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_wrapper(struct blob_descriptor *blob, void *_ctx)
361 struct apply_ctx *ctx = _ctx;
363 ctx->cur_blob = blob;
364 ctx->cur_blob_offset = 0;
366 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
367 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
369 return call_begin_blob(blob, ctx->saved_cbs);
373 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
375 struct apply_ctx *ctx = _ctx;
376 union wimlib_progress_info *progress = &ctx->progress;
379 ctx->cur_blob_offset += size;
381 if (likely(ctx->supported_features.hard_links)) {
382 progress->extract.completed_bytes +=
383 (u64)size * ctx->cur_blob->out_refcnt;
384 if (ctx->cur_blob_offset == ctx->cur_blob->size)
385 progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
387 const struct blob_extraction_target *targets =
388 blob_extraction_targets(ctx->cur_blob);
389 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
390 const struct wim_inode *inode = targets[i].inode;
391 const struct wim_dentry *dentry;
393 inode_for_each_extraction_alias(dentry, inode) {
394 progress->extract.completed_bytes += size;
395 if (ctx->cur_blob_offset == ctx->cur_blob->size)
396 progress->extract.completed_streams++;
400 if (progress->extract.completed_bytes >= ctx->next_progress) {
402 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
406 set_next_progress(progress->extract.completed_bytes,
407 progress->extract.total_bytes,
408 &ctx->next_progress);
411 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
412 /* Just extracting to temporary file for now. */
413 ret = full_write(&ctx->tmpfile_fd, chunk, size);
415 ERROR_WITH_ERRNO("Error writing data to "
416 "temporary file \"%"TS"\"",
422 return call_consume_chunk(chunk, size, ctx->saved_cbs);
425 /* Copy the blob's data from the temporary file to each of its targets.
427 * This is executed only in the very uncommon case that a blob is being
428 * extracted to more than MAX_OPEN_FILES targets! */
430 extract_from_tmpfile(const tchar *tmpfile_name,
431 const struct blob_descriptor *orig_blob,
432 const struct read_blob_callbacks *cbs)
434 struct blob_descriptor tmpfile_blob;
435 const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
438 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
439 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
440 tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
441 tmpfile_blob.out_refcnt = 1;
443 for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
444 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
445 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
453 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
455 struct apply_ctx *ctx = _ctx;
457 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
458 filedes_close(&ctx->tmpfile_fd);
460 status = extract_from_tmpfile(ctx->tmpfile_name, blob,
462 filedes_invalidate(&ctx->tmpfile_fd);
463 tunlink(ctx->tmpfile_name);
464 FREE(ctx->tmpfile_name);
468 return call_end_blob(blob, status, ctx->saved_cbs);
472 * Read the list of blobs to extract and feed their data into the specified
473 * callback functions.
475 * This handles checksumming each blob.
477 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
479 * This also works if the WIM is being read from a pipe.
481 * This also will split up blobs that will need to be extracted to more than
482 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
483 * Therefore, the apply_operations implementation need not worry about running
484 * out of file descriptors, unless it might open more than one file descriptor
485 * per 'blob_extraction_target' (e.g. Win32 currently might because the
486 * destination file system might not support hard links).
489 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
491 struct read_blob_callbacks wrapper_cbs = {
492 .begin_blob = begin_extract_blob_wrapper,
493 .consume_chunk = extract_chunk_wrapper,
494 .end_blob = end_extract_blob_wrapper,
497 ctx->saved_cbs = cbs;
498 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
499 return read_blobs_from_pipe(ctx, &wrapper_cbs);
501 return read_blob_list(&ctx->blob_list,
502 offsetof(struct blob_descriptor,
504 &wrapper_cbs, VERIFY_BLOB_HASHES);
508 /* Extract a WIM dentry to standard output.
510 * This obviously doesn't make sense in all cases. We return an error if the
511 * dentry does not correspond to a regular file. Otherwise we extract the
512 * unnamed data stream only. */
514 extract_dentry_to_stdout(struct wim_dentry *dentry,
515 const struct blob_table *blob_table)
517 struct wim_inode *inode = dentry->d_inode;
518 struct blob_descriptor *blob;
519 struct filedes _stdout;
521 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
522 FILE_ATTRIBUTE_DIRECTORY |
523 FILE_ATTRIBUTE_ENCRYPTED))
525 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
526 "extracted to standard output", dentry_full_path(dentry));
527 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
530 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
532 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
533 if (!is_zero_hash(hash))
534 return blob_not_found_error(inode, hash);
538 filedes_init(&_stdout, STDOUT_FILENO);
539 return extract_blob_to_fd(blob, &_stdout);
543 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
544 const struct blob_table *blob_table)
546 for (size_t i = 0; i < num_dentries; i++) {
547 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
554 /**********************************************************************/
557 * Removes duplicate dentries from the array.
559 * Returns the new number of dentries, packed at the front of the array.
562 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
565 for (i = 0; i < num_trees; i++) {
566 if (!trees[i]->d_tmp_flag) {
567 /* Found distinct dentry. */
568 trees[i]->d_tmp_flag = 1;
569 trees[j++] = trees[i];
572 for (i = 0; i < j; i++)
573 trees[i]->d_tmp_flag = 0;
578 * Remove dentries that are descendants of other dentries in the array.
580 * Returns the new number of dentries, packed at the front of the array.
583 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
586 for (i = 0; i < num_trees; i++)
587 trees[i]->d_tmp_flag = 1;
588 for (i = 0; i < num_trees; i++) {
589 struct wim_dentry *d = trees[i];
590 while (!dentry_is_root(d)) {
595 trees[j++] = trees[i];
599 trees[i]->d_tmp_flag = 0;
602 for (i = 0; i < j; i++)
603 trees[i]->d_tmp_flag = 0;
608 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
610 struct list_head *dentry_list = _dentry_list;
611 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
616 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
618 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
622 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
624 if (will_extract_dentry(dentry)) {
625 list_del(&dentry->d_extraction_list_node);
626 dentry_reset_extraction_list_node(dentry);
632 * Build the preliminary list of dentries to be extracted.
634 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
635 * an ancestor of d2, then d1 appears before d2 in the list.
638 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
639 size_t num_trees, bool add_ancestors)
641 INIT_LIST_HEAD(dentry_list);
643 /* Add the trees recursively. */
644 for (size_t i = 0; i < num_trees; i++)
645 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
647 /* If requested, add ancestors of the trees. */
649 for (size_t i = 0; i < num_trees; i++) {
650 struct wim_dentry *dentry = trees[i];
651 struct wim_dentry *ancestor;
652 struct list_head *place_after;
654 if (dentry_is_root(dentry))
657 place_after = dentry_list;
660 ancestor = ancestor->d_parent;
661 if (will_extract_dentry(ancestor)) {
662 place_after = &ancestor->d_extraction_list_node;
665 } while (!dentry_is_root(ancestor));
669 ancestor = ancestor->d_parent;
670 if (will_extract_dentry(ancestor))
672 list_add(&ancestor->d_extraction_list_node, place_after);
673 } while (!dentry_is_root(ancestor));
679 destroy_dentry_list(struct list_head *dentry_list)
681 struct wim_dentry *dentry, *tmp;
682 struct wim_inode *inode;
684 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
685 inode = dentry->d_inode;
686 dentry_reset_extraction_list_node(dentry);
687 inode->i_visited = 0;
688 inode->i_can_externally_back = 0;
689 if ((void *)dentry->d_extraction_name != (void *)dentry->d_name)
690 FREE(dentry->d_extraction_name);
691 dentry->d_extraction_name = NULL;
692 dentry->d_extraction_name_nchars = 0;
697 destroy_blob_list(struct list_head *blob_list)
699 struct blob_descriptor *blob;
701 list_for_each_entry(blob, blob_list, extraction_list)
702 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
703 FREE(blob->blob_extraction_targets);
707 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
709 static const utf16lechar replacement_char = cpu_to_le16('?');
713 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
719 for (i = 0; i < num_chars; i++) {
720 switch (le16_to_cpu(name[i])) {
722 case '\x01'...'\x1F':
735 name[i] = replacement_char;
745 dentry_calculate_extraction_name(struct wim_dentry *dentry,
746 struct apply_ctx *ctx)
750 if (dentry_is_root(dentry))
754 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
755 dentry->d_extraction_name = dentry->d_name;
756 dentry->d_extraction_name_nchars = dentry->d_name_nbytes /
762 if (!ctx->supported_features.case_sensitive_filenames) {
763 struct wim_dentry *other;
764 list_for_each_entry(other, &dentry->d_ci_conflict_list,
767 if (will_extract_dentry(other)) {
768 if (ctx->extract_flags &
769 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
770 WARNING("\"%"TS"\" has the same "
771 "case-insensitive name as "
772 "\"%"TS"\"; extracting "
773 "dummy name instead",
774 dentry_full_path(dentry),
775 dentry_full_path(other));
778 WARNING("Not extracting \"%"TS"\": "
779 "has same case-insensitive "
781 dentry_full_path(dentry),
782 dentry_full_path(other));
789 if (file_name_valid(dentry->d_name, dentry->d_name_nbytes / 2, false)) {
791 ret = utf16le_get_tstr(dentry->d_name,
792 dentry->d_name_nbytes,
793 (const tchar **)&dentry->d_extraction_name,
795 dentry->d_extraction_name_nchars = nbytes / sizeof(tchar);
798 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
800 WARNING("\"%"TS"\" has an invalid filename "
801 "that is not supported on this platform; "
802 "extracting dummy name instead",
803 dentry_full_path(dentry));
806 WARNING("Not extracting \"%"TS"\": has an invalid filename "
807 "that is not supported on this platform",
808 dentry_full_path(dentry));
815 utf16lechar utf16_name_copy[dentry->d_name_nbytes / 2];
817 memcpy(utf16_name_copy, dentry->d_name, dentry->d_name_nbytes);
818 file_name_valid(utf16_name_copy, dentry->d_name_nbytes / 2, true);
820 const tchar *tchar_name;
823 ret = utf16le_get_tstr(utf16_name_copy,
824 dentry->d_name_nbytes,
825 &tchar_name, &tchar_nchars);
829 tchar_nchars /= sizeof(tchar);
831 size_t fixed_name_num_chars = tchar_nchars;
832 tchar fixed_name[tchar_nchars + 50];
834 tmemcpy(fixed_name, tchar_name, tchar_nchars);
835 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
836 T(" (invalid filename #%lu)"),
837 ++ctx->invalid_sequence);
839 utf16le_put_tstr(tchar_name);
841 dentry->d_extraction_name = TSTRDUP(fixed_name);
842 if (!dentry->d_extraction_name)
843 return WIMLIB_ERR_NOMEM;
844 dentry->d_extraction_name_nchars = fixed_name_num_chars;
849 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
854 * Calculate the actual filename component at which each WIM dentry will be
855 * extracted, with special handling for dentries that are unsupported by the
856 * extraction backend or have invalid names.
858 * ctx->supported_features must be filled in.
860 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
863 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
864 struct apply_ctx *ctx)
866 struct list_head *prev, *cur;
868 /* Can't use list_for_each_entry() because a call to
869 * dentry_calculate_extraction_name() may delete the current dentry and
870 * its children from the list. */
874 struct wim_dentry *dentry;
878 if (cur == dentry_list)
881 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
883 ret = dentry_calculate_extraction_name(dentry, ctx);
887 if (prev->next == cur)
890 ; /* Current dentry and its children (which follow in
891 the list) were deleted. prev stays the same. */
897 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
898 struct blob_table *blob_table)
900 struct wim_inode *inode = dentry->d_inode;
901 struct blob_descriptor *blob;
905 /* Special case: when extracting from a pipe, the WIM blob table is
906 * initially empty, so "resolving" an inode's streams is initially not
907 * possible. However, we still need to keep track of which blobs,
908 * identified by SHA-1 message digests, need to be extracted, so we
909 * "resolve" the inode's streams anyway by allocating a 'struct
910 * blob_descriptor' for each one. */
911 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
913 ret = inode_resolve_streams(inode, blob_table, force);
916 for (unsigned i = 0; i < inode->i_num_streams; i++) {
917 blob = stream_blob_resolved(&inode->i_streams[i]);
919 blob->out_refcnt = 0;
925 * For each dentry to be extracted, resolve all streams in the corresponding
926 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
928 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
931 dentry_list_resolve_streams(struct list_head *dentry_list,
932 struct apply_ctx *ctx)
934 struct wim_dentry *dentry;
937 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
938 ret = dentry_resolve_streams(dentry,
940 ctx->wim->blob_table);
948 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
949 struct apply_ctx *ctx)
951 struct wim_inode *inode = dentry->d_inode;
952 struct blob_descriptor *blob = stream_blob_resolved(strm);
953 struct blob_extraction_target *targets;
958 /* Tally the size only for each actual extraction of the stream (not
959 * additional hard links to the inode). */
960 if (inode->i_visited && ctx->supported_features.hard_links)
963 ctx->progress.extract.total_bytes += blob->size;
964 ctx->progress.extract.total_streams++;
966 if (inode->i_visited)
969 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
970 * many extraction targets it will have. */
971 if (blob->out_refcnt == 0) {
972 list_add_tail(&blob->extraction_list, &ctx->blob_list);
973 ctx->num_blobs_remaining++;
976 /* Set this stream as an extraction target of 'blob'. */
978 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
979 targets = blob->inline_blob_extraction_targets;
981 struct blob_extraction_target *prev_targets;
982 size_t alloc_blob_extraction_targets;
984 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
986 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
988 prev_targets = blob->blob_extraction_targets;
989 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
992 if (blob->out_refcnt == alloc_blob_extraction_targets) {
993 alloc_blob_extraction_targets *= 2;
994 targets = REALLOC(prev_targets,
995 alloc_blob_extraction_targets *
998 return WIMLIB_ERR_NOMEM;
1001 blob->inline_blob_extraction_targets,
1002 sizeof(blob->inline_blob_extraction_targets));
1004 blob->blob_extraction_targets = targets;
1005 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1007 targets = blob->blob_extraction_targets;
1009 targets[blob->out_refcnt].inode = inode;
1010 targets[blob->out_refcnt].stream = strm;
1016 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1017 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1019 bool need_stream = false;
1020 switch (strm->stream_type) {
1021 case STREAM_TYPE_DATA:
1022 if (stream_is_named(strm)) {
1023 /* Named data stream */
1024 if (ctx->supported_features.named_data_streams)
1026 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1027 FILE_ATTRIBUTE_ENCRYPTED))
1028 && !(inode_is_symlink(inode)
1029 && !ctx->supported_features.reparse_points
1030 && ctx->supported_features.symlink_reparse_points))
1033 * Unnamed data stream. Skip if any of the following is true:
1035 * - file is a directory
1036 * - file is encrypted
1037 * - backend needs to create the file as UNIX symlink
1038 * - backend will extract the stream as externally
1039 * backed from the WIM archive itself
1041 if (ctx->apply_ops->will_back_from_wim) {
1042 int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx);
1043 if (ret > 0) /* Error? */
1045 if (ret < 0) /* Won't externally back? */
1052 case STREAM_TYPE_REPARSE_POINT:
1053 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1054 if (ctx->supported_features.reparse_points ||
1055 (inode_is_symlink(inode) &&
1056 ctx->supported_features.symlink_reparse_points))
1059 case STREAM_TYPE_EFSRPC_RAW_DATA:
1060 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1061 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1062 if (ctx->supported_features.encrypted_directories)
1065 if (ctx->supported_features.encrypted_files)
1071 return ref_stream(strm, dentry, ctx);
1076 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1078 struct wim_inode *inode = dentry->d_inode;
1079 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1080 int ret = ref_stream_if_needed(dentry, inode,
1081 &inode->i_streams[i], ctx);
1085 inode->i_visited = 1;
1090 * Given a list of dentries to be extracted, build the list of blobs that need
1091 * to be extracted, and for each blob determine the streams to which that blob
1092 * will be extracted.
1094 * This also initializes the extract progress info with byte and blob
1097 * ctx->supported_features must be filled in.
1100 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1102 struct wim_dentry *dentry;
1105 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1106 ret = dentry_ref_streams(dentry, ctx);
1110 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1111 dentry->d_inode->i_visited = 0;
1116 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1118 struct wim_dentry *dentry;
1120 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1121 dentry->d_inode->i_first_extraction_alias = NULL;
1123 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1124 dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias;
1125 dentry->d_inode->i_first_extraction_alias = dentry;
1130 inode_tally_features(const struct wim_inode *inode,
1131 struct wim_features *features)
1133 if (inode->i_attributes & FILE_ATTRIBUTE_READONLY)
1134 features->readonly_files++;
1135 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1136 features->hidden_files++;
1137 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1138 features->system_files++;
1139 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1140 features->archive_files++;
1141 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1142 features->compressed_files++;
1143 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1144 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1145 features->encrypted_directories++;
1147 features->encrypted_files++;
1149 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1150 features->not_context_indexed_files++;
1151 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1152 features->sparse_files++;
1153 if (inode_has_named_data_stream(inode))
1154 features->named_data_streams++;
1155 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1156 features->reparse_points++;
1157 if (inode_is_symlink(inode))
1158 features->symlink_reparse_points++;
1160 features->other_reparse_points++;
1162 if (inode_has_security_descriptor(inode))
1163 features->security_descriptors++;
1164 if (inode_has_unix_data(inode))
1165 features->unix_data++;
1166 if (inode_has_object_id(inode))
1167 features->object_ids++;
1170 /* Tally features necessary to extract a dentry and the corresponding inode. */
1172 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1174 struct wim_inode *inode = dentry->d_inode;
1176 if (dentry_has_short_name(dentry))
1177 features->short_names++;
1179 if (inode->i_visited) {
1180 features->hard_links++;
1182 inode_tally_features(inode, features);
1183 inode->i_visited = 1;
1187 /* Tally the features necessary to extract the specified dentries. */
1189 dentry_list_get_features(struct list_head *dentry_list,
1190 struct wim_features *features)
1192 struct wim_dentry *dentry;
1194 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1195 dentry_tally_features(dentry, features);
1197 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1198 dentry->d_inode->i_visited = 0;
1202 do_feature_check(const struct wim_features *required_features,
1203 const struct wim_features *supported_features,
1206 /* Encrypted files. */
1207 if (required_features->encrypted_files &&
1208 !supported_features->encrypted_files)
1209 WARNING("Ignoring EFS-encrypted data of %lu files",
1210 required_features->encrypted_files);
1212 /* Named data streams. */
1213 if (required_features->named_data_streams &&
1214 !supported_features->named_data_streams)
1215 WARNING("Ignoring named data streams of %lu files",
1216 required_features->named_data_streams);
1218 /* File attributes. */
1219 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1221 if (required_features->readonly_files &&
1222 !supported_features->readonly_files)
1223 WARNING("Ignoring FILE_ATTRIBUTE_READONLY of %lu files",
1224 required_features->readonly_files);
1226 if (required_features->hidden_files &&
1227 !supported_features->hidden_files)
1228 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1229 required_features->hidden_files);
1231 if (required_features->system_files &&
1232 !supported_features->system_files)
1233 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1234 required_features->system_files);
1236 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1237 * We're an archive program, so theoretically we can do what we
1240 if (required_features->compressed_files &&
1241 !supported_features->compressed_files)
1242 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1243 required_features->compressed_files);
1245 if (required_features->not_context_indexed_files &&
1246 !supported_features->not_context_indexed_files)
1247 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1248 required_features->not_context_indexed_files);
1250 if (required_features->sparse_files &&
1251 !supported_features->sparse_files)
1252 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1253 required_features->sparse_files);
1255 if (required_features->encrypted_directories &&
1256 !supported_features->encrypted_directories)
1257 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1258 required_features->encrypted_directories);
1262 if (required_features->hard_links && !supported_features->hard_links)
1263 WARNING("Extracting %lu hard links as independent files",
1264 required_features->hard_links);
1266 /* Symbolic links and reparse points. */
1267 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1268 required_features->symlink_reparse_points &&
1269 !supported_features->symlink_reparse_points &&
1270 !supported_features->reparse_points)
1272 ERROR("Extraction backend does not support symbolic links!");
1273 return WIMLIB_ERR_UNSUPPORTED;
1275 if (required_features->reparse_points &&
1276 !supported_features->reparse_points)
1278 if (supported_features->symlink_reparse_points) {
1279 if (required_features->other_reparse_points) {
1280 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1281 required_features->other_reparse_points);
1284 WARNING("Ignoring reparse data of %lu files",
1285 required_features->reparse_points);
1289 /* Security descriptors. */
1290 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1291 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1292 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1293 required_features->security_descriptors &&
1294 !supported_features->security_descriptors)
1296 ERROR("Extraction backend does not support security descriptors!");
1297 return WIMLIB_ERR_UNSUPPORTED;
1299 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1300 required_features->security_descriptors &&
1301 !supported_features->security_descriptors)
1302 WARNING("Ignoring Windows NT security descriptors of %lu files",
1303 required_features->security_descriptors);
1306 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1307 required_features->unix_data && !supported_features->unix_data)
1309 ERROR("Extraction backend does not support UNIX data!");
1310 return WIMLIB_ERR_UNSUPPORTED;
1313 if (required_features->unix_data &&
1314 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1316 WARNING("Ignoring UNIX metadata of %lu files",
1317 required_features->unix_data);
1321 if (required_features->object_ids && !supported_features->object_ids) {
1322 WARNING("Ignoring object IDs of %lu files",
1323 required_features->object_ids);
1327 if (required_features->short_names &&
1328 !supported_features->short_names)
1330 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1331 ERROR("Extraction backend does not support DOS names!");
1332 return WIMLIB_ERR_UNSUPPORTED;
1334 WARNING("Ignoring DOS names of %lu files",
1335 required_features->short_names);
1339 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1340 !supported_features->timestamps)
1342 ERROR("Extraction backend does not support timestamps!");
1343 return WIMLIB_ERR_UNSUPPORTED;
1349 static const struct apply_operations *
1350 select_apply_operations(int extract_flags)
1353 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1354 return &ntfs_3g_apply_ops;
1357 return &win32_apply_ops;
1359 return &unix_apply_ops;
1364 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1365 const tchar *target, int extract_flags)
1367 const struct apply_operations *ops;
1368 struct apply_ctx *ctx;
1370 LIST_HEAD(dentry_list);
1372 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1373 ret = extract_dentries_to_stdout(trees, num_trees,
1378 num_trees = remove_duplicate_trees(trees, num_trees);
1379 num_trees = remove_contained_trees(trees, num_trees);
1381 ops = select_apply_operations(extract_flags);
1383 if (num_trees > 1 && ops->single_tree_only) {
1384 ERROR("Extracting multiple directory trees "
1385 "at once is not supported in %s extraction mode!",
1387 ret = WIMLIB_ERR_UNSUPPORTED;
1391 ctx = CALLOC(1, ops->context_size);
1393 ret = WIMLIB_ERR_NOMEM;
1398 ctx->target = target;
1399 ctx->target_nchars = tstrlen(target);
1400 ctx->extract_flags = extract_flags;
1401 if (ctx->wim->progfunc) {
1402 ctx->progfunc = ctx->wim->progfunc;
1403 ctx->progctx = ctx->wim->progctx;
1404 ctx->progress.extract.image = wim->current_image;
1405 ctx->progress.extract.extract_flags = (extract_flags &
1406 WIMLIB_EXTRACT_MASK_PUBLIC);
1407 ctx->progress.extract.wimfile_name = wim->filename;
1408 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1409 wim->current_image);
1410 ctx->progress.extract.target = target;
1412 INIT_LIST_HEAD(&ctx->blob_list);
1413 filedes_invalidate(&ctx->tmpfile_fd);
1414 ctx->apply_ops = ops;
1416 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1420 build_dentry_list(&dentry_list, trees, num_trees,
1422 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1424 dentry_list_get_features(&dentry_list, &ctx->required_features);
1426 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1427 ctx->extract_flags);
1431 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1435 if (unlikely(list_empty(&dentry_list))) {
1436 WARNING("There is nothing to extract!");
1440 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1444 dentry_list_build_inode_alias_lists(&dentry_list);
1446 ret = dentry_list_ref_streams(&dentry_list, ctx);
1450 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1451 /* When extracting from a pipe, the number of bytes of data to
1452 * extract can't be determined in the normal way (examining the
1453 * blob table), since at this point all we have is a set of
1454 * SHA-1 message digests of blobs that need to be extracted.
1455 * However, we can get a reasonably accurate estimate by taking
1456 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1457 * data. This does assume that a full image is being extracted,
1458 * but currently there is no API for doing otherwise. (Also,
1459 * subtract <HARDLINKBYTES> from this if hard links are
1460 * supported by the extraction mode.) */
1461 ctx->progress.extract.total_bytes =
1462 xml_get_image_total_bytes(wim->xml_info,
1463 wim->current_image);
1464 if (ctx->supported_features.hard_links) {
1465 ctx->progress.extract.total_bytes -=
1466 xml_get_image_hard_link_bytes(wim->xml_info,
1467 wim->current_image);
1471 ret = extract_progress(ctx,
1472 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1473 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1474 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1478 ret = (*ops->extract)(&dentry_list, ctx);
1482 if (ctx->progress.extract.completed_bytes <
1483 ctx->progress.extract.total_bytes)
1485 ctx->progress.extract.completed_bytes =
1486 ctx->progress.extract.total_bytes;
1487 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1492 ret = extract_progress(ctx,
1493 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1494 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1495 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1497 destroy_blob_list(&ctx->blob_list);
1498 destroy_dentry_list(&dentry_list);
1505 mkdir_if_needed(const tchar *target)
1507 if (!tmkdir(target, 0755))
1510 if (errno == EEXIST)
1514 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1515 if (errno == EACCES)
1519 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1520 return WIMLIB_ERR_MKDIR;
1523 /* Make sure the extraction flags make sense, and update them if needed. */
1525 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1527 int extract_flags = *extract_flags_p;
1529 /* Check for invalid flag combinations */
1531 if ((extract_flags &
1532 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1533 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1534 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1535 return WIMLIB_ERR_INVALID_PARAM;
1537 if ((extract_flags &
1538 (WIMLIB_EXTRACT_FLAG_RPFIX |
1539 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1540 WIMLIB_EXTRACT_FLAG_NORPFIX))
1541 return WIMLIB_ERR_INVALID_PARAM;
1543 #ifndef WITH_NTFS_3G
1544 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1545 ERROR("wimlib was compiled without support for NTFS-3G, so\n"
1546 " it cannot apply a WIM image directly to an NTFS volume.");
1547 return WIMLIB_ERR_UNSUPPORTED;
1551 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1554 return WIMLIB_ERR_NO_FILENAME;
1556 ERROR("WIMBoot extraction is only supported on Windows!");
1557 return WIMLIB_ERR_UNSUPPORTED;
1561 if (extract_flags & (WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K |
1562 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K |
1563 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K |
1564 WIMLIB_EXTRACT_FLAG_COMPACT_LZX))
1568 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K) != 0);
1569 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K) != 0);
1570 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K) != 0);
1571 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_LZX) != 0);
1573 ERROR("Only one compression format can be specified "
1574 "for compact-mode extraction!");
1575 return WIMLIB_ERR_INVALID_PARAM;
1577 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1578 ERROR("Compact-mode extraction and WIMBoot-mode "
1579 "extraction are mutually exclusive!");
1580 return WIMLIB_ERR_INVALID_PARAM;
1583 ERROR("Compact-mode extraction (System Compression) "
1584 "is only supported on Windows!");
1585 return WIMLIB_ERR_UNSUPPORTED;
1590 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1591 WIMLIB_EXTRACT_FLAG_NORPFIX |
1592 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1593 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1595 /* For full-image extraction, do reparse point fixups by default
1596 * if the WIM header says they are enabled. */
1597 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1598 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1601 *extract_flags_p = extract_flags;
1605 struct append_dentry_ctx {
1606 struct wim_dentry **dentries;
1607 size_t num_dentries;
1608 size_t num_alloc_dentries;
1612 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1614 struct append_dentry_ctx *ctx = _ctx;
1616 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1617 struct wim_dentry **new_dentries;
1620 new_length = max(ctx->num_alloc_dentries + 8,
1621 ctx->num_alloc_dentries * 3 / 2);
1622 new_dentries = REALLOC(ctx->dentries,
1623 new_length * sizeof(ctx->dentries[0]));
1624 if (new_dentries == NULL)
1625 return WIMLIB_ERR_NOMEM;
1626 ctx->dentries = new_dentries;
1627 ctx->num_alloc_dentries = new_length;
1629 ctx->dentries[ctx->num_dentries++] = dentry;
1633 /* Append dentries matched by a path which can contain wildcard characters. */
1635 append_matched_dentries(WIMStruct *wim, const tchar *orig_pattern,
1636 int extract_flags, struct append_dentry_ctx *ctx)
1638 const size_t count_before = ctx->num_dentries;
1642 pattern = canonicalize_wim_path(orig_pattern);
1644 return WIMLIB_ERR_NOMEM;
1645 ret = expand_path_pattern(wim_get_current_root_dentry(wim), pattern,
1646 append_dentry_cb, ctx);
1648 if (ret || ctx->num_dentries > count_before)
1650 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) {
1651 ERROR("No matches for path pattern \"%"TS"\"", orig_pattern);
1652 return WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1654 WARNING("No matches for path pattern \"%"TS"\"", orig_pattern);
1659 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1660 const tchar * const *paths, size_t num_paths,
1664 struct wim_dentry **trees;
1667 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1668 (num_paths != 0 && paths == NULL))
1669 return WIMLIB_ERR_INVALID_PARAM;
1671 ret = check_extract_flags(wim, &extract_flags);
1675 ret = select_wim_image(wim, image);
1679 ret = wim_checksum_unhashed_blobs(wim);
1683 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1684 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1685 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1687 ret = mkdir_if_needed(target);
1692 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1694 struct append_dentry_ctx append_dentry_ctx = {
1697 .num_alloc_dentries = 0,
1700 for (size_t i = 0; i < num_paths; i++) {
1701 ret = append_matched_dentries(wim, paths[i],
1703 &append_dentry_ctx);
1705 trees = append_dentry_ctx.dentries;
1706 goto out_free_trees;
1709 trees = append_dentry_ctx.dentries;
1710 num_trees = append_dentry_ctx.num_dentries;
1712 trees = MALLOC(num_paths * sizeof(trees[0]));
1714 return WIMLIB_ERR_NOMEM;
1716 for (size_t i = 0; i < num_paths; i++) {
1718 tchar *path = canonicalize_wim_path(paths[i]);
1720 ret = WIMLIB_ERR_NOMEM;
1721 goto out_free_trees;
1724 trees[i] = get_dentry(wim, path,
1725 WIMLIB_CASE_PLATFORM_DEFAULT);
1727 if (trees[i] == NULL) {
1728 ERROR("Path \"%"TS"\" does not exist "
1730 paths[i], wim->current_image);
1731 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1732 goto out_free_trees;
1735 num_trees = num_paths;
1738 if (num_trees == 0) {
1740 goto out_free_trees;
1743 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1750 extract_single_image(WIMStruct *wim, int image,
1751 const tchar *target, int extract_flags)
1753 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1754 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1755 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1758 static const tchar * const filename_forbidden_chars =
1767 /* This function checks if it is okay to use a WIM image's name as a directory
1770 image_name_ok_as_dir(const tchar *image_name)
1772 return image_name && *image_name &&
1773 !tstrpbrk(image_name, filename_forbidden_chars) &&
1774 tstrcmp(image_name, T(".")) &&
1775 tstrcmp(image_name, T("..")) &&
1776 tstrlen(image_name) <= 128;
1779 /* Extracts all images from the WIM to the directory @target, with the images
1780 * placed in subdirectories named by their image names. */
1782 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1784 size_t output_path_len = tstrlen(target);
1785 tchar buf[output_path_len + 1 + 128 + 1];
1788 const tchar *image_name;
1790 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1791 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1792 return WIMLIB_ERR_INVALID_PARAM;
1795 ret = mkdir_if_needed(target);
1798 tmemcpy(buf, target, output_path_len);
1799 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1800 for (image = 1; image <= wim->hdr.image_count; image++) {
1801 image_name = wimlib_get_image_name(wim, image);
1802 if (image_name_ok_as_dir(image_name)) {
1803 tstrcpy(buf + output_path_len + 1, image_name);
1805 /* Image name is empty or contains forbidden characters.
1806 * Use image number instead. */
1807 tsprintf(buf + output_path_len + 1, T("%d"), image);
1809 ret = extract_single_image(wim, image, buf, extract_flags);
1817 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1820 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1821 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1822 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1823 return WIMLIB_ERR_INVALID_PARAM;
1825 if (image == WIMLIB_ALL_IMAGES)
1826 return extract_all_images(wim, target, extract_flags);
1828 return extract_single_image(wim, image, target, extract_flags);
1832 /****************************************************************************
1834 ****************************************************************************/
1837 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1838 const tchar * const *paths, size_t num_paths,
1841 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1842 return WIMLIB_ERR_INVALID_PARAM;
1844 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1849 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1850 const tchar *path_list_file, int extract_flags)
1857 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1859 ERROR("Failed to read path list file \"%"TS"\"",
1864 ret = wimlib_extract_paths(wim, image, target,
1865 (const tchar * const *)paths, num_paths,
1873 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1874 const tchar *image_num_or_name,
1875 const tchar *target,
1877 wimlib_progress_func_t progfunc,
1882 struct filedes *in_fd;
1886 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1887 return WIMLIB_ERR_INVALID_PARAM;
1889 /* Read the WIM header from the pipe and get a WIMStruct to represent
1890 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1891 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1892 * empty blob table, no XML data read, and no filename set. */
1893 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1898 /* Sanity check to make sure this is a pipable WIM. */
1899 if (pwm->hdr.magic != PWM_MAGIC) {
1900 ERROR("The WIM being read from file descriptor %d "
1901 "is not pipable!", pipe_fd);
1902 ret = WIMLIB_ERR_NOT_PIPABLE;
1903 goto out_wimlib_free;
1906 /* Sanity check to make sure the first part of a pipable split WIM is
1907 * sent over the pipe first. */
1908 if (pwm->hdr.part_number != 1) {
1909 ERROR("The first part of the split WIM must be "
1910 "sent over the pipe first.");
1911 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1912 goto out_wimlib_free;
1915 in_fd = &pwm->in_fd;
1916 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1918 /* As mentioned, the WIMStruct we created from the pipe does not have
1919 * XML data yet. Fix this by reading the extra copy of the XML data
1920 * that directly follows the header in pipable WIMs. (Note: see
1921 * write_pipable_wim() for more details about the format of pipable
1924 u8 hash[SHA1_HASH_SIZE];
1926 ret = read_pwm_blob_header(pwm, hash,
1927 &pwm->hdr.xml_data_reshdr, NULL);
1929 goto out_wimlib_free;
1931 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1932 ERROR("Expected XML data, but found non-metadata resource.");
1933 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1934 goto out_wimlib_free;
1937 ret = read_wim_xml_data(pwm);
1939 goto out_wimlib_free;
1941 if (xml_get_image_count(pwm->xml_info) != pwm->hdr.image_count) {
1942 ERROR("Image count in XML data is not the same as in WIM header.");
1943 ret = WIMLIB_ERR_IMAGE_COUNT;
1944 goto out_wimlib_free;
1948 /* Get image index (this may use the XML data that was just read to
1949 * resolve an image name). */
1950 if (image_num_or_name) {
1951 image = wimlib_resolve_image(pwm, image_num_or_name);
1952 if (image == WIMLIB_NO_IMAGE) {
1953 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1955 ret = WIMLIB_ERR_INVALID_IMAGE;
1956 goto out_wimlib_free;
1957 } else if (image == WIMLIB_ALL_IMAGES) {
1958 ERROR("Applying all images from a pipe is not supported!");
1959 ret = WIMLIB_ERR_INVALID_IMAGE;
1960 goto out_wimlib_free;
1963 if (pwm->hdr.image_count != 1) {
1964 ERROR("No image was specified, but the pipable WIM "
1965 "did not contain exactly 1 image");
1966 ret = WIMLIB_ERR_INVALID_IMAGE;
1967 goto out_wimlib_free;
1972 /* Load the needed metadata resource. */
1973 for (i = 1; i <= pwm->hdr.image_count; i++) {
1974 ret = handle_pwm_metadata_resource(pwm, i, i == image);
1976 goto out_wimlib_free;
1978 /* Extract the image. */
1979 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1980 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1981 /* Clean up and return. */
1989 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1990 const tchar *target, int extract_flags)
1992 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2001 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2004 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2005 return WIMLIB_ERR_INVALID_PARAM;
2006 return do_wimlib_extract_image(wim, image, target, extract_flags);