4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012, 2013 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
13 * wimlib is free software; you can redistribute it and/or modify it under the
14 * terms of the GNU General Public License as published by the Free
15 * Software Foundation; either version 3 of the License, or (at your option)
18 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20 * A PARTICULAR PURPOSE. See the GNU General Public License for more
23 * You should have received a copy of the GNU General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
28 * This file provides the API functions wimlib_extract_image(),
29 * wimlib_extract_files(), and wimlib_extract_image_from_pipe(). Internally,
30 * all end up calling extract_tree() zero or more times to extract a tree of
31 * files from the currently selected WIM image to the specified target directory
34 * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
35 * Win32), this file does not itself have code to extract files or directories
36 * to any specific target; instead, it handles generic functionality and relies
37 * on lower-level callback functions declared in `struct apply_operations' to do
38 * the actual extraction.
45 #include "wimlib/apply.h"
46 #include "wimlib/dentry.h"
47 #include "wimlib/encoding.h"
48 #include "wimlib/endianness.h"
49 #include "wimlib/error.h"
50 #include "wimlib/lookup_table.h"
51 #include "wimlib/metadata.h"
52 #include "wimlib/paths.h"
53 #include "wimlib/reparse.h"
54 #include "wimlib/resource.h"
55 #include "wimlib/security.h"
56 #include "wimlib/swm.h"
58 # include "wimlib/win32.h" /* for realpath() equivalent */
60 #include "wimlib/xml.h"
61 #include "wimlib/wim.h"
69 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000
71 #define WIMLIB_EXTRACT_MASK_PUBLIC 0x3fffffff
73 /* Given a WIM dentry in the tree to be extracted, resolve all streams in the
74 * corresponding inode and set 'out_refcnt' in each to 0. */
76 dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_ctx)
78 struct apply_ctx *ctx = _ctx;
79 struct wim_inode *inode = dentry->d_inode;
80 struct wim_lookup_table_entry *lte;
84 if (dentry->extraction_skipped)
87 /* Special case: when extracting from a pipe, the WIM lookup table is
88 * initially empty, so "resolving" an inode's streams is initially not
89 * possible. However, we still need to keep track of which streams,
90 * identified by SHA1 message digests, need to be extracted, so we
91 * "resolve" the inode's streams anyway by allocating new entries. */
92 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
94 ret = inode_resolve_ltes(inode, ctx->wim->lookup_table, force);
97 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
98 lte = inode_stream_lte_resolved(inode, i);
106 is_linked_extraction(const struct apply_ctx *ctx)
108 return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
109 WIMLIB_EXTRACT_FLAG_SYMLINK));
113 can_extract_named_data_streams(const struct apply_ctx *ctx)
115 return ctx->supported_features.named_data_streams &&
116 !is_linked_extraction(ctx);
120 ref_stream_to_extract(struct wim_lookup_table_entry *lte,
121 struct wim_dentry *dentry, struct apply_ctx *ctx)
126 if (likely(!is_linked_extraction(ctx)) || (lte->out_refcnt == 0 &&
127 lte->extracted_file == NULL))
129 ctx->progress.extract.total_bytes += wim_resource_size(lte);
130 ctx->progress.extract.num_streams++;
133 if (lte->out_refcnt == 0) {
134 list_add_tail(<e->extraction_list, &ctx->stream_list);
135 ctx->num_streams_remaining++;
138 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
139 struct wim_dentry **lte_dentries;
141 /* Append dentry to this stream's array of dentries referencing
142 * it. Use inline array to avoid memory allocation until the
143 * number of dentries becomes too large. */
144 if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) {
145 lte_dentries = lte->inline_lte_dentries;
147 struct wim_dentry **prev_lte_dentries;
148 size_t alloc_lte_dentries;
150 if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) {
151 prev_lte_dentries = NULL;
152 alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries);
154 prev_lte_dentries = lte->lte_dentries;
155 alloc_lte_dentries = lte->alloc_lte_dentries;
158 if (lte->out_refcnt == alloc_lte_dentries) {
159 alloc_lte_dentries *= 2;
160 lte_dentries = REALLOC(prev_lte_dentries,
162 sizeof(lte_dentries[0]));
164 return WIMLIB_ERR_NOMEM;
165 if (prev_lte_dentries == NULL) {
167 lte->inline_lte_dentries,
168 sizeof(lte->inline_lte_dentries));
170 lte->lte_dentries = lte_dentries;
171 lte->alloc_lte_dentries = alloc_lte_dentries;
173 lte_dentries = lte->lte_dentries;
175 lte_dentries[lte->out_refcnt] = dentry;
181 /* Given a WIM dentry in the tree to be extracted, iterate through streams that
182 * need to be extracted. For each one, add it to the list of streams to be
183 * extracted (ctx->stream_list) if not already done so, and also update the
184 * progress information (ctx->progress) with the stream. Furthermore, if doing
185 * a sequential extraction, build a mapping from each the stream to the dentries
188 dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx)
190 struct apply_ctx *ctx = _ctx;
191 struct wim_inode *inode = dentry->d_inode;
194 /* Don't process dentries marked as skipped. */
195 if (dentry->extraction_skipped)
198 /* Don't process additional hard links. */
199 if (inode->i_visited && ctx->supported_features.hard_links)
202 /* The unnamed data stream will always be extracted, except in an
204 if (!inode_is_encrypted_directory(inode)) {
205 ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode),
211 /* Named data streams will be extracted only if supported in the current
212 * extraction mode and volume, and to avoid complications, if not doing
213 * a linked extraction. */
214 if (can_extract_named_data_streams(ctx)) {
215 for (u16 i = 0; i < inode->i_num_ads; i++) {
216 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
218 ret = ref_stream_to_extract(inode->i_ads_entries[i].lte,
224 inode->i_visited = 1;
228 /* Inform library user of progress of stream extraction following the successful
229 * extraction of a copy of the stream specified by @lte. */
231 update_extract_progress(struct apply_ctx *ctx,
232 const struct wim_lookup_table_entry *lte)
234 wimlib_progress_func_t progress_func = ctx->progress_func;
235 union wimlib_progress_info *progress = &ctx->progress;
237 progress->extract.completed_bytes += wim_resource_size(lte);
239 progress->extract.completed_bytes >= ctx->next_progress)
241 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress);
242 if (progress->extract.completed_bytes >=
243 progress->extract.total_bytes)
245 ctx->next_progress = ~0ULL;
247 ctx->next_progress += progress->extract.total_bytes / 128;
248 if (ctx->next_progress > progress->extract.total_bytes)
249 ctx->next_progress = progress->extract.total_bytes;
255 /* Extract a symbolic link (not directly as reparse data), handling fixing up
256 * the target of absolute symbolic links and updating the extract progress.
258 * @inode must specify the WIM inode for a symbolic link or junction reparse
261 * @lte_override overrides the resource used as the reparse data for the
264 extract_symlink(const tchar *path, struct apply_ctx *ctx,
265 struct wim_inode *inode,
266 struct wim_lookup_table_entry *lte_override)
268 ssize_t bufsize = ctx->ops->path_max;
269 tchar target[bufsize];
275 /* If absolute symbolic link fixups requested, reserve space in the link
276 * target buffer for the absolute path of the target directory. */
277 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
279 buf += ctx->realtarget_nchars;
280 bufsize -= ctx->realtarget_nchars;
283 /* Translate the WIM inode's reparse data into the link target. */
284 sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override);
287 return WIMLIB_ERR_READLINK;
291 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
294 /* Fix absolute symbolic link target to point into the
295 * actual extraction destination. */
296 tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars);
297 fixed_target = target;
299 /* Keep same link target. */
303 /* Call into the apply_operations to create the symbolic link. */
304 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
306 ret = ctx->ops->create_symlink(fixed_target, path, ctx);
308 ERROR_WITH_ERRNO("Failed to create symlink "
309 "\"%"TS"\" => \"%"TS"\"", path, fixed_target);
313 /* Account for reparse data consumed. */
314 update_extract_progress(ctx,
315 (lte_override ? lte_override :
316 inode_unnamed_lte_resolved(inode)));
319 #endif /* !__WIN32__ */
321 /* Create a file, directory, or symbolic link. */
323 extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode)
328 if (ctx->supported_features.symlink_reparse_points &&
329 !ctx->supported_features.reparse_points &&
330 inode_is_symlink(inode))
332 ret = extract_symlink(path, ctx, inode, NULL);
334 #endif /* !__WIN32__ */
335 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
336 ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie);
338 ERROR_WITH_ERRNO("Failed to create the directory "
342 ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie);
344 ERROR_WITH_ERRNO("Failed to create the file "
352 extract_hardlink(const tchar *oldpath, const tchar *newpath,
353 struct apply_ctx *ctx)
357 DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath);
358 ret = ctx->ops->create_hardlink(oldpath, newpath, ctx);
360 ERROR_WITH_ERRNO("Failed to create hardlink "
361 "\"%"TS"\" => \"%"TS"\"",
369 try_extract_rpfix(u8 *rpbuf,
371 const wchar_t *extract_root_realpath,
372 unsigned extract_root_realpath_nchars)
374 struct reparse_data rpdata;
376 size_t target_nchars;
377 size_t stripped_nchars;
378 wchar_t *stripped_target;
379 wchar_t stripped_target_nchars;
382 utf16lechar *new_target;
383 utf16lechar *new_print_name;
384 size_t new_target_nchars;
385 size_t new_print_name_nchars;
388 ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata);
392 if (extract_root_realpath[0] == L'\0' ||
393 extract_root_realpath[1] != L':' ||
394 extract_root_realpath[2] != L'\\')
395 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
397 ret = parse_substitute_name(rpdata.substitute_name,
398 rpdata.substitute_name_nbytes,
402 stripped_nchars = ret;
403 target = rpdata.substitute_name;
404 target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar);
405 stripped_target = target + stripped_nchars;
406 stripped_target_nchars = target_nchars - stripped_nchars;
408 new_target = alloca((6 + extract_root_realpath_nchars +
409 stripped_target_nchars) * sizeof(utf16lechar));
412 if (stripped_nchars == 6) {
413 /* Include \??\ prefix if it was present before */
414 p = wmempcpy(p, L"\\??\\", 4);
417 /* Print name excludes the \??\ if present. */
419 if (stripped_nchars != 0) {
420 /* Get drive letter from real path to extract root, if a drive
421 * letter was present before. */
422 *p++ = extract_root_realpath[0];
423 *p++ = extract_root_realpath[1];
425 /* Copy the rest of the extract root */
426 p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2);
428 /* Append the stripped target */
429 p = wmempcpy(p, stripped_target, stripped_target_nchars);
430 new_target_nchars = p - new_target;
431 new_print_name_nchars = p - new_print_name;
433 if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE ||
434 new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE)
435 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
437 rpdata.substitute_name = new_target;
438 rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar);
439 rpdata.print_name = new_print_name;
440 rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar);
441 return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p);
443 #endif /* __WIN32__ */
445 /* Set reparse data on extracted file or directory that has
446 * FILE_ATTRIBUTE_REPARSE_POINT set. */
448 extract_reparse_data(const tchar *path, struct apply_ctx *ctx,
449 struct wim_inode *inode,
450 struct wim_lookup_table_entry *lte_override)
453 u8 rpbuf[REPARSE_POINT_MAX_SIZE];
456 ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override);
461 /* Fix up target of absolute symbolic link or junction points so
462 * that they point into the actual extraction target. */
463 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
464 (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
465 inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) &&
466 !inode->i_not_rpfixed)
468 ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget,
469 ctx->realtarget_nchars);
470 if (ret && !(ctx->extract_flags &
471 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
473 WARNING("Reparse point fixup of \"%"TS"\" "
482 ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx);
484 /* On Windows, the SeCreateSymbolicLink privilege is required to create
485 * symbolic links. To be more friendly towards non-Administrator users,
486 * we merely warn the user if symbolic links cannot be created due to
487 * insufficient permissions or privileges, unless
488 * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided. */
490 if (ret && inode_is_symlink(inode) &&
491 (errno == EACCES || errno == EPERM) &&
492 !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
494 WARNING("Can't set reparse data on \"%"TS"\": "
496 " You may be trying to "
497 "extract a symbolic link without the\n"
498 " SeCreateSymbolicLink privilege, "
499 "which by default non-Administrator\n"
500 " accounts do not have.",
508 /* Account for reparse data consumed. */
509 update_extract_progress(ctx,
510 (lte_override ? lte_override :
511 inode_unnamed_lte_resolved(inode)));
515 ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path);
520 * Extract zero or more streams to a file.
522 * This function operates slightly differently depending on whether @lte_spec is
523 * NULL or not. When @lte_spec is NULL, the behavior is to extract the default
524 * file contents (unnamed stream), and, if named data streams are supported in
525 * the extract mode and volume, any named data streams. When @lte_spec is NULL,
526 * the behavior is to extract only all copies of the stream @lte_spec, and in
527 * addition use @lte_spec to set the reparse data or create the symbolic link if
531 * Path to file to extract (as can be passed to apply_operations
536 * WIM dentry that corresponds to the file being extracted.
538 * If non-NULL, specifies the lookup table entry for a stream to extract,
539 * and only that stream will be extracted (although there may be more than
540 * one instance of it).
542 * Used only if @lte_spec != NULL; it is passed to the extraction functions
543 * rather than @lte_spec, allowing the location of the stream to be
544 * overridden. (This is used when the WIM is being read from a nonseekable
545 * file, such as a pipe, when streams need to be used more than once; each
546 * such stream is extracted to a temporary file.)
549 extract_streams(const tchar *path, struct apply_ctx *ctx,
550 struct wim_dentry *dentry,
551 struct wim_lookup_table_entry *lte_spec,
552 struct wim_lookup_table_entry *lte_override)
554 struct wim_inode *inode = dentry->d_inode;
555 struct wim_lookup_table_entry *lte;
556 file_spec_t file_spec;
559 if (dentry->was_hardlinked)
566 for (unsigned i = 0; i < SHA1_HASH_SIZE; i++)
567 p += sprintf(p, "%02x", lte_override->hash[i]);
568 DEBUG("Extracting stream SHA1=%s to \"%"TS"\"",
569 sha1_str, path, inode->i_ino);
571 DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino);
575 if (ctx->ops->uses_cookies)
576 file_spec.cookie = inode->extract_cookie;
578 file_spec.path = path;
580 /* Unnamed data stream. */
581 lte = inode_unnamed_lte_resolved(inode);
582 if (lte && (!lte_spec || lte == lte_spec)) {
585 if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
586 FILE_ATTRIBUTE_REPARSE_POINT)))
588 if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) &&
589 ctx->supported_features.encrypted_files)
590 ret = ctx->ops->extract_encrypted_stream(file_spec, lte, ctx);
592 ret = ctx->ops->extract_unnamed_stream(file_spec, lte, ctx);
595 update_extract_progress(ctx, lte);
597 else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
600 if (ctx->supported_features.reparse_points)
601 ret = extract_reparse_data(path, ctx, inode, lte);
603 else if ((inode_is_symlink(inode) &&
604 ctx->supported_features.symlink_reparse_points))
605 ret = extract_symlink(path, ctx, inode, lte);
612 /* Named data streams. */
613 if (can_extract_named_data_streams(ctx)) {
614 for (u16 i = 0; i < inode->i_num_ads; i++) {
615 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
617 if (!ads_entry_is_named_stream(entry))
622 if (lte_spec && lte_spec != lte)
626 ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name,
627 entry->stream_name_nbytes / 2,
631 update_extract_progress(ctx, lte);
637 ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path);
641 /* Set attributes on an extracted file or directory if supported by the
642 * extraction mode. */
644 extract_file_attributes(const tchar *path, struct apply_ctx *ctx,
645 struct wim_dentry *dentry)
649 if (ctx->ops->set_file_attributes) {
650 if (dentry == ctx->extract_root && ctx->root_dentry_is_special)
652 ret = ctx->ops->set_file_attributes(path,
653 dentry->d_inode->i_attributes,
656 ERROR_WITH_ERRNO("Failed to set attributes on "
665 /* Set or remove the short (DOS) name on an extracted file or directory if
666 * supported by the extraction mode. Since DOS names are unimportant and it's
667 * easy to run into problems setting them on Windows (SetFileShortName()
668 * requires SE_RESTORE privilege, which only the Administrator can request, and
669 * also requires DELETE access to the file), failure is ignored unless
670 * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set. */
672 extract_short_name(const tchar *path, struct apply_ctx *ctx,
673 struct wim_dentry *dentry)
677 /* The root of the dentry tree being extracted may not be extracted to
678 * its original name, so its short name should be ignored. */
679 if (dentry == ctx->extract_root)
682 if (ctx->supported_features.short_names) {
683 ret = ctx->ops->set_short_name(path,
685 dentry->short_name_nbytes / 2,
687 if (ret && (ctx->extract_flags &
688 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES))
690 ERROR_WITH_ERRNO("Failed to set short name of "
698 /* Set security descriptor, UNIX data, or neither on an extracted file, taking
699 * into account the current extraction mode and flags. */
701 extract_security(const tchar *path, struct apply_ctx *ctx,
702 struct wim_dentry *dentry)
705 struct wim_inode *inode = dentry->d_inode;
707 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS)
710 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
714 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
715 struct wimlib_unix_data data;
717 ret = inode_get_unix_data(inode, &data, NULL);
721 ret = ctx->ops->set_unix_data(path, &data, ctx);
723 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
724 ERROR_WITH_ERRNO("Failed to set UNIX owner, "
725 "group, and/or mode on "
729 WARNING_WITH_ERRNO("Failed to set UNIX owner, "
730 "group, and/or/mode on "
736 #endif /* __WIN32__ */
737 if (ctx->supported_features.security_descriptors &&
738 inode->i_security_id != -1)
740 const struct wim_security_data *sd;
744 sd = wim_const_security_data(ctx->wim);
745 desc = sd->descriptors[inode->i_security_id];
746 desc_size = sd->sizes[inode->i_security_id];
748 ret = ctx->ops->set_security_descriptor(path, desc,
751 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
752 ERROR_WITH_ERRNO("Failed to set security "
753 "descriptor on \"%"TS"\"", path);
756 if (errno != EACCES) {
757 WARNING_WITH_ERRNO("Failed to set "
758 "security descriptor "
759 "on \"%"TS"\"", path);
767 /* Set timestamps on an extracted file. Failure is warning-only unless
768 * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */
770 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
771 struct wim_dentry *dentry)
773 struct wim_inode *inode = dentry->d_inode;
776 if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special)
779 if (ctx->ops->set_timestamps) {
780 ret = ctx->ops->set_timestamps(path,
781 inode->i_creation_time,
782 inode->i_last_write_time,
783 inode->i_last_access_time,
786 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
787 ERROR_WITH_ERRNO("Failed to set timestamps "
788 "on \"%"TS"\"", path);
791 WARNING_WITH_ERRNO("Failed to set timestamps "
792 "on \"%"TS"\"", path);
799 /* Check whether the extraction of a dentry should be skipped completely. */
801 dentry_is_supported(struct wim_dentry *dentry,
802 const struct wim_features *supported_features)
804 struct wim_inode *inode = dentry->d_inode;
806 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
807 if (supported_features->reparse_points)
809 if (supported_features->symlink_reparse_points &&
810 inode_is_symlink(inode))
817 /* Given a WIM dentry to extract, build the path to which to extract it, in the
818 * format understood by the callbacks in the apply_operations being used.
820 * Write the resulting path into @path, which must have room for at least
821 * ctx->ops->max_path characters including the null-terminator.
823 * Return %true if successful; %false if this WIM dentry doesn't actually need
824 * to be extracted or if the calculated path exceeds ctx->ops->max_path
827 * This function clobbers the tmp_list member of @dentry and its ancestors up
828 * until the extraction root. */
830 build_extraction_path(tchar path[], struct wim_dentry *dentry,
831 struct apply_ctx *ctx)
834 LIST_HEAD(ancestor_list);
836 const tchar *target_prefix;
837 size_t target_prefix_nchars;
838 struct wim_dentry *d;
840 if (dentry->extraction_skipped)
843 path_nchars = ctx->ops->path_prefix_nchars;
845 if (ctx->ops->requires_realtarget_in_paths) {
846 target_prefix = ctx->realtarget;
847 target_prefix_nchars = ctx->realtarget_nchars;
848 } else if (ctx->ops->requires_target_in_paths) {
849 target_prefix = ctx->target;
850 target_prefix_nchars = ctx->target_nchars;
852 target_prefix = NULL;
853 target_prefix_nchars = 0;
855 path_nchars += target_prefix_nchars;
857 for (d = dentry; d != ctx->extract_root; d = d->parent) {
858 path_nchars += d->extraction_name_nchars + 1;
859 list_add(&d->tmp_list, &ancestor_list);
862 path_nchars++; /* null terminator */
864 if (path_nchars > ctx->ops->path_max) {
865 WARNING("\"%"TS"\": Path too long to extract",
866 dentry_full_path(dentry));
870 p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
871 p = tmempcpy(p, target_prefix, target_prefix_nchars);
872 list_for_each_entry(d, &ancestor_list, tmp_list) {
873 *p++ = ctx->ops->path_separator;
874 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
877 wimlib_assert(p - path == path_nchars);
882 get_num_path_components(const tchar *path, tchar path_separator)
884 unsigned num_components = 0;
887 while (*path == path_separator)
891 while (*path && *path != path_separator)
894 return num_components;
898 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
899 struct apply_ctx *ctx, struct wim_dentry *dentry)
901 size_t num_raw_path_components;
902 const struct wim_dentry *d;
903 size_t num_target_path_components;
908 num_raw_path_components = 0;
909 for (d = dentry; d != ctx->extract_root; d = d->parent)
910 num_raw_path_components++;
912 if (ctx->ops->requires_realtarget_in_paths)
913 num_target_path_components = get_num_path_components(ctx->realtarget,
914 ctx->ops->path_separator);
915 else if (ctx->ops->requires_target_in_paths)
916 num_target_path_components = get_num_path_components(ctx->target,
917 ctx->ops->path_separator);
919 num_target_path_components = 0;
921 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
922 wimlib_assert(num_target_path_components > 0);
923 num_raw_path_components++;
924 num_target_path_components--;
928 while (*p_old == ctx->ops->path_separator)
930 while (--num_target_path_components) {
931 while (*p_old != ctx->ops->path_separator)
933 while (*p_old == ctx->ops->path_separator)
937 tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
939 p = &symlink_target[0];
940 while (num_raw_path_components--) {
943 *p++ = ctx->ops->path_separator;
946 DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
947 newpath, symlink_target);
948 ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
950 ERROR_WITH_ERRNO("Failed to create symlink "
951 "\"%"TS"\" => \"%"TS"\"",
952 newpath, symlink_target);
957 /* Create the "skeleton" of an extracted file or directory. Don't yet extract
958 * data streams, reparse data (including symbolic links), timestamps, and
959 * security descriptors. Basically, everything that doesn't require reading
960 * non-metadata resources from the WIM file and isn't delayed until the final
963 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
964 struct apply_ctx *ctx)
966 struct wim_inode *inode = dentry->d_inode;
968 const tchar *oldpath;
970 if (unlikely(is_linked_extraction(ctx))) {
971 struct wim_lookup_table_entry *unnamed_lte;
973 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
974 if (unnamed_lte && unnamed_lte->extracted_file) {
975 oldpath = unnamed_lte->extracted_file;
976 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
983 /* Create hard link if this dentry corresponds to an already-extracted
985 if (inode->i_extracted_file) {
986 oldpath = inode->i_extracted_file;
990 /* Skip symlinks unless they can be extracted as reparse points rather
991 * than created directly. */
992 if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
995 /* Create this file or directory unless it's the extraction root, which
996 * was already created if necessary. */
997 if (dentry != ctx->extract_root) {
998 ret = extract_inode(path, ctx, inode);
1003 /* Create empty named data streams. */
1004 if (can_extract_named_data_streams(ctx)) {
1005 for (u16 i = 0; i < inode->i_num_ads; i++) {
1006 file_spec_t file_spec;
1007 struct wim_ads_entry *entry = &inode->i_ads_entries[i];
1009 if (!ads_entry_is_named_stream(entry))
1013 if (ctx->ops->uses_cookies)
1014 file_spec.cookie = inode->extract_cookie;
1016 file_spec.path = path;
1017 ret = ctx->ops->extract_named_stream(file_spec,
1019 entry->stream_name_nbytes / 2,
1022 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
1023 "empty named data stream",
1030 /* Set file attributes (if supported). */
1031 ret = extract_file_attributes(path, ctx, dentry);
1035 /* Set or remove file short name (if supported). */
1036 ret = extract_short_name(path, ctx, dentry);
1040 /* If inode has multiple links and hard links are supported in this
1041 * extraction mode and volume, save the path to the extracted file in
1042 * case it's needed to create a hard link. */
1043 if (unlikely(is_linked_extraction(ctx))) {
1044 struct wim_lookup_table_entry *unnamed_lte;
1046 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
1048 unnamed_lte->extracted_file = TSTRDUP(path);
1049 if (!unnamed_lte->extracted_file)
1050 return WIMLIB_ERR_NOMEM;
1052 } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
1053 inode->i_extracted_file = TSTRDUP(path);
1054 if (!inode->i_extracted_file)
1055 return WIMLIB_ERR_NOMEM;
1060 ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
1063 dentry->was_hardlinked = 1;
1067 ret = extract_hardlink(oldpath, path, ctx);
1070 dentry->was_hardlinked = 1;
1075 dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx)
1077 struct apply_ctx *ctx = _ctx;
1078 tchar path[ctx->ops->path_max];
1079 struct wim_dentry *orig_dentry;
1080 struct wim_dentry *other_dentry;
1083 /* Here we may re-order the extraction of multiple names (hard links)
1084 * for the same file in the same directory in order to ensure the short
1085 * (DOS) name is set correctly. A short name is always associated with
1086 * exactly one long name, and at least on NTFS, only one long name for a
1087 * file can have a short name associated with it. (More specifically,
1088 * there can be unlimited names in the POSIX namespace, but only one
1089 * name can be in the Win32+DOS namespace, or one name in the Win32
1090 * namespace with a corresponding name in the DOS namespace.) To ensure
1091 * the short name of a file is associated with the correct long name in
1092 * a directory, we extract the long name with a corresponding short name
1093 * before any additional names. This can affect NTFS-3g extraction
1094 * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying
1095 * the long name to associate with a short name) and may affect Win32
1096 * extraction as well (which uses SetFileShortName()). */
1098 if (dentry->skeleton_extracted)
1101 if (ctx->supported_features.short_names
1102 && !dentry_has_short_name(dentry)
1103 && !dentry->d_inode->i_dos_name_extracted)
1105 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1106 if (dentry_has_short_name(other_dentry)
1107 && !other_dentry->skeleton_extracted
1108 && other_dentry->parent == dentry->parent)
1110 DEBUG("Creating %"TS" before %"TS" "
1111 "to guarantee correct DOS name extraction",
1112 dentry_full_path(other_dentry),
1113 dentry_full_path(dentry));
1114 orig_dentry = dentry;
1115 dentry = other_dentry;
1121 if (!build_extraction_path(path, dentry, ctx))
1123 ret = do_dentry_extract_skeleton(path, dentry, ctx);
1127 dentry->skeleton_extracted = 1;
1130 dentry = orig_dentry;
1134 dentry->d_inode->i_dos_name_extracted = 1;
1138 /* Create a file or directory, then immediately extract all streams. This
1139 * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM
1140 * may not be read sequentially by this function. */
1142 dentry_extract(struct wim_dentry *dentry, void *_ctx)
1144 struct apply_ctx *ctx = _ctx;
1145 tchar path[ctx->ops->path_max];
1148 ret = dentry_extract_skeleton(dentry, ctx);
1152 if (!build_extraction_path(path, dentry, ctx))
1155 return extract_streams(path, ctx, dentry, NULL, NULL);
1158 /* Extract all instances of the stream @lte that are being extracted in this
1159 * call of extract_tree(). @can_seek specifies whether the WIM file descriptor
1160 * is seekable or not (e.g. is a pipe). If not and the stream needs to be
1161 * extracted multiple times, it is extracted to a temporary file first.
1163 * This is intended for use with sequential extraction of a WIM image
1164 * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */
1166 extract_stream_instances(struct wim_lookup_table_entry *lte,
1167 struct apply_ctx *ctx, bool can_seek)
1169 struct wim_dentry **lte_dentries;
1170 struct wim_lookup_table_entry *lte_tmp = NULL;
1171 struct wim_lookup_table_entry *lte_override;
1172 tchar *stream_tmp_filename = NULL;
1173 tchar path[ctx->ops->path_max];
1177 if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1178 lte_dentries = lte->inline_lte_dentries;
1180 lte_dentries = lte->lte_dentries;
1182 if (likely(can_seek || lte->out_refcnt < 2)) {
1185 /* Need to extract stream to temporary file. */
1189 stream_tmp_filename = ttempnam(NULL, T("wimlib"));
1190 if (!stream_tmp_filename) {
1191 ERROR_WITH_ERRNO("Failed to create temporary filename");
1192 ret = WIMLIB_ERR_OPEN;
1196 lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
1198 ret = WIMLIB_ERR_NOMEM;
1199 goto out_free_stream_tmp_filename;
1201 lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
1202 lte_tmp->file_on_disk = stream_tmp_filename;
1203 lte_override = lte_tmp;
1205 raw_fd = topen(stream_tmp_filename,
1206 O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
1208 ERROR_WITH_ERRNO("Failed to open temporary file");
1209 ret = WIMLIB_ERR_OPEN;
1210 goto out_free_lte_tmp;
1212 filedes_init(&fd, raw_fd);
1213 ret = extract_wim_resource_to_fd(lte, &fd,
1214 wim_resource_size(lte));
1215 if (filedes_close(&fd) && !ret)
1216 ret = WIMLIB_ERR_WRITE;
1218 goto out_unlink_stream_tmp_file;
1221 /* Extract all instances of the stream, reading either from the stream
1222 * in the WIM file or from the temporary file containing the stream.
1223 * dentry->tmp_flag is used to ensure that each dentry is processed only
1224 * once regardless of how many times this stream appears in the streams
1225 * of the corresponding inode. */
1226 for (i = 0; i < lte->out_refcnt; i++) {
1227 struct wim_dentry *dentry = lte_dentries[i];
1229 if (dentry->tmp_flag)
1231 if (!build_extraction_path(path, dentry, ctx))
1233 ret = extract_streams(path, ctx, dentry,
1236 goto out_clear_tmp_flags;
1237 dentry->tmp_flag = 1;
1240 out_clear_tmp_flags:
1241 for (i = 0; i < lte->out_refcnt; i++)
1242 lte_dentries[i]->tmp_flag = 0;
1243 out_unlink_stream_tmp_file:
1244 if (stream_tmp_filename)
1245 tunlink(stream_tmp_filename);
1248 out_free_stream_tmp_filename:
1249 FREE(stream_tmp_filename);
1254 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1255 * structure and empty files were already created. This relies on the
1256 * per-`struct wim_lookup_table_entry' list of dentries that reference each
1257 * stream that was constructed earlier. Streams are extracted exactly in the
1258 * order of the stream list; however, unless the WIM's file descriptor is
1259 * detected to be non-seekable, streams may be read from the WIM file more than
1260 * one time if multiple copies need to be extracted. */
1262 extract_stream_list(struct apply_ctx *ctx)
1264 struct wim_lookup_table_entry *lte;
1268 can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
1269 list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
1270 ret = extract_stream_instances(lte, ctx, can_seek);
1277 #define PWM_ALLOW_WIM_HDR 0x00001
1278 #define PWM_SILENT_EOF 0x00002
1280 /* Read the header from a stream in a pipable WIM. */
1282 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1283 int flags, struct wim_header_disk *hdr_ret)
1286 struct pwm_stream_hdr stream_hdr;
1287 struct wim_header_disk pwm_hdr;
1291 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
1295 if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
1296 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
1297 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
1298 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
1302 lte->resource_location = RESOURCE_NONEXISTENT;
1303 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
1307 if (buf.stream_hdr.magic != PWM_STREAM_MAGIC) {
1308 ERROR("Data read on pipe is invalid (expected stream header).");
1309 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1312 lte->resource_entry.original_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
1313 copy_hash(lte->hash, buf.stream_hdr.hash);
1314 lte->resource_entry.flags = le32_to_cpu(buf.stream_hdr.flags);
1315 lte->resource_entry.offset = pwm->in_fd.offset;
1316 lte->resource_location = RESOURCE_IN_WIM;
1318 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) {
1319 lte->compression_type = pwm->compression_type;
1320 lte->resource_entry.size = 0;
1322 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
1323 lte->resource_entry.size = lte->resource_entry.original_size;
1325 lte->is_pipable = 1;
1329 if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1330 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1334 /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */
1336 skip_pwm_stream(struct wim_lookup_table_entry *lte)
1338 return read_partial_wim_resource(lte, wim_resource_size(lte),
1340 WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY,
1345 extract_streams_from_pipe(struct apply_ctx *ctx)
1347 struct wim_lookup_table_entry *found_lte;
1348 struct wim_lookup_table_entry *needed_lte;
1349 struct wim_lookup_table *lookup_table;
1350 struct wim_header_disk pwm_hdr;
1354 ret = WIMLIB_ERR_NOMEM;
1355 found_lte = new_lookup_table_entry();
1359 lookup_table = ctx->wim->lookup_table;
1360 pwm_flags = PWM_ALLOW_WIM_HDR;
1361 if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1362 pwm_flags |= PWM_SILENT_EOF;
1363 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
1364 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
1365 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
1366 if (ctx->progress_func)
1367 ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1369 while (ctx->num_streams_remaining) {
1370 ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags,
1373 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1374 (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1378 goto out_free_found_lte;
1381 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1382 && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
1383 && (needed_lte = __lookup_resource(lookup_table, found_lte->hash))
1384 && (needed_lte->out_refcnt))
1386 copy_resource_entry(&needed_lte->resource_entry,
1387 &found_lte->resource_entry);
1388 needed_lte->resource_location = found_lte->resource_location;
1389 needed_lte->wim = found_lte->wim;
1390 needed_lte->compression_type = found_lte->compression_type;
1391 needed_lte->is_pipable = found_lte->is_pipable;
1393 ret = extract_stream_instances(needed_lte, ctx, false);
1395 goto out_free_found_lte;
1396 ctx->num_streams_remaining--;
1397 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1398 ret = skip_pwm_stream(found_lte);
1400 goto out_free_found_lte;
1402 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
1403 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
1405 if (part_number != ctx->progress.extract.part_number ||
1406 total_parts != ctx->progress.extract.total_parts ||
1407 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
1410 ctx->progress.extract.part_number = part_number;
1411 ctx->progress.extract.total_parts = total_parts;
1412 memcpy(ctx->progress.extract.guid,
1413 pwm_hdr.guid, WIM_GID_LEN);
1414 if (ctx->progress_func) {
1416 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1425 free_lookup_table_entry(found_lte);
1434 /* Finish extracting a file, directory, or symbolic link by setting file
1435 * security and timestamps. */
1437 dentry_extract_final(struct wim_dentry *dentry, void *_ctx)
1439 struct apply_ctx *ctx = _ctx;
1441 tchar path[ctx->ops->path_max];
1443 if (!build_extraction_path(path, dentry, ctx))
1446 ret = extract_security(path, ctx, dentry);
1450 return extract_timestamps(path, ctx, dentry);
1454 * Extract a WIM dentry to standard output.
1456 * This obviously doesn't make sense in all cases. We return an error if the
1457 * dentry does not correspond to a regular file. Otherwise we extract the
1458 * unnamed data stream only.
1461 extract_dentry_to_stdout(struct wim_dentry *dentry)
1464 if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1465 FILE_ATTRIBUTE_DIRECTORY))
1467 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1468 "extracted to standard output", dentry_full_path(dentry));
1469 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
1471 struct wim_lookup_table_entry *lte;
1473 lte = inode_unnamed_lte_resolved(dentry->d_inode);
1475 struct filedes _stdout;
1476 filedes_init(&_stdout, STDOUT_FILENO);
1477 ret = extract_wim_resource_to_fd(lte, &_stdout,
1478 wim_resource_size(lte));
1485 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1487 static const utf16lechar replacement_char = cpu_to_le16('?');
1491 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1497 for (i = 0; i < num_chars; i++) {
1500 case cpu_to_le16('\\'):
1501 case cpu_to_le16(':'):
1502 case cpu_to_le16('*'):
1503 case cpu_to_le16('?'):
1504 case cpu_to_le16('"'):
1505 case cpu_to_le16('<'):
1506 case cpu_to_le16('>'):
1507 case cpu_to_le16('|'):
1509 case cpu_to_le16('/'):
1510 case cpu_to_le16('\0'):
1512 name[i] = replacement_char;
1519 if (name[num_chars - 1] == cpu_to_le16(' ') ||
1520 name[num_chars - 1] == cpu_to_le16('.'))
1523 name[num_chars - 1] = replacement_char;
1532 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
1534 const utf16lechar *file_name = dentry->file_name;
1535 return file_name != NULL &&
1536 file_name[0] == cpu_to_le16('.') &&
1537 (file_name[1] == cpu_to_le16('\0') ||
1538 (file_name[1] == cpu_to_le16('.') &&
1539 file_name[2] == cpu_to_le16('\0')));
1543 dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore)
1545 dentry->extraction_skipped = 1;
1550 * dentry_calculate_extraction_path-
1552 * Calculate the actual filename component at which a WIM dentry will be
1553 * extracted, handling invalid filenames "properly".
1555 * dentry->extraction_name usually will be set the same as dentry->file_name (on
1556 * UNIX, converted into the platform's multibyte encoding). However, if the
1557 * file name contains characters that are not valid on the current platform or
1558 * has some other format that is not valid, leave dentry->extraction_name as
1559 * NULL and set dentry->extraction_skipped to indicate that this dentry should
1560 * not be extracted, unless the appropriate flag
1561 * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
1562 * which case a substitute filename will be created and set instead.
1564 * Conflicts with case-insensitive names on Windows are handled similarly; see
1568 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
1570 struct apply_ctx *ctx = _args;
1573 if (dentry == ctx->extract_root || dentry->extraction_skipped)
1576 if (!dentry_is_supported(dentry, &ctx->supported_features))
1579 if (dentry_is_dot_or_dotdot(dentry)) {
1580 /* WIM files shouldn't contain . or .. entries. But if they are
1581 * there, don't attempt to extract them. */
1582 WARNING("Skipping extraction of unexpected . or .. file "
1583 "\"%"TS"\"", dentry_full_path(dentry));
1588 if (!ctx->ops->supports_case_sensitive_filenames)
1590 struct wim_dentry *other;
1591 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1592 case_insensitive_conflict_list)
1594 if (ctx->extract_flags &
1595 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1596 WARNING("\"%"TS"\" has the same "
1597 "case-insensitive name as "
1598 "\"%"TS"\"; extracting "
1599 "dummy name instead",
1600 dentry_full_path(dentry),
1601 dentry_full_path(other));
1604 WARNING("Not extracting \"%"TS"\": "
1605 "has same case-insensitive "
1606 "name as \"%"TS"\"",
1607 dentry_full_path(dentry),
1608 dentry_full_path(other));
1613 #else /* __WIN32__ */
1614 wimlib_assert(ctx->ops->supports_case_sensitive_filenames);
1615 #endif /* !__WIN32__ */
1617 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1619 dentry->extraction_name = dentry->file_name;
1620 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1623 return utf16le_to_tstr(dentry->file_name,
1624 dentry->file_name_nbytes,
1625 &dentry->extraction_name,
1626 &dentry->extraction_name_nchars);
1629 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1631 WARNING("\"%"TS"\" has an invalid filename "
1632 "that is not supported on this platform; "
1633 "extracting dummy name instead",
1634 dentry_full_path(dentry));
1637 WARNING("Not extracting \"%"TS"\": has an invalid filename "
1638 "that is not supported on this platform",
1639 dentry_full_path(dentry));
1646 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1648 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1649 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1652 size_t tchar_nchars;
1654 tchar_name = utf16_name_copy;
1655 tchar_nchars = dentry->file_name_nbytes / 2;
1657 ret = utf16le_to_tstr(utf16_name_copy,
1658 dentry->file_name_nbytes,
1659 &tchar_name, &tchar_nchars);
1663 size_t fixed_name_num_chars = tchar_nchars;
1664 tchar fixed_name[tchar_nchars + 50];
1666 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1667 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1668 T(" (invalid filename #%lu)"),
1669 ++ctx->invalid_sequence);
1673 dentry->extraction_name = memdup(fixed_name,
1674 2 * fixed_name_num_chars + 2);
1675 if (!dentry->extraction_name)
1676 return WIMLIB_ERR_NOMEM;
1677 dentry->extraction_name_nchars = fixed_name_num_chars;
1682 for_dentry_in_tree(dentry, dentry_mark_skipped, NULL);
1686 /* Clean up dentry and inode structure after extraction. */
1688 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
1690 struct wim_inode *inode = dentry->d_inode;
1692 dentry->extraction_skipped = 0;
1693 dentry->was_hardlinked = 0;
1694 dentry->skeleton_extracted = 0;
1695 inode->i_visited = 0;
1696 FREE(inode->i_extracted_file);
1697 inode->i_extracted_file = NULL;
1698 inode->i_dos_name_extracted = 0;
1699 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
1700 FREE(dentry->extraction_name);
1701 dentry->extraction_name = NULL;
1705 /* Tally features necessary to extract a dentry and the corresponding inode. */
1707 dentry_tally_features(struct wim_dentry *dentry, void *_features)
1709 struct wim_features *features = _features;
1710 struct wim_inode *inode = dentry->d_inode;
1712 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1713 features->archive_files++;
1714 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1715 features->hidden_files++;
1716 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1717 features->system_files++;
1718 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1719 features->compressed_files++;
1720 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED)
1721 features->encrypted_files++;
1722 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1723 features->not_context_indexed_files++;
1724 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1725 features->sparse_files++;
1726 if (inode_has_named_stream(inode))
1727 features->named_data_streams++;
1728 if (inode->i_visited)
1729 features->hard_links++;
1730 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1731 features->reparse_points++;
1732 if (inode_is_symlink(inode))
1733 features->symlink_reparse_points++;
1735 features->other_reparse_points++;
1737 if (inode->i_security_id != -1)
1738 features->security_descriptors++;
1739 if (dentry->short_name_nbytes)
1740 features->short_names++;
1741 if (inode_has_unix_data(inode))
1742 features->unix_data++;
1743 inode->i_visited = 1;
1748 dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
1750 dentry->d_inode->i_visited = 0;
1754 /* Tally the features necessary to extract a dentry tree. */
1756 dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features)
1758 memset(features, 0, sizeof(struct wim_features));
1759 for_dentry_in_tree(root, dentry_tally_features, features);
1760 for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
1764 do_feature_check(const struct wim_features *required_features,
1765 const struct wim_features *supported_features,
1767 const struct apply_operations *ops,
1768 const tchar *wim_source_path)
1771 const tchar *mode = "this extraction mode";
1773 if (wim_source_path[0] == '\0')
1774 loc = "the WIM image";
1776 loc = wim_source_path;
1778 /* We're an archive program, so theoretically we can do what we want
1779 * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't
1780 * bother the user about it. */
1782 if (required_features->archive_files && !supported_features->archive_files)
1785 "%lu files in %"TS" are marked as archived, but this attribute\n"
1786 " is not supported in %"TS".",
1787 required_features->archive_files, loc, mode);
1791 if (required_features->hidden_files && !supported_features->hidden_files)
1794 "%lu files in %"TS" are marked as hidden, but this\n"
1795 " attribute is not supported in %"TS".",
1796 required_features->hidden_files, loc, mode);
1799 if (required_features->system_files && !supported_features->system_files)
1802 "%lu files in %"TS" are marked as system files,\n"
1803 " but this attribute is not supported in %"TS".",
1804 required_features->system_files, loc, mode);
1807 if (required_features->compressed_files && !supported_features->compressed_files)
1810 "%lu files in %"TS" are marked as being transparently\n"
1811 " compressed, but transparent compression is not supported in\n"
1812 " %"TS". These files will be extracted as uncompressed.",
1813 required_features->compressed_files, loc, mode);
1816 if (required_features->encrypted_files && !supported_features->encrypted_files)
1819 "%lu files in %"TS" are marked as being encrypted,\n"
1820 " but encryption is not supported in %"TS". These files\n"
1821 " will be extracted as raw encrypted data instead.",
1822 required_features->encrypted_files, loc, mode);
1825 if (required_features->not_context_indexed_files &&
1826 !supported_features->not_context_indexed_files)
1829 "%lu files in %"TS" are marked as not content indexed,\n"
1830 " but this attribute is not supported in %"TS".",
1831 required_features->not_context_indexed_files, loc, mode);
1834 if (required_features->sparse_files && !supported_features->sparse_files)
1837 "%lu files in %"TS" are marked as sparse, but creating\n"
1838 " sparse files is not supported in %"TS". These files\n"
1839 " will be extracted as non-sparse.",
1840 required_features->sparse_files, loc, mode);
1843 if (required_features->named_data_streams &&
1844 !supported_features->named_data_streams)
1847 "%lu files in %"TS" contain one or more alternate (named)\n"
1848 " data streams, which are not supported in %"TS".\n"
1849 " Alternate data streams will NOT be extracted.",
1850 required_features->named_data_streams, loc, mode);
1853 if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
1854 WIMLIB_EXTRACT_FLAG_SYMLINK)) &&
1855 required_features->named_data_streams &&
1856 supported_features->named_data_streams)
1859 "%lu files in %"TS" contain one or more alternate (named)\n"
1860 " data streams, which are not supported in linked extraction mode.\n"
1861 " Alternate data streams will NOT be extracted.",
1862 required_features->named_data_streams, loc);
1865 if (required_features->hard_links && !supported_features->hard_links)
1868 "%lu files in %"TS" are hard links, but hard links are\n"
1869 " not supported in %"TS". Hard links will be extracted as\n"
1870 " duplicate copies of the linked files.",
1871 required_features->hard_links, loc, mode);
1874 if (required_features->reparse_points && !supported_features->reparse_points)
1876 if (supported_features->symlink_reparse_points) {
1877 if (required_features->other_reparse_points) {
1879 "%lu files in %"TS" are reparse points that are neither\n"
1880 " symbolic links nor junction points and are not supported in\n"
1881 " %"TS". These reparse points will not be extracted.",
1882 required_features->other_reparse_points, loc,
1887 "%lu files in %"TS" are reparse points, which are\n"
1888 " not supported in %"TS" and will not be extracted.",
1889 required_features->reparse_points, loc, mode);
1893 if (required_features->security_descriptors &&
1894 !supported_features->security_descriptors)
1897 "%lu files in %"TS" have Windows NT security descriptors,\n"
1898 " but extracting security descriptors is not supported in\n"
1899 " %"TS". No security descriptors will be extracted.",
1900 required_features->security_descriptors, loc, mode);
1903 if (required_features->short_names && !supported_features->short_names)
1906 "%lu files in %"TS" have short (DOS) names, but\n"
1907 " extracting short names is not supported in %"TS".\n"
1908 " Short names will not be extracted.\n",
1909 required_features->short_names, loc, mode);
1912 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1913 required_features->unix_data && !supported_features->unix_data)
1915 ERROR("Extracting UNIX data is not supported in %"TS, mode);
1916 return WIMLIB_ERR_UNSUPPORTED;
1918 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) &&
1919 required_features->short_names && !supported_features->short_names)
1921 ERROR("Extracting short names is not supported in %"TS"", mode);
1922 return WIMLIB_ERR_UNSUPPORTED;
1924 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1925 !ops->set_timestamps)
1927 ERROR("Extracting timestamps is not supported in %"TS"", mode);
1928 return WIMLIB_ERR_UNSUPPORTED;
1930 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1931 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1932 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1933 required_features->security_descriptors &&
1934 !supported_features->security_descriptors)
1936 ERROR("Extracting security descriptors is not supported in %"TS, mode);
1937 return WIMLIB_ERR_UNSUPPORTED;
1940 if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
1941 !supported_features->hard_links)
1943 ERROR("Hard link extraction mode requested, but "
1944 "%"TS" does not support hard links!", mode);
1945 return WIMLIB_ERR_UNSUPPORTED;
1948 if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
1949 !supported_features->symlink_reparse_points)
1951 ERROR("Symbolic link extraction mode requested, but "
1952 "%"TS" does not support symbolic "
1954 return WIMLIB_ERR_UNSUPPORTED;
1960 * extract_tree - Extract a file or directory tree from the currently selected
1963 * @wim: WIMStruct for the WIM file, with the desired image selected
1964 * (as wim->current_image).
1967 * "Canonical" (i.e. no leading or trailing slashes, path
1968 * separators WIM_PATH_SEPARATOR) path inside the WIM image to
1969 * extract. An empty string means the full image.
1972 * Filesystem path to extract the file or directory tree to.
1973 * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.)
1976 * WIMLIB_EXTRACT_FLAG_*. Also, the private flag
1977 * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
1978 * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
1982 * If non-NULL, progress function for the extraction. The messages
1983 * that may be sent in this function are:
1985 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
1986 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
1987 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
1988 * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
1989 * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
1990 * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
1991 * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
1992 * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
1993 * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
1995 * Returns 0 on success; a positive WIMLIB_ERR_* code on failure.
1998 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
1999 int extract_flags, wimlib_progress_func_t progress_func)
2001 struct wim_dentry *root;
2002 struct wim_features required_features;
2003 struct apply_ctx ctx;
2005 struct wim_lookup_table_entry *lte;
2007 /* Start initializing the apply_ctx. */
2008 memset(&ctx, 0, sizeof(struct apply_ctx));
2010 ctx.extract_flags = extract_flags;
2011 ctx.target = target;
2012 ctx.target_nchars = tstrlen(target);
2013 ctx.progress_func = progress_func;
2014 if (progress_func) {
2015 ctx.progress.extract.wimfile_name = wim->filename;
2016 ctx.progress.extract.image = wim->current_image;
2017 ctx.progress.extract.extract_flags = (extract_flags &
2018 WIMLIB_EXTRACT_MASK_PUBLIC);
2019 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
2020 wim->current_image);
2021 ctx.progress.extract.extract_root_wim_source_path = wim_source_path;
2022 ctx.progress.extract.target = target;
2024 INIT_LIST_HEAD(&ctx.stream_list);
2026 /* Translate the path to extract into the corresponding
2027 * `struct wim_dentry', which will be the root of the
2028 * "dentry tree" to extract. */
2029 root = get_dentry(wim, wim_source_path);
2031 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
2032 wim_source_path, wim->current_image);
2033 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
2037 ctx.extract_root = root;
2039 /* Select the appropriate apply_operations based on the
2040 * platform and extract_flags. */
2042 ctx.ops = &win32_apply_ops;
2044 ctx.ops = &unix_apply_ops;
2048 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
2049 ctx.ops = &ntfs_3g_apply_ops;
2052 /* Call the start_extract() callback. This gives the apply_operations
2053 * implementation a chance to do any setup needed to access the volume.
2054 * Furthermore, it's expected to set the supported features of this
2055 * extraction mode (ctx.supported_features), which are determined at
2056 * runtime as they may vary depending on the actual volume. These
2057 * features are then compared with the actual features extracting this
2058 * dentry tree requires. Some mismatches will merely produce warnings
2059 * and the unsupported data will be ignored; others will produce errors.
2061 ret = ctx.ops->start_extract(target, &ctx);
2065 dentry_tree_get_features(root, &required_features);
2066 ret = do_feature_check(&required_features, &ctx.supported_features,
2067 extract_flags, ctx.ops, wim_source_path);
2069 goto out_finish_or_abort_extract;
2071 /* Figure out whether the root dentry is being extracted to the root of
2072 * a volume and therefore needs to be treated "specially", for example
2073 * not being explicitly created and not having attributes set. */
2074 if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2075 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2077 /* Calculate the actual filename component of each extracted dentry. In
2078 * the process, set the dentry->extraction_skipped flag on dentries that
2079 * are being skipped for some reason (e.g. invalid filename). */
2080 ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx);
2082 goto out_dentry_reset_needs_extraction;
2084 /* Build the list of the streams that need to be extracted and
2085 * initialize ctx.progress.extract with stream information. */
2086 ret = for_dentry_in_tree(ctx.extract_root,
2087 dentry_resolve_and_zero_lte_refcnt, &ctx);
2089 goto out_dentry_reset_needs_extraction;
2091 ret = for_dentry_in_tree(ctx.extract_root,
2092 dentry_add_streams_to_extract, &ctx);
2094 goto out_teardown_stream_list;
2096 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2097 /* When extracting from a pipe, the number of bytes of data to
2098 * extract can't be determined in the normal way (examining the
2099 * lookup table), since at this point all we have is a set of
2100 * SHA1 message digests of streams that need to be extracted.
2101 * However, we can get a reasonably accurate estimate by taking
2102 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2103 * data. This does assume that a full image is being extracted,
2104 * but currently there is no API for doing otherwise. (Also,
2105 * subtract <HARDLINKBYTES> from this if hard links are
2106 * supported by the extraction mode.) */
2107 ctx.progress.extract.total_bytes =
2108 wim_info_get_image_total_bytes(wim->wim_info,
2109 wim->current_image);
2110 if (ctx.supported_features.hard_links) {
2111 ctx.progress.extract.total_bytes -=
2112 wim_info_get_image_hard_link_bytes(wim->wim_info,
2113 wim->current_image);
2117 /* Handle the special case of extracting a file to standard
2118 * output. In that case, "root" should be a single file, not a
2119 * directory tree. (If not, extract_dentry_to_stdout() will
2120 * return an error.) */
2121 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
2122 ret = extract_dentry_to_stdout(root);
2123 goto out_teardown_stream_list;
2126 /* If a sequential extraction was specified, sort the streams to be
2127 * extracted by their position in the WIM file so that the WIM file can
2128 * be read sequentially. */
2129 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2130 WIMLIB_EXTRACT_FLAG_FROM_PIPE))
2131 == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2133 ret = sort_stream_list_by_sequential_order(
2135 offsetof(struct wim_lookup_table_entry,
2138 goto out_teardown_stream_list;
2141 if (ctx.ops->realpath_works_on_nonexisting_files &&
2142 ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2143 ctx.ops->requires_realtarget_in_paths))
2145 ctx.realtarget = realpath(target, NULL);
2146 if (!ctx.realtarget) {
2147 ret = WIMLIB_ERR_NOMEM;
2148 goto out_teardown_stream_list;
2150 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2153 if (progress_func) {
2154 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
2155 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
2159 if (!ctx.root_dentry_is_special)
2161 tchar path[ctx.ops->path_max];
2162 if (build_extraction_path(path, root, &ctx))
2164 ret = extract_inode(path, &ctx, root->d_inode);
2166 goto out_free_realtarget;
2170 /* If we need to fix up the targets of absolute symbolic links
2171 * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2172 * be absolute, use realpath() (or its replacement on Windows) to get
2173 * the absolute path to the extraction target. Note that this requires
2174 * the target directory to exist, unless
2175 * realpath_works_on_nonexisting_files is set in the apply_operations.
2177 if (!ctx.realtarget &&
2178 (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2179 required_features.symlink_reparse_points) ||
2180 ctx.ops->requires_realtarget_in_paths))
2182 ctx.realtarget = realpath(target, NULL);
2183 if (!ctx.realtarget) {
2184 ret = WIMLIB_ERR_NOMEM;
2185 goto out_free_realtarget;
2187 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2190 /* Finally, the important part: extract the tree of files. */
2191 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
2192 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) {
2193 /* Sequential extraction requested, so two passes are needed
2194 * (one for directory structure, one for streams.) */
2196 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2199 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2200 ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx);
2202 goto out_free_realtarget;
2205 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2207 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2208 ret = extract_streams_from_pipe(&ctx);
2210 ret = extract_stream_list(&ctx);
2212 goto out_free_realtarget;
2214 /* Sequential extraction was not requested, so we can make do
2215 * with one pass where we both create the files and extract
2218 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2220 ret = for_dentry_in_tree(root, dentry_extract, &ctx);
2222 goto out_free_realtarget;
2224 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2228 /* If the total number of bytes to extract was miscalculated, just jump
2229 * to the calculated number in order to avoid confusing the progress
2230 * function. This should only occur when extracting from a pipe. */
2231 if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2233 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2234 "extracted %"PRIu64,
2235 ctx.progress.extract.total_bytes,
2236 ctx.progress.extract.completed_bytes);
2238 if (progress_func &&
2239 ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2241 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2242 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2245 /* Apply security descriptors and timestamps. This is done at the end,
2246 * and in a depth-first manner, to prevent timestamps from getting
2247 * changed by subsequent extract operations and to minimize the chance
2248 * of the restored security descriptors getting in our way. */
2250 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2252 ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx);
2254 goto out_free_realtarget;
2256 if (progress_func) {
2257 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
2258 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
2263 out_free_realtarget:
2264 FREE(ctx.realtarget);
2265 out_teardown_stream_list:
2266 /* Free memory allocated as part of the mapping from each
2267 * wim_lookup_table_entry to the dentries that reference it. */
2268 if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
2269 list_for_each_entry(lte, &ctx.stream_list, extraction_list)
2270 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2271 FREE(lte->lte_dentries);
2272 out_dentry_reset_needs_extraction:
2273 for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
2274 out_finish_or_abort_extract:
2276 if (ctx.ops->abort_extract)
2277 ctx.ops->abort_extract(&ctx);
2279 if (ctx.ops->finish_extract)
2280 ret = ctx.ops->finish_extract(&ctx);
2286 /* Validates a single wimlib_extract_command, mostly checking to make sure the
2287 * extract flags make sense. */
2289 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
2293 /* Empty destination path? */
2294 if (cmd->fs_dest_path[0] == T('\0'))
2295 return WIMLIB_ERR_INVALID_PARAM;
2297 extract_flags = cmd->extract_flags;
2299 /* Check for invalid flag combinations */
2300 if ((extract_flags &
2301 (WIMLIB_EXTRACT_FLAG_SYMLINK |
2302 WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2303 WIMLIB_EXTRACT_FLAG_HARDLINK))
2304 return WIMLIB_ERR_INVALID_PARAM;
2306 if ((extract_flags &
2307 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2308 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2309 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2310 return WIMLIB_ERR_INVALID_PARAM;
2312 if ((extract_flags &
2313 (WIMLIB_EXTRACT_FLAG_RPFIX |
2314 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2315 WIMLIB_EXTRACT_FLAG_NORPFIX))
2316 return WIMLIB_ERR_INVALID_PARAM;
2318 if ((extract_flags &
2319 (WIMLIB_EXTRACT_FLAG_RESUME |
2320 WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2321 return WIMLIB_ERR_INVALID_PARAM;
2323 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2324 #ifndef WITH_NTFS_3G
2325 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2326 " we cannot apply a WIM image directly to a NTFS volume.");
2327 return WIMLIB_ERR_UNSUPPORTED;
2331 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2332 WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
2334 /* Do reparse point fixups by default if the WIM header says
2335 * they are enabled and we are extracting a full image. */
2336 if (wim_header_flags & WIM_HDR_FLAG_RP_FIX)
2337 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2340 /* TODO: Since UNIX data entries are stored in the file resources, in a
2341 * completely sequential extraction they may come up before the
2342 * corresponding file or symbolic link data. This needs to be handled
2344 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2345 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2346 == (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2347 WIMLIB_EXTRACT_FLAG_SEQUENTIAL))
2349 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2350 WARNING("Setting UNIX file/owner group may "
2351 "be impossible on some\n"
2353 "when applying from a pipe.");
2355 extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2356 WARNING("Disabling sequential extraction for "
2361 cmd->extract_flags = extract_flags;
2366 /* Internal function to execute extraction commands for a WIM image. The paths
2367 * in the extract commands are expected to be already "canonicalized". */
2369 do_wimlib_extract_files(WIMStruct *wim,
2371 struct wimlib_extract_command *cmds,
2373 wimlib_progress_func_t progress_func)
2376 bool found_link_cmd = false;
2377 bool found_nolink_cmd = false;
2379 /* Select the image from which we are extracting files */
2380 ret = select_wim_image(wim, image);
2384 /* Make sure there are no streams in the WIM that have not been
2385 * checksummed yet. */
2386 ret = wim_checksum_unhashed_streams(wim);
2390 /* Check for problems with the extraction commands */
2391 for (size_t i = 0; i < num_cmds; i++) {
2392 ret = check_extract_command(&cmds[i], wim->hdr.flags);
2395 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2396 WIMLIB_EXTRACT_FLAG_HARDLINK)) {
2397 found_link_cmd = true;
2399 found_nolink_cmd = true;
2401 if (found_link_cmd && found_nolink_cmd) {
2402 ERROR("Symlink or hardlink extraction mode must "
2403 "be set on all extraction commands");
2404 return WIMLIB_ERR_INVALID_PARAM;
2408 /* Execute the extraction commands */
2409 for (size_t i = 0; i < num_cmds; i++) {
2410 ret = extract_tree(wim,
2411 cmds[i].wim_source_path,
2412 cmds[i].fs_dest_path,
2413 cmds[i].extract_flags,
2421 /* API function documented in wimlib.h */
2423 wimlib_extract_files(WIMStruct *wim,
2425 const struct wimlib_extract_command *cmds,
2427 int default_extract_flags,
2428 WIMStruct **additional_swms,
2429 unsigned num_additional_swms,
2430 wimlib_progress_func_t progress_func)
2433 struct wimlib_extract_command *cmds_copy;
2436 default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2438 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2445 if (num_additional_swms)
2446 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2448 cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
2450 ret = WIMLIB_ERR_NOMEM;
2451 goto out_restore_lookup_table;
2454 for (size_t i = 0; i < num_cmds; i++) {
2455 cmds_copy[i].extract_flags = (default_extract_flags |
2456 cmds[i].extract_flags)
2457 & WIMLIB_EXTRACT_MASK_PUBLIC;
2458 all_flags |= cmds_copy[i].extract_flags;
2460 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
2461 if (!cmds_copy[i].wim_source_path) {
2462 ret = WIMLIB_ERR_NOMEM;
2463 goto out_free_cmds_copy;
2466 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
2467 if (!cmds_copy[i].fs_dest_path) {
2468 ret = WIMLIB_ERR_NOMEM;
2469 goto out_free_cmds_copy;
2473 ret = do_wimlib_extract_files(wim, image,
2474 cmds_copy, num_cmds,
2477 if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2478 WIMLIB_EXTRACT_FLAG_HARDLINK))
2480 for_lookup_table_entry(wim->lookup_table,
2481 lte_free_extracted_file, NULL);
2484 for (size_t i = 0; i < num_cmds; i++) {
2485 FREE(cmds_copy[i].wim_source_path);
2486 FREE(cmds_copy[i].fs_dest_path);
2489 out_restore_lookup_table:
2490 if (num_additional_swms)
2491 unmerge_lookup_table(wim);
2497 * Extracts an image from a WIM file.
2499 * @wim: WIMStruct for the WIM file.
2501 * @image: Number of the single image to extract.
2503 * @target: Directory or NTFS volume to extract the image to.
2505 * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*.
2507 * @progress_func: If non-NULL, a progress function to be called
2510 * Returns 0 on success; nonzero on failure.
2513 extract_single_image(WIMStruct *wim, int image,
2514 const tchar *target, int extract_flags,
2515 wimlib_progress_func_t progress_func)
2518 tchar *target_copy = canonicalize_fs_path(target);
2520 return WIMLIB_ERR_NOMEM;
2521 struct wimlib_extract_command cmd = {
2522 .wim_source_path = T(""),
2523 .fs_dest_path = target_copy,
2524 .extract_flags = extract_flags,
2526 ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
2531 static const tchar * const filename_forbidden_chars =
2540 /* This function checks if it is okay to use a WIM image's name as a directory
2543 image_name_ok_as_dir(const tchar *image_name)
2545 return image_name && *image_name &&
2546 !tstrpbrk(image_name, filename_forbidden_chars) &&
2547 tstrcmp(image_name, T(".")) &&
2548 tstrcmp(image_name, T(".."));
2551 /* Extracts all images from the WIM to the directory @target, with the images
2552 * placed in subdirectories named by their image names. */
2554 extract_all_images(WIMStruct *wim,
2555 const tchar *target,
2557 wimlib_progress_func_t progress_func)
2559 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2560 size_t output_path_len = tstrlen(target);
2561 tchar buf[output_path_len + 1 + image_name_max_len + 1];
2564 const tchar *image_name;
2567 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2569 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2570 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2571 return WIMLIB_ERR_INVALID_PARAM;
2574 if (tstat(target, &stbuf)) {
2575 if (errno == ENOENT) {
2576 if (tmkdir(target, 0755)) {
2577 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
2578 return WIMLIB_ERR_MKDIR;
2581 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2582 return WIMLIB_ERR_STAT;
2584 } else if (!S_ISDIR(stbuf.st_mode)) {
2585 ERROR("\"%"TS"\" is not a directory", target);
2586 return WIMLIB_ERR_NOTDIR;
2589 tmemcpy(buf, target, output_path_len);
2590 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2591 for (image = 1; image <= wim->hdr.image_count; image++) {
2592 image_name = wimlib_get_image_name(wim, image);
2593 if (image_name_ok_as_dir(image_name)) {
2594 tstrcpy(buf + output_path_len + 1, image_name);
2596 /* Image name is empty or contains forbidden characters.
2597 * Use image number instead. */
2598 tsprintf(buf + output_path_len + 1, T("%d"), image);
2600 ret = extract_single_image(wim, image, buf, extract_flags,
2609 do_wimlib_extract_image(WIMStruct *wim,
2611 const tchar *target,
2613 WIMStruct **additional_swms,
2614 unsigned num_additional_swms,
2615 wimlib_progress_func_t progress_func)
2619 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2620 wimlib_assert(wim->hdr.part_number == 1);
2621 wimlib_assert(num_additional_swms == 0);
2623 ret = verify_swm_set(wim, additional_swms, num_additional_swms);
2627 if (num_additional_swms)
2628 merge_lookup_tables(wim, additional_swms, num_additional_swms);
2631 if (image == WIMLIB_ALL_IMAGES) {
2632 ret = extract_all_images(wim, target, extract_flags,
2635 ret = extract_single_image(wim, image, target, extract_flags,
2639 if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2640 WIMLIB_EXTRACT_FLAG_HARDLINK))
2642 for_lookup_table_entry(wim->lookup_table,
2643 lte_free_extracted_file,
2646 if (num_additional_swms)
2647 unmerge_lookup_table(wim);
2651 /* API function documented in wimlib.h */
2653 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2654 const tchar *target, int extract_flags,
2655 wimlib_progress_func_t progress_func)
2659 struct filedes *in_fd;
2663 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2665 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2666 return WIMLIB_ERR_INVALID_PARAM;
2668 extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
2670 /* Read the WIM header from the pipe and get a WIMStruct to represent
2671 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
2672 * wimlib_open_wim(), getting a WIMStruct in this way will result in
2673 * an empty lookup table, no XML data read, and no filename set. */
2674 ret = open_wim_as_WIMStruct(&pipe_fd,
2675 WIMLIB_OPEN_FLAG_FROM_PIPE |
2676 WIMLIB_OPEN_FLAG_SPLIT_OK,
2677 &pwm, progress_func);
2681 /* Sanity check to make sure this is a pipable WIM. */
2682 if (pwm->hdr.magic != PWM_MAGIC) {
2683 ERROR("The WIM being read from file descriptor %d "
2684 "is not pipable!", pipe_fd);
2685 ret = WIMLIB_ERR_NOT_PIPABLE;
2686 goto out_wimlib_free;
2689 /* Sanity check to make sure the first part of a pipable split WIM is
2690 * sent over the pipe first. */
2691 if (pwm->hdr.part_number != 1) {
2692 ERROR("The first part of the split WIM must be "
2693 "sent over the pipe first.");
2694 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2695 goto out_wimlib_free;
2698 in_fd = &pwm->in_fd;
2699 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
2701 /* As mentioned, the WIMStruct we created from the pipe does not have
2702 * XML data yet. Fix this by reading the extra copy of the XML data
2703 * that directly follows the header in pipable WIMs. (Note: see
2704 * write_pipable_wim() for more details about the format of pipable
2707 struct wim_lookup_table_entry xml_lte;
2708 ret = read_pwm_stream_header(pwm, &xml_lte, 0, NULL);
2710 goto out_wimlib_free;
2712 if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
2714 ERROR("Expected XML data, but found non-metadata "
2716 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2717 goto out_wimlib_free;
2720 copy_resource_entry(&pwm->hdr.xml_res_entry,
2721 &xml_lte.resource_entry);
2723 ret = read_wim_xml_data(pwm);
2725 goto out_wimlib_free;
2726 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
2727 ERROR("Image count in XML data is not the same as in WIM header.");
2728 ret = WIMLIB_ERR_XML;
2729 goto out_wimlib_free;
2733 /* Get image index (this may use the XML data that was just read to
2734 * resolve an image name). */
2735 if (image_num_or_name) {
2736 image = wimlib_resolve_image(pwm, image_num_or_name);
2737 if (image == WIMLIB_NO_IMAGE) {
2738 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2740 ret = WIMLIB_ERR_INVALID_IMAGE;
2741 goto out_wimlib_free;
2742 } else if (image == WIMLIB_ALL_IMAGES) {
2743 ERROR("Applying all images from a pipe is not supported.");
2744 ret = WIMLIB_ERR_INVALID_IMAGE;
2745 goto out_wimlib_free;
2748 if (pwm->hdr.image_count != 1) {
2749 ERROR("No image was specified, but the pipable WIM "
2750 "did not contain exactly 1 image");
2751 ret = WIMLIB_ERR_INVALID_IMAGE;
2752 goto out_wimlib_free;
2757 /* Load the needed metadata resource. */
2758 for (i = 1; i <= pwm->hdr.image_count; i++) {
2759 struct wim_lookup_table_entry *metadata_lte;
2760 struct wim_image_metadata *imd;
2762 metadata_lte = new_lookup_table_entry();
2763 if (!metadata_lte) {
2764 ret = WIMLIB_ERR_NOMEM;
2765 goto out_wimlib_free;
2768 ret = read_pwm_stream_header(pwm, metadata_lte, 0, NULL);
2769 imd = pwm->image_metadata[i - 1];
2770 imd->metadata_lte = metadata_lte;
2772 goto out_wimlib_free;
2774 if (!(metadata_lte->resource_entry.flags &
2775 WIM_RESHDR_FLAG_METADATA))
2777 ERROR("Expected metadata resource, but found "
2778 "non-metadata stream.");
2779 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2780 goto out_wimlib_free;
2784 /* Metadata resource is for the images being extracted.
2785 * Parse it and save the metadata in memory. */
2786 ret = read_metadata_resource(pwm, imd);
2788 goto out_wimlib_free;
2791 /* Metadata resource is not for the image being
2792 * extracted. Skip over it. */
2793 ret = skip_pwm_stream(metadata_lte);
2795 goto out_wimlib_free;
2798 /* Extract the image. */
2799 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2800 ret = do_wimlib_extract_image(pwm, image, target,
2801 extract_flags, NULL, 0, progress_func);
2802 /* Clean up and return. */
2808 /* API function documented in wimlib.h */
2810 wimlib_extract_image(WIMStruct *wim,
2812 const tchar *target,
2814 WIMStruct **additional_swms,
2815 unsigned num_additional_swms,
2816 wimlib_progress_func_t progress_func)
2818 extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
2819 return do_wimlib_extract_image(wim, image, target, extract_flags,
2820 additional_swms, num_additional_swms,