4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012, 2013, 2014, 2015 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.
47 #include "wimlib/apply.h"
48 #include "wimlib/assert.h"
49 #include "wimlib/blob_table.h"
50 #include "wimlib/dentry.h"
51 #include "wimlib/encoding.h"
52 #include "wimlib/endianness.h"
53 #include "wimlib/error.h"
54 #include "wimlib/metadata.h"
55 #include "wimlib/pathlist.h"
56 #include "wimlib/paths.h"
57 #include "wimlib/reparse.h"
58 #include "wimlib/resource.h"
59 #include "wimlib/security.h"
60 #include "wimlib/unix_data.h"
61 #include "wimlib/wildcard.h"
62 #include "wimlib/wim.h"
63 #include "wimlib/win32.h" /* for realpath() equivalent */
64 #include "wimlib/xml.h"
66 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
67 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
69 /* Keep in sync with wimlib.h */
70 #define WIMLIB_EXTRACT_MASK_PUBLIC \
71 (WIMLIB_EXTRACT_FLAG_NTFS | \
72 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
73 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
74 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
75 WIMLIB_EXTRACT_FLAG_RPFIX | \
76 WIMLIB_EXTRACT_FLAG_NORPFIX | \
77 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
78 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
79 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
80 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
81 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
82 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
83 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
84 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
85 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
86 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
87 WIMLIB_EXTRACT_FLAG_WIMBOOT)
89 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
90 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
92 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
94 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
95 return extract_progress(ctx, msg);
99 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
101 ctx->progress.extract.current_file_count = 0;
102 ctx->progress.extract.end_file_count = end_file_count;
103 return do_file_extract_progress(ctx, msg);
107 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
109 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
113 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
115 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
119 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
121 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
122 return do_file_extract_progress(ctx, msg);
126 end_file_structure_phase(struct apply_ctx *ctx)
128 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
132 end_file_metadata_phase(struct apply_ctx *ctx)
134 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
137 #define PWM_FOUND_WIM_HDR (-1)
139 /* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
140 * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */
142 read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
143 struct wim_reshdr *reshdr_ret,
144 struct wim_header_disk *pwm_hdr_ret)
147 struct pwm_blob_hdr blob_hdr;
150 ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
154 magic = le64_to_cpu(blob_hdr.magic);
156 if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
157 memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
158 ret = full_read(&pwm->in_fd,
159 (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
160 sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
163 return PWM_FOUND_WIM_HDR;
166 if (unlikely(magic != PWM_BLOB_MAGIC)) {
167 ERROR("Data read on pipe is invalid (expected blob header)");
168 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
171 copy_hash(hash_ret, blob_hdr.hash);
173 reshdr_ret->size_in_wim = 0; /* Not available */
174 reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
175 reshdr_ret->offset_in_wim = pwm->in_fd.offset;
176 reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
178 if (unlikely(reshdr_ret->uncompressed_size == 0)) {
179 ERROR("Data read on pipe is invalid (resource is of 0 size)");
180 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
186 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
187 ERROR("The pipe ended before all needed data was sent!");
189 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
194 read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
197 u8 hash[SHA1_HASH_SIZE];
198 struct wim_reshdr reshdr;
199 struct wim_header_disk pwm_hdr;
200 struct wim_resource_descriptor rdesc;
201 struct blob_descriptor *blob;
203 copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
204 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
205 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
206 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
210 while (ctx->num_blobs_remaining) {
212 ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
214 if (ret == PWM_FOUND_WIM_HDR) {
215 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
216 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
218 if (part_number == ctx->progress.extract.part_number &&
219 total_parts == ctx->progress.extract.total_parts &&
220 guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
223 copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
224 ctx->progress.extract.part_number = part_number;
225 ctx->progress.extract.total_parts = total_parts;
226 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
236 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
237 && (blob = lookup_blob(ctx->wim->blob_table, hash))
238 && (blob->out_refcnt))
240 wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
241 ret = read_blob_with_sha1(blob, cbs);
242 blob_unset_is_located_in_wim_resource(blob);
245 ctx->num_blobs_remaining--;
247 wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
248 ret = skip_wim_resource(&rdesc);
257 /* Creates a temporary file opened for writing. The open file descriptor is
258 * returned in @fd_ret and its name is returned in @name_ret (dynamically
261 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
268 name = ttempnam(NULL, T("wimlib"));
270 ERROR_WITH_ERRNO("Failed to create temporary filename");
271 return WIMLIB_ERR_NOMEM;
274 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
276 open_flags |= _O_SHORT_LIVED;
278 raw_fd = topen(name, open_flags, 0600);
281 if (errno == EEXIST) {
285 ERROR_WITH_ERRNO("Failed to create temporary file "
288 return WIMLIB_ERR_OPEN;
291 filedes_init(fd_ret, raw_fd);
297 begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
299 struct apply_ctx *ctx = _ctx;
301 ctx->cur_blob = blob;
302 ctx->cur_blob_offset = 0;
304 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
305 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
307 return call_begin_blob(blob, ctx->saved_cbs);
311 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
313 struct apply_ctx *ctx = _ctx;
314 union wimlib_progress_info *progress = &ctx->progress;
317 ctx->cur_blob_offset += size;
319 if (likely(ctx->supported_features.hard_links)) {
320 progress->extract.completed_bytes +=
321 (u64)size * ctx->cur_blob->out_refcnt;
322 if (ctx->cur_blob_offset == ctx->cur_blob->size)
323 progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
325 const struct blob_extraction_target *targets =
326 blob_extraction_targets(ctx->cur_blob);
327 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
328 const struct wim_inode *inode = targets[i].inode;
329 const struct wim_dentry *dentry;
331 list_for_each_entry(dentry,
332 &inode->i_extraction_aliases,
333 d_extraction_alias_node)
335 progress->extract.completed_bytes += size;
336 if (ctx->cur_blob_offset == ctx->cur_blob->size)
337 progress->extract.completed_streams++;
341 if (progress->extract.completed_bytes >= ctx->next_progress) {
343 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
347 if (progress->extract.completed_bytes >=
348 progress->extract.total_bytes)
350 ctx->next_progress = UINT64_MAX;
352 /* Send new message as soon as another 1/128 of the
353 * total has been extracted. (Arbitrary number.) */
355 progress->extract.completed_bytes +
356 progress->extract.total_bytes / 128;
358 /* ... Unless that would be more than 5000000 bytes, in
359 * which case send the next after the next 5000000
360 * bytes. (Another arbitrary number.) */
361 if (progress->extract.completed_bytes + 5000000 <
364 progress->extract.completed_bytes + 5000000;
366 /* ... But always send a message as soon as we're
367 * completely done. */
368 if (progress->extract.total_bytes < ctx->next_progress)
369 ctx->next_progress = progress->extract.total_bytes;
373 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
374 /* Just extracting to temporary file for now. */
375 ret = full_write(&ctx->tmpfile_fd, chunk, size);
377 ERROR_WITH_ERRNO("Error writing data to "
378 "temporary file \"%"TS"\"",
384 return call_consume_chunk(chunk, size, ctx->saved_cbs);
387 /* Copy the blob's data from the temporary file to each of its targets.
389 * This is executed only in the very uncommon case that a blob is being
390 * extracted to more than MAX_OPEN_FILES targets! */
392 extract_from_tmpfile(const tchar *tmpfile_name,
393 const struct blob_descriptor *orig_blob,
394 const struct read_blob_callbacks *cbs)
396 struct blob_descriptor tmpfile_blob;
397 const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
400 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
401 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
402 tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
403 tmpfile_blob.out_refcnt = 1;
405 for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
406 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
407 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
415 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
417 struct apply_ctx *ctx = _ctx;
419 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
420 filedes_close(&ctx->tmpfile_fd);
422 status = extract_from_tmpfile(ctx->tmpfile_name, blob,
424 filedes_invalidate(&ctx->tmpfile_fd);
425 tunlink(ctx->tmpfile_name);
426 FREE(ctx->tmpfile_name);
430 return call_end_blob(blob, status, ctx->saved_cbs);
434 * Read the list of blobs to extract and feed their data into the specified
435 * callback functions.
437 * This handles checksumming each blob.
439 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
441 * This also works if the WIM is being read from a pipe.
443 * This also will split up blobs that will need to be extracted to more than
444 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
445 * Therefore, the apply_operations implementation need not worry about running
446 * out of file descriptors, unless it might open more than one file descriptor
447 * per 'blob_extraction_target' (e.g. Win32 currently might because the
448 * destination file system might not support hard links).
451 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
453 struct read_blob_callbacks wrapper_cbs = {
454 .begin_blob = begin_extract_blob_wrapper,
455 .consume_chunk = extract_chunk_wrapper,
456 .end_blob = end_extract_blob_wrapper,
459 ctx->saved_cbs = cbs;
460 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
461 return read_blobs_from_pipe(ctx, &wrapper_cbs);
463 return read_blob_list(&ctx->blob_list,
464 offsetof(struct blob_descriptor,
466 &wrapper_cbs, VERIFY_BLOB_HASHES);
470 /* Extract a WIM dentry to standard output.
472 * This obviously doesn't make sense in all cases. We return an error if the
473 * dentry does not correspond to a regular file. Otherwise we extract the
474 * unnamed data stream only. */
476 extract_dentry_to_stdout(struct wim_dentry *dentry,
477 const struct blob_table *blob_table)
479 struct wim_inode *inode = dentry->d_inode;
480 struct blob_descriptor *blob;
481 struct filedes _stdout;
483 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
484 FILE_ATTRIBUTE_DIRECTORY |
485 FILE_ATTRIBUTE_ENCRYPTED))
487 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
488 "extracted to standard output", dentry_full_path(dentry));
489 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
492 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
494 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
495 if (!is_zero_hash(hash))
496 return blob_not_found_error(inode, hash);
500 filedes_init(&_stdout, STDOUT_FILENO);
501 return extract_blob_to_fd(blob, &_stdout);
505 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
506 const struct blob_table *blob_table)
508 for (size_t i = 0; i < num_dentries; i++) {
509 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
516 /**********************************************************************/
519 * Removes duplicate dentries from the array.
521 * Returns the new number of dentries, packed at the front of the array.
524 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
527 for (i = 0; i < num_trees; i++) {
528 if (!trees[i]->tmp_flag) {
529 /* Found distinct dentry. */
530 trees[i]->tmp_flag = 1;
531 trees[j++] = trees[i];
534 for (i = 0; i < j; i++)
535 trees[i]->tmp_flag = 0;
540 * Remove dentries that are descendants of other dentries in the array.
542 * Returns the new number of dentries, packed at the front of the array.
545 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
548 for (i = 0; i < num_trees; i++)
549 trees[i]->tmp_flag = 1;
550 for (i = 0; i < num_trees; i++) {
551 struct wim_dentry *d = trees[i];
552 while (!dentry_is_root(d)) {
557 trees[j++] = trees[i];
561 trees[i]->tmp_flag = 0;
564 for (i = 0; i < j; i++)
565 trees[i]->tmp_flag = 0;
570 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
572 struct list_head *dentry_list = _dentry_list;
573 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
578 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
580 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
584 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
586 list_del(&dentry->d_extraction_list_node);
587 dentry_reset_extraction_list_node(dentry);
592 * Build the preliminary list of dentries to be extracted.
594 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
595 * an ancestor of d2, then d1 appears before d2 in the list.
598 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
599 size_t num_trees, bool add_ancestors)
601 INIT_LIST_HEAD(dentry_list);
603 /* Add the trees recursively. */
604 for (size_t i = 0; i < num_trees; i++)
605 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
607 /* If requested, add ancestors of the trees. */
609 for (size_t i = 0; i < num_trees; i++) {
610 struct wim_dentry *dentry = trees[i];
611 struct wim_dentry *ancestor;
612 struct list_head *place_after;
614 if (dentry_is_root(dentry))
617 place_after = dentry_list;
620 ancestor = ancestor->d_parent;
621 if (will_extract_dentry(ancestor)) {
622 place_after = &ancestor->d_extraction_list_node;
625 } while (!dentry_is_root(ancestor));
629 ancestor = ancestor->d_parent;
630 if (will_extract_dentry(ancestor))
632 list_add(&ancestor->d_extraction_list_node, place_after);
633 } while (!dentry_is_root(ancestor));
639 destroy_dentry_list(struct list_head *dentry_list)
641 struct wim_dentry *dentry, *tmp;
642 struct wim_inode *inode;
644 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
645 inode = dentry->d_inode;
646 dentry_reset_extraction_list_node(dentry);
647 inode->i_visited = 0;
648 inode->i_can_externally_back = 0;
649 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
650 FREE(dentry->d_extraction_name);
651 dentry->d_extraction_name = NULL;
652 dentry->d_extraction_name_nchars = 0;
657 destroy_blob_list(struct list_head *blob_list)
659 struct blob_descriptor *blob;
661 list_for_each_entry(blob, blob_list, extraction_list)
662 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
663 FREE(blob->blob_extraction_targets);
667 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
669 static const utf16lechar replacement_char = cpu_to_le16('?');
673 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
679 for (i = 0; i < num_chars; i++) {
682 case cpu_to_le16('\\'):
683 case cpu_to_le16(':'):
684 case cpu_to_le16('*'):
685 case cpu_to_le16('?'):
686 case cpu_to_le16('"'):
687 case cpu_to_le16('<'):
688 case cpu_to_le16('>'):
689 case cpu_to_le16('|'):
691 case cpu_to_le16('/'):
692 case cpu_to_le16('\0'):
694 name[i] = replacement_char;
701 if (name[num_chars - 1] == cpu_to_le16(' ') ||
702 name[num_chars - 1] == cpu_to_le16('.'))
705 name[num_chars - 1] = replacement_char;
714 dentry_calculate_extraction_name(struct wim_dentry *dentry,
715 struct apply_ctx *ctx)
719 if (dentry_is_root(dentry))
723 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
724 dentry->d_extraction_name = dentry->file_name;
725 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
731 if (!ctx->supported_features.case_sensitive_filenames) {
732 struct wim_dentry *other;
733 list_for_each_entry(other, &dentry->d_ci_conflict_list,
736 if (will_extract_dentry(other)) {
737 if (ctx->extract_flags &
738 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
739 WARNING("\"%"TS"\" has the same "
740 "case-insensitive name as "
741 "\"%"TS"\"; extracting "
742 "dummy name instead",
743 dentry_full_path(dentry),
744 dentry_full_path(other));
747 WARNING("Not extracting \"%"TS"\": "
748 "has same case-insensitive "
750 dentry_full_path(dentry),
751 dentry_full_path(other));
758 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
759 ret = utf16le_get_tstr(dentry->file_name,
760 dentry->file_name_nbytes,
761 (const tchar **)&dentry->d_extraction_name,
762 &dentry->d_extraction_name_nchars);
763 dentry->d_extraction_name_nchars /= sizeof(tchar);
766 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
768 WARNING("\"%"TS"\" has an invalid filename "
769 "that is not supported on this platform; "
770 "extracting dummy name instead",
771 dentry_full_path(dentry));
774 WARNING("Not extracting \"%"TS"\": has an invalid filename "
775 "that is not supported on this platform",
776 dentry_full_path(dentry));
783 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
785 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
786 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
788 const tchar *tchar_name;
791 ret = utf16le_get_tstr(utf16_name_copy,
792 dentry->file_name_nbytes,
793 &tchar_name, &tchar_nchars);
797 tchar_nchars /= sizeof(tchar);
799 size_t fixed_name_num_chars = tchar_nchars;
800 tchar fixed_name[tchar_nchars + 50];
802 tmemcpy(fixed_name, tchar_name, tchar_nchars);
803 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
804 T(" (invalid filename #%lu)"),
805 ++ctx->invalid_sequence);
807 utf16le_put_tstr(tchar_name);
809 dentry->d_extraction_name = TSTRDUP(fixed_name);
810 if (!dentry->d_extraction_name)
811 return WIMLIB_ERR_NOMEM;
812 dentry->d_extraction_name_nchars = fixed_name_num_chars;
817 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
822 * Calculate the actual filename component at which each WIM dentry will be
823 * extracted, with special handling for dentries that are unsupported by the
824 * extraction backend or have invalid names.
826 * ctx->supported_features must be filled in.
828 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
831 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
832 struct apply_ctx *ctx)
834 struct list_head *prev, *cur;
836 /* Can't use list_for_each_entry() because a call to
837 * dentry_calculate_extraction_name() may delete the current dentry and
838 * its children from the list. */
842 struct wim_dentry *dentry;
846 if (cur == dentry_list)
849 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
851 ret = dentry_calculate_extraction_name(dentry, ctx);
855 if (prev->next == cur)
858 ; /* Current dentry and its children (which follow in
859 the list) were deleted. prev stays the same. */
865 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
866 struct blob_table *blob_table)
868 struct wim_inode *inode = dentry->d_inode;
869 struct blob_descriptor *blob;
873 /* Special case: when extracting from a pipe, the WIM blob table is
874 * initially empty, so "resolving" an inode's streams is initially not
875 * possible. However, we still need to keep track of which blobs,
876 * identified by SHA-1 message digests, need to be extracted, so we
877 * "resolve" the inode's streams anyway by allocating a 'struct
878 * blob_descriptor' for each one. */
879 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
881 ret = inode_resolve_streams(inode, blob_table, force);
884 for (unsigned i = 0; i < inode->i_num_streams; i++) {
885 blob = stream_blob_resolved(&inode->i_streams[i]);
887 blob->out_refcnt = 0;
893 * For each dentry to be extracted, resolve all streams in the corresponding
894 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
896 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
899 dentry_list_resolve_streams(struct list_head *dentry_list,
900 struct apply_ctx *ctx)
902 struct wim_dentry *dentry;
905 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
906 ret = dentry_resolve_streams(dentry,
908 ctx->wim->blob_table);
916 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
917 struct apply_ctx *ctx)
919 struct wim_inode *inode = dentry->d_inode;
920 struct blob_descriptor *blob = stream_blob_resolved(strm);
921 struct blob_extraction_target *targets;
926 /* Tally the size only for each actual extraction of the stream (not
927 * additional hard links to the inode). */
928 if (inode->i_visited && ctx->supported_features.hard_links)
931 ctx->progress.extract.total_bytes += blob->size;
932 ctx->progress.extract.total_streams++;
934 if (inode->i_visited)
937 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
938 * many extraction targets it will have. */
939 if (blob->out_refcnt == 0) {
940 list_add_tail(&blob->extraction_list, &ctx->blob_list);
941 ctx->num_blobs_remaining++;
944 /* Set this stream as an extraction target of 'blob'. */
946 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
947 targets = blob->inline_blob_extraction_targets;
949 struct blob_extraction_target *prev_targets;
950 size_t alloc_blob_extraction_targets;
952 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
954 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
956 prev_targets = blob->blob_extraction_targets;
957 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
960 if (blob->out_refcnt == alloc_blob_extraction_targets) {
961 alloc_blob_extraction_targets *= 2;
962 targets = REALLOC(prev_targets,
963 alloc_blob_extraction_targets *
966 return WIMLIB_ERR_NOMEM;
969 blob->inline_blob_extraction_targets,
970 sizeof(blob->inline_blob_extraction_targets));
972 blob->blob_extraction_targets = targets;
973 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
975 targets = blob->blob_extraction_targets;
977 targets[blob->out_refcnt].inode = inode;
978 targets[blob->out_refcnt].stream = strm;
984 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
985 struct wim_inode_stream *strm, struct apply_ctx *ctx)
987 bool need_stream = false;
988 switch (strm->stream_type) {
989 case STREAM_TYPE_DATA:
990 if (stream_is_named(strm)) {
991 /* Named data stream */
992 if (ctx->supported_features.named_data_streams)
994 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
995 FILE_ATTRIBUTE_ENCRYPTED))
996 && !(inode_is_symlink(inode)
997 && !ctx->supported_features.reparse_points
998 && ctx->supported_features.symlink_reparse_points))
1001 * Unnamed data stream. Skip if any of the following is true:
1003 * - file is a directory
1004 * - file is encrypted
1005 * - backend needs to create the file as UNIX symlink
1006 * - backend will extract the stream as externally backed
1008 if (ctx->apply_ops->will_externally_back) {
1009 int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1010 if (ret > 0) /* Error? */
1012 if (ret < 0) /* Won't externally back? */
1019 case STREAM_TYPE_REPARSE_POINT:
1020 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1021 if (ctx->supported_features.reparse_points ||
1022 (inode_is_symlink(inode) &&
1023 ctx->supported_features.symlink_reparse_points))
1026 case STREAM_TYPE_EFSRPC_RAW_DATA:
1027 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1028 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1029 if (ctx->supported_features.encrypted_directories)
1032 if (ctx->supported_features.encrypted_files)
1038 return ref_stream(strm, dentry, ctx);
1043 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1045 struct wim_inode *inode = dentry->d_inode;
1046 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1047 int ret = ref_stream_if_needed(dentry, inode,
1048 &inode->i_streams[i], ctx);
1052 inode->i_visited = 1;
1057 * Given a list of dentries to be extracted, build the list of blobs that need
1058 * to be extracted, and for each blob determine the streams to which that blob
1059 * will be extracted.
1061 * This also initializes the extract progress info with byte and blob
1064 * ctx->supported_features must be filled in.
1067 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1069 struct wim_dentry *dentry;
1072 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1073 ret = dentry_ref_streams(dentry, ctx);
1077 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1078 dentry->d_inode->i_visited = 0;
1083 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1085 struct wim_dentry *dentry;
1086 struct wim_inode *inode;
1088 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1089 inode = dentry->d_inode;
1090 if (!inode->i_visited)
1091 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1092 list_add_tail(&dentry->d_extraction_alias_node,
1093 &inode->i_extraction_aliases);
1094 inode->i_visited = 1;
1096 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1097 dentry->d_inode->i_visited = 0;
1101 inode_tally_features(const struct wim_inode *inode,
1102 struct wim_features *features)
1104 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1105 features->archive_files++;
1106 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1107 features->hidden_files++;
1108 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1109 features->system_files++;
1110 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1111 features->compressed_files++;
1112 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1113 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1114 features->encrypted_directories++;
1116 features->encrypted_files++;
1118 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1119 features->not_context_indexed_files++;
1120 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1121 features->sparse_files++;
1122 if (inode_has_named_data_stream(inode))
1123 features->named_data_streams++;
1124 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1125 features->reparse_points++;
1126 if (inode_is_symlink(inode))
1127 features->symlink_reparse_points++;
1129 features->other_reparse_points++;
1131 if (inode_has_security_descriptor(inode))
1132 features->security_descriptors++;
1133 if (inode_has_unix_data(inode))
1134 features->unix_data++;
1137 /* Tally features necessary to extract a dentry and the corresponding inode. */
1139 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1141 struct wim_inode *inode = dentry->d_inode;
1143 if (dentry_has_short_name(dentry))
1144 features->short_names++;
1146 if (inode->i_visited) {
1147 features->hard_links++;
1149 inode_tally_features(inode, features);
1150 inode->i_visited = 1;
1154 /* Tally the features necessary to extract the specified dentries. */
1156 dentry_list_get_features(struct list_head *dentry_list,
1157 struct wim_features *features)
1159 struct wim_dentry *dentry;
1161 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1162 dentry_tally_features(dentry, features);
1164 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1165 dentry->d_inode->i_visited = 0;
1169 do_feature_check(const struct wim_features *required_features,
1170 const struct wim_features *supported_features,
1173 /* Encrypted files. */
1174 if (required_features->encrypted_files &&
1175 !supported_features->encrypted_files)
1176 WARNING("Ignoring EFS-encrypted data of %lu files",
1177 required_features->encrypted_files);
1179 /* Named data streams. */
1180 if (required_features->named_data_streams &&
1181 !supported_features->named_data_streams)
1182 WARNING("Ignoring named data streams of %lu files",
1183 required_features->named_data_streams);
1185 /* File attributes. */
1186 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1187 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1188 * We're an archive program, so theoretically we can do what we
1191 if (required_features->hidden_files &&
1192 !supported_features->hidden_files)
1193 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1194 required_features->hidden_files);
1196 if (required_features->system_files &&
1197 !supported_features->system_files)
1198 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1199 required_features->system_files);
1201 if (required_features->compressed_files &&
1202 !supported_features->compressed_files)
1203 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1204 required_features->compressed_files);
1206 if (required_features->not_context_indexed_files &&
1207 !supported_features->not_context_indexed_files)
1208 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1209 required_features->not_context_indexed_files);
1211 if (required_features->sparse_files &&
1212 !supported_features->sparse_files)
1213 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1214 required_features->sparse_files);
1216 if (required_features->encrypted_directories &&
1217 !supported_features->encrypted_directories)
1218 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1219 required_features->encrypted_directories);
1223 if (required_features->hard_links && !supported_features->hard_links)
1224 WARNING("Extracting %lu hard links as independent files",
1225 required_features->hard_links);
1227 /* Symbolic links and reparse points. */
1228 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1229 required_features->symlink_reparse_points &&
1230 !supported_features->symlink_reparse_points &&
1231 !supported_features->reparse_points)
1233 ERROR("Extraction backend does not support symbolic links!");
1234 return WIMLIB_ERR_UNSUPPORTED;
1236 if (required_features->reparse_points &&
1237 !supported_features->reparse_points)
1239 if (supported_features->symlink_reparse_points) {
1240 if (required_features->other_reparse_points) {
1241 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1242 required_features->other_reparse_points);
1245 WARNING("Ignoring reparse data of %lu files",
1246 required_features->reparse_points);
1250 /* Security descriptors. */
1251 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1252 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1253 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1254 required_features->security_descriptors &&
1255 !supported_features->security_descriptors)
1257 ERROR("Extraction backend does not support security descriptors!");
1258 return WIMLIB_ERR_UNSUPPORTED;
1260 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1261 required_features->security_descriptors &&
1262 !supported_features->security_descriptors)
1263 WARNING("Ignoring Windows NT security descriptors of %lu files",
1264 required_features->security_descriptors);
1267 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1268 required_features->unix_data && !supported_features->unix_data)
1270 ERROR("Extraction backend does not support UNIX data!");
1271 return WIMLIB_ERR_UNSUPPORTED;
1274 if (required_features->unix_data &&
1275 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1277 WARNING("Ignoring UNIX metadata of %lu files",
1278 required_features->unix_data);
1282 if (required_features->short_names &&
1283 !supported_features->short_names)
1285 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1286 ERROR("Extraction backend does not support DOS names!");
1287 return WIMLIB_ERR_UNSUPPORTED;
1289 WARNING("Ignoring DOS names of %lu files",
1290 required_features->short_names);
1294 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1295 !supported_features->timestamps)
1297 ERROR("Extraction backend does not support timestamps!");
1298 return WIMLIB_ERR_UNSUPPORTED;
1304 static const struct apply_operations *
1305 select_apply_operations(int extract_flags)
1308 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1309 return &ntfs_3g_apply_ops;
1312 return &win32_apply_ops;
1314 return &unix_apply_ops;
1319 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1320 const tchar *target, int extract_flags)
1322 const struct apply_operations *ops;
1323 struct apply_ctx *ctx;
1325 LIST_HEAD(dentry_list);
1327 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1328 ret = extract_dentries_to_stdout(trees, num_trees,
1333 num_trees = remove_duplicate_trees(trees, num_trees);
1334 num_trees = remove_contained_trees(trees, num_trees);
1336 ops = select_apply_operations(extract_flags);
1338 if (num_trees > 1 && ops->single_tree_only) {
1339 ERROR("Extracting multiple directory trees "
1340 "at once is not supported in %s extraction mode!",
1342 ret = WIMLIB_ERR_UNSUPPORTED;
1346 ctx = CALLOC(1, ops->context_size);
1348 ret = WIMLIB_ERR_NOMEM;
1353 ctx->target = target;
1354 ctx->target_nchars = tstrlen(target);
1355 ctx->extract_flags = extract_flags;
1356 if (ctx->wim->progfunc) {
1357 ctx->progfunc = ctx->wim->progfunc;
1358 ctx->progctx = ctx->wim->progctx;
1359 ctx->progress.extract.image = wim->current_image;
1360 ctx->progress.extract.extract_flags = (extract_flags &
1361 WIMLIB_EXTRACT_MASK_PUBLIC);
1362 ctx->progress.extract.wimfile_name = wim->filename;
1363 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1364 wim->current_image);
1365 ctx->progress.extract.target = target;
1367 INIT_LIST_HEAD(&ctx->blob_list);
1368 filedes_invalidate(&ctx->tmpfile_fd);
1369 ctx->apply_ops = ops;
1371 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1375 build_dentry_list(&dentry_list, trees, num_trees,
1377 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1379 dentry_list_get_features(&dentry_list, &ctx->required_features);
1381 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1382 ctx->extract_flags);
1386 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1390 if (unlikely(list_empty(&dentry_list))) {
1391 WARNING("There is nothing to extract!");
1395 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1399 dentry_list_build_inode_alias_lists(&dentry_list);
1401 ret = dentry_list_ref_streams(&dentry_list, ctx);
1405 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1406 /* When extracting from a pipe, the number of bytes of data to
1407 * extract can't be determined in the normal way (examining the
1408 * blob table), since at this point all we have is a set of
1409 * SHA-1 message digests of blobs that need to be extracted.
1410 * However, we can get a reasonably accurate estimate by taking
1411 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1412 * data. This does assume that a full image is being extracted,
1413 * but currently there is no API for doing otherwise. (Also,
1414 * subtract <HARDLINKBYTES> from this if hard links are
1415 * supported by the extraction mode.) */
1416 ctx->progress.extract.total_bytes =
1417 wim_info_get_image_total_bytes(wim->wim_info,
1418 wim->current_image);
1419 if (ctx->supported_features.hard_links) {
1420 ctx->progress.extract.total_bytes -=
1421 wim_info_get_image_hard_link_bytes(wim->wim_info,
1422 wim->current_image);
1426 ret = extract_progress(ctx,
1427 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1428 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1429 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1433 ret = (*ops->extract)(&dentry_list, ctx);
1437 if (ctx->progress.extract.completed_bytes <
1438 ctx->progress.extract.total_bytes)
1440 ctx->progress.extract.completed_bytes =
1441 ctx->progress.extract.total_bytes;
1442 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1447 ret = extract_progress(ctx,
1448 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1449 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1450 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1452 destroy_blob_list(&ctx->blob_list);
1453 destroy_dentry_list(&dentry_list);
1460 mkdir_if_needed(const tchar *target)
1462 if (!tmkdir(target, 0755))
1465 if (errno == EEXIST)
1469 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1470 if (errno == EACCES)
1474 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1475 return WIMLIB_ERR_MKDIR;
1478 /* Make sure the extraction flags make sense, and update them if needed. */
1480 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1482 int extract_flags = *extract_flags_p;
1484 /* Check for invalid flag combinations */
1486 if ((extract_flags &
1487 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1488 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1489 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1490 return WIMLIB_ERR_INVALID_PARAM;
1492 if ((extract_flags &
1493 (WIMLIB_EXTRACT_FLAG_RPFIX |
1494 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1495 WIMLIB_EXTRACT_FLAG_NORPFIX))
1496 return WIMLIB_ERR_INVALID_PARAM;
1498 #ifndef WITH_NTFS_3G
1499 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1500 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1501 " it cannot apply a WIM image directly to an NTFS volume.");
1502 return WIMLIB_ERR_UNSUPPORTED;
1506 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1509 return WIMLIB_ERR_NO_FILENAME;
1511 ERROR("WIMBoot extraction is only supported on Windows!");
1512 return WIMLIB_ERR_UNSUPPORTED;
1517 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1518 WIMLIB_EXTRACT_FLAG_NORPFIX |
1519 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1520 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1522 /* For full-image extraction, do reparse point fixups by default
1523 * if the WIM header says they are enabled. */
1524 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1525 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1528 *extract_flags_p = extract_flags;
1533 get_wildcard_flags(int extract_flags)
1535 u32 wildcard_flags = 0;
1537 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1538 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1540 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1542 if (default_ignore_case)
1543 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1545 return wildcard_flags;
1548 struct append_dentry_ctx {
1549 struct wim_dentry **dentries;
1550 size_t num_dentries;
1551 size_t num_alloc_dentries;
1555 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1557 struct append_dentry_ctx *ctx = _ctx;
1559 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1560 struct wim_dentry **new_dentries;
1563 new_length = max(ctx->num_alloc_dentries + 8,
1564 ctx->num_alloc_dentries * 3 / 2);
1565 new_dentries = REALLOC(ctx->dentries,
1566 new_length * sizeof(ctx->dentries[0]));
1567 if (new_dentries == NULL)
1568 return WIMLIB_ERR_NOMEM;
1569 ctx->dentries = new_dentries;
1570 ctx->num_alloc_dentries = new_length;
1572 ctx->dentries[ctx->num_dentries++] = dentry;
1577 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1578 const tchar * const *paths, size_t num_paths,
1582 struct wim_dentry **trees;
1585 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1586 (num_paths != 0 && paths == NULL))
1587 return WIMLIB_ERR_INVALID_PARAM;
1589 ret = check_extract_flags(wim, &extract_flags);
1593 ret = select_wim_image(wim, image);
1597 ret = wim_checksum_unhashed_blobs(wim);
1601 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1602 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1603 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1605 ret = mkdir_if_needed(target);
1610 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1612 struct append_dentry_ctx append_dentry_ctx = {
1615 .num_alloc_dentries = 0,
1618 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1620 for (size_t i = 0; i < num_paths; i++) {
1621 tchar *path = canonicalize_wim_path(paths[i]);
1623 ret = WIMLIB_ERR_NOMEM;
1624 trees = append_dentry_ctx.dentries;
1625 goto out_free_trees;
1627 ret = expand_wildcard(wim, path,
1633 trees = append_dentry_ctx.dentries;
1634 goto out_free_trees;
1637 trees = append_dentry_ctx.dentries;
1638 num_trees = append_dentry_ctx.num_dentries;
1640 trees = MALLOC(num_paths * sizeof(trees[0]));
1642 return WIMLIB_ERR_NOMEM;
1644 for (size_t i = 0; i < num_paths; i++) {
1646 tchar *path = canonicalize_wim_path(paths[i]);
1648 ret = WIMLIB_ERR_NOMEM;
1649 goto out_free_trees;
1652 trees[i] = get_dentry(wim, path,
1653 WIMLIB_CASE_PLATFORM_DEFAULT);
1655 if (trees[i] == NULL) {
1656 ERROR("Path \"%"TS"\" does not exist "
1658 paths[i], wim->current_image);
1659 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1660 goto out_free_trees;
1663 num_trees = num_paths;
1666 if (num_trees == 0) {
1668 goto out_free_trees;
1671 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1678 extract_single_image(WIMStruct *wim, int image,
1679 const tchar *target, int extract_flags)
1681 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1682 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1683 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1686 static const tchar * const filename_forbidden_chars =
1695 /* This function checks if it is okay to use a WIM image's name as a directory
1698 image_name_ok_as_dir(const tchar *image_name)
1700 return image_name && *image_name &&
1701 !tstrpbrk(image_name, filename_forbidden_chars) &&
1702 tstrcmp(image_name, T(".")) &&
1703 tstrcmp(image_name, T(".."));
1706 /* Extracts all images from the WIM to the directory @target, with the images
1707 * placed in subdirectories named by their image names. */
1709 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1711 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1712 size_t output_path_len = tstrlen(target);
1713 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1716 const tchar *image_name;
1718 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1719 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1720 return WIMLIB_ERR_INVALID_PARAM;
1723 ret = mkdir_if_needed(target);
1726 tmemcpy(buf, target, output_path_len);
1727 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1728 for (image = 1; image <= wim->hdr.image_count; image++) {
1729 image_name = wimlib_get_image_name(wim, image);
1730 if (image_name_ok_as_dir(image_name)) {
1731 tstrcpy(buf + output_path_len + 1, image_name);
1733 /* Image name is empty or contains forbidden characters.
1734 * Use image number instead. */
1735 tsprintf(buf + output_path_len + 1, T("%d"), image);
1737 ret = extract_single_image(wim, image, buf, extract_flags);
1745 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1748 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1749 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1750 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1751 return WIMLIB_ERR_INVALID_PARAM;
1753 if (image == WIMLIB_ALL_IMAGES)
1754 return extract_all_images(wim, target, extract_flags);
1756 return extract_single_image(wim, image, target, extract_flags);
1760 /****************************************************************************
1762 ****************************************************************************/
1765 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1766 const tchar * const *paths, size_t num_paths,
1769 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1770 return WIMLIB_ERR_INVALID_PARAM;
1772 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1777 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1778 const tchar *path_list_file, int extract_flags)
1785 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1787 ERROR("Failed to read path list file \"%"TS"\"",
1792 ret = wimlib_extract_paths(wim, image, target,
1793 (const tchar * const *)paths, num_paths,
1801 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1802 const tchar *image_num_or_name,
1803 const tchar *target,
1805 wimlib_progress_func_t progfunc,
1810 struct filedes *in_fd;
1814 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1815 return WIMLIB_ERR_INVALID_PARAM;
1817 /* Read the WIM header from the pipe and get a WIMStruct to represent
1818 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1819 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1820 * empty blob table, no XML data read, and no filename set. */
1821 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1826 /* Sanity check to make sure this is a pipable WIM. */
1827 if (pwm->hdr.magic != PWM_MAGIC) {
1828 ERROR("The WIM being read from file descriptor %d "
1829 "is not pipable!", pipe_fd);
1830 ret = WIMLIB_ERR_NOT_PIPABLE;
1831 goto out_wimlib_free;
1834 /* Sanity check to make sure the first part of a pipable split WIM is
1835 * sent over the pipe first. */
1836 if (pwm->hdr.part_number != 1) {
1837 ERROR("The first part of the split WIM must be "
1838 "sent over the pipe first.");
1839 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1840 goto out_wimlib_free;
1843 in_fd = &pwm->in_fd;
1844 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1846 /* As mentioned, the WIMStruct we created from the pipe does not have
1847 * XML data yet. Fix this by reading the extra copy of the XML data
1848 * that directly follows the header in pipable WIMs. (Note: see
1849 * write_pipable_wim() for more details about the format of pipable
1852 u8 hash[SHA1_HASH_SIZE];
1854 ret = read_pwm_blob_header(pwm, hash,
1855 &pwm->hdr.xml_data_reshdr, NULL);
1857 goto out_wimlib_free;
1859 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1860 ERROR("Expected XML data, but found non-metadata resource.");
1861 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1862 goto out_wimlib_free;
1865 ret = read_wim_xml_data(pwm);
1867 goto out_wimlib_free;
1869 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1870 ERROR("Image count in XML data is not the same as in WIM header.");
1871 ret = WIMLIB_ERR_IMAGE_COUNT;
1872 goto out_wimlib_free;
1876 /* Get image index (this may use the XML data that was just read to
1877 * resolve an image name). */
1878 if (image_num_or_name) {
1879 image = wimlib_resolve_image(pwm, image_num_or_name);
1880 if (image == WIMLIB_NO_IMAGE) {
1881 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1883 ret = WIMLIB_ERR_INVALID_IMAGE;
1884 goto out_wimlib_free;
1885 } else if (image == WIMLIB_ALL_IMAGES) {
1886 ERROR("Applying all images from a pipe is not supported!");
1887 ret = WIMLIB_ERR_INVALID_IMAGE;
1888 goto out_wimlib_free;
1891 if (pwm->hdr.image_count != 1) {
1892 ERROR("No image was specified, but the pipable WIM "
1893 "did not contain exactly 1 image");
1894 ret = WIMLIB_ERR_INVALID_IMAGE;
1895 goto out_wimlib_free;
1900 /* Load the needed metadata resource. */
1901 for (i = 1; i <= pwm->hdr.image_count; i++) {
1902 struct wim_image_metadata *imd;
1903 struct wim_reshdr reshdr;
1904 struct wim_resource_descriptor *metadata_rdesc;
1906 imd = pwm->image_metadata[i - 1];
1908 ret = WIMLIB_ERR_NOMEM;
1909 imd->metadata_blob = new_blob_descriptor();
1910 if (!imd->metadata_blob)
1911 goto out_wimlib_free;
1913 imd->metadata_blob->is_metadata = 1;
1915 ret = read_pwm_blob_header(pwm, imd->metadata_blob->hash,
1918 goto out_wimlib_free;
1920 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1921 ERROR("Expected metadata resource, but found "
1922 "non-metadata resource");
1923 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1924 goto out_wimlib_free;
1927 ret = WIMLIB_ERR_NOMEM;
1928 metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
1929 if (!metadata_rdesc)
1930 goto out_wimlib_free;
1931 wim_reshdr_to_desc_and_blob(&reshdr, pwm, metadata_rdesc,
1932 imd->metadata_blob);
1935 /* Metadata resource is for the image being extracted.
1936 * Parse it and save the metadata in memory. */
1937 ret = read_metadata_resource(imd);
1939 goto out_wimlib_free;
1942 /* Metadata resource is not for the image being
1943 * extracted. Skip over it. */
1944 ret = skip_wim_resource(metadata_rdesc);
1946 goto out_wimlib_free;
1949 /* Extract the image. */
1950 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1951 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1952 /* Clean up and return. */
1960 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1961 const tchar *target, int extract_flags)
1963 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1972 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1975 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1976 return WIMLIB_ERR_INVALID_PARAM;
1977 return do_wimlib_extract_image(wim, image, target, extract_flags);