not supported on this platform; "
1588 "extracting dummy name instead",
1589 dentry_full_path(dentry));
1592 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1593 "that is not supported on this platform",
1594 dentry_full_path(dentry));
1601 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1603 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1604 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1607 size_t tchar_nchars;
1609 tchar_name = utf16_name_copy;
1610 tchar_nchars = dentry->file_name_nbytes / 2;
1612 ret = utf16le_to_tstr(utf16_name_copy,
1613 dentry->file_name_nbytes,
1614 &tchar_name, &tchar_nchars);
1618 size_t fixed_name_num_chars = tchar_nchars;
1619 tchar fixed_name[tchar_nchars + 50];
1621 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1622 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1623 T(" (invalid filename #%lu)"),
1624 ++ctx->invalid_sequence);
1628 dentry->extraction_name = memdup(fixed_name,
1629 2 * fixed_name_num_chars + 2);
1630 if (!dentry->extraction_name)
1631 return WIMLIB_ERR_NOMEM;
1632 dentry->extraction_name_nchars = fixed_name_num_chars;
1637 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1641 /* Clean up dentry and inode structure after extraction. */
1643 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1645 struct wim_inode *inode = dentry->d_inode;
1647 dentry->extraction_skipped = 0;
1648 dentry->was_hardlinked = 0;
1649 dentry->skeleton_extracted = 0;
1650 inode->i_visited = 0;
1651 FREE(inode->i_extracted_file);
1652 inode->i_extracted_file = NULL;
1653 inode->i_dos_name_extracted = 0;
1654 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1655 FREE(dentry->extraction_name);
1656 dentry->extraction_name = NULL;
1660 /* Tally features necessary to extract a dentry and the corresponding inode. */
1662 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1664 struct wim_features *features = _features;
1665 struct wim_inode *inode = dentry->d_inode;
1667 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1668 features->archive_files++;
1669 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1670 features->hidden_files++;
1671 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1672 features->system_files++;
1673 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1674 features->compressed_files++;
1675 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED)
1676 features->encrypted_files++;
1677 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1678 features->not_context_indexed_files++;
1679 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1680 features->sparse_files++;
1681 if (inode_has_named_stream(inode))
1682 features->named_data_streams++;
1683 if (inode->i_visited)
1684 features->hard_links++;
1685 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1686 features->reparse_points++;
1687 if (inode_is_symlink(inode))
1688 features->symlink_reparse_points++;
1690 features->other_reparse_points++;
1692 if (inode->i_security_id != -1)
1693 features->security_descriptors++;
1694 if (dentry->short_name_nbytes)
1695 features->short_names++;
1696 if (inode_has_unix_data(inode))
1697 features->unix_data++;
1698 inode->i_visited = 1;
1703 dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
1705 dentry->d_inode->i_visited = 0;
1709 /* Tally the features necessary to extract a dentry tree. */
1711 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1713 memset(features, 0, sizeof(struct wim_features));
1714 for_dentry_in_tree(root, dentry_tally_features, features);
1715 for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
1719 do_feature_check(const struct wim_features *required_features,
1720 const struct wim_features *supported_features,
1722 const struct apply_operations *ops,
1723 const tchar *wim_source_path)
1726 const tchar *mode = "this extraction mode";
1728 if (wim_source_path[0] == '\0')
1729 loc = "the WIM image";
1731 loc = wim_source_path;
1733 /* We're an archive program, so theoretically we can do what we want
1734 * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't
1735 * bother the user about it. */
1737 if (required_features->archive_files && !supported_features->archive_files)
1740 "%lu files in %"TS" are marked as archived, but this attribute\n"
1741 " is not supported in %"TS".",
1742 required_features->archive_files, loc, mode);
1746 if (required_features->hidden_files && !supported_features->hidden_files)
1749 "%lu files in %"TS" are marked as hidden, but this\n"
1750 " attribute is not supported in %"TS".",
1751 required_features->hidden_files, loc, mode);
1754 if (required_features->system_files && !supported_features->system_files)
1757 "%lu files in %"TS" are marked as system files,\n"
1758 " but this attribute is not supported in %"TS".",
1759 required_features->system_files, loc, mode);
1762 if (required_features->compressed_files && !supported_features->compressed_files)
1765 "%lu files in %"TS" are marked as being transparently\n"
1766 " compressed, but transparent compression is not supported in\n"
1767 " %"TS". These files will be extracted as uncompressed.",
1768 required_features->compressed_files, loc, mode);
1771 if (required_features->encrypted_files && !supported_features->encrypted_files)
1774 "%lu files in %"TS" are marked as being encrypted,\n"
1775 " but encryption is not supported in %"TS". These files\n"
1776 " will be extracted as raw encrypted data instead.",
1777 required_features->encrypted_files, loc, mode);
1780 if (required_features->not_context_indexed_files &&
1781 !supported_features->not_context_indexed_files)
1784 "%lu files in %"TS" are marked as not content indexed,\n"
1785 " but this attribute is not supported in %"TS".",
1786 required_features->not_context_indexed_files, loc, mode);
1789 if (required_features->sparse_files && !supported_features->sparse_files)
1792 "%lu files in %"TS" are marked as sparse, but creating\n"
1793 " sparse files is not supported in %"TS". These files\n"
1794 " will be extracted as non-sparse.",
1795 required_features->sparse_files, loc, mode);
1798 if (required_features->named_data_streams &&
1799 !supported_features->named_data_streams)
1802 "%lu files in %"TS" contain one or more alternate (named)\n"
1803 " data streams, which are not supported in %"TS".\n"
1804 " Alternate data streams will NOT be extracted.",
1805 required_features->named_data_streams, loc, mode);
1808 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1809 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1810 required_features->named_data_streams &&
1811 supported_features->named_data_streams)
1814 "%lu files in %"TS" contain one or more alternate (named)\n"
1815 " data streams, which are not supported in linked extraction mode.\n"
1816 " Alternate data streams will NOT be extracted.",
1817 required_features->named_data_streams, loc);
1820 if (required_features->hard_links && !supported_features->hard_links)
1823 "%lu files in %"TS" are hard links, but hard links are\n"
1824 " not supported in %"TS". Hard links will be extracted as\n"
1825 " duplicate copies of the linked files.",
1826 required_features->hard_links, loc, mode);
1829 if (required_features->reparse_points && !supported_features->reparse_points)
1831 if (supported_features->symlink_reparse_points) {
1832 if (required_features->other_reparse_points) {
1834 "%lu files in %"TS" are reparse points that are neither\n"
1835 " symbolic links nor junction points and are not supported in\n"
1836 " %"TS". These reparse points will not be extracted.",
1837 required_features->other_reparse_points, loc,
1842 "%lu files in %"TS" are reparse points, which are\n"
1843 " not supported in %"TS" and will not be extracted.",
1844 required_features->reparse_points, loc, mode);
1848 if (required_features->security_descriptors &&
1849 !supported_features->security_descriptors)
1852 "%lu files in %"TS" have Windows NT security descriptors,\n"
1853 " but extracting security descriptors is not supported in\n"
1854 " %"TS". No security descriptors will be extracted.",
1855 required_features->security_descriptors, loc, mode);
1858 if (required_features->short_names && !supported_features->short_names)
1861 "%lu files in %"TS" have short (DOS) names, but\n"
1862 " extracting short names is not supported in %"TS".\n"
1863 " Short names will not be extracted.\n",
1864 required_features->short_names, loc, mode);
1867 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1868 required_features->unix_data && !supported_features->unix_data)
1870 ERROR("Extracting UNIX data is not supported in %"TS, mode);
1871 return WIMLIB_ERR_UNSUPPORTED;
1873 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
1874 required_features->short_names && !supported_features->short_names)
1876 ERROR("Extracting short names is not supported in %"TS"", mode);
1877 return WIMLIB_ERR_UNSUPPORTED;
1879 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1880 !ops->set_timestamps)
1882 ERROR("Extracting timestamps is not supported in %"TS"", mode);
1883 return WIMLIB_ERR_UNSUPPORTED;
1885 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1886 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1887 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1888 required_features->security_descriptors &&
1889 !supported_features->security_descriptors)
1891 ERROR("Extracting security descriptors is not supported in %"TS, mode);
1892 return WIMLIB_ERR_UNSUPPORTED;
1895 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
1896 !supported_features->hard_links)
1898 ERROR("Hard link extraction mode requested, but "
1899 "%"TS" does not support hard links!", mode);
1900 return WIMLIB_ERR_UNSUPPORTED;
1903 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
1904 !supported_features->symlink_reparse_points)
1906 ERROR("Symbolic link extraction mode requested, but "
1907 "%"TS" does not support symbolic "
1909 return WIMLIB_ERR_UNSUPPORTED;
1915 * extract_tree - Extract a file or directory tree from the currently selected
1918 * @wim: WIMStruct for the WIM file, with the desired image selected
1919 * (as wim->current_image).
1922 * "Canonical" (i.e. no leading or trailing slashes, path
1923 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
1924 * extract. An empty string means the full image.
1927 * Filesystem path to extract the file or directory tree to.
1928 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
1931 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
1932 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
1933 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
1937 * If non-NULL, progress function for the extraction. The messages
1938 * that may be sent in this function are:
1940 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
1941 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
1942 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
1943 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
1944 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
1945 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
1946 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
1947 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
1948 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
1950 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
1953 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
1954 int extract_flags, wimlib_progress_func_t progress_func)
1956 struct wim_dentry *root;
1957 struct wim_features required_features;
1958 struct apply_ctx ctx;
1960 struct wim_lookup_table_entry *lte;
1962 /* Start initializing the apply_ctx. */
1963 memset(&ctx, 0, sizeof(struct apply_ctx));
1965 ctx.extract_flags = extract_flags;
1966 ctx.target = target;
1967 ctx.target_nchars = tstrlen(target);
1968 ctx.progress_func = progress_func;
1969 if (progress_func) {
1970 ctx.progress.extract.wimfile_name = wim->filename;
1971 ctx.progress.extract.image = wim->current_image;
1972 ctx.progress.extract.extract_flags = (extract_flags &
1973 WIMLIB_EXTRACT_MASK_PUBLIC);
1974 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
1975 wim->current_image);
1976 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
1977 ctx.progress.extract.target = target;
1979 INIT_LIST_HEAD(&ctx.stream_list);
1981 /* Translate the path to extract into the corresponding
1982 * `struct wim_dentry', which will be the root of the
1983 * "dentry tree" to extract. */
1984 root = get_dentry(wim, wim_source_path);
1986 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
1987 wim_source_path, wim->current_image);
1988 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1992 ctx.extract_root = root;
1994 /* Select the appropriate apply_operations based on the
1995 * platform and extract_flags. */
1997 ctx.ops = &win32_apply_ops;
1999 ctx.ops = &unix_apply_ops;
2003 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
2004 ctx.ops = &ntfs_3g_apply_ops;
2007 /* Call the start_extract() callback. This gives the apply_operations
2008 * implementation a chance to do any setup needed to access the volume.
2009 * Furthermore, it's expected to set the supported features of this
2010 * extraction mode (ctx.supported_features), which are determined at
2011 * runtime as they may vary depending on the actual volume. These
2012 * features are then compared with the actual features extracting this
2013 * dentry tree requires. Some mismatches will merely produce warnings
2014 * and the unsupported data will be ignored; others will produce errors.
2016 ret = ctx.ops->start_extract(target, &ctx);
2020 dentry_tree_get_features(root, &required_features);
2021 ret = do_feature_check(&required_features, &ctx.supported_features,
2022 extract_flags, ctx.ops, wim_source_path);
2024 goto out_finish_or_abort_extract;
2026 /* Figure out whether the root dentry is being extracted to the root of
2027 * a volume and therefore needs to be treated "specially", for example
2028 * not being explicitly created and not having attributes set. */
2029 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2030 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2032 /* Calculate the actual filename component of each extracted dentry. In
2033 * the process, set the dentry->extraction_skipped flag on dentries that
2034 * are being skipped for some reason (e.g. invalid filename). */
2035 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
2037 goto out_dentry_reset_needs_extraction;
2039 /* Build the list of the streams that need to be extracted and
2040 * initialize ctx.progress.extract with stream information. */
2041 ret = for_dentry_in_tree(ctx.extract_root,
2042 dentry_resolve_and_zero_lte_refcnt, &ctx);
2044 goto out_dentry_reset_needs_extraction;
2046 ret = for_dentry_in_tree(ctx.extract_root,
2047 dentry_add_streams_to_extract, &ctx);
2049 goto out_teardown_stream_list;
2051 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2052 /* When extracting from a pipe, the number of bytes of data to
2053 * extract can't be determined in the normal way (examining the
2054 * lookup table), since at this point all we have is a set of
2055 * SHA1 message digests of streams that need to be extracted.
2056 * However, we can get a reasonably accurate estimate by taking
2057 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2058 * data. This does assume that a full image is being extracted,
2059 * but currently there is no API for doing otherwise. (Also,
2060 * subtract <HARDLINKBYTES> from this if hard links are
2061 * supported by the extraction mode.) */
2062 ctx.progress.extract.total_bytes =
2063 wim_info_get_image_total_bytes(wim->wim_info,
2064 wim->current_image);
2065 if (ctx.supported_features.hard_links) {
2066 ctx.progress.extract.total_bytes -=
2067 wim_info_get_image_hard_link_bytes(wim->wim_info,
2068 wim->current_image);
2072 /* Handle the special case of extracting a file to standard
2073 * output. In that case, "root" should be a single file, not a
2074 * directory tree. (If not, extract_dentry_to_stdout() will
2075 * return an error.) */
2076 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
2077 ret = extract_dentry_to_stdout(root);
2078 goto out_teardown_stream_list;
2081 /* If a sequential extraction was specified, sort the streams to be
2082 * extracted by their position in the WIM file so that the WIM file can
2083 * be read sequentially. */
2084 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2085 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2086 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2088 ret = sort_stream_list_by_sequential_order(
2090 offsetof(struct wim_lookup_table_entry,
2093 goto out_teardown_stream_list;
2096 if (ctx.ops->realpath_works_on_nonexisting_files &&
2097 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2098 ctx.ops->requires_realtarget_in_paths))
2100 ctx.realtarget = realpath(target, NULL);
2101 if (!ctx.realtarget) {
2102 ret = WIMLIB_ERR_NOMEM;
2103 goto out_teardown_stream_list;
2105 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2108 if (progress_func) {
2109 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2110 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2114 if (!ctx.root_dentry_is_special)
2116 tchar path[ctx.ops->path_max];
2117 if (build_extraction_path(path, root, &ctx))
2119 ret = extract_inode(path, &ctx, root->d_inode);
2121 goto out_free_realtarget;
2125 /* If we need to fix up the targets of absolute symbolic links
2126 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2127 * be absolute, use realpath() (or its replacement on Windows) to get
2128 * the absolute path to the extraction target. Note that this requires
2129 * the target directory to exist, unless
2130 * realpath_works_on_nonexisting_files is set in the apply_operations.
2132 if (!ctx.realtarget &&
2133 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2134 required_features.symlink_reparse_points) ||
2135 ctx.ops->requires_realtarget_in_paths))
2137 ctx.realtarget = realpath(target, NULL);
2138 if (!ctx.realtarget) {
2139 ret = WIMLIB_ERR_NOMEM;
2140 goto out_free_realtarget;
2142 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2145 /* Finally, the important part: extract the tree of files. */
2146 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2147 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2148 /* Sequential extraction requested, so two passes are needed
2149 * (one for directory structure, one for streams.) */
2151 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2154 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2155 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2157 goto out_free_realtarget;
2160 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2162 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2163 ret = extract_streams_from_pipe(&ctx);
2165 ret = extract_stream_list(&ctx);
2167 goto out_free_realtarget;
2169 /* Sequential extraction was not requested, so we can make do
2170 * with one pass where we both create the files and extract
2173 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2175 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2177 goto out_free_realtarget;
2179 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2183 /* If the total number of bytes to extract was miscalculated, just jump
2184 * to the calculated number in order to avoid confusing the progress
2185 * function. This should only occur when extracting from a pipe. */
2186 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2188 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2189 "extracted %"PRIu64,
2190 ctx.progress.extract.total_bytes,
2191 ctx.progress.extract.completed_bytes);
2193 if (progress_func &&
2194 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2196 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2197 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2200 /* Apply security descriptors and timestamps. This is done at the end,
2201 * and in a depth-first manner, to prevent timestamps from getting
2202 * changed by subsequent extract operations and to minimize the chance
2203 * of the restored security descriptors getting in our way. */
2205 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2207 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2209 goto out_free_realtarget;
2211 if (progress_func) {
2212 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2213 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2218 out_free_realtarget:
2219 FREE(ctx.realtarget);
2220 out_teardown_stream_list:
2221 /* Free memory allocated as part of the mapping from each
2222 * wim_lookup_table_entry to the dentries that reference it. */
2223 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2224 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2225 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2226 FREE(lte->lte_dentries);
2227 out_dentry_reset_needs_extraction:
2228 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2229 out_finish_or_abort_extract:
2231 if (ctx.ops->abort_extract)
2232 ctx.ops->abort_extract(&ctx);
2234 if (ctx.ops->finish_extract)
2235 ret = ctx.ops->finish_extract(&ctx);
2241 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2242 * extract flags make sense. */
2244 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2248 /* Empty destination path? */
2249 if (cmd->fs_dest_path[0] == T('\0'))
2250 return WIMLIB_ERR_INVALID_PARAM;
2252 extract_flags = cmd->extract_flags;
2254 /* Check for invalid flag combinations */
2255 if ((extract_flags &
2256 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2257 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2258 WIMLIB_EXTRACT_FLAG_HARDLINK))
2259 return WIMLIB_ERR_INVALID_PARAM;
2261 if ((extract_flags &
2262 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2263 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2264 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2265 return WIMLIB_ERR_INVALID_PARAM;
2267 if ((extract_flags &
2268 (WIMLIB_EXTRACT_FLAG_RPFIX |
2269 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2270 WIMLIB_EXTRACT_FLAG_NORPFIX))
2271 return WIMLIB_ERR_INVALID_PARAM;
2273 if ((extract_flags &
2274 (WIMLIB_EXTRACT_FLAG_RESUME |
2275 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2276 return WIMLIB_ERR_INVALID_PARAM;
2278 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2279 #ifndef WITH_NTFS_3G
2280 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2281 " we cannot apply a WIM image directly to a NTFS volume.");
2282 return WIMLIB_ERR_UNSUPPORTED;
2286 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2287 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2289 /* Do reparse point fixups by default if the WIM header says
2290 * they are enabled and we are extracting a full image. */
2291 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2292 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2295 /* TODO: Since UNIX data entries are stored in the file resources, in a
2296 * completely sequential extraction they may come up before the
2297 * corresponding file or symbolic link data. This needs to be handled
2299 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2300 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2301 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2302 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2304 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2305 WARNING("Setting UNIX file/owner group may "
2306 "be impossible on some\n"
2308 "when applying from a pipe.");
2310 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2311 WARNING("Disabling sequential extraction for "
2316 cmd->extract_flags = extract_flags;
2321 /* Internal function to execute extraction commands for a WIM image. The paths
2322 * in the extract commands are expected to be already "canonicalized". */
2324 do_wimlib_extract_files(WIMStruct *wim,
2326 struct wimlib_extract_command *cmds,
2328 wimlib_progress_func_t progress_func)
2331 bool found_link_cmd = false;
2332 bool found_nolink_cmd = false;
2334 /* Select the image from which we are extracting files */
2335 ret = select_wim_image(wim, image);
2339 /* Make sure there are no streams in the WIM that have not been
2340 * checksummed yet. */
2341 ret = wim_checksum_unhashed_streams(wim);
2345 /* Check for problems with the extraction commands */
2346 for (size_t i = 0; i < num_cmds; i++) {
2347 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2350 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2351 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2352 found_link_cmd = true;
2354 found_nolink_cmd = true;
2356 if (found_link_cmd && found_nolink_cmd) {
2357 ERROR("Symlink or hardlink extraction mode must "
2358 "be set on all extraction commands");
2359 return WIMLIB_ERR_INVALID_PARAM;
2363 /* Execute the extraction commands */
2364 for (size_t i = 0; i < num_cmds; i++) {
2365 ret = extract_tree(wim,
2366 cmds[i].wim_source_path,
2367 cmds[i].fs_dest_path,
2368 cmds[i].extract_flags,
2376 /* API function documented in wimlib.h */
2378 wimlib_extract_files(WIMStruct *wim,
2380 const struct wimlib_extract_command *cmds,
2382 int default_extract_flags,
2383 WIMStruct **additional_swms,
2384 unsigned num_additional_swms,
2385 wimlib_progress_func_t progress_func)
2388 struct wimlib_extract_command *cmds_copy;
2391 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2393 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2400 if (num_additional_swms)
2401 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2403 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2405 ret = WIMLIB_ERR_NOMEM;
2406 goto out_restore_lookup_table;
2409 for (size_t i = 0; i < num_cmds; i++) {
2410 cmds_copy[i].extract_flags = (default_extract_flags |
2411 cmds[i].extract_flags)
2412 & WIMLIB_EXTRACT_MASK_PUBLIC;
2413 all_flags |= cmds_copy[i].extract_flags;
2415 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
2416 if (!cmds_copy[i].wim_source_path) {
2417 ret = WIMLIB_ERR_NOMEM;
2418 goto out_free_cmds_copy;
2421 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
2422 if (!cmds_copy[i].fs_dest_path) {
2423 ret = WIMLIB_ERR_NOMEM;
2424 goto out_free_cmds_copy;
2428 ret = do_wimlib_extract_files(wim, image,
2429 cmds_copy, num_cmds,
2432 if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2433 WIMLIB_EXTRACT_FLAG_HARDLINK))
2435 for_lookup_table_entry(wim->lookup_table,
2436 lte_free_extracted_file, NULL);
2439 for (size_t i = 0; i < num_cmds; i++) {
2440 FREE(cmds_copy[i].wim_source_path);
2441 FREE(cmds_copy[i].fs_dest_path);
2444 out_restore_lookup_table:
2445 if (num_additional_swms)
2446 unmerge_lookup_table(wim);
2452 * Extracts an image from a WIM file.
2454 * @wim: WIMStruct for the WIM file.
2456 * @image: Number of the single image to extract.
2458 * @target: Directory or NTFS volume to extract the image to.
2460 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2462 * @progress_func: If non-NULL, a progress function to be called
2465 * Returns 0 on success; nonzero on failure.
2468 extract_single_image(WIMStruct *wim, int image,
2469 const tchar *target, int extract_flags,
2470 wimlib_progress_func_t progress_func)
2473 tchar *target_copy = canonicalize_fs_path(target);
2475 return WIMLIB_ERR_NOMEM;
2476 struct wimlib_extract_command cmd = {
2477 .wim_source_path = T(""),
2478 .fs_dest_path = target_copy,
2479 .extract_flags = extract_flags,
2481 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2486 static const tchar * const filename_forbidden_chars =
2495 /* This function checks if it is okay to use a WIM image's name as a directory
2498 image_name_ok_as_dir(const tchar *image_name)
2500 return image_name && *image_name &&
2501 !tstrpbrk(image_name, filename_forbidden_chars) &&
2502 tstrcmp(image_name, T(".")) &&
2503 tstrcmp(image_name, T(".."));
2506 /* Extracts all images from the WIM to the directory @target, with the images
2507 * placed in subdirectories named by their image names. */
2509 extract_all_images(WIMStruct *wim,
2510 const tchar *target,
2512 wimlib_progress_func_t progress_func)
2514 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2515 size_t output_path_len = tstrlen(target);
2516 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2519 const tchar *image_name;
2522 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2524 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2525 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2526 return WIMLIB_ERR_INVALID_PARAM;
2529 if (tstat(target, &stbuf)) {
2530 if (errno == ENOENT) {
2531 if (tmkdir(target, 0755)) {
2532 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2533 return WIMLIB_ERR_MKDIR;
2536 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2537 return WIMLIB_ERR_STAT;
2539 } else if (!S_ISDIR(stbuf.st_mode)) {
2540 ERROR("\"%"TS"\" is not a directory", target);
2541 return WIMLIB_ERR_NOTDIR;
2544 tmemcpy(buf, target, output_path_len);
2545 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2546 for (image = 1; image <= wim->hdr.image_count; image++) {
2547 image_name = wimlib_get_image_name(wim, image);
2548 if (image_name_ok_as_dir(image_name)) {
2549 tstrcpy(buf + output_path_len + 1, image_name);
2551 /* Image name is empty or contains forbidden characters.
2552 * Use image number instead. */
2553 tsprintf(buf + output_path_len + 1, T("%d"), image);
2555 ret = extract_single_image(wim, image, buf, extract_flags,
2564 do_wimlib_extract_image(WIMStruct *wim,
2566 const tchar *target,
2568 WIMStruct **additional_swms,
2569 unsigned num_additional_swms,
2570 wimlib_progress_func_t progress_func)
2574 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2575 wimlib_assert(wim->hdr.part_number == 1);
2576 wimlib_assert(num_additional_swms == 0);
2578 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2582 if (num_additional_swms)
2583 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2586 if (image == WIMLIB_ALL_IMAGES) {
2587 ret = extract_all_images(wim, target, extract_flags,
2590 ret = extract_single_image(wim, image, target, extract_flags,
2594 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2595 WIMLIB_EXTRACT_FLAG_HARDLINK))
2597 for_lookup_table_entry(wim->lookup_table,
2598 lte_free_extracted_file,
2601 if (num_additional_swms)
2602 unmerge_lookup_table(wim);
2606 /* API function documented in wimlib.h */
2608 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2609 const tchar *target, int extract_flags,
2610 wimlib_progress_func_t progress_func)
2614 struct filedes *in_fd;
2618 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2620 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2621 return WIMLIB_ERR_INVALID_PARAM;
2623 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2625 /* Read the WIM header from the pipe and get a WIMStruct to represent
2626 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2627 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2628 * an empty lookup table, no XML data read, and no filename set. */
2629 ret = open_wim_as_WIMStruct(&pipe_fd,
2630 WIMLIB_OPEN_FLAG_FROM_PIPE |
2631 WIMLIB_OPEN_FLAG_SPLIT_OK,
2632 &pwm, progress_func);
2636 /* Sanity check to make sure this is a pipable WIM. */
2637 if (pwm->hdr.magic != PWM_MAGIC) {
2638 ERROR("The WIM being read from file descriptor %d "
2639 "is not pipable!", pipe_fd);
2640 ret = WIMLIB_ERR_NOT_PIPABLE;
2641 goto out_wimlib_free;
2644 /* Sanity check to make sure the first part of a pipable split WIM is
2645 * sent over the pipe first. */
2646 if (pwm->hdr.part_number != 1) {
2647 ERROR("The first part of the split WIM must be "
2648 "sent over the pipe first.");
2649 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2650 goto out_wimlib_free;
2653 in_fd = &pwm->in_fd;
2654 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2656 /* As mentioned, the WIMStruct we created from the pipe does not have
2657 * XML data yet. Fix this by reading the extra copy of the XML data
2658 * that directly follows the header in pipable WIMs. (Note: see
2659 * write_pipable_wim() for more details about the format of pipable
2662 struct wim_lookup_table_entry xml_lte;
2663 ret = read_pwm_stream_header(pwm, &xml_lte, 0);
2665 goto out_wimlib_free;
2667 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2669 ERROR("Expected XML data, but found non-metadata "
2671 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2672 goto out_wimlib_free;
2675 copy_resource_entry(&pwm->hdr.xml_res_entry,
2676 &xml_lte.resource_entry);
2678 ret = read_wim_xml_data(pwm);
2680 goto out_wimlib_free;
2681 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2682 ERROR("Image count in XML data is not the same as in WIM header.");
2683 ret = WIMLIB_ERR_XML;
2684 goto out_wimlib_free;
2688 /* Get image index (this may use the XML data that was just read to
2689 * resolve an image name). */
2690 if (image_num_or_name) {
2691 image = wimlib_resolve_image(pwm, image_num_or_name);
2692 if (image == WIMLIB_NO_IMAGE) {
2693 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2695 ret = WIMLIB_ERR_INVALID_IMAGE;
2696 goto out_wimlib_free;
2697 } else if (image == WIMLIB_ALL_IMAGES) {
2698 ERROR("Applying all images from a pipe is not supported.");
2699 ret = WIMLIB_ERR_INVALID_IMAGE;
2700 goto out_wimlib_free;
2703 if (pwm->hdr.image_count != 1) {
2704 ERROR("No image was specified, but the pipable WIM "
2705 "did not contain exactly 1 image");
2706 ret = WIMLIB_ERR_INVALID_IMAGE;
2707 goto out_wimlib_free;
2712 /* Load the needed metadata resource. */
2713 for (i = 1; i <= pwm->hdr.image_count; i++) {
2714 struct wim_lookup_table_entry *metadata_lte;
2715 struct wim_image_metadata *imd;
2717 metadata_lte = new_lookup_table_entry();
2718 if (!metadata_lte) {
2719 ret = WIMLIB_ERR_NOMEM;
2720 goto out_wimlib_free;
2723 ret = read_pwm_stream_header(pwm, metadata_lte, 0);
2724 imd = pwm->image_metadata[i - 1];
2725 imd->metadata_lte = metadata_lte;
2727 goto out_wimlib_free;
2729 if (!(metadata_lte->resource_entry.flags &
2730 WIM_RESHDR_FLAG_METADATA))
2732 ERROR("Expected metadata resource, but found "
2733 "non-metadata stream.");
2734 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2735 goto out_wimlib_free;
2739 /* Metadata resource is for the images being extracted.
2740 * Parse it and save the metadata in memory. */
2741 ret = read_metadata_resource(pwm, imd);
2743 goto out_wimlib_free;
2746 /* Metadata resource is not for the image being
2747 * extracted. Skip over it. */
2748 ret = skip_pwm_stream(metadata_lte);
2750 goto out_wimlib_free;
2753 /* Extract the image. */
2754 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2755 ret = do_wimlib_extract_image(pwm, image, target,
2756 extract_flags, NULL, 0, progress_func);
2757 /* Clean up and return. */
2763 /* API function documented in wimlib.h */
2765 wimlib_extract_image(WIMStruct *wim,
2767 const tchar *target,
2769 WIMStruct **additional_swms,
2770 unsigned num_additional_swms,
2771 wimlib_progress_func_t progress_func)
2773 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2774 return do_wimlib_extract_image(wim, image, target, extract_flags,
2775 additional_swms, num_additional_swms,