4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012-2016 Eric Biggers
11 * This file is free software; you can redistribute it and/or modify it under
12 * the terms of the GNU Lesser General Public License as published by the Free
13 * Software Foundation; either version 3 of the License, or (at your option) any
16 * This file is distributed in the hope that it will be useful, but WITHOUT
17 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
18 * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
21 * You should have received a copy of the GNU Lesser General Public License
22 * along with this file; if not, see http://www.gnu.org/licenses/.
26 * This file provides the API functions wimlib_extract_image(),
27 * wimlib_extract_image_from_pipe(), wimlib_extract_paths(), and
28 * wimlib_extract_pathlist(). Internally, all end up calling
29 * do_wimlib_extract_paths() and extract_trees().
31 * Although wimlib supports multiple extraction modes/backends (NTFS-3G, UNIX,
32 * Win32), this file does not itself have code to extract files or directories
33 * to any specific target; instead, it handles generic functionality and relies
34 * on lower-level callback functions declared in `struct apply_operations' to do
35 * the actual extraction.
48 #include "wimlib/apply.h"
49 #include "wimlib/assert.h"
50 #include "wimlib/blob_table.h"
51 #include "wimlib/dentry.h"
52 #include "wimlib/encoding.h"
53 #include "wimlib/endianness.h"
54 #include "wimlib/error.h"
55 #include "wimlib/metadata.h"
56 #include "wimlib/object_id.h"
57 #include "wimlib/pathlist.h"
58 #include "wimlib/paths.h"
59 #include "wimlib/pattern.h"
60 #include "wimlib/reparse.h"
61 #include "wimlib/resource.h"
62 #include "wimlib/security.h"
63 #include "wimlib/unix_data.h"
64 #include "wimlib/wim.h"
65 #include "wimlib/win32.h" /* for realpath() equivalent */
66 #include "wimlib/xml.h"
68 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
69 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
71 /* Keep in sync with wimlib.h */
72 #define WIMLIB_EXTRACT_MASK_PUBLIC \
73 (WIMLIB_EXTRACT_FLAG_NTFS | \
74 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
75 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
76 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
77 WIMLIB_EXTRACT_FLAG_RPFIX | \
78 WIMLIB_EXTRACT_FLAG_NORPFIX | \
79 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
80 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
81 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
82 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
83 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
84 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
85 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
86 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
87 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
88 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
89 WIMLIB_EXTRACT_FLAG_WIMBOOT | \
90 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | \
91 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | \
92 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | \
93 WIMLIB_EXTRACT_FLAG_COMPACT_LZX \
96 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
97 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
99 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
101 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
102 return extract_progress(ctx, msg);
106 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
108 ctx->progress.extract.current_file_count = 0;
109 ctx->progress.extract.end_file_count = end_file_count;
110 return do_file_extract_progress(ctx, msg);
114 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
116 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
120 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
122 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
126 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
128 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
129 return do_file_extract_progress(ctx, msg);
133 end_file_structure_phase(struct apply_ctx *ctx)
135 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
139 end_file_metadata_phase(struct apply_ctx *ctx)
141 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
144 #define PWM_FOUND_WIM_HDR (-1)
146 /* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
147 * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */
149 read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
150 struct wim_reshdr *reshdr_ret,
151 struct wim_header_disk *pwm_hdr_ret)
154 struct pwm_blob_hdr blob_hdr;
157 ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
161 magic = le64_to_cpu(blob_hdr.magic);
163 if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
164 memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
165 ret = full_read(&pwm->in_fd,
166 (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
167 sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
170 return PWM_FOUND_WIM_HDR;
173 if (unlikely(magic != PWM_BLOB_MAGIC)) {
174 ERROR("Data read on pipe is invalid (expected blob header)");
175 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
178 copy_hash(hash_ret, blob_hdr.hash);
180 reshdr_ret->size_in_wim = 0; /* Not available */
181 reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
182 reshdr_ret->offset_in_wim = pwm->in_fd.offset;
183 reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
185 if (unlikely(reshdr_ret->uncompressed_size == 0)) {
186 ERROR("Data read on pipe is invalid (resource is of 0 size)");
187 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
193 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
194 ERROR("The pipe ended before all needed data was sent!");
196 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
201 read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
204 u8 hash[SHA1_HASH_SIZE];
205 struct wim_reshdr reshdr;
206 struct wim_header_disk pwm_hdr;
207 struct wim_resource_descriptor rdesc;
208 struct blob_descriptor *blob;
210 copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
211 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
212 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
213 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
217 while (ctx->num_blobs_remaining) {
219 ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
221 if (ret == PWM_FOUND_WIM_HDR) {
222 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
223 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
225 if (part_number == ctx->progress.extract.part_number &&
226 total_parts == ctx->progress.extract.total_parts &&
227 guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
230 copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
231 ctx->progress.extract.part_number = part_number;
232 ctx->progress.extract.total_parts = total_parts;
233 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
243 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
244 && (blob = lookup_blob(ctx->wim->blob_table, hash))
245 && (blob->out_refcnt))
247 wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
248 ret = read_blob_with_sha1(blob, cbs);
249 blob_unset_is_located_in_wim_resource(blob);
252 ctx->num_blobs_remaining--;
254 wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
255 ret = skip_wim_resource(&rdesc);
265 handle_pwm_metadata_resource(WIMStruct *pwm, int image, bool is_needed)
267 struct blob_descriptor *blob;
268 struct wim_reshdr reshdr;
269 struct wim_resource_descriptor *rdesc;
272 ret = WIMLIB_ERR_NOMEM;
273 blob = new_blob_descriptor();
277 ret = read_pwm_blob_header(pwm, blob->hash, &reshdr, NULL);
281 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
282 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
283 ERROR("Expected metadata resource, but found non-metadata "
288 ret = WIMLIB_ERR_NOMEM;
289 rdesc = MALLOC(sizeof(*rdesc));
293 wim_reshdr_to_desc_and_blob(&reshdr, pwm, rdesc, blob);
296 ret = WIMLIB_ERR_NOMEM;
297 pwm->image_metadata[image - 1] = new_unloaded_image_metadata(blob);
298 if (!pwm->image_metadata[image - 1])
302 /* If the metadata resource is for the image being extracted, then parse
303 * it and save the metadata in memory. Otherwise, skip over it. */
305 ret = select_wim_image(pwm, image);
307 ret = skip_wim_resource(rdesc);
309 free_blob_descriptor(blob);
313 /* Creates a temporary file opened for writing. The open file descriptor is
314 * returned in @fd_ret and its name is returned in @name_ret (dynamically
317 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
324 name = _wtempnam(NULL, L"wimlib");
326 ERROR_WITH_ERRNO("Failed to create temporary filename");
327 return WIMLIB_ERR_NOMEM;
329 raw_fd = _wopen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY |
330 _O_SHORT_LIVED, 0600);
331 if (raw_fd < 0 && errno == EEXIST) {
335 #else /* __WIN32__ */
336 const char *tmpdir = getenv("TMPDIR");
339 name = MALLOC(strlen(tmpdir) + 1 + 6 + 6 + 1);
341 return WIMLIB_ERR_NOMEM;
342 sprintf(name, "%s/wimlibXXXXXX", tmpdir);
343 raw_fd = mkstemp(name);
344 #endif /* !__WIN32__ */
347 ERROR_WITH_ERRNO("Failed to create temporary file "
350 return WIMLIB_ERR_OPEN;
353 filedes_init(fd_ret, raw_fd);
359 begin_extract_blob(struct blob_descriptor *blob, void *_ctx)
361 struct apply_ctx *ctx = _ctx;
363 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
364 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
366 return call_begin_blob(blob, ctx->saved_cbs);
370 extract_chunk(const struct blob_descriptor *blob, u64 offset,
371 const void *chunk, size_t size, void *_ctx)
373 struct apply_ctx *ctx = _ctx;
374 union wimlib_progress_info *progress = &ctx->progress;
375 bool last = (offset + size == blob->size);
378 if (likely(ctx->supported_features.hard_links)) {
379 progress->extract.completed_bytes +=
380 (u64)size * blob->out_refcnt;
382 progress->extract.completed_streams += blob->out_refcnt;
384 const struct blob_extraction_target *targets =
385 blob_extraction_targets(blob);
386 for (u32 i = 0; i < blob->out_refcnt; i++) {
387 const struct wim_inode *inode = targets[i].inode;
388 const struct wim_dentry *dentry;
390 inode_for_each_extraction_alias(dentry, inode) {
391 progress->extract.completed_bytes += size;
393 progress->extract.completed_streams++;
397 if (progress->extract.completed_bytes >= ctx->next_progress) {
399 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
403 set_next_progress(progress->extract.completed_bytes,
404 progress->extract.total_bytes,
405 &ctx->next_progress);
408 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
409 /* Just extracting to temporary file for now. */
410 ret = full_write(&ctx->tmpfile_fd, chunk, size);
412 ERROR_WITH_ERRNO("Error writing data to "
413 "temporary file \"%"TS"\"",
419 return call_continue_blob(blob, offset, chunk, size, ctx->saved_cbs);
422 /* Copy the blob's data from the temporary file to each of its targets.
424 * This is executed only in the very uncommon case that a blob is being
425 * extracted to more than MAX_OPEN_FILES targets! */
427 extract_from_tmpfile(const tchar *tmpfile_name,
428 const struct blob_descriptor *orig_blob,
429 const struct read_blob_callbacks *cbs)
431 struct blob_descriptor tmpfile_blob;
432 const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
435 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
436 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
437 tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
438 tmpfile_blob.out_refcnt = 1;
440 for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
441 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
442 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
450 end_extract_blob(struct blob_descriptor *blob, int status, void *_ctx)
452 struct apply_ctx *ctx = _ctx;
454 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
455 filedes_close(&ctx->tmpfile_fd);
457 status = extract_from_tmpfile(ctx->tmpfile_name, blob,
459 filedes_invalidate(&ctx->tmpfile_fd);
460 tunlink(ctx->tmpfile_name);
461 FREE(ctx->tmpfile_name);
465 return call_end_blob(blob, status, ctx->saved_cbs);
469 * Read the list of blobs to extract and feed their data into the specified
470 * callback functions.
472 * This handles checksumming each blob.
474 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
476 * This also works if the WIM is being read from a pipe.
478 * This also will split up blobs that will need to be extracted to more than
479 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
480 * Therefore, the apply_operations implementation need not worry about running
481 * out of file descriptors, unless it might open more than one file descriptor
482 * per 'blob_extraction_target' (e.g. Win32 currently might because the
483 * destination file system might not support hard links).
486 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
488 struct read_blob_callbacks wrapper_cbs = {
489 .begin_blob = begin_extract_blob,
490 .continue_blob = extract_chunk,
491 .end_blob = end_extract_blob,
494 ctx->saved_cbs = cbs;
495 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
496 return read_blobs_from_pipe(ctx, &wrapper_cbs);
498 return read_blob_list(&ctx->blob_list,
499 offsetof(struct blob_descriptor,
501 &wrapper_cbs, VERIFY_BLOB_HASHES);
505 /* Extract a WIM dentry to standard output.
507 * This obviously doesn't make sense in all cases. We return an error if the
508 * dentry does not correspond to a regular file. Otherwise we extract the
509 * unnamed data stream only. */
511 extract_dentry_to_stdout(struct wim_dentry *dentry,
512 const struct blob_table *blob_table)
514 struct wim_inode *inode = dentry->d_inode;
515 struct blob_descriptor *blob;
516 struct filedes _stdout;
518 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
519 FILE_ATTRIBUTE_DIRECTORY |
520 FILE_ATTRIBUTE_ENCRYPTED))
522 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
523 "extracted to standard output", dentry_full_path(dentry));
524 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
527 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
529 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
530 if (!is_zero_hash(hash))
531 return blob_not_found_error(inode, hash);
535 filedes_init(&_stdout, STDOUT_FILENO);
536 return extract_blob_to_fd(blob, &_stdout);
540 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
541 const struct blob_table *blob_table)
543 for (size_t i = 0; i < num_dentries; i++) {
544 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
551 /**********************************************************************/
554 * Removes duplicate dentries from the array.
556 * Returns the new number of dentries, packed at the front of the array.
559 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
562 for (i = 0; i < num_trees; i++) {
563 if (!trees[i]->d_tmp_flag) {
564 /* Found distinct dentry. */
565 trees[i]->d_tmp_flag = 1;
566 trees[j++] = trees[i];
569 for (i = 0; i < j; i++)
570 trees[i]->d_tmp_flag = 0;
575 * Remove dentries that are descendants of other dentries in the array.
577 * Returns the new number of dentries, packed at the front of the array.
580 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
583 for (i = 0; i < num_trees; i++)
584 trees[i]->d_tmp_flag = 1;
585 for (i = 0; i < num_trees; i++) {
586 struct wim_dentry *d = trees[i];
587 while (!dentry_is_root(d)) {
592 trees[j++] = trees[i];
596 trees[i]->d_tmp_flag = 0;
599 for (i = 0; i < j; i++)
600 trees[i]->d_tmp_flag = 0;
605 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
607 struct list_head *dentry_list = _dentry_list;
608 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
613 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
615 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
619 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
621 if (will_extract_dentry(dentry)) {
622 list_del(&dentry->d_extraction_list_node);
623 dentry_reset_extraction_list_node(dentry);
629 * Build the preliminary list of dentries to be extracted.
631 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
632 * an ancestor of d2, then d1 appears before d2 in the list.
635 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
636 size_t num_trees, bool add_ancestors)
638 INIT_LIST_HEAD(dentry_list);
640 /* Add the trees recursively. */
641 for (size_t i = 0; i < num_trees; i++)
642 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
644 /* If requested, add ancestors of the trees. */
646 for (size_t i = 0; i < num_trees; i++) {
647 struct wim_dentry *dentry = trees[i];
648 struct wim_dentry *ancestor;
649 struct list_head *place_after;
651 if (dentry_is_root(dentry))
654 place_after = dentry_list;
657 ancestor = ancestor->d_parent;
658 if (will_extract_dentry(ancestor)) {
659 place_after = &ancestor->d_extraction_list_node;
662 } while (!dentry_is_root(ancestor));
666 ancestor = ancestor->d_parent;
667 if (will_extract_dentry(ancestor))
669 list_add(&ancestor->d_extraction_list_node, place_after);
670 } while (!dentry_is_root(ancestor));
676 destroy_dentry_list(struct list_head *dentry_list)
678 struct wim_dentry *dentry, *tmp;
679 struct wim_inode *inode;
681 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
682 inode = dentry->d_inode;
683 dentry_reset_extraction_list_node(dentry);
684 inode->i_visited = 0;
685 inode->i_can_externally_back = 0;
686 if ((void *)dentry->d_extraction_name != (void *)dentry->d_name)
687 FREE(dentry->d_extraction_name);
688 dentry->d_extraction_name = NULL;
689 dentry->d_extraction_name_nchars = 0;
694 destroy_blob_list(struct list_head *blob_list)
696 struct blob_descriptor *blob;
698 list_for_each_entry(blob, blob_list, extraction_list)
699 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
700 FREE(blob->blob_extraction_targets);
704 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
706 static const utf16lechar replacement_char = cpu_to_le16('?');
710 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
716 for (i = 0; i < num_chars; i++) {
717 switch (le16_to_cpu(name[i])) {
719 case '\x01'...'\x1F':
732 name[i] = replacement_char;
742 dentry_calculate_extraction_name(struct wim_dentry *dentry,
743 struct apply_ctx *ctx)
747 if (dentry_is_root(dentry))
751 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
752 dentry->d_extraction_name = dentry->d_name;
753 dentry->d_extraction_name_nchars = dentry->d_name_nbytes /
759 if (!ctx->supported_features.case_sensitive_filenames) {
760 struct wim_dentry *other;
761 dentry_for_each_ci_match(other, dentry) {
762 if (will_extract_dentry(other)) {
763 if (ctx->extract_flags &
764 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
765 WARNING("\"%"TS"\" has the same "
766 "case-insensitive name as "
767 "\"%"TS"\"; extracting "
768 "dummy name instead",
769 dentry_full_path(dentry),
770 dentry_full_path(other));
773 WARNING("Not extracting \"%"TS"\": "
774 "has same case-insensitive "
776 dentry_full_path(dentry),
777 dentry_full_path(other));
784 if (file_name_valid(dentry->d_name, dentry->d_name_nbytes / 2, false)) {
786 ret = utf16le_get_tstr(dentry->d_name,
787 dentry->d_name_nbytes,
788 (const tchar **)&dentry->d_extraction_name,
790 dentry->d_extraction_name_nchars = nbytes / sizeof(tchar);
793 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
795 WARNING("\"%"TS"\" has an invalid filename "
796 "that is not supported on this platform; "
797 "extracting dummy name instead",
798 dentry_full_path(dentry));
801 WARNING("Not extracting \"%"TS"\": has an invalid filename "
802 "that is not supported on this platform",
803 dentry_full_path(dentry));
810 utf16lechar utf16_name_copy[dentry->d_name_nbytes / 2];
812 memcpy(utf16_name_copy, dentry->d_name, dentry->d_name_nbytes);
813 file_name_valid(utf16_name_copy, dentry->d_name_nbytes / 2, true);
815 const tchar *tchar_name;
818 ret = utf16le_get_tstr(utf16_name_copy,
819 dentry->d_name_nbytes,
820 &tchar_name, &tchar_nchars);
824 tchar_nchars /= sizeof(tchar);
826 size_t fixed_name_num_chars = tchar_nchars;
827 tchar fixed_name[tchar_nchars + 50];
829 tmemcpy(fixed_name, tchar_name, tchar_nchars);
830 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
831 T(" (invalid filename #%lu)"),
832 ++ctx->invalid_sequence);
834 utf16le_put_tstr(tchar_name);
836 dentry->d_extraction_name = TSTRDUP(fixed_name);
837 if (!dentry->d_extraction_name)
838 return WIMLIB_ERR_NOMEM;
839 dentry->d_extraction_name_nchars = fixed_name_num_chars;
844 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
849 * Calculate the actual filename component at which each WIM dentry will be
850 * extracted, with special handling for dentries that are unsupported by the
851 * extraction backend or have invalid names.
853 * ctx->supported_features must be filled in.
855 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
858 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
859 struct apply_ctx *ctx)
861 struct list_head *prev, *cur;
863 /* Can't use list_for_each_entry() because a call to
864 * dentry_calculate_extraction_name() may delete the current dentry and
865 * its children from the list. */
869 struct wim_dentry *dentry;
873 if (cur == dentry_list)
876 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
878 ret = dentry_calculate_extraction_name(dentry, ctx);
882 if (prev->next == cur)
885 ; /* Current dentry and its children (which follow in
886 the list) were deleted. prev stays the same. */
892 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
893 struct blob_table *blob_table)
895 struct wim_inode *inode = dentry->d_inode;
896 struct blob_descriptor *blob;
900 /* Special case: when extracting from a pipe, the WIM blob table is
901 * initially empty, so "resolving" an inode's streams is initially not
902 * possible. However, we still need to keep track of which blobs,
903 * identified by SHA-1 message digests, need to be extracted, so we
904 * "resolve" the inode's streams anyway by allocating a 'struct
905 * blob_descriptor' for each one. */
906 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
908 ret = inode_resolve_streams(inode, blob_table, force);
911 for (unsigned i = 0; i < inode->i_num_streams; i++) {
912 blob = stream_blob_resolved(&inode->i_streams[i]);
914 blob->out_refcnt = 0;
920 * For each dentry to be extracted, resolve all streams in the corresponding
921 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
923 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
926 dentry_list_resolve_streams(struct list_head *dentry_list,
927 struct apply_ctx *ctx)
929 struct wim_dentry *dentry;
932 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
933 ret = dentry_resolve_streams(dentry,
935 ctx->wim->blob_table);
943 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
944 struct apply_ctx *ctx)
946 struct wim_inode *inode = dentry->d_inode;
947 struct blob_descriptor *blob = stream_blob_resolved(strm);
948 struct blob_extraction_target *targets;
953 /* Tally the size only for each actual extraction of the stream (not
954 * additional hard links to the inode). */
955 if (inode->i_visited && ctx->supported_features.hard_links)
958 ctx->progress.extract.total_bytes += blob->size;
959 ctx->progress.extract.total_streams++;
961 if (inode->i_visited)
964 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
965 * many extraction targets it will have. */
966 if (blob->out_refcnt == 0) {
967 list_add_tail(&blob->extraction_list, &ctx->blob_list);
968 ctx->num_blobs_remaining++;
971 /* Set this stream as an extraction target of 'blob'. */
973 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
974 targets = blob->inline_blob_extraction_targets;
976 struct blob_extraction_target *prev_targets;
977 size_t alloc_blob_extraction_targets;
979 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
981 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
983 prev_targets = blob->blob_extraction_targets;
984 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
987 if (blob->out_refcnt == alloc_blob_extraction_targets) {
988 alloc_blob_extraction_targets *= 2;
989 targets = REALLOC(prev_targets,
990 alloc_blob_extraction_targets *
993 return WIMLIB_ERR_NOMEM;
996 blob->inline_blob_extraction_targets,
997 sizeof(blob->inline_blob_extraction_targets));
999 blob->blob_extraction_targets = targets;
1000 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1002 targets = blob->blob_extraction_targets;
1004 targets[blob->out_refcnt].inode = inode;
1005 targets[blob->out_refcnt].stream = strm;
1011 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1012 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1014 bool need_stream = false;
1015 switch (strm->stream_type) {
1016 case STREAM_TYPE_DATA:
1017 if (stream_is_named(strm)) {
1018 /* Named data stream */
1019 if (ctx->supported_features.named_data_streams)
1021 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1022 FILE_ATTRIBUTE_ENCRYPTED))
1023 && !(inode_is_symlink(inode)
1024 && !ctx->supported_features.reparse_points
1025 && ctx->supported_features.symlink_reparse_points))
1028 * Unnamed data stream. Skip if any of the following is true:
1030 * - file is a directory
1031 * - file is encrypted
1032 * - backend needs to create the file as UNIX symlink
1033 * - backend will extract the stream as externally
1034 * backed from the WIM archive itself
1036 if (ctx->apply_ops->will_back_from_wim) {
1037 int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx);
1038 if (ret > 0) /* Error? */
1040 if (ret < 0) /* Won't externally back? */
1047 case STREAM_TYPE_REPARSE_POINT:
1048 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1049 if (ctx->supported_features.reparse_points ||
1050 (inode_is_symlink(inode) &&
1051 ctx->supported_features.symlink_reparse_points))
1054 case STREAM_TYPE_EFSRPC_RAW_DATA:
1055 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1056 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1057 if (ctx->supported_features.encrypted_directories)
1060 if (ctx->supported_features.encrypted_files)
1066 return ref_stream(strm, dentry, ctx);
1071 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1073 struct wim_inode *inode = dentry->d_inode;
1074 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1075 int ret = ref_stream_if_needed(dentry, inode,
1076 &inode->i_streams[i], ctx);
1080 inode->i_visited = 1;
1085 * Given a list of dentries to be extracted, build the list of blobs that need
1086 * to be extracted, and for each blob determine the streams to which that blob
1087 * will be extracted.
1089 * This also initializes the extract progress info with byte and blob
1092 * ctx->supported_features must be filled in.
1095 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1097 struct wim_dentry *dentry;
1100 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1101 ret = dentry_ref_streams(dentry, ctx);
1105 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1106 dentry->d_inode->i_visited = 0;
1111 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1113 struct wim_dentry *dentry;
1115 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1116 dentry->d_inode->i_first_extraction_alias = NULL;
1118 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1119 dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias;
1120 dentry->d_inode->i_first_extraction_alias = dentry;
1125 inode_tally_features(const struct wim_inode *inode,
1126 struct wim_features *features)
1128 if (inode->i_attributes & FILE_ATTRIBUTE_READONLY)
1129 features->readonly_files++;
1130 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1131 features->hidden_files++;
1132 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1133 features->system_files++;
1134 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1135 features->archive_files++;
1136 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1137 features->compressed_files++;
1138 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1139 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1140 features->encrypted_directories++;
1142 features->encrypted_files++;
1144 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1145 features->not_context_indexed_files++;
1146 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1147 features->sparse_files++;
1148 if (inode_has_named_data_stream(inode))
1149 features->named_data_streams++;
1150 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1151 features->reparse_points++;
1152 if (inode_is_symlink(inode))
1153 features->symlink_reparse_points++;
1155 features->other_reparse_points++;
1157 if (inode_has_security_descriptor(inode))
1158 features->security_descriptors++;
1159 if (inode_has_unix_data(inode))
1160 features->unix_data++;
1161 if (inode_has_object_id(inode))
1162 features->object_ids++;
1165 /* Tally features necessary to extract a dentry and the corresponding inode. */
1167 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1169 struct wim_inode *inode = dentry->d_inode;
1171 if (dentry_has_short_name(dentry))
1172 features->short_names++;
1174 if (inode->i_visited) {
1175 features->hard_links++;
1177 inode_tally_features(inode, features);
1178 inode->i_visited = 1;
1182 /* Tally the features necessary to extract the specified dentries. */
1184 dentry_list_get_features(struct list_head *dentry_list,
1185 struct wim_features *features)
1187 struct wim_dentry *dentry;
1189 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1190 dentry_tally_features(dentry, features);
1192 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1193 dentry->d_inode->i_visited = 0;
1197 do_feature_check(const struct wim_features *required_features,
1198 const struct wim_features *supported_features,
1201 /* Encrypted files. */
1202 if (required_features->encrypted_files &&
1203 !supported_features->encrypted_files)
1204 WARNING("Ignoring EFS-encrypted data of %lu files",
1205 required_features->encrypted_files);
1207 /* Named data streams. */
1208 if (required_features->named_data_streams &&
1209 !supported_features->named_data_streams)
1210 WARNING("Ignoring named data streams of %lu files",
1211 required_features->named_data_streams);
1213 /* File attributes. */
1214 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1216 if (required_features->readonly_files &&
1217 !supported_features->readonly_files)
1218 WARNING("Ignoring FILE_ATTRIBUTE_READONLY of %lu files",
1219 required_features->readonly_files);
1221 if (required_features->hidden_files &&
1222 !supported_features->hidden_files)
1223 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1224 required_features->hidden_files);
1226 if (required_features->system_files &&
1227 !supported_features->system_files)
1228 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1229 required_features->system_files);
1231 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1232 * We're an archive program, so theoretically we can do what we
1235 if (required_features->compressed_files &&
1236 !supported_features->compressed_files)
1237 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1238 required_features->compressed_files);
1240 if (required_features->not_context_indexed_files &&
1241 !supported_features->not_context_indexed_files)
1242 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1243 required_features->not_context_indexed_files);
1245 if (required_features->sparse_files &&
1246 !supported_features->sparse_files)
1247 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1248 required_features->sparse_files);
1250 if (required_features->encrypted_directories &&
1251 !supported_features->encrypted_directories)
1252 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1253 required_features->encrypted_directories);
1257 if (required_features->hard_links && !supported_features->hard_links)
1258 WARNING("Extracting %lu hard links as independent files",
1259 required_features->hard_links);
1261 /* Symbolic links and reparse points. */
1262 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1263 required_features->symlink_reparse_points &&
1264 !supported_features->symlink_reparse_points &&
1265 !supported_features->reparse_points)
1267 ERROR("Extraction backend does not support symbolic links!");
1268 return WIMLIB_ERR_UNSUPPORTED;
1270 if (required_features->reparse_points &&
1271 !supported_features->reparse_points)
1273 if (supported_features->symlink_reparse_points) {
1274 if (required_features->other_reparse_points) {
1275 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1276 required_features->other_reparse_points);
1279 WARNING("Ignoring reparse data of %lu files",
1280 required_features->reparse_points);
1284 /* Security descriptors. */
1285 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1286 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1287 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1288 required_features->security_descriptors &&
1289 !supported_features->security_descriptors)
1291 ERROR("Extraction backend does not support security descriptors!");
1292 return WIMLIB_ERR_UNSUPPORTED;
1294 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1295 required_features->security_descriptors &&
1296 !supported_features->security_descriptors)
1297 WARNING("Ignoring Windows NT security descriptors of %lu files",
1298 required_features->security_descriptors);
1301 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1302 required_features->unix_data && !supported_features->unix_data)
1304 ERROR("Extraction backend does not support UNIX data!");
1305 return WIMLIB_ERR_UNSUPPORTED;
1308 if (required_features->unix_data &&
1309 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1311 WARNING("Ignoring UNIX metadata of %lu files",
1312 required_features->unix_data);
1316 if (required_features->object_ids && !supported_features->object_ids) {
1317 WARNING("Ignoring object IDs of %lu files",
1318 required_features->object_ids);
1322 if (required_features->short_names &&
1323 !supported_features->short_names)
1325 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1326 ERROR("Extraction backend does not support DOS names!");
1327 return WIMLIB_ERR_UNSUPPORTED;
1329 WARNING("Ignoring DOS names of %lu files",
1330 required_features->short_names);
1334 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1335 !supported_features->timestamps)
1337 ERROR("Extraction backend does not support timestamps!");
1338 return WIMLIB_ERR_UNSUPPORTED;
1344 static const struct apply_operations *
1345 select_apply_operations(int extract_flags)
1348 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1349 return &ntfs_3g_apply_ops;
1352 return &win32_apply_ops;
1354 return &unix_apply_ops;
1359 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1360 const tchar *target, int extract_flags)
1362 const struct apply_operations *ops;
1363 struct apply_ctx *ctx;
1365 LIST_HEAD(dentry_list);
1367 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1368 ret = extract_dentries_to_stdout(trees, num_trees,
1373 num_trees = remove_duplicate_trees(trees, num_trees);
1374 num_trees = remove_contained_trees(trees, num_trees);
1376 ops = select_apply_operations(extract_flags);
1378 if (num_trees > 1 && ops->single_tree_only) {
1379 ERROR("Extracting multiple directory trees "
1380 "at once is not supported in %s extraction mode!",
1382 ret = WIMLIB_ERR_UNSUPPORTED;
1386 ctx = CALLOC(1, ops->context_size);
1388 ret = WIMLIB_ERR_NOMEM;
1393 ctx->target = target;
1394 ctx->target_nchars = tstrlen(target);
1395 ctx->extract_flags = extract_flags;
1396 if (ctx->wim->progfunc) {
1397 ctx->progfunc = ctx->wim->progfunc;
1398 ctx->progctx = ctx->wim->progctx;
1399 ctx->progress.extract.image = wim->current_image;
1400 ctx->progress.extract.extract_flags = (extract_flags &
1401 WIMLIB_EXTRACT_MASK_PUBLIC);
1402 ctx->progress.extract.wimfile_name = wim->filename;
1403 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1404 wim->current_image);
1405 ctx->progress.extract.target = target;
1407 INIT_LIST_HEAD(&ctx->blob_list);
1408 filedes_invalidate(&ctx->tmpfile_fd);
1409 ctx->apply_ops = ops;
1411 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1415 build_dentry_list(&dentry_list, trees, num_trees,
1417 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1419 dentry_list_get_features(&dentry_list, &ctx->required_features);
1421 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1422 ctx->extract_flags);
1426 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1430 if (unlikely(list_empty(&dentry_list))) {
1431 WARNING("There is nothing to extract!");
1435 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1439 dentry_list_build_inode_alias_lists(&dentry_list);
1441 ret = dentry_list_ref_streams(&dentry_list, ctx);
1445 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1446 /* When extracting from a pipe, the number of bytes of data to
1447 * extract can't be determined in the normal way (examining the
1448 * blob table), since at this point all we have is a set of
1449 * SHA-1 message digests of blobs that need to be extracted.
1450 * However, we can get a reasonably accurate estimate by taking
1451 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1452 * data. This does assume that a full image is being extracted,
1453 * but currently there is no API for doing otherwise. (Also,
1454 * subtract <HARDLINKBYTES> from this if hard links are
1455 * supported by the extraction mode.) */
1456 ctx->progress.extract.total_bytes =
1457 xml_get_image_total_bytes(wim->xml_info,
1458 wim->current_image);
1459 if (ctx->supported_features.hard_links) {
1460 ctx->progress.extract.total_bytes -=
1461 xml_get_image_hard_link_bytes(wim->xml_info,
1462 wim->current_image);
1466 ret = extract_progress(ctx,
1467 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1468 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1469 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1473 ret = (*ops->extract)(&dentry_list, ctx);
1477 if (ctx->progress.extract.completed_bytes <
1478 ctx->progress.extract.total_bytes)
1480 ctx->progress.extract.completed_bytes =
1481 ctx->progress.extract.total_bytes;
1482 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1487 ret = extract_progress(ctx,
1488 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1489 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1490 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1492 destroy_blob_list(&ctx->blob_list);
1493 destroy_dentry_list(&dentry_list);
1500 mkdir_if_needed(const tchar *target)
1502 if (!tmkdir(target, 0755))
1505 if (errno == EEXIST)
1509 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1510 if (errno == EACCES)
1514 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1515 return WIMLIB_ERR_MKDIR;
1518 /* Make sure the extraction flags make sense, and update them if needed. */
1520 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1522 int extract_flags = *extract_flags_p;
1524 /* Check for invalid flag combinations */
1526 if ((extract_flags &
1527 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1528 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1529 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1530 return WIMLIB_ERR_INVALID_PARAM;
1532 if ((extract_flags &
1533 (WIMLIB_EXTRACT_FLAG_RPFIX |
1534 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1535 WIMLIB_EXTRACT_FLAG_NORPFIX))
1536 return WIMLIB_ERR_INVALID_PARAM;
1538 #ifndef WITH_NTFS_3G
1539 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1540 ERROR("wimlib was compiled without support for NTFS-3G, so\n"
1541 " it cannot apply a WIM image directly to an NTFS volume.");
1542 return WIMLIB_ERR_UNSUPPORTED;
1546 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1549 return WIMLIB_ERR_NO_FILENAME;
1551 ERROR("WIMBoot extraction is only supported on Windows!");
1552 return WIMLIB_ERR_UNSUPPORTED;
1556 if (extract_flags & (WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K |
1557 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K |
1558 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K |
1559 WIMLIB_EXTRACT_FLAG_COMPACT_LZX))
1563 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K) != 0);
1564 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K) != 0);
1565 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K) != 0);
1566 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_LZX) != 0);
1568 ERROR("Only one compression format can be specified "
1569 "for compact-mode extraction!");
1570 return WIMLIB_ERR_INVALID_PARAM;
1572 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1573 ERROR("Compact-mode extraction and WIMBoot-mode "
1574 "extraction are mutually exclusive!");
1575 return WIMLIB_ERR_INVALID_PARAM;
1578 ERROR("Compact-mode extraction (System Compression) "
1579 "is only supported on Windows!");
1580 return WIMLIB_ERR_UNSUPPORTED;
1585 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1586 WIMLIB_EXTRACT_FLAG_NORPFIX |
1587 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1588 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1590 /* For full-image extraction, do reparse point fixups by default
1591 * if the WIM header says they are enabled. */
1592 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1593 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1596 *extract_flags_p = extract_flags;
1600 struct append_dentry_ctx {
1601 struct wim_dentry **dentries;
1602 size_t num_dentries;
1603 size_t num_alloc_dentries;
1607 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1609 struct append_dentry_ctx *ctx = _ctx;
1611 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1612 struct wim_dentry **new_dentries;
1615 new_length = max(ctx->num_alloc_dentries + 8,
1616 ctx->num_alloc_dentries * 3 / 2);
1617 new_dentries = REALLOC(ctx->dentries,
1618 new_length * sizeof(ctx->dentries[0]));
1619 if (new_dentries == NULL)
1620 return WIMLIB_ERR_NOMEM;
1621 ctx->dentries = new_dentries;
1622 ctx->num_alloc_dentries = new_length;
1624 ctx->dentries[ctx->num_dentries++] = dentry;
1628 /* Append dentries matched by a path which can contain wildcard characters. */
1630 append_matched_dentries(WIMStruct *wim, const tchar *orig_pattern,
1631 int extract_flags, struct append_dentry_ctx *ctx)
1633 const size_t count_before = ctx->num_dentries;
1637 pattern = canonicalize_wim_path(orig_pattern);
1639 return WIMLIB_ERR_NOMEM;
1640 ret = expand_path_pattern(wim_get_current_root_dentry(wim), pattern,
1641 append_dentry_cb, ctx);
1643 if (ret || ctx->num_dentries > count_before)
1645 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) {
1646 ERROR("No matches for path pattern \"%"TS"\"", orig_pattern);
1647 return WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1649 WARNING("No matches for path pattern \"%"TS"\"", orig_pattern);
1654 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1655 const tchar * const *paths, size_t num_paths,
1659 struct wim_dentry **trees;
1662 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1663 (num_paths != 0 && paths == NULL))
1664 return WIMLIB_ERR_INVALID_PARAM;
1666 ret = check_extract_flags(wim, &extract_flags);
1670 ret = select_wim_image(wim, image);
1674 ret = wim_checksum_unhashed_blobs(wim);
1678 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1679 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1680 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1682 ret = mkdir_if_needed(target);
1687 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1689 struct append_dentry_ctx append_dentry_ctx = {
1692 .num_alloc_dentries = 0,
1695 for (size_t i = 0; i < num_paths; i++) {
1696 ret = append_matched_dentries(wim, paths[i],
1698 &append_dentry_ctx);
1700 trees = append_dentry_ctx.dentries;
1701 goto out_free_trees;
1704 trees = append_dentry_ctx.dentries;
1705 num_trees = append_dentry_ctx.num_dentries;
1707 trees = MALLOC(num_paths * sizeof(trees[0]));
1709 return WIMLIB_ERR_NOMEM;
1711 for (size_t i = 0; i < num_paths; i++) {
1713 tchar *path = canonicalize_wim_path(paths[i]);
1715 ret = WIMLIB_ERR_NOMEM;
1716 goto out_free_trees;
1719 trees[i] = get_dentry(wim, path,
1720 WIMLIB_CASE_PLATFORM_DEFAULT);
1722 if (trees[i] == NULL) {
1723 ERROR("Path \"%"TS"\" does not exist "
1725 paths[i], wim->current_image);
1726 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1727 goto out_free_trees;
1730 num_trees = num_paths;
1733 if (num_trees == 0) {
1735 goto out_free_trees;
1738 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1745 extract_single_image(WIMStruct *wim, int image,
1746 const tchar *target, int extract_flags)
1748 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1749 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1750 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1753 static const tchar * const filename_forbidden_chars =
1760 /* This function checks if it is okay to use a WIM image's name as a directory
1763 image_name_ok_as_dir(const tchar *image_name)
1765 return image_name && *image_name &&
1766 !tstrpbrk(image_name, filename_forbidden_chars) &&
1767 tstrcmp(image_name, T(".")) &&
1768 tstrcmp(image_name, T("..")) &&
1769 tstrlen(image_name) <= 128;
1772 /* Extracts all images from the WIM to the directory @target, with the images
1773 * placed in subdirectories named by their image names. */
1775 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1777 size_t output_path_len = tstrlen(target);
1778 tchar buf[output_path_len + 1 + 128 + 1];
1781 const tchar *image_name;
1783 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1784 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1785 return WIMLIB_ERR_INVALID_PARAM;
1788 ret = mkdir_if_needed(target);
1791 tmemcpy(buf, target, output_path_len);
1792 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1793 for (image = 1; image <= wim->hdr.image_count; image++) {
1794 image_name = wimlib_get_image_name(wim, image);
1795 if (image_name_ok_as_dir(image_name)) {
1796 tstrcpy(buf + output_path_len + 1, image_name);
1798 /* Image name is empty or contains forbidden characters.
1799 * Use image number instead. */
1800 tsprintf(buf + output_path_len + 1, T("%d"), image);
1802 ret = extract_single_image(wim, image, buf, extract_flags);
1810 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1813 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1814 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1815 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1816 return WIMLIB_ERR_INVALID_PARAM;
1818 if (image == WIMLIB_ALL_IMAGES)
1819 return extract_all_images(wim, target, extract_flags);
1821 return extract_single_image(wim, image, target, extract_flags);
1825 /****************************************************************************
1827 ****************************************************************************/
1830 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1831 const tchar * const *paths, size_t num_paths,
1834 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1835 return WIMLIB_ERR_INVALID_PARAM;
1837 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1842 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1843 const tchar *path_list_file, int extract_flags)
1850 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1852 ERROR("Failed to read path list file \"%"TS"\"",
1857 ret = wimlib_extract_paths(wim, image, target,
1858 (const tchar * const *)paths, num_paths,
1866 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1867 const tchar *image_num_or_name,
1868 const tchar *target,
1870 wimlib_progress_func_t progfunc,
1875 struct filedes *in_fd;
1879 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1880 return WIMLIB_ERR_INVALID_PARAM;
1882 /* Read the WIM header from the pipe and get a WIMStruct to represent
1883 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1884 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1885 * empty blob table, no XML data read, and no filename set. */
1886 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1891 /* Sanity check to make sure this is a pipable WIM. */
1892 if (pwm->hdr.magic != PWM_MAGIC) {
1893 ERROR("The WIM being read from file descriptor %d "
1894 "is not pipable!", pipe_fd);
1895 ret = WIMLIB_ERR_NOT_PIPABLE;
1896 goto out_wimlib_free;
1899 /* Sanity check to make sure the first part of a pipable split WIM is
1900 * sent over the pipe first. */
1901 if (pwm->hdr.part_number != 1) {
1902 ERROR("The first part of the split WIM must be "
1903 "sent over the pipe first.");
1904 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1905 goto out_wimlib_free;
1908 in_fd = &pwm->in_fd;
1909 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1911 /* As mentioned, the WIMStruct we created from the pipe does not have
1912 * XML data yet. Fix this by reading the extra copy of the XML data
1913 * that directly follows the header in pipable WIMs. (Note: see
1914 * write_pipable_wim() for more details about the format of pipable
1917 u8 hash[SHA1_HASH_SIZE];
1919 ret = read_pwm_blob_header(pwm, hash,
1920 &pwm->hdr.xml_data_reshdr, NULL);
1922 goto out_wimlib_free;
1924 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1925 ERROR("Expected XML data, but found non-metadata resource.");
1926 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1927 goto out_wimlib_free;
1930 ret = read_wim_xml_data(pwm);
1932 goto out_wimlib_free;
1934 if (xml_get_image_count(pwm->xml_info) != pwm->hdr.image_count) {
1935 ERROR("Image count in XML data is not the same as in WIM header.");
1936 ret = WIMLIB_ERR_IMAGE_COUNT;
1937 goto out_wimlib_free;
1941 /* Get image index (this may use the XML data that was just read to
1942 * resolve an image name). */
1943 if (image_num_or_name) {
1944 image = wimlib_resolve_image(pwm, image_num_or_name);
1945 if (image == WIMLIB_NO_IMAGE) {
1946 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1948 ret = WIMLIB_ERR_INVALID_IMAGE;
1949 goto out_wimlib_free;
1950 } else if (image == WIMLIB_ALL_IMAGES) {
1951 ERROR("Applying all images from a pipe is not supported!");
1952 ret = WIMLIB_ERR_INVALID_IMAGE;
1953 goto out_wimlib_free;
1956 if (pwm->hdr.image_count != 1) {
1957 ERROR("No image was specified, but the pipable WIM "
1958 "did not contain exactly 1 image");
1959 ret = WIMLIB_ERR_INVALID_IMAGE;
1960 goto out_wimlib_free;
1965 /* Load the needed metadata resource. */
1966 for (i = 1; i <= pwm->hdr.image_count; i++) {
1967 ret = handle_pwm_metadata_resource(pwm, i, i == image);
1969 goto out_wimlib_free;
1971 /* Extract the image. */
1972 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1973 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1974 /* Clean up and return. */
1982 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1983 const tchar *target, int extract_flags)
1985 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1994 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1997 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1998 return WIMLIB_ERR_INVALID_PARAM;
1999 return do_wimlib_extract_image(wim, image, target, extract_flags);