4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012-2017 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/xattr.h"
67 #include "wimlib/xml.h"
69 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
72 /* Keep in sync with wimlib.h */
73 #define WIMLIB_EXTRACT_MASK_PUBLIC \
74 (WIMLIB_EXTRACT_FLAG_NTFS | \
75 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
76 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
77 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
78 WIMLIB_EXTRACT_FLAG_RPFIX | \
79 WIMLIB_EXTRACT_FLAG_NORPFIX | \
80 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
81 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
82 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
83 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
84 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
85 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
86 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
87 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
88 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
89 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
90 WIMLIB_EXTRACT_FLAG_WIMBOOT | \
91 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | \
92 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | \
93 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | \
94 WIMLIB_EXTRACT_FLAG_COMPACT_LZX \
97 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
98 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
100 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
102 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
103 return extract_progress(ctx, msg);
107 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
109 ctx->progress.extract.current_file_count = 0;
110 ctx->progress.extract.end_file_count = end_file_count;
111 return do_file_extract_progress(ctx, msg);
115 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
117 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
121 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
123 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
127 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
129 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
130 return do_file_extract_progress(ctx, msg);
134 end_file_structure_phase(struct apply_ctx *ctx)
136 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
140 end_file_metadata_phase(struct apply_ctx *ctx)
142 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
145 /* Are all bytes in the specified buffer zero? */
147 is_all_zeroes(const u8 *p, const size_t size)
149 const u8 * const end = p + size;
151 for (; (uintptr_t)p % WORDBYTES && p != end; p++)
155 for (; end - p >= WORDBYTES; p += WORDBYTES)
156 if (*(const machine_word_t *)p)
159 for (; p != end; p++)
167 * Sparse regions should be detected at the granularity of the filesystem block
168 * size. For now just assume 4096 bytes, which is the default block size on
169 * NTFS and most Linux filesystems.
171 #define SPARSE_UNIT 4096
174 * Detect whether the specified buffer begins with a region of all zero bytes.
175 * Return %true if a zero region was found or %false if a nonzero region was
176 * found, and sets *len_ret to the length of the region. This operates at a
177 * granularity of SPARSE_UNIT bytes, meaning that to extend a zero region, there
178 * must be SPARSE_UNIT zero bytes with no interruption, but to extend a nonzero
179 * region, just one nonzero byte in the next SPARSE_UNIT bytes is sufficient.
181 * Note: besides compression, the WIM format doesn't yet have a way to
182 * efficiently represent zero regions, so that's why we need to detect them
183 * ourselves. Things will still fall apart badly on extremely large sparse
184 * files, but this is a start...
187 detect_sparse_region(const void *data, size_t size, size_t *len_ret)
189 const void *p = data;
190 const void * const end = data + size;
195 size_t n = min(end - p, SPARSE_UNIT);
196 bool z = is_all_zeroes(p, n);
198 if (len != 0 && z != zeroes)
209 #define PWM_FOUND_WIM_HDR (-1)
211 /* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
212 * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */
214 read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
215 struct wim_reshdr *reshdr_ret,
216 struct wim_header_disk *pwm_hdr_ret)
219 struct pwm_blob_hdr blob_hdr;
222 ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
226 magic = le64_to_cpu(blob_hdr.magic);
228 if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
229 memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
230 ret = full_read(&pwm->in_fd,
231 (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
232 sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
235 return PWM_FOUND_WIM_HDR;
238 if (unlikely(magic != PWM_BLOB_MAGIC)) {
239 ERROR("Data read on pipe is invalid (expected blob header)");
240 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
243 copy_hash(hash_ret, blob_hdr.hash);
245 reshdr_ret->size_in_wim = 0; /* Not available */
246 reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
247 reshdr_ret->offset_in_wim = pwm->in_fd.offset;
248 reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
250 if (unlikely(reshdr_ret->uncompressed_size == 0)) {
251 ERROR("Data read on pipe is invalid (resource is of 0 size)");
252 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
258 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
259 ERROR("The pipe ended before all needed data was sent!");
261 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
266 read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
269 u8 hash[SHA1_HASH_SIZE];
270 struct wim_reshdr reshdr;
271 struct wim_header_disk pwm_hdr;
272 struct wim_resource_descriptor rdesc;
273 struct blob_descriptor *blob;
275 copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
276 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
277 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
278 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
282 while (ctx->num_blobs_remaining) {
284 ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
286 if (ret == PWM_FOUND_WIM_HDR) {
287 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
288 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
290 if (part_number == ctx->progress.extract.part_number &&
291 total_parts == ctx->progress.extract.total_parts &&
292 guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
295 copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
296 ctx->progress.extract.part_number = part_number;
297 ctx->progress.extract.total_parts = total_parts;
298 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
308 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
309 && (blob = lookup_blob(ctx->wim->blob_table, hash))
310 && (blob->out_refcnt))
312 wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
313 ret = read_blob_with_sha1(blob, cbs);
314 blob_unset_is_located_in_wim_resource(blob);
317 ctx->num_blobs_remaining--;
319 wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
320 ret = skip_wim_resource(&rdesc);
330 handle_pwm_metadata_resource(WIMStruct *pwm, int image, bool is_needed)
332 struct blob_descriptor *blob;
333 struct wim_reshdr reshdr;
334 struct wim_resource_descriptor *rdesc;
337 ret = WIMLIB_ERR_NOMEM;
338 blob = new_blob_descriptor();
342 ret = read_pwm_blob_header(pwm, blob->hash, &reshdr, NULL);
346 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
347 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
348 ERROR("Expected metadata resource, but found non-metadata "
353 ret = WIMLIB_ERR_NOMEM;
354 rdesc = MALLOC(sizeof(*rdesc));
358 wim_reshdr_to_desc_and_blob(&reshdr, pwm, rdesc, blob);
361 ret = WIMLIB_ERR_NOMEM;
362 pwm->image_metadata[image - 1] = new_unloaded_image_metadata(blob);
363 if (!pwm->image_metadata[image - 1])
367 /* If the metadata resource is for the image being extracted, then parse
368 * it and save the metadata in memory. Otherwise, skip over it. */
370 ret = select_wim_image(pwm, image);
372 ret = skip_wim_resource(rdesc);
374 free_blob_descriptor(blob);
378 /* Creates a temporary file opened for writing. The open file descriptor is
379 * returned in @fd_ret and its name is returned in @name_ret (dynamically
382 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
389 name = _wtempnam(NULL, L"wimlib");
391 ERROR_WITH_ERRNO("Failed to create temporary filename");
392 return WIMLIB_ERR_NOMEM;
394 raw_fd = _wopen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY |
395 _O_SHORT_LIVED, 0600);
396 if (raw_fd < 0 && errno == EEXIST) {
400 #else /* __WIN32__ */
401 const char *tmpdir = getenv("TMPDIR");
404 name = MALLOC(strlen(tmpdir) + 1 + 6 + 6 + 1);
406 return WIMLIB_ERR_NOMEM;
407 sprintf(name, "%s/wimlibXXXXXX", tmpdir);
408 raw_fd = mkstemp(name);
409 #endif /* !__WIN32__ */
412 ERROR_WITH_ERRNO("Failed to create temporary file "
415 return WIMLIB_ERR_OPEN;
418 filedes_init(fd_ret, raw_fd);
424 begin_extract_blob(struct blob_descriptor *blob, void *_ctx)
426 struct apply_ctx *ctx = _ctx;
428 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
429 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
431 return call_begin_blob(blob, ctx->saved_cbs);
435 extract_chunk(const struct blob_descriptor *blob, u64 offset,
436 const void *chunk, size_t size, void *_ctx)
438 struct apply_ctx *ctx = _ctx;
439 union wimlib_progress_info *progress = &ctx->progress;
440 bool last = (offset + size == blob->size);
443 if (likely(ctx->supported_features.hard_links)) {
444 progress->extract.completed_bytes +=
445 (u64)size * blob->out_refcnt;
447 progress->extract.completed_streams += blob->out_refcnt;
449 const struct blob_extraction_target *targets =
450 blob_extraction_targets(blob);
451 for (u32 i = 0; i < blob->out_refcnt; i++) {
452 const struct wim_inode *inode = targets[i].inode;
453 const struct wim_dentry *dentry;
455 inode_for_each_extraction_alias(dentry, inode) {
456 progress->extract.completed_bytes += size;
458 progress->extract.completed_streams++;
462 if (progress->extract.completed_bytes >= ctx->next_progress) {
464 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
468 set_next_progress(progress->extract.completed_bytes,
469 progress->extract.total_bytes,
470 &ctx->next_progress);
473 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
474 /* Just extracting to temporary file for now. */
475 ret = full_write(&ctx->tmpfile_fd, chunk, size);
477 ERROR_WITH_ERRNO("Error writing data to "
478 "temporary file \"%"TS"\"",
484 return call_continue_blob(blob, offset, chunk, size, ctx->saved_cbs);
487 /* Copy the blob's data from the temporary file to each of its targets.
489 * This is executed only in the very uncommon case that a blob is being
490 * extracted to more than MAX_OPEN_FILES targets! */
492 extract_from_tmpfile(const tchar *tmpfile_name,
493 const struct blob_descriptor *orig_blob,
494 const struct read_blob_callbacks *cbs)
496 struct blob_descriptor tmpfile_blob;
497 const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
500 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
501 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
502 tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
503 tmpfile_blob.out_refcnt = 1;
505 for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
506 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
507 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
515 end_extract_blob(struct blob_descriptor *blob, int status, void *_ctx)
517 struct apply_ctx *ctx = _ctx;
519 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
520 filedes_close(&ctx->tmpfile_fd);
522 status = extract_from_tmpfile(ctx->tmpfile_name, blob,
524 filedes_invalidate(&ctx->tmpfile_fd);
525 tunlink(ctx->tmpfile_name);
526 FREE(ctx->tmpfile_name);
530 return call_end_blob(blob, status, ctx->saved_cbs);
534 * Read the list of blobs to extract and feed their data into the specified
535 * callback functions.
537 * This handles checksumming each blob.
539 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
541 * This also works if the WIM is being read from a pipe.
543 * This also will split up blobs that will need to be extracted to more than
544 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
545 * Therefore, the apply_operations implementation need not worry about running
546 * out of file descriptors, unless it might open more than one file descriptor
547 * per 'blob_extraction_target' (e.g. Win32 currently might because the
548 * destination file system might not support hard links).
551 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
553 struct read_blob_callbacks wrapper_cbs = {
554 .begin_blob = begin_extract_blob,
555 .continue_blob = extract_chunk,
556 .end_blob = end_extract_blob,
559 ctx->saved_cbs = cbs;
560 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
561 return read_blobs_from_pipe(ctx, &wrapper_cbs);
563 return read_blob_list(&ctx->blob_list,
564 offsetof(struct blob_descriptor,
566 &wrapper_cbs, VERIFY_BLOB_HASHES);
570 /* Extract a WIM dentry to standard output.
572 * This obviously doesn't make sense in all cases. We return an error if the
573 * dentry does not correspond to a regular file. Otherwise we extract the
574 * unnamed data stream only. */
576 extract_dentry_to_stdout(struct wim_dentry *dentry,
577 const struct blob_table *blob_table)
579 struct wim_inode *inode = dentry->d_inode;
580 struct blob_descriptor *blob;
581 struct filedes _stdout;
583 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
584 FILE_ATTRIBUTE_DIRECTORY |
585 FILE_ATTRIBUTE_ENCRYPTED))
587 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
588 "extracted to standard output", dentry_full_path(dentry));
589 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
592 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
594 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
595 if (!is_zero_hash(hash))
596 return blob_not_found_error(inode, hash);
600 filedes_init(&_stdout, STDOUT_FILENO);
601 return extract_blob_to_fd(blob, &_stdout);
605 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
606 const struct blob_table *blob_table)
608 for (size_t i = 0; i < num_dentries; i++) {
609 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
616 /**********************************************************************/
619 * Removes duplicate dentries from the array.
621 * Returns the new number of dentries, packed at the front of the array.
624 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
627 for (i = 0; i < num_trees; i++) {
628 if (!trees[i]->d_tmp_flag) {
629 /* Found distinct dentry. */
630 trees[i]->d_tmp_flag = 1;
631 trees[j++] = trees[i];
634 for (i = 0; i < j; i++)
635 trees[i]->d_tmp_flag = 0;
640 * Remove dentries that are descendants of other dentries in the array.
642 * Returns the new number of dentries, packed at the front of the array.
645 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
648 for (i = 0; i < num_trees; i++)
649 trees[i]->d_tmp_flag = 1;
650 for (i = 0; i < num_trees; i++) {
651 struct wim_dentry *d = trees[i];
652 while (!dentry_is_root(d)) {
657 trees[j++] = trees[i];
661 trees[i]->d_tmp_flag = 0;
664 for (i = 0; i < j; i++)
665 trees[i]->d_tmp_flag = 0;
670 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
672 struct list_head *dentry_list = _dentry_list;
673 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
678 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
680 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
684 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
686 if (will_extract_dentry(dentry)) {
687 list_del(&dentry->d_extraction_list_node);
688 dentry_reset_extraction_list_node(dentry);
694 * Build the preliminary list of dentries to be extracted.
696 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
697 * an ancestor of d2, then d1 appears before d2 in the list.
700 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
701 size_t num_trees, bool add_ancestors)
703 INIT_LIST_HEAD(dentry_list);
705 /* Add the trees recursively. */
706 for (size_t i = 0; i < num_trees; i++)
707 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
709 /* If requested, add ancestors of the trees. */
711 for (size_t i = 0; i < num_trees; i++) {
712 struct wim_dentry *dentry = trees[i];
713 struct wim_dentry *ancestor;
714 struct list_head *place_after;
716 if (dentry_is_root(dentry))
719 place_after = dentry_list;
722 ancestor = ancestor->d_parent;
723 if (will_extract_dentry(ancestor)) {
724 place_after = &ancestor->d_extraction_list_node;
727 } while (!dentry_is_root(ancestor));
731 ancestor = ancestor->d_parent;
732 if (will_extract_dentry(ancestor))
734 list_add(&ancestor->d_extraction_list_node, place_after);
735 } while (!dentry_is_root(ancestor));
741 destroy_dentry_list(struct list_head *dentry_list)
743 struct wim_dentry *dentry, *tmp;
744 struct wim_inode *inode;
746 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
747 inode = dentry->d_inode;
748 dentry_reset_extraction_list_node(dentry);
749 inode->i_visited = 0;
750 inode->i_can_externally_back = 0;
751 if ((void *)dentry->d_extraction_name != (void *)dentry->d_name)
752 FREE(dentry->d_extraction_name);
753 dentry->d_extraction_name = NULL;
754 dentry->d_extraction_name_nchars = 0;
759 destroy_blob_list(struct list_head *blob_list)
761 struct blob_descriptor *blob;
763 list_for_each_entry(blob, blob_list, extraction_list)
764 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
765 FREE(blob->blob_extraction_targets);
769 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
771 static const utf16lechar replacement_char = cpu_to_le16('?');
775 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
781 for (i = 0; i < num_chars; i++) {
782 switch (le16_to_cpu(name[i])) {
784 case '\x01'...'\x1F':
797 name[i] = replacement_char;
807 dentry_calculate_extraction_name(struct wim_dentry *dentry,
808 struct apply_ctx *ctx)
812 if (dentry_is_root(dentry))
816 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
817 dentry->d_extraction_name = dentry->d_name;
818 dentry->d_extraction_name_nchars = dentry->d_name_nbytes /
824 if (!ctx->supported_features.case_sensitive_filenames) {
825 struct wim_dentry *other;
826 dentry_for_each_ci_match(other, dentry) {
827 if (will_extract_dentry(other)) {
828 if (ctx->extract_flags &
829 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
830 WARNING("\"%"TS"\" has the same "
831 "case-insensitive name as "
832 "\"%"TS"\"; extracting "
833 "dummy name instead",
834 dentry_full_path(dentry),
835 dentry_full_path(other));
838 WARNING("Not extracting \"%"TS"\": "
839 "has same case-insensitive "
841 dentry_full_path(dentry),
842 dentry_full_path(other));
849 if (file_name_valid(dentry->d_name, dentry->d_name_nbytes / 2, false)) {
851 ret = utf16le_get_tstr(dentry->d_name,
852 dentry->d_name_nbytes,
853 (const tchar **)&dentry->d_extraction_name,
855 dentry->d_extraction_name_nchars = nbytes / sizeof(tchar);
858 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
860 WARNING("\"%"TS"\" has an invalid filename "
861 "that is not supported on this platform; "
862 "extracting dummy name instead",
863 dentry_full_path(dentry));
866 WARNING("Not extracting \"%"TS"\": has an invalid filename "
867 "that is not supported on this platform",
868 dentry_full_path(dentry));
875 utf16lechar utf16_name_copy[dentry->d_name_nbytes / 2];
877 memcpy(utf16_name_copy, dentry->d_name, dentry->d_name_nbytes);
878 file_name_valid(utf16_name_copy, dentry->d_name_nbytes / 2, true);
880 const tchar *tchar_name;
883 ret = utf16le_get_tstr(utf16_name_copy,
884 dentry->d_name_nbytes,
885 &tchar_name, &tchar_nchars);
889 tchar_nchars /= sizeof(tchar);
891 size_t fixed_name_num_chars = tchar_nchars;
892 tchar fixed_name[tchar_nchars + 50];
894 tmemcpy(fixed_name, tchar_name, tchar_nchars);
895 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
896 T(" (invalid filename #%lu)"),
897 ++ctx->invalid_sequence);
899 utf16le_put_tstr(tchar_name);
901 dentry->d_extraction_name = TSTRDUP(fixed_name);
902 if (!dentry->d_extraction_name)
903 return WIMLIB_ERR_NOMEM;
904 dentry->d_extraction_name_nchars = fixed_name_num_chars;
909 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
914 * Calculate the actual filename component at which each WIM dentry will be
915 * extracted, with special handling for dentries that are unsupported by the
916 * extraction backend or have invalid names.
918 * ctx->supported_features must be filled in.
920 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
923 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
924 struct apply_ctx *ctx)
926 struct list_head *prev, *cur;
928 /* Can't use list_for_each_entry() because a call to
929 * dentry_calculate_extraction_name() may delete the current dentry and
930 * its children from the list. */
934 struct wim_dentry *dentry;
938 if (cur == dentry_list)
941 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
943 ret = dentry_calculate_extraction_name(dentry, ctx);
947 if (prev->next == cur)
950 ; /* Current dentry and its children (which follow in
951 the list) were deleted. prev stays the same. */
957 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
958 struct blob_table *blob_table)
960 struct wim_inode *inode = dentry->d_inode;
961 struct blob_descriptor *blob;
965 /* Special case: when extracting from a pipe, the WIM blob table is
966 * initially empty, so "resolving" an inode's streams is initially not
967 * possible. However, we still need to keep track of which blobs,
968 * identified by SHA-1 message digests, need to be extracted, so we
969 * "resolve" the inode's streams anyway by allocating a 'struct
970 * blob_descriptor' for each one. */
971 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
973 ret = inode_resolve_streams(inode, blob_table, force);
976 for (unsigned i = 0; i < inode->i_num_streams; i++) {
977 blob = stream_blob_resolved(&inode->i_streams[i]);
979 blob->out_refcnt = 0;
985 * For each dentry to be extracted, resolve all streams in the corresponding
986 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
988 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
991 dentry_list_resolve_streams(struct list_head *dentry_list,
992 struct apply_ctx *ctx)
994 struct wim_dentry *dentry;
997 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
998 ret = dentry_resolve_streams(dentry,
1000 ctx->wim->blob_table);
1008 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
1009 struct apply_ctx *ctx)
1011 struct wim_inode *inode = dentry->d_inode;
1012 struct blob_descriptor *blob = stream_blob_resolved(strm);
1013 struct blob_extraction_target *targets;
1018 /* Tally the size only for each actual extraction of the stream (not
1019 * additional hard links to the inode). */
1020 if (inode->i_visited && ctx->supported_features.hard_links)
1023 ctx->progress.extract.total_bytes += blob->size;
1024 ctx->progress.extract.total_streams++;
1026 if (inode->i_visited)
1029 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1030 * many extraction targets it will have. */
1031 if (blob->out_refcnt == 0) {
1032 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1033 ctx->num_blobs_remaining++;
1036 /* Set this stream as an extraction target of 'blob'. */
1038 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1039 targets = blob->inline_blob_extraction_targets;
1041 struct blob_extraction_target *prev_targets;
1042 size_t alloc_blob_extraction_targets;
1044 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1045 prev_targets = NULL;
1046 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1048 prev_targets = blob->blob_extraction_targets;
1049 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1052 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1053 alloc_blob_extraction_targets *= 2;
1054 targets = REALLOC(prev_targets,
1055 alloc_blob_extraction_targets *
1056 sizeof(targets[0]));
1058 return WIMLIB_ERR_NOMEM;
1059 if (!prev_targets) {
1061 blob->inline_blob_extraction_targets,
1062 sizeof(blob->inline_blob_extraction_targets));
1064 blob->blob_extraction_targets = targets;
1065 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1067 targets = blob->blob_extraction_targets;
1069 targets[blob->out_refcnt].inode = inode;
1070 targets[blob->out_refcnt].stream = strm;
1076 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1077 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1079 bool need_stream = false;
1080 switch (strm->stream_type) {
1081 case STREAM_TYPE_DATA:
1082 if (stream_is_named(strm)) {
1083 /* Named data stream */
1084 if (ctx->supported_features.named_data_streams)
1086 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1087 FILE_ATTRIBUTE_ENCRYPTED))
1088 && !(inode_is_symlink(inode)
1089 && !ctx->supported_features.reparse_points
1090 && ctx->supported_features.symlink_reparse_points))
1093 * Unnamed data stream. Skip if any of the following is true:
1095 * - file is a directory
1096 * - file is encrypted
1097 * - backend needs to create the file as UNIX symlink
1098 * - backend will extract the stream as externally
1099 * backed from the WIM archive itself
1101 if (ctx->apply_ops->will_back_from_wim) {
1102 int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx);
1103 if (ret > 0) /* Error? */
1105 if (ret < 0) /* Won't externally back? */
1112 case STREAM_TYPE_REPARSE_POINT:
1113 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1114 if (ctx->supported_features.reparse_points ||
1115 (inode_is_symlink(inode) &&
1116 ctx->supported_features.symlink_reparse_points))
1119 case STREAM_TYPE_EFSRPC_RAW_DATA:
1120 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1121 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1122 if (ctx->supported_features.encrypted_directories)
1125 if (ctx->supported_features.encrypted_files)
1131 return ref_stream(strm, dentry, ctx);
1136 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1138 struct wim_inode *inode = dentry->d_inode;
1139 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1140 int ret = ref_stream_if_needed(dentry, inode,
1141 &inode->i_streams[i], ctx);
1145 inode->i_visited = 1;
1150 * Given a list of dentries to be extracted, build the list of blobs that need
1151 * to be extracted, and for each blob determine the streams to which that blob
1152 * will be extracted.
1154 * This also initializes the extract progress info with byte and blob
1157 * ctx->supported_features must be filled in.
1160 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1162 struct wim_dentry *dentry;
1165 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1166 ret = dentry_ref_streams(dentry, ctx);
1170 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1171 dentry->d_inode->i_visited = 0;
1176 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1178 struct wim_dentry *dentry;
1180 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1181 dentry->d_inode->i_first_extraction_alias = NULL;
1183 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1184 dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias;
1185 dentry->d_inode->i_first_extraction_alias = dentry;
1190 inode_tally_features(const struct wim_inode *inode,
1191 struct wim_features *features)
1193 if (inode->i_attributes & FILE_ATTRIBUTE_READONLY)
1194 features->readonly_files++;
1195 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1196 features->hidden_files++;
1197 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1198 features->system_files++;
1199 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1200 features->archive_files++;
1201 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1202 features->compressed_files++;
1203 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1204 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1205 features->encrypted_directories++;
1207 features->encrypted_files++;
1209 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1210 features->not_context_indexed_files++;
1211 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1212 features->sparse_files++;
1213 if (inode_has_named_data_stream(inode))
1214 features->named_data_streams++;
1215 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1216 features->reparse_points++;
1217 if (inode_is_symlink(inode))
1218 features->symlink_reparse_points++;
1220 features->other_reparse_points++;
1222 if (inode_has_security_descriptor(inode))
1223 features->security_descriptors++;
1224 if (inode_has_unix_data(inode))
1225 features->unix_data++;
1226 if (inode_has_object_id(inode))
1227 features->object_ids++;
1228 if (inode_has_linux_xattrs(inode))
1229 features->linux_xattrs++;
1232 /* Tally features necessary to extract a dentry and the corresponding inode. */
1234 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1236 struct wim_inode *inode = dentry->d_inode;
1238 if (dentry_has_short_name(dentry))
1239 features->short_names++;
1241 if (inode->i_visited) {
1242 features->hard_links++;
1244 inode_tally_features(inode, features);
1245 inode->i_visited = 1;
1249 /* Tally the features necessary to extract the specified dentries. */
1251 dentry_list_get_features(struct list_head *dentry_list,
1252 struct wim_features *features)
1254 struct wim_dentry *dentry;
1256 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1257 dentry_tally_features(dentry, features);
1259 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1260 dentry->d_inode->i_visited = 0;
1264 do_feature_check(const struct wim_features *required_features,
1265 const struct wim_features *supported_features,
1268 /* Encrypted files. */
1269 if (required_features->encrypted_files &&
1270 !supported_features->encrypted_files)
1271 WARNING("Ignoring EFS-encrypted data of %lu files",
1272 required_features->encrypted_files);
1274 /* Named data streams. */
1275 if (required_features->named_data_streams &&
1276 !supported_features->named_data_streams)
1277 WARNING("Ignoring named data streams of %lu files",
1278 required_features->named_data_streams);
1280 /* File attributes. */
1281 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1283 if (required_features->readonly_files &&
1284 !supported_features->readonly_files)
1285 WARNING("Ignoring FILE_ATTRIBUTE_READONLY of %lu files",
1286 required_features->readonly_files);
1288 if (required_features->hidden_files &&
1289 !supported_features->hidden_files)
1290 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1291 required_features->hidden_files);
1293 if (required_features->system_files &&
1294 !supported_features->system_files)
1295 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1296 required_features->system_files);
1298 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1299 * We're an archive program, so theoretically we can do what we
1302 if (required_features->compressed_files &&
1303 !supported_features->compressed_files)
1304 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1305 required_features->compressed_files);
1307 if (required_features->not_context_indexed_files &&
1308 !supported_features->not_context_indexed_files)
1309 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1310 required_features->not_context_indexed_files);
1312 if (required_features->sparse_files &&
1313 !supported_features->sparse_files)
1314 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1315 required_features->sparse_files);
1317 if (required_features->encrypted_directories &&
1318 !supported_features->encrypted_directories)
1319 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1320 required_features->encrypted_directories);
1324 if (required_features->hard_links && !supported_features->hard_links)
1325 WARNING("Extracting %lu hard links as independent files",
1326 required_features->hard_links);
1328 /* Symbolic links and reparse points. */
1329 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1330 required_features->symlink_reparse_points &&
1331 !supported_features->symlink_reparse_points &&
1332 !supported_features->reparse_points)
1334 ERROR("Extraction backend does not support symbolic links!");
1335 return WIMLIB_ERR_UNSUPPORTED;
1337 if (required_features->reparse_points &&
1338 !supported_features->reparse_points)
1340 if (supported_features->symlink_reparse_points) {
1341 if (required_features->other_reparse_points) {
1342 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1343 required_features->other_reparse_points);
1346 WARNING("Ignoring reparse data of %lu files",
1347 required_features->reparse_points);
1351 /* Security descriptors. */
1352 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1353 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1354 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1355 required_features->security_descriptors &&
1356 !supported_features->security_descriptors)
1358 ERROR("Extraction backend does not support security descriptors!");
1359 return WIMLIB_ERR_UNSUPPORTED;
1361 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1362 required_features->security_descriptors &&
1363 !supported_features->security_descriptors)
1364 WARNING("Ignoring Windows NT security descriptors of %lu files",
1365 required_features->security_descriptors);
1367 /* Standard UNIX metadata */
1368 if (required_features->unix_data &&
1369 (!supported_features->unix_data ||
1370 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA)))
1372 if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
1373 ERROR("Requested UNIX metadata extraction, but "
1374 "extraction backend does not support it!");
1375 return WIMLIB_ERR_UNSUPPORTED;
1377 WARNING("Ignoring UNIX metadata (uid/gid/mode/rdev) of %lu files%"TS,
1378 required_features->unix_data,
1379 (supported_features->unix_data ?
1380 T("\n (use --unix-data mode to extract these)") : T("")));
1383 /* Linux-style extended attributes */
1384 if (required_features->linux_xattrs &&
1385 (!supported_features->linux_xattrs ||
1386 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA)))
1388 WARNING("Ignoring Linux-style extended attributes of %lu files%"TS,
1389 required_features->linux_xattrs,
1390 (supported_features->linux_xattrs ?
1391 T("\n (use --unix-data mode to extract these)") : T("")));
1395 if (required_features->object_ids && !supported_features->object_ids) {
1396 WARNING("Ignoring object IDs of %lu files",
1397 required_features->object_ids);
1401 if (required_features->short_names &&
1402 !supported_features->short_names)
1404 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1405 ERROR("Extraction backend does not support DOS names!");
1406 return WIMLIB_ERR_UNSUPPORTED;
1408 WARNING("Ignoring DOS names of %lu files",
1409 required_features->short_names);
1413 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1414 !supported_features->timestamps)
1416 ERROR("Extraction backend does not support timestamps!");
1417 return WIMLIB_ERR_UNSUPPORTED;
1423 static const struct apply_operations *
1424 select_apply_operations(int extract_flags)
1427 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1428 return &ntfs_3g_apply_ops;
1431 return &win32_apply_ops;
1433 return &unix_apply_ops;
1438 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1439 const tchar *target, int extract_flags)
1441 const struct apply_operations *ops;
1442 struct apply_ctx *ctx;
1444 LIST_HEAD(dentry_list);
1446 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1447 ret = extract_dentries_to_stdout(trees, num_trees,
1452 num_trees = remove_duplicate_trees(trees, num_trees);
1453 num_trees = remove_contained_trees(trees, num_trees);
1455 ops = select_apply_operations(extract_flags);
1457 if (num_trees > 1 && ops->single_tree_only) {
1458 ERROR("Extracting multiple directory trees "
1459 "at once is not supported in %s extraction mode!",
1461 ret = WIMLIB_ERR_UNSUPPORTED;
1465 ctx = CALLOC(1, ops->context_size);
1467 ret = WIMLIB_ERR_NOMEM;
1472 ctx->target = target;
1473 ctx->target_nchars = tstrlen(target);
1474 ctx->extract_flags = extract_flags;
1475 if (ctx->wim->progfunc) {
1476 ctx->progfunc = ctx->wim->progfunc;
1477 ctx->progctx = ctx->wim->progctx;
1478 ctx->progress.extract.image = wim->current_image;
1479 ctx->progress.extract.extract_flags = (extract_flags &
1480 WIMLIB_EXTRACT_MASK_PUBLIC);
1481 ctx->progress.extract.wimfile_name = wim->filename;
1482 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1483 wim->current_image);
1484 ctx->progress.extract.target = target;
1486 INIT_LIST_HEAD(&ctx->blob_list);
1487 filedes_invalidate(&ctx->tmpfile_fd);
1488 ctx->apply_ops = ops;
1490 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1494 build_dentry_list(&dentry_list, trees, num_trees,
1496 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1498 dentry_list_get_features(&dentry_list, &ctx->required_features);
1500 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1501 ctx->extract_flags);
1505 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1509 if (unlikely(list_empty(&dentry_list))) {
1510 WARNING("There is nothing to extract!");
1514 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1518 dentry_list_build_inode_alias_lists(&dentry_list);
1520 ret = dentry_list_ref_streams(&dentry_list, ctx);
1524 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1525 /* When extracting from a pipe, the number of bytes of data to
1526 * extract can't be determined in the normal way (examining the
1527 * blob table), since at this point all we have is a set of
1528 * SHA-1 message digests of blobs that need to be extracted.
1529 * However, we can get a reasonably accurate estimate by taking
1530 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1531 * data. This does assume that a full image is being extracted,
1532 * but currently there is no API for doing otherwise. (Also,
1533 * subtract <HARDLINKBYTES> from this if hard links are
1534 * supported by the extraction mode.) */
1535 ctx->progress.extract.total_bytes =
1536 xml_get_image_total_bytes(wim->xml_info,
1537 wim->current_image);
1538 if (ctx->supported_features.hard_links) {
1539 ctx->progress.extract.total_bytes -=
1540 xml_get_image_hard_link_bytes(wim->xml_info,
1541 wim->current_image);
1545 ret = extract_progress(ctx,
1546 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1547 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1548 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1552 ret = (*ops->extract)(&dentry_list, ctx);
1556 if (ctx->progress.extract.completed_bytes <
1557 ctx->progress.extract.total_bytes)
1559 ctx->progress.extract.completed_bytes =
1560 ctx->progress.extract.total_bytes;
1561 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1566 ret = extract_progress(ctx,
1567 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1568 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1569 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1571 destroy_blob_list(&ctx->blob_list);
1572 destroy_dentry_list(&dentry_list);
1579 mkdir_if_needed(const tchar *target)
1581 if (!tmkdir(target, 0755))
1584 if (errno == EEXIST)
1588 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1589 if (errno == EACCES)
1593 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1594 return WIMLIB_ERR_MKDIR;
1597 /* Make sure the extraction flags make sense, and update them if needed. */
1599 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1601 int extract_flags = *extract_flags_p;
1603 /* Check for invalid flag combinations */
1605 if ((extract_flags &
1606 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1607 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1608 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1609 return WIMLIB_ERR_INVALID_PARAM;
1611 if ((extract_flags &
1612 (WIMLIB_EXTRACT_FLAG_RPFIX |
1613 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1614 WIMLIB_EXTRACT_FLAG_NORPFIX))
1615 return WIMLIB_ERR_INVALID_PARAM;
1617 #ifndef WITH_NTFS_3G
1618 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1619 ERROR("wimlib was compiled without support for NTFS-3G, so\n"
1620 " it cannot apply a WIM image directly to an NTFS volume.");
1621 return WIMLIB_ERR_UNSUPPORTED;
1625 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1628 return WIMLIB_ERR_NO_FILENAME;
1630 ERROR("WIMBoot extraction is only supported on Windows!");
1631 return WIMLIB_ERR_UNSUPPORTED;
1635 if (extract_flags & (WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K |
1636 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K |
1637 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K |
1638 WIMLIB_EXTRACT_FLAG_COMPACT_LZX))
1642 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K) != 0);
1643 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K) != 0);
1644 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K) != 0);
1645 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_LZX) != 0);
1647 ERROR("Only one compression format can be specified "
1648 "for compact-mode extraction!");
1649 return WIMLIB_ERR_INVALID_PARAM;
1651 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1652 ERROR("Compact-mode extraction and WIMBoot-mode "
1653 "extraction are mutually exclusive!");
1654 return WIMLIB_ERR_INVALID_PARAM;
1657 ERROR("Compact-mode extraction (System Compression) "
1658 "is only supported on Windows!");
1659 return WIMLIB_ERR_UNSUPPORTED;
1664 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1665 WIMLIB_EXTRACT_FLAG_NORPFIX |
1666 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1667 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1669 /* For full-image extraction, do reparse point fixups by default
1670 * if the WIM header says they are enabled. */
1671 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1672 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1675 *extract_flags_p = extract_flags;
1679 struct append_dentry_ctx {
1680 struct wim_dentry **dentries;
1681 size_t num_dentries;
1682 size_t num_alloc_dentries;
1686 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1688 struct append_dentry_ctx *ctx = _ctx;
1690 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1691 struct wim_dentry **new_dentries;
1694 new_length = max(ctx->num_alloc_dentries + 8,
1695 ctx->num_alloc_dentries * 3 / 2);
1696 new_dentries = REALLOC(ctx->dentries,
1697 new_length * sizeof(ctx->dentries[0]));
1698 if (new_dentries == NULL)
1699 return WIMLIB_ERR_NOMEM;
1700 ctx->dentries = new_dentries;
1701 ctx->num_alloc_dentries = new_length;
1703 ctx->dentries[ctx->num_dentries++] = dentry;
1707 /* Append dentries matched by a path which can contain wildcard characters. */
1709 append_matched_dentries(WIMStruct *wim, const tchar *orig_pattern,
1710 int extract_flags, struct append_dentry_ctx *ctx)
1712 const size_t count_before = ctx->num_dentries;
1716 pattern = canonicalize_wim_path(orig_pattern);
1718 return WIMLIB_ERR_NOMEM;
1719 ret = expand_path_pattern(wim_get_current_root_dentry(wim), pattern,
1720 append_dentry_cb, ctx);
1722 if (ret || ctx->num_dentries > count_before)
1724 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) {
1725 ERROR("No matches for path pattern \"%"TS"\"", orig_pattern);
1726 return WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1728 WARNING("No matches for path pattern \"%"TS"\"", orig_pattern);
1733 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1734 const tchar * const *paths, size_t num_paths,
1738 struct wim_dentry **trees;
1741 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1742 (num_paths != 0 && paths == NULL))
1743 return WIMLIB_ERR_INVALID_PARAM;
1745 ret = check_extract_flags(wim, &extract_flags);
1749 ret = select_wim_image(wim, image);
1753 ret = wim_checksum_unhashed_blobs(wim);
1757 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1758 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1759 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1761 ret = mkdir_if_needed(target);
1766 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1768 struct append_dentry_ctx append_dentry_ctx = {
1771 .num_alloc_dentries = 0,
1774 for (size_t i = 0; i < num_paths; i++) {
1775 ret = append_matched_dentries(wim, paths[i],
1777 &append_dentry_ctx);
1779 trees = append_dentry_ctx.dentries;
1780 goto out_free_trees;
1783 trees = append_dentry_ctx.dentries;
1784 num_trees = append_dentry_ctx.num_dentries;
1786 trees = MALLOC(num_paths * sizeof(trees[0]));
1788 return WIMLIB_ERR_NOMEM;
1790 for (size_t i = 0; i < num_paths; i++) {
1792 tchar *path = canonicalize_wim_path(paths[i]);
1794 ret = WIMLIB_ERR_NOMEM;
1795 goto out_free_trees;
1798 trees[i] = get_dentry(wim, path,
1799 WIMLIB_CASE_PLATFORM_DEFAULT);
1801 if (trees[i] == NULL) {
1802 ERROR("Path \"%"TS"\" does not exist "
1804 paths[i], wim->current_image);
1805 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1806 goto out_free_trees;
1809 num_trees = num_paths;
1812 if (num_trees == 0) {
1814 goto out_free_trees;
1817 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1824 extract_single_image(WIMStruct *wim, int image,
1825 const tchar *target, int extract_flags)
1827 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1828 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1829 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1832 static const tchar * const filename_forbidden_chars =
1839 /* This function checks if it is okay to use a WIM image's name as a directory
1842 image_name_ok_as_dir(const tchar *image_name)
1844 return image_name && *image_name &&
1845 !tstrpbrk(image_name, filename_forbidden_chars) &&
1846 tstrcmp(image_name, T(".")) &&
1847 tstrcmp(image_name, T("..")) &&
1848 tstrlen(image_name) <= 128;
1851 /* Extracts all images from the WIM to the directory @target, with the images
1852 * placed in subdirectories named by their image names. */
1854 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1856 size_t output_path_len = tstrlen(target);
1857 tchar buf[output_path_len + 1 + 128 + 1];
1860 const tchar *image_name;
1862 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1863 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1864 return WIMLIB_ERR_INVALID_PARAM;
1867 ret = mkdir_if_needed(target);
1870 tmemcpy(buf, target, output_path_len);
1871 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1872 for (image = 1; image <= wim->hdr.image_count; image++) {
1873 image_name = wimlib_get_image_name(wim, image);
1874 if (image_name_ok_as_dir(image_name)) {
1875 tstrcpy(buf + output_path_len + 1, image_name);
1877 /* Image name is empty or contains forbidden characters.
1878 * Use image number instead. */
1879 tsprintf(buf + output_path_len + 1, T("%d"), image);
1881 ret = extract_single_image(wim, image, buf, extract_flags);
1889 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1892 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1893 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1894 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1895 return WIMLIB_ERR_INVALID_PARAM;
1897 if (image == WIMLIB_ALL_IMAGES)
1898 return extract_all_images(wim, target, extract_flags);
1900 return extract_single_image(wim, image, target, extract_flags);
1904 /****************************************************************************
1906 ****************************************************************************/
1909 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1910 const tchar * const *paths, size_t num_paths,
1913 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1914 return WIMLIB_ERR_INVALID_PARAM;
1916 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1921 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1922 const tchar *path_list_file, int extract_flags)
1929 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1931 ERROR("Failed to read path list file \"%"TS"\"",
1936 ret = wimlib_extract_paths(wim, image, target,
1937 (const tchar * const *)paths, num_paths,
1945 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1946 const tchar *image_num_or_name,
1947 const tchar *target,
1949 wimlib_progress_func_t progfunc,
1954 struct filedes *in_fd;
1958 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1959 return WIMLIB_ERR_INVALID_PARAM;
1961 /* Read the WIM header from the pipe and get a WIMStruct to represent
1962 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1963 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1964 * empty blob table, no XML data read, and no filename set. */
1965 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1970 /* Sanity check to make sure this is a pipable WIM. */
1971 if (pwm->hdr.magic != PWM_MAGIC) {
1972 ERROR("The WIM being read from file descriptor %d "
1973 "is not pipable!", pipe_fd);
1974 ret = WIMLIB_ERR_NOT_PIPABLE;
1975 goto out_wimlib_free;
1978 /* Sanity check to make sure the first part of a pipable split WIM is
1979 * sent over the pipe first. */
1980 if (pwm->hdr.part_number != 1) {
1981 ERROR("The first part of the split WIM must be "
1982 "sent over the pipe first.");
1983 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1984 goto out_wimlib_free;
1987 in_fd = &pwm->in_fd;
1988 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1990 /* As mentioned, the WIMStruct we created from the pipe does not have
1991 * XML data yet. Fix this by reading the extra copy of the XML data
1992 * that directly follows the header in pipable WIMs. (Note: see
1993 * write_pipable_wim() for more details about the format of pipable
1996 u8 hash[SHA1_HASH_SIZE];
1998 ret = read_pwm_blob_header(pwm, hash,
1999 &pwm->hdr.xml_data_reshdr, NULL);
2001 goto out_wimlib_free;
2003 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
2004 ERROR("Expected XML data, but found non-metadata resource.");
2005 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2006 goto out_wimlib_free;
2009 ret = read_wim_xml_data(pwm);
2011 goto out_wimlib_free;
2013 if (xml_get_image_count(pwm->xml_info) != pwm->hdr.image_count) {
2014 ERROR("Image count in XML data is not the same as in WIM header.");
2015 ret = WIMLIB_ERR_IMAGE_COUNT;
2016 goto out_wimlib_free;
2020 /* Get image index (this may use the XML data that was just read to
2021 * resolve an image name). */
2022 if (image_num_or_name) {
2023 image = wimlib_resolve_image(pwm, image_num_or_name);
2024 if (image == WIMLIB_NO_IMAGE) {
2025 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2027 ret = WIMLIB_ERR_INVALID_IMAGE;
2028 goto out_wimlib_free;
2029 } else if (image == WIMLIB_ALL_IMAGES) {
2030 ERROR("Applying all images from a pipe is not supported!");
2031 ret = WIMLIB_ERR_INVALID_IMAGE;
2032 goto out_wimlib_free;
2035 if (pwm->hdr.image_count != 1) {
2036 ERROR("No image was specified, but the pipable WIM "
2037 "did not contain exactly 1 image");
2038 ret = WIMLIB_ERR_INVALID_IMAGE;
2039 goto out_wimlib_free;
2044 /* Load the needed metadata resource. */
2045 for (i = 1; i <= pwm->hdr.image_count; i++) {
2046 ret = handle_pwm_metadata_resource(pwm, i, i == image);
2048 goto out_wimlib_free;
2050 /* Extract the image. */
2051 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2052 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2053 /* Clean up and return. */
2061 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2062 const tchar *target, int extract_flags)
2064 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2073 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2076 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2077 return WIMLIB_ERR_INVALID_PARAM;
2078 return do_wimlib_extract_image(wim, image, target, extract_flags);